Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / llvm / f33b84582978b8d35462d261c789893634eee5f3 / . / docs / LangRef.html
blob: e0945b1c5a4e82df5ce7821c33a453f94131830f [file] [log] [blame]
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
3<html>
4<head>
5 <title>LLVM Assembly Language Reference Manual</title>
6 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
8 <meta name="description"
9 content="LLVM Assembly Language Reference Manual.">
10 <link rel="stylesheet" href="llvm.css" type="text/css">
11</head>
12
13<body>
14
15<div class="doc_title"> LLVM Language Reference Manual </div>
16<ol>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
20 <li><a href="#highlevel">High Level Structure</a>
21 <ol>
22 <li><a href="#modulestructure">Module Structure</a></li>
23 <li><a href="#linkage">Linkage Types</a></li>
24 <li><a href="#callingconv">Calling Conventions</a></li>
25 <li><a href="#globalvars">Global Variables</a></li>
26 <li><a href="#functionstructure">Functions</a></li>
27 <li><a href="#aliasstructure">Aliases</a>
28 <li><a href="#paramattrs">Parameter Attributes</a></li>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]29 <li><a href="#fnattrs">Function Attributes</a></li>
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]30 <li><a href="#gc">Garbage Collector Names</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]31 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
32 <li><a href="#datalayout">Data Layout</a></li>
33 </ol>
34 </li>
35 <li><a href="#typesystem">Type System</a>
36 <ol>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]37 <li><a href="#t_classifications">Type Classifications</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]38 <li><a href="#t_primitive">Primitive Types</a>
39 <ol>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]40 <li><a href="#t_floating">Floating Point Types</a></li>
41 <li><a href="#t_void">Void Type</a></li>
42 <li><a href="#t_label">Label Type</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]43 </ol>
44 </li>
45 <li><a href="#t_derived">Derived Types</a>
46 <ol>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]47 <li><a href="#t_integer">Integer Type</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]48 <li><a href="#t_array">Array Type</a></li>
49 <li><a href="#t_function">Function Type</a></li>
50 <li><a href="#t_pointer">Pointer Type</a></li>
51 <li><a href="#t_struct">Structure Type</a></li>
52 <li><a href="#t_pstruct">Packed Structure Type</a></li>
53 <li><a href="#t_vector">Vector Type</a></li>
54 <li><a href="#t_opaque">Opaque Type</a></li>
55 </ol>
56 </li>
57 </ol>
58 </li>
59 <li><a href="#constants">Constants</a>
60 <ol>
61 <li><a href="#simpleconstants">Simple Constants</a>
62 <li><a href="#aggregateconstants">Aggregate Constants</a>
63 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
64 <li><a href="#undefvalues">Undefined Values</a>
65 <li><a href="#constantexprs">Constant Expressions</a>
66 </ol>
67 </li>
68 <li><a href="#othervalues">Other Values</a>
69 <ol>
70 <li><a href="#inlineasm">Inline Assembler Expressions</a>
71 </ol>
72 </li>
73 <li><a href="#instref">Instruction Reference</a>
74 <ol>
75 <li><a href="#terminators">Terminator Instructions</a>
76 <ol>
77 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
78 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
79 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
80 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
81 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
82 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
83 </ol>
84 </li>
85 <li><a href="#binaryops">Binary Operations</a>
86 <ol>
87 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
88 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
89 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
90 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
91 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
92 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
93 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
94 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
95 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
96 </ol>
97 </li>
98 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
99 <ol>
100 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
101 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
102 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
103 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
104 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
105 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
106 </ol>
107 </li>
108 <li><a href="#vectorops">Vector Operations</a>
109 <ol>
110 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
111 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
112 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
113 </ol>
114 </li>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]115 <li><a href="#aggregateops">Aggregate Operations</a>
116 <ol>
117 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
118 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
119 </ol>
120 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]121 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
122 <ol>
123 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
124 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
125 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
126 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
127 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
128 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
129 </ol>
130 </li>
131 <li><a href="#convertops">Conversion Operations</a>
132 <ol>
133 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
134 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
135 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
136 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
137 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
138 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
139 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
140 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
141 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
142 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
143 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
144 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
145 </ol>
146 <li><a href="#otherops">Other Operations</a>
147 <ol>
148 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
149 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]150 <li><a href="#i_vicmp">'<tt>vicmp</tt>' Instruction</a></li>
151 <li><a href="#i_vfcmp">'<tt>vfcmp</tt>' Instruction</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]152 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
153 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
154 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
155 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]156 <li><a href="#i_getresult">'<tt>getresult</tt>' Instruction</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]157 </ol>
158 </li>
159 </ol>
160 </li>
161 <li><a href="#intrinsics">Intrinsic Functions</a>
162 <ol>
163 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
164 <ol>
165 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
166 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
167 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
168 </ol>
169 </li>
170 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
171 <ol>
172 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
173 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
174 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
175 </ol>
176 </li>
177 <li><a href="#int_codegen">Code Generator Intrinsics</a>
178 <ol>
179 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
180 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
181 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
182 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
183 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
184 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
185 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
186 </ol>
187 </li>
188 <li><a href="#int_libc">Standard C Library Intrinsics</a>
189 <ol>
190 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
191 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
192 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
193 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
194 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]195 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
196 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
197 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]198 </ol>
199 </li>
200 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
201 <ol>
202 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
203 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
204 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
205 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
206 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
207 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
208 </ol>
209 </li>
210 <li><a href="#int_debugger">Debugger intrinsics</a></li>
211 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]212 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]213 <ol>
214 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]215 </ol>
216 </li>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]217 <li><a href="#int_atomics">Atomic intrinsics</a>
218 <ol>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]219 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]220 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]221 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]222 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
223 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
224 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
225 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
226 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
227 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
228 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
229 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
230 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
231 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]232 </ol>
233 </li>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]234 <li><a href="#int_general">General intrinsics</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]235 <ol>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]236 <li><a href="#int_var_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]237 <tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]238 <li><a href="#int_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]239 <tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]240 <li><a href="#int_trap">
241 <tt>llvm.trap</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]242 </ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]243 </li>
244 </ol>
245 </li>
246</ol>
247
248<div class="doc_author">
249 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
250 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
251</div>
252
253<!-- *********************************************************************** -->
254<div class="doc_section"> <a name="abstract">Abstract </a></div>
255<!-- *********************************************************************** -->
256
257<div class="doc_text">
258<p>This document is a reference manual for the LLVM assembly language.
Bill Wendlinge7846a52008-08-05 22:29:16 +0000[diff] [blame]259LLVM is a Static Single Assignment (SSA) based representation that provides
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]260type safety, low-level operations, flexibility, and the capability of
261representing 'all' high-level languages cleanly. It is the common code
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]262representation used throughout all phases of the LLVM compilation
263strategy.</p>
264</div>
265
266<!-- *********************************************************************** -->
267<div class="doc_section"> <a name="introduction">Introduction</a> </div>
268<!-- *********************************************************************** -->
269
270<div class="doc_text">
271
272<p>The LLVM code representation is designed to be used in three
273different forms: as an in-memory compiler IR, as an on-disk bitcode
274representation (suitable for fast loading by a Just-In-Time compiler),
275and as a human readable assembly language representation. This allows
276LLVM to provide a powerful intermediate representation for efficient
277compiler transformations and analysis, while providing a natural means
278to debug and visualize the transformations. The three different forms
279of LLVM are all equivalent. This document describes the human readable
280representation and notation.</p>
281
282<p>The LLVM representation aims to be light-weight and low-level
283while being expressive, typed, and extensible at the same time. It
284aims to be a "universal IR" of sorts, by being at a low enough level
285that high-level ideas may be cleanly mapped to it (similar to how
286microprocessors are "universal IR's", allowing many source languages to
287be mapped to them). By providing type information, LLVM can be used as
288the target of optimizations: for example, through pointer analysis, it
289can be proven that a C automatic variable is never accessed outside of
290the current function... allowing it to be promoted to a simple SSA
291value instead of a memory location.</p>
292
293</div>
294
295<!-- _______________________________________________________________________ -->
296<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
297
298<div class="doc_text">
299
300<p>It is important to note that this document describes 'well formed'
301LLVM assembly language. There is a difference between what the parser
302accepts and what is considered 'well formed'. For example, the
303following instruction is syntactically okay, but not well formed:</p>
304
305<div class="doc_code">
306<pre>
307%x = <a href="#i_add">add</a> i32 1, %x
308</pre>
309</div>
310
311<p>...because the definition of <tt>%x</tt> does not dominate all of
312its uses. The LLVM infrastructure provides a verification pass that may
313be used to verify that an LLVM module is well formed. This pass is
314automatically run by the parser after parsing input assembly and by
315the optimizer before it outputs bitcode. The violations pointed out
316by the verifier pass indicate bugs in transformation passes or input to
317the parser.</p>
318</div>
319
Chris Lattnera83fdc02007-10-03 17:34:29 +0000[diff] [blame]320<!-- Describe the typesetting conventions here. -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]321
322<!-- *********************************************************************** -->
323<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
324<!-- *********************************************************************** -->
325
326<div class="doc_text">
327
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]328 <p>LLVM identifiers come in two basic types: global and local. Global
329 identifiers (functions, global variables) begin with the @ character. Local
330 identifiers (register names, types) begin with the % character. Additionally,
331 there are three different formats for identifiers, for different purposes:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]332
333<ol>
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]334 <li>Named values are represented as a string of characters with their prefix.
335 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
336 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]337 Identifiers which require other characters in their names can be surrounded
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]338 with quotes. In this way, anything except a <tt>&quot;</tt> character can
339 be used in a named value.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]340
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]341 <li>Unnamed values are represented as an unsigned numeric value with their
342 prefix. For example, %12, @2, %44.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]343
344 <li>Constants, which are described in a <a href="#constants">section about
345 constants</a>, below.</li>
346</ol>
347
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]348<p>LLVM requires that values start with a prefix for two reasons: Compilers
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]349don't need to worry about name clashes with reserved words, and the set of
350reserved words may be expanded in the future without penalty. Additionally,
351unnamed identifiers allow a compiler to quickly come up with a temporary
352variable without having to avoid symbol table conflicts.</p>
353
354<p>Reserved words in LLVM are very similar to reserved words in other
355languages. There are keywords for different opcodes
356('<tt><a href="#i_add">add</a></tt>',
357 '<tt><a href="#i_bitcast">bitcast</a></tt>',
358 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
359href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
360and others. These reserved words cannot conflict with variable names, because
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]361none of them start with a prefix character ('%' or '@').</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]362
363<p>Here is an example of LLVM code to multiply the integer variable
364'<tt>%X</tt>' by 8:</p>
365
366<p>The easy way:</p>
367
368<div class="doc_code">
369<pre>
370%result = <a href="#i_mul">mul</a> i32 %X, 8
371</pre>
372</div>
373
374<p>After strength reduction:</p>
375
376<div class="doc_code">
377<pre>
378%result = <a href="#i_shl">shl</a> i32 %X, i8 3
379</pre>
380</div>
381
382<p>And the hard way:</p>
383
384<div class="doc_code">
385<pre>
386<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
387<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
388%result = <a href="#i_add">add</a> i32 %1, %1
389</pre>
390</div>
391
392<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
393important lexical features of LLVM:</p>
394
395<ol>
396
397 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
398 line.</li>
399
400 <li>Unnamed temporaries are created when the result of a computation is not
401 assigned to a named value.</li>
402
403 <li>Unnamed temporaries are numbered sequentially</li>
404
405</ol>
406
407<p>...and it also shows a convention that we follow in this document. When
408demonstrating instructions, we will follow an instruction with a comment that
409defines the type and name of value produced. Comments are shown in italic
410text.</p>
411
412</div>
413
414<!-- *********************************************************************** -->
415<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
416<!-- *********************************************************************** -->
417
418<!-- ======================================================================= -->
419<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
420</div>
421
422<div class="doc_text">
423
424<p>LLVM programs are composed of "Module"s, each of which is a
425translation unit of the input programs. Each module consists of
426functions, global variables, and symbol table entries. Modules may be
427combined together with the LLVM linker, which merges function (and
428global variable) definitions, resolves forward declarations, and merges
429symbol table entries. Here is an example of the "hello world" module:</p>
430
431<div class="doc_code">
432<pre><i>; Declare the string constant as a global constant...</i>
433<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
434 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
435
436<i>; External declaration of the puts function</i>
437<a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
438
439<i>; Definition of main function</i>
440define i32 @main() { <i>; i32()* </i>
441 <i>; Convert [13x i8 ]* to i8 *...</i>
442 %cast210 = <a
443 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
444
445 <i>; Call puts function to write out the string to stdout...</i>
446 <a
447 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
448 <a
449 href="#i_ret">ret</a> i32 0<br>}<br>
450</pre>
451</div>
452
453<p>This example is made up of a <a href="#globalvars">global variable</a>
454named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
455function, and a <a href="#functionstructure">function definition</a>
456for "<tt>main</tt>".</p>
457
458<p>In general, a module is made up of a list of global values,
459where both functions and global variables are global values. Global values are
460represented by a pointer to a memory location (in this case, a pointer to an
461array of char, and a pointer to a function), and have one of the following <a
462href="#linkage">linkage types</a>.</p>
463
464</div>
465
466<!-- ======================================================================= -->
467<div class="doc_subsection">
468 <a name="linkage">Linkage Types</a>
469</div>
470
471<div class="doc_text">
472
473<p>
474All Global Variables and Functions have one of the following types of linkage:
475</p>
476
477<dl>
478
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]479 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]480
481 <dd>Global values with internal linkage are only directly accessible by
482 objects in the current module. In particular, linking code into a module with
483 an internal global value may cause the internal to be renamed as necessary to
484 avoid collisions. Because the symbol is internal to the module, all
485 references can be updated. This corresponds to the notion of the
486 '<tt>static</tt>' keyword in C.
487 </dd>
488
489 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
490
491 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
492 the same name when linkage occurs. This is typically used to implement
493 inline functions, templates, or other code which must be generated in each
494 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
495 allowed to be discarded.
496 </dd>
497
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]498 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
499
500 <dd>"<tt>common</tt>" linkage is exactly the same as <tt>linkonce</tt>
501 linkage, except that unreferenced <tt>common</tt> globals may not be
502 discarded. This is used for globals that may be emitted in multiple
503 translation units, but that are not guaranteed to be emitted into every
504 translation unit that uses them. One example of this is tentative
505 definitions in C, such as "<tt>int X;</tt>" at global scope.
506 </dd>
507
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]508 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
509
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]510 <dd>"<tt>weak</tt>" linkage is the same as <tt>common</tt> linkage, except
511 that some targets may choose to emit different assembly sequences for them
512 for target-dependent reasons. This is used for globals that are declared
513 "weak" in C source code.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]514 </dd>
515
516 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
517
518 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
519 pointer to array type. When two global variables with appending linkage are
520 linked together, the two global arrays are appended together. This is the
521 LLVM, typesafe, equivalent of having the system linker append together
522 "sections" with identical names when .o files are linked.
523 </dd>
524
525 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]526 <dd>The semantics of this linkage follow the ELF object file model: the
527 symbol is weak until linked, if not linked, the symbol becomes null instead
528 of being an undefined reference.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]529 </dd>
530
531 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
532
533 <dd>If none of the above identifiers are used, the global is externally
534 visible, meaning that it participates in linkage and can be used to resolve
535 external symbol references.
536 </dd>
537</dl>
538
539 <p>
540 The next two types of linkage are targeted for Microsoft Windows platform
541 only. They are designed to support importing (exporting) symbols from (to)
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]542 DLLs (Dynamic Link Libraries).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]543 </p>
544
545 <dl>
546 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
547
548 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
549 or variable via a global pointer to a pointer that is set up by the DLL
550 exporting the symbol. On Microsoft Windows targets, the pointer name is
551 formed by combining <code>_imp__</code> and the function or variable name.
552 </dd>
553
554 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
555
556 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
557 pointer to a pointer in a DLL, so that it can be referenced with the
558 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
559 name is formed by combining <code>_imp__</code> and the function or variable
560 name.
561 </dd>
562
563</dl>
564
565<p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
566variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
567variable and was linked with this one, one of the two would be renamed,
568preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
569external (i.e., lacking any linkage declarations), they are accessible
570outside of the current module.</p>
571<p>It is illegal for a function <i>declaration</i>
572to have any linkage type other than "externally visible", <tt>dllimport</tt>,
573or <tt>extern_weak</tt>.</p>
574<p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
575linkages.
576</div>
577
578<!-- ======================================================================= -->
579<div class="doc_subsection">
580 <a name="callingconv">Calling Conventions</a>
581</div>
582
583<div class="doc_text">
584
585<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
586and <a href="#i_invoke">invokes</a> can all have an optional calling convention
587specified for the call. The calling convention of any pair of dynamic
588caller/callee must match, or the behavior of the program is undefined. The
589following calling conventions are supported by LLVM, and more may be added in
590the future:</p>
591
592<dl>
593 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
594
595 <dd>This calling convention (the default if no other calling convention is
596 specified) matches the target C calling conventions. This calling convention
597 supports varargs function calls and tolerates some mismatch in the declared
598 prototype and implemented declaration of the function (as does normal C).
599 </dd>
600
601 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
602
603 <dd>This calling convention attempts to make calls as fast as possible
604 (e.g. by passing things in registers). This calling convention allows the
605 target to use whatever tricks it wants to produce fast code for the target,
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]606 without having to conform to an externally specified ABI (Application Binary
607 Interface). Implementations of this convention should allow arbitrary
Arnold Schwaighofer07444922008-05-14 09:17:12 +0000[diff] [blame]608 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> to be
609 supported. This calling convention does not support varargs and requires the
610 prototype of all callees to exactly match the prototype of the function
611 definition.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]612 </dd>
613
614 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
615
616 <dd>This calling convention attempts to make code in the caller as efficient
617 as possible under the assumption that the call is not commonly executed. As
618 such, these calls often preserve all registers so that the call does not break
619 any live ranges in the caller side. This calling convention does not support
620 varargs and requires the prototype of all callees to exactly match the
621 prototype of the function definition.
622 </dd>
623
624 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
625
626 <dd>Any calling convention may be specified by number, allowing
627 target-specific calling conventions to be used. Target specific calling
628 conventions start at 64.
629 </dd>
630</dl>
631
632<p>More calling conventions can be added/defined on an as-needed basis, to
633support pascal conventions or any other well-known target-independent
634convention.</p>
635
636</div>
637
638<!-- ======================================================================= -->
639<div class="doc_subsection">
640 <a name="visibility">Visibility Styles</a>
641</div>
642
643<div class="doc_text">
644
645<p>
646All Global Variables and Functions have one of the following visibility styles:
647</p>
648
649<dl>
650 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
651
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]652 <dd>On targets that use the ELF object file format, default visibility means
653 that the declaration is visible to other
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]654 modules and, in shared libraries, means that the declared entity may be
655 overridden. On Darwin, default visibility means that the declaration is
656 visible to other modules. Default visibility corresponds to "external
657 linkage" in the language.
658 </dd>
659
660 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
661
662 <dd>Two declarations of an object with hidden visibility refer to the same
663 object if they are in the same shared object. Usually, hidden visibility
664 indicates that the symbol will not be placed into the dynamic symbol table,
665 so no other module (executable or shared library) can reference it
666 directly.
667 </dd>
668
669 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
670
671 <dd>On ELF, protected visibility indicates that the symbol will be placed in
672 the dynamic symbol table, but that references within the defining module will
673 bind to the local symbol. That is, the symbol cannot be overridden by another
674 module.
675 </dd>
676</dl>
677
678</div>
679
680<!-- ======================================================================= -->
681<div class="doc_subsection">
682 <a name="globalvars">Global Variables</a>
683</div>
684
685<div class="doc_text">
686
687<p>Global variables define regions of memory allocated at compilation time
688instead of run-time. Global variables may optionally be initialized, may have
689an explicit section to be placed in, and may have an optional explicit alignment
690specified. A variable may be defined as "thread_local", which means that it
691will not be shared by threads (each thread will have a separated copy of the
692variable). A variable may be defined as a global "constant," which indicates
693that the contents of the variable will <b>never</b> be modified (enabling better
694optimization, allowing the global data to be placed in the read-only section of
695an executable, etc). Note that variables that need runtime initialization
696cannot be marked "constant" as there is a store to the variable.</p>
697
698<p>
699LLVM explicitly allows <em>declarations</em> of global variables to be marked
700constant, even if the final definition of the global is not. This capability
701can be used to enable slightly better optimization of the program, but requires
702the language definition to guarantee that optimizations based on the
703'constantness' are valid for the translation units that do not include the
704definition.
705</p>
706
707<p>As SSA values, global variables define pointer values that are in
708scope (i.e. they dominate) all basic blocks in the program. Global
709variables always define a pointer to their "content" type because they
710describe a region of memory, and all memory objects in LLVM are
711accessed through pointers.</p>
712
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]713<p>A global variable may be declared to reside in a target-specifc numbered
714address space. For targets that support them, address spaces may affect how
715optimizations are performed and/or what target instructions are used to access
Christopher Lamb20a39e92007-12-12 08:44:39 +0000[diff] [blame]716the variable. The default address space is zero. The address space qualifier
717must precede any other attributes.</p>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]718
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]719<p>LLVM allows an explicit section to be specified for globals. If the target
720supports it, it will emit globals to the section specified.</p>
721
722<p>An explicit alignment may be specified for a global. If not present, or if
723the alignment is set to zero, the alignment of the global is set by the target
724to whatever it feels convenient. If an explicit alignment is specified, the
725global is forced to have at least that much alignment. All alignments must be
726a power of 2.</p>
727
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]728<p>For example, the following defines a global in a numbered address space with
729an initializer, section, and alignment:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]730
731<div class="doc_code">
732<pre>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]733@G = constant float 1.0 addrspace(5), section "foo", align 4
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]734</pre>
735</div>
736
737</div>
738
739
740<!-- ======================================================================= -->
741<div class="doc_subsection">
742 <a name="functionstructure">Functions</a>
743</div>
744
745<div class="doc_text">
746
747<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
748an optional <a href="#linkage">linkage type</a>, an optional
749<a href="#visibility">visibility style</a>, an optional
750<a href="#callingconv">calling convention</a>, a return type, an optional
751<a href="#paramattrs">parameter attribute</a> for the return type, a function
752name, a (possibly empty) argument list (each with optional
753<a href="#paramattrs">parameter attributes</a>), an optional section, an
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]754optional alignment, an optional <a href="#gc">garbage collector name</a>,
Chris Lattneree202542008-10-04 18:10:21 +0000[diff] [blame]755an opening curly brace, a list of basic blocks, and a closing curly brace.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]756
757LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
758optional <a href="#linkage">linkage type</a>, an optional
759<a href="#visibility">visibility style</a>, an optional
760<a href="#callingconv">calling convention</a>, a return type, an optional
761<a href="#paramattrs">parameter attribute</a> for the return type, a function
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]762name, a possibly empty list of arguments, an optional alignment, and an optional
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]763<a href="#gc">garbage collector name</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]764
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]765<p>A function definition contains a list of basic blocks, forming the CFG
766(Control Flow Graph) for
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]767the function. Each basic block may optionally start with a label (giving the
768basic block a symbol table entry), contains a list of instructions, and ends
769with a <a href="#terminators">terminator</a> instruction (such as a branch or
770function return).</p>
771
772<p>The first basic block in a function is special in two ways: it is immediately
773executed on entrance to the function, and it is not allowed to have predecessor
774basic blocks (i.e. there can not be any branches to the entry block of a
775function). Because the block can have no predecessors, it also cannot have any
776<a href="#i_phi">PHI nodes</a>.</p>
777
778<p>LLVM allows an explicit section to be specified for functions. If the target
779supports it, it will emit functions to the section specified.</p>
780
781<p>An explicit alignment may be specified for a function. If not present, or if
782the alignment is set to zero, the alignment of the function is set by the target
783to whatever it feels convenient. If an explicit alignment is specified, the
784function is forced to have at least that much alignment. All alignments must be
785a power of 2.</p>
786
787</div>
788
789
790<!-- ======================================================================= -->
791<div class="doc_subsection">
792 <a name="aliasstructure">Aliases</a>
793</div>
794<div class="doc_text">
795 <p>Aliases act as "second name" for the aliasee value (which can be either
Anton Korobeynikov96822812008-03-22 08:36:14 +0000[diff] [blame]796 function, global variable, another alias or bitcast of global value). Aliases
797 may have an optional <a href="#linkage">linkage type</a>, and an
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]798 optional <a href="#visibility">visibility style</a>.</p>
799
800 <h5>Syntax:</h5>
801
802<div class="doc_code">
803<pre>
Duncan Sandsd7bfabf2008-09-12 20:48:21 +0000[diff] [blame]804@&lt;Name&gt; = alias [Linkage] [Visibility] &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]805</pre>
806</div>
807
808</div>
809
810
811
812<!-- ======================================================================= -->
813<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
814<div class="doc_text">
815 <p>The return type and each parameter of a function type may have a set of
816 <i>parameter attributes</i> associated with them. Parameter attributes are
817 used to communicate additional information about the result or parameters of
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]818 a function. Parameter attributes are considered to be part of the function,
819 not of the function type, so functions with different parameter attributes
820 can have the same function type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]821
822 <p>Parameter attributes are simple keywords that follow the type specified. If
823 multiple parameter attributes are needed, they are space separated. For
824 example:</p>
825
826<div class="doc_code">
827<pre>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]828declare i32 @printf(i8* noalias , ...)
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame^]829declare i32 @atoi(i8 zeroext)
830declare signext i8 @returns_signed_char()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]831</pre>
832</div>
833
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]834 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
835 <tt>readonly</tt>) come immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]836
837 <p>Currently, only the following parameter attributes are defined:</p>
838 <dl>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]839 <dt><tt>zeroext</tt></dt>
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame^]840 <dd>This indicates to the code generator that the parameter or return value
841 should be zero-extended to a 32-bit value by the caller (for a parameter)
842 or the callee (for a return value).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]843
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]844 <dt><tt>signext</tt></dt>
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame^]845 <dd>This indicates to the code generator that the parameter or return value
846 should be sign-extended to a 32-bit value by the caller (for a parameter)
847 or the callee (for a return value).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]848
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]849 <dt><tt>inreg</tt></dt>
Dale Johannesenc08a0e22008-09-25 20:47:45 +0000[diff] [blame]850 <dd>This indicates that this parameter or return value should be treated
851 in a special target-dependent fashion during while emitting code for a
852 function call or return (usually, by putting it in a register as opposed
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame^]853 to memory, though some targets use it to distinguish between two different
854 kinds of registers). Use of this attribute is target-specific.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]855
856 <dt><tt>byval</tt></dt>
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]857 <dd>This indicates that the pointer parameter should really be passed by
858 value to the function. The attribute implies that a hidden copy of the
859 pointee is made between the caller and the callee, so the callee is unable
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]860 to modify the value in the callee. This attribute is only valid on LLVM
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]861 pointer arguments. It is generally used to pass structs and arrays by
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame^]862 value, but is also valid on pointers to scalars.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]863
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]864 <dt><tt>sret</tt></dt>
Duncan Sands616cc032008-02-18 04:19:38 +0000[diff] [blame]865 <dd>This indicates that the pointer parameter specifies the address of a
866 structure that is the return value of the function in the source program.
Chris Lattnerf33b8452008-10-04 18:33:34 +0000[diff] [blame^]867 This pointer must be guaranteed by the caller to be valid: loads and stores
868 to the structure may be assumed by the callee to not to trap. This may only
869 be applied to the first parameter.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]870
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]871 <dt><tt>noalias</tt></dt>
Owen Andersonc4fc4cd2008-02-18 04:09:01 +0000[diff] [blame]872 <dd>This indicates that the parameter does not alias any global or any other
873 parameter. The caller is responsible for ensuring that this is the case,
874 usually by placing the value in a stack allocation.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]875
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]876 <dt><tt>nest</tt></dt>
Duncan Sandsf1a7d4c2008-07-08 09:27:25 +0000[diff] [blame]877 <dd>This indicates that the pointer parameter can be excised using the
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]878 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]879 </dl>
880
881</div>
882
883<!-- ======================================================================= -->
884<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]885 <a name="gc">Garbage Collector Names</a>
886</div>
887
888<div class="doc_text">
889<p>Each function may specify a garbage collector name, which is simply a
890string.</p>
891
892<div class="doc_code"><pre
893>define void @f() gc "name" { ...</pre></div>
894
895<p>The compiler declares the supported values of <i>name</i>. Specifying a
896collector which will cause the compiler to alter its output in order to support
897the named garbage collection algorithm.</p>
898</div>
899
900<!-- ======================================================================= -->
901<div class="doc_subsection">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]902 <a name="fnattrs">Function Attributes</a>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]903</div>
904
905<div class="doc_text">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]906
907<p>Function attributes are set to communicate additional information about
908 a function. Function attributes are considered to be part of the function,
909 not of the function type, so functions with different parameter attributes
910 can have the same function type.</p>
911
912 <p>Function attributes are simple keywords that follow the type specified. If
913 multiple attributes are needed, they are space separated. For
914 example:</p>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]915
916<div class="doc_code">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]917<pre>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]918define void @f() noinline { ... }
919define void @f() alwaysinline { ... }
920define void @f() alwaysinline optsize { ... }
921define void @f() optsize
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]922</pre>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]923</div>
924
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]925<dl>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]926<dt><tt>alwaysinline</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]927<dd>This attribute indicates that the inliner should attempt to inline this
928function into callers whenever possible, ignoring any active inlining size
929threshold for this caller.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]930
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]931<dt><tt>noinline</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]932<dd>This attribute indicates that the inliner should never inline this function
933in any situation. This attribute may not be used together with
934<tt>alwaysinline</tt> attribute.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]935
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]936<dt><tt>optsize</tt></dt>
Devang Patelc29099b2008-09-29 18:34:44 +0000[diff] [blame]937<dd>This attribute suggests that optimization passes and code generator passes
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]938make choices that keep the code size of this function low, and otherwise do
939optimizations specifically to reduce code size.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]940
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]941<dt><tt>noreturn</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]942<dd>This function attribute indicates that the function never returns normally.
943This produces undefined behavior at runtime if the function ever does
944dynamically return.</dd>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]945
946<dt><tt>nounwind</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]947<dd>This function attribute indicates that the function never returns with an
948unwind or exceptional control flow. If the function does unwind, its runtime
949behavior is undefined.</dd>
950
951<dt><tt>readnone</tt></dt>
952<dd>This attribute indicates that the function computes its result (or its
953thrown exception) based strictly on its arguments. It does not read any global
954mutable state (e.g. memory, control registers, etc) visible to caller functions.
955Furthermore, <tt>readnone</tt> functions never change any state visible to their
956caller.
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame]957
958<dt><tt>readonly</tt></dt>
Chris Lattner82d70f92008-10-04 18:23:17 +0000[diff] [blame]959<dd>This function attribute indicates that the function has no side-effects on
960the calling function, but that it depends on state (memory state, control
961register state, etc) that may be set in the caller. A readonly function always
962returns the same value (or throws the same exception) whenever it is called with
963a particular set of arguments and global state.</dd>
964
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]965</dl>
966
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]967</div>
968
969<!-- ======================================================================= -->
970<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]971 <a name="moduleasm">Module-Level Inline Assembly</a>
972</div>
973
974<div class="doc_text">
975<p>
976Modules may contain "module-level inline asm" blocks, which corresponds to the
977GCC "file scope inline asm" blocks. These blocks are internally concatenated by
978LLVM and treated as a single unit, but may be separated in the .ll file if
979desired. The syntax is very simple:
980</p>
981
982<div class="doc_code">
983<pre>
984module asm "inline asm code goes here"
985module asm "more can go here"
986</pre>
987</div>
988
989<p>The strings can contain any character by escaping non-printable characters.
990 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
991 for the number.
992</p>
993
994<p>
995 The inline asm code is simply printed to the machine code .s file when
996 assembly code is generated.
997</p>
998</div>
999
1000<!-- ======================================================================= -->
1001<div class="doc_subsection">
1002 <a name="datalayout">Data Layout</a>
1003</div>
1004
1005<div class="doc_text">
1006<p>A module may specify a target specific data layout string that specifies how
1007data is to be laid out in memory. The syntax for the data layout is simply:</p>
1008<pre> target datalayout = "<i>layout specification</i>"</pre>
1009<p>The <i>layout specification</i> consists of a list of specifications
1010separated by the minus sign character ('-'). Each specification starts with a
1011letter and may include other information after the letter to define some
1012aspect of the data layout. The specifications accepted are as follows: </p>
1013<dl>
1014 <dt><tt>E</tt></dt>
1015 <dd>Specifies that the target lays out data in big-endian form. That is, the
1016 bits with the most significance have the lowest address location.</dd>
1017 <dt><tt>e</tt></dt>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1018 <dd>Specifies that the target lays out data in little-endian form. That is,
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1019 the bits with the least significance have the lowest address location.</dd>
1020 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1021 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1022 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1023 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1024 too.</dd>
1025 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1026 <dd>This specifies the alignment for an integer type of a given bit
1027 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1028 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1029 <dd>This specifies the alignment for a vector type of a given bit
1030 <i>size</i>.</dd>
1031 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1032 <dd>This specifies the alignment for a floating point type of a given bit
1033 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1034 (double).</dd>
1035 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1036 <dd>This specifies the alignment for an aggregate type of a given bit
1037 <i>size</i>.</dd>
1038</dl>
1039<p>When constructing the data layout for a given target, LLVM starts with a
1040default set of specifications which are then (possibly) overriden by the
1041specifications in the <tt>datalayout</tt> keyword. The default specifications
1042are given in this list:</p>
1043<ul>
1044 <li><tt>E</tt> - big endian</li>
1045 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1046 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1047 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1048 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1049 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1050 <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]1051 alignment of 64-bits</li>
1052 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1053 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1054 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1055 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1056 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1057</ul>
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]1058<p>When LLVM is determining the alignment for a given type, it uses the
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1059following rules:
1060<ol>
1061 <li>If the type sought is an exact match for one of the specifications, that
1062 specification is used.</li>
1063 <li>If no match is found, and the type sought is an integer type, then the
1064 smallest integer type that is larger than the bitwidth of the sought type is
1065 used. If none of the specifications are larger than the bitwidth then the the
1066 largest integer type is used. For example, given the default specifications
1067 above, the i7 type will use the alignment of i8 (next largest) while both
1068 i65 and i256 will use the alignment of i64 (largest specified).</li>
1069 <li>If no match is found, and the type sought is a vector type, then the
1070 largest vector type that is smaller than the sought vector type will be used
1071 as a fall back. This happens because <128 x double> can be implemented in
1072 terms of 64 <2 x double>, for example.</li>
1073</ol>
1074</div>
1075
1076<!-- *********************************************************************** -->
1077<div class="doc_section"> <a name="typesystem">Type System</a> </div>
1078<!-- *********************************************************************** -->
1079
1080<div class="doc_text">
1081
1082<p>The LLVM type system is one of the most important features of the
1083intermediate representation. Being typed enables a number of
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1084optimizations to be performed on the intermediate representation directly,
1085without having to do
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1086extra analyses on the side before the transformation. A strong type
1087system makes it easier to read the generated code and enables novel
1088analyses and transformations that are not feasible to perform on normal
1089three address code representations.</p>
1090
1091</div>
1092
1093<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1094<div class="doc_subsection"> <a name="t_classifications">Type
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1095Classifications</a> </div>
1096<div class="doc_text">
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1097<p>The types fall into a few useful
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1098classifications:</p>
1099
1100<table border="1" cellspacing="0" cellpadding="4">
1101 <tbody>
1102 <tr><th>Classification</th><th>Types</th></tr>
1103 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1104 <td><a href="#t_integer">integer</a></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1105 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1106 </tr>
1107 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1108 <td><a href="#t_floating">floating point</a></td>
1109 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1110 </tr>
1111 <tr>
1112 <td><a name="t_firstclass">first class</a></td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1113 <td><a href="#t_integer">integer</a>,
1114 <a href="#t_floating">floating point</a>,
1115 <a href="#t_pointer">pointer</a>,
Dan Gohmanf6237db2008-06-18 18:42:13 +0000[diff] [blame]1116 <a href="#t_vector">vector</a>,
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]1117 <a href="#t_struct">structure</a>,
1118 <a href="#t_array">array</a>,
Dan Gohmand13951e2008-05-23 22:50:26 +0000[diff] [blame]1119 <a href="#t_label">label</a>.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1120 </td>
1121 </tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1122 <tr>
1123 <td><a href="#t_primitive">primitive</a></td>
1124 <td><a href="#t_label">label</a>,
1125 <a href="#t_void">void</a>,
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1126 <a href="#t_floating">floating point</a>.</td>
1127 </tr>
1128 <tr>
1129 <td><a href="#t_derived">derived</a></td>
1130 <td><a href="#t_integer">integer</a>,
1131 <a href="#t_array">array</a>,
1132 <a href="#t_function">function</a>,
1133 <a href="#t_pointer">pointer</a>,
1134 <a href="#t_struct">structure</a>,
1135 <a href="#t_pstruct">packed structure</a>,
1136 <a href="#t_vector">vector</a>,
1137 <a href="#t_opaque">opaque</a>.
1138 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1139 </tbody>
1140</table>
1141
1142<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1143most important. Values of these types are the only ones which can be
1144produced by instructions, passed as arguments, or used as operands to
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]1145instructions.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1146</div>
1147
1148<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1149<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1150
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1151<div class="doc_text">
1152<p>The primitive types are the fundamental building blocks of the LLVM
1153system.</p>
1154
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1155</div>
1156
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1157<!-- _______________________________________________________________________ -->
1158<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1159
1160<div class="doc_text">
1161 <table>
1162 <tbody>
1163 <tr><th>Type</th><th>Description</th></tr>
1164 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1165 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1166 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1167 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1168 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1169 </tbody>
1170 </table>
1171</div>
1172
1173<!-- _______________________________________________________________________ -->
1174<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1175
1176<div class="doc_text">
1177<h5>Overview:</h5>
1178<p>The void type does not represent any value and has no size.</p>
1179
1180<h5>Syntax:</h5>
1181
1182<pre>
1183 void
1184</pre>
1185</div>
1186
1187<!-- _______________________________________________________________________ -->
1188<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1189
1190<div class="doc_text">
1191<h5>Overview:</h5>
1192<p>The label type represents code labels.</p>
1193
1194<h5>Syntax:</h5>
1195
1196<pre>
1197 label
1198</pre>
1199</div>
1200
1201
1202<!-- ======================================================================= -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1203<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1204
1205<div class="doc_text">
1206
1207<p>The real power in LLVM comes from the derived types in the system.
1208This is what allows a programmer to represent arrays, functions,
1209pointers, and other useful types. Note that these derived types may be
1210recursive: For example, it is possible to have a two dimensional array.</p>
1211
1212</div>
1213
1214<!-- _______________________________________________________________________ -->
1215<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1216
1217<div class="doc_text">
1218
1219<h5>Overview:</h5>
1220<p>The integer type is a very simple derived type that simply specifies an
1221arbitrary bit width for the integer type desired. Any bit width from 1 bit to
12222^23-1 (about 8 million) can be specified.</p>
1223
1224<h5>Syntax:</h5>
1225
1226<pre>
1227 iN
1228</pre>
1229
1230<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1231value.</p>
1232
1233<h5>Examples:</h5>
1234<table class="layout">
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1235 <tbody>
1236 <tr>
1237 <td><tt>i1</tt></td>
1238 <td>a single-bit integer.</td>
1239 </tr><tr>
1240 <td><tt>i32</tt></td>
1241 <td>a 32-bit integer.</td>
1242 </tr><tr>
1243 <td><tt>i1942652</tt></td>
1244 <td>a really big integer of over 1 million bits.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1245 </tr>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1246 </tbody>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1247</table>
1248</div>
1249
1250<!-- _______________________________________________________________________ -->
1251<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1252
1253<div class="doc_text">
1254
1255<h5>Overview:</h5>
1256
1257<p>The array type is a very simple derived type that arranges elements
1258sequentially in memory. The array type requires a size (number of
1259elements) and an underlying data type.</p>
1260
1261<h5>Syntax:</h5>
1262
1263<pre>
1264 [&lt;# elements&gt; x &lt;elementtype&gt;]
1265</pre>
1266
1267<p>The number of elements is a constant integer value; elementtype may
1268be any type with a size.</p>
1269
1270<h5>Examples:</h5>
1271<table class="layout">
1272 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1273 <td class="left"><tt>[40 x i32]</tt></td>
1274 <td class="left">Array of 40 32-bit integer values.</td>
1275 </tr>
1276 <tr class="layout">
1277 <td class="left"><tt>[41 x i32]</tt></td>
1278 <td class="left">Array of 41 32-bit integer values.</td>
1279 </tr>
1280 <tr class="layout">
1281 <td class="left"><tt>[4 x i8]</tt></td>
1282 <td class="left">Array of 4 8-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1283 </tr>
1284</table>
1285<p>Here are some examples of multidimensional arrays:</p>
1286<table class="layout">
1287 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1288 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1289 <td class="left">3x4 array of 32-bit integer values.</td>
1290 </tr>
1291 <tr class="layout">
1292 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1293 <td class="left">12x10 array of single precision floating point values.</td>
1294 </tr>
1295 <tr class="layout">
1296 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1297 <td class="left">2x3x4 array of 16-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1298 </tr>
1299</table>
1300
1301<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1302length array. Normally, accesses past the end of an array are undefined in
1303LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1304As a special case, however, zero length arrays are recognized to be variable
1305length. This allows implementation of 'pascal style arrays' with the LLVM
1306type "{ i32, [0 x float]}", for example.</p>
1307
1308</div>
1309
1310<!-- _______________________________________________________________________ -->
1311<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1312<div class="doc_text">
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1313
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1314<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1315
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1316<p>The function type can be thought of as a function signature. It
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1317consists of a return type and a list of formal parameter types. The
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1318return 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]1319If 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]1320class types, and the struct must have at least one element.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1321
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1322<h5>Syntax:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1323
1324<pre>
1325 &lt;returntype list&gt; (&lt;parameter list&gt;)
1326</pre>
1327
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1328<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1329specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1330which indicates that the function takes a variable number of arguments.
1331Variable argument functions can access their arguments with the <a
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1332 href="#int_varargs">variable argument handling intrinsic</a> functions.
1333'<tt>&lt;returntype list&gt;</tt>' is a comma-separated list of
1334<a href="#t_firstclass">first class</a> type specifiers.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1335
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1336<h5>Examples:</h5>
1337<table class="layout">
1338 <tr class="layout">
1339 <td class="left"><tt>i32 (i32)</tt></td>
1340 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1341 </td>
1342 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1343 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1344 </tt></td>
1345 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1346 an <tt>i16</tt> that should be sign extended and a
1347 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1348 <tt>float</tt>.
1349 </td>
1350 </tr><tr class="layout">
1351 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1352 <td class="left">A vararg function that takes at least one
1353 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1354 which returns an integer. This is the signature for <tt>printf</tt> in
1355 LLVM.
1356 </td>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1357 </tr><tr class="layout">
1358 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Devang Patel6dddba22008-03-24 18:10:52 +0000[diff] [blame]1359 <td class="left">A function taking an <tt>i32></tt>, returning two
1360 <tt> i32 </tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1361 </td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1362 </tr>
1363</table>
1364
1365</div>
1366<!-- _______________________________________________________________________ -->
1367<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1368<div class="doc_text">
1369<h5>Overview:</h5>
1370<p>The structure type is used to represent a collection of data members
1371together in memory. The packing of the field types is defined to match
1372the ABI of the underlying processor. The elements of a structure may
1373be any type that has a size.</p>
1374<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1375and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1376field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1377instruction.</p>
1378<h5>Syntax:</h5>
1379<pre> { &lt;type list&gt; }<br></pre>
1380<h5>Examples:</h5>
1381<table class="layout">
1382 <tr class="layout">
1383 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1384 <td class="left">A triple of three <tt>i32</tt> values</td>
1385 </tr><tr class="layout">
1386 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1387 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1388 second element is a <a href="#t_pointer">pointer</a> to a
1389 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1390 an <tt>i32</tt>.</td>
1391 </tr>
1392</table>
1393</div>
1394
1395<!-- _______________________________________________________________________ -->
1396<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1397</div>
1398<div class="doc_text">
1399<h5>Overview:</h5>
1400<p>The packed structure type is used to represent a collection of data members
1401together in memory. There is no padding between fields. Further, the alignment
1402of a packed structure is 1 byte. The elements of a packed structure may
1403be any type that has a size.</p>
1404<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1405and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1406field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1407instruction.</p>
1408<h5>Syntax:</h5>
1409<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1410<h5>Examples:</h5>
1411<table class="layout">
1412 <tr class="layout">
1413 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1414 <td class="left">A triple of three <tt>i32</tt> values</td>
1415 </tr><tr class="layout">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1416 <td class="left">
1417<tt>&lt;&nbsp;{&nbsp;float,&nbsp;i32&nbsp;(i32)*&nbsp;}&nbsp;&gt;</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1418 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1419 second element is a <a href="#t_pointer">pointer</a> to a
1420 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1421 an <tt>i32</tt>.</td>
1422 </tr>
1423</table>
1424</div>
1425
1426<!-- _______________________________________________________________________ -->
1427<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1428<div class="doc_text">
1429<h5>Overview:</h5>
1430<p>As in many languages, the pointer type represents a pointer or
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1431reference to another object, which must live in memory. Pointer types may have
1432an optional address space attribute defining the target-specific numbered
1433address space where the pointed-to object resides. The default address space is
1434zero.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1435<h5>Syntax:</h5>
1436<pre> &lt;type&gt; *<br></pre>
1437<h5>Examples:</h5>
1438<table class="layout">
1439 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1440 <td class="left"><tt>[4x i32]*</tt></td>
1441 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1442 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1443 </tr>
1444 <tr class="layout">
1445 <td class="left"><tt>i32 (i32 *) *</tt></td>
1446 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1447 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1448 <tt>i32</tt>.</td>
1449 </tr>
1450 <tr class="layout">
1451 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1452 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1453 that resides in address space #5.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1454 </tr>
1455</table>
1456</div>
1457
1458<!-- _______________________________________________________________________ -->
1459<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1460<div class="doc_text">
1461
1462<h5>Overview:</h5>
1463
1464<p>A vector type is a simple derived type that represents a vector
1465of elements. Vector types are used when multiple primitive data
1466are operated in parallel using a single instruction (SIMD).
1467A vector type requires a size (number of
1468elements) and an underlying primitive data type. Vectors must have a power
1469of two length (1, 2, 4, 8, 16 ...). Vector types are
1470considered <a href="#t_firstclass">first class</a>.</p>
1471
1472<h5>Syntax:</h5>
1473
1474<pre>
1475 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1476</pre>
1477
1478<p>The number of elements is a constant integer value; elementtype may
1479be any integer or floating point type.</p>
1480
1481<h5>Examples:</h5>
1482
1483<table class="layout">
1484 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1485 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1486 <td class="left">Vector of 4 32-bit integer values.</td>
1487 </tr>
1488 <tr class="layout">
1489 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1490 <td class="left">Vector of 8 32-bit floating-point values.</td>
1491 </tr>
1492 <tr class="layout">
1493 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1494 <td class="left">Vector of 2 64-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1495 </tr>
1496</table>
1497</div>
1498
1499<!-- _______________________________________________________________________ -->
1500<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1501<div class="doc_text">
1502
1503<h5>Overview:</h5>
1504
1505<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1506corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1507In LLVM, opaque types can eventually be resolved to any type (not just a
1508structure type).</p>
1509
1510<h5>Syntax:</h5>
1511
1512<pre>
1513 opaque
1514</pre>
1515
1516<h5>Examples:</h5>
1517
1518<table class="layout">
1519 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1520 <td class="left"><tt>opaque</tt></td>
1521 <td class="left">An opaque type.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1522 </tr>
1523</table>
1524</div>
1525
1526
1527<!-- *********************************************************************** -->
1528<div class="doc_section"> <a name="constants">Constants</a> </div>
1529<!-- *********************************************************************** -->
1530
1531<div class="doc_text">
1532
1533<p>LLVM has several different basic types of constants. This section describes
1534them all and their syntax.</p>
1535
1536</div>
1537
1538<!-- ======================================================================= -->
1539<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1540
1541<div class="doc_text">
1542
1543<dl>
1544 <dt><b>Boolean constants</b></dt>
1545
1546 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1547 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1548 </dd>
1549
1550 <dt><b>Integer constants</b></dt>
1551
1552 <dd>Standard integers (such as '4') are constants of the <a
1553 href="#t_integer">integer</a> type. Negative numbers may be used with
1554 integer types.
1555 </dd>
1556
1557 <dt><b>Floating point constants</b></dt>
1558
1559 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1560 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]1561 notation (see below). The assembler requires the exact decimal value of
1562 a floating-point constant. For example, the assembler accepts 1.25 but
1563 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1564 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1565
1566 <dt><b>Null pointer constants</b></dt>
1567
1568 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1569 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1570
1571</dl>
1572
1573<p>The one non-intuitive notation for constants is the optional hexadecimal form
1574of floating point constants. For example, the form '<tt>double
15750x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
15764.5e+15</tt>'. The only time hexadecimal floating point constants are required
1577(and the only time that they are generated by the disassembler) is when a
1578floating point constant must be emitted but it cannot be represented as a
1579decimal floating point number. For example, NaN's, infinities, and other
1580special values are represented in their IEEE hexadecimal format so that
1581assembly and disassembly do not cause any bits to change in the constants.</p>
1582
1583</div>
1584
1585<!-- ======================================================================= -->
1586<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1587</div>
1588
1589<div class="doc_text">
1590<p>Aggregate constants arise from aggregation of simple constants
1591and smaller aggregate constants.</p>
1592
1593<dl>
1594 <dt><b>Structure constants</b></dt>
1595
1596 <dd>Structure constants are represented with notation similar to structure
1597 type definitions (a comma separated list of elements, surrounded by braces
Chris Lattner6c8de962007-12-25 20:34:52 +0000[diff] [blame]1598 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1599 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]1600 must have <a href="#t_struct">structure type</a>, and the number and
1601 types of elements must match those specified by the type.
1602 </dd>
1603
1604 <dt><b>Array constants</b></dt>
1605
1606 <dd>Array constants are represented with notation similar to array type
1607 definitions (a comma separated list of elements, surrounded by square brackets
1608 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1609 constants must have <a href="#t_array">array type</a>, and the number and
1610 types of elements must match those specified by the type.
1611 </dd>
1612
1613 <dt><b>Vector constants</b></dt>
1614
1615 <dd>Vector constants are represented with notation similar to vector type
1616 definitions (a comma separated list of elements, surrounded by
1617 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1618 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1619 href="#t_vector">vector type</a>, and the number and types of elements must
1620 match those specified by the type.
1621 </dd>
1622
1623 <dt><b>Zero initialization</b></dt>
1624
1625 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1626 value to zero of <em>any</em> type, including scalar and aggregate types.
1627 This is often used to avoid having to print large zero initializers (e.g. for
1628 large arrays) and is always exactly equivalent to using explicit zero
1629 initializers.
1630 </dd>
1631</dl>
1632
1633</div>
1634
1635<!-- ======================================================================= -->
1636<div class="doc_subsection">
1637 <a name="globalconstants">Global Variable and Function Addresses</a>
1638</div>
1639
1640<div class="doc_text">
1641
1642<p>The addresses of <a href="#globalvars">global variables</a> and <a
1643href="#functionstructure">functions</a> are always implicitly valid (link-time)
1644constants. These constants are explicitly referenced when the <a
1645href="#identifiers">identifier for the global</a> is used and always have <a
1646href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1647file:</p>
1648
1649<div class="doc_code">
1650<pre>
1651@X = global i32 17
1652@Y = global i32 42
1653@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1654</pre>
1655</div>
1656
1657</div>
1658
1659<!-- ======================================================================= -->
1660<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1661<div class="doc_text">
1662 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1663 no specific value. Undefined values may be of any type and be used anywhere
1664 a constant is permitted.</p>
1665
1666 <p>Undefined values indicate to the compiler that the program is well defined
1667 no matter what value is used, giving the compiler more freedom to optimize.
1668 </p>
1669</div>
1670
1671<!-- ======================================================================= -->
1672<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1673</div>
1674
1675<div class="doc_text">
1676
1677<p>Constant expressions are used to allow expressions involving other constants
1678to be used as constants. Constant expressions may be of any <a
1679href="#t_firstclass">first class</a> type and may involve any LLVM operation
1680that does not have side effects (e.g. load and call are not supported). The
1681following is the syntax for constant expressions:</p>
1682
1683<dl>
1684 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1685 <dd>Truncate a constant to another type. The bit size of CST must be larger
1686 than the bit size of TYPE. Both types must be integers.</dd>
1687
1688 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1689 <dd>Zero extend a constant to another type. The bit size of CST must be
1690 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1691
1692 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1693 <dd>Sign extend a constant to another type. The bit size of CST must be
1694 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1695
1696 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1697 <dd>Truncate a floating point constant to another floating point type. The
1698 size of CST must be larger than the size of TYPE. Both types must be
1699 floating point.</dd>
1700
1701 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1702 <dd>Floating point extend a constant to another type. The size of CST must be
1703 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1704
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1705 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1706 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1707 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1708 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1709 of the same number of elements. If the value won't fit in the integer type,
1710 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1711
1712 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1713 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1714 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1715 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1716 of the same number of elements. If the value won't fit in the integer type,
1717 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1718
1719 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1720 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1721 constant. TYPE must be a scalar or vector floating point type. CST must be of
1722 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1723 of the same number of elements. If the value won't fit in the floating point
1724 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1725
1726 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1727 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1728 constant. TYPE must be a scalar or vector floating point type. CST must be of
1729 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1730 of the same number of elements. If the value won't fit in the floating point
1731 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1732
1733 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1734 <dd>Convert a pointer typed constant to the corresponding integer constant
1735 TYPE must be an integer type. CST must be of pointer type. The CST value is
1736 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1737
1738 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1739 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1740 pointer type. CST must be of integer type. The CST value is zero extended,
1741 truncated, or unchanged to make it fit in a pointer size. This one is
1742 <i>really</i> dangerous!</dd>
1743
1744 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1745 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1746 identical (same number of bits). The conversion is done as if the CST value
1747 was stored to memory and read back as TYPE. In other words, no bits change
1748 with this operator, just the type. This can be used for conversion of
1749 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]1750 pointers it is only valid to cast to another pointer type. It is not valid
1751 to bitcast to or from an aggregate type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1752 </dd>
1753
1754 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1755
1756 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1757 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1758 instruction, the index list may have zero or more indexes, which are required
1759 to make sense for the type of "CSTPTR".</dd>
1760
1761 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1762
1763 <dd>Perform the <a href="#i_select">select operation</a> on
1764 constants.</dd>
1765
1766 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1767 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1768
1769 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1770 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1771
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]1772 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
1773 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
1774
1775 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
1776 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
1777
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1778 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1779
1780 <dd>Perform the <a href="#i_extractelement">extractelement
1781 operation</a> on constants.
1782
1783 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1784
1785 <dd>Perform the <a href="#i_insertelement">insertelement
1786 operation</a> on constants.</dd>
1787
1788
1789 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1790
1791 <dd>Perform the <a href="#i_shufflevector">shufflevector
1792 operation</a> on constants.</dd>
1793
1794 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1795
1796 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1797 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1798 binary</a> operations. The constraints on operands are the same as those for
1799 the corresponding instruction (e.g. no bitwise operations on floating point
1800 values are allowed).</dd>
1801</dl>
1802</div>
1803
1804<!-- *********************************************************************** -->
1805<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1806<!-- *********************************************************************** -->
1807
1808<!-- ======================================================================= -->
1809<div class="doc_subsection">
1810<a name="inlineasm">Inline Assembler Expressions</a>
1811</div>
1812
1813<div class="doc_text">
1814
1815<p>
1816LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1817Module-Level Inline Assembly</a>) through the use of a special value. This
1818value represents the inline assembler as a string (containing the instructions
1819to emit), a list of operand constraints (stored as a string), and a flag that
1820indicates whether or not the inline asm expression has side effects. An example
1821inline assembler expression is:
1822</p>
1823
1824<div class="doc_code">
1825<pre>
1826i32 (i32) asm "bswap $0", "=r,r"
1827</pre>
1828</div>
1829
1830<p>
1831Inline assembler expressions may <b>only</b> be used as the callee operand of
1832a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1833</p>
1834
1835<div class="doc_code">
1836<pre>
1837%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1838</pre>
1839</div>
1840
1841<p>
1842Inline asms with side effects not visible in the constraint list must be marked
1843as having side effects. This is done through the use of the
1844'<tt>sideeffect</tt>' keyword, like so:
1845</p>
1846
1847<div class="doc_code">
1848<pre>
1849call void asm sideeffect "eieio", ""()
1850</pre>
1851</div>
1852
1853<p>TODO: The format of the asm and constraints string still need to be
1854documented here. Constraints on what can be done (e.g. duplication, moving, etc
1855need to be documented).
1856</p>
1857
1858</div>
1859
1860<!-- *********************************************************************** -->
1861<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1862<!-- *********************************************************************** -->
1863
1864<div class="doc_text">
1865
1866<p>The LLVM instruction set consists of several different
1867classifications of instructions: <a href="#terminators">terminator
1868instructions</a>, <a href="#binaryops">binary instructions</a>,
1869<a href="#bitwiseops">bitwise binary instructions</a>, <a
1870 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1871instructions</a>.</p>
1872
1873</div>
1874
1875<!-- ======================================================================= -->
1876<div class="doc_subsection"> <a name="terminators">Terminator
1877Instructions</a> </div>
1878
1879<div class="doc_text">
1880
1881<p>As mentioned <a href="#functionstructure">previously</a>, every
1882basic block in a program ends with a "Terminator" instruction, which
1883indicates which block should be executed after the current block is
1884finished. These terminator instructions typically yield a '<tt>void</tt>'
1885value: they produce control flow, not values (the one exception being
1886the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1887<p>There are six different terminator instructions: the '<a
1888 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1889instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1890the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1891 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1892 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1893
1894</div>
1895
1896<!-- _______________________________________________________________________ -->
1897<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1898Instruction</a> </div>
1899<div class="doc_text">
1900<h5>Syntax:</h5>
1901<pre> ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
1902 ret void <i>; Return from void function</i>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1903 ret &lt;type&gt; &lt;value&gt;, &lt;type&gt; &lt;value&gt; <i>; Return two values from a non-void function </i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1904</pre>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1905
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1906<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1907
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1908<p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1909value) from a function back to the caller.</p>
1910<p>There are two forms of the '<tt>ret</tt>' instruction: one that
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1911returns value(s) and then causes control flow, and one that just causes
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1912control flow to occur.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1913
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1914<h5>Arguments:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1915
1916<p>The '<tt>ret</tt>' instruction may return zero, one or multiple values.
1917The type of each return value must be a '<a href="#t_firstclass">first
1918class</a>' type. Note that a function is not <a href="#wellformed">well
1919formed</a> if there exists a '<tt>ret</tt>' instruction inside of the
1920function that returns values that do not match the return type of the
1921function.</p>
1922
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1923<h5>Semantics:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1924
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1925<p>When the '<tt>ret</tt>' instruction is executed, control flow
1926returns back to the calling function's context. If the caller is a "<a
1927 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1928the instruction after the call. If the caller was an "<a
1929 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1930at the beginning of the "normal" destination block. If the instruction
1931returns a value, that value shall set the call or invoke instruction's
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1932return value. If the instruction returns multiple values then these
Devang Patelec8a5b02008-03-11 05:51:59 +0000[diff] [blame]1933values can only be accessed through a '<a href="#i_getresult"><tt>getresult</tt>
1934</a>' instruction.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1935
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1936<h5>Example:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1937
1938<pre>
1939 ret i32 5 <i>; Return an integer value of 5</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1940 ret void <i>; Return from a void function</i>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1941 ret i32 4, i8 2 <i>; Return two values 4 and 2 </i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1942</pre>
1943</div>
1944<!-- _______________________________________________________________________ -->
1945<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1946<div class="doc_text">
1947<h5>Syntax:</h5>
1948<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
1949</pre>
1950<h5>Overview:</h5>
1951<p>The '<tt>br</tt>' instruction is used to cause control flow to
1952transfer to a different basic block in the current function. There are
1953two forms of this instruction, corresponding to a conditional branch
1954and an unconditional branch.</p>
1955<h5>Arguments:</h5>
1956<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1957single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1958unconditional form of the '<tt>br</tt>' instruction takes a single
1959'<tt>label</tt>' value as a target.</p>
1960<h5>Semantics:</h5>
1961<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1962argument is evaluated. If the value is <tt>true</tt>, control flows
1963to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1964control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1965<h5>Example:</h5>
1966<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
1967 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1968</div>
1969<!-- _______________________________________________________________________ -->
1970<div class="doc_subsubsection">
1971 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1972</div>
1973
1974<div class="doc_text">
1975<h5>Syntax:</h5>
1976
1977<pre>
1978 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
1979</pre>
1980
1981<h5>Overview:</h5>
1982
1983<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1984several different places. It is a generalization of the '<tt>br</tt>'
1985instruction, allowing a branch to occur to one of many possible
1986destinations.</p>
1987
1988
1989<h5>Arguments:</h5>
1990
1991<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1992comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1993an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1994table is not allowed to contain duplicate constant entries.</p>
1995
1996<h5>Semantics:</h5>
1997
1998<p>The <tt>switch</tt> instruction specifies a table of values and
1999destinations. When the '<tt>switch</tt>' instruction is executed, this
2000table is searched for the given value. If the value is found, control flow is
2001transfered to the corresponding destination; otherwise, control flow is
2002transfered to the default destination.</p>
2003
2004<h5>Implementation:</h5>
2005
2006<p>Depending on properties of the target machine and the particular
2007<tt>switch</tt> instruction, this instruction may be code generated in different
2008ways. For example, it could be generated as a series of chained conditional
2009branches or with a lookup table.</p>
2010
2011<h5>Example:</h5>
2012
2013<pre>
2014 <i>; Emulate a conditional br instruction</i>
2015 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2016 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
2017
2018 <i>; Emulate an unconditional br instruction</i>
2019 switch i32 0, label %dest [ ]
2020
2021 <i>; Implement a jump table:</i>
2022 switch i32 %val, label %otherwise [ i32 0, label %onzero
2023 i32 1, label %onone
2024 i32 2, label %ontwo ]
2025</pre>
2026</div>
2027
2028<!-- _______________________________________________________________________ -->
2029<div class="doc_subsubsection">
2030 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2031</div>
2032
2033<div class="doc_text">
2034
2035<h5>Syntax:</h5>
2036
2037<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2038 &lt;result&gt; = invoke [<a href="#callingconv">cconv</a>] &lt;ptr to function ty&gt; &lt;function ptr val&gt;(&lt;function args&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2039 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
2040</pre>
2041
2042<h5>Overview:</h5>
2043
2044<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2045function, with the possibility of control flow transfer to either the
2046'<tt>normal</tt>' label or the
2047'<tt>exception</tt>' label. If the callee function returns with the
2048"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2049"normal" label. If the callee (or any indirect callees) returns with the "<a
2050href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]2051continued at the dynamically nearest "exception" label. If the callee function
Devang Patelec8a5b02008-03-11 05:51:59 +0000[diff] [blame]2052returns multiple values then individual return values are only accessible through
2053a '<tt><a href="#i_getresult">getresult</a></tt>' instruction.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2054
2055<h5>Arguments:</h5>
2056
2057<p>This instruction requires several arguments:</p>
2058
2059<ol>
2060 <li>
2061 The optional "cconv" marker indicates which <a href="#callingconv">calling
2062 convention</a> the call should use. If none is specified, the call defaults
2063 to using C calling conventions.
2064 </li>
2065 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2066 function value being invoked. In most cases, this is a direct function
2067 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2068 an arbitrary pointer to function value.
2069 </li>
2070
2071 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2072 function to be invoked. </li>
2073
2074 <li>'<tt>function args</tt>': argument list whose types match the function
2075 signature argument types. If the function signature indicates the function
2076 accepts a variable number of arguments, the extra arguments can be
2077 specified. </li>
2078
2079 <li>'<tt>normal label</tt>': the label reached when the called function
2080 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2081
2082 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2083 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2084
2085</ol>
2086
2087<h5>Semantics:</h5>
2088
2089<p>This instruction is designed to operate as a standard '<tt><a
2090href="#i_call">call</a></tt>' instruction in most regards. The primary
2091difference is that it establishes an association with a label, which is used by
2092the runtime library to unwind the stack.</p>
2093
2094<p>This instruction is used in languages with destructors to ensure that proper
2095cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2096exception. Additionally, this is important for implementation of
2097'<tt>catch</tt>' clauses in high-level languages that support them.</p>
2098
2099<h5>Example:</h5>
2100<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2101 %retval = invoke i32 @Test(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2102 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2103 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2104 unwind label %TestCleanup <i>; {i32}:retval set</i>
2105</pre>
2106</div>
2107
2108
2109<!-- _______________________________________________________________________ -->
2110
2111<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2112Instruction</a> </div>
2113
2114<div class="doc_text">
2115
2116<h5>Syntax:</h5>
2117<pre>
2118 unwind
2119</pre>
2120
2121<h5>Overview:</h5>
2122
2123<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2124at the first callee in the dynamic call stack which used an <a
2125href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2126primarily used to implement exception handling.</p>
2127
2128<h5>Semantics:</h5>
2129
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2130<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2131immediately halt. The dynamic call stack is then searched for the first <a
2132href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2133execution continues at the "exceptional" destination block specified by the
2134<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2135dynamic call chain, undefined behavior results.</p>
2136</div>
2137
2138<!-- _______________________________________________________________________ -->
2139
2140<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2141Instruction</a> </div>
2142
2143<div class="doc_text">
2144
2145<h5>Syntax:</h5>
2146<pre>
2147 unreachable
2148</pre>
2149
2150<h5>Overview:</h5>
2151
2152<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2153instruction is used to inform the optimizer that a particular portion of the
2154code is not reachable. This can be used to indicate that the code after a
2155no-return function cannot be reached, and other facts.</p>
2156
2157<h5>Semantics:</h5>
2158
2159<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2160</div>
2161
2162
2163
2164<!-- ======================================================================= -->
2165<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2166<div class="doc_text">
2167<p>Binary operators are used to do most of the computation in a
Chris Lattnerab596d92008-04-01 18:47:32 +0000[diff] [blame]2168program. They require two operands of the same type, execute an operation on them, and
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2169produce a single value. The operands might represent
2170multiple 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]2171The result value has the same type as its operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2172<p>There are several different binary operators:</p>
2173</div>
2174<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2175<div class="doc_subsubsection">
2176 <a name="i_add">'<tt>add</tt>' Instruction</a>
2177</div>
2178
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2179<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2180
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2181<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2182
2183<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2184 &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]2185</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2186
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2187<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2188
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2189<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2190
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2191<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2192
2193<p>The two arguments to the '<tt>add</tt>' instruction must be <a
2194 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>, or
2195 <a href="#t_vector">vector</a> values. Both arguments must have identical
2196 types.</p>
2197
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2198<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2199
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2200<p>The value produced is the integer or floating point sum of the two
2201operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2202
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2203<p>If an integer sum has unsigned overflow, the result returned is the
2204mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2205the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2206
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2207<p>Because LLVM integers use a two's complement representation, this
2208instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2209
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2210<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2211
2212<pre>
2213 &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2214</pre>
2215</div>
2216<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2217<div class="doc_subsubsection">
2218 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2219</div>
2220
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2221<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2222
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2223<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2224
2225<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2226 &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]2227</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2228
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2229<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2230
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2231<p>The '<tt>sub</tt>' instruction returns the difference of its two
2232operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2233
2234<p>Note that the '<tt>sub</tt>' instruction is used to represent the
2235'<tt>neg</tt>' instruction present in most other intermediate
2236representations.</p>
2237
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2238<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2239
2240<p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2241 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2242 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2243 types.</p>
2244
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2245<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2246
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2247<p>The value produced is the integer or floating point difference of
2248the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2249
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2250<p>If an integer difference has unsigned overflow, the result returned is the
2251mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2252the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2253
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2254<p>Because LLVM integers use a two's complement representation, this
2255instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2256
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2257<h5>Example:</h5>
2258<pre>
2259 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2260 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2261</pre>
2262</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2263
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2264<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2265<div class="doc_subsubsection">
2266 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2267</div>
2268
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2269<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2270
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2271<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2272<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]2273</pre>
2274<h5>Overview:</h5>
2275<p>The '<tt>mul</tt>' instruction returns the product of its two
2276operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2277
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2278<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2279
2280<p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2281href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2282or <a href="#t_vector">vector</a> values. Both arguments must have identical
2283types.</p>
2284
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2285<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2286
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2287<p>The value produced is the integer or floating point product of the
2288two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2289
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2290<p>If the result of an integer multiplication has unsigned overflow,
2291the result returned is the mathematical result modulo
22922<sup>n</sup>, where n is the bit width of the result.</p>
2293<p>Because LLVM integers use a two's complement representation, and the
2294result is the same width as the operands, this instruction returns the
2295correct result for both signed and unsigned integers. If a full product
2296(e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2297should be sign-extended or zero-extended as appropriate to the
2298width of the full product.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2299<h5>Example:</h5>
2300<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2301</pre>
2302</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2303
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2304<!-- _______________________________________________________________________ -->
2305<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2306</a></div>
2307<div class="doc_text">
2308<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2309<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]2310</pre>
2311<h5>Overview:</h5>
2312<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2313operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2314
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2315<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2316
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2317<p>The two arguments to the '<tt>udiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2318<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2319values. Both arguments must have identical types.</p>
2320
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2321<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2322
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2323<p>The value produced is the unsigned integer quotient of the two operands.</p>
2324<p>Note that unsigned integer division and signed integer division are distinct
2325operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2326<p>Division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2327<h5>Example:</h5>
2328<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2329</pre>
2330</div>
2331<!-- _______________________________________________________________________ -->
2332<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2333</a> </div>
2334<div class="doc_text">
2335<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2336<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2337 &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]2338</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2339
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2340<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2341
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2342<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2343operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2344
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2345<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2346
2347<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2348<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2349values. Both arguments must have identical types.</p>
2350
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2351<h5>Semantics:</h5>
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2352<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]2353<p>Note that signed integer division and unsigned integer division are distinct
2354operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2355<p>Division by zero leads to undefined behavior. Overflow also leads to
2356undefined behavior; this is a rare case, but can occur, for example,
2357by doing a 32-bit division of -2147483648 by -1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2358<h5>Example:</h5>
2359<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2360</pre>
2361</div>
2362<!-- _______________________________________________________________________ -->
2363<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2364Instruction</a> </div>
2365<div class="doc_text">
2366<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2367<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2368 &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]2369</pre>
2370<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2371
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2372<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2373operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2374
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2375<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2376
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2377<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2378<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2379of floating point values. Both arguments must have identical types.</p>
2380
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2381<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2382
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2383<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2384
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2385<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2386
2387<pre>
2388 &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]2389</pre>
2390</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2391
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2392<!-- _______________________________________________________________________ -->
2393<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2394</div>
2395<div class="doc_text">
2396<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2397<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]2398</pre>
2399<h5>Overview:</h5>
2400<p>The '<tt>urem</tt>' instruction returns the remainder from the
2401unsigned division of its two arguments.</p>
2402<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2403<p>The two arguments to the '<tt>urem</tt>' instruction must be
2404<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2405values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2406<h5>Semantics:</h5>
2407<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2408This instruction always performs an unsigned division to get the remainder.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2409<p>Note that unsigned integer remainder and signed integer remainder are
2410distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2411<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2412<h5>Example:</h5>
2413<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2414</pre>
2415
2416</div>
2417<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2418<div class="doc_subsubsection">
2419 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2420</div>
2421
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2422<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2423
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2424<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2425
2426<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2427 &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]2428</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2429
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2430<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2431
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2432<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2433signed division of its two operands. This instruction can also take
2434<a href="#t_vector">vector</a> versions of the values in which case
2435the elements must be integers.</p>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]2436
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2437<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2438
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2439<p>The two arguments to the '<tt>srem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2440<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2441values. Both arguments must have identical types.</p>
2442
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2443<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2444
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2445<p>This instruction returns the <i>remainder</i> of a division (where the result
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2446has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2447operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2448a value. For more information about the difference, see <a
2449 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2450Math Forum</a>. For a table of how this is implemented in various languages,
2451please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2452Wikipedia: modulo operation</a>.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2453<p>Note that signed integer remainder and unsigned integer remainder are
2454distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2455<p>Taking the remainder of a division by zero leads to undefined behavior.
2456Overflow also leads to undefined behavior; this is a rare case, but can occur,
2457for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2458(The remainder doesn't actually overflow, but this rule lets srem be
2459implemented using instructions that return both the result of the division
2460and the remainder.)</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2461<h5>Example:</h5>
2462<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2463</pre>
2464
2465</div>
2466<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2467<div class="doc_subsubsection">
2468 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2469
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2470<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2471
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2472<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2473<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]2474</pre>
2475<h5>Overview:</h5>
2476<p>The '<tt>frem</tt>' instruction returns the remainder from the
2477division of its two operands.</p>
2478<h5>Arguments:</h5>
2479<p>The two arguments to the '<tt>frem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2480<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2481of floating point values. Both arguments must have identical types.</p>
2482
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2483<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2484
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2485<p>This instruction returns the <i>remainder</i> of a division.
2486The remainder has the same sign as the dividend.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2487
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2488<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2489
2490<pre>
2491 &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]2492</pre>
2493</div>
2494
2495<!-- ======================================================================= -->
2496<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2497Operations</a> </div>
2498<div class="doc_text">
2499<p>Bitwise binary operators are used to do various forms of
2500bit-twiddling in a program. They are generally very efficient
2501instructions and can commonly be strength reduced from other
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2502instructions. They require two operands of the same type, execute an operation on them,
2503and 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]2504</div>
2505
2506<!-- _______________________________________________________________________ -->
2507<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2508Instruction</a> </div>
2509<div class="doc_text">
2510<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2511<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]2512</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2513
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2514<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2515
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2516<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2517the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2518
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2519<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2520
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2521<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2522 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2523type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2524
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2525<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2526
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2527<p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2528where n is the width of the result. If <tt>op2</tt> is (statically or dynamically) negative or
2529equal to or larger than the number of bits in <tt>op1</tt>, the result is undefined.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2530
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2531<h5>Example:</h5><pre>
2532 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2533 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2534 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2535 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2536</pre>
2537</div>
2538<!-- _______________________________________________________________________ -->
2539<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2540Instruction</a> </div>
2541<div class="doc_text">
2542<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2543<pre> &lt;result&gt; = 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]2544</pre>
2545
2546<h5>Overview:</h5>
2547<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2548operand shifted to the right a specified number of bits with zero fill.</p>
2549
2550<h5>Arguments:</h5>
2551<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2552<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]2553type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2554
2555<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2556
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2557<p>This instruction always performs a logical shift right operation. The most
2558significant bits of the result will be filled with zero bits after the
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2559shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
2560the number of bits in <tt>op1</tt>, the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2561
2562<h5>Example:</h5>
2563<pre>
2564 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2565 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2566 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2567 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2568 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2569</pre>
2570</div>
2571
2572<!-- _______________________________________________________________________ -->
2573<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2574Instruction</a> </div>
2575<div class="doc_text">
2576
2577<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2578<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]2579</pre>
2580
2581<h5>Overview:</h5>
2582<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2583operand shifted to the right a specified number of bits with sign extension.</p>
2584
2585<h5>Arguments:</h5>
2586<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2587<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]2588type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2589
2590<h5>Semantics:</h5>
2591<p>This instruction always performs an arithmetic shift right operation,
2592The most significant bits of the result will be filled with the sign bit
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2593of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
2594larger than the number of bits in <tt>op1</tt>, the result is undefined.
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2595</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2596
2597<h5>Example:</h5>
2598<pre>
2599 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2600 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2601 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2602 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2603 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2604</pre>
2605</div>
2606
2607<!-- _______________________________________________________________________ -->
2608<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2609Instruction</a> </div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2610
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2611<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2612
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2613<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2614
2615<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2616 &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]2617</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2618
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2619<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2620
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2621<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2622its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2623
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2624<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2625
2626<p>The two arguments to the '<tt>and</tt>' instruction must be
2627<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2628values. Both arguments must have identical types.</p>
2629
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2630<h5>Semantics:</h5>
2631<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2632<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2633<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2634<table border="1" cellspacing="0" cellpadding="4">
2635 <tbody>
2636 <tr>
2637 <td>In0</td>
2638 <td>In1</td>
2639 <td>Out</td>
2640 </tr>
2641 <tr>
2642 <td>0</td>
2643 <td>0</td>
2644 <td>0</td>
2645 </tr>
2646 <tr>
2647 <td>0</td>
2648 <td>1</td>
2649 <td>0</td>
2650 </tr>
2651 <tr>
2652 <td>1</td>
2653 <td>0</td>
2654 <td>0</td>
2655 </tr>
2656 <tr>
2657 <td>1</td>
2658 <td>1</td>
2659 <td>1</td>
2660 </tr>
2661 </tbody>
2662</table>
2663</div>
2664<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2665<pre>
2666 &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]2667 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2668 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2669</pre>
2670</div>
2671<!-- _______________________________________________________________________ -->
2672<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2673<div class="doc_text">
2674<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2675<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]2676</pre>
2677<h5>Overview:</h5>
2678<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2679or of its two operands.</p>
2680<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2681
2682<p>The two arguments to the '<tt>or</tt>' instruction must be
2683<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2684values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2685<h5>Semantics:</h5>
2686<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2687<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2688<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2689<table border="1" cellspacing="0" cellpadding="4">
2690 <tbody>
2691 <tr>
2692 <td>In0</td>
2693 <td>In1</td>
2694 <td>Out</td>
2695 </tr>
2696 <tr>
2697 <td>0</td>
2698 <td>0</td>
2699 <td>0</td>
2700 </tr>
2701 <tr>
2702 <td>0</td>
2703 <td>1</td>
2704 <td>1</td>
2705 </tr>
2706 <tr>
2707 <td>1</td>
2708 <td>0</td>
2709 <td>1</td>
2710 </tr>
2711 <tr>
2712 <td>1</td>
2713 <td>1</td>
2714 <td>1</td>
2715 </tr>
2716 </tbody>
2717</table>
2718</div>
2719<h5>Example:</h5>
2720<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2721 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2722 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2723</pre>
2724</div>
2725<!-- _______________________________________________________________________ -->
2726<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2727Instruction</a> </div>
2728<div class="doc_text">
2729<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2730<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]2731</pre>
2732<h5>Overview:</h5>
2733<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2734or of its two operands. The <tt>xor</tt> is used to implement the
2735"one's complement" operation, which is the "~" operator in C.</p>
2736<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2737<p>The two arguments to the '<tt>xor</tt>' instruction must be
2738<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2739values. Both arguments must have identical types.</p>
2740
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2741<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2742
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2743<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2744<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2745<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2746<table border="1" cellspacing="0" cellpadding="4">
2747 <tbody>
2748 <tr>
2749 <td>In0</td>
2750 <td>In1</td>
2751 <td>Out</td>
2752 </tr>
2753 <tr>
2754 <td>0</td>
2755 <td>0</td>
2756 <td>0</td>
2757 </tr>
2758 <tr>
2759 <td>0</td>
2760 <td>1</td>
2761 <td>1</td>
2762 </tr>
2763 <tr>
2764 <td>1</td>
2765 <td>0</td>
2766 <td>1</td>
2767 </tr>
2768 <tr>
2769 <td>1</td>
2770 <td>1</td>
2771 <td>0</td>
2772 </tr>
2773 </tbody>
2774</table>
2775</div>
2776<p> </p>
2777<h5>Example:</h5>
2778<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2779 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2780 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2781 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2782</pre>
2783</div>
2784
2785<!-- ======================================================================= -->
2786<div class="doc_subsection">
2787 <a name="vectorops">Vector Operations</a>
2788</div>
2789
2790<div class="doc_text">
2791
2792<p>LLVM supports several instructions to represent vector operations in a
2793target-independent manner. These instructions cover the element-access and
2794vector-specific operations needed to process vectors effectively. While LLVM
2795does directly support these vector operations, many sophisticated algorithms
2796will want to use target-specific intrinsics to take full advantage of a specific
2797target.</p>
2798
2799</div>
2800
2801<!-- _______________________________________________________________________ -->
2802<div class="doc_subsubsection">
2803 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2804</div>
2805
2806<div class="doc_text">
2807
2808<h5>Syntax:</h5>
2809
2810<pre>
2811 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
2812</pre>
2813
2814<h5>Overview:</h5>
2815
2816<p>
2817The '<tt>extractelement</tt>' instruction extracts a single scalar
2818element from a vector at a specified index.
2819</p>
2820
2821
2822<h5>Arguments:</h5>
2823
2824<p>
2825The first operand of an '<tt>extractelement</tt>' instruction is a
2826value of <a href="#t_vector">vector</a> type. The second operand is
2827an index indicating the position from which to extract the element.
2828The index may be a variable.</p>
2829
2830<h5>Semantics:</h5>
2831
2832<p>
2833The result is a scalar of the same type as the element type of
2834<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2835<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2836results are undefined.
2837</p>
2838
2839<h5>Example:</h5>
2840
2841<pre>
2842 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
2843</pre>
2844</div>
2845
2846
2847<!-- _______________________________________________________________________ -->
2848<div class="doc_subsubsection">
2849 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2850</div>
2851
2852<div class="doc_text">
2853
2854<h5>Syntax:</h5>
2855
2856<pre>
Dan Gohmanbcc3c502008-05-12 23:38:42 +0000[diff] [blame]2857 &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]2858</pre>
2859
2860<h5>Overview:</h5>
2861
2862<p>
2863The '<tt>insertelement</tt>' instruction inserts a scalar
2864element into a vector at a specified index.
2865</p>
2866
2867
2868<h5>Arguments:</h5>
2869
2870<p>
2871The first operand of an '<tt>insertelement</tt>' instruction is a
2872value of <a href="#t_vector">vector</a> type. The second operand is a
2873scalar value whose type must equal the element type of the first
2874operand. The third operand is an index indicating the position at
2875which to insert the value. The index may be a variable.</p>
2876
2877<h5>Semantics:</h5>
2878
2879<p>
2880The result is a vector of the same type as <tt>val</tt>. Its
2881element values are those of <tt>val</tt> except at position
2882<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2883exceeds the length of <tt>val</tt>, the results are undefined.
2884</p>
2885
2886<h5>Example:</h5>
2887
2888<pre>
2889 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
2890</pre>
2891</div>
2892
2893<!-- _______________________________________________________________________ -->
2894<div class="doc_subsubsection">
2895 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2896</div>
2897
2898<div class="doc_text">
2899
2900<h5>Syntax:</h5>
2901
2902<pre>
2903 &lt;result&gt; = shufflevector &lt;n x &lt;ty&gt;&gt; &lt;v1&gt;, &lt;n x &lt;ty&gt;&gt; &lt;v2&gt;, &lt;n x i32&gt; &lt;mask&gt; <i>; yields &lt;n x &lt;ty&gt;&gt;</i>
2904</pre>
2905
2906<h5>Overview:</h5>
2907
2908<p>
2909The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2910from two input vectors, returning a vector of the same type.
2911</p>
2912
2913<h5>Arguments:</h5>
2914
2915<p>
2916The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2917with types that match each other and types that match the result of the
2918instruction. The third argument is a shuffle mask, which has the same number
2919of elements as the other vector type, but whose element type is always 'i32'.
2920</p>
2921
2922<p>
2923The shuffle mask operand is required to be a constant vector with either
2924constant integer or undef values.
2925</p>
2926
2927<h5>Semantics:</h5>
2928
2929<p>
2930The elements of the two input vectors are numbered from left to right across
2931both of the vectors. The shuffle mask operand specifies, for each element of
2932the result vector, which element of the two input registers the result element
2933gets. The element selector may be undef (meaning "don't care") and the second
2934operand may be undef if performing a shuffle from only one vector.
2935</p>
2936
2937<h5>Example:</h5>
2938
2939<pre>
2940 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
2941 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
2942 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
2943 &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.
2944</pre>
2945</div>
2946
2947
2948<!-- ======================================================================= -->
2949<div class="doc_subsection">
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2950 <a name="aggregateops">Aggregate Operations</a>
2951</div>
2952
2953<div class="doc_text">
2954
2955<p>LLVM supports several instructions for working with aggregate values.
2956</p>
2957
2958</div>
2959
2960<!-- _______________________________________________________________________ -->
2961<div class="doc_subsubsection">
2962 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
2963</div>
2964
2965<div class="doc_text">
2966
2967<h5>Syntax:</h5>
2968
2969<pre>
2970 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
2971</pre>
2972
2973<h5>Overview:</h5>
2974
2975<p>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2976The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
2977or array element from an aggregate value.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2978</p>
2979
2980
2981<h5>Arguments:</h5>
2982
2983<p>
2984The first operand of an '<tt>extractvalue</tt>' instruction is a
2985value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2986type. The operands are constant indices to specify which value to extract
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]2987in a similar manner as indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2988'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
2989</p>
2990
2991<h5>Semantics:</h5>
2992
2993<p>
2994The result is the value at the position in the aggregate specified by
2995the index operands.
2996</p>
2997
2998<h5>Example:</h5>
2999
3000<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3001 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3002</pre>
3003</div>
3004
3005
3006<!-- _______________________________________________________________________ -->
3007<div class="doc_subsubsection">
3008 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3009</div>
3010
3011<div class="doc_text">
3012
3013<h5>Syntax:</h5>
3014
3015<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3016 &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]3017</pre>
3018
3019<h5>Overview:</h5>
3020
3021<p>
3022The '<tt>insertvalue</tt>' instruction inserts a value
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3023into a struct field or array element in an aggregate.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3024</p>
3025
3026
3027<h5>Arguments:</h5>
3028
3029<p>
3030The first operand of an '<tt>insertvalue</tt>' instruction is a
3031value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3032The second operand is a first-class value to insert.
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]3033The following operands are constant indices
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3034indicating the position at which to insert the value in a similar manner as
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3035indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3036'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3037The value to insert must have the same type as the value identified
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3038by the indices.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3039
3040<h5>Semantics:</h5>
3041
3042<p>
3043The result is an aggregate of the same type as <tt>val</tt>. Its
3044value is that of <tt>val</tt> except that the value at the position
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3045specified by the indices is that of <tt>elt</tt>.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3046</p>
3047
3048<h5>Example:</h5>
3049
3050<pre>
Dan Gohmanb1aab4e2008-06-23 15:26:37 +0000[diff] [blame]3051 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3052</pre>
3053</div>
3054
3055
3056<!-- ======================================================================= -->
3057<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3058 <a name="memoryops">Memory Access and Addressing Operations</a>
3059</div>
3060
3061<div class="doc_text">
3062
3063<p>A key design point of an SSA-based representation is how it
3064represents memory. In LLVM, no memory locations are in SSA form, which
3065makes things very simple. This section describes how to read, write,
3066allocate, and free memory in LLVM.</p>
3067
3068</div>
3069
3070<!-- _______________________________________________________________________ -->
3071<div class="doc_subsubsection">
3072 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3073</div>
3074
3075<div class="doc_text">
3076
3077<h5>Syntax:</h5>
3078
3079<pre>
3080 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3081</pre>
3082
3083<h5>Overview:</h5>
3084
3085<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3086heap and returns a pointer to it. The object is always allocated in the generic
3087address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3088
3089<h5>Arguments:</h5>
3090
3091<p>The '<tt>malloc</tt>' instruction allocates
3092<tt>sizeof(&lt;type&gt;)*NumElements</tt>
3093bytes of memory from the operating system and returns a pointer of the
3094appropriate type to the program. If "NumElements" is specified, it is the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3095number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3096If a constant alignment is specified, the value result of the allocation is guaranteed to
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3097be aligned to at least that boundary. If not specified, or if zero, the target can
3098choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3099
3100<p>'<tt>type</tt>' must be a sized type.</p>
3101
3102<h5>Semantics:</h5>
3103
3104<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3105a pointer is returned. The result of a zero byte allocattion is undefined. The
3106result is null if there is insufficient memory available.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3107
3108<h5>Example:</h5>
3109
3110<pre>
3111 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
3112
3113 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3114 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3115 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3116 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3117 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3118</pre>
3119</div>
3120
3121<!-- _______________________________________________________________________ -->
3122<div class="doc_subsubsection">
3123 <a name="i_free">'<tt>free</tt>' Instruction</a>
3124</div>
3125
3126<div class="doc_text">
3127
3128<h5>Syntax:</h5>
3129
3130<pre>
3131 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
3132</pre>
3133
3134<h5>Overview:</h5>
3135
3136<p>The '<tt>free</tt>' instruction returns memory back to the unused
3137memory heap to be reallocated in the future.</p>
3138
3139<h5>Arguments:</h5>
3140
3141<p>'<tt>value</tt>' shall be a pointer value that points to a value
3142that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3143instruction.</p>
3144
3145<h5>Semantics:</h5>
3146
3147<p>Access to the memory pointed to by the pointer is no longer defined
Chris Lattner329d6532008-04-19 22:41:32 +0000[diff] [blame]3148after this instruction executes. If the pointer is null, the operation
3149is a noop.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3150
3151<h5>Example:</h5>
3152
3153<pre>
3154 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
3155 free [4 x i8]* %array
3156</pre>
3157</div>
3158
3159<!-- _______________________________________________________________________ -->
3160<div class="doc_subsubsection">
3161 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3162</div>
3163
3164<div class="doc_text">
3165
3166<h5>Syntax:</h5>
3167
3168<pre>
3169 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3170</pre>
3171
3172<h5>Overview:</h5>
3173
3174<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3175currently executing function, to be automatically released when this function
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3176returns to its caller. The object is always allocated in the generic address
3177space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3178
3179<h5>Arguments:</h5>
3180
3181<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
3182bytes of memory on the runtime stack, returning a pointer of the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3183appropriate type to the program. If "NumElements" is specified, it is the
3184number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3185If a constant alignment is specified, the value result of the allocation is guaranteed
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3186to be aligned to at least that boundary. If not specified, or if zero, the target
3187can choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3188
3189<p>'<tt>type</tt>' may be any sized type.</p>
3190
3191<h5>Semantics:</h5>
3192
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3193<p>Memory is allocated; a pointer is returned. The operation is undefiend if
3194there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3195memory is automatically released when the function returns. The '<tt>alloca</tt>'
3196instruction is commonly used to represent automatic variables that must
3197have an address available. When the function returns (either with the <tt><a
3198 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]3199instructions), the memory is reclaimed. Allocating zero bytes
3200is legal, but the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3201
3202<h5>Example:</h5>
3203
3204<pre>
3205 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3206 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3207 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3208 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3209</pre>
3210</div>
3211
3212<!-- _______________________________________________________________________ -->
3213<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3214Instruction</a> </div>
3215<div class="doc_text">
3216<h5>Syntax:</h5>
3217<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>
3218<h5>Overview:</h5>
3219<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3220<h5>Arguments:</h5>
3221<p>The argument to the '<tt>load</tt>' instruction specifies the memory
3222address from which to load. The pointer must point to a <a
3223 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3224marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3225the number or order of execution of this <tt>load</tt> with other
3226volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3227instructions. </p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3228<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3229The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3230(that is, the alignment of the memory address). A value of 0 or an
3231omitted "align" argument means that the operation has the preferential
3232alignment for the target. It is the responsibility of the code emitter
3233to ensure that the alignment information is correct. Overestimating
3234the alignment results in an undefined behavior. Underestimating the
3235alignment may produce less efficient code. An alignment of 1 is always
3236safe.
3237</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3238<h5>Semantics:</h5>
3239<p>The location of memory pointed to is loaded.</p>
3240<h5>Examples:</h5>
3241<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3242 <a
3243 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3244 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3245</pre>
3246</div>
3247<!-- _______________________________________________________________________ -->
3248<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3249Instruction</a> </div>
3250<div class="doc_text">
3251<h5>Syntax:</h5>
3252<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3253 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3254</pre>
3255<h5>Overview:</h5>
3256<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3257<h5>Arguments:</h5>
3258<p>There are two arguments to the '<tt>store</tt>' instruction: a value
3259to 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]3260operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3261of the '<tt>&lt;value&gt;</tt>'
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3262operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3263optimizer is not allowed to modify the number or order of execution of
3264this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3265 href="#i_store">store</a></tt> instructions.</p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3266<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3267The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3268(that is, the alignment of the memory address). A value of 0 or an
3269omitted "align" argument means that the operation has the preferential
3270alignment for the target. It is the responsibility of the code emitter
3271to ensure that the alignment information is correct. Overestimating
3272the alignment results in an undefined behavior. Underestimating the
3273alignment may produce less efficient code. An alignment of 1 is always
3274safe.
3275</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3276<h5>Semantics:</h5>
3277<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
3278at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
3279<h5>Example:</h5>
3280<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]3281 store i32 3, i32* %ptr <i>; yields {void}</i>
3282 %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]3283</pre>
3284</div>
3285
3286<!-- _______________________________________________________________________ -->
3287<div class="doc_subsubsection">
3288 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3289</div>
3290
3291<div class="doc_text">
3292<h5>Syntax:</h5>
3293<pre>
3294 &lt;result&gt; = getelementptr &lt;ty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
3295</pre>
3296
3297<h5>Overview:</h5>
3298
3299<p>
3300The '<tt>getelementptr</tt>' instruction is used to get the address of a
3301subelement of an aggregate data structure.</p>
3302
3303<h5>Arguments:</h5>
3304
3305<p>This instruction takes a list of integer operands that indicate what
3306elements of the aggregate object to index to. The actual types of the arguments
3307provided depend on the type of the first pointer argument. The
3308'<tt>getelementptr</tt>' instruction is used to index down through the type
3309levels of a structure or to a specific index in an array. When indexing into a
3310structure, only <tt>i32</tt> integer constants are allowed. When indexing
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]3311into an array or pointer, only integers of 32 or 64 bits are allowed; 32-bit
3312values will be sign extended to 64-bits if required.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3313
3314<p>For example, let's consider a C code fragment and how it gets
3315compiled to LLVM:</p>
3316
3317<div class="doc_code">
3318<pre>
3319struct RT {
3320 char A;
3321 int B[10][20];
3322 char C;
3323};
3324struct ST {
3325 int X;
3326 double Y;
3327 struct RT Z;
3328};
3329
3330int *foo(struct ST *s) {
3331 return &amp;s[1].Z.B[5][13];
3332}
3333</pre>
3334</div>
3335
3336<p>The LLVM code generated by the GCC frontend is:</p>
3337
3338<div class="doc_code">
3339<pre>
3340%RT = type { i8 , [10 x [20 x i32]], i8 }
3341%ST = type { i32, double, %RT }
3342
3343define i32* %foo(%ST* %s) {
3344entry:
3345 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3346 ret i32* %reg
3347}
3348</pre>
3349</div>
3350
3351<h5>Semantics:</h5>
3352
3353<p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
3354on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
3355and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
3356<a href="#t_integer">integer</a> type but the value will always be sign extended
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3357to 64-bits. <a href="#t_struct">Structure</a> and <a href="#t_pstruct">packed
3358structure</a> types require <tt>i32</tt> <b>constants</b>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3359
3360<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3361type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3362}</tt>' type, a structure. The second index indexes into the third element of
3363the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3364i8 }</tt>' type, another structure. The third index indexes into the second
3365element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3366array. The two dimensions of the array are subscripted into, yielding an
3367'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3368to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3369
3370<p>Note that it is perfectly legal to index partially through a
3371structure, returning a pointer to an inner element. Because of this,
3372the LLVM code for the given testcase is equivalent to:</p>
3373
3374<pre>
3375 define i32* %foo(%ST* %s) {
3376 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3377 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3378 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3379 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3380 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3381 ret i32* %t5
3382 }
3383</pre>
3384
3385<p>Note that it is undefined to access an array out of bounds: array and
3386pointer indexes must always be within the defined bounds of the array type.
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]3387The one exception for this rule is zero length arrays. These arrays are
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3388defined to be accessible as variable length arrays, which requires access
3389beyond the zero'th element.</p>
3390
3391<p>The getelementptr instruction is often confusing. For some more insight
3392into how it works, see <a href="GetElementPtr.html">the getelementptr
3393FAQ</a>.</p>
3394
3395<h5>Example:</h5>
3396
3397<pre>
3398 <i>; yields [12 x i8]*:aptr</i>
3399 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
3400</pre>
3401</div>
3402
3403<!-- ======================================================================= -->
3404<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3405</div>
3406<div class="doc_text">
3407<p>The instructions in this category are the conversion instructions (casting)
3408which all take a single operand and a type. They perform various bit conversions
3409on the operand.</p>
3410</div>
3411
3412<!-- _______________________________________________________________________ -->
3413<div class="doc_subsubsection">
3414 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3415</div>
3416<div class="doc_text">
3417
3418<h5>Syntax:</h5>
3419<pre>
3420 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3421</pre>
3422
3423<h5>Overview:</h5>
3424<p>
3425The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3426</p>
3427
3428<h5>Arguments:</h5>
3429<p>
3430The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3431be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3432and type of the result, which must be an <a href="#t_integer">integer</a>
3433type. The bit size of <tt>value</tt> must be larger than the bit size of
3434<tt>ty2</tt>. Equal sized types are not allowed.</p>
3435
3436<h5>Semantics:</h5>
3437<p>
3438The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3439and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3440larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3441It will always truncate bits.</p>
3442
3443<h5>Example:</h5>
3444<pre>
3445 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3446 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3447 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3448</pre>
3449</div>
3450
3451<!-- _______________________________________________________________________ -->
3452<div class="doc_subsubsection">
3453 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3454</div>
3455<div class="doc_text">
3456
3457<h5>Syntax:</h5>
3458<pre>
3459 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3460</pre>
3461
3462<h5>Overview:</h5>
3463<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3464<tt>ty2</tt>.</p>
3465
3466
3467<h5>Arguments:</h5>
3468<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3469<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3470also be of <a href="#t_integer">integer</a> type. The bit size of the
3471<tt>value</tt> must be smaller than the bit size of the destination type,
3472<tt>ty2</tt>.</p>
3473
3474<h5>Semantics:</h5>
3475<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3476bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3477
3478<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3479
3480<h5>Example:</h5>
3481<pre>
3482 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3483 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3484</pre>
3485</div>
3486
3487<!-- _______________________________________________________________________ -->
3488<div class="doc_subsubsection">
3489 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3490</div>
3491<div class="doc_text">
3492
3493<h5>Syntax:</h5>
3494<pre>
3495 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3496</pre>
3497
3498<h5>Overview:</h5>
3499<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3500
3501<h5>Arguments:</h5>
3502<p>
3503The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3504<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3505also be of <a href="#t_integer">integer</a> type. The bit size of the
3506<tt>value</tt> must be smaller than the bit size of the destination type,
3507<tt>ty2</tt>.</p>
3508
3509<h5>Semantics:</h5>
3510<p>
3511The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3512bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3513the type <tt>ty2</tt>.</p>
3514
3515<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3516
3517<h5>Example:</h5>
3518<pre>
3519 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3520 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3521</pre>
3522</div>
3523
3524<!-- _______________________________________________________________________ -->
3525<div class="doc_subsubsection">
3526 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3527</div>
3528
3529<div class="doc_text">
3530
3531<h5>Syntax:</h5>
3532
3533<pre>
3534 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3535</pre>
3536
3537<h5>Overview:</h5>
3538<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3539<tt>ty2</tt>.</p>
3540
3541
3542<h5>Arguments:</h5>
3543<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3544 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3545cast it to. The size of <tt>value</tt> must be larger than the size of
3546<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3547<i>no-op cast</i>.</p>
3548
3549<h5>Semantics:</h5>
3550<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3551<a href="#t_floating">floating point</a> type to a smaller
3552<a href="#t_floating">floating point</a> type. If the value cannot fit within
3553the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3554
3555<h5>Example:</h5>
3556<pre>
3557 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3558 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3559</pre>
3560</div>
3561
3562<!-- _______________________________________________________________________ -->
3563<div class="doc_subsubsection">
3564 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3565</div>
3566<div class="doc_text">
3567
3568<h5>Syntax:</h5>
3569<pre>
3570 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3571</pre>
3572
3573<h5>Overview:</h5>
3574<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3575floating point value.</p>
3576
3577<h5>Arguments:</h5>
3578<p>The '<tt>fpext</tt>' instruction takes a
3579<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3580and a <a href="#t_floating">floating point</a> type to cast it to. The source
3581type must be smaller than the destination type.</p>
3582
3583<h5>Semantics:</h5>
3584<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3585<a href="#t_floating">floating point</a> type to a larger
3586<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3587used to make a <i>no-op cast</i> because it always changes bits. Use
3588<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3589
3590<h5>Example:</h5>
3591<pre>
3592 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3593 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3594</pre>
3595</div>
3596
3597<!-- _______________________________________________________________________ -->
3598<div class="doc_subsubsection">
3599 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3600</div>
3601<div class="doc_text">
3602
3603<h5>Syntax:</h5>
3604<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3605 &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]3606</pre>
3607
3608<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3609<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3610unsigned integer equivalent of type <tt>ty2</tt>.
3611</p>
3612
3613<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3614<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]3615scalar or vector <a href="#t_floating">floating point</a> value, and a type
3616to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3617type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3618vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3619
3620<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3621<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3622<a href="#t_floating">floating point</a> operand into the nearest (rounding
3623towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3624the results are undefined.</p>
3625
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3626<h5>Example:</h5>
3627<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3628 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3629 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3630 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3631</pre>
3632</div>
3633
3634<!-- _______________________________________________________________________ -->
3635<div class="doc_subsubsection">
3636 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3637</div>
3638<div class="doc_text">
3639
3640<h5>Syntax:</h5>
3641<pre>
3642 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3643</pre>
3644
3645<h5>Overview:</h5>
3646<p>The '<tt>fptosi</tt>' instruction converts
3647<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3648</p>
3649
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3650<h5>Arguments:</h5>
3651<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]3652scalar or vector <a href="#t_floating">floating point</a> value, and a type
3653to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3654type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3655vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3656
3657<h5>Semantics:</h5>
3658<p>The '<tt>fptosi</tt>' instruction converts its
3659<a href="#t_floating">floating point</a> operand into the nearest (rounding
3660towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3661the results are undefined.</p>
3662
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3663<h5>Example:</h5>
3664<pre>
3665 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3666 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3667 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3668</pre>
3669</div>
3670
3671<!-- _______________________________________________________________________ -->
3672<div class="doc_subsubsection">
3673 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3674</div>
3675<div class="doc_text">
3676
3677<h5>Syntax:</h5>
3678<pre>
3679 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3680</pre>
3681
3682<h5>Overview:</h5>
3683<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3684integer and converts that value to the <tt>ty2</tt> type.</p>
3685
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3686<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3687<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3688scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3689to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3690type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3691floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3692
3693<h5>Semantics:</h5>
3694<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3695integer quantity and converts it to the corresponding floating point value. If
3696the value cannot fit in the floating point value, the results are undefined.</p>
3697
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3698<h5>Example:</h5>
3699<pre>
3700 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3701 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3702</pre>
3703</div>
3704
3705<!-- _______________________________________________________________________ -->
3706<div class="doc_subsubsection">
3707 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3708</div>
3709<div class="doc_text">
3710
3711<h5>Syntax:</h5>
3712<pre>
3713 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3714</pre>
3715
3716<h5>Overview:</h5>
3717<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3718integer and converts that value to the <tt>ty2</tt> type.</p>
3719
3720<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3721<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3722scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3723to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3724type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3725floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3726
3727<h5>Semantics:</h5>
3728<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3729integer quantity and converts it to the corresponding floating point value. If
3730the value cannot fit in the floating point value, the results are undefined.</p>
3731
3732<h5>Example:</h5>
3733<pre>
3734 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3735 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3736</pre>
3737</div>
3738
3739<!-- _______________________________________________________________________ -->
3740<div class="doc_subsubsection">
3741 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3742</div>
3743<div class="doc_text">
3744
3745<h5>Syntax:</h5>
3746<pre>
3747 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3748</pre>
3749
3750<h5>Overview:</h5>
3751<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3752the integer type <tt>ty2</tt>.</p>
3753
3754<h5>Arguments:</h5>
3755<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3756must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3757<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3758
3759<h5>Semantics:</h5>
3760<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3761<tt>ty2</tt> by interpreting the pointer value as an integer and either
3762truncating or zero extending that value to the size of the integer type. If
3763<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3764<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3765are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3766change.</p>
3767
3768<h5>Example:</h5>
3769<pre>
3770 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3771 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3772</pre>
3773</div>
3774
3775<!-- _______________________________________________________________________ -->
3776<div class="doc_subsubsection">
3777 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3778</div>
3779<div class="doc_text">
3780
3781<h5>Syntax:</h5>
3782<pre>
3783 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3784</pre>
3785
3786<h5>Overview:</h5>
3787<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3788a pointer type, <tt>ty2</tt>.</p>
3789
3790<h5>Arguments:</h5>
3791<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3792value to cast, and a type to cast it to, which must be a
3793<a href="#t_pointer">pointer</a> type.
3794
3795<h5>Semantics:</h5>
3796<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3797<tt>ty2</tt> by applying either a zero extension or a truncation depending on
3798the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3799size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3800the size of a pointer then a zero extension is done. If they are the same size,
3801nothing is done (<i>no-op cast</i>).</p>
3802
3803<h5>Example:</h5>
3804<pre>
3805 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3806 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3807 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3808</pre>
3809</div>
3810
3811<!-- _______________________________________________________________________ -->
3812<div class="doc_subsubsection">
3813 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3814</div>
3815<div class="doc_text">
3816
3817<h5>Syntax:</h5>
3818<pre>
3819 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3820</pre>
3821
3822<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3823
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3824<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3825<tt>ty2</tt> without changing any bits.</p>
3826
3827<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3828
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3829<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
Dan Gohman7305fa02008-09-08 16:45:59 +0000[diff] [blame]3830a non-aggregate first class value, and a type to cast it to, which must also be
3831a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
3832<tt>value</tt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3833and the destination type, <tt>ty2</tt>, must be identical. If the source
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3834type is a pointer, the destination type must also be a pointer. This
3835instruction supports bitwise conversion of vectors to integers and to vectors
3836of other types (as long as they have the same size).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3837
3838<h5>Semantics:</h5>
3839<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3840<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3841this conversion. The conversion is done as if the <tt>value</tt> had been
3842stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3843converted to other pointer types with this instruction. To convert pointers to
3844other types, use the <a href="#i_inttoptr">inttoptr</a> or
3845<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3846
3847<h5>Example:</h5>
3848<pre>
3849 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3850 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3851 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3852</pre>
3853</div>
3854
3855<!-- ======================================================================= -->
3856<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3857<div class="doc_text">
3858<p>The instructions in this category are the "miscellaneous"
3859instructions, which defy better classification.</p>
3860</div>
3861
3862<!-- _______________________________________________________________________ -->
3863<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3864</div>
3865<div class="doc_text">
3866<h5>Syntax:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3867<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]3868</pre>
3869<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3870<p>The '<tt>icmp</tt>' instruction returns a boolean value or
3871a vector of boolean values based on comparison
3872of its two integer, integer vector, or pointer operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3873<h5>Arguments:</h5>
3874<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3875the condition code indicating the kind of comparison to perform. It is not
3876a value, just a keyword. The possible condition code are:
3877<ol>
3878 <li><tt>eq</tt>: equal</li>
3879 <li><tt>ne</tt>: not equal </li>
3880 <li><tt>ugt</tt>: unsigned greater than</li>
3881 <li><tt>uge</tt>: unsigned greater or equal</li>
3882 <li><tt>ult</tt>: unsigned less than</li>
3883 <li><tt>ule</tt>: unsigned less or equal</li>
3884 <li><tt>sgt</tt>: signed greater than</li>
3885 <li><tt>sge</tt>: signed greater or equal</li>
3886 <li><tt>slt</tt>: signed less than</li>
3887 <li><tt>sle</tt>: signed less or equal</li>
3888</ol>
3889<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3890<a href="#t_pointer">pointer</a>
3891or integer <a href="#t_vector">vector</a> typed.
3892They must also be identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3893<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3894<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]3895the condition code given as <tt>cond</tt>. The comparison performed always
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3896yields either an <a href="#t_primitive"><tt>i1</tt></a> or vector of <tt>i1</tt> result, as follows:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3897<ol>
3898 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3899 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3900 </li>
3901 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3902 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3903 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3904 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3905 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3906 <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]3907 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3908 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3909 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3910 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3911 <li><tt>sgt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3912 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3913 <li><tt>sge</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3914 <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]3915 <li><tt>slt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3916 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3917 <li><tt>sle</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3918 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3919</ol>
3920<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3921values are compared as if they were integers.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3922<p>If the operands are integer vectors, then they are compared
3923element by element. The result is an <tt>i1</tt> vector with
3924the same number of elements as the values being compared.
3925Otherwise, the result is an <tt>i1</tt>.
3926</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3927
3928<h5>Example:</h5>
3929<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
3930 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
3931 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
3932 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3933 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3934 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
3935</pre>
3936</div>
3937
3938<!-- _______________________________________________________________________ -->
3939<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3940</div>
3941<div class="doc_text">
3942<h5>Syntax:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3943<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]3944</pre>
3945<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3946<p>The '<tt>fcmp</tt>' instruction returns a boolean value
3947or vector of boolean values based on comparison
3948of its operands.
3949<p>
3950If the operands are floating point scalars, then the result
3951type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
3952</p>
3953<p>If the operands are floating point vectors, then the result type
3954is a vector of boolean with the same number of elements as the
3955operands being compared.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3956<h5>Arguments:</h5>
3957<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3958the condition code indicating the kind of comparison to perform. It is not
3959a value, just a keyword. The possible condition code are:
3960<ol>
3961 <li><tt>false</tt>: no comparison, always returns false</li>
3962 <li><tt>oeq</tt>: ordered and equal</li>
3963 <li><tt>ogt</tt>: ordered and greater than </li>
3964 <li><tt>oge</tt>: ordered and greater than or equal</li>
3965 <li><tt>olt</tt>: ordered and less than </li>
3966 <li><tt>ole</tt>: ordered and less than or equal</li>
3967 <li><tt>one</tt>: ordered and not equal</li>
3968 <li><tt>ord</tt>: ordered (no nans)</li>
3969 <li><tt>ueq</tt>: unordered or equal</li>
3970 <li><tt>ugt</tt>: unordered or greater than </li>
3971 <li><tt>uge</tt>: unordered or greater than or equal</li>
3972 <li><tt>ult</tt>: unordered or less than </li>
3973 <li><tt>ule</tt>: unordered or less than or equal</li>
3974 <li><tt>une</tt>: unordered or not equal</li>
3975 <li><tt>uno</tt>: unordered (either nans)</li>
3976 <li><tt>true</tt>: no comparison, always returns true</li>
3977</ol>
3978<p><i>Ordered</i> means that neither operand is a QNAN while
3979<i>unordered</i> means that either operand may be a QNAN.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3980<p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
3981either a <a href="#t_floating">floating point</a> type
3982or a <a href="#t_vector">vector</a> of floating point type.
3983They must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3984<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3985<p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3986according to the condition code given as <tt>cond</tt>.
3987If the operands are vectors, then the vectors are compared
3988element by element.
3989Each comparison performed
3990always yields an <a href="#t_primitive">i1</a> result, as follows:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3991<ol>
3992 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3993 <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]3994 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3995 <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]3996 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3997 <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]3998 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3999 <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]4000 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4001 <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]4002 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4003 <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]4004 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4005 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4006 <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]4007 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4008 <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]4009 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4010 <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]4011 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4012 <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]4013 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4014 <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]4015 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4016 <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]4017 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4018 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4019 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4020</ol>
4021
4022<h5>Example:</h5>
4023<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]4024 &lt;result&gt; = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4025 &lt;result&gt; = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4026 &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]4027</pre>
4028</div>
4029
4030<!-- _______________________________________________________________________ -->
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4031<div class="doc_subsubsection">
4032 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
4033</div>
4034<div class="doc_text">
4035<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4036<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]4037</pre>
4038<h5>Overview:</h5>
4039<p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
4040element-wise comparison of its two integer vector operands.</p>
4041<h5>Arguments:</h5>
4042<p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
4043the condition code indicating the kind of comparison to perform. It is not
4044a value, just a keyword. The possible condition code are:
4045<ol>
4046 <li><tt>eq</tt>: equal</li>
4047 <li><tt>ne</tt>: not equal </li>
4048 <li><tt>ugt</tt>: unsigned greater than</li>
4049 <li><tt>uge</tt>: unsigned greater or equal</li>
4050 <li><tt>ult</tt>: unsigned less than</li>
4051 <li><tt>ule</tt>: unsigned less or equal</li>
4052 <li><tt>sgt</tt>: signed greater than</li>
4053 <li><tt>sge</tt>: signed greater or equal</li>
4054 <li><tt>slt</tt>: signed less than</li>
4055 <li><tt>sle</tt>: signed less or equal</li>
4056</ol>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4057<p>The remaining two arguments must be <a href="#t_vector">vector</a> or
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4058<a href="#t_integer">integer</a> typed. They must also be identical types.</p>
4059<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4060<p>The '<tt>vicmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4061according to the condition code given as <tt>cond</tt>. The comparison yields a
4062<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
4063identical type as the values being compared. The most significant bit in each
4064element is 1 if the element-wise comparison evaluates to true, and is 0
4065otherwise. All other bits of the result are undefined. The condition codes
4066are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
4067instruction</a>.
4068
4069<h5>Example:</h5>
4070<pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4071 &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>
4072 &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]4073</pre>
4074</div>
4075
4076<!-- _______________________________________________________________________ -->
4077<div class="doc_subsubsection">
4078 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4079</div>
4080<div class="doc_text">
4081<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4082<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]4083<h5>Overview:</h5>
4084<p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4085element-wise comparison of its two floating point vector operands. The output
4086elements have the same width as the input elements.</p>
4087<h5>Arguments:</h5>
4088<p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4089the condition code indicating the kind of comparison to perform. It is not
4090a value, just a keyword. The possible condition code are:
4091<ol>
4092 <li><tt>false</tt>: no comparison, always returns false</li>
4093 <li><tt>oeq</tt>: ordered and equal</li>
4094 <li><tt>ogt</tt>: ordered and greater than </li>
4095 <li><tt>oge</tt>: ordered and greater than or equal</li>
4096 <li><tt>olt</tt>: ordered and less than </li>
4097 <li><tt>ole</tt>: ordered and less than or equal</li>
4098 <li><tt>one</tt>: ordered and not equal</li>
4099 <li><tt>ord</tt>: ordered (no nans)</li>
4100 <li><tt>ueq</tt>: unordered or equal</li>
4101 <li><tt>ugt</tt>: unordered or greater than </li>
4102 <li><tt>uge</tt>: unordered or greater than or equal</li>
4103 <li><tt>ult</tt>: unordered or less than </li>
4104 <li><tt>ule</tt>: unordered or less than or equal</li>
4105 <li><tt>une</tt>: unordered or not equal</li>
4106 <li><tt>uno</tt>: unordered (either nans)</li>
4107 <li><tt>true</tt>: no comparison, always returns true</li>
4108</ol>
4109<p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4110<a href="#t_floating">floating point</a> typed. They must also be identical
4111types.</p>
4112<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4113<p>The '<tt>vfcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4114according to the condition code given as <tt>cond</tt>. The comparison yields a
4115<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4116an identical number of elements as the values being compared, and each element
4117having identical with to the width of the floating point elements. The most
4118significant bit in each element is 1 if the element-wise comparison evaluates to
4119true, and is 0 otherwise. All other bits of the result are undefined. The
4120condition codes are evaluated identically to the
4121<a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.
4122
4123<h5>Example:</h5>
4124<pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4125 &lt;result&gt; = vfcmp oeq &lt;2 x float&gt; &lt; float 4, float 0 &gt;, &lt; float 5, float 0 &gt; <i>; yields: result=&lt;2 x i32&gt; &lt; i32 0, i32 -1 &gt;</i>
4126 &lt;result&gt; = vfcmp ult &lt;2 x double&gt; &lt; double 1, double 2 &gt;, &lt; double 2, double 2&gt; <i>; yields: result=&lt;2 x i64&gt; &lt; i64 -1, i64 0 &gt;</i>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4127</pre>
4128</div>
4129
4130<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4131<div class="doc_subsubsection">
4132 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4133</div>
4134
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4135<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4136
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4137<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4138
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4139<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
4140<h5>Overview:</h5>
4141<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
4142the SSA graph representing the function.</p>
4143<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4144
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4145<p>The type of the incoming values is specified with the first type
4146field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4147as arguments, with one pair for each predecessor basic block of the
4148current block. Only values of <a href="#t_firstclass">first class</a>
4149type may be used as the value arguments to the PHI node. Only labels
4150may be used as the label arguments.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4151
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4152<p>There must be no non-phi instructions between the start of a basic
4153block and the PHI instructions: i.e. PHI instructions must be first in
4154a basic block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4155
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4156<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4157
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4158<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4159specified by the pair corresponding to the predecessor basic block that executed
4160just prior to the current block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4161
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4162<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4163<pre>
4164Loop: ; Infinite loop that counts from 0 on up...
4165 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4166 %nextindvar = add i32 %indvar, 1
4167 br label %Loop
4168</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4169</div>
4170
4171<!-- _______________________________________________________________________ -->
4172<div class="doc_subsubsection">
4173 <a name="i_select">'<tt>select</tt>' Instruction</a>
4174</div>
4175
4176<div class="doc_text">
4177
4178<h5>Syntax:</h5>
4179
4180<pre>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4181 &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>
4182
4183 <i>selty</i> is either i1 or {&lt;N x i1&gt}
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4184</pre>
4185
4186<h5>Overview:</h5>
4187
4188<p>
4189The '<tt>select</tt>' instruction is used to choose one value based on a
4190condition, without branching.
4191</p>
4192
4193
4194<h5>Arguments:</h5>
4195
4196<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4197The '<tt>select</tt>' instruction requires an 'i1' value or
4198a vector of 'i1' values indicating the
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4199condition, and two values of the same <a href="#t_firstclass">first class</a>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4200type. If the val1/val2 are vectors and
4201the condition is a scalar, then entire vectors are selected, not
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4202individual elements.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4203</p>
4204
4205<h5>Semantics:</h5>
4206
4207<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4208If 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]4209value argument; otherwise, it returns the second value argument.
4210</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4211<p>
4212If the condition is a vector of i1, then the value arguments must
4213be vectors of the same size, and the selection is done element
4214by element.
4215</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4216
4217<h5>Example:</h5>
4218
4219<pre>
4220 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4221</pre>
4222</div>
4223
4224
4225<!-- _______________________________________________________________________ -->
4226<div class="doc_subsubsection">
4227 <a name="i_call">'<tt>call</tt>' Instruction</a>
4228</div>
4229
4230<div class="doc_text">
4231
4232<h5>Syntax:</h5>
4233<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4234 &lt;result&gt; = [tail] call [<a href="#callingconv">cconv</a>] &lt;ty&gt; [&lt;fnty&gt;*] &lt;fnptrval&gt;(&lt;param list&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4235</pre>
4236
4237<h5>Overview:</h5>
4238
4239<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4240
4241<h5>Arguments:</h5>
4242
4243<p>This instruction requires several arguments:</p>
4244
4245<ol>
4246 <li>
4247 <p>The optional "tail" marker indicates whether the callee function accesses
4248 any allocas or varargs in the caller. If the "tail" marker is present, the
4249 function call is eligible for tail call optimization. Note that calls may
4250 be marked "tail" even if they do not occur before a <a
4251 href="#i_ret"><tt>ret</tt></a> instruction.
4252 </li>
4253 <li>
4254 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4255 convention</a> the call should use. If none is specified, the call defaults
4256 to using C calling conventions.
4257 </li>
4258 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4259 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4260 the type of the return value. Functions that return no value are marked
4261 <tt><a href="#t_void">void</a></tt>.</p>
4262 </li>
4263 <li>
4264 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4265 value being invoked. The argument types must match the types implied by
4266 this signature. This type can be omitted if the function is not varargs
4267 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4268 </li>
4269 <li>
4270 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4271 be invoked. In most cases, this is a direct function invocation, but
4272 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4273 to function value.</p>
4274 </li>
4275 <li>
4276 <p>'<tt>function args</tt>': argument list whose types match the
4277 function signature argument types. All arguments must be of
4278 <a href="#t_firstclass">first class</a> type. If the function signature
4279 indicates the function accepts a variable number of arguments, the extra
4280 arguments can be specified.</p>
4281 </li>
4282</ol>
4283
4284<h5>Semantics:</h5>
4285
4286<p>The '<tt>call</tt>' instruction is used to cause control flow to
4287transfer to a specified function, with its incoming arguments bound to
4288the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4289instruction in the called function, control flow continues with the
4290instruction after the function call, and the return value of the
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4291function is bound to the result argument. If the callee returns multiple
4292values then the return values of the function are only accessible through
4293the '<tt><a href="#i_getresult">getresult</a></tt>' instruction.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4294
4295<h5>Example:</h5>
4296
4297<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4298 %retval = call i32 @test(i32 %argc)
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4299 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4300 %X = tail call i32 @foo() <i>; yields i32</i>
4301 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4302 call void %foo(i8 97 signext)
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4303
4304 %struct.A = type { i32, i8 }
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4305 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
4306 %gr = getresult %struct.A %r, 0 <i>; yields i32</i>
4307 %gr1 = getresult %struct.A %r, 1 <i>; yields i8</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4308</pre>
4309
4310</div>
4311
4312<!-- _______________________________________________________________________ -->
4313<div class="doc_subsubsection">
4314 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4315</div>
4316
4317<div class="doc_text">
4318
4319<h5>Syntax:</h5>
4320
4321<pre>
4322 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
4323</pre>
4324
4325<h5>Overview:</h5>
4326
4327<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4328the "variable argument" area of a function call. It is used to implement the
4329<tt>va_arg</tt> macro in C.</p>
4330
4331<h5>Arguments:</h5>
4332
4333<p>This instruction takes a <tt>va_list*</tt> value and the type of
4334the argument. It returns a value of the specified argument type and
4335increments the <tt>va_list</tt> to point to the next argument. The
4336actual type of <tt>va_list</tt> is target specific.</p>
4337
4338<h5>Semantics:</h5>
4339
4340<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4341type from the specified <tt>va_list</tt> and causes the
4342<tt>va_list</tt> to point to the next argument. For more information,
4343see the variable argument handling <a href="#int_varargs">Intrinsic
4344Functions</a>.</p>
4345
4346<p>It is legal for this instruction to be called in a function which does not
4347take a variable number of arguments, for example, the <tt>vfprintf</tt>
4348function.</p>
4349
4350<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4351href="#intrinsics">intrinsic function</a> because it takes a type as an
4352argument.</p>
4353
4354<h5>Example:</h5>
4355
4356<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4357
4358</div>
4359
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4360<!-- _______________________________________________________________________ -->
4361<div class="doc_subsubsection">
4362 <a name="i_getresult">'<tt>getresult</tt>' Instruction</a>
4363</div>
4364
4365<div class="doc_text">
4366
4367<h5>Syntax:</h5>
4368<pre>
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4369 &lt;resultval&gt; = getresult &lt;type&gt; &lt;retval&gt;, &lt;index&gt;
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4370</pre>
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4371
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4372<h5>Overview:</h5>
4373
4374<p> The '<tt>getresult</tt>' instruction is used to extract individual values
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4375from a '<tt><a href="#i_call">call</a></tt>'
4376or '<tt><a href="#i_invoke">invoke</a></tt>' instruction that returns multiple
4377results.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4378
4379<h5>Arguments:</h5>
4380
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4381<p>The '<tt>getresult</tt>' instruction takes a call or invoke value as its
Chris Lattnerd8dd3522008-04-23 04:06:52 +0000[diff] [blame]4382first argument, or an undef value. The value must have <a
4383href="#t_struct">structure type</a>. The second argument is a constant
4384unsigned index value which must be in range for the number of values returned
4385by the call.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4386
4387<h5>Semantics:</h5>
4388
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4389<p>The '<tt>getresult</tt>' instruction extracts the element identified by
4390'<tt>index</tt>' from the aggregate value.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4391
4392<h5>Example:</h5>
4393
4394<pre>
4395 %struct.A = type { i32, i8 }
4396
4397 %r = call %struct.A @foo()
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4398 %gr = getresult %struct.A %r, 0 <i>; yields i32:%gr</i>
4399 %gr1 = getresult %struct.A %r, 1 <i>; yields i8:%gr1</i>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4400 add i32 %gr, 42
4401 add i8 %gr1, 41
4402</pre>
4403
4404</div>
4405
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4406<!-- *********************************************************************** -->
4407<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4408<!-- *********************************************************************** -->
4409
4410<div class="doc_text">
4411
4412<p>LLVM supports the notion of an "intrinsic function". These functions have
4413well known names and semantics and are required to follow certain restrictions.
4414Overall, these intrinsics represent an extension mechanism for the LLVM
4415language that does not require changing all of the transformations in LLVM when
4416adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4417
4418<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4419prefix is reserved in LLVM for intrinsic names; thus, function names may not
4420begin with this prefix. Intrinsic functions must always be external functions:
4421you cannot define the body of intrinsic functions. Intrinsic functions may
4422only be used in call or invoke instructions: it is illegal to take the address
4423of an intrinsic function. Additionally, because intrinsic functions are part
4424of the LLVM language, it is required if any are added that they be documented
4425here.</p>
4426
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4427<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4428a family of functions that perform the same operation but on different data
4429types. Because LLVM can represent over 8 million different integer types,
4430overloading is used commonly to allow an intrinsic function to operate on any
4431integer type. One or more of the argument types or the result type can be
4432overloaded to accept any integer type. Argument types may also be defined as
4433exactly matching a previous argument's type or the result type. This allows an
4434intrinsic function which accepts multiple arguments, but needs all of them to
4435be of the same type, to only be overloaded with respect to a single argument or
4436the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4437
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4438<p>Overloaded intrinsics will have the names of its overloaded argument types
4439encoded into its function name, each preceded by a period. Only those types
4440which are overloaded result in a name suffix. Arguments whose type is matched
4441against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4442take an integer of any width and returns an integer of exactly the same integer
4443width. This leads to a family of functions such as
4444<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4445Only one type, the return type, is overloaded, and only one type suffix is
4446required. Because the argument's type is matched against the return type, it
4447does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4448
4449<p>To learn how to add an intrinsic function, please see the
4450<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4451</p>
4452
4453</div>
4454
4455<!-- ======================================================================= -->
4456<div class="doc_subsection">
4457 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4458</div>
4459
4460<div class="doc_text">
4461
4462<p>Variable argument support is defined in LLVM with the <a
4463 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4464intrinsic functions. These functions are related to the similarly
4465named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
4466
4467<p>All of these functions operate on arguments that use a
4468target-specific value type "<tt>va_list</tt>". The LLVM assembly
4469language reference manual does not define what this type is, so all
4470transformations should be prepared to handle these functions regardless of
4471the type used.</p>
4472
4473<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4474instruction and the variable argument handling intrinsic functions are
4475used.</p>
4476
4477<div class="doc_code">
4478<pre>
4479define i32 @test(i32 %X, ...) {
4480 ; Initialize variable argument processing
4481 %ap = alloca i8*
4482 %ap2 = bitcast i8** %ap to i8*
4483 call void @llvm.va_start(i8* %ap2)
4484
4485 ; Read a single integer argument
4486 %tmp = va_arg i8** %ap, i32
4487
4488 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4489 %aq = alloca i8*
4490 %aq2 = bitcast i8** %aq to i8*
4491 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4492 call void @llvm.va_end(i8* %aq2)
4493
4494 ; Stop processing of arguments.
4495 call void @llvm.va_end(i8* %ap2)
4496 ret i32 %tmp
4497}
4498
4499declare void @llvm.va_start(i8*)
4500declare void @llvm.va_copy(i8*, i8*)
4501declare void @llvm.va_end(i8*)
4502</pre>
4503</div>
4504
4505</div>
4506
4507<!-- _______________________________________________________________________ -->
4508<div class="doc_subsubsection">
4509 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4510</div>
4511
4512
4513<div class="doc_text">
4514<h5>Syntax:</h5>
4515<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
4516<h5>Overview:</h5>
4517<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
4518<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
4519href="#i_va_arg">va_arg</a></tt>.</p>
4520
4521<h5>Arguments:</h5>
4522
4523<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
4524
4525<h5>Semantics:</h5>
4526
4527<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
4528macro available in C. In a target-dependent way, it initializes the
4529<tt>va_list</tt> element to which the argument points, so that the next call to
4530<tt>va_arg</tt> will produce the first variable argument passed to the function.
4531Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4532last argument of the function as the compiler can figure that out.</p>
4533
4534</div>
4535
4536<!-- _______________________________________________________________________ -->
4537<div class="doc_subsubsection">
4538 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4539</div>
4540
4541<div class="doc_text">
4542<h5>Syntax:</h5>
4543<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
4544<h5>Overview:</h5>
4545
4546<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
4547which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4548or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4549
4550<h5>Arguments:</h5>
4551
4552<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4553
4554<h5>Semantics:</h5>
4555
4556<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4557macro available in C. In a target-dependent way, it destroys the
4558<tt>va_list</tt> element to which the argument points. Calls to <a
4559href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4560<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4561<tt>llvm.va_end</tt>.</p>
4562
4563</div>
4564
4565<!-- _______________________________________________________________________ -->
4566<div class="doc_subsubsection">
4567 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4568</div>
4569
4570<div class="doc_text">
4571
4572<h5>Syntax:</h5>
4573
4574<pre>
4575 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
4576</pre>
4577
4578<h5>Overview:</h5>
4579
4580<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4581from the source argument list to the destination argument list.</p>
4582
4583<h5>Arguments:</h5>
4584
4585<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4586The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4587
4588
4589<h5>Semantics:</h5>
4590
4591<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4592macro available in C. In a target-dependent way, it copies the source
4593<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4594intrinsic is necessary because the <tt><a href="#int_va_start">
4595llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4596example, memory allocation.</p>
4597
4598</div>
4599
4600<!-- ======================================================================= -->
4601<div class="doc_subsection">
4602 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4603</div>
4604
4605<div class="doc_text">
4606
4607<p>
4608LLVM support for <a href="GarbageCollection.html">Accurate Garbage
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]4609Collection</a> (GC) requires the implementation and generation of these
4610intrinsics.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4611These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4612stack</a>, as well as garbage collector implementations that require <a
4613href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4614Front-ends for type-safe garbage collected languages should generate these
4615intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4616href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4617</p>
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]4618
4619<p>The garbage collection intrinsics only operate on objects in the generic
4620 address space (address space zero).</p>
4621
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4622</div>
4623
4624<!-- _______________________________________________________________________ -->
4625<div class="doc_subsubsection">
4626 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4627</div>
4628
4629<div class="doc_text">
4630
4631<h5>Syntax:</h5>
4632
4633<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4634 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4635</pre>
4636
4637<h5>Overview:</h5>
4638
4639<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4640the code generator, and allows some metadata to be associated with it.</p>
4641
4642<h5>Arguments:</h5>
4643
4644<p>The first argument specifies the address of a stack object that contains the
4645root pointer. The second pointer (which must be either a constant or a global
4646value address) contains the meta-data to be associated with the root.</p>
4647
4648<h5>Semantics:</h5>
4649
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]4650<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]4651location. At compile-time, the code generator generates information to allow
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4652the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4653intrinsic may only be used in a function which <a href="#gc">specifies a GC
4654algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4655
4656</div>
4657
4658
4659<!-- _______________________________________________________________________ -->
4660<div class="doc_subsubsection">
4661 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4662</div>
4663
4664<div class="doc_text">
4665
4666<h5>Syntax:</h5>
4667
4668<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4669 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4670</pre>
4671
4672<h5>Overview:</h5>
4673
4674<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4675locations, allowing garbage collector implementations that require read
4676barriers.</p>
4677
4678<h5>Arguments:</h5>
4679
4680<p>The second argument is the address to read from, which should be an address
4681allocated from the garbage collector. The first object is a pointer to the
4682start of the referenced object, if needed by the language runtime (otherwise
4683null).</p>
4684
4685<h5>Semantics:</h5>
4686
4687<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4688instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4689garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4690may only be used in a function which <a href="#gc">specifies a GC
4691algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4692
4693</div>
4694
4695
4696<!-- _______________________________________________________________________ -->
4697<div class="doc_subsubsection">
4698 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4699</div>
4700
4701<div class="doc_text">
4702
4703<h5>Syntax:</h5>
4704
4705<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4706 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4707</pre>
4708
4709<h5>Overview:</h5>
4710
4711<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4712locations, allowing garbage collector implementations that require write
4713barriers (such as generational or reference counting collectors).</p>
4714
4715<h5>Arguments:</h5>
4716
4717<p>The first argument is the reference to store, the second is the start of the
4718object to store it to, and the third is the address of the field of Obj to
4719store to. If the runtime does not require a pointer to the object, Obj may be
4720null.</p>
4721
4722<h5>Semantics:</h5>
4723
4724<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4725instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4726garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4727may only be used in a function which <a href="#gc">specifies a GC
4728algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4729
4730</div>
4731
4732
4733
4734<!-- ======================================================================= -->
4735<div class="doc_subsection">
4736 <a name="int_codegen">Code Generator Intrinsics</a>
4737</div>
4738
4739<div class="doc_text">
4740<p>
4741These intrinsics are provided by LLVM to expose special features that may only
4742be implemented with code generator support.
4743</p>
4744
4745</div>
4746
4747<!-- _______________________________________________________________________ -->
4748<div class="doc_subsubsection">
4749 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4750</div>
4751
4752<div class="doc_text">
4753
4754<h5>Syntax:</h5>
4755<pre>
4756 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
4757</pre>
4758
4759<h5>Overview:</h5>
4760
4761<p>
4762The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4763target-specific value indicating the return address of the current function
4764or one of its callers.
4765</p>
4766
4767<h5>Arguments:</h5>
4768
4769<p>
4770The argument to this intrinsic indicates which function to return the address
4771for. Zero indicates the calling function, one indicates its caller, etc. The
4772argument is <b>required</b> to be a constant integer value.
4773</p>
4774
4775<h5>Semantics:</h5>
4776
4777<p>
4778The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4779the return address of the specified call frame, or zero if it cannot be
4780identified. The value returned by this intrinsic is likely to be incorrect or 0
4781for arguments other than zero, so it should only be used for debugging purposes.
4782</p>
4783
4784<p>
4785Note that calling this intrinsic does not prevent function inlining or other
4786aggressive transformations, so the value returned may not be that of the obvious
4787source-language caller.
4788</p>
4789</div>
4790
4791
4792<!-- _______________________________________________________________________ -->
4793<div class="doc_subsubsection">
4794 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4795</div>
4796
4797<div class="doc_text">
4798
4799<h5>Syntax:</h5>
4800<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4801 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4802</pre>
4803
4804<h5>Overview:</h5>
4805
4806<p>
4807The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4808target-specific frame pointer value for the specified stack frame.
4809</p>
4810
4811<h5>Arguments:</h5>
4812
4813<p>
4814The argument to this intrinsic indicates which function to return the frame
4815pointer for. Zero indicates the calling function, one indicates its caller,
4816etc. The argument is <b>required</b> to be a constant integer value.
4817</p>
4818
4819<h5>Semantics:</h5>
4820
4821<p>
4822The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4823the frame address of the specified call frame, or zero if it cannot be
4824identified. The value returned by this intrinsic is likely to be incorrect or 0
4825for arguments other than zero, so it should only be used for debugging purposes.
4826</p>
4827
4828<p>
4829Note that calling this intrinsic does not prevent function inlining or other
4830aggressive transformations, so the value returned may not be that of the obvious
4831source-language caller.
4832</p>
4833</div>
4834
4835<!-- _______________________________________________________________________ -->
4836<div class="doc_subsubsection">
4837 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4838</div>
4839
4840<div class="doc_text">
4841
4842<h5>Syntax:</h5>
4843<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4844 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4845</pre>
4846
4847<h5>Overview:</h5>
4848
4849<p>
4850The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4851the function stack, for use with <a href="#int_stackrestore">
4852<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4853features like scoped automatic variable sized arrays in C99.
4854</p>
4855
4856<h5>Semantics:</h5>
4857
4858<p>
4859This intrinsic returns a opaque pointer value that can be passed to <a
4860href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4861<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4862<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4863state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4864practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4865that were allocated after the <tt>llvm.stacksave</tt> was executed.
4866</p>
4867
4868</div>
4869
4870<!-- _______________________________________________________________________ -->
4871<div class="doc_subsubsection">
4872 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4873</div>
4874
4875<div class="doc_text">
4876
4877<h5>Syntax:</h5>
4878<pre>
4879 declare void @llvm.stackrestore(i8 * %ptr)
4880</pre>
4881
4882<h5>Overview:</h5>
4883
4884<p>
4885The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4886the function stack to the state it was in when the corresponding <a
4887href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4888useful for implementing language features like scoped automatic variable sized
4889arrays in C99.
4890</p>
4891
4892<h5>Semantics:</h5>
4893
4894<p>
4895See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4896</p>
4897
4898</div>
4899
4900
4901<!-- _______________________________________________________________________ -->
4902<div class="doc_subsubsection">
4903 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4904</div>
4905
4906<div class="doc_text">
4907
4908<h5>Syntax:</h5>
4909<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4910 declare void @llvm.prefetch(i8* &lt;address&gt;, i32 &lt;rw&gt;, i32 &lt;locality&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4911</pre>
4912
4913<h5>Overview:</h5>
4914
4915
4916<p>
4917The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4918a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4919no
4920effect on the behavior of the program but can change its performance
4921characteristics.
4922</p>
4923
4924<h5>Arguments:</h5>
4925
4926<p>
4927<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4928determining if the fetch should be for a read (0) or write (1), and
4929<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4930locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4931<tt>locality</tt> arguments must be constant integers.
4932</p>
4933
4934<h5>Semantics:</h5>
4935
4936<p>
4937This intrinsic does not modify the behavior of the program. In particular,
4938prefetches cannot trap and do not produce a value. On targets that support this
4939intrinsic, the prefetch can provide hints to the processor cache for better
4940performance.
4941</p>
4942
4943</div>
4944
4945<!-- _______________________________________________________________________ -->
4946<div class="doc_subsubsection">
4947 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4948</div>
4949
4950<div class="doc_text">
4951
4952<h5>Syntax:</h5>
4953<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4954 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4955</pre>
4956
4957<h5>Overview:</h5>
4958
4959
4960<p>
4961The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]4962(PC) in a region of
4963code to simulators and other tools. The method is target specific, but it is
4964expected that the marker will use exported symbols to transmit the PC of the
4965marker.
4966The marker makes no guarantees that it will remain with any specific instruction
4967after optimizations. It is possible that the presence of a marker will inhibit
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4968optimizations. The intended use is to be inserted after optimizations to allow
4969correlations of simulation runs.
4970</p>
4971
4972<h5>Arguments:</h5>
4973
4974<p>
4975<tt>id</tt> is a numerical id identifying the marker.
4976</p>
4977
4978<h5>Semantics:</h5>
4979
4980<p>
4981This intrinsic does not modify the behavior of the program. Backends that do not
4982support this intrinisic may ignore it.
4983</p>
4984
4985</div>
4986
4987<!-- _______________________________________________________________________ -->
4988<div class="doc_subsubsection">
4989 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4990</div>
4991
4992<div class="doc_text">
4993
4994<h5>Syntax:</h5>
4995<pre>
4996 declare i64 @llvm.readcyclecounter( )
4997</pre>
4998
4999<h5>Overview:</h5>
5000
5001
5002<p>
5003The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5004counter register (or similar low latency, high accuracy clocks) on those targets
5005that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
5006As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
5007should only be used for small timings.
5008</p>
5009
5010<h5>Semantics:</h5>
5011
5012<p>
5013When directly supported, reading the cycle counter should not modify any memory.
5014Implementations are allowed to either return a application specific value or a
5015system wide value. On backends without support, this is lowered to a constant 0.
5016</p>
5017
5018</div>
5019
5020<!-- ======================================================================= -->
5021<div class="doc_subsection">
5022 <a name="int_libc">Standard C Library Intrinsics</a>
5023</div>
5024
5025<div class="doc_text">
5026<p>
5027LLVM provides intrinsics for a few important standard C library functions.
5028These intrinsics allow source-language front-ends to pass information about the
5029alignment of the pointer arguments to the code generator, providing opportunity
5030for more efficient code generation.
5031</p>
5032
5033</div>
5034
5035<!-- _______________________________________________________________________ -->
5036<div class="doc_subsubsection">
5037 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5038</div>
5039
5040<div class="doc_text">
5041
5042<h5>Syntax:</h5>
5043<pre>
5044 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5045 i32 &lt;len&gt;, i32 &lt;align&gt;)
5046 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5047 i64 &lt;len&gt;, i32 &lt;align&gt;)
5048</pre>
5049
5050<h5>Overview:</h5>
5051
5052<p>
5053The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5054location to the destination location.
5055</p>
5056
5057<p>
5058Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5059intrinsics do not return a value, and takes an extra alignment argument.
5060</p>
5061
5062<h5>Arguments:</h5>
5063
5064<p>
5065The first argument is a pointer to the destination, the second is a pointer to
5066the source. The third argument is an integer argument
5067specifying the number of bytes to copy, and the fourth argument is the alignment
5068of the source and destination locations.
5069</p>
5070
5071<p>
5072If the call to this intrinisic has an alignment value that is not 0 or 1, then
5073the caller guarantees that both the source and destination pointers are aligned
5074to that boundary.
5075</p>
5076
5077<h5>Semantics:</h5>
5078
5079<p>
5080The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5081location to the destination location, which are not allowed to overlap. It
5082copies "len" bytes of memory over. If the argument is known to be aligned to
5083some boundary, this can be specified as the fourth argument, otherwise it should
5084be set to 0 or 1.
5085</p>
5086</div>
5087
5088
5089<!-- _______________________________________________________________________ -->
5090<div class="doc_subsubsection">
5091 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5092</div>
5093
5094<div class="doc_text">
5095
5096<h5>Syntax:</h5>
5097<pre>
5098 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5099 i32 &lt;len&gt;, i32 &lt;align&gt;)
5100 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5101 i64 &lt;len&gt;, i32 &lt;align&gt;)
5102</pre>
5103
5104<h5>Overview:</h5>
5105
5106<p>
5107The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5108location to the destination location. It is similar to the
Chris Lattnerdba16ea2008-01-06 19:51:52 +0000[diff] [blame]5109'<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5110</p>
5111
5112<p>
5113Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5114intrinsics do not return a value, and takes an extra alignment argument.
5115</p>
5116
5117<h5>Arguments:</h5>
5118
5119<p>
5120The first argument is a pointer to the destination, the second is a pointer to
5121the source. The third argument is an integer argument
5122specifying the number of bytes to copy, and the fourth argument is the alignment
5123of the source and destination locations.
5124</p>
5125
5126<p>
5127If the call to this intrinisic has an alignment value that is not 0 or 1, then
5128the caller guarantees that the source and destination pointers are aligned to
5129that boundary.
5130</p>
5131
5132<h5>Semantics:</h5>
5133
5134<p>
5135The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5136location to the destination location, which may overlap. It
5137copies "len" bytes of memory over. If the argument is known to be aligned to
5138some boundary, this can be specified as the fourth argument, otherwise it should
5139be set to 0 or 1.
5140</p>
5141</div>
5142
5143
5144<!-- _______________________________________________________________________ -->
5145<div class="doc_subsubsection">
5146 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5147</div>
5148
5149<div class="doc_text">
5150
5151<h5>Syntax:</h5>
5152<pre>
5153 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5154 i32 &lt;len&gt;, i32 &lt;align&gt;)
5155 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5156 i64 &lt;len&gt;, i32 &lt;align&gt;)
5157</pre>
5158
5159<h5>Overview:</h5>
5160
5161<p>
5162The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5163byte value.
5164</p>
5165
5166<p>
5167Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5168does not return a value, and takes an extra alignment argument.
5169</p>
5170
5171<h5>Arguments:</h5>
5172
5173<p>
5174The first argument is a pointer to the destination to fill, the second is the
5175byte value to fill it with, the third argument is an integer
5176argument specifying the number of bytes to fill, and the fourth argument is the
5177known alignment of destination location.
5178</p>
5179
5180<p>
5181If the call to this intrinisic has an alignment value that is not 0 or 1, then
5182the caller guarantees that the destination pointer is aligned to that boundary.
5183</p>
5184
5185<h5>Semantics:</h5>
5186
5187<p>
5188The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5189the
5190destination location. If the argument is known to be aligned to some boundary,
5191this can be specified as the fourth argument, otherwise it should be set to 0 or
51921.
5193</p>
5194</div>
5195
5196
5197<!-- _______________________________________________________________________ -->
5198<div class="doc_subsubsection">
5199 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5200</div>
5201
5202<div class="doc_text">
5203
5204<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5205<p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5206floating point or vector of floating point type. Not all targets support all
5207types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5208<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5209 declare float @llvm.sqrt.f32(float %Val)
5210 declare double @llvm.sqrt.f64(double %Val)
5211 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5212 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5213 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5214</pre>
5215
5216<h5>Overview:</h5>
5217
5218<p>
5219The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5220returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5221<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
Chris Lattnerf914a002008-01-29 07:00:44 +0000[diff] [blame]5222negative numbers other than -0.0 (which allows for better optimization, because
5223there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5224defined to return -0.0 like IEEE sqrt.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5225</p>
5226
5227<h5>Arguments:</h5>
5228
5229<p>
5230The argument and return value are floating point numbers of the same type.
5231</p>
5232
5233<h5>Semantics:</h5>
5234
5235<p>
5236This function returns the sqrt of the specified operand if it is a nonnegative
5237floating point number.
5238</p>
5239</div>
5240
5241<!-- _______________________________________________________________________ -->
5242<div class="doc_subsubsection">
5243 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5244</div>
5245
5246<div class="doc_text">
5247
5248<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5249<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]5250floating point or vector of floating point type. Not all targets support all
5251types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5252<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5253 declare float @llvm.powi.f32(float %Val, i32 %power)
5254 declare double @llvm.powi.f64(double %Val, i32 %power)
5255 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5256 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5257 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5258</pre>
5259
5260<h5>Overview:</h5>
5261
5262<p>
5263The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5264specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5265multiplications is not defined. When a vector of floating point type is
5266used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5267</p>
5268
5269<h5>Arguments:</h5>
5270
5271<p>
5272The second argument is an integer power, and the first is a value to raise to
5273that power.
5274</p>
5275
5276<h5>Semantics:</h5>
5277
5278<p>
5279This function returns the first value raised to the second power with an
5280unspecified sequence of rounding operations.</p>
5281</div>
5282
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5283<!-- _______________________________________________________________________ -->
5284<div class="doc_subsubsection">
5285 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5286</div>
5287
5288<div class="doc_text">
5289
5290<h5>Syntax:</h5>
5291<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5292floating point or vector of floating point type. Not all targets support all
5293types however.
5294<pre>
5295 declare float @llvm.sin.f32(float %Val)
5296 declare double @llvm.sin.f64(double %Val)
5297 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5298 declare fp128 @llvm.sin.f128(fp128 %Val)
5299 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5300</pre>
5301
5302<h5>Overview:</h5>
5303
5304<p>
5305The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5306</p>
5307
5308<h5>Arguments:</h5>
5309
5310<p>
5311The argument and return value are floating point numbers of the same type.
5312</p>
5313
5314<h5>Semantics:</h5>
5315
5316<p>
5317This function returns the sine of the specified operand, returning the
5318same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5319conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5320</div>
5321
5322<!-- _______________________________________________________________________ -->
5323<div class="doc_subsubsection">
5324 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5325</div>
5326
5327<div class="doc_text">
5328
5329<h5>Syntax:</h5>
5330<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5331floating point or vector of floating point type. Not all targets support all
5332types however.
5333<pre>
5334 declare float @llvm.cos.f32(float %Val)
5335 declare double @llvm.cos.f64(double %Val)
5336 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5337 declare fp128 @llvm.cos.f128(fp128 %Val)
5338 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5339</pre>
5340
5341<h5>Overview:</h5>
5342
5343<p>
5344The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5345</p>
5346
5347<h5>Arguments:</h5>
5348
5349<p>
5350The argument and return value are floating point numbers of the same type.
5351</p>
5352
5353<h5>Semantics:</h5>
5354
5355<p>
5356This function returns the cosine of the specified operand, returning the
5357same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5358conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5359</div>
5360
5361<!-- _______________________________________________________________________ -->
5362<div class="doc_subsubsection">
5363 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5364</div>
5365
5366<div class="doc_text">
5367
5368<h5>Syntax:</h5>
5369<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5370floating point or vector of floating point type. Not all targets support all
5371types however.
5372<pre>
5373 declare float @llvm.pow.f32(float %Val, float %Power)
5374 declare double @llvm.pow.f64(double %Val, double %Power)
5375 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5376 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5377 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5378</pre>
5379
5380<h5>Overview:</h5>
5381
5382<p>
5383The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5384specified (positive or negative) power.
5385</p>
5386
5387<h5>Arguments:</h5>
5388
5389<p>
5390The second argument is a floating point power, and the first is a value to
5391raise to that power.
5392</p>
5393
5394<h5>Semantics:</h5>
5395
5396<p>
5397This function returns the first value raised to the second power,
5398returning the
5399same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5400conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5401</div>
5402
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5403
5404<!-- ======================================================================= -->
5405<div class="doc_subsection">
5406 <a name="int_manip">Bit Manipulation Intrinsics</a>
5407</div>
5408
5409<div class="doc_text">
5410<p>
5411LLVM provides intrinsics for a few important bit manipulation operations.
5412These allow efficient code generation for some algorithms.
5413</p>
5414
5415</div>
5416
5417<!-- _______________________________________________________________________ -->
5418<div class="doc_subsubsection">
5419 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5420</div>
5421
5422<div class="doc_text">
5423
5424<h5>Syntax:</h5>
5425<p>This is an overloaded intrinsic function. You can use bswap on any integer
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5426type that is an even number of bytes (i.e. BitWidth % 16 == 0).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5427<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5428 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
5429 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
5430 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5431</pre>
5432
5433<h5>Overview:</h5>
5434
5435<p>
5436The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5437values with an even number of bytes (positive multiple of 16 bits). These are
5438useful for performing operations on data that is not in the target's native
5439byte order.
5440</p>
5441
5442<h5>Semantics:</h5>
5443
5444<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5445The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5446and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5447intrinsic returns an i32 value that has the four bytes of the input i32
5448swapped, 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]5449i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5450<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5451additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5452</p>
5453
5454</div>
5455
5456<!-- _______________________________________________________________________ -->
5457<div class="doc_subsubsection">
5458 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5459</div>
5460
5461<div class="doc_text">
5462
5463<h5>Syntax:</h5>
5464<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
5465width. Not all targets support all bit widths however.
5466<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5467 declare i8 @llvm.ctpop.i8 (i8 &lt;src&gt;)
5468 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5469 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5470 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
5471 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5472</pre>
5473
5474<h5>Overview:</h5>
5475
5476<p>
5477The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5478value.
5479</p>
5480
5481<h5>Arguments:</h5>
5482
5483<p>
5484The only argument is the value to be counted. The argument may be of any
5485integer type. The return type must match the argument type.
5486</p>
5487
5488<h5>Semantics:</h5>
5489
5490<p>
5491The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5492</p>
5493</div>
5494
5495<!-- _______________________________________________________________________ -->
5496<div class="doc_subsubsection">
5497 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5498</div>
5499
5500<div class="doc_text">
5501
5502<h5>Syntax:</h5>
5503<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5504integer bit width. Not all targets support all bit widths however.
5505<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5506 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
5507 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5508 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5509 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
5510 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5511</pre>
5512
5513<h5>Overview:</h5>
5514
5515<p>
5516The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5517leading zeros in a variable.
5518</p>
5519
5520<h5>Arguments:</h5>
5521
5522<p>
5523The only argument is the value to be counted. The argument may be of any
5524integer type. The return type must match the argument type.
5525</p>
5526
5527<h5>Semantics:</h5>
5528
5529<p>
5530The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5531in a variable. If the src == 0 then the result is the size in bits of the type
5532of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5533</p>
5534</div>
5535
5536
5537
5538<!-- _______________________________________________________________________ -->
5539<div class="doc_subsubsection">
5540 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5541</div>
5542
5543<div class="doc_text">
5544
5545<h5>Syntax:</h5>
5546<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5547integer bit width. Not all targets support all bit widths however.
5548<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5549 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
5550 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5551 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5552 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
5553 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5554</pre>
5555
5556<h5>Overview:</h5>
5557
5558<p>
5559The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5560trailing zeros.
5561</p>
5562
5563<h5>Arguments:</h5>
5564
5565<p>
5566The only argument is the value to be counted. The argument may be of any
5567integer type. The return type must match the argument type.
5568</p>
5569
5570<h5>Semantics:</h5>
5571
5572<p>
5573The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5574in a variable. If the src == 0 then the result is the size in bits of the type
5575of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5576</p>
5577</div>
5578
5579<!-- _______________________________________________________________________ -->
5580<div class="doc_subsubsection">
5581 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5582</div>
5583
5584<div class="doc_text">
5585
5586<h5>Syntax:</h5>
5587<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
5588on any integer bit width.
5589<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5590 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5591 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5592</pre>
5593
5594<h5>Overview:</h5>
5595<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5596range of bits from an integer value and returns them in the same bit width as
5597the original value.</p>
5598
5599<h5>Arguments:</h5>
5600<p>The first argument, <tt>%val</tt> and the result may be integer types of
5601any bit width but they must have the same bit width. The second and third
5602arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5603
5604<h5>Semantics:</h5>
5605<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5606of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5607<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5608operates in forward mode.</p>
5609<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5610right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5611only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5612<ol>
5613 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5614 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5615 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5616 to determine the number of bits to retain.</li>
5617 <li>A mask of the retained bits is created by shifting a -1 value.</li>
5618 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
5619</ol>
5620<p>In reverse mode, a similar computation is made except that the bits are
5621returned in the reverse order. So, for example, if <tt>X</tt> has the value
5622<tt>i16 0x0ACF (101011001111)</tt> and we apply
5623<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5624<tt>i16 0x0026 (000000100110)</tt>.</p>
5625</div>
5626
5627<div class="doc_subsubsection">
5628 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5629</div>
5630
5631<div class="doc_text">
5632
5633<h5>Syntax:</h5>
5634<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
5635on any integer bit width.
5636<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5637 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5638 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]5639</pre>
5640
5641<h5>Overview:</h5>
5642<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5643of bits in an integer value with another integer value. It returns the integer
5644with the replaced bits.</p>
5645
5646<h5>Arguments:</h5>
5647<p>The first argument, <tt>%val</tt> and the result may be integer types of
5648any bit width but they must have the same bit width. <tt>%val</tt> is the value
5649whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5650integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5651type since they specify only a bit index.</p>
5652
5653<h5>Semantics:</h5>
5654<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5655of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5656<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5657operates in forward mode.</p>
5658<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5659truncating it down to the size of the replacement area or zero extending it
5660up to that size.</p>
5661<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5662are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5663in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5664to the <tt>%hi</tt>th bit.
5665<p>In reverse mode, a similar computation is made except that the bits are
5666reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5667<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
5668<h5>Examples:</h5>
5669<pre>
5670 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
5671 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
5672 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
5673 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
5674 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
5675</pre>
5676</div>
5677
5678<!-- ======================================================================= -->
5679<div class="doc_subsection">
5680 <a name="int_debugger">Debugger Intrinsics</a>
5681</div>
5682
5683<div class="doc_text">
5684<p>
5685The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5686are described in the <a
5687href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5688Debugging</a> document.
5689</p>
5690</div>
5691
5692
5693<!-- ======================================================================= -->
5694<div class="doc_subsection">
5695 <a name="int_eh">Exception Handling Intrinsics</a>
5696</div>
5697
5698<div class="doc_text">
5699<p> The LLVM exception handling intrinsics (which all start with
5700<tt>llvm.eh.</tt> prefix), are described in the <a
5701href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5702Handling</a> document. </p>
5703</div>
5704
5705<!-- ======================================================================= -->
5706<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5707 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5708</div>
5709
5710<div class="doc_text">
5711<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5712 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5713 the <tt>nest</tt> attribute, from a function. The result is a callable
5714 function pointer lacking the nest parameter - the caller does not need
5715 to provide a value for it. Instead, the value to use is stored in
5716 advance in a "trampoline", a block of memory usually allocated
5717 on the stack, which also contains code to splice the nest value into the
5718 argument list. This is used to implement the GCC nested function address
5719 extension.
5720</p>
5721<p>
5722 For example, if the function is
5723 <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]5724 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]5725<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5726 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5727 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5728 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5729 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5730</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]5731 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5732 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5733</div>
5734
5735<!-- _______________________________________________________________________ -->
5736<div class="doc_subsubsection">
5737 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5738</div>
5739<div class="doc_text">
5740<h5>Syntax:</h5>
5741<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5742declare 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]5743</pre>
5744<h5>Overview:</h5>
5745<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5746 This fills the memory pointed to by <tt>tramp</tt> with code
5747 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5748</p>
5749<h5>Arguments:</h5>
5750<p>
5751 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5752 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5753 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]5754 intrinsic. Note that the size and the alignment are target-specific - LLVM
5755 currently provides no portable way of determining them, so a front-end that
5756 generates this intrinsic needs to have some target-specific knowledge.
5757 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]5758</p>
5759<h5>Semantics:</h5>
5760<p>
5761 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5762 dependent code, turning it into a function. A pointer to this function is
5763 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5764 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5765 before being called. The new function's signature is the same as that of
5766 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5767 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5768 of pointer type. Calling the new function is equivalent to calling
5769 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5770 missing <tt>nest</tt> argument. If, after calling
5771 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5772 modified, then the effect of any later call to the returned function pointer is
5773 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5774</p>
5775</div>
5776
5777<!-- ======================================================================= -->
5778<div class="doc_subsection">
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]5779 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
5780</div>
5781
5782<div class="doc_text">
5783<p>
5784 These intrinsic functions expand the "universal IR" of LLVM to represent
5785 hardware constructs for atomic operations and memory synchronization. This
5786 provides an interface to the hardware, not an interface to the programmer. It
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]5787 is aimed at a low enough level to allow any programming models or APIs
5788 (Application Programming Interfaces) which
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]5789 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
5790 hardware behavior. Just as hardware provides a "universal IR" for source
5791 languages, it also provides a starting point for developing a "universal"
5792 atomic operation and synchronization IR.
5793</p>
5794<p>
5795 These do <em>not</em> form an API such as high-level threading libraries,
5796 software transaction memory systems, atomic primitives, and intrinsic
5797 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
5798 application libraries. The hardware interface provided by LLVM should allow
5799 a clean implementation of all of these APIs and parallel programming models.
5800 No one model or paradigm should be selected above others unless the hardware
5801 itself ubiquitously does so.
5802
5803</p>
5804</div>
5805
5806<!-- _______________________________________________________________________ -->
5807<div class="doc_subsubsection">
5808 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
5809</div>
5810<div class="doc_text">
5811<h5>Syntax:</h5>
5812<pre>
5813declare void @llvm.memory.barrier( i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt;,
5814i1 &lt;device&gt; )
5815
5816</pre>
5817<h5>Overview:</h5>
5818<p>
5819 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
5820 specific pairs of memory access types.
5821</p>
5822<h5>Arguments:</h5>
5823<p>
5824 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
5825 The first four arguments enables a specific barrier as listed below. The fith
5826 argument specifies that the barrier applies to io or device or uncached memory.
5827
5828</p>
5829 <ul>
5830 <li><tt>ll</tt>: load-load barrier</li>
5831 <li><tt>ls</tt>: load-store barrier</li>
5832 <li><tt>sl</tt>: store-load barrier</li>
5833 <li><tt>ss</tt>: store-store barrier</li>
5834 <li><tt>device</tt>: barrier applies to device and uncached memory also.
5835 </ul>
5836<h5>Semantics:</h5>
5837<p>
5838 This intrinsic causes the system to enforce some ordering constraints upon
5839 the loads and stores of the program. This barrier does not indicate
5840 <em>when</em> any events will occur, it only enforces an <em>order</em> in
5841 which they occur. For any of the specified pairs of load and store operations
5842 (f.ex. load-load, or store-load), all of the first operations preceding the
5843 barrier will complete before any of the second operations succeeding the
5844 barrier begin. Specifically the semantics for each pairing is as follows:
5845</p>
5846 <ul>
5847 <li><tt>ll</tt>: All loads before the barrier must complete before any load
5848 after the barrier begins.</li>
5849
5850 <li><tt>ls</tt>: All loads before the barrier must complete before any
5851 store after the barrier begins.</li>
5852 <li><tt>ss</tt>: All stores before the barrier must complete before any
5853 store after the barrier begins.</li>
5854 <li><tt>sl</tt>: All stores before the barrier must complete before any
5855 load after the barrier begins.</li>
5856 </ul>
5857<p>
5858 These semantics are applied with a logical "and" behavior when more than one
5859 is enabled in a single memory barrier intrinsic.
5860</p>
5861<p>
5862 Backends may implement stronger barriers than those requested when they do not
5863 support as fine grained a barrier as requested. Some architectures do not
5864 need all types of barriers and on such architectures, these become noops.
5865</p>
5866<h5>Example:</h5>
5867<pre>
5868%ptr = malloc i32
5869 store i32 4, %ptr
5870
5871%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
5872 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
5873 <i>; guarantee the above finishes</i>
5874 store i32 8, %ptr <i>; before this begins</i>
5875</pre>
5876</div>
5877
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5878<!-- _______________________________________________________________________ -->
5879<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5880 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5881</div>
5882<div class="doc_text">
5883<h5>Syntax:</h5>
5884<p>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5885 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
5886 any integer bit width and for different address spaces. Not all targets
5887 support all bit widths however.</p>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5888
5889<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5890declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt; )
5891declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt; )
5892declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt; )
5893declare 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]5894
5895</pre>
5896<h5>Overview:</h5>
5897<p>
5898 This loads a value in memory and compares it to a given value. If they are
5899 equal, it stores a new value into the memory.
5900</p>
5901<h5>Arguments:</h5>
5902<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5903 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5904 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
5905 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
5906 this integer type. While any bit width integer may be used, targets may only
5907 lower representations they support in hardware.
5908
5909</p>
5910<h5>Semantics:</h5>
5911<p>
5912 This entire intrinsic must be executed atomically. It first loads the value
5913 in memory pointed to by <tt>ptr</tt> and compares it with the value
5914 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
5915 loaded value is yielded in all cases. This provides the equivalent of an
5916 atomic compare-and-swap operation within the SSA framework.
5917</p>
5918<h5>Examples:</h5>
5919
5920<pre>
5921%ptr = malloc i32
5922 store i32 4, %ptr
5923
5924%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5925%result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5926 <i>; yields {i32}:result1 = 4</i>
5927%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5928%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5929
5930%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5931%result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5932 <i>; yields {i32}:result2 = 8</i>
5933%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
5934
5935%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
5936</pre>
5937</div>
5938
5939<!-- _______________________________________________________________________ -->
5940<div class="doc_subsubsection">
5941 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
5942</div>
5943<div class="doc_text">
5944<h5>Syntax:</h5>
5945
5946<p>
5947 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
5948 integer bit width. Not all targets support all bit widths however.</p>
5949<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5950declare i8 @llvm.atomic.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;val&gt; )
5951declare i16 @llvm.atomic.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;val&gt; )
5952declare i32 @llvm.atomic.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;val&gt; )
5953declare i64 @llvm.atomic.swap.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;val&gt; )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5954
5955</pre>
5956<h5>Overview:</h5>
5957<p>
5958 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
5959 the value from memory. It then stores the value in <tt>val</tt> in the memory
5960 at <tt>ptr</tt>.
5961</p>
5962<h5>Arguments:</h5>
5963
5964<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5965 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5966 <tt>val</tt> argument and the result must be integers of the same bit width.
5967 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
5968 integer type. The targets may only lower integer representations they
5969 support.
5970</p>
5971<h5>Semantics:</h5>
5972<p>
5973 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
5974 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
5975 equivalent of an atomic swap operation within the SSA framework.
5976
5977</p>
5978<h5>Examples:</h5>
5979<pre>
5980%ptr = malloc i32
5981 store i32 4, %ptr
5982
5983%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5984%result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5985 <i>; yields {i32}:result1 = 4</i>
5986%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5987%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5988
5989%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5990%result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5991 <i>; yields {i32}:result2 = 8</i>
5992
5993%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
5994%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
5995</pre>
5996</div>
5997
5998<!-- _______________________________________________________________________ -->
5999<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6000 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6001
6002</div>
6003<div class="doc_text">
6004<h5>Syntax:</h5>
6005<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6006 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]6007 integer bit width. Not all targets support all bit widths however.</p>
6008<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6009declare i8 @llvm.atomic.load.add.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6010declare i16 @llvm.atomic.load.add.i16..p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6011declare i32 @llvm.atomic.load.add.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6012declare 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]6013
6014</pre>
6015<h5>Overview:</h5>
6016<p>
6017 This intrinsic adds <tt>delta</tt> to the value stored in memory at
6018 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6019</p>
6020<h5>Arguments:</h5>
6021<p>
6022
6023 The intrinsic takes two arguments, the first a pointer to an integer value
6024 and the second an integer value. The result is also an integer value. These
6025 integer types can have any bit width, but they must all have the same bit
6026 width. The targets may only lower integer representations they support.
6027</p>
6028<h5>Semantics:</h5>
6029<p>
6030 This intrinsic does a series of operations atomically. It first loads the
6031 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6032 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6033</p>
6034
6035<h5>Examples:</h5>
6036<pre>
6037%ptr = malloc i32
6038 store i32 4, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6039%result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6040 <i>; yields {i32}:result1 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6041%result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6042 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6043%result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6044 <i>; yields {i32}:result3 = 10</i>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6045%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6046</pre>
6047</div>
6048
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6049<!-- _______________________________________________________________________ -->
6050<div class="doc_subsubsection">
6051 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6052
6053</div>
6054<div class="doc_text">
6055<h5>Syntax:</h5>
6056<p>
6057 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]6058 any integer bit width and for different address spaces. Not all targets
6059 support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6060<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6061declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6062declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6063declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6064declare 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]6065
6066</pre>
6067<h5>Overview:</h5>
6068<p>
6069 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6070 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6071</p>
6072<h5>Arguments:</h5>
6073<p>
6074
6075 The intrinsic takes two arguments, the first a pointer to an integer value
6076 and the second an integer value. The result is also an integer value. These
6077 integer types can have any bit width, but they must all have the same bit
6078 width. The targets may only lower integer representations they support.
6079</p>
6080<h5>Semantics:</h5>
6081<p>
6082 This intrinsic does a series of operations atomically. It first loads the
6083 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6084 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6085</p>
6086
6087<h5>Examples:</h5>
6088<pre>
6089%ptr = malloc i32
6090 store i32 8, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6091%result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6092 <i>; yields {i32}:result1 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6093%result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6094 <i>; yields {i32}:result2 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6095%result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6096 <i>; yields {i32}:result3 = 2</i>
6097%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6098</pre>
6099</div>
6100
6101<!-- _______________________________________________________________________ -->
6102<div class="doc_subsubsection">
6103 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6104 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6105 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6106 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6107
6108</div>
6109<div class="doc_text">
6110<h5>Syntax:</h5>
6111<p>
6112 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6113 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6114 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6115 address spaces. Not all targets support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6116<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6117declare i8 @llvm.atomic.load.and.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6118declare i16 @llvm.atomic.load.and.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6119declare i32 @llvm.atomic.load.and.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6120declare 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]6121
6122</pre>
6123
6124<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6125declare i8 @llvm.atomic.load.or.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6126declare i16 @llvm.atomic.load.or.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6127declare i32 @llvm.atomic.load.or.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6128declare 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]6129
6130</pre>
6131
6132<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6133declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6134declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6135declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6136declare 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]6137
6138</pre>
6139
6140<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6141declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6142declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6143declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6144declare 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]6145
6146</pre>
6147<h5>Overview:</h5>
6148<p>
6149 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6150 the value stored in memory at <tt>ptr</tt>. It yields the original value
6151 at <tt>ptr</tt>.
6152</p>
6153<h5>Arguments:</h5>
6154<p>
6155
6156 These intrinsics take two arguments, the first a pointer to an integer value
6157 and the second an integer value. The result is also an integer value. These
6158 integer types can have any bit width, but they must all have the same bit
6159 width. The targets may only lower integer representations they support.
6160</p>
6161<h5>Semantics:</h5>
6162<p>
6163 These intrinsics does a series of operations atomically. They first load the
6164 value stored at <tt>ptr</tt>. They then do the bitwise operation
6165 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6166 value stored at <tt>ptr</tt>.
6167</p>
6168
6169<h5>Examples:</h5>
6170<pre>
6171%ptr = malloc i32
6172 store i32 0x0F0F, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6173%result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6174 <i>; yields {i32}:result0 = 0x0F0F</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6175%result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6176 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6177%result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6178 <i>; yields {i32}:result2 = 0xF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6179%result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6180 <i>; yields {i32}:result3 = FF</i>
6181%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6182</pre>
6183</div>
6184
6185
6186<!-- _______________________________________________________________________ -->
6187<div class="doc_subsubsection">
6188 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6189 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6190 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6191 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6192
6193</div>
6194<div class="doc_text">
6195<h5>Syntax:</h5>
6196<p>
6197 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6198 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6199 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6200 address spaces. Not all targets
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6201 support all bit widths however.</p>
6202<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6203declare i8 @llvm.atomic.load.max.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6204declare i16 @llvm.atomic.load.max.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6205declare i32 @llvm.atomic.load.max.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6206declare 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]6207
6208</pre>
6209
6210<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6211declare i8 @llvm.atomic.load.min.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6212declare i16 @llvm.atomic.load.min.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6213declare i32 @llvm.atomic.load.min.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6214declare 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]6215
6216</pre>
6217
6218<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6219declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6220declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6221declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6222declare 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]6223
6224</pre>
6225
6226<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6227declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6228declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6229declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6230declare 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]6231
6232</pre>
6233<h5>Overview:</h5>
6234<p>
6235 These intrinsics takes the signed or unsigned minimum or maximum of
6236 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6237 original value at <tt>ptr</tt>.
6238</p>
6239<h5>Arguments:</h5>
6240<p>
6241
6242 These intrinsics take two arguments, the first a pointer to an integer value
6243 and the second an integer value. The result is also an integer value. These
6244 integer types can have any bit width, but they must all have the same bit
6245 width. The targets may only lower integer representations they support.
6246</p>
6247<h5>Semantics:</h5>
6248<p>
6249 These intrinsics does a series of operations atomically. They first load the
6250 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6251 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6252 the original value stored at <tt>ptr</tt>.
6253</p>
6254
6255<h5>Examples:</h5>
6256<pre>
6257%ptr = malloc i32
6258 store i32 7, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6259%result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6260 <i>; yields {i32}:result0 = 7</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6261%result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6262 <i>; yields {i32}:result1 = -2</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6263%result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6264 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6265%result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6266 <i>; yields {i32}:result3 = 8</i>
6267%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6268</pre>
6269</div>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6270
6271<!-- ======================================================================= -->
6272<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6273 <a name="int_general">General Intrinsics</a>
6274</div>
6275
6276<div class="doc_text">
6277<p> This class of intrinsics is designed to be generic and has
6278no specific purpose. </p>
6279</div>
6280
6281<!-- _______________________________________________________________________ -->
6282<div class="doc_subsubsection">
6283 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6284</div>
6285
6286<div class="doc_text">
6287
6288<h5>Syntax:</h5>
6289<pre>
6290 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6291</pre>
6292
6293<h5>Overview:</h5>
6294
6295<p>
6296The '<tt>llvm.var.annotation</tt>' intrinsic
6297</p>
6298
6299<h5>Arguments:</h5>
6300
6301<p>
6302The first argument is a pointer to a value, the second is a pointer to a
6303global string, the third is a pointer to a global string which is the source
6304file name, and the last argument is the line number.
6305</p>
6306
6307<h5>Semantics:</h5>
6308
6309<p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6310This intrinsic allows annotation of local variables with arbitrary strings.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6311This can be useful for special purpose optimizations that want to look for these
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6312annotations. These have no other defined use, they are ignored by code
6313generation and optimization.
6314</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6315</div>
6316
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6317<!-- _______________________________________________________________________ -->
6318<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]6319 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6320</div>
6321
6322<div class="doc_text">
6323
6324<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6325<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6326any integer bit width.
6327</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6328<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]6329 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6330 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6331 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6332 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6333 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]6334</pre>
6335
6336<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6337
6338<p>
6339The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6340</p>
6341
6342<h5>Arguments:</h5>
6343
6344<p>
6345The first argument is an integer value (result of some expression),
6346the second is a pointer to a global string, the third is a pointer to a global
6347string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6348It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6349</p>
6350
6351<h5>Semantics:</h5>
6352
6353<p>
6354This intrinsic allows annotations to be put on arbitrary expressions
6355with arbitrary strings. This can be useful for special purpose optimizations
6356that want to look for these annotations. These have no other defined use, they
6357are ignored by code generation and optimization.
6358</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6359
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6360<!-- _______________________________________________________________________ -->
6361<div class="doc_subsubsection">
6362 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6363</div>
6364
6365<div class="doc_text">
6366
6367<h5>Syntax:</h5>
6368<pre>
6369 declare void @llvm.trap()
6370</pre>
6371
6372<h5>Overview:</h5>
6373
6374<p>
6375The '<tt>llvm.trap</tt>' intrinsic
6376</p>
6377
6378<h5>Arguments:</h5>
6379
6380<p>
6381None
6382</p>
6383
6384<h5>Semantics:</h5>
6385
6386<p>
6387This intrinsics is lowered to the target dependent trap instruction. If the
6388target does not have a trap instruction, this intrinsic will be lowered to the
6389call of the abort() function.
6390</p>
6391</div>
6392
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6393<!-- *********************************************************************** -->
6394<hr>
6395<address>
6396 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
6397 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
6398 <a href="http://validator.w3.org/check/referer"><img
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]6399 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6400
6401 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
6402 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
6403 Last modified: $Date$
6404</address>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]6405
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6406</body>
6407</html>