Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / bda98b63105b4ed36d217a400bb4bab9b49b4ea6 / . / docs / LangRef.html
blob: ada3eacc3b898bc8487eb0dfe4366d789db199ce [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>
64 <li><a href="#aggregateconstants">Aggregate Constants</a></li>
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>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]548 <dd>The semantics of this linkage follow the ELF object file model: the
549 symbol is weak until linked, if not linked, the symbol becomes null instead
550 of being an undefined reference.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]551 </dd>
552
553 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
554
555 <dd>If none of the above identifiers are used, the global is externally
556 visible, meaning that it participates in linkage and can be used to resolve
557 external symbol references.
558 </dd>
559</dl>
560
561 <p>
562 The next two types of linkage are targeted for Microsoft Windows platform
563 only. They are designed to support importing (exporting) symbols from (to)
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]564 DLLs (Dynamic Link Libraries).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]565 </p>
566
567 <dl>
568 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
569
570 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
571 or variable via a global pointer to a pointer that is set up by the DLL
572 exporting the symbol. On Microsoft Windows targets, the pointer name is
Dan Gohman5ec99832009-01-12 21:35:55 +0000[diff] [blame]573 formed by combining <code>__imp_</code> and the function or variable name.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]574 </dd>
575
576 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
577
578 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
579 pointer to a pointer in a DLL, so that it can be referenced with the
580 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
Dan Gohman5ec99832009-01-12 21:35:55 +0000[diff] [blame]581 name is formed by combining <code>__imp_</code> and the function or variable
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]582 name.
583 </dd>
584
585</dl>
586
Dan Gohman4dfac702008-11-24 17:18:39 +0000[diff] [blame]587<p>For example, since the "<tt>.LC0</tt>"
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]588variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
589variable and was linked with this one, one of the two would be renamed,
590preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
591external (i.e., lacking any linkage declarations), they are accessible
592outside of the current module.</p>
593<p>It is illegal for a function <i>declaration</i>
594to have any linkage type other than "externally visible", <tt>dllimport</tt>,
595or <tt>extern_weak</tt>.</p>
596<p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]597linkages.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]598</div>
599
600<!-- ======================================================================= -->
601<div class="doc_subsection">
602 <a name="callingconv">Calling Conventions</a>
603</div>
604
605<div class="doc_text">
606
607<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
608and <a href="#i_invoke">invokes</a> can all have an optional calling convention
609specified for the call. The calling convention of any pair of dynamic
610caller/callee must match, or the behavior of the program is undefined. The
611following calling conventions are supported by LLVM, and more may be added in
612the future:</p>
613
614<dl>
615 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
616
617 <dd>This calling convention (the default if no other calling convention is
618 specified) matches the target C calling conventions. This calling convention
619 supports varargs function calls and tolerates some mismatch in the declared
620 prototype and implemented declaration of the function (as does normal C).
621 </dd>
622
623 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
624
625 <dd>This calling convention attempts to make calls as fast as possible
626 (e.g. by passing things in registers). This calling convention allows the
627 target to use whatever tricks it wants to produce fast code for the target,
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]628 without having to conform to an externally specified ABI (Application Binary
629 Interface). Implementations of this convention should allow arbitrary
Arnold Schwaighofer07444922008-05-14 09:17:12 +0000[diff] [blame]630 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> to be
631 supported. This calling convention does not support varargs and requires the
632 prototype of all callees to exactly match the prototype of the function
633 definition.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]634 </dd>
635
636 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
637
638 <dd>This calling convention attempts to make code in the caller as efficient
639 as possible under the assumption that the call is not commonly executed. As
640 such, these calls often preserve all registers so that the call does not break
641 any live ranges in the caller side. This calling convention does not support
642 varargs and requires the prototype of all callees to exactly match the
643 prototype of the function definition.
644 </dd>
645
646 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
647
648 <dd>Any calling convention may be specified by number, allowing
649 target-specific calling conventions to be used. Target specific calling
650 conventions start at 64.
651 </dd>
652</dl>
653
654<p>More calling conventions can be added/defined on an as-needed basis, to
655support pascal conventions or any other well-known target-independent
656convention.</p>
657
658</div>
659
660<!-- ======================================================================= -->
661<div class="doc_subsection">
662 <a name="visibility">Visibility Styles</a>
663</div>
664
665<div class="doc_text">
666
667<p>
668All Global Variables and Functions have one of the following visibility styles:
669</p>
670
671<dl>
672 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
673
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]674 <dd>On targets that use the ELF object file format, default visibility means
675 that the declaration is visible to other
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]676 modules and, in shared libraries, means that the declared entity may be
677 overridden. On Darwin, default visibility means that the declaration is
678 visible to other modules. Default visibility corresponds to "external
679 linkage" in the language.
680 </dd>
681
682 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
683
684 <dd>Two declarations of an object with hidden visibility refer to the same
685 object if they are in the same shared object. Usually, hidden visibility
686 indicates that the symbol will not be placed into the dynamic symbol table,
687 so no other module (executable or shared library) can reference it
688 directly.
689 </dd>
690
691 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
692
693 <dd>On ELF, protected visibility indicates that the symbol will be placed in
694 the dynamic symbol table, but that references within the defining module will
695 bind to the local symbol. That is, the symbol cannot be overridden by another
696 module.
697 </dd>
698</dl>
699
700</div>
701
702<!-- ======================================================================= -->
703<div class="doc_subsection">
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000[diff] [blame]704 <a name="namedtypes">Named Types</a>
705</div>
706
707<div class="doc_text">
708
709<p>LLVM IR allows you to specify name aliases for certain types. This can make
710it easier to read the IR and make the IR more condensed (particularly when
711recursive types are involved). An example of a name specification is:
712</p>
713
714<div class="doc_code">
715<pre>
716%mytype = type { %mytype*, i32 }
717</pre>
718</div>
719
720<p>You may give a name to any <a href="#typesystem">type</a> except "<a
721href="t_void">void</a>". Type name aliases may be used anywhere a type is
722expected with the syntax "%mytype".</p>
723
724<p>Note that type names are aliases for the structural type that they indicate,
725and that you can therefore specify multiple names for the same type. This often
726leads to confusing behavior when dumping out a .ll file. Since LLVM IR uses
727structural typing, the name is not part of the type. When printing out LLVM IR,
728the printer will pick <em>one name</em> to render all types of a particular
729shape. This means that if you have code where two different source types end up
730having the same LLVM type, that the dumper will sometimes print the "wrong" or
731unexpected type. This is an important design point and isn't going to
732change.</p>
733
734</div>
735
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000[diff] [blame]736<!-- ======================================================================= -->
737<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]738 <a name="globalvars">Global Variables</a>
739</div>
740
741<div class="doc_text">
742
743<p>Global variables define regions of memory allocated at compilation time
744instead of run-time. Global variables may optionally be initialized, may have
745an explicit section to be placed in, and may have an optional explicit alignment
746specified. A variable may be defined as "thread_local", which means that it
747will not be shared by threads (each thread will have a separated copy of the
748variable). A variable may be defined as a global "constant," which indicates
749that the contents of the variable will <b>never</b> be modified (enabling better
750optimization, allowing the global data to be placed in the read-only section of
751an executable, etc). Note that variables that need runtime initialization
752cannot be marked "constant" as there is a store to the variable.</p>
753
754<p>
755LLVM explicitly allows <em>declarations</em> of global variables to be marked
756constant, even if the final definition of the global is not. This capability
757can be used to enable slightly better optimization of the program, but requires
758the language definition to guarantee that optimizations based on the
759'constantness' are valid for the translation units that do not include the
760definition.
761</p>
762
763<p>As SSA values, global variables define pointer values that are in
764scope (i.e. they dominate) all basic blocks in the program. Global
765variables always define a pointer to their "content" type because they
766describe a region of memory, and all memory objects in LLVM are
767accessed through pointers.</p>
768
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]769<p>A global variable may be declared to reside in a target-specifc numbered
770address space. For targets that support them, address spaces may affect how
771optimizations are performed and/or what target instructions are used to access
Christopher Lamb20a39e92007-12-12 08:44:39 +0000[diff] [blame]772the variable. The default address space is zero. The address space qualifier
773must precede any other attributes.</p>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]774
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]775<p>LLVM allows an explicit section to be specified for globals. If the target
776supports it, it will emit globals to the section specified.</p>
777
778<p>An explicit alignment may be specified for a global. If not present, or if
779the alignment is set to zero, the alignment of the global is set by the target
780to whatever it feels convenient. If an explicit alignment is specified, the
781global is forced to have at least that much alignment. All alignments must be
782a power of 2.</p>
783
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]784<p>For example, the following defines a global in a numbered address space with
785an initializer, section, and alignment:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]786
787<div class="doc_code">
788<pre>
Dan Gohman21ef02c2009-01-11 00:40:00 +0000[diff] [blame]789@G = addrspace(5) constant float 1.0, section "foo", align 4
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]790</pre>
791</div>
792
793</div>
794
795
796<!-- ======================================================================= -->
797<div class="doc_subsection">
798 <a name="functionstructure">Functions</a>
799</div>
800
801<div class="doc_text">
802
803<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
804an optional <a href="#linkage">linkage type</a>, an optional
805<a href="#visibility">visibility style</a>, an optional
806<a href="#callingconv">calling convention</a>, a return type, an optional
807<a href="#paramattrs">parameter attribute</a> for the return type, a function
808name, a (possibly empty) argument list (each with optional
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]809<a href="#paramattrs">parameter attributes</a>), optional
810<a href="#fnattrs">function attributes</a>, an optional section,
811an optional alignment, an optional <a href="#gc">garbage collector name</a>,
Chris Lattneree202542008-10-04 18:10:21 +0000[diff] [blame]812an opening curly brace, a list of basic blocks, and a closing curly brace.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]813
814LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
815optional <a href="#linkage">linkage type</a>, an optional
816<a href="#visibility">visibility style</a>, an optional
817<a href="#callingconv">calling convention</a>, a return type, an optional
818<a href="#paramattrs">parameter attribute</a> for the return type, a function
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]819name, a possibly empty list of arguments, an optional alignment, and an optional
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]820<a href="#gc">garbage collector name</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]821
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]822<p>A function definition contains a list of basic blocks, forming the CFG
823(Control Flow Graph) for
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]824the function. Each basic block may optionally start with a label (giving the
825basic block a symbol table entry), contains a list of instructions, and ends
826with a <a href="#terminators">terminator</a> instruction (such as a branch or
827function return).</p>
828
829<p>The first basic block in a function is special in two ways: it is immediately
830executed on entrance to the function, and it is not allowed to have predecessor
831basic blocks (i.e. there can not be any branches to the entry block of a
832function). Because the block can have no predecessors, it also cannot have any
833<a href="#i_phi">PHI nodes</a>.</p>
834
835<p>LLVM allows an explicit section to be specified for functions. If the target
836supports it, it will emit functions to the section specified.</p>
837
838<p>An explicit alignment may be specified for a function. If not present, or if
839the alignment is set to zero, the alignment of the function is set by the target
840to whatever it feels convenient. If an explicit alignment is specified, the
841function is forced to have at least that much alignment. All alignments must be
842a power of 2.</p>
843
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]844 <h5>Syntax:</h5>
845
846<div class="doc_code">
Chris Lattner1e5c5cd02008-10-13 16:55:18 +0000[diff] [blame]847<tt>
848define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
849 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
850 &lt;ResultType&gt; @&lt;FunctionName&gt; ([argument list])
851 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
852 [<a href="#gc">gc</a>] { ... }
853</tt>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]854</div>
855
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]856</div>
857
858
859<!-- ======================================================================= -->
860<div class="doc_subsection">
861 <a name="aliasstructure">Aliases</a>
862</div>
863<div class="doc_text">
864 <p>Aliases act as "second name" for the aliasee value (which can be either
Anton Korobeynikov96822812008-03-22 08:36:14 +0000[diff] [blame]865 function, global variable, another alias or bitcast of global value). Aliases
866 may have an optional <a href="#linkage">linkage type</a>, and an
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]867 optional <a href="#visibility">visibility style</a>.</p>
868
869 <h5>Syntax:</h5>
870
871<div class="doc_code">
872<pre>
Duncan Sandsd7bfabf2008-09-12 20:48:21 +0000[diff] [blame]873@&lt;Name&gt; = alias [Linkage] [Visibility] &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]874</pre>
875</div>
876
877</div>
878
879
880
881<!-- ======================================================================= -->
882<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
883<div class="doc_text">
884 <p>The return type and each parameter of a function type may have a set of
885 <i>parameter attributes</i> associated with them. Parameter attributes are
886 used to communicate additional information about the result or parameters of
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]887 a function. Parameter attributes are considered to be part of the function,
888 not of the function type, so functions with different parameter attributes
889 can have the same function type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]890
891 <p>Parameter attributes are simple keywords that follow the type specified. If
892 multiple parameter attributes are needed, they are space separated. For
893 example:</p>
894
895<div class="doc_code">
896<pre>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]897declare i32 @printf(i8* noalias , ...)
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]898declare i32 @atoi(i8 zeroext)
899declare signext i8 @returns_signed_char()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]900</pre>
901</div>
902
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]903 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
904 <tt>readonly</tt>) come immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]905
906 <p>Currently, only the following parameter attributes are defined:</p>
907 <dl>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]908 <dt><tt>zeroext</tt></dt>
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]909 <dd>This indicates to the code generator that the parameter or return value
910 should be zero-extended to a 32-bit value by the caller (for a parameter)
911 or the callee (for a return value).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]912
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]913 <dt><tt>signext</tt></dt>
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]914 <dd>This indicates to the code generator that the parameter or return value
915 should be sign-extended to a 32-bit value by the caller (for a parameter)
916 or the callee (for a return value).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]917
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]918 <dt><tt>inreg</tt></dt>
Dale Johannesenc08a0e22008-09-25 20:47:45 +0000[diff] [blame]919 <dd>This indicates that this parameter or return value should be treated
920 in a special target-dependent fashion during while emitting code for a
921 function call or return (usually, by putting it in a register as opposed
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]922 to memory, though some targets use it to distinguish between two different
923 kinds of registers). Use of this attribute is target-specific.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]924
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]925 <dt><tt><a name="byval">byval</a></tt></dt>
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]926 <dd>This indicates that the pointer parameter should really be passed by
927 value to the function. The attribute implies that a hidden copy of the
928 pointee is made between the caller and the callee, so the callee is unable
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]929 to modify the value in the callee. This attribute is only valid on LLVM
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]930 pointer arguments. It is generally used to pass structs and arrays by
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]931 value, but is also valid on pointers to scalars. The copy is considered to
932 belong to the caller not the callee (for example,
933 <tt><a href="#readonly">readonly</a></tt> functions should not write to
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]934 <tt>byval</tt> parameters). This is not a valid attribute for return
Chris Lattnere39c5bf2009-02-05 05:42:28 +0000[diff] [blame]935 values. The byval attribute also supports specifying an alignment with the
936 align attribute. This has a target-specific effect on the code generator
937 that usually indicates a desired alignment for the synthesized stack
938 slot.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]939
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]940 <dt><tt>sret</tt></dt>
Duncan Sands616cc032008-02-18 04:19:38 +0000[diff] [blame]941 <dd>This indicates that the pointer parameter specifies the address of a
942 structure that is the return value of the function in the source program.
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame]943 This pointer must be guaranteed by the caller to be valid: loads and stores
944 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]945 be applied to the first parameter. This is not a valid attribute for
946 return values. </dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]947
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]948 <dt><tt>noalias</tt></dt>
Nick Lewyckyff384472008-11-24 03:41:24 +0000[diff] [blame]949 <dd>This indicates that the pointer does not alias any global or any other
950 parameter. The caller is responsible for ensuring that this is the
Nick Lewyckya604a942008-11-24 05:00:44 +0000[diff] [blame]951 case. On a function return value, <tt>noalias</tt> additionally indicates
952 that the pointer does not alias any other pointers visible to the
Nick Lewycky40fb6f22008-12-19 06:39:12 +0000[diff] [blame]953 caller. For further details, please see the discussion of the NoAlias
954 response in
955 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
956 analysis</a>.</dd>
957
958 <dt><tt>nocapture</tt></dt>
959 <dd>This indicates that the callee does not make any copies of the pointer
960 that outlive the callee itself. This is not a valid attribute for return
961 values.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]962
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]963 <dt><tt>nest</tt></dt>
Duncan Sandsf1a7d4c2008-07-08 09:27:25 +0000[diff] [blame]964 <dd>This indicates that the pointer parameter can be excised using the
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]965 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
966 attribute for return values.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]967 </dl>
968
969</div>
970
971<!-- ======================================================================= -->
972<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]973 <a name="gc">Garbage Collector Names</a>
974</div>
975
976<div class="doc_text">
977<p>Each function may specify a garbage collector name, which is simply a
978string.</p>
979
980<div class="doc_code"><pre
981>define void @f() gc "name" { ...</pre></div>
982
983<p>The compiler declares the supported values of <i>name</i>. Specifying a
984collector which will cause the compiler to alter its output in order to support
985the named garbage collection algorithm.</p>
986</div>
987
988<!-- ======================================================================= -->
989<div class="doc_subsection">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]990 <a name="fnattrs">Function Attributes</a>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]991</div>
992
993<div class="doc_text">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]994
995<p>Function attributes are set to communicate additional information about
996 a function. Function attributes are considered to be part of the function,
997 not of the function type, so functions with different parameter attributes
998 can have the same function type.</p>
999
1000 <p>Function attributes are simple keywords that follow the type specified. If
1001 multiple attributes are needed, they are space separated. For
1002 example:</p>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]1003
1004<div class="doc_code">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1005<pre>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1006define void @f() noinline { ... }
1007define void @f() alwaysinline { ... }
1008define void @f() alwaysinline optsize { ... }
1009define void @f() optsize
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1010</pre>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]1011</div>
1012
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1013<dl>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1014<dt><tt>alwaysinline</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1015<dd>This attribute indicates that the inliner should attempt to inline this
1016function into callers whenever possible, ignoring any active inlining size
1017threshold for this caller.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1018
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1019<dt><tt>noinline</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1020<dd>This attribute indicates that the inliner should never inline this function
Chris Lattner377f3ae2008-10-05 17:14:59 +0000[diff] [blame]1021in any situation. This attribute may not be used together with the
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1022<tt>alwaysinline</tt> attribute.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1023
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1024<dt><tt>optsize</tt></dt>
Devang Patelc29099b2008-09-29 18:34:44 +0000[diff] [blame]1025<dd>This attribute suggests that optimization passes and code generator passes
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1026make choices that keep the code size of this function low, and otherwise do
1027optimizations specifically to reduce code size.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1028
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1029<dt><tt>noreturn</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1030<dd>This function attribute indicates that the function never returns normally.
1031This produces undefined behavior at runtime if the function ever does
1032dynamically return.</dd>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1033
1034<dt><tt>nounwind</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]1035<dd>This function attribute indicates that the function never returns with an
1036unwind or exceptional control flow. If the function does unwind, its runtime
1037behavior is undefined.</dd>
1038
1039<dt><tt>readnone</tt></dt>
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]1040<dd>This attribute indicates that the function computes its result (or the
1041exception it throws) based strictly on its arguments, without dereferencing any
1042pointer arguments or otherwise accessing any mutable state (e.g. memory, control
1043registers, etc) visible to caller functions. It does not write through any
1044pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments) and
1045never changes any state visible to callers.</dd>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]1046
Duncan Sands7d94e5a2008-10-06 08:14:18 +0000[diff] [blame]1047<dt><tt><a name="readonly">readonly</a></tt></dt>
1048<dd>This attribute indicates that the function does not write through any
1049pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments)
1050or otherwise modify any state (e.g. memory, control registers, etc) visible to
1051caller functions. It may dereference pointer arguments and read state that may
1052be set in the caller. A readonly function always returns the same value (or
1053throws the same exception) when called with the same set of arguments and global
1054state.</dd>
Bill Wendlingdfaabba2008-11-13 01:02:51 +0000[diff] [blame]1055
1056<dt><tt><a name="ssp">ssp</a></tt></dt>
Bill Wendling34f70d82008-11-26 19:19:05 +0000[diff] [blame]1057<dd>This attribute indicates that the function should emit a stack smashing
Bill Wendlingdfaabba2008-11-13 01:02:51 +0000[diff] [blame]1058protector. It is in the form of a "canary"&mdash;a random value placed on the
1059stack before the local variables that's checked upon return from the function to
1060see if it has been overwritten. A heuristic is used to determine if a function
Bill Wendling34f70d82008-11-26 19:19:05 +0000[diff] [blame]1061needs stack protectors or not.
Bill Wendlingdfaabba2008-11-13 01:02:51 +0000[diff] [blame]1062
Bill Wendlingbe9ec3f2008-11-26 19:07:40 +0000[diff] [blame]1063<p>If a function that has an <tt>ssp</tt> attribute is inlined into a function
1064that doesn't have an <tt>ssp</tt> attribute, then the resulting function will
1065have an <tt>ssp</tt> attribute.</p></dd>
1066
1067<dt><tt>sspreq</tt></dt>
Bill Wendling34f70d82008-11-26 19:19:05 +0000[diff] [blame]1068<dd>This attribute indicates that the function should <em>always</em> emit a
Bill Wendlingdfaabba2008-11-13 01:02:51 +0000[diff] [blame]1069stack smashing protector. This overrides the <tt><a href="#ssp">ssp</a></tt>
Bill Wendling34f70d82008-11-26 19:19:05 +0000[diff] [blame]1070function attribute.
Bill Wendlingbe9ec3f2008-11-26 19:07:40 +0000[diff] [blame]1071
1072<p>If a function that has an <tt>sspreq</tt> attribute is inlined into a
1073function that doesn't have an <tt>sspreq</tt> attribute or which has
1074an <tt>ssp</tt> attribute, then the resulting function will have
1075an <tt>sspreq</tt> attribute.</p></dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1076</dl>
1077
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]1078</div>
1079
1080<!-- ======================================================================= -->
1081<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1082 <a name="moduleasm">Module-Level Inline Assembly</a>
1083</div>
1084
1085<div class="doc_text">
1086<p>
1087Modules may contain "module-level inline asm" blocks, which corresponds to the
1088GCC "file scope inline asm" blocks. These blocks are internally concatenated by
1089LLVM and treated as a single unit, but may be separated in the .ll file if
1090desired. The syntax is very simple:
1091</p>
1092
1093<div class="doc_code">
1094<pre>
1095module asm "inline asm code goes here"
1096module asm "more can go here"
1097</pre>
1098</div>
1099
1100<p>The strings can contain any character by escaping non-printable characters.
1101 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1102 for the number.
1103</p>
1104
1105<p>
1106 The inline asm code is simply printed to the machine code .s file when
1107 assembly code is generated.
1108</p>
1109</div>
1110
1111<!-- ======================================================================= -->
1112<div class="doc_subsection">
1113 <a name="datalayout">Data Layout</a>
1114</div>
1115
1116<div class="doc_text">
1117<p>A module may specify a target specific data layout string that specifies how
1118data is to be laid out in memory. The syntax for the data layout is simply:</p>
1119<pre> target datalayout = "<i>layout specification</i>"</pre>
1120<p>The <i>layout specification</i> consists of a list of specifications
1121separated by the minus sign character ('-'). Each specification starts with a
1122letter and may include other information after the letter to define some
1123aspect of the data layout. The specifications accepted are as follows: </p>
1124<dl>
1125 <dt><tt>E</tt></dt>
1126 <dd>Specifies that the target lays out data in big-endian form. That is, the
1127 bits with the most significance have the lowest address location.</dd>
1128 <dt><tt>e</tt></dt>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1129 <dd>Specifies that the target lays out data in little-endian form. That is,
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1130 the bits with the least significance have the lowest address location.</dd>
1131 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1132 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1133 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1134 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1135 too.</dd>
1136 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1137 <dd>This specifies the alignment for an integer type of a given bit
1138 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1139 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1140 <dd>This specifies the alignment for a vector type of a given bit
1141 <i>size</i>.</dd>
1142 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1143 <dd>This specifies the alignment for a floating point type of a given bit
1144 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1145 (double).</dd>
1146 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1147 <dd>This specifies the alignment for an aggregate type of a given bit
1148 <i>size</i>.</dd>
1149</dl>
1150<p>When constructing the data layout for a given target, LLVM starts with a
1151default set of specifications which are then (possibly) overriden by the
1152specifications in the <tt>datalayout</tt> keyword. The default specifications
1153are given in this list:</p>
1154<ul>
1155 <li><tt>E</tt> - big endian</li>
1156 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1157 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1158 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1159 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1160 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1161 <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]1162 alignment of 64-bits</li>
1163 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1164 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1165 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1166 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1167 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1168</ul>
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]1169<p>When LLVM is determining the alignment for a given type, it uses the
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]1170following rules:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1171<ol>
1172 <li>If the type sought is an exact match for one of the specifications, that
1173 specification is used.</li>
1174 <li>If no match is found, and the type sought is an integer type, then the
1175 smallest integer type that is larger than the bitwidth of the sought type is
1176 used. If none of the specifications are larger than the bitwidth then the the
1177 largest integer type is used. For example, given the default specifications
1178 above, the i7 type will use the alignment of i8 (next largest) while both
1179 i65 and i256 will use the alignment of i64 (largest specified).</li>
1180 <li>If no match is found, and the type sought is a vector type, then the
1181 largest vector type that is smaller than the sought vector type will be used
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]1182 as a fall back. This happens because &lt;128 x double&gt; can be implemented
1183 in terms of 64 &lt;2 x double&gt;, for example.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1184</ol>
1185</div>
1186
1187<!-- *********************************************************************** -->
1188<div class="doc_section"> <a name="typesystem">Type System</a> </div>
1189<!-- *********************************************************************** -->
1190
1191<div class="doc_text">
1192
1193<p>The LLVM type system is one of the most important features of the
1194intermediate representation. Being typed enables a number of
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1195optimizations to be performed on the intermediate representation directly,
1196without having to do
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1197extra analyses on the side before the transformation. A strong type
1198system makes it easier to read the generated code and enables novel
1199analyses and transformations that are not feasible to perform on normal
1200three address code representations.</p>
1201
1202</div>
1203
1204<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1205<div class="doc_subsection"> <a name="t_classifications">Type
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1206Classifications</a> </div>
1207<div class="doc_text">
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1208<p>The types fall into a few useful
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1209classifications:</p>
1210
1211<table border="1" cellspacing="0" cellpadding="4">
1212 <tbody>
1213 <tr><th>Classification</th><th>Types</th></tr>
1214 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1215 <td><a href="#t_integer">integer</a></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1216 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1217 </tr>
1218 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1219 <td><a href="#t_floating">floating point</a></td>
1220 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1221 </tr>
1222 <tr>
1223 <td><a name="t_firstclass">first class</a></td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1224 <td><a href="#t_integer">integer</a>,
1225 <a href="#t_floating">floating point</a>,
1226 <a href="#t_pointer">pointer</a>,
Dan Gohmanf6237db2008-06-18 18:42:13 +0000[diff] [blame]1227 <a href="#t_vector">vector</a>,
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]1228 <a href="#t_struct">structure</a>,
1229 <a href="#t_array">array</a>,
Dan Gohmand13951e2008-05-23 22:50:26 +0000[diff] [blame]1230 <a href="#t_label">label</a>.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1231 </td>
1232 </tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1233 <tr>
1234 <td><a href="#t_primitive">primitive</a></td>
1235 <td><a href="#t_label">label</a>,
1236 <a href="#t_void">void</a>,
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1237 <a href="#t_floating">floating point</a>.</td>
1238 </tr>
1239 <tr>
1240 <td><a href="#t_derived">derived</a></td>
1241 <td><a href="#t_integer">integer</a>,
1242 <a href="#t_array">array</a>,
1243 <a href="#t_function">function</a>,
1244 <a href="#t_pointer">pointer</a>,
1245 <a href="#t_struct">structure</a>,
1246 <a href="#t_pstruct">packed structure</a>,
1247 <a href="#t_vector">vector</a>,
1248 <a href="#t_opaque">opaque</a>.
Dan Gohman032ba852008-10-14 16:32:04 +0000[diff] [blame]1249 </td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1250 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1251 </tbody>
1252</table>
1253
1254<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1255most important. Values of these types are the only ones which can be
1256produced by instructions, passed as arguments, or used as operands to
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]1257instructions.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1258</div>
1259
1260<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1261<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1262
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1263<div class="doc_text">
1264<p>The primitive types are the fundamental building blocks of the LLVM
1265system.</p>
1266
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1267</div>
1268
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1269<!-- _______________________________________________________________________ -->
1270<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1271
1272<div class="doc_text">
1273 <table>
1274 <tbody>
1275 <tr><th>Type</th><th>Description</th></tr>
1276 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1277 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1278 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1279 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1280 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1281 </tbody>
1282 </table>
1283</div>
1284
1285<!-- _______________________________________________________________________ -->
1286<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1287
1288<div class="doc_text">
1289<h5>Overview:</h5>
1290<p>The void type does not represent any value and has no size.</p>
1291
1292<h5>Syntax:</h5>
1293
1294<pre>
1295 void
1296</pre>
1297</div>
1298
1299<!-- _______________________________________________________________________ -->
1300<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1301
1302<div class="doc_text">
1303<h5>Overview:</h5>
1304<p>The label type represents code labels.</p>
1305
1306<h5>Syntax:</h5>
1307
1308<pre>
1309 label
1310</pre>
1311</div>
1312
1313
1314<!-- ======================================================================= -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1315<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1316
1317<div class="doc_text">
1318
1319<p>The real power in LLVM comes from the derived types in the system.
1320This is what allows a programmer to represent arrays, functions,
1321pointers, and other useful types. Note that these derived types may be
1322recursive: For example, it is possible to have a two dimensional array.</p>
1323
1324</div>
1325
1326<!-- _______________________________________________________________________ -->
1327<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1328
1329<div class="doc_text">
1330
1331<h5>Overview:</h5>
1332<p>The integer type is a very simple derived type that simply specifies an
1333arbitrary bit width for the integer type desired. Any bit width from 1 bit to
13342^23-1 (about 8 million) can be specified.</p>
1335
1336<h5>Syntax:</h5>
1337
1338<pre>
1339 iN
1340</pre>
1341
1342<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1343value.</p>
1344
1345<h5>Examples:</h5>
1346<table class="layout">
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1347 <tbody>
1348 <tr>
1349 <td><tt>i1</tt></td>
1350 <td>a single-bit integer.</td>
1351 </tr><tr>
1352 <td><tt>i32</tt></td>
1353 <td>a 32-bit integer.</td>
1354 </tr><tr>
1355 <td><tt>i1942652</tt></td>
1356 <td>a really big integer of over 1 million bits.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1357 </tr>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1358 </tbody>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1359</table>
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]1360
1361<p>Note that the code generator does not yet support large integer types
1362to be used as function return types. The specific limit on how large a
1363return type the code generator can currently handle is target-dependent;
1364currently it's often 64 bits for 32-bit targets and 128 bits for 64-bit
1365targets.</p>
1366
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1367</div>
1368
1369<!-- _______________________________________________________________________ -->
1370<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1371
1372<div class="doc_text">
1373
1374<h5>Overview:</h5>
1375
1376<p>The array type is a very simple derived type that arranges elements
1377sequentially in memory. The array type requires a size (number of
1378elements) and an underlying data type.</p>
1379
1380<h5>Syntax:</h5>
1381
1382<pre>
1383 [&lt;# elements&gt; x &lt;elementtype&gt;]
1384</pre>
1385
1386<p>The number of elements is a constant integer value; elementtype may
1387be any type with a size.</p>
1388
1389<h5>Examples:</h5>
1390<table class="layout">
1391 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1392 <td class="left"><tt>[40 x i32]</tt></td>
1393 <td class="left">Array of 40 32-bit integer values.</td>
1394 </tr>
1395 <tr class="layout">
1396 <td class="left"><tt>[41 x i32]</tt></td>
1397 <td class="left">Array of 41 32-bit integer values.</td>
1398 </tr>
1399 <tr class="layout">
1400 <td class="left"><tt>[4 x i8]</tt></td>
1401 <td class="left">Array of 4 8-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1402 </tr>
1403</table>
1404<p>Here are some examples of multidimensional arrays:</p>
1405<table class="layout">
1406 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1407 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1408 <td class="left">3x4 array of 32-bit integer values.</td>
1409 </tr>
1410 <tr class="layout">
1411 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1412 <td class="left">12x10 array of single precision floating point values.</td>
1413 </tr>
1414 <tr class="layout">
1415 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1416 <td class="left">2x3x4 array of 16-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1417 </tr>
1418</table>
1419
1420<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1421length array. Normally, accesses past the end of an array are undefined in
1422LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1423As a special case, however, zero length arrays are recognized to be variable
1424length. This allows implementation of 'pascal style arrays' with the LLVM
1425type "{ i32, [0 x float]}", for example.</p>
1426
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]1427<p>Note that the code generator does not yet support large aggregate types
1428to be used as function return types. The specific limit on how large an
1429aggregate return type the code generator can currently handle is
1430target-dependent, and also dependent on the aggregate element types.</p>
1431
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1432</div>
1433
1434<!-- _______________________________________________________________________ -->
1435<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1436<div class="doc_text">
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1437
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1438<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1439
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1440<p>The function type can be thought of as a function signature. It
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1441consists of a return type and a list of formal parameter types. The
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1442return 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]1443If 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]1444class types, and the struct must have at least one element.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1445
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1446<h5>Syntax:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1447
1448<pre>
1449 &lt;returntype list&gt; (&lt;parameter list&gt;)
1450</pre>
1451
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1452<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1453specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1454which indicates that the function takes a variable number of arguments.
1455Variable argument functions can access their arguments with the <a
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1456 href="#int_varargs">variable argument handling intrinsic</a> functions.
1457'<tt>&lt;returntype list&gt;</tt>' is a comma-separated list of
1458<a href="#t_firstclass">first class</a> type specifiers.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1459
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1460<h5>Examples:</h5>
1461<table class="layout">
1462 <tr class="layout">
1463 <td class="left"><tt>i32 (i32)</tt></td>
1464 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1465 </td>
1466 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1467 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1468 </tt></td>
1469 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1470 an <tt>i16</tt> that should be sign extended and a
1471 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1472 <tt>float</tt>.
1473 </td>
1474 </tr><tr class="layout">
1475 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1476 <td class="left">A vararg function that takes at least one
1477 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1478 which returns an integer. This is the signature for <tt>printf</tt> in
1479 LLVM.
1480 </td>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1481 </tr><tr class="layout">
1482 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Misha Brukmanafc88b02008-11-27 06:41:20 +0000[diff] [blame]1483 <td class="left">A function taking an <tt>i32</tt>, returning two
1484 <tt>i32</tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1485 </td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1486 </tr>
1487</table>
1488
1489</div>
1490<!-- _______________________________________________________________________ -->
1491<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1492<div class="doc_text">
1493<h5>Overview:</h5>
1494<p>The structure type is used to represent a collection of data members
1495together in memory. The packing of the field types is defined to match
1496the ABI of the underlying processor. The elements of a structure may
1497be any type that has a size.</p>
1498<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1499and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1500field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1501instruction.</p>
1502<h5>Syntax:</h5>
1503<pre> { &lt;type list&gt; }<br></pre>
1504<h5>Examples:</h5>
1505<table class="layout">
1506 <tr class="layout">
1507 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1508 <td class="left">A triple of three <tt>i32</tt> values</td>
1509 </tr><tr class="layout">
1510 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1511 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1512 second element is a <a href="#t_pointer">pointer</a> to a
1513 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1514 an <tt>i32</tt>.</td>
1515 </tr>
1516</table>
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]1517
1518<p>Note that the code generator does not yet support large aggregate types
1519to be used as function return types. The specific limit on how large an
1520aggregate return type the code generator can currently handle is
1521target-dependent, and also dependent on the aggregate element types.</p>
1522
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1523</div>
1524
1525<!-- _______________________________________________________________________ -->
1526<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1527</div>
1528<div class="doc_text">
1529<h5>Overview:</h5>
1530<p>The packed structure type is used to represent a collection of data members
1531together in memory. There is no padding between fields. Further, the alignment
1532of a packed structure is 1 byte. The elements of a packed structure may
1533be any type that has a size.</p>
1534<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1535and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1536field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1537instruction.</p>
1538<h5>Syntax:</h5>
1539<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1540<h5>Examples:</h5>
1541<table class="layout">
1542 <tr class="layout">
1543 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1544 <td class="left">A triple of three <tt>i32</tt> values</td>
1545 </tr><tr class="layout">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1546 <td class="left">
1547<tt>&lt;&nbsp;{&nbsp;float,&nbsp;i32&nbsp;(i32)*&nbsp;}&nbsp;&gt;</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1548 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1549 second element is a <a href="#t_pointer">pointer</a> to a
1550 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1551 an <tt>i32</tt>.</td>
1552 </tr>
1553</table>
1554</div>
1555
1556<!-- _______________________________________________________________________ -->
1557<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1558<div class="doc_text">
1559<h5>Overview:</h5>
1560<p>As in many languages, the pointer type represents a pointer or
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1561reference to another object, which must live in memory. Pointer types may have
1562an optional address space attribute defining the target-specific numbered
1563address space where the pointed-to object resides. The default address space is
1564zero.</p>
Chris Lattner96edbd32009-02-08 19:53:29 +0000[diff] [blame]1565
1566<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]1567it permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
Chris Lattner96edbd32009-02-08 19:53:29 +0000[diff] [blame]1568
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1569<h5>Syntax:</h5>
1570<pre> &lt;type&gt; *<br></pre>
1571<h5>Examples:</h5>
1572<table class="layout">
1573 <tr class="layout">
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]1574 <td class="left"><tt>[4 x i32]*</tt></td>
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1575 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1576 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1577 </tr>
1578 <tr class="layout">
1579 <td class="left"><tt>i32 (i32 *) *</tt></td>
1580 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1581 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1582 <tt>i32</tt>.</td>
1583 </tr>
1584 <tr class="layout">
1585 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1586 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1587 that resides in address space #5.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1588 </tr>
1589</table>
1590</div>
1591
1592<!-- _______________________________________________________________________ -->
1593<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1594<div class="doc_text">
1595
1596<h5>Overview:</h5>
1597
1598<p>A vector type is a simple derived type that represents a vector
1599of elements. Vector types are used when multiple primitive data
1600are operated in parallel using a single instruction (SIMD).
1601A vector type requires a size (number of
1602elements) and an underlying primitive data type. Vectors must have a power
1603of two length (1, 2, 4, 8, 16 ...). Vector types are
1604considered <a href="#t_firstclass">first class</a>.</p>
1605
1606<h5>Syntax:</h5>
1607
1608<pre>
1609 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1610</pre>
1611
1612<p>The number of elements is a constant integer value; elementtype may
1613be any integer or floating point type.</p>
1614
1615<h5>Examples:</h5>
1616
1617<table class="layout">
1618 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1619 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1620 <td class="left">Vector of 4 32-bit integer values.</td>
1621 </tr>
1622 <tr class="layout">
1623 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1624 <td class="left">Vector of 8 32-bit floating-point values.</td>
1625 </tr>
1626 <tr class="layout">
1627 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1628 <td class="left">Vector of 2 64-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1629 </tr>
1630</table>
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]1631
1632<p>Note that the code generator does not yet support large vector types
1633to be used as function return types. The specific limit on how large a
1634vector return type codegen can currently handle is target-dependent;
1635currently it's often a few times longer than a hardware vector register.</p>
1636
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1637</div>
1638
1639<!-- _______________________________________________________________________ -->
1640<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1641<div class="doc_text">
1642
1643<h5>Overview:</h5>
1644
1645<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1646corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1647In LLVM, opaque types can eventually be resolved to any type (not just a
1648structure type).</p>
1649
1650<h5>Syntax:</h5>
1651
1652<pre>
1653 opaque
1654</pre>
1655
1656<h5>Examples:</h5>
1657
1658<table class="layout">
1659 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1660 <td class="left"><tt>opaque</tt></td>
1661 <td class="left">An opaque type.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1662 </tr>
1663</table>
1664</div>
1665
Chris Lattner515195a2009-02-02 07:32:36 +0000[diff] [blame]1666<!-- ======================================================================= -->
1667<div class="doc_subsection">
1668 <a name="t_uprefs">Type Up-references</a>
1669</div>
1670
1671<div class="doc_text">
1672<h5>Overview:</h5>
1673<p>
1674An "up reference" allows you to refer to a lexically enclosing type without
1675requiring it to have a name. For instance, a structure declaration may contain a
1676pointer to any of the types it is lexically a member of. Example of up
1677references (with their equivalent as named type declarations) include:</p>
1678
1679<pre>
1680 { \2 * } %x = type { %t* }
1681 { \2 }* %y = type { %y }*
1682 \1* %z = type %z*
1683</pre>
1684
1685<p>
1686An up reference is needed by the asmprinter for printing out cyclic types when
1687there is no declared name for a type in the cycle. Because the asmprinter does
1688not want to print out an infinite type string, it needs a syntax to handle
1689recursive types that have no names (all names are optional in llvm IR).
1690</p>
1691
1692<h5>Syntax:</h5>
1693<pre>
1694 \&lt;level&gt;
1695</pre>
1696
1697<p>
1698The level is the count of the lexical type that is being referred to.
1699</p>
1700
1701<h5>Examples:</h5>
1702
1703<table class="layout">
1704 <tr class="layout">
1705 <td class="left"><tt>\1*</tt></td>
1706 <td class="left">Self-referential pointer.</td>
1707 </tr>
1708 <tr class="layout">
1709 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
1710 <td class="left">Recursive structure where the upref refers to the out-most
1711 structure.</td>
1712 </tr>
1713</table>
1714</div>
1715
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1716
1717<!-- *********************************************************************** -->
1718<div class="doc_section"> <a name="constants">Constants</a> </div>
1719<!-- *********************************************************************** -->
1720
1721<div class="doc_text">
1722
1723<p>LLVM has several different basic types of constants. This section describes
1724them all and their syntax.</p>
1725
1726</div>
1727
1728<!-- ======================================================================= -->
1729<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1730
1731<div class="doc_text">
1732
1733<dl>
1734 <dt><b>Boolean constants</b></dt>
1735
1736 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1737 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1738 </dd>
1739
1740 <dt><b>Integer constants</b></dt>
1741
1742 <dd>Standard integers (such as '4') are constants of the <a
1743 href="#t_integer">integer</a> type. Negative numbers may be used with
1744 integer types.
1745 </dd>
1746
1747 <dt><b>Floating point constants</b></dt>
1748
1749 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1750 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]1751 notation (see below). The assembler requires the exact decimal value of
1752 a floating-point constant. For example, the assembler accepts 1.25 but
1753 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1754 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1755
1756 <dt><b>Null pointer constants</b></dt>
1757
1758 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1759 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1760
1761</dl>
1762
1763<p>The one non-intuitive notation for constants is the optional hexadecimal form
1764of floating point constants. For example, the form '<tt>double
17650x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
17664.5e+15</tt>'. The only time hexadecimal floating point constants are required
1767(and the only time that they are generated by the disassembler) is when a
1768floating point constant must be emitted but it cannot be represented as a
1769decimal floating point number. For example, NaN's, infinities, and other
1770special values are represented in their IEEE hexadecimal format so that
1771assembly and disassembly do not cause any bits to change in the constants.</p>
1772
1773</div>
1774
1775<!-- ======================================================================= -->
1776<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1777</div>
1778
1779<div class="doc_text">
1780<p>Aggregate constants arise from aggregation of simple constants
1781and smaller aggregate constants.</p>
1782
1783<dl>
1784 <dt><b>Structure constants</b></dt>
1785
1786 <dd>Structure constants are represented with notation similar to structure
1787 type definitions (a comma separated list of elements, surrounded by braces
Chris Lattner6c8de962007-12-25 20:34:52 +0000[diff] [blame]1788 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1789 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]1790 must have <a href="#t_struct">structure type</a>, and the number and
1791 types of elements must match those specified by the type.
1792 </dd>
1793
1794 <dt><b>Array constants</b></dt>
1795
1796 <dd>Array constants are represented with notation similar to array type
1797 definitions (a comma separated list of elements, surrounded by square brackets
1798 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1799 constants must have <a href="#t_array">array type</a>, and the number and
1800 types of elements must match those specified by the type.
1801 </dd>
1802
1803 <dt><b>Vector constants</b></dt>
1804
1805 <dd>Vector constants are represented with notation similar to vector type
1806 definitions (a comma separated list of elements, surrounded by
1807 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1808 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1809 href="#t_vector">vector type</a>, and the number and types of elements must
1810 match those specified by the type.
1811 </dd>
1812
1813 <dt><b>Zero initialization</b></dt>
1814
1815 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1816 value to zero of <em>any</em> type, including scalar and aggregate types.
1817 This is often used to avoid having to print large zero initializers (e.g. for
1818 large arrays) and is always exactly equivalent to using explicit zero
1819 initializers.
1820 </dd>
1821</dl>
1822
1823</div>
1824
1825<!-- ======================================================================= -->
1826<div class="doc_subsection">
1827 <a name="globalconstants">Global Variable and Function Addresses</a>
1828</div>
1829
1830<div class="doc_text">
1831
1832<p>The addresses of <a href="#globalvars">global variables</a> and <a
1833href="#functionstructure">functions</a> are always implicitly valid (link-time)
1834constants. These constants are explicitly referenced when the <a
1835href="#identifiers">identifier for the global</a> is used and always have <a
1836href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1837file:</p>
1838
1839<div class="doc_code">
1840<pre>
1841@X = global i32 17
1842@Y = global i32 42
1843@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1844</pre>
1845</div>
1846
1847</div>
1848
1849<!-- ======================================================================= -->
1850<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1851<div class="doc_text">
1852 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1853 no specific value. Undefined values may be of any type and be used anywhere
1854 a constant is permitted.</p>
1855
1856 <p>Undefined values indicate to the compiler that the program is well defined
1857 no matter what value is used, giving the compiler more freedom to optimize.
1858 </p>
1859</div>
1860
1861<!-- ======================================================================= -->
1862<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1863</div>
1864
1865<div class="doc_text">
1866
1867<p>Constant expressions are used to allow expressions involving other constants
1868to be used as constants. Constant expressions may be of any <a
1869href="#t_firstclass">first class</a> type and may involve any LLVM operation
1870that does not have side effects (e.g. load and call are not supported). The
1871following is the syntax for constant expressions:</p>
1872
1873<dl>
1874 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1875 <dd>Truncate a constant to another type. The bit size of CST must be larger
1876 than the bit size of TYPE. Both types must be integers.</dd>
1877
1878 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1879 <dd>Zero extend a constant to another type. The bit size of CST must be
1880 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1881
1882 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1883 <dd>Sign extend a constant to another type. The bit size of CST must be
1884 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1885
1886 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1887 <dd>Truncate a floating point constant to another floating point type. The
1888 size of CST must be larger than the size of TYPE. Both types must be
1889 floating point.</dd>
1890
1891 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1892 <dd>Floating point extend a constant to another type. The size of CST must be
1893 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1894
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1895 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1896 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1897 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1898 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1899 of the same number of elements. If the value won't fit in the integer type,
1900 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1901
1902 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1903 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1904 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1905 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1906 of the same number of elements. If the value won't fit in the integer type,
1907 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1908
1909 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1910 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1911 constant. TYPE must be a scalar or vector floating point type. CST must be of
1912 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1913 of the same number of elements. If the value won't fit in the floating point
1914 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1915
1916 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1917 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1918 constant. TYPE must be a scalar or vector floating point type. CST must be of
1919 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1920 of the same number of elements. If the value won't fit in the floating point
1921 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1922
1923 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1924 <dd>Convert a pointer typed constant to the corresponding integer constant
1925 TYPE must be an integer type. CST must be of pointer type. The CST value is
1926 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1927
1928 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1929 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1930 pointer type. CST must be of integer type. The CST value is zero extended,
1931 truncated, or unchanged to make it fit in a pointer size. This one is
1932 <i>really</i> dangerous!</dd>
1933
1934 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1935 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1936 identical (same number of bits). The conversion is done as if the CST value
1937 was stored to memory and read back as TYPE. In other words, no bits change
1938 with this operator, just the type. This can be used for conversion of
1939 vector types to any other type, as long as they have the same bit width. For
Dan Gohman7305fa02008-09-08 16:45:59 +0000[diff] [blame]1940 pointers it is only valid to cast to another pointer type. It is not valid
1941 to bitcast to or from an aggregate type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1942 </dd>
1943
1944 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1945
1946 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1947 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1948 instruction, the index list may have zero or more indexes, which are required
1949 to make sense for the type of "CSTPTR".</dd>
1950
1951 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1952
1953 <dd>Perform the <a href="#i_select">select operation</a> on
1954 constants.</dd>
1955
1956 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1957 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1958
1959 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1960 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1961
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]1962 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
1963 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
1964
1965 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
1966 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
1967
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1968 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1969
1970 <dd>Perform the <a href="#i_extractelement">extractelement
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]1971 operation</a> on constants.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1972
1973 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1974
1975 <dd>Perform the <a href="#i_insertelement">insertelement
1976 operation</a> on constants.</dd>
1977
1978
1979 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1980
1981 <dd>Perform the <a href="#i_shufflevector">shufflevector
1982 operation</a> on constants.</dd>
1983
1984 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1985
1986 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1987 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1988 binary</a> operations. The constraints on operands are the same as those for
1989 the corresponding instruction (e.g. no bitwise operations on floating point
1990 values are allowed).</dd>
1991</dl>
1992</div>
1993
1994<!-- *********************************************************************** -->
1995<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1996<!-- *********************************************************************** -->
1997
1998<!-- ======================================================================= -->
1999<div class="doc_subsection">
2000<a name="inlineasm">Inline Assembler Expressions</a>
2001</div>
2002
2003<div class="doc_text">
2004
2005<p>
2006LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
2007Module-Level Inline Assembly</a>) through the use of a special value. This
2008value represents the inline assembler as a string (containing the instructions
2009to emit), a list of operand constraints (stored as a string), and a flag that
2010indicates whether or not the inline asm expression has side effects. An example
2011inline assembler expression is:
2012</p>
2013
2014<div class="doc_code">
2015<pre>
2016i32 (i32) asm "bswap $0", "=r,r"
2017</pre>
2018</div>
2019
2020<p>
2021Inline assembler expressions may <b>only</b> be used as the callee operand of
2022a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
2023</p>
2024
2025<div class="doc_code">
2026<pre>
2027%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2028</pre>
2029</div>
2030
2031<p>
2032Inline asms with side effects not visible in the constraint list must be marked
2033as having side effects. This is done through the use of the
2034'<tt>sideeffect</tt>' keyword, like so:
2035</p>
2036
2037<div class="doc_code">
2038<pre>
2039call void asm sideeffect "eieio", ""()
2040</pre>
2041</div>
2042
2043<p>TODO: The format of the asm and constraints string still need to be
2044documented here. Constraints on what can be done (e.g. duplication, moving, etc
Chris Lattner1cdd64e2008-10-04 18:36:02 +0000[diff] [blame]2045need to be documented). This is probably best done by reference to another
2046document that covers inline asm from a holistic perspective.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2047</p>
2048
2049</div>
2050
2051<!-- *********************************************************************** -->
2052<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2053<!-- *********************************************************************** -->
2054
2055<div class="doc_text">
2056
2057<p>The LLVM instruction set consists of several different
2058classifications of instructions: <a href="#terminators">terminator
2059instructions</a>, <a href="#binaryops">binary instructions</a>,
2060<a href="#bitwiseops">bitwise binary instructions</a>, <a
2061 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
2062instructions</a>.</p>
2063
2064</div>
2065
2066<!-- ======================================================================= -->
2067<div class="doc_subsection"> <a name="terminators">Terminator
2068Instructions</a> </div>
2069
2070<div class="doc_text">
2071
2072<p>As mentioned <a href="#functionstructure">previously</a>, every
2073basic block in a program ends with a "Terminator" instruction, which
2074indicates which block should be executed after the current block is
2075finished. These terminator instructions typically yield a '<tt>void</tt>'
2076value: they produce control flow, not values (the one exception being
2077the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2078<p>There are six different terminator instructions: the '<a
2079 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
2080instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
2081the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
2082 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
2083 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2084
2085</div>
2086
2087<!-- _______________________________________________________________________ -->
2088<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2089Instruction</a> </div>
2090<div class="doc_text">
2091<h5>Syntax:</h5>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2092<pre>
2093 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]2094 ret void <i>; Return from void function</i>
2095</pre>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2096
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2097<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2098
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2099<p>The '<tt>ret</tt>' instruction is used to return control flow (and
2100optionally a value) from a function back to the caller.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2101<p>There are two forms of the '<tt>ret</tt>' instruction: one that
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2102returns a value and then causes control flow, and one that just causes
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2103control flow to occur.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2104
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2105<h5>Arguments:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2106
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2107<p>The '<tt>ret</tt>' instruction optionally accepts a single argument,
2108the return value. The type of the return value must be a
2109'<a href="#t_firstclass">first class</a>' type.</p>
2110
2111<p>A function is not <a href="#wellformed">well formed</a> if
2112it it has a non-void return type and contains a '<tt>ret</tt>'
2113instruction with no return value or a return value with a type that
2114does not match its type, or if it has a void return type and contains
2115a '<tt>ret</tt>' instruction with a return value.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2116
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2117<h5>Semantics:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2118
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2119<p>When the '<tt>ret</tt>' instruction is executed, control flow
2120returns back to the calling function's context. If the caller is a "<a
2121 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
2122the instruction after the call. If the caller was an "<a
2123 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
2124at the beginning of the "normal" destination block. If the instruction
2125returns a value, that value shall set the call or invoke instruction's
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]2126return value.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2127
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2128<h5>Example:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]2129
2130<pre>
2131 ret i32 5 <i>; Return an integer value of 5</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2132 ret void <i>; Return from a void function</i>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]2133 ret { i32, i8 } { i32 4, i8 2 } <i>; Return an aggregate of values 4 and 2</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2134</pre>
Dan Gohman60967192009-01-12 23:12:39 +0000[diff] [blame]2135
djge93155c2009-01-24 15:58:40 +0000[diff] [blame]2136<p>Note that the code generator does not yet fully support large
2137 return values. The specific sizes that are currently supported are
2138 dependent on the target. For integers, on 32-bit targets the limit
2139 is often 64 bits, and on 64-bit targets the limit is often 128 bits.
2140 For aggregate types, the current limits are dependent on the element
2141 types; for example targets are often limited to 2 total integer
2142 elements and 2 total floating-point elements.</p>
Dan Gohman60967192009-01-12 23:12:39 +0000[diff] [blame]2143
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2144</div>
2145<!-- _______________________________________________________________________ -->
2146<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2147<div class="doc_text">
2148<h5>Syntax:</h5>
2149<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
2150</pre>
2151<h5>Overview:</h5>
2152<p>The '<tt>br</tt>' instruction is used to cause control flow to
2153transfer to a different basic block in the current function. There are
2154two forms of this instruction, corresponding to a conditional branch
2155and an unconditional branch.</p>
2156<h5>Arguments:</h5>
2157<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
2158single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
2159unconditional form of the '<tt>br</tt>' instruction takes a single
2160'<tt>label</tt>' value as a target.</p>
2161<h5>Semantics:</h5>
2162<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2163argument is evaluated. If the value is <tt>true</tt>, control flows
2164to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2165control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2166<h5>Example:</h5>
2167<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
2168 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
2169</div>
2170<!-- _______________________________________________________________________ -->
2171<div class="doc_subsubsection">
2172 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2173</div>
2174
2175<div class="doc_text">
2176<h5>Syntax:</h5>
2177
2178<pre>
2179 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
2180</pre>
2181
2182<h5>Overview:</h5>
2183
2184<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2185several different places. It is a generalization of the '<tt>br</tt>'
2186instruction, allowing a branch to occur to one of many possible
2187destinations.</p>
2188
2189
2190<h5>Arguments:</h5>
2191
2192<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2193comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
2194an array of pairs of comparison value constants and '<tt>label</tt>'s. The
2195table is not allowed to contain duplicate constant entries.</p>
2196
2197<h5>Semantics:</h5>
2198
2199<p>The <tt>switch</tt> instruction specifies a table of values and
2200destinations. When the '<tt>switch</tt>' instruction is executed, this
2201table is searched for the given value. If the value is found, control flow is
2202transfered to the corresponding destination; otherwise, control flow is
2203transfered to the default destination.</p>
2204
2205<h5>Implementation:</h5>
2206
2207<p>Depending on properties of the target machine and the particular
2208<tt>switch</tt> instruction, this instruction may be code generated in different
2209ways. For example, it could be generated as a series of chained conditional
2210branches or with a lookup table.</p>
2211
2212<h5>Example:</h5>
2213
2214<pre>
2215 <i>; Emulate a conditional br instruction</i>
2216 %Val = <a href="#i_zext">zext</a> i1 %value to i32
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]2217 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2218
2219 <i>; Emulate an unconditional br instruction</i>
2220 switch i32 0, label %dest [ ]
2221
2222 <i>; Implement a jump table:</i>
Dan Gohman01852382009-01-04 23:44:43 +0000[diff] [blame]2223 switch i32 %val, label %otherwise [ i32 0, label %onzero
2224 i32 1, label %onone
2225 i32 2, label %ontwo ]
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2226</pre>
2227</div>
2228
2229<!-- _______________________________________________________________________ -->
2230<div class="doc_subsubsection">
2231 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2232</div>
2233
2234<div class="doc_text">
2235
2236<h5>Syntax:</h5>
2237
2238<pre>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]2239 &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]2240 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
2241</pre>
2242
2243<h5>Overview:</h5>
2244
2245<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2246function, with the possibility of control flow transfer to either the
2247'<tt>normal</tt>' label or the
2248'<tt>exception</tt>' label. If the callee function returns with the
2249"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2250"normal" label. If the callee (or any indirect callees) returns with the "<a
2251href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]2252continued at the dynamically nearest "exception" label.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2253
2254<h5>Arguments:</h5>
2255
2256<p>This instruction requires several arguments:</p>
2257
2258<ol>
2259 <li>
2260 The optional "cconv" marker indicates which <a href="#callingconv">calling
2261 convention</a> the call should use. If none is specified, the call defaults
2262 to using C calling conventions.
2263 </li>
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]2264
2265 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2266 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
2267 and '<tt>inreg</tt>' attributes are valid here.</li>
2268
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2269 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2270 function value being invoked. In most cases, this is a direct function
2271 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2272 an arbitrary pointer to function value.
2273 </li>
2274
2275 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2276 function to be invoked. </li>
2277
2278 <li>'<tt>function args</tt>': argument list whose types match the function
2279 signature argument types. If the function signature indicates the function
2280 accepts a variable number of arguments, the extra arguments can be
2281 specified. </li>
2282
2283 <li>'<tt>normal label</tt>': the label reached when the called function
2284 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2285
2286 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2287 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2288
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]2289 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]2290 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2291 '<tt>readnone</tt>' attributes are valid here.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2292</ol>
2293
2294<h5>Semantics:</h5>
2295
2296<p>This instruction is designed to operate as a standard '<tt><a
2297href="#i_call">call</a></tt>' instruction in most regards. The primary
2298difference is that it establishes an association with a label, which is used by
2299the runtime library to unwind the stack.</p>
2300
2301<p>This instruction is used in languages with destructors to ensure that proper
2302cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2303exception. Additionally, this is important for implementation of
2304'<tt>catch</tt>' clauses in high-level languages that support them.</p>
2305
2306<h5>Example:</h5>
2307<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2308 %retval = invoke i32 @Test(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2309 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2310 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2311 unwind label %TestCleanup <i>; {i32}:retval set</i>
2312</pre>
2313</div>
2314
2315
2316<!-- _______________________________________________________________________ -->
2317
2318<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2319Instruction</a> </div>
2320
2321<div class="doc_text">
2322
2323<h5>Syntax:</h5>
2324<pre>
2325 unwind
2326</pre>
2327
2328<h5>Overview:</h5>
2329
2330<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2331at the first callee in the dynamic call stack which used an <a
2332href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2333primarily used to implement exception handling.</p>
2334
2335<h5>Semantics:</h5>
2336
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2337<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2338immediately halt. The dynamic call stack is then searched for the first <a
2339href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2340execution continues at the "exceptional" destination block specified by the
2341<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2342dynamic call chain, undefined behavior results.</p>
2343</div>
2344
2345<!-- _______________________________________________________________________ -->
2346
2347<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2348Instruction</a> </div>
2349
2350<div class="doc_text">
2351
2352<h5>Syntax:</h5>
2353<pre>
2354 unreachable
2355</pre>
2356
2357<h5>Overview:</h5>
2358
2359<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2360instruction is used to inform the optimizer that a particular portion of the
2361code is not reachable. This can be used to indicate that the code after a
2362no-return function cannot be reached, and other facts.</p>
2363
2364<h5>Semantics:</h5>
2365
2366<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2367</div>
2368
2369
2370
2371<!-- ======================================================================= -->
2372<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2373<div class="doc_text">
2374<p>Binary operators are used to do most of the computation in a
Chris Lattnerab596d92008-04-01 18:47:32 +0000[diff] [blame]2375program. They require two operands of the same type, execute an operation on them, and
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2376produce a single value. The operands might represent
2377multiple 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]2378The result value has the same type as its operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2379<p>There are several different binary operators:</p>
2380</div>
2381<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2382<div class="doc_subsubsection">
2383 <a name="i_add">'<tt>add</tt>' Instruction</a>
2384</div>
2385
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2386<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2387
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2388<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2389
2390<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2391 &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]2392</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2393
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2394<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2395
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2396<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2397
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2398<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2399
2400<p>The two arguments to the '<tt>add</tt>' instruction must be <a
2401 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>, or
2402 <a href="#t_vector">vector</a> values. Both arguments must have identical
2403 types.</p>
2404
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2405<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2406
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2407<p>The value produced is the integer or floating point sum of the two
2408operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2409
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2410<p>If an integer sum has unsigned overflow, the result returned is the
2411mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2412the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2413
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2414<p>Because LLVM integers use a two's complement representation, this
2415instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2416
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2417<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2418
2419<pre>
2420 &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2421</pre>
2422</div>
2423<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2424<div class="doc_subsubsection">
2425 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2426</div>
2427
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2428<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2429
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2430<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2431
2432<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2433 &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]2434</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2435
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2436<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2437
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2438<p>The '<tt>sub</tt>' instruction returns the difference of its two
2439operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2440
2441<p>Note that the '<tt>sub</tt>' instruction is used to represent the
2442'<tt>neg</tt>' instruction present in most other intermediate
2443representations.</p>
2444
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2445<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2446
2447<p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2448 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2449 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2450 types.</p>
2451
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2452<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2453
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2454<p>The value produced is the integer or floating point difference of
2455the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2456
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2457<p>If an integer difference has unsigned overflow, the result returned is the
2458mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2459the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2460
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2461<p>Because LLVM integers use a two's complement representation, this
2462instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2463
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2464<h5>Example:</h5>
2465<pre>
2466 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2467 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2468</pre>
2469</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2470
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2471<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2472<div class="doc_subsubsection">
2473 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2474</div>
2475
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2476<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2477
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2478<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2479<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]2480</pre>
2481<h5>Overview:</h5>
2482<p>The '<tt>mul</tt>' instruction returns the product of its two
2483operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2484
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2485<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2486
2487<p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2488href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2489or <a href="#t_vector">vector</a> values. Both arguments must have identical
2490types.</p>
2491
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2492<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2493
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2494<p>The value produced is the integer or floating point product of the
2495two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2496
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2497<p>If the result of an integer multiplication has unsigned overflow,
2498the result returned is the mathematical result modulo
24992<sup>n</sup>, where n is the bit width of the result.</p>
2500<p>Because LLVM integers use a two's complement representation, and the
2501result is the same width as the operands, this instruction returns the
2502correct result for both signed and unsigned integers. If a full product
2503(e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2504should be sign-extended or zero-extended as appropriate to the
2505width of the full product.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2506<h5>Example:</h5>
2507<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2508</pre>
2509</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2510
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2511<!-- _______________________________________________________________________ -->
2512<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2513</a></div>
2514<div class="doc_text">
2515<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2516<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]2517</pre>
2518<h5>Overview:</h5>
2519<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2520operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2521
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2522<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2523
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2524<p>The two arguments to the '<tt>udiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2525<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2526values. Both arguments must have identical types.</p>
2527
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2528<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2529
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2530<p>The value produced is the unsigned integer quotient of the two operands.</p>
2531<p>Note that unsigned integer division and signed integer division are distinct
2532operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2533<p>Division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2534<h5>Example:</h5>
2535<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2536</pre>
2537</div>
2538<!-- _______________________________________________________________________ -->
2539<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2540</a> </div>
2541<div class="doc_text">
2542<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2543<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2544 &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]2545</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2546
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2547<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2548
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2549<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2550operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2551
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2552<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2553
2554<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2555<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2556values. Both arguments must have identical types.</p>
2557
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2558<h5>Semantics:</h5>
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2559<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]2560<p>Note that signed integer division and unsigned integer division are distinct
2561operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2562<p>Division by zero leads to undefined behavior. Overflow also leads to
2563undefined behavior; this is a rare case, but can occur, for example,
2564by doing a 32-bit division of -2147483648 by -1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2565<h5>Example:</h5>
2566<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2567</pre>
2568</div>
2569<!-- _______________________________________________________________________ -->
2570<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2571Instruction</a> </div>
2572<div class="doc_text">
2573<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2574<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2575 &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]2576</pre>
2577<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2578
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2579<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2580operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2581
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2582<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2583
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2584<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2585<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2586of floating point values. Both arguments must have identical types.</p>
2587
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2588<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2589
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2590<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2591
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2592<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2593
2594<pre>
2595 &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]2596</pre>
2597</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2598
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2599<!-- _______________________________________________________________________ -->
2600<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2601</div>
2602<div class="doc_text">
2603<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2604<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]2605</pre>
2606<h5>Overview:</h5>
2607<p>The '<tt>urem</tt>' instruction returns the remainder from the
2608unsigned division of its two arguments.</p>
2609<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2610<p>The two arguments to the '<tt>urem</tt>' instruction must be
2611<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2612values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2613<h5>Semantics:</h5>
2614<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2615This instruction always performs an unsigned division to get the remainder.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2616<p>Note that unsigned integer remainder and signed integer remainder are
2617distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2618<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2619<h5>Example:</h5>
2620<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2621</pre>
2622
2623</div>
2624<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2625<div class="doc_subsubsection">
2626 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2627</div>
2628
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2629<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2630
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2631<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2632
2633<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2634 &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]2635</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2636
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2637<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2638
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2639<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2640signed division of its two operands. This instruction can also take
2641<a href="#t_vector">vector</a> versions of the values in which case
2642the elements must be integers.</p>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]2643
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2644<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2645
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2646<p>The two arguments to the '<tt>srem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2647<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2648values. Both arguments must have identical types.</p>
2649
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2650<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2651
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2652<p>This instruction returns the <i>remainder</i> of a division (where the result
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2653has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2654operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2655a value. For more information about the difference, see <a
2656 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2657Math Forum</a>. For a table of how this is implemented in various languages,
2658please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2659Wikipedia: modulo operation</a>.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2660<p>Note that signed integer remainder and unsigned integer remainder are
2661distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2662<p>Taking the remainder of a division by zero leads to undefined behavior.
2663Overflow also leads to undefined behavior; this is a rare case, but can occur,
2664for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2665(The remainder doesn't actually overflow, but this rule lets srem be
2666implemented using instructions that return both the result of the division
2667and the remainder.)</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2668<h5>Example:</h5>
2669<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2670</pre>
2671
2672</div>
2673<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2674<div class="doc_subsubsection">
2675 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2676
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2677<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2678
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2679<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2680<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]2681</pre>
2682<h5>Overview:</h5>
2683<p>The '<tt>frem</tt>' instruction returns the remainder from the
2684division of its two operands.</p>
2685<h5>Arguments:</h5>
2686<p>The two arguments to the '<tt>frem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2687<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2688of floating point values. Both arguments must have identical types.</p>
2689
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2690<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2691
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2692<p>This instruction returns the <i>remainder</i> of a division.
2693The remainder has the same sign as the dividend.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2694
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2695<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2696
2697<pre>
2698 &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]2699</pre>
2700</div>
2701
2702<!-- ======================================================================= -->
2703<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2704Operations</a> </div>
2705<div class="doc_text">
2706<p>Bitwise binary operators are used to do various forms of
2707bit-twiddling in a program. They are generally very efficient
2708instructions and can commonly be strength reduced from other
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2709instructions. They require two operands of the same type, execute an operation on them,
2710and 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]2711</div>
2712
2713<!-- _______________________________________________________________________ -->
2714<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2715Instruction</a> </div>
2716<div class="doc_text">
2717<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2718<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]2719</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2720
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2721<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2722
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2723<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2724the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2725
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2726<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2727
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2728<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2729 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2730type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2731
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2732<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2733
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2734<p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2735where 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]2736equal to or larger than the number of bits in <tt>op1</tt>, the result is undefined.
2737If the arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2738corresponding shift amount in <tt>op2</tt>.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2739
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2740<h5>Example:</h5><pre>
2741 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2742 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2743 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2744 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Mon P Wang9901e732008-12-09 05:46:39 +0000[diff] [blame]2745 &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]2746</pre>
2747</div>
2748<!-- _______________________________________________________________________ -->
2749<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2750Instruction</a> </div>
2751<div class="doc_text">
2752<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2753<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]2754</pre>
2755
2756<h5>Overview:</h5>
2757<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2758operand shifted to the right a specified number of bits with zero fill.</p>
2759
2760<h5>Arguments:</h5>
2761<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2762<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]2763type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2764
2765<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2766
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2767<p>This instruction always performs a logical shift right operation. The most
2768significant bits of the result will be filled with zero bits after the
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2769shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
Mon P Wangb0f51822008-12-10 08:55:09 +0000[diff] [blame]2770the number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
2771vectors, each vector element of <tt>op1</tt> is shifted by the corresponding shift
2772amount in <tt>op2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2773
2774<h5>Example:</h5>
2775<pre>
2776 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2777 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2778 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2779 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2780 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Mon P Wang9901e732008-12-09 05:46:39 +0000[diff] [blame]2781 &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]2782</pre>
2783</div>
2784
2785<!-- _______________________________________________________________________ -->
2786<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2787Instruction</a> </div>
2788<div class="doc_text">
2789
2790<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2791<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]2792</pre>
2793
2794<h5>Overview:</h5>
2795<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2796operand shifted to the right a specified number of bits with sign extension.</p>
2797
2798<h5>Arguments:</h5>
2799<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2800<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]2801type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2802
2803<h5>Semantics:</h5>
2804<p>This instruction always performs an arithmetic shift right operation,
2805The most significant bits of the result will be filled with the sign bit
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2806of <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]2807larger than the number of bits in <tt>op1</tt>, the result is undefined. If the
2808arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2809corresponding shift amount in <tt>op2</tt>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2810
2811<h5>Example:</h5>
2812<pre>
2813 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2814 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2815 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2816 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2817 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Mon P Wang9901e732008-12-09 05:46:39 +0000[diff] [blame]2818 &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]2819</pre>
2820</div>
2821
2822<!-- _______________________________________________________________________ -->
2823<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2824Instruction</a> </div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2825
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2826<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2827
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2828<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2829
2830<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2831 &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]2832</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2833
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2834<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2835
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2836<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2837its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2838
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2839<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2840
2841<p>The two arguments to the '<tt>and</tt>' instruction must be
2842<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2843values. Both arguments must have identical types.</p>
2844
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2845<h5>Semantics:</h5>
2846<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2847<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2848<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2849<table border="1" cellspacing="0" cellpadding="4">
2850 <tbody>
2851 <tr>
2852 <td>In0</td>
2853 <td>In1</td>
2854 <td>Out</td>
2855 </tr>
2856 <tr>
2857 <td>0</td>
2858 <td>0</td>
2859 <td>0</td>
2860 </tr>
2861 <tr>
2862 <td>0</td>
2863 <td>1</td>
2864 <td>0</td>
2865 </tr>
2866 <tr>
2867 <td>1</td>
2868 <td>0</td>
2869 <td>0</td>
2870 </tr>
2871 <tr>
2872 <td>1</td>
2873 <td>1</td>
2874 <td>1</td>
2875 </tr>
2876 </tbody>
2877</table>
2878</div>
2879<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2880<pre>
2881 &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]2882 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2883 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2884</pre>
2885</div>
2886<!-- _______________________________________________________________________ -->
2887<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2888<div class="doc_text">
2889<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2890<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]2891</pre>
2892<h5>Overview:</h5>
2893<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2894or of its two operands.</p>
2895<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2896
2897<p>The two arguments to the '<tt>or</tt>' instruction must be
2898<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2899values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2900<h5>Semantics:</h5>
2901<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2902<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2903<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2904<table border="1" cellspacing="0" cellpadding="4">
2905 <tbody>
2906 <tr>
2907 <td>In0</td>
2908 <td>In1</td>
2909 <td>Out</td>
2910 </tr>
2911 <tr>
2912 <td>0</td>
2913 <td>0</td>
2914 <td>0</td>
2915 </tr>
2916 <tr>
2917 <td>0</td>
2918 <td>1</td>
2919 <td>1</td>
2920 </tr>
2921 <tr>
2922 <td>1</td>
2923 <td>0</td>
2924 <td>1</td>
2925 </tr>
2926 <tr>
2927 <td>1</td>
2928 <td>1</td>
2929 <td>1</td>
2930 </tr>
2931 </tbody>
2932</table>
2933</div>
2934<h5>Example:</h5>
2935<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2936 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2937 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2938</pre>
2939</div>
2940<!-- _______________________________________________________________________ -->
2941<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2942Instruction</a> </div>
2943<div class="doc_text">
2944<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2945<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]2946</pre>
2947<h5>Overview:</h5>
2948<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2949or of its two operands. The <tt>xor</tt> is used to implement the
2950"one's complement" operation, which is the "~" operator in C.</p>
2951<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2952<p>The two arguments to the '<tt>xor</tt>' instruction must be
2953<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2954values. Both arguments must have identical types.</p>
2955
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2956<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2957
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2958<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2959<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2960<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2961<table border="1" cellspacing="0" cellpadding="4">
2962 <tbody>
2963 <tr>
2964 <td>In0</td>
2965 <td>In1</td>
2966 <td>Out</td>
2967 </tr>
2968 <tr>
2969 <td>0</td>
2970 <td>0</td>
2971 <td>0</td>
2972 </tr>
2973 <tr>
2974 <td>0</td>
2975 <td>1</td>
2976 <td>1</td>
2977 </tr>
2978 <tr>
2979 <td>1</td>
2980 <td>0</td>
2981 <td>1</td>
2982 </tr>
2983 <tr>
2984 <td>1</td>
2985 <td>1</td>
2986 <td>0</td>
2987 </tr>
2988 </tbody>
2989</table>
2990</div>
2991<p> </p>
2992<h5>Example:</h5>
2993<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2994 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2995 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2996 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2997</pre>
2998</div>
2999
3000<!-- ======================================================================= -->
3001<div class="doc_subsection">
3002 <a name="vectorops">Vector Operations</a>
3003</div>
3004
3005<div class="doc_text">
3006
3007<p>LLVM supports several instructions to represent vector operations in a
3008target-independent manner. These instructions cover the element-access and
3009vector-specific operations needed to process vectors effectively. While LLVM
3010does directly support these vector operations, many sophisticated algorithms
3011will want to use target-specific intrinsics to take full advantage of a specific
3012target.</p>
3013
3014</div>
3015
3016<!-- _______________________________________________________________________ -->
3017<div class="doc_subsubsection">
3018 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3019</div>
3020
3021<div class="doc_text">
3022
3023<h5>Syntax:</h5>
3024
3025<pre>
3026 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
3027</pre>
3028
3029<h5>Overview:</h5>
3030
3031<p>
3032The '<tt>extractelement</tt>' instruction extracts a single scalar
3033element from a vector at a specified index.
3034</p>
3035
3036
3037<h5>Arguments:</h5>
3038
3039<p>
3040The first operand of an '<tt>extractelement</tt>' instruction is a
3041value of <a href="#t_vector">vector</a> type. The second operand is
3042an index indicating the position from which to extract the element.
3043The index may be a variable.</p>
3044
3045<h5>Semantics:</h5>
3046
3047<p>
3048The result is a scalar of the same type as the element type of
3049<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3050<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3051results are undefined.
3052</p>
3053
3054<h5>Example:</h5>
3055
3056<pre>
3057 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
3058</pre>
3059</div>
3060
3061
3062<!-- _______________________________________________________________________ -->
3063<div class="doc_subsubsection">
3064 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
3065</div>
3066
3067<div class="doc_text">
3068
3069<h5>Syntax:</h5>
3070
3071<pre>
Dan Gohmanbcc3c502008-05-12 23:38:42 +0000[diff] [blame]3072 &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]3073</pre>
3074
3075<h5>Overview:</h5>
3076
3077<p>
3078The '<tt>insertelement</tt>' instruction inserts a scalar
3079element into a vector at a specified index.
3080</p>
3081
3082
3083<h5>Arguments:</h5>
3084
3085<p>
3086The first operand of an '<tt>insertelement</tt>' instruction is a
3087value of <a href="#t_vector">vector</a> type. The second operand is a
3088scalar value whose type must equal the element type of the first
3089operand. The third operand is an index indicating the position at
3090which to insert the value. The index may be a variable.</p>
3091
3092<h5>Semantics:</h5>
3093
3094<p>
3095The result is a vector of the same type as <tt>val</tt>. Its
3096element values are those of <tt>val</tt> except at position
3097<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
3098exceeds the length of <tt>val</tt>, the results are undefined.
3099</p>
3100
3101<h5>Example:</h5>
3102
3103<pre>
3104 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
3105</pre>
3106</div>
3107
3108<!-- _______________________________________________________________________ -->
3109<div class="doc_subsubsection">
3110 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3111</div>
3112
3113<div class="doc_text">
3114
3115<h5>Syntax:</h5>
3116
3117<pre>
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3118 &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]3119</pre>
3120
3121<h5>Overview:</h5>
3122
3123<p>
3124The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3125from two input vectors, returning a vector with the same element type as
3126the input and length that is the same as the shuffle mask.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3127</p>
3128
3129<h5>Arguments:</h5>
3130
3131<p>
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3132The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3133with types that match each other. The third argument is a shuffle mask whose
3134element type is always 'i32'. The result of the instruction is a vector whose
3135length is the same as the shuffle mask and whose element type is the same as
3136the element type of the first two operands.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3137</p>
3138
3139<p>
3140The shuffle mask operand is required to be a constant vector with either
3141constant integer or undef values.
3142</p>
3143
3144<h5>Semantics:</h5>
3145
3146<p>
3147The elements of the two input vectors are numbered from left to right across
3148both of the vectors. The shuffle mask operand specifies, for each element of
Mon P Wangbff5d9c2008-11-10 04:46:22 +0000[diff] [blame]3149the result vector, which element of the two input vectors the result element
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3150gets. The element selector may be undef (meaning "don't care") and the second
3151operand may be undef if performing a shuffle from only one vector.
3152</p>
3153
3154<h5>Example:</h5>
3155
3156<pre>
3157 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
3158 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
3159 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
3160 &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]3161 %result = shufflevector &lt;8 x i32&gt; %v1, &lt;8 x i32&gt; undef,
3162 &lt;4 x i32&gt; &lt;i32 0, i32 1, i32 2, i32 3&gt; <i>; yields &lt;4 x i32&gt;</i>
3163 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
3164 &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]3165</pre>
3166</div>
3167
3168
3169<!-- ======================================================================= -->
3170<div class="doc_subsection">
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3171 <a name="aggregateops">Aggregate Operations</a>
3172</div>
3173
3174<div class="doc_text">
3175
3176<p>LLVM supports several instructions for working with aggregate values.
3177</p>
3178
3179</div>
3180
3181<!-- _______________________________________________________________________ -->
3182<div class="doc_subsubsection">
3183 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3184</div>
3185
3186<div class="doc_text">
3187
3188<h5>Syntax:</h5>
3189
3190<pre>
3191 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
3192</pre>
3193
3194<h5>Overview:</h5>
3195
3196<p>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3197The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3198or array element from an aggregate value.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3199</p>
3200
3201
3202<h5>Arguments:</h5>
3203
3204<p>
3205The first operand of an '<tt>extractvalue</tt>' instruction is a
3206value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3207type. The operands are constant indices to specify which value to extract
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3208in a similar manner as indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3209'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3210</p>
3211
3212<h5>Semantics:</h5>
3213
3214<p>
3215The result is the value at the position in the aggregate specified by
3216the index operands.
3217</p>
3218
3219<h5>Example:</h5>
3220
3221<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3222 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3223</pre>
3224</div>
3225
3226
3227<!-- _______________________________________________________________________ -->
3228<div class="doc_subsubsection">
3229 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3230</div>
3231
3232<div class="doc_text">
3233
3234<h5>Syntax:</h5>
3235
3236<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3237 &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]3238</pre>
3239
3240<h5>Overview:</h5>
3241
3242<p>
3243The '<tt>insertvalue</tt>' instruction inserts a value
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3244into a struct field or array element in an aggregate.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3245</p>
3246
3247
3248<h5>Arguments:</h5>
3249
3250<p>
3251The first operand of an '<tt>insertvalue</tt>' instruction is a
3252value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3253The second operand is a first-class value to insert.
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]3254The following operands are constant indices
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3255indicating the position at which to insert the value in a similar manner as
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3256indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3257'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3258The value to insert must have the same type as the value identified
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3259by the indices.
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]3260</p>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3261
3262<h5>Semantics:</h5>
3263
3264<p>
3265The result is an aggregate of the same type as <tt>val</tt>. Its
3266value is that of <tt>val</tt> except that the value at the position
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3267specified by the indices is that of <tt>elt</tt>.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3268</p>
3269
3270<h5>Example:</h5>
3271
3272<pre>
Dan Gohmanb1aab4e2008-06-23 15:26:37 +0000[diff] [blame]3273 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3274</pre>
3275</div>
3276
3277
3278<!-- ======================================================================= -->
3279<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3280 <a name="memoryops">Memory Access and Addressing Operations</a>
3281</div>
3282
3283<div class="doc_text">
3284
3285<p>A key design point of an SSA-based representation is how it
3286represents memory. In LLVM, no memory locations are in SSA form, which
3287makes things very simple. This section describes how to read, write,
3288allocate, and free memory in LLVM.</p>
3289
3290</div>
3291
3292<!-- _______________________________________________________________________ -->
3293<div class="doc_subsubsection">
3294 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3295</div>
3296
3297<div class="doc_text">
3298
3299<h5>Syntax:</h5>
3300
3301<pre>
3302 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3303</pre>
3304
3305<h5>Overview:</h5>
3306
3307<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3308heap and returns a pointer to it. The object is always allocated in the generic
3309address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3310
3311<h5>Arguments:</h5>
3312
3313<p>The '<tt>malloc</tt>' instruction allocates
3314<tt>sizeof(&lt;type&gt;)*NumElements</tt>
3315bytes of memory from the operating system and returns a pointer of the
3316appropriate type to the program. If "NumElements" is specified, it is the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3317number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3318If a constant alignment is specified, the value result of the allocation is guaranteed to
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3319be aligned to at least that boundary. If not specified, or if zero, the target can
3320choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3321
3322<p>'<tt>type</tt>' must be a sized type.</p>
3323
3324<h5>Semantics:</h5>
3325
3326<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
Nick Lewyckyff384472008-11-24 03:41:24 +0000[diff] [blame]3327a pointer is returned. The result of a zero byte allocation is undefined. The
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3328result is null if there is insufficient memory available.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3329
3330<h5>Example:</h5>
3331
3332<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +0000[diff] [blame]3333 %array = malloc [4 x i8] <i>; yields {[%4 x i8]*}:array</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3334
3335 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3336 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3337 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3338 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3339 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3340</pre>
Dan Gohman60967192009-01-12 23:12:39 +0000[diff] [blame]3341
3342<p>Note that the code generator does not yet respect the
3343 alignment value.</p>
3344
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3345</div>
3346
3347<!-- _______________________________________________________________________ -->
3348<div class="doc_subsubsection">
3349 <a name="i_free">'<tt>free</tt>' Instruction</a>
3350</div>
3351
3352<div class="doc_text">
3353
3354<h5>Syntax:</h5>
3355
3356<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +0000[diff] [blame]3357 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3358</pre>
3359
3360<h5>Overview:</h5>
3361
3362<p>The '<tt>free</tt>' instruction returns memory back to the unused
3363memory heap to be reallocated in the future.</p>
3364
3365<h5>Arguments:</h5>
3366
3367<p>'<tt>value</tt>' shall be a pointer value that points to a value
3368that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3369instruction.</p>
3370
3371<h5>Semantics:</h5>
3372
3373<p>Access to the memory pointed to by the pointer is no longer defined
Chris Lattner329d6532008-04-19 22:41:32 +0000[diff] [blame]3374after this instruction executes. If the pointer is null, the operation
3375is a noop.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3376
3377<h5>Example:</h5>
3378
3379<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +0000[diff] [blame]3380 %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]3381 free [4 x i8]* %array
3382</pre>
3383</div>
3384
3385<!-- _______________________________________________________________________ -->
3386<div class="doc_subsubsection">
3387 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3388</div>
3389
3390<div class="doc_text">
3391
3392<h5>Syntax:</h5>
3393
3394<pre>
3395 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3396</pre>
3397
3398<h5>Overview:</h5>
3399
3400<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3401currently executing function, to be automatically released when this function
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3402returns to its caller. The object is always allocated in the generic address
3403space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3404
3405<h5>Arguments:</h5>
3406
3407<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
3408bytes of memory on the runtime stack, returning a pointer of the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3409appropriate type to the program. If "NumElements" is specified, it is the
3410number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3411If a constant alignment is specified, the value result of the allocation is guaranteed
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3412to be aligned to at least that boundary. If not specified, or if zero, the target
3413can choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3414
3415<p>'<tt>type</tt>' may be any sized type.</p>
3416
3417<h5>Semantics:</h5>
3418
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3419<p>Memory is allocated; a pointer is returned. The operation is undefiend if
3420there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3421memory is automatically released when the function returns. The '<tt>alloca</tt>'
3422instruction is commonly used to represent automatic variables that must
3423have an address available. When the function returns (either with the <tt><a
3424 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]3425instructions), the memory is reclaimed. Allocating zero bytes
3426is legal, but the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3427
3428<h5>Example:</h5>
3429
3430<pre>
Dan Gohmanf54f50a2009-01-04 23:49:44 +0000[diff] [blame]3431 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3432 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3433 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3434 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3435</pre>
3436</div>
3437
3438<!-- _______________________________________________________________________ -->
3439<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3440Instruction</a> </div>
3441<div class="doc_text">
3442<h5>Syntax:</h5>
3443<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>
3444<h5>Overview:</h5>
3445<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3446<h5>Arguments:</h5>
3447<p>The argument to the '<tt>load</tt>' instruction specifies the memory
3448address from which to load. The pointer must point to a <a
3449 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3450marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3451the number or order of execution of this <tt>load</tt> with other
3452volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3453instructions. </p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3454<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3455The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3456(that is, the alignment of the memory address). A value of 0 or an
3457omitted "align" argument means that the operation has the preferential
3458alignment for the target. It is the responsibility of the code emitter
3459to ensure that the alignment information is correct. Overestimating
3460the alignment results in an undefined behavior. Underestimating the
3461alignment may produce less efficient code. An alignment of 1 is always
3462safe.
3463</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3464<h5>Semantics:</h5>
3465<p>The location of memory pointed to is loaded.</p>
3466<h5>Examples:</h5>
3467<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3468 <a
3469 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3470 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3471</pre>
3472</div>
3473<!-- _______________________________________________________________________ -->
3474<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3475Instruction</a> </div>
3476<div class="doc_text">
3477<h5>Syntax:</h5>
3478<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3479 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3480</pre>
3481<h5>Overview:</h5>
3482<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3483<h5>Arguments:</h5>
3484<p>There are two arguments to the '<tt>store</tt>' instruction: a value
3485to 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]3486operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3487of the '<tt>&lt;value&gt;</tt>'
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3488operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3489optimizer is not allowed to modify the number or order of execution of
3490this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3491 href="#i_store">store</a></tt> instructions.</p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3492<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3493The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3494(that is, the alignment of the memory address). A value of 0 or an
3495omitted "align" argument means that the operation has the preferential
3496alignment for the target. It is the responsibility of the code emitter
3497to ensure that the alignment information is correct. Overestimating
3498the alignment results in an undefined behavior. Underestimating the
3499alignment may produce less efficient code. An alignment of 1 is always
3500safe.
3501</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3502<h5>Semantics:</h5>
3503<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
3504at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
3505<h5>Example:</h5>
3506<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]3507 store i32 3, i32* %ptr <i>; yields {void}</i>
3508 %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]3509</pre>
3510</div>
3511
3512<!-- _______________________________________________________________________ -->
3513<div class="doc_subsubsection">
3514 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3515</div>
3516
3517<div class="doc_text">
3518<h5>Syntax:</h5>
3519<pre>
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3520 &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]3521</pre>
3522
3523<h5>Overview:</h5>
3524
3525<p>
3526The '<tt>getelementptr</tt>' instruction is used to get the address of a
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3527subelement of an aggregate data structure. It performs address calculation only
3528and does not access memory.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3529
3530<h5>Arguments:</h5>
3531
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3532<p>The first argument is always a pointer, and forms the basis of the
3533calculation. The remaining arguments are indices, that indicate which of the
3534elements of the aggregate object are indexed. The interpretation of each index
3535is dependent on the type being indexed into. The first index always indexes the
3536pointer value given as the first argument, the second index indexes a value of
3537the type pointed to (not necessarily the value directly pointed to, since the
3538first index can be non-zero), etc. The first type indexed into must be a pointer
3539value, subsequent types can be arrays, vectors and structs. Note that subsequent
3540types being indexed into can never be pointers, since that would require loading
3541the pointer before continuing calculation.</p>
3542
3543<p>The type of each index argument depends on the type it is indexing into.
3544When indexing into a (packed) structure, only <tt>i32</tt> integer
3545<b>constants</b> are allowed. When indexing into an array, pointer or vector,
3546only integers of 32 or 64 bits are allowed (also non-constants). 32-bit values
3547will be sign extended to 64-bits if required.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3548
3549<p>For example, let's consider a C code fragment and how it gets
3550compiled to LLVM:</p>
3551
3552<div class="doc_code">
3553<pre>
3554struct RT {
3555 char A;
3556 int B[10][20];
3557 char C;
3558};
3559struct ST {
3560 int X;
3561 double Y;
3562 struct RT Z;
3563};
3564
3565int *foo(struct ST *s) {
3566 return &amp;s[1].Z.B[5][13];
3567}
3568</pre>
3569</div>
3570
3571<p>The LLVM code generated by the GCC frontend is:</p>
3572
3573<div class="doc_code">
3574<pre>
Chris Lattner5b6dc6e2009-01-11 20:53:49 +0000[diff] [blame]3575%RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
3576%ST = <a href="#namedtypes">type</a> { i32, double, %RT }
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3577
3578define i32* %foo(%ST* %s) {
3579entry:
3580 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3581 ret i32* %reg
3582}
3583</pre>
3584</div>
3585
3586<h5>Semantics:</h5>
3587
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3588<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3589type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3590}</tt>' type, a structure. The second index indexes into the third element of
3591the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3592i8 }</tt>' type, another structure. The third index indexes into the second
3593element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3594array. The two dimensions of the array are subscripted into, yielding an
3595'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3596to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3597
3598<p>Note that it is perfectly legal to index partially through a
3599structure, returning a pointer to an inner element. Because of this,
3600the LLVM code for the given testcase is equivalent to:</p>
3601
3602<pre>
3603 define i32* %foo(%ST* %s) {
3604 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3605 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3606 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3607 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3608 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3609 ret i32* %t5
3610 }
3611</pre>
3612
3613<p>Note that it is undefined to access an array out of bounds: array and
3614pointer indexes must always be within the defined bounds of the array type.
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]3615The one exception for this rule is zero length arrays. These arrays are
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3616defined to be accessible as variable length arrays, which requires access
3617beyond the zero'th element.</p>
3618
3619<p>The getelementptr instruction is often confusing. For some more insight
3620into how it works, see <a href="GetElementPtr.html">the getelementptr
3621FAQ</a>.</p>
3622
3623<h5>Example:</h5>
3624
3625<pre>
3626 <i>; yields [12 x i8]*:aptr</i>
Matthijs Kooijman32a080e2008-10-13 13:44:15 +0000[diff] [blame]3627 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
3628 <i>; yields i8*:vptr</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]3629 %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]3630 <i>; yields i8*:eptr</i>
3631 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3632</pre>
3633</div>
3634
3635<!-- ======================================================================= -->
3636<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3637</div>
3638<div class="doc_text">
3639<p>The instructions in this category are the conversion instructions (casting)
3640which all take a single operand and a type. They perform various bit conversions
3641on the operand.</p>
3642</div>
3643
3644<!-- _______________________________________________________________________ -->
3645<div class="doc_subsubsection">
3646 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3647</div>
3648<div class="doc_text">
3649
3650<h5>Syntax:</h5>
3651<pre>
3652 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3653</pre>
3654
3655<h5>Overview:</h5>
3656<p>
3657The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3658</p>
3659
3660<h5>Arguments:</h5>
3661<p>
3662The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3663be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3664and type of the result, which must be an <a href="#t_integer">integer</a>
3665type. The bit size of <tt>value</tt> must be larger than the bit size of
3666<tt>ty2</tt>. Equal sized types are not allowed.</p>
3667
3668<h5>Semantics:</h5>
3669<p>
3670The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3671and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3672larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3673It will always truncate bits.</p>
3674
3675<h5>Example:</h5>
3676<pre>
3677 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3678 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3679 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3680</pre>
3681</div>
3682
3683<!-- _______________________________________________________________________ -->
3684<div class="doc_subsubsection">
3685 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3686</div>
3687<div class="doc_text">
3688
3689<h5>Syntax:</h5>
3690<pre>
3691 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3692</pre>
3693
3694<h5>Overview:</h5>
3695<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3696<tt>ty2</tt>.</p>
3697
3698
3699<h5>Arguments:</h5>
3700<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3701<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3702also be of <a href="#t_integer">integer</a> type. The bit size of the
3703<tt>value</tt> must be smaller than the bit size of the destination type,
3704<tt>ty2</tt>.</p>
3705
3706<h5>Semantics:</h5>
3707<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3708bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3709
3710<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3711
3712<h5>Example:</h5>
3713<pre>
3714 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3715 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3716</pre>
3717</div>
3718
3719<!-- _______________________________________________________________________ -->
3720<div class="doc_subsubsection">
3721 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3722</div>
3723<div class="doc_text">
3724
3725<h5>Syntax:</h5>
3726<pre>
3727 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3728</pre>
3729
3730<h5>Overview:</h5>
3731<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3732
3733<h5>Arguments:</h5>
3734<p>
3735The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3736<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3737also be of <a href="#t_integer">integer</a> type. The bit size of the
3738<tt>value</tt> must be smaller than the bit size of the destination type,
3739<tt>ty2</tt>.</p>
3740
3741<h5>Semantics:</h5>
3742<p>
3743The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3744bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3745the type <tt>ty2</tt>.</p>
3746
3747<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3748
3749<h5>Example:</h5>
3750<pre>
3751 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3752 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3753</pre>
3754</div>
3755
3756<!-- _______________________________________________________________________ -->
3757<div class="doc_subsubsection">
3758 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3759</div>
3760
3761<div class="doc_text">
3762
3763<h5>Syntax:</h5>
3764
3765<pre>
3766 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3767</pre>
3768
3769<h5>Overview:</h5>
3770<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3771<tt>ty2</tt>.</p>
3772
3773
3774<h5>Arguments:</h5>
3775<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3776 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3777cast it to. The size of <tt>value</tt> must be larger than the size of
3778<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3779<i>no-op cast</i>.</p>
3780
3781<h5>Semantics:</h5>
3782<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3783<a href="#t_floating">floating point</a> type to a smaller
3784<a href="#t_floating">floating point</a> type. If the value cannot fit within
3785the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3786
3787<h5>Example:</h5>
3788<pre>
3789 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3790 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3791</pre>
3792</div>
3793
3794<!-- _______________________________________________________________________ -->
3795<div class="doc_subsubsection">
3796 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3797</div>
3798<div class="doc_text">
3799
3800<h5>Syntax:</h5>
3801<pre>
3802 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3803</pre>
3804
3805<h5>Overview:</h5>
3806<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3807floating point value.</p>
3808
3809<h5>Arguments:</h5>
3810<p>The '<tt>fpext</tt>' instruction takes a
3811<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3812and a <a href="#t_floating">floating point</a> type to cast it to. The source
3813type must be smaller than the destination type.</p>
3814
3815<h5>Semantics:</h5>
3816<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3817<a href="#t_floating">floating point</a> type to a larger
3818<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3819used to make a <i>no-op cast</i> because it always changes bits. Use
3820<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3821
3822<h5>Example:</h5>
3823<pre>
3824 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3825 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3826</pre>
3827</div>
3828
3829<!-- _______________________________________________________________________ -->
3830<div class="doc_subsubsection">
3831 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3832</div>
3833<div class="doc_text">
3834
3835<h5>Syntax:</h5>
3836<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3837 &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]3838</pre>
3839
3840<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3841<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3842unsigned integer equivalent of type <tt>ty2</tt>.
3843</p>
3844
3845<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3846<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]3847scalar or vector <a href="#t_floating">floating point</a> value, and a type
3848to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3849type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3850vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3851
3852<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3853<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3854<a href="#t_floating">floating point</a> operand into the nearest (rounding
3855towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3856the results are undefined.</p>
3857
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3858<h5>Example:</h5>
3859<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3860 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3861 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3862 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3863</pre>
3864</div>
3865
3866<!-- _______________________________________________________________________ -->
3867<div class="doc_subsubsection">
3868 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3869</div>
3870<div class="doc_text">
3871
3872<h5>Syntax:</h5>
3873<pre>
3874 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3875</pre>
3876
3877<h5>Overview:</h5>
3878<p>The '<tt>fptosi</tt>' instruction converts
3879<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3880</p>
3881
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3882<h5>Arguments:</h5>
3883<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]3884scalar or vector <a href="#t_floating">floating point</a> value, and a type
3885to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3886type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3887vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3888
3889<h5>Semantics:</h5>
3890<p>The '<tt>fptosi</tt>' instruction converts its
3891<a href="#t_floating">floating point</a> operand into the nearest (rounding
3892towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3893the results are undefined.</p>
3894
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3895<h5>Example:</h5>
3896<pre>
3897 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3898 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3899 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3900</pre>
3901</div>
3902
3903<!-- _______________________________________________________________________ -->
3904<div class="doc_subsubsection">
3905 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3906</div>
3907<div class="doc_text">
3908
3909<h5>Syntax:</h5>
3910<pre>
3911 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3912</pre>
3913
3914<h5>Overview:</h5>
3915<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3916integer and converts that value to the <tt>ty2</tt> type.</p>
3917
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3918<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3919<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3920scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3921to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3922type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3923floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3924
3925<h5>Semantics:</h5>
3926<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3927integer quantity and converts it to the corresponding floating point value. If
3928the value cannot fit in the floating point value, the results are undefined.</p>
3929
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3930<h5>Example:</h5>
3931<pre>
3932 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]3933 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3934</pre>
3935</div>
3936
3937<!-- _______________________________________________________________________ -->
3938<div class="doc_subsubsection">
3939 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3940</div>
3941<div class="doc_text">
3942
3943<h5>Syntax:</h5>
3944<pre>
3945 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3946</pre>
3947
3948<h5>Overview:</h5>
3949<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3950integer and converts that value to the <tt>ty2</tt> type.</p>
3951
3952<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3953<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3954scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3955to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3956type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3957floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3958
3959<h5>Semantics:</h5>
3960<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3961integer quantity and converts it to the corresponding floating point value. If
3962the value cannot fit in the floating point value, the results are undefined.</p>
3963
3964<h5>Example:</h5>
3965<pre>
3966 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]3967 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3968</pre>
3969</div>
3970
3971<!-- _______________________________________________________________________ -->
3972<div class="doc_subsubsection">
3973 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3974</div>
3975<div class="doc_text">
3976
3977<h5>Syntax:</h5>
3978<pre>
3979 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3980</pre>
3981
3982<h5>Overview:</h5>
3983<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3984the integer type <tt>ty2</tt>.</p>
3985
3986<h5>Arguments:</h5>
3987<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3988must 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]3989<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]3990
3991<h5>Semantics:</h5>
3992<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3993<tt>ty2</tt> by interpreting the pointer value as an integer and either
3994truncating or zero extending that value to the size of the integer type. If
3995<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3996<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3997are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3998change.</p>
3999
4000<h5>Example:</h5>
4001<pre>
4002 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4003 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4004</pre>
4005</div>
4006
4007<!-- _______________________________________________________________________ -->
4008<div class="doc_subsubsection">
4009 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4010</div>
4011<div class="doc_text">
4012
4013<h5>Syntax:</h5>
4014<pre>
4015 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4016</pre>
4017
4018<h5>Overview:</h5>
4019<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
4020a pointer type, <tt>ty2</tt>.</p>
4021
4022<h5>Arguments:</h5>
4023<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4024value to cast, and a type to cast it to, which must be a
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4025<a href="#t_pointer">pointer</a> type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4026
4027<h5>Semantics:</h5>
4028<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4029<tt>ty2</tt> by applying either a zero extension or a truncation depending on
4030the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4031size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
4032the size of a pointer then a zero extension is done. If they are the same size,
4033nothing is done (<i>no-op cast</i>).</p>
4034
4035<h5>Example:</h5>
4036<pre>
4037 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4038 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4039 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4040</pre>
4041</div>
4042
4043<!-- _______________________________________________________________________ -->
4044<div class="doc_subsubsection">
4045 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4046</div>
4047<div class="doc_text">
4048
4049<h5>Syntax:</h5>
4050<pre>
4051 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4052</pre>
4053
4054<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4055
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4056<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4057<tt>ty2</tt> without changing any bits.</p>
4058
4059<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4060
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4061<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
Dan Gohman7305fa02008-09-08 16:45:59 +0000[diff] [blame]4062a non-aggregate first class value, and a type to cast it to, which must also be
4063a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
4064<tt>value</tt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4065and the destination type, <tt>ty2</tt>, must be identical. If the source
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4066type is a pointer, the destination type must also be a pointer. This
4067instruction supports bitwise conversion of vectors to integers and to vectors
4068of other types (as long as they have the same size).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4069
4070<h5>Semantics:</h5>
4071<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4072<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4073this conversion. The conversion is done as if the <tt>value</tt> had been
4074stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
4075converted to other pointer types with this instruction. To convert pointers to
4076other types, use the <a href="#i_inttoptr">inttoptr</a> or
4077<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4078
4079<h5>Example:</h5>
4080<pre>
4081 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4082 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4083 %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]4084</pre>
4085</div>
4086
4087<!-- ======================================================================= -->
4088<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4089<div class="doc_text">
4090<p>The instructions in this category are the "miscellaneous"
4091instructions, which defy better classification.</p>
4092</div>
4093
4094<!-- _______________________________________________________________________ -->
4095<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4096</div>
4097<div class="doc_text">
4098<h5>Syntax:</h5>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4099<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]4100</pre>
4101<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4102<p>The '<tt>icmp</tt>' instruction returns a boolean value or
4103a vector of boolean values based on comparison
4104of its two integer, integer vector, or pointer operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4105<h5>Arguments:</h5>
4106<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
4107the condition code indicating the kind of comparison to perform. It is not
4108a value, just a keyword. The possible condition code are:
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4109</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4110<ol>
4111 <li><tt>eq</tt>: equal</li>
4112 <li><tt>ne</tt>: not equal </li>
4113 <li><tt>ugt</tt>: unsigned greater than</li>
4114 <li><tt>uge</tt>: unsigned greater or equal</li>
4115 <li><tt>ult</tt>: unsigned less than</li>
4116 <li><tt>ule</tt>: unsigned less or equal</li>
4117 <li><tt>sgt</tt>: signed greater than</li>
4118 <li><tt>sge</tt>: signed greater or equal</li>
4119 <li><tt>slt</tt>: signed less than</li>
4120 <li><tt>sle</tt>: signed less or equal</li>
4121</ol>
4122<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4123<a href="#t_pointer">pointer</a>
4124or integer <a href="#t_vector">vector</a> typed.
4125They must also be identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4126<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4127<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]4128the condition code given as <tt>cond</tt>. The comparison performed always
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4129yields 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]4130</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4131<ol>
4132 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
4133 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
4134 </li>
4135 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4136 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4137 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4138 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4139 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4140 <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]4141 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4142 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4143 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4144 <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]4145 <li><tt>sgt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4146 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4147 <li><tt>sge</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4148 <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]4149 <li><tt>slt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4150 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4151 <li><tt>sle</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4152 <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]4153</ol>
4154<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
4155values are compared as if they were integers.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4156<p>If the operands are integer vectors, then they are compared
4157element by element. The result is an <tt>i1</tt> vector with
4158the same number of elements as the values being compared.
4159Otherwise, the result is an <tt>i1</tt>.
4160</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4161
4162<h5>Example:</h5>
4163<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
4164 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
4165 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
4166 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4167 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4168 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
4169</pre>
Dan Gohmana53eb382009-01-22 01:39:38 +0000[diff] [blame]4170
4171<p>Note that the code generator does not yet support vector types with
4172 the <tt>icmp</tt> instruction.</p>
4173
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4174</div>
4175
4176<!-- _______________________________________________________________________ -->
4177<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4178</div>
4179<div class="doc_text">
4180<h5>Syntax:</h5>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4181<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]4182</pre>
4183<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4184<p>The '<tt>fcmp</tt>' instruction returns a boolean value
4185or vector of boolean values based on comparison
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4186of its operands.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4187<p>
4188If the operands are floating point scalars, then the result
4189type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
4190</p>
4191<p>If the operands are floating point vectors, then the result type
4192is a vector of boolean with the same number of elements as the
4193operands being compared.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4194<h5>Arguments:</h5>
4195<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4196the condition code indicating the kind of comparison to perform. It is not
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4197a value, just a keyword. The possible condition code are:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4198<ol>
4199 <li><tt>false</tt>: no comparison, always returns false</li>
4200 <li><tt>oeq</tt>: ordered and equal</li>
4201 <li><tt>ogt</tt>: ordered and greater than </li>
4202 <li><tt>oge</tt>: ordered and greater than or equal</li>
4203 <li><tt>olt</tt>: ordered and less than </li>
4204 <li><tt>ole</tt>: ordered and less than or equal</li>
4205 <li><tt>one</tt>: ordered and not equal</li>
4206 <li><tt>ord</tt>: ordered (no nans)</li>
4207 <li><tt>ueq</tt>: unordered or equal</li>
4208 <li><tt>ugt</tt>: unordered or greater than </li>
4209 <li><tt>uge</tt>: unordered or greater than or equal</li>
4210 <li><tt>ult</tt>: unordered or less than </li>
4211 <li><tt>ule</tt>: unordered or less than or equal</li>
4212 <li><tt>une</tt>: unordered or not equal</li>
4213 <li><tt>uno</tt>: unordered (either nans)</li>
4214 <li><tt>true</tt>: no comparison, always returns true</li>
4215</ol>
4216<p><i>Ordered</i> means that neither operand is a QNAN while
4217<i>unordered</i> means that either operand may be a QNAN.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4218<p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
4219either a <a href="#t_floating">floating point</a> type
4220or a <a href="#t_vector">vector</a> of floating point type.
4221They must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4222<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4223<p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4224according to the condition code given as <tt>cond</tt>.
4225If the operands are vectors, then the vectors are compared
4226element by element.
4227Each comparison performed
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4228always yields an <a href="#t_primitive">i1</a> result, as follows:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4229<ol>
4230 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4231 <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]4232 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4233 <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]4234 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4235 <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]4236 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4237 <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]4238 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4239 <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]4240 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4241 <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]4242 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4243 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4244 <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]4245 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4246 <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]4247 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4248 <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]4249 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4250 <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]4251 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4252 <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]4253 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4254 <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]4255 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4256 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4257 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4258</ol>
4259
4260<h5>Example:</h5>
4261<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]4262 &lt;result&gt; = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4263 &lt;result&gt; = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4264 &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]4265</pre>
Dan Gohmana53eb382009-01-22 01:39:38 +0000[diff] [blame]4266
4267<p>Note that the code generator does not yet support vector types with
4268 the <tt>fcmp</tt> instruction.</p>
4269
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4270</div>
4271
4272<!-- _______________________________________________________________________ -->
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4273<div class="doc_subsubsection">
4274 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
4275</div>
4276<div class="doc_text">
4277<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4278<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]4279</pre>
4280<h5>Overview:</h5>
4281<p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
4282element-wise comparison of its two integer vector operands.</p>
4283<h5>Arguments:</h5>
4284<p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
4285the condition code indicating the kind of comparison to perform. It is not
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4286a value, just a keyword. The possible condition code are:</p>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4287<ol>
4288 <li><tt>eq</tt>: equal</li>
4289 <li><tt>ne</tt>: not equal </li>
4290 <li><tt>ugt</tt>: unsigned greater than</li>
4291 <li><tt>uge</tt>: unsigned greater or equal</li>
4292 <li><tt>ult</tt>: unsigned less than</li>
4293 <li><tt>ule</tt>: unsigned less or equal</li>
4294 <li><tt>sgt</tt>: signed greater than</li>
4295 <li><tt>sge</tt>: signed greater or equal</li>
4296 <li><tt>slt</tt>: signed less than</li>
4297 <li><tt>sle</tt>: signed less or equal</li>
4298</ol>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4299<p>The remaining two arguments must be <a href="#t_vector">vector</a> or
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4300<a href="#t_integer">integer</a> typed. They must also be identical types.</p>
4301<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4302<p>The '<tt>vicmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4303according to the condition code given as <tt>cond</tt>. The comparison yields a
4304<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
4305identical type as the values being compared. The most significant bit in each
4306element is 1 if the element-wise comparison evaluates to true, and is 0
4307otherwise. All other bits of the result are undefined. The condition codes
4308are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4309instruction</a>.</p>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4310
4311<h5>Example:</h5>
4312<pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4313 &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>
4314 &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]4315</pre>
4316</div>
4317
4318<!-- _______________________________________________________________________ -->
4319<div class="doc_subsubsection">
4320 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4321</div>
4322<div class="doc_text">
4323<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4324<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]4325<h5>Overview:</h5>
4326<p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4327element-wise comparison of its two floating point vector operands. The output
4328elements have the same width as the input elements.</p>
4329<h5>Arguments:</h5>
4330<p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4331the condition code indicating the kind of comparison to perform. It is not
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4332a value, just a keyword. The possible condition code are:</p>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4333<ol>
4334 <li><tt>false</tt>: no comparison, always returns false</li>
4335 <li><tt>oeq</tt>: ordered and equal</li>
4336 <li><tt>ogt</tt>: ordered and greater than </li>
4337 <li><tt>oge</tt>: ordered and greater than or equal</li>
4338 <li><tt>olt</tt>: ordered and less than </li>
4339 <li><tt>ole</tt>: ordered and less than or equal</li>
4340 <li><tt>one</tt>: ordered and not equal</li>
4341 <li><tt>ord</tt>: ordered (no nans)</li>
4342 <li><tt>ueq</tt>: unordered or equal</li>
4343 <li><tt>ugt</tt>: unordered or greater than </li>
4344 <li><tt>uge</tt>: unordered or greater than or equal</li>
4345 <li><tt>ult</tt>: unordered or less than </li>
4346 <li><tt>ule</tt>: unordered or less than or equal</li>
4347 <li><tt>une</tt>: unordered or not equal</li>
4348 <li><tt>uno</tt>: unordered (either nans)</li>
4349 <li><tt>true</tt>: no comparison, always returns true</li>
4350</ol>
4351<p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4352<a href="#t_floating">floating point</a> typed. They must also be identical
4353types.</p>
4354<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4355<p>The '<tt>vfcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4356according to the condition code given as <tt>cond</tt>. The comparison yields a
4357<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4358an identical number of elements as the values being compared, and each element
4359having identical with to the width of the floating point elements. The most
4360significant bit in each element is 1 if the element-wise comparison evaluates to
4361true, and is 0 otherwise. All other bits of the result are undefined. The
4362condition codes are evaluated identically to the
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4363<a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.</p>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4364
4365<h5>Example:</h5>
4366<pre>
Chris Lattner1e5c5cd02008-10-13 16:55:18 +0000[diff] [blame]4367 <i>; yields: result=&lt;2 x i32&gt; &lt; i32 0, i32 -1 &gt;</i>
4368 &lt;result&gt; = vfcmp oeq &lt;2 x float&gt; &lt; float 4, float 0 &gt;, &lt; float 5, float 0 &gt;
4369
4370 <i>; yields: result=&lt;2 x i64&gt; &lt; i64 -1, i64 0 &gt;</i>
4371 &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]4372</pre>
4373</div>
4374
4375<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4376<div class="doc_subsubsection">
4377 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4378</div>
4379
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4380<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4381
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4382<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4383
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4384<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
4385<h5>Overview:</h5>
4386<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
4387the SSA graph representing the function.</p>
4388<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4389
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4390<p>The type of the incoming values is specified with the first type
4391field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4392as arguments, with one pair for each predecessor basic block of the
4393current block. Only values of <a href="#t_firstclass">first class</a>
4394type may be used as the value arguments to the PHI node. Only labels
4395may be used as the label arguments.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4396
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4397<p>There must be no non-phi instructions between the start of a basic
4398block and the PHI instructions: i.e. PHI instructions must be first in
4399a basic block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4400
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4401<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4402
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4403<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4404specified by the pair corresponding to the predecessor basic block that executed
4405just prior to the current block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4406
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4407<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4408<pre>
4409Loop: ; Infinite loop that counts from 0 on up...
4410 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4411 %nextindvar = add i32 %indvar, 1
4412 br label %Loop
4413</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4414</div>
4415
4416<!-- _______________________________________________________________________ -->
4417<div class="doc_subsubsection">
4418 <a name="i_select">'<tt>select</tt>' Instruction</a>
4419</div>
4420
4421<div class="doc_text">
4422
4423<h5>Syntax:</h5>
4424
4425<pre>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4426 &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>
4427
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4428 <i>selty</i> is either i1 or {&lt;N x i1&gt;}
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4429</pre>
4430
4431<h5>Overview:</h5>
4432
4433<p>
4434The '<tt>select</tt>' instruction is used to choose one value based on a
4435condition, without branching.
4436</p>
4437
4438
4439<h5>Arguments:</h5>
4440
4441<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4442The '<tt>select</tt>' instruction requires an 'i1' value or
4443a vector of 'i1' values indicating the
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4444condition, and two values of the same <a href="#t_firstclass">first class</a>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4445type. If the val1/val2 are vectors and
4446the condition is a scalar, then entire vectors are selected, not
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4447individual elements.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4448</p>
4449
4450<h5>Semantics:</h5>
4451
4452<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4453If 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]4454value argument; otherwise, it returns the second value argument.
4455</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4456<p>
4457If the condition is a vector of i1, then the value arguments must
4458be vectors of the same size, and the selection is done element
4459by element.
4460</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4461
4462<h5>Example:</h5>
4463
4464<pre>
4465 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4466</pre>
Dan Gohmana53eb382009-01-22 01:39:38 +0000[diff] [blame]4467
4468<p>Note that the code generator does not yet support conditions
4469 with vector type.</p>
4470
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4471</div>
4472
4473
4474<!-- _______________________________________________________________________ -->
4475<div class="doc_subsubsection">
4476 <a name="i_call">'<tt>call</tt>' Instruction</a>
4477</div>
4478
4479<div class="doc_text">
4480
4481<h5>Syntax:</h5>
4482<pre>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]4483 &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]4484</pre>
4485
4486<h5>Overview:</h5>
4487
4488<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4489
4490<h5>Arguments:</h5>
4491
4492<p>This instruction requires several arguments:</p>
4493
4494<ol>
4495 <li>
4496 <p>The optional "tail" marker indicates whether the callee function accesses
4497 any allocas or varargs in the caller. If the "tail" marker is present, the
4498 function call is eligible for tail call optimization. Note that calls may
4499 be marked "tail" even if they do not occur before a <a
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4500 href="#i_ret"><tt>ret</tt></a> instruction.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4501 </li>
4502 <li>
4503 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4504 convention</a> the call should use. If none is specified, the call defaults
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4505 to using C calling conventions.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4506 </li>
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4507
4508 <li>
4509 <p>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4510 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
4511 and '<tt>inreg</tt>' attributes are valid here.</p>
4512 </li>
4513
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4514 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4515 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4516 the type of the return value. Functions that return no value are marked
4517 <tt><a href="#t_void">void</a></tt>.</p>
4518 </li>
4519 <li>
4520 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4521 value being invoked. The argument types must match the types implied by
4522 this signature. This type can be omitted if the function is not varargs
4523 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4524 </li>
4525 <li>
4526 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4527 be invoked. In most cases, this is a direct function invocation, but
4528 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4529 to function value.</p>
4530 </li>
4531 <li>
4532 <p>'<tt>function args</tt>': argument list whose types match the
4533 function signature argument types. All arguments must be of
4534 <a href="#t_firstclass">first class</a> type. If the function signature
4535 indicates the function accepts a variable number of arguments, the extra
4536 arguments can be specified.</p>
4537 </li>
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4538 <li>
Devang Pateld0bfcc72008-10-07 17:48:33 +0000[diff] [blame]4539 <p>The optional <a href="#fnattrs">function attributes</a> list. Only
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4540 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
4541 '<tt>readnone</tt>' attributes are valid here.</p>
4542 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4543</ol>
4544
4545<h5>Semantics:</h5>
4546
4547<p>The '<tt>call</tt>' instruction is used to cause control flow to
4548transfer to a specified function, with its incoming arguments bound to
4549the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4550instruction in the called function, control flow continues with the
4551instruction after the function call, and the return value of the
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4552function is bound to the result argument.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4553
4554<h5>Example:</h5>
4555
4556<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4557 %retval = call i32 @test(i32 %argc)
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4558 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4559 %X = tail call i32 @foo() <i>; yields i32</i>
4560 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4561 call void %foo(i8 97 signext)
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4562
4563 %struct.A = type { i32, i8 }
Devang Patelac2fc272008-10-06 18:50:38 +0000[diff] [blame]4564 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
Dan Gohman3e700032008-10-04 19:00:07 +0000[diff] [blame]4565 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4566 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
Chris Lattnerac454b32008-10-08 06:26:11 +0000[diff] [blame]4567 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
Matthijs Kooijman2c4e05a2008-10-07 10:03:45 +0000[diff] [blame]4568 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4569</pre>
4570
4571</div>
4572
4573<!-- _______________________________________________________________________ -->
4574<div class="doc_subsubsection">
4575 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4576</div>
4577
4578<div class="doc_text">
4579
4580<h5>Syntax:</h5>
4581
4582<pre>
4583 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
4584</pre>
4585
4586<h5>Overview:</h5>
4587
4588<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4589the "variable argument" area of a function call. It is used to implement the
4590<tt>va_arg</tt> macro in C.</p>
4591
4592<h5>Arguments:</h5>
4593
4594<p>This instruction takes a <tt>va_list*</tt> value and the type of
4595the argument. It returns a value of the specified argument type and
4596increments the <tt>va_list</tt> to point to the next argument. The
4597actual type of <tt>va_list</tt> is target specific.</p>
4598
4599<h5>Semantics:</h5>
4600
4601<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4602type from the specified <tt>va_list</tt> and causes the
4603<tt>va_list</tt> to point to the next argument. For more information,
4604see the variable argument handling <a href="#int_varargs">Intrinsic
4605Functions</a>.</p>
4606
4607<p>It is legal for this instruction to be called in a function which does not
4608take a variable number of arguments, for example, the <tt>vfprintf</tt>
4609function.</p>
4610
4611<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4612href="#intrinsics">intrinsic function</a> because it takes a type as an
4613argument.</p>
4614
4615<h5>Example:</h5>
4616
4617<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4618
Dan Gohman60967192009-01-12 23:12:39 +0000[diff] [blame]4619<p>Note that the code generator does not yet fully support va_arg
4620 on many targets. Also, it does not currently support va_arg with
4621 aggregate types on any target.</p>
4622
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4623</div>
4624
4625<!-- *********************************************************************** -->
4626<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4627<!-- *********************************************************************** -->
4628
4629<div class="doc_text">
4630
4631<p>LLVM supports the notion of an "intrinsic function". These functions have
4632well known names and semantics and are required to follow certain restrictions.
4633Overall, these intrinsics represent an extension mechanism for the LLVM
4634language that does not require changing all of the transformations in LLVM when
4635adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4636
4637<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4638prefix is reserved in LLVM for intrinsic names; thus, function names may not
4639begin with this prefix. Intrinsic functions must always be external functions:
4640you cannot define the body of intrinsic functions. Intrinsic functions may
4641only be used in call or invoke instructions: it is illegal to take the address
4642of an intrinsic function. Additionally, because intrinsic functions are part
4643of the LLVM language, it is required if any are added that they be documented
4644here.</p>
4645
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4646<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4647a family of functions that perform the same operation but on different data
4648types. Because LLVM can represent over 8 million different integer types,
4649overloading is used commonly to allow an intrinsic function to operate on any
4650integer type. One or more of the argument types or the result type can be
4651overloaded to accept any integer type. Argument types may also be defined as
4652exactly matching a previous argument's type or the result type. This allows an
4653intrinsic function which accepts multiple arguments, but needs all of them to
4654be of the same type, to only be overloaded with respect to a single argument or
4655the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4656
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4657<p>Overloaded intrinsics will have the names of its overloaded argument types
4658encoded into its function name, each preceded by a period. Only those types
4659which are overloaded result in a name suffix. Arguments whose type is matched
4660against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4661take an integer of any width and returns an integer of exactly the same integer
4662width. This leads to a family of functions such as
4663<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4664Only one type, the return type, is overloaded, and only one type suffix is
4665required. Because the argument's type is matched against the return type, it
4666does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4667
4668<p>To learn how to add an intrinsic function, please see the
4669<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4670</p>
4671
4672</div>
4673
4674<!-- ======================================================================= -->
4675<div class="doc_subsection">
4676 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4677</div>
4678
4679<div class="doc_text">
4680
4681<p>Variable argument support is defined in LLVM with the <a
4682 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4683intrinsic functions. These functions are related to the similarly
4684named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
4685
4686<p>All of these functions operate on arguments that use a
4687target-specific value type "<tt>va_list</tt>". The LLVM assembly
4688language reference manual does not define what this type is, so all
4689transformations should be prepared to handle these functions regardless of
4690the type used.</p>
4691
4692<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4693instruction and the variable argument handling intrinsic functions are
4694used.</p>
4695
4696<div class="doc_code">
4697<pre>
4698define i32 @test(i32 %X, ...) {
4699 ; Initialize variable argument processing
4700 %ap = alloca i8*
4701 %ap2 = bitcast i8** %ap to i8*
4702 call void @llvm.va_start(i8* %ap2)
4703
4704 ; Read a single integer argument
4705 %tmp = va_arg i8** %ap, i32
4706
4707 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4708 %aq = alloca i8*
4709 %aq2 = bitcast i8** %aq to i8*
4710 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4711 call void @llvm.va_end(i8* %aq2)
4712
4713 ; Stop processing of arguments.
4714 call void @llvm.va_end(i8* %ap2)
4715 ret i32 %tmp
4716}
4717
4718declare void @llvm.va_start(i8*)
4719declare void @llvm.va_copy(i8*, i8*)
4720declare void @llvm.va_end(i8*)
4721</pre>
4722</div>
4723
4724</div>
4725
4726<!-- _______________________________________________________________________ -->
4727<div class="doc_subsubsection">
4728 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4729</div>
4730
4731
4732<div class="doc_text">
4733<h5>Syntax:</h5>
4734<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
4735<h5>Overview:</h5>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4736<p>The '<tt>llvm.va_start</tt>' intrinsic initializes
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4737<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
4738href="#i_va_arg">va_arg</a></tt>.</p>
4739
4740<h5>Arguments:</h5>
4741
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4742<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]4743
4744<h5>Semantics:</h5>
4745
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]4746<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]4747macro available in C. In a target-dependent way, it initializes the
4748<tt>va_list</tt> element to which the argument points, so that the next call to
4749<tt>va_arg</tt> will produce the first variable argument passed to the function.
4750Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4751last argument of the function as the compiler can figure that out.</p>
4752
4753</div>
4754
4755<!-- _______________________________________________________________________ -->
4756<div class="doc_subsubsection">
4757 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4758</div>
4759
4760<div class="doc_text">
4761<h5>Syntax:</h5>
4762<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
4763<h5>Overview:</h5>
4764
4765<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
4766which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4767or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4768
4769<h5>Arguments:</h5>
4770
4771<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4772
4773<h5>Semantics:</h5>
4774
4775<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4776macro available in C. In a target-dependent way, it destroys the
4777<tt>va_list</tt> element to which the argument points. Calls to <a
4778href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4779<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4780<tt>llvm.va_end</tt>.</p>
4781
4782</div>
4783
4784<!-- _______________________________________________________________________ -->
4785<div class="doc_subsubsection">
4786 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4787</div>
4788
4789<div class="doc_text">
4790
4791<h5>Syntax:</h5>
4792
4793<pre>
4794 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
4795</pre>
4796
4797<h5>Overview:</h5>
4798
4799<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4800from the source argument list to the destination argument list.</p>
4801
4802<h5>Arguments:</h5>
4803
4804<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4805The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4806
4807
4808<h5>Semantics:</h5>
4809
4810<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4811macro available in C. In a target-dependent way, it copies the source
4812<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4813intrinsic is necessary because the <tt><a href="#int_va_start">
4814llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4815example, memory allocation.</p>
4816
4817</div>
4818
4819<!-- ======================================================================= -->
4820<div class="doc_subsection">
4821 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4822</div>
4823
4824<div class="doc_text">
4825
4826<p>
4827LLVM support for <a href="GarbageCollection.html">Accurate Garbage
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]4828Collection</a> (GC) requires the implementation and generation of these
4829intrinsics.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4830These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4831stack</a>, as well as garbage collector implementations that require <a
4832href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4833Front-ends for type-safe garbage collected languages should generate these
4834intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4835href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4836</p>
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]4837
4838<p>The garbage collection intrinsics only operate on objects in the generic
4839 address space (address space zero).</p>
4840
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4841</div>
4842
4843<!-- _______________________________________________________________________ -->
4844<div class="doc_subsubsection">
4845 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4846</div>
4847
4848<div class="doc_text">
4849
4850<h5>Syntax:</h5>
4851
4852<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4853 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4854</pre>
4855
4856<h5>Overview:</h5>
4857
4858<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4859the code generator, and allows some metadata to be associated with it.</p>
4860
4861<h5>Arguments:</h5>
4862
4863<p>The first argument specifies the address of a stack object that contains the
4864root pointer. The second pointer (which must be either a constant or a global
4865value address) contains the meta-data to be associated with the root.</p>
4866
4867<h5>Semantics:</h5>
4868
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]4869<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]4870location. At compile-time, the code generator generates information to allow
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4871the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4872intrinsic may only be used in a function which <a href="#gc">specifies a GC
4873algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4874
4875</div>
4876
4877
4878<!-- _______________________________________________________________________ -->
4879<div class="doc_subsubsection">
4880 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4881</div>
4882
4883<div class="doc_text">
4884
4885<h5>Syntax:</h5>
4886
4887<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4888 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4889</pre>
4890
4891<h5>Overview:</h5>
4892
4893<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4894locations, allowing garbage collector implementations that require read
4895barriers.</p>
4896
4897<h5>Arguments:</h5>
4898
4899<p>The second argument is the address to read from, which should be an address
4900allocated from the garbage collector. The first object is a pointer to the
4901start of the referenced object, if needed by the language runtime (otherwise
4902null).</p>
4903
4904<h5>Semantics:</h5>
4905
4906<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4907instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4908garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4909may only be used in a function which <a href="#gc">specifies a GC
4910algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4911
4912</div>
4913
4914
4915<!-- _______________________________________________________________________ -->
4916<div class="doc_subsubsection">
4917 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4918</div>
4919
4920<div class="doc_text">
4921
4922<h5>Syntax:</h5>
4923
4924<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4925 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4926</pre>
4927
4928<h5>Overview:</h5>
4929
4930<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4931locations, allowing garbage collector implementations that require write
4932barriers (such as generational or reference counting collectors).</p>
4933
4934<h5>Arguments:</h5>
4935
4936<p>The first argument is the reference to store, the second is the start of the
4937object to store it to, and the third is the address of the field of Obj to
4938store to. If the runtime does not require a pointer to the object, Obj may be
4939null.</p>
4940
4941<h5>Semantics:</h5>
4942
4943<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4944instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4945garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4946may only be used in a function which <a href="#gc">specifies a GC
4947algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4948
4949</div>
4950
4951
4952
4953<!-- ======================================================================= -->
4954<div class="doc_subsection">
4955 <a name="int_codegen">Code Generator Intrinsics</a>
4956</div>
4957
4958<div class="doc_text">
4959<p>
4960These intrinsics are provided by LLVM to expose special features that may only
4961be implemented with code generator support.
4962</p>
4963
4964</div>
4965
4966<!-- _______________________________________________________________________ -->
4967<div class="doc_subsubsection">
4968 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4969</div>
4970
4971<div class="doc_text">
4972
4973<h5>Syntax:</h5>
4974<pre>
4975 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
4976</pre>
4977
4978<h5>Overview:</h5>
4979
4980<p>
4981The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4982target-specific value indicating the return address of the current function
4983or one of its callers.
4984</p>
4985
4986<h5>Arguments:</h5>
4987
4988<p>
4989The argument to this intrinsic indicates which function to return the address
4990for. Zero indicates the calling function, one indicates its caller, etc. The
4991argument is <b>required</b> to be a constant integer value.
4992</p>
4993
4994<h5>Semantics:</h5>
4995
4996<p>
4997The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4998the return address of the specified call frame, or zero if it cannot be
4999identified. The value returned by this intrinsic is likely to be incorrect or 0
5000for arguments other than zero, so it should only be used for debugging purposes.
5001</p>
5002
5003<p>
5004Note that calling this intrinsic does not prevent function inlining or other
5005aggressive transformations, so the value returned may not be that of the obvious
5006source-language caller.
5007</p>
5008</div>
5009
5010
5011<!-- _______________________________________________________________________ -->
5012<div class="doc_subsubsection">
5013 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5014</div>
5015
5016<div class="doc_text">
5017
5018<h5>Syntax:</h5>
5019<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5020 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5021</pre>
5022
5023<h5>Overview:</h5>
5024
5025<p>
5026The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5027target-specific frame pointer value for the specified stack frame.
5028</p>
5029
5030<h5>Arguments:</h5>
5031
5032<p>
5033The argument to this intrinsic indicates which function to return the frame
5034pointer for. Zero indicates the calling function, one indicates its caller,
5035etc. The argument is <b>required</b> to be a constant integer value.
5036</p>
5037
5038<h5>Semantics:</h5>
5039
5040<p>
5041The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
5042the frame address of the specified call frame, or zero if it cannot be
5043identified. The value returned by this intrinsic is likely to be incorrect or 0
5044for arguments other than zero, so it should only be used for debugging purposes.
5045</p>
5046
5047<p>
5048Note that calling this intrinsic does not prevent function inlining or other
5049aggressive transformations, so the value returned may not be that of the obvious
5050source-language caller.
5051</p>
5052</div>
5053
5054<!-- _______________________________________________________________________ -->
5055<div class="doc_subsubsection">
5056 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5057</div>
5058
5059<div class="doc_text">
5060
5061<h5>Syntax:</h5>
5062<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5063 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5064</pre>
5065
5066<h5>Overview:</h5>
5067
5068<p>
5069The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
5070the function stack, for use with <a href="#int_stackrestore">
5071<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
5072features like scoped automatic variable sized arrays in C99.
5073</p>
5074
5075<h5>Semantics:</h5>
5076
5077<p>
5078This intrinsic returns a opaque pointer value that can be passed to <a
5079href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
5080<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
5081<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
5082state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
5083practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
5084that were allocated after the <tt>llvm.stacksave</tt> was executed.
5085</p>
5086
5087</div>
5088
5089<!-- _______________________________________________________________________ -->
5090<div class="doc_subsubsection">
5091 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5092</div>
5093
5094<div class="doc_text">
5095
5096<h5>Syntax:</h5>
5097<pre>
5098 declare void @llvm.stackrestore(i8 * %ptr)
5099</pre>
5100
5101<h5>Overview:</h5>
5102
5103<p>
5104The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5105the function stack to the state it was in when the corresponding <a
5106href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
5107useful for implementing language features like scoped automatic variable sized
5108arrays in C99.
5109</p>
5110
5111<h5>Semantics:</h5>
5112
5113<p>
5114See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
5115</p>
5116
5117</div>
5118
5119
5120<!-- _______________________________________________________________________ -->
5121<div class="doc_subsubsection">
5122 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5123</div>
5124
5125<div class="doc_text">
5126
5127<h5>Syntax:</h5>
5128<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5129 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]5130</pre>
5131
5132<h5>Overview:</h5>
5133
5134
5135<p>
5136The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
5137a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
5138no
5139effect on the behavior of the program but can change its performance
5140characteristics.
5141</p>
5142
5143<h5>Arguments:</h5>
5144
5145<p>
5146<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
5147determining if the fetch should be for a read (0) or write (1), and
5148<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5149locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
5150<tt>locality</tt> arguments must be constant integers.
5151</p>
5152
5153<h5>Semantics:</h5>
5154
5155<p>
5156This intrinsic does not modify the behavior of the program. In particular,
5157prefetches cannot trap and do not produce a value. On targets that support this
5158intrinsic, the prefetch can provide hints to the processor cache for better
5159performance.
5160</p>
5161
5162</div>
5163
5164<!-- _______________________________________________________________________ -->
5165<div class="doc_subsubsection">
5166 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5167</div>
5168
5169<div class="doc_text">
5170
5171<h5>Syntax:</h5>
5172<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]5173 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5174</pre>
5175
5176<h5>Overview:</h5>
5177
5178
5179<p>
5180The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]5181(PC) in a region of
5182code to simulators and other tools. The method is target specific, but it is
5183expected that the marker will use exported symbols to transmit the PC of the
5184marker.
5185The marker makes no guarantees that it will remain with any specific instruction
5186after optimizations. It is possible that the presence of a marker will inhibit
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5187optimizations. The intended use is to be inserted after optimizations to allow
5188correlations of simulation runs.
5189</p>
5190
5191<h5>Arguments:</h5>
5192
5193<p>
5194<tt>id</tt> is a numerical id identifying the marker.
5195</p>
5196
5197<h5>Semantics:</h5>
5198
5199<p>
5200This intrinsic does not modify the behavior of the program. Backends that do not
5201support this intrinisic may ignore it.
5202</p>
5203
5204</div>
5205
5206<!-- _______________________________________________________________________ -->
5207<div class="doc_subsubsection">
5208 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5209</div>
5210
5211<div class="doc_text">
5212
5213<h5>Syntax:</h5>
5214<pre>
5215 declare i64 @llvm.readcyclecounter( )
5216</pre>
5217
5218<h5>Overview:</h5>
5219
5220
5221<p>
5222The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5223counter register (or similar low latency, high accuracy clocks) on those targets
5224that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
5225As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
5226should only be used for small timings.
5227</p>
5228
5229<h5>Semantics:</h5>
5230
5231<p>
5232When directly supported, reading the cycle counter should not modify any memory.
5233Implementations are allowed to either return a application specific value or a
5234system wide value. On backends without support, this is lowered to a constant 0.
5235</p>
5236
5237</div>
5238
5239<!-- ======================================================================= -->
5240<div class="doc_subsection">
5241 <a name="int_libc">Standard C Library Intrinsics</a>
5242</div>
5243
5244<div class="doc_text">
5245<p>
5246LLVM provides intrinsics for a few important standard C library functions.
5247These intrinsics allow source-language front-ends to pass information about the
5248alignment of the pointer arguments to the code generator, providing opportunity
5249for more efficient code generation.
5250</p>
5251
5252</div>
5253
5254<!-- _______________________________________________________________________ -->
5255<div class="doc_subsubsection">
5256 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5257</div>
5258
5259<div class="doc_text">
5260
5261<h5>Syntax:</h5>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5262<p>This is an overloaded intrinsic. You can use llvm.memcpy on any integer bit
5263width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5264<pre>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5265 declare void @llvm.memcpy.i8(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5266 i8 &lt;len&gt;, i32 &lt;align&gt;)
5267 declare void @llvm.memcpy.i16(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5268 i16 &lt;len&gt;, i32 &lt;align&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5269 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5270 i32 &lt;len&gt;, i32 &lt;align&gt;)
5271 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5272 i64 &lt;len&gt;, i32 &lt;align&gt;)
5273</pre>
5274
5275<h5>Overview:</h5>
5276
5277<p>
5278The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5279location to the destination location.
5280</p>
5281
5282<p>
5283Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5284intrinsics do not return a value, and takes an extra alignment argument.
5285</p>
5286
5287<h5>Arguments:</h5>
5288
5289<p>
5290The first argument is a pointer to the destination, the second is a pointer to
5291the source. The third argument is an integer argument
5292specifying the number of bytes to copy, and the fourth argument is the alignment
5293of the source and destination locations.
5294</p>
5295
5296<p>
5297If the call to this intrinisic has an alignment value that is not 0 or 1, then
5298the caller guarantees that both the source and destination pointers are aligned
5299to that boundary.
5300</p>
5301
5302<h5>Semantics:</h5>
5303
5304<p>
5305The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5306location to the destination location, which are not allowed to overlap. It
5307copies "len" bytes of memory over. If the argument is known to be aligned to
5308some boundary, this can be specified as the fourth argument, otherwise it should
5309be set to 0 or 1.
5310</p>
5311</div>
5312
5313
5314<!-- _______________________________________________________________________ -->
5315<div class="doc_subsubsection">
5316 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5317</div>
5318
5319<div class="doc_text">
5320
5321<h5>Syntax:</h5>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5322<p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5323width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5324<pre>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5325 declare void @llvm.memmove.i8(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5326 i8 &lt;len&gt;, i32 &lt;align&gt;)
5327 declare void @llvm.memmove.i16(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5328 i16 &lt;len&gt;, i32 &lt;align&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5329 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5330 i32 &lt;len&gt;, i32 &lt;align&gt;)
5331 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5332 i64 &lt;len&gt;, i32 &lt;align&gt;)
5333</pre>
5334
5335<h5>Overview:</h5>
5336
5337<p>
5338The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5339location to the destination location. It is similar to the
Chris Lattnerdba16ea2008-01-06 19:51:52 +0000[diff] [blame]5340'<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5341</p>
5342
5343<p>
5344Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5345intrinsics do not return a value, and takes an extra alignment argument.
5346</p>
5347
5348<h5>Arguments:</h5>
5349
5350<p>
5351The first argument is a pointer to the destination, the second is a pointer to
5352the source. The third argument is an integer argument
5353specifying the number of bytes to copy, and the fourth argument is the alignment
5354of the source and destination locations.
5355</p>
5356
5357<p>
5358If the call to this intrinisic has an alignment value that is not 0 or 1, then
5359the caller guarantees that the source and destination pointers are aligned to
5360that boundary.
5361</p>
5362
5363<h5>Semantics:</h5>
5364
5365<p>
5366The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5367location to the destination location, which may overlap. It
5368copies "len" bytes of memory over. If the argument is known to be aligned to
5369some boundary, this can be specified as the fourth argument, otherwise it should
5370be set to 0 or 1.
5371</p>
5372</div>
5373
5374
5375<!-- _______________________________________________________________________ -->
5376<div class="doc_subsubsection">
5377 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5378</div>
5379
5380<div class="doc_text">
5381
5382<h5>Syntax:</h5>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5383<p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5384width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5385<pre>
Chris Lattner82c2e432008-11-21 16:42:48 +0000[diff] [blame]5386 declare void @llvm.memset.i8(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5387 i8 &lt;len&gt;, i32 &lt;align&gt;)
5388 declare void @llvm.memset.i16(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5389 i16 &lt;len&gt;, i32 &lt;align&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5390 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5391 i32 &lt;len&gt;, i32 &lt;align&gt;)
5392 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5393 i64 &lt;len&gt;, i32 &lt;align&gt;)
5394</pre>
5395
5396<h5>Overview:</h5>
5397
5398<p>
5399The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5400byte value.
5401</p>
5402
5403<p>
5404Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5405does not return a value, and takes an extra alignment argument.
5406</p>
5407
5408<h5>Arguments:</h5>
5409
5410<p>
5411The first argument is a pointer to the destination to fill, the second is the
5412byte value to fill it with, the third argument is an integer
5413argument specifying the number of bytes to fill, and the fourth argument is the
5414known alignment of destination location.
5415</p>
5416
5417<p>
5418If the call to this intrinisic has an alignment value that is not 0 or 1, then
5419the caller guarantees that the destination pointer is aligned to that boundary.
5420</p>
5421
5422<h5>Semantics:</h5>
5423
5424<p>
5425The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5426the
5427destination location. If the argument is known to be aligned to some boundary,
5428this can be specified as the fourth argument, otherwise it should be set to 0 or
54291.
5430</p>
5431</div>
5432
5433
5434<!-- _______________________________________________________________________ -->
5435<div class="doc_subsubsection">
5436 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5437</div>
5438
5439<div class="doc_text">
5440
5441<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5442<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]5443floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5444types however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5445<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5446 declare float @llvm.sqrt.f32(float %Val)
5447 declare double @llvm.sqrt.f64(double %Val)
5448 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5449 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5450 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5451</pre>
5452
5453<h5>Overview:</h5>
5454
5455<p>
5456The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5457returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5458<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
Chris Lattnerf914a002008-01-29 07:00:44 +0000[diff] [blame]5459negative numbers other than -0.0 (which allows for better optimization, because
5460there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5461defined to return -0.0 like IEEE sqrt.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5462</p>
5463
5464<h5>Arguments:</h5>
5465
5466<p>
5467The argument and return value are floating point numbers of the same type.
5468</p>
5469
5470<h5>Semantics:</h5>
5471
5472<p>
5473This function returns the sqrt of the specified operand if it is a nonnegative
5474floating point number.
5475</p>
5476</div>
5477
5478<!-- _______________________________________________________________________ -->
5479<div class="doc_subsubsection">
5480 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5481</div>
5482
5483<div class="doc_text">
5484
5485<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5486<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]5487floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5488types however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5489<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5490 declare float @llvm.powi.f32(float %Val, i32 %power)
5491 declare double @llvm.powi.f64(double %Val, i32 %power)
5492 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5493 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5494 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5495</pre>
5496
5497<h5>Overview:</h5>
5498
5499<p>
5500The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5501specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5502multiplications is not defined. When a vector of floating point type is
5503used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5504</p>
5505
5506<h5>Arguments:</h5>
5507
5508<p>
5509The second argument is an integer power, and the first is a value to raise to
5510that power.
5511</p>
5512
5513<h5>Semantics:</h5>
5514
5515<p>
5516This function returns the first value raised to the second power with an
5517unspecified sequence of rounding operations.</p>
5518</div>
5519
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5520<!-- _______________________________________________________________________ -->
5521<div class="doc_subsubsection">
5522 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5523</div>
5524
5525<div class="doc_text">
5526
5527<h5>Syntax:</h5>
5528<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5529floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5530types however.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5531<pre>
5532 declare float @llvm.sin.f32(float %Val)
5533 declare double @llvm.sin.f64(double %Val)
5534 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5535 declare fp128 @llvm.sin.f128(fp128 %Val)
5536 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5537</pre>
5538
5539<h5>Overview:</h5>
5540
5541<p>
5542The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5543</p>
5544
5545<h5>Arguments:</h5>
5546
5547<p>
5548The argument and return value are floating point numbers of the same type.
5549</p>
5550
5551<h5>Semantics:</h5>
5552
5553<p>
5554This function returns the sine of the specified operand, returning the
5555same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5556conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5557</div>
5558
5559<!-- _______________________________________________________________________ -->
5560<div class="doc_subsubsection">
5561 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5562</div>
5563
5564<div class="doc_text">
5565
5566<h5>Syntax:</h5>
5567<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5568floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5569types however.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5570<pre>
5571 declare float @llvm.cos.f32(float %Val)
5572 declare double @llvm.cos.f64(double %Val)
5573 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5574 declare fp128 @llvm.cos.f128(fp128 %Val)
5575 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5576</pre>
5577
5578<h5>Overview:</h5>
5579
5580<p>
5581The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5582</p>
5583
5584<h5>Arguments:</h5>
5585
5586<p>
5587The argument and return value are floating point numbers of the same type.
5588</p>
5589
5590<h5>Semantics:</h5>
5591
5592<p>
5593This function returns the cosine of the specified operand, returning the
5594same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5595conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5596</div>
5597
5598<!-- _______________________________________________________________________ -->
5599<div class="doc_subsubsection">
5600 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5601</div>
5602
5603<div class="doc_text">
5604
5605<h5>Syntax:</h5>
5606<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5607floating point or vector of floating point type. Not all targets support all
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5608types however.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5609<pre>
5610 declare float @llvm.pow.f32(float %Val, float %Power)
5611 declare double @llvm.pow.f64(double %Val, double %Power)
5612 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5613 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5614 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5615</pre>
5616
5617<h5>Overview:</h5>
5618
5619<p>
5620The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5621specified (positive or negative) power.
5622</p>
5623
5624<h5>Arguments:</h5>
5625
5626<p>
5627The second argument is a floating point power, and the first is a value to
5628raise to that power.
5629</p>
5630
5631<h5>Semantics:</h5>
5632
5633<p>
5634This function returns the first value raised to the second power,
5635returning the
5636same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5637conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5638</div>
5639
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5640
5641<!-- ======================================================================= -->
5642<div class="doc_subsection">
5643 <a name="int_manip">Bit Manipulation Intrinsics</a>
5644</div>
5645
5646<div class="doc_text">
5647<p>
5648LLVM provides intrinsics for a few important bit manipulation operations.
5649These allow efficient code generation for some algorithms.
5650</p>
5651
5652</div>
5653
5654<!-- _______________________________________________________________________ -->
5655<div class="doc_subsubsection">
5656 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5657</div>
5658
5659<div class="doc_text">
5660
5661<h5>Syntax:</h5>
5662<p>This is an overloaded intrinsic function. You can use bswap on any integer
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5663type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5664<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5665 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
5666 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
5667 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5668</pre>
5669
5670<h5>Overview:</h5>
5671
5672<p>
5673The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5674values with an even number of bytes (positive multiple of 16 bits). These are
5675useful for performing operations on data that is not in the target's native
5676byte order.
5677</p>
5678
5679<h5>Semantics:</h5>
5680
5681<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5682The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5683and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5684intrinsic returns an i32 value that has the four bytes of the input i32
5685swapped, 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]5686i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5687<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5688additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5689</p>
5690
5691</div>
5692
5693<!-- _______________________________________________________________________ -->
5694<div class="doc_subsubsection">
5695 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5696</div>
5697
5698<div class="doc_text">
5699
5700<h5>Syntax:</h5>
5701<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]5702width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5703<pre>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5704 declare i8 @llvm.ctpop.i8(i8 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5705 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5706 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5707 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
5708 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5709</pre>
5710
5711<h5>Overview:</h5>
5712
5713<p>
5714The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5715value.
5716</p>
5717
5718<h5>Arguments:</h5>
5719
5720<p>
5721The only argument is the value to be counted. The argument may be of any
5722integer type. The return type must match the argument type.
5723</p>
5724
5725<h5>Semantics:</h5>
5726
5727<p>
5728The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5729</p>
5730</div>
5731
5732<!-- _______________________________________________________________________ -->
5733<div class="doc_subsubsection">
5734 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5735</div>
5736
5737<div class="doc_text">
5738
5739<h5>Syntax:</h5>
5740<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]5741integer bit width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5742<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5743 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
5744 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5745 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5746 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
5747 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5748</pre>
5749
5750<h5>Overview:</h5>
5751
5752<p>
5753The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5754leading zeros in a variable.
5755</p>
5756
5757<h5>Arguments:</h5>
5758
5759<p>
5760The only argument is the value to be counted. The argument may be of any
5761integer type. The return type must match the argument type.
5762</p>
5763
5764<h5>Semantics:</h5>
5765
5766<p>
5767The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5768in a variable. If the src == 0 then the result is the size in bits of the type
5769of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5770</p>
5771</div>
5772
5773
5774
5775<!-- _______________________________________________________________________ -->
5776<div class="doc_subsubsection">
5777 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5778</div>
5779
5780<div class="doc_text">
5781
5782<h5>Syntax:</h5>
5783<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]5784integer bit width. Not all targets support all bit widths however.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5785<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5786 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
5787 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5788 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5789 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
5790 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5791</pre>
5792
5793<h5>Overview:</h5>
5794
5795<p>
5796The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5797trailing zeros.
5798</p>
5799
5800<h5>Arguments:</h5>
5801
5802<p>
5803The only argument is the value to be counted. The argument may be of any
5804integer type. The return type must match the argument type.
5805</p>
5806
5807<h5>Semantics:</h5>
5808
5809<p>
5810The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5811in a variable. If the src == 0 then the result is the size in bits of the type
5812of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5813</p>
5814</div>
5815
5816<!-- _______________________________________________________________________ -->
5817<div class="doc_subsubsection">
5818 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5819</div>
5820
5821<div class="doc_text">
5822
5823<h5>Syntax:</h5>
5824<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5825on any integer bit width.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5826<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5827 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5828 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5829</pre>
5830
5831<h5>Overview:</h5>
5832<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5833range of bits from an integer value and returns them in the same bit width as
5834the original value.</p>
5835
5836<h5>Arguments:</h5>
5837<p>The first argument, <tt>%val</tt> and the result may be integer types of
5838any bit width but they must have the same bit width. The second and third
5839arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5840
5841<h5>Semantics:</h5>
5842<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5843of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5844<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5845operates in forward mode.</p>
5846<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5847right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5848only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5849<ol>
5850 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5851 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5852 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5853 to determine the number of bits to retain.</li>
5854 <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]5855 <li>The mask is ANDed with <tt>%val</tt> to produce the result.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5856</ol>
5857<p>In reverse mode, a similar computation is made except that the bits are
5858returned in the reverse order. So, for example, if <tt>X</tt> has the value
5859<tt>i16 0x0ACF (101011001111)</tt> and we apply
5860<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5861<tt>i16 0x0026 (000000100110)</tt>.</p>
5862</div>
5863
5864<div class="doc_subsubsection">
5865 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5866</div>
5867
5868<div class="doc_text">
5869
5870<h5>Syntax:</h5>
5871<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5872on any integer bit width.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5873<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5874 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5875 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]5876</pre>
5877
5878<h5>Overview:</h5>
5879<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5880of bits in an integer value with another integer value. It returns the integer
5881with the replaced bits.</p>
5882
5883<h5>Arguments:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5884<p>The first argument, <tt>%val</tt>, and the result may be integer types of
5885any 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]5886whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5887integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5888type since they specify only a bit index.</p>
5889
5890<h5>Semantics:</h5>
5891<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5892of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5893<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5894operates in forward mode.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5895
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5896<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5897truncating it down to the size of the replacement area or zero extending it
5898up to that size.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5899
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5900<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5901are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5902in <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]5903to the <tt>%hi</tt>th bit.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5904
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5905<p>In reverse mode, a similar computation is made except that the bits are
5906reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]5907<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]5908
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5909<h5>Examples:</h5>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5910
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5911<pre>
5912 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
5913 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
5914 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
5915 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
5916 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
5917</pre>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5918
5919</div>
5920
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]5921<!-- ======================================================================= -->
5922<div class="doc_subsection">
5923 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
5924</div>
5925
5926<div class="doc_text">
5927<p>
5928LLVM provides intrinsics for some arithmetic with overflow operations.
5929</p>
5930
5931</div>
5932
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5933<!-- _______________________________________________________________________ -->
5934<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]5935 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5936</div>
5937
5938<div class="doc_text">
5939
5940<h5>Syntax:</h5>
5941
5942<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]5943on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5944
5945<pre>
5946 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
5947 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
5948 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
5949</pre>
5950
5951<h5>Overview:</h5>
5952
5953<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
5954a signed addition of the two arguments, and indicate whether an overflow
5955occurred during the signed summation.</p>
5956
5957<h5>Arguments:</h5>
5958
5959<p>The arguments (%a and %b) and the first element of the result structure may
5960be of integer types of any bit width, but they must have the same bit width. The
5961second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
5962and <tt>%b</tt> are the two values that will undergo signed addition.</p>
5963
5964<h5>Semantics:</h5>
5965
5966<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
5967a signed addition of the two variables. They return a structure &mdash; the
5968first element of which is the signed summation, and the second element of which
5969is a bit specifying if the signed summation resulted in an overflow.</p>
5970
5971<h5>Examples:</h5>
5972<pre>
5973 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
5974 %sum = extractvalue {i32, i1} %res, 0
5975 %obit = extractvalue {i32, i1} %res, 1
5976 br i1 %obit, label %overflow, label %normal
5977</pre>
5978
5979</div>
5980
5981<!-- _______________________________________________________________________ -->
5982<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]5983 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5984</div>
5985
5986<div class="doc_text">
5987
5988<h5>Syntax:</h5>
5989
5990<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]5991on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]5992
5993<pre>
5994 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
5995 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
5996 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
5997</pre>
5998
5999<h5>Overview:</h5>
6000
6001<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6002an unsigned addition of the two arguments, and indicate whether a carry occurred
6003during the unsigned summation.</p>
6004
6005<h5>Arguments:</h5>
6006
6007<p>The arguments (%a and %b) and the first element of the result structure may
6008be of integer types of any bit width, but they must have the same bit width. The
6009second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6010and <tt>%b</tt> are the two values that will undergo unsigned addition.</p>
6011
6012<h5>Semantics:</h5>
6013
6014<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6015an unsigned addition of the two arguments. They return a structure &mdash; the
6016first element of which is the sum, and the second element of which is a bit
6017specifying if the unsigned summation resulted in a carry.</p>
6018
6019<h5>Examples:</h5>
6020<pre>
6021 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6022 %sum = extractvalue {i32, i1} %res, 0
6023 %obit = extractvalue {i32, i1} %res, 1
6024 br i1 %obit, label %carry, label %normal
6025</pre>
6026
6027</div>
6028
6029<!-- _______________________________________________________________________ -->
6030<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6031 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6032</div>
6033
6034<div class="doc_text">
6035
6036<h5>Syntax:</h5>
6037
6038<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]6039on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6040
6041<pre>
6042 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6043 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6044 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6045</pre>
6046
6047<h5>Overview:</h5>
6048
6049<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6050a signed subtraction of the two arguments, and indicate whether an overflow
6051occurred during the signed subtraction.</p>
6052
6053<h5>Arguments:</h5>
6054
6055<p>The arguments (%a and %b) and the first element of the result structure may
6056be of integer types of any bit width, but they must have the same bit width. The
6057second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6058and <tt>%b</tt> are the two values that will undergo signed subtraction.</p>
6059
6060<h5>Semantics:</h5>
6061
6062<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6063a signed subtraction of the two arguments. They return a structure &mdash; the
6064first element of which is the subtraction, and the second element of which is a bit
6065specifying if the signed subtraction resulted in an overflow.</p>
6066
6067<h5>Examples:</h5>
6068<pre>
6069 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6070 %sum = extractvalue {i32, i1} %res, 0
6071 %obit = extractvalue {i32, i1} %res, 1
6072 br i1 %obit, label %overflow, label %normal
6073</pre>
6074
6075</div>
6076
6077<!-- _______________________________________________________________________ -->
6078<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6079 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6080</div>
6081
6082<div class="doc_text">
6083
6084<h5>Syntax:</h5>
6085
6086<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]6087on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6088
6089<pre>
6090 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6091 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6092 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6093</pre>
6094
6095<h5>Overview:</h5>
6096
6097<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6098an unsigned subtraction of the two arguments, and indicate whether an overflow
6099occurred during the unsigned subtraction.</p>
6100
6101<h5>Arguments:</h5>
6102
6103<p>The arguments (%a and %b) and the first element of the result structure may
6104be of integer types of any bit width, but they must have the same bit width. The
6105second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6106and <tt>%b</tt> are the two values that will undergo unsigned subtraction.</p>
6107
6108<h5>Semantics:</h5>
6109
6110<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6111an unsigned subtraction of the two arguments. They return a structure &mdash; the
6112first element of which is the subtraction, and the second element of which is a bit
6113specifying if the unsigned subtraction resulted in an overflow.</p>
6114
6115<h5>Examples:</h5>
6116<pre>
6117 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6118 %sum = extractvalue {i32, i1} %res, 0
6119 %obit = extractvalue {i32, i1} %res, 1
6120 br i1 %obit, label %overflow, label %normal
6121</pre>
6122
6123</div>
6124
6125<!-- _______________________________________________________________________ -->
6126<div class="doc_subsubsection">
Bill Wendling3e1258b2009-02-08 04:04:40 +0000[diff] [blame]6127 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6128</div>
6129
6130<div class="doc_text">
6131
6132<h5>Syntax:</h5>
6133
6134<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]6135on any integer bit width.</p>
Bill Wendling3f8cebe2009-02-08 01:40:31 +0000[diff] [blame]6136
6137<pre>
6138 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6139 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6140 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6141</pre>
6142
6143<h5>Overview:</h5>
6144
6145<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6146a signed multiplication of the two arguments, and indicate whether an overflow
6147occurred during the signed multiplication.</p>
6148
6149<h5>Arguments:</h5>
6150
6151<p>The arguments (%a and %b) and the first element of the result structure may
6152be of integer types of any bit width, but they must have the same bit width. The
6153second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6154and <tt>%b</tt> are the two values that will undergo signed multiplication.</p>
6155
6156<h5>Semantics:</h5>
6157
6158<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6159a signed multiplication of the two arguments. They return a structure &mdash;
6160the first element of which is the multiplication, and the second element of
6161which is a bit specifying if the signed multiplication resulted in an
6162overflow.</p>
6163
6164<h5>Examples:</h5>
6165<pre>
6166 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6167 %sum = extractvalue {i32, i1} %res, 0
6168 %obit = extractvalue {i32, i1} %res, 1
6169 br i1 %obit, label %overflow, label %normal
6170</pre>
6171
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6172</div>
6173
Bill Wendlingbda98b62009-02-08 23:00:09 +0000[diff] [blame^]6174<!-- _______________________________________________________________________ -->
6175<div class="doc_subsubsection">
6176 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6177</div>
6178
6179<div class="doc_text">
6180
6181<h5>Syntax:</h5>
6182
6183<p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6184on any integer bit width.</p>
6185
6186<pre>
6187 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6188 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6189 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6190</pre>
6191
6192<h5>Overview:</h5>
6193
6194<p><i><b>Warning:</b> '<tt>llvm.umul.with.overflow</tt>' is badly broken. It is
6195actively being fixed, but it should not currently be used!</i></p>
6196
6197<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6198a unsigned multiplication of the two arguments, and indicate whether an overflow
6199occurred during the unsigned multiplication.</p>
6200
6201<h5>Arguments:</h5>
6202
6203<p>The arguments (%a and %b) and the first element of the result structure may
6204be of integer types of any bit width, but they must have the same bit width. The
6205second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6206and <tt>%b</tt> are the two values that will undergo unsigned
6207multiplication.</p>
6208
6209<h5>Semantics:</h5>
6210
6211<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6212an unsigned multiplication of the two arguments. They return a structure &mdash;
6213the first element of which is the multiplication, and the second element of
6214which is a bit specifying if the unsigned multiplication resulted in an
6215overflow.</p>
6216
6217<h5>Examples:</h5>
6218<pre>
6219 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6220 %sum = extractvalue {i32, i1} %res, 0
6221 %obit = extractvalue {i32, i1} %res, 1
6222 br i1 %obit, label %overflow, label %normal
6223</pre>
6224
6225</div>
6226
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6227<!-- ======================================================================= -->
6228<div class="doc_subsection">
6229 <a name="int_debugger">Debugger Intrinsics</a>
6230</div>
6231
6232<div class="doc_text">
6233<p>
6234The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
6235are described in the <a
6236href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
6237Debugging</a> document.
6238</p>
6239</div>
6240
6241
6242<!-- ======================================================================= -->
6243<div class="doc_subsection">
6244 <a name="int_eh">Exception Handling Intrinsics</a>
6245</div>
6246
6247<div class="doc_text">
6248<p> The LLVM exception handling intrinsics (which all start with
6249<tt>llvm.eh.</tt> prefix), are described in the <a
6250href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6251Handling</a> document. </p>
6252</div>
6253
6254<!-- ======================================================================= -->
6255<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6256 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6257</div>
6258
6259<div class="doc_text">
6260<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6261 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6262 the <tt>nest</tt> attribute, from a function. The result is a callable
6263 function pointer lacking the nest parameter - the caller does not need
6264 to provide a value for it. Instead, the value to use is stored in
6265 advance in a "trampoline", a block of memory usually allocated
6266 on the stack, which also contains code to splice the nest value into the
6267 argument list. This is used to implement the GCC nested function address
6268 extension.
6269</p>
6270<p>
6271 For example, if the function is
6272 <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]6273 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]6274<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6275 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6276 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6277 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
6278 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6279</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]6280 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
6281 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6282</div>
6283
6284<!-- _______________________________________________________________________ -->
6285<div class="doc_subsubsection">
6286 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6287</div>
6288<div class="doc_text">
6289<h5>Syntax:</h5>
6290<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6291declare 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]6292</pre>
6293<h5>Overview:</h5>
6294<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6295 This fills the memory pointed to by <tt>tramp</tt> with code
6296 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6297</p>
6298<h5>Arguments:</h5>
6299<p>
6300 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6301 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
6302 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]6303 intrinsic. Note that the size and the alignment are target-specific - LLVM
6304 currently provides no portable way of determining them, so a front-end that
6305 generates this intrinsic needs to have some target-specific knowledge.
6306 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]6307</p>
6308<h5>Semantics:</h5>
6309<p>
6310 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6311 dependent code, turning it into a function. A pointer to this function is
6312 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6313 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]6314 before being called. The new function's signature is the same as that of
6315 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
6316 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
6317 of pointer type. Calling the new function is equivalent to calling
6318 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
6319 missing <tt>nest</tt> argument. If, after calling
6320 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
6321 modified, then the effect of any later call to the returned function pointer is
6322 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]6323</p>
6324</div>
6325
6326<!-- ======================================================================= -->
6327<div class="doc_subsection">
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6328 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6329</div>
6330
6331<div class="doc_text">
6332<p>
6333 These intrinsic functions expand the "universal IR" of LLVM to represent
6334 hardware constructs for atomic operations and memory synchronization. This
6335 provides an interface to the hardware, not an interface to the programmer. It
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]6336 is aimed at a low enough level to allow any programming models or APIs
6337 (Application Programming Interfaces) which
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6338 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
6339 hardware behavior. Just as hardware provides a "universal IR" for source
6340 languages, it also provides a starting point for developing a "universal"
6341 atomic operation and synchronization IR.
6342</p>
6343<p>
6344 These do <em>not</em> form an API such as high-level threading libraries,
6345 software transaction memory systems, atomic primitives, and intrinsic
6346 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6347 application libraries. The hardware interface provided by LLVM should allow
6348 a clean implementation of all of these APIs and parallel programming models.
6349 No one model or paradigm should be selected above others unless the hardware
6350 itself ubiquitously does so.
6351
6352</p>
6353</div>
6354
6355<!-- _______________________________________________________________________ -->
6356<div class="doc_subsubsection">
6357 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6358</div>
6359<div class="doc_text">
6360<h5>Syntax:</h5>
6361<pre>
6362declare void @llvm.memory.barrier( i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt;,
6363i1 &lt;device&gt; )
6364
6365</pre>
6366<h5>Overview:</h5>
6367<p>
6368 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6369 specific pairs of memory access types.
6370</p>
6371<h5>Arguments:</h5>
6372<p>
6373 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6374 The first four arguments enables a specific barrier as listed below. The fith
6375 argument specifies that the barrier applies to io or device or uncached memory.
6376
6377</p>
6378 <ul>
6379 <li><tt>ll</tt>: load-load barrier</li>
6380 <li><tt>ls</tt>: load-store barrier</li>
6381 <li><tt>sl</tt>: store-load barrier</li>
6382 <li><tt>ss</tt>: store-store barrier</li>
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]6383 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6384 </ul>
6385<h5>Semantics:</h5>
6386<p>
6387 This intrinsic causes the system to enforce some ordering constraints upon
6388 the loads and stores of the program. This barrier does not indicate
6389 <em>when</em> any events will occur, it only enforces an <em>order</em> in
6390 which they occur. For any of the specified pairs of load and store operations
6391 (f.ex. load-load, or store-load), all of the first operations preceding the
6392 barrier will complete before any of the second operations succeeding the
6393 barrier begin. Specifically the semantics for each pairing is as follows:
6394</p>
6395 <ul>
6396 <li><tt>ll</tt>: All loads before the barrier must complete before any load
6397 after the barrier begins.</li>
6398
6399 <li><tt>ls</tt>: All loads before the barrier must complete before any
6400 store after the barrier begins.</li>
6401 <li><tt>ss</tt>: All stores before the barrier must complete before any
6402 store after the barrier begins.</li>
6403 <li><tt>sl</tt>: All stores before the barrier must complete before any
6404 load after the barrier begins.</li>
6405 </ul>
6406<p>
6407 These semantics are applied with a logical "and" behavior when more than one
6408 is enabled in a single memory barrier intrinsic.
6409</p>
6410<p>
6411 Backends may implement stronger barriers than those requested when they do not
6412 support as fine grained a barrier as requested. Some architectures do not
6413 need all types of barriers and on such architectures, these become noops.
6414</p>
6415<h5>Example:</h5>
6416<pre>
6417%ptr = malloc i32
6418 store i32 4, %ptr
6419
6420%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
6421 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
6422 <i>; guarantee the above finishes</i>
6423 store i32 8, %ptr <i>; before this begins</i>
6424</pre>
6425</div>
6426
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6427<!-- _______________________________________________________________________ -->
6428<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6429 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6430</div>
6431<div class="doc_text">
6432<h5>Syntax:</h5>
6433<p>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6434 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
6435 any integer bit width and for different address spaces. Not all targets
6436 support all bit widths however.</p>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6437
6438<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6439declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt; )
6440declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt; )
6441declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt; )
6442declare 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]6443
6444</pre>
6445<h5>Overview:</h5>
6446<p>
6447 This loads a value in memory and compares it to a given value. If they are
6448 equal, it stores a new value into the memory.
6449</p>
6450<h5>Arguments:</h5>
6451<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6452 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6453 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6454 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6455 this integer type. While any bit width integer may be used, targets may only
6456 lower representations they support in hardware.
6457
6458</p>
6459<h5>Semantics:</h5>
6460<p>
6461 This entire intrinsic must be executed atomically. It first loads the value
6462 in memory pointed to by <tt>ptr</tt> and compares it with the value
6463 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
6464 loaded value is yielded in all cases. This provides the equivalent of an
6465 atomic compare-and-swap operation within the SSA framework.
6466</p>
6467<h5>Examples:</h5>
6468
6469<pre>
6470%ptr = malloc i32
6471 store i32 4, %ptr
6472
6473%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6474%result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6475 <i>; yields {i32}:result1 = 4</i>
6476%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6477%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6478
6479%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6480%result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6481 <i>; yields {i32}:result2 = 8</i>
6482%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6483
6484%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6485</pre>
6486</div>
6487
6488<!-- _______________________________________________________________________ -->
6489<div class="doc_subsubsection">
6490 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6491</div>
6492<div class="doc_text">
6493<h5>Syntax:</h5>
6494
6495<p>
6496 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6497 integer bit width. Not all targets support all bit widths however.</p>
6498<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6499declare i8 @llvm.atomic.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;val&gt; )
6500declare i16 @llvm.atomic.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;val&gt; )
6501declare i32 @llvm.atomic.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;val&gt; )
6502declare i64 @llvm.atomic.swap.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;val&gt; )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6503
6504</pre>
6505<h5>Overview:</h5>
6506<p>
6507 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6508 the value from memory. It then stores the value in <tt>val</tt> in the memory
6509 at <tt>ptr</tt>.
6510</p>
6511<h5>Arguments:</h5>
6512
6513<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6514 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6515 <tt>val</tt> argument and the result must be integers of the same bit width.
6516 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6517 integer type. The targets may only lower integer representations they
6518 support.
6519</p>
6520<h5>Semantics:</h5>
6521<p>
6522 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6523 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6524 equivalent of an atomic swap operation within the SSA framework.
6525
6526</p>
6527<h5>Examples:</h5>
6528<pre>
6529%ptr = malloc i32
6530 store i32 4, %ptr
6531
6532%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6533%result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6534 <i>; yields {i32}:result1 = 4</i>
6535%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6536%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6537
6538%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6539%result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6540 <i>; yields {i32}:result2 = 8</i>
6541
6542%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6543%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6544</pre>
6545</div>
6546
6547<!-- _______________________________________________________________________ -->
6548<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6549 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6550
6551</div>
6552<div class="doc_text">
6553<h5>Syntax:</h5>
6554<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6555 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]6556 integer bit width. Not all targets support all bit widths however.</p>
6557<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6558declare i8 @llvm.atomic.load.add.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6559declare i16 @llvm.atomic.load.add.i16..p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6560declare i32 @llvm.atomic.load.add.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6561declare 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]6562
6563</pre>
6564<h5>Overview:</h5>
6565<p>
6566 This intrinsic adds <tt>delta</tt> to the value stored in memory at
6567 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6568</p>
6569<h5>Arguments:</h5>
6570<p>
6571
6572 The intrinsic takes two arguments, the first a pointer to an integer value
6573 and the second an integer value. The result is also an integer value. These
6574 integer types can have any bit width, but they must all have the same bit
6575 width. The targets may only lower integer representations they support.
6576</p>
6577<h5>Semantics:</h5>
6578<p>
6579 This intrinsic does a series of operations atomically. It first loads the
6580 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6581 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6582</p>
6583
6584<h5>Examples:</h5>
6585<pre>
6586%ptr = malloc i32
6587 store i32 4, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6588%result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6589 <i>; yields {i32}:result1 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6590%result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6591 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6592%result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6593 <i>; yields {i32}:result3 = 10</i>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6594%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6595</pre>
6596</div>
6597
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6598<!-- _______________________________________________________________________ -->
6599<div class="doc_subsubsection">
6600 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6601
6602</div>
6603<div class="doc_text">
6604<h5>Syntax:</h5>
6605<p>
6606 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]6607 any integer bit width and for different address spaces. Not all targets
6608 support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6609<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6610declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6611declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6612declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6613declare 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]6614
6615</pre>
6616<h5>Overview:</h5>
6617<p>
6618 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6619 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6620</p>
6621<h5>Arguments:</h5>
6622<p>
6623
6624 The intrinsic takes two arguments, the first a pointer to an integer value
6625 and the second an integer value. The result is also an integer value. These
6626 integer types can have any bit width, but they must all have the same bit
6627 width. The targets may only lower integer representations they support.
6628</p>
6629<h5>Semantics:</h5>
6630<p>
6631 This intrinsic does a series of operations atomically. It first loads the
6632 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6633 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6634</p>
6635
6636<h5>Examples:</h5>
6637<pre>
6638%ptr = malloc i32
6639 store i32 8, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6640%result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6641 <i>; yields {i32}:result1 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6642%result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6643 <i>; yields {i32}:result2 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6644%result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6645 <i>; yields {i32}:result3 = 2</i>
6646%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6647</pre>
6648</div>
6649
6650<!-- _______________________________________________________________________ -->
6651<div class="doc_subsubsection">
6652 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6653 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6654 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6655 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6656
6657</div>
6658<div class="doc_text">
6659<h5>Syntax:</h5>
6660<p>
6661 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6662 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6663 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6664 address spaces. Not all targets support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6665<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6666declare i8 @llvm.atomic.load.and.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6667declare i16 @llvm.atomic.load.and.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6668declare i32 @llvm.atomic.load.and.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6669declare 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]6670
6671</pre>
6672
6673<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6674declare i8 @llvm.atomic.load.or.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6675declare i16 @llvm.atomic.load.or.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6676declare i32 @llvm.atomic.load.or.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6677declare 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]6678
6679</pre>
6680
6681<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6682declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6683declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6684declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6685declare 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]6686
6687</pre>
6688
6689<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6690declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6691declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6692declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6693declare 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]6694
6695</pre>
6696<h5>Overview:</h5>
6697<p>
6698 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6699 the value stored in memory at <tt>ptr</tt>. It yields the original value
6700 at <tt>ptr</tt>.
6701</p>
6702<h5>Arguments:</h5>
6703<p>
6704
6705 These intrinsics take two arguments, the first a pointer to an integer value
6706 and the second an integer value. The result is also an integer value. These
6707 integer types can have any bit width, but they must all have the same bit
6708 width. The targets may only lower integer representations they support.
6709</p>
6710<h5>Semantics:</h5>
6711<p>
6712 These intrinsics does a series of operations atomically. They first load the
6713 value stored at <tt>ptr</tt>. They then do the bitwise operation
6714 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6715 value stored at <tt>ptr</tt>.
6716</p>
6717
6718<h5>Examples:</h5>
6719<pre>
6720%ptr = malloc i32
6721 store i32 0x0F0F, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6722%result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6723 <i>; yields {i32}:result0 = 0x0F0F</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6724%result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6725 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6726%result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6727 <i>; yields {i32}:result2 = 0xF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6728%result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6729 <i>; yields {i32}:result3 = FF</i>
6730%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6731</pre>
6732</div>
6733
6734
6735<!-- _______________________________________________________________________ -->
6736<div class="doc_subsubsection">
6737 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6738 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6739 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6740 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6741
6742</div>
6743<div class="doc_text">
6744<h5>Syntax:</h5>
6745<p>
6746 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6747 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6748 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6749 address spaces. Not all targets
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6750 support all bit widths however.</p>
6751<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6752declare i8 @llvm.atomic.load.max.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6753declare i16 @llvm.atomic.load.max.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6754declare i32 @llvm.atomic.load.max.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6755declare 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]6756
6757</pre>
6758
6759<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6760declare i8 @llvm.atomic.load.min.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6761declare i16 @llvm.atomic.load.min.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6762declare i32 @llvm.atomic.load.min.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6763declare 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]6764
6765</pre>
6766
6767<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6768declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6769declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6770declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6771declare 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]6772
6773</pre>
6774
6775<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6776declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6777declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6778declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6779declare 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]6780
6781</pre>
6782<h5>Overview:</h5>
6783<p>
6784 These intrinsics takes the signed or unsigned minimum or maximum of
6785 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6786 original value at <tt>ptr</tt>.
6787</p>
6788<h5>Arguments:</h5>
6789<p>
6790
6791 These intrinsics take two arguments, the first a pointer to an integer value
6792 and the second an integer value. The result is also an integer value. These
6793 integer types can have any bit width, but they must all have the same bit
6794 width. The targets may only lower integer representations they support.
6795</p>
6796<h5>Semantics:</h5>
6797<p>
6798 These intrinsics does a series of operations atomically. They first load the
6799 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6800 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6801 the original value stored at <tt>ptr</tt>.
6802</p>
6803
6804<h5>Examples:</h5>
6805<pre>
6806%ptr = malloc i32
6807 store i32 7, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6808%result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6809 <i>; yields {i32}:result0 = 7</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6810%result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6811 <i>; yields {i32}:result1 = -2</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6812%result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6813 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6814%result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6815 <i>; yields {i32}:result3 = 8</i>
6816%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6817</pre>
6818</div>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6819
6820<!-- ======================================================================= -->
6821<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6822 <a name="int_general">General Intrinsics</a>
6823</div>
6824
6825<div class="doc_text">
6826<p> This class of intrinsics is designed to be generic and has
6827no specific purpose. </p>
6828</div>
6829
6830<!-- _______________________________________________________________________ -->
6831<div class="doc_subsubsection">
6832 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6833</div>
6834
6835<div class="doc_text">
6836
6837<h5>Syntax:</h5>
6838<pre>
6839 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6840</pre>
6841
6842<h5>Overview:</h5>
6843
6844<p>
6845The '<tt>llvm.var.annotation</tt>' intrinsic
6846</p>
6847
6848<h5>Arguments:</h5>
6849
6850<p>
6851The first argument is a pointer to a value, the second is a pointer to a
6852global string, the third is a pointer to a global string which is the source
6853file name, and the last argument is the line number.
6854</p>
6855
6856<h5>Semantics:</h5>
6857
6858<p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6859This intrinsic allows annotation of local variables with arbitrary strings.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6860This can be useful for special purpose optimizations that want to look for these
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6861annotations. These have no other defined use, they are ignored by code
6862generation and optimization.
6863</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6864</div>
6865
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6866<!-- _______________________________________________________________________ -->
6867<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]6868 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6869</div>
6870
6871<div class="doc_text">
6872
6873<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6874<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6875any integer bit width.
6876</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6877<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]6878 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6879 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6880 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6881 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6882 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]6883</pre>
6884
6885<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6886
6887<p>
6888The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6889</p>
6890
6891<h5>Arguments:</h5>
6892
6893<p>
6894The first argument is an integer value (result of some expression),
6895the second is a pointer to a global string, the third is a pointer to a global
6896string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6897It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6898</p>
6899
6900<h5>Semantics:</h5>
6901
6902<p>
6903This intrinsic allows annotations to be put on arbitrary expressions
6904with arbitrary strings. This can be useful for special purpose optimizations
6905that want to look for these annotations. These have no other defined use, they
6906are ignored by code generation and optimization.
Dan Gohman2672f3e2008-10-14 16:51:45 +0000[diff] [blame]6907</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6908</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6909
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6910<!-- _______________________________________________________________________ -->
6911<div class="doc_subsubsection">
6912 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6913</div>
6914
6915<div class="doc_text">
6916
6917<h5>Syntax:</h5>
6918<pre>
6919 declare void @llvm.trap()
6920</pre>
6921
6922<h5>Overview:</h5>
6923
6924<p>
6925The '<tt>llvm.trap</tt>' intrinsic
6926</p>
6927
6928<h5>Arguments:</h5>
6929
6930<p>
6931None
6932</p>
6933
6934<h5>Semantics:</h5>
6935
6936<p>
6937This intrinsics is lowered to the target dependent trap instruction. If the
6938target does not have a trap instruction, this intrinsic will be lowered to the
6939call of the abort() function.
6940</p>
6941</div>
6942
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]6943<!-- _______________________________________________________________________ -->
6944<div class="doc_subsubsection">
Misha Brukman5dd7f4d2008-11-22 23:55:29 +0000[diff] [blame]6945 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
Bill Wendlinge4164592008-11-19 05:56:17 +0000[diff] [blame]6946</div>
6947<div class="doc_text">
6948<h5>Syntax:</h5>
6949<pre>
6950declare void @llvm.stackprotector( i8* &lt;guard&gt;, i8** &lt;slot&gt; )
6951
6952</pre>
6953<h5>Overview:</h5>
6954<p>
6955 The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and stores
6956 it onto the stack at <tt>slot</tt>. The stack slot is adjusted to ensure that
6957 it is placed on the stack before local variables.
6958</p>
6959<h5>Arguments:</h5>
6960<p>
6961 The <tt>llvm.stackprotector</tt> intrinsic requires two pointer arguments. The
6962 first argument is the value loaded from the stack guard
6963 <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt> that
6964 has enough space to hold the value of the guard.
6965</p>
6966<h5>Semantics:</h5>
6967<p>
6968 This intrinsic causes the prologue/epilogue inserter to force the position of
6969 the <tt>AllocaInst</tt> stack slot to be before local variables on the
6970 stack. This is to ensure that if a local variable on the stack is overwritten,
6971 it will destroy the value of the guard. When the function exits, the guard on
6972 the stack is checked against the original guard. If they're different, then
6973 the program aborts by calling the <tt>__stack_chk_fail()</tt> function.
6974</p>
6975</div>
6976
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6977<!-- *********************************************************************** -->
6978<hr>
6979<address>
6980 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman947321d2008-12-11 17:34:48 +0000[diff] [blame]6981 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6982 <a href="http://validator.w3.org/check/referer"><img
Misha Brukman947321d2008-12-11 17:34:48 +0000[diff] [blame]6983 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]6984
6985 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
6986 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
6987 Last modified: $Date$
6988</address>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]6989
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6990</body>
6991</html>