Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 008cd3ecea2958540d8d8f5370deffb6103c6758 / . / docs / LangRef.html
blob: 24f08eb92a17b4eb7699cd6471b1994fc8431b4d [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>,
755an optional <a href="#notes">function notes</a>, an
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]756opening curly brace, a list of basic blocks, and a closing curly brace.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]757
758LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
759optional <a href="#linkage">linkage type</a>, an optional
760<a href="#visibility">visibility style</a>, an optional
761<a href="#callingconv">calling convention</a>, a return type, an optional
762<a href="#paramattrs">parameter attribute</a> for the return type, a function
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]763name, a possibly empty list of arguments, an optional alignment, and an optional
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]764<a href="#gc">garbage collector name</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]765
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]766<p>A function definition contains a list of basic blocks, forming the CFG
767(Control Flow Graph) for
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]768the function. Each basic block may optionally start with a label (giving the
769basic block a symbol table entry), contains a list of instructions, and ends
770with a <a href="#terminators">terminator</a> instruction (such as a branch or
771function return).</p>
772
773<p>The first basic block in a function is special in two ways: it is immediately
774executed on entrance to the function, and it is not allowed to have predecessor
775basic blocks (i.e. there can not be any branches to the entry block of a
776function). Because the block can have no predecessors, it also cannot have any
777<a href="#i_phi">PHI nodes</a>.</p>
778
779<p>LLVM allows an explicit section to be specified for functions. If the target
780supports it, it will emit functions to the section specified.</p>
781
782<p>An explicit alignment may be specified for a function. If not present, or if
783the alignment is set to zero, the alignment of the function is set by the target
784to whatever it feels convenient. If an explicit alignment is specified, the
785function is forced to have at least that much alignment. All alignments must be
786a power of 2.</p>
787
788</div>
789
790
791<!-- ======================================================================= -->
792<div class="doc_subsection">
793 <a name="aliasstructure">Aliases</a>
794</div>
795<div class="doc_text">
796 <p>Aliases act as "second name" for the aliasee value (which can be either
Anton Korobeynikov96822812008-03-22 08:36:14 +0000[diff] [blame]797 function, global variable, another alias or bitcast of global value). Aliases
798 may have an optional <a href="#linkage">linkage type</a>, and an
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]799 optional <a href="#visibility">visibility style</a>.</p>
800
801 <h5>Syntax:</h5>
802
803<div class="doc_code">
804<pre>
Duncan Sandsd7bfabf2008-09-12 20:48:21 +0000[diff] [blame]805@&lt;Name&gt; = alias [Linkage] [Visibility] &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]806</pre>
807</div>
808
809</div>
810
811
812
813<!-- ======================================================================= -->
814<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
815<div class="doc_text">
816 <p>The return type and each parameter of a function type may have a set of
817 <i>parameter attributes</i> associated with them. Parameter attributes are
818 used to communicate additional information about the result or parameters of
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]819 a function. Parameter attributes are considered to be part of the function,
820 not of the function type, so functions with different parameter attributes
821 can have the same function type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]822
823 <p>Parameter attributes are simple keywords that follow the type specified. If
824 multiple parameter attributes are needed, they are space separated. For
825 example:</p>
826
827<div class="doc_code">
828<pre>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame^]829declare i32 @printf(i8* noalias , ...)
830declare i32 @atoi(i8 zeroext*)
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>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]840 <dd>This indicates that the parameter should be zero extended just before
841 a call to this function.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]842
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]843 <dt><tt>signext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]844 <dd>This indicates that the parameter should be sign extended just before
845 a call to this function.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]846
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]847 <dt><tt>inreg</tt></dt>
Dale Johannesenc08a0e22008-09-25 20:47:45 +0000[diff] [blame]848 <dd>This indicates that this parameter or return value should be treated
849 in a special target-dependent fashion during while emitting code for a
850 function call or return (usually, by putting it in a register as opposed
851 to memory; in some places it is used to distinguish between two different
852 kinds of registers). Use of this attribute is target-specific</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]853
854 <dt><tt>byval</tt></dt>
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]855 <dd>This indicates that the pointer parameter should really be passed by
856 value to the function. The attribute implies that a hidden copy of the
857 pointee is made between the caller and the callee, so the callee is unable
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]858 to modify the value in the callee. This attribute is only valid on LLVM
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]859 pointer arguments. It is generally used to pass structs and arrays by
860 value, but is also valid on scalars (even though this is silly).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]861
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]862 <dt><tt>sret</tt></dt>
Duncan Sands616cc032008-02-18 04:19:38 +0000[diff] [blame]863 <dd>This indicates that the pointer parameter specifies the address of a
864 structure that is the return value of the function in the source program.
Duncan Sandsef0e9e42008-03-17 12:17:41 +0000[diff] [blame]865 Loads and stores to the structure are assumed not to trap.
Duncan Sands616cc032008-02-18 04:19:38 +0000[diff] [blame]866 May only be applied to the first parameter.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]867
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]868 <dt><tt>noalias</tt></dt>
Owen Andersonc4fc4cd2008-02-18 04:09:01 +0000[diff] [blame]869 <dd>This indicates that the parameter does not alias any global or any other
870 parameter. The caller is responsible for ensuring that this is the case,
871 usually by placing the value in a stack allocation.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]872
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]873 <dt><tt>nest</tt></dt>
Duncan Sandsf1a7d4c2008-07-08 09:27:25 +0000[diff] [blame]874 <dd>This indicates that the pointer parameter can be excised using the
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]875 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]876 </dl>
877
878</div>
879
880<!-- ======================================================================= -->
881<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]882 <a name="gc">Garbage Collector Names</a>
883</div>
884
885<div class="doc_text">
886<p>Each function may specify a garbage collector name, which is simply a
887string.</p>
888
889<div class="doc_code"><pre
890>define void @f() gc "name" { ...</pre></div>
891
892<p>The compiler declares the supported values of <i>name</i>. Specifying a
893collector which will cause the compiler to alter its output in order to support
894the named garbage collection algorithm.</p>
895</div>
896
897<!-- ======================================================================= -->
898<div class="doc_subsection">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame^]899 <a name="fnattrs">Function Attributes</a>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]900</div>
901
902<div class="doc_text">
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame^]903
904<p>Function attributes are set to communicate additional information about
905 a function. Function attributes are considered to be part of the function,
906 not of the function type, so functions with different parameter attributes
907 can have the same function type.</p>
908
909 <p>Function attributes are simple keywords that follow the type specified. If
910 multiple attributes are needed, they are space separated. For
911 example:</p>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]912
913<div class="doc_code">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]914<pre>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame^]915define void @f() noinline { ... }
916define void @f() alwaysinline { ... }
917define void @f() alwaysinline optsize { ... }
918define void @f() optsize
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]919</pre>
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]920</div>
921
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]922<dl>
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame^]923<dt><tt>alwaysinline</tt></dt>
924<dd>This attribute requests inliner to inline this function irrespective of
925inlining size threshold for this function.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]926
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame^]927<dt><tt>noinline</tt></dt>
928<dd>This attributes requests inliner to never inline this function in any
929situation. This attribute may not be used together with <tt>alwaysinline</tt>
930 attribute.</dd>
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]931
Devang Patel008cd3e2008-09-26 23:51:19 +0000[diff] [blame^]932<dt><tt>optsize</tt></dt>
933<dd>This attribute suggests optimization passes and code generator passes to
934make choices that help reduce code size.</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>noreturn</tt></dt>
937<dd>This function attribute indicates that the function never returns. This
938 indicates to LLVM that every call to this function should be treated as if
939 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
940
941<dt><tt>nounwind</tt></dt>
942<dd>This function attribute indicates that no exceptions unwind out of the
943 function. Usually this is because the function makes no use of exceptions,
944 but it may also be that the function catches any exceptions thrown when
945 executing it.</dd>
946
947<dt><tt>readonly</tt></dt>
948<dd>This function attribute indicates that the function has no side-effects
949 except for producing a return value or throwing an exception. The value
950 returned must only depend on the function arguments and/or global variables.
951 It may use values obtained by dereferencing pointers.</dd>
952<dt><tt>readnone</tt></dt>
953<dd>A <tt>readnone</tt> function has the same restrictions as a <tt>readonly</tt>
954 function, but in addition it is not allowed to dereference any pointer arguments
955 or global variables.
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]956</dl>
957
Devang Pateld468f1c2008-09-04 23:05:13 +0000[diff] [blame]958</div>
959
960<!-- ======================================================================= -->
961<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]962 <a name="moduleasm">Module-Level Inline Assembly</a>
963</div>
964
965<div class="doc_text">
966<p>
967Modules may contain "module-level inline asm" blocks, which corresponds to the
968GCC "file scope inline asm" blocks. These blocks are internally concatenated by
969LLVM and treated as a single unit, but may be separated in the .ll file if
970desired. The syntax is very simple:
971</p>
972
973<div class="doc_code">
974<pre>
975module asm "inline asm code goes here"
976module asm "more can go here"
977</pre>
978</div>
979
980<p>The strings can contain any character by escaping non-printable characters.
981 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
982 for the number.
983</p>
984
985<p>
986 The inline asm code is simply printed to the machine code .s file when
987 assembly code is generated.
988</p>
989</div>
990
991<!-- ======================================================================= -->
992<div class="doc_subsection">
993 <a name="datalayout">Data Layout</a>
994</div>
995
996<div class="doc_text">
997<p>A module may specify a target specific data layout string that specifies how
998data is to be laid out in memory. The syntax for the data layout is simply:</p>
999<pre> target datalayout = "<i>layout specification</i>"</pre>
1000<p>The <i>layout specification</i> consists of a list of specifications
1001separated by the minus sign character ('-'). Each specification starts with a
1002letter and may include other information after the letter to define some
1003aspect of the data layout. The specifications accepted are as follows: </p>
1004<dl>
1005 <dt><tt>E</tt></dt>
1006 <dd>Specifies that the target lays out data in big-endian form. That is, the
1007 bits with the most significance have the lowest address location.</dd>
1008 <dt><tt>e</tt></dt>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1009 <dd>Specifies that the target lays out data in little-endian form. That is,
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1010 the bits with the least significance have the lowest address location.</dd>
1011 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1012 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1013 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1014 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1015 too.</dd>
1016 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1017 <dd>This specifies the alignment for an integer type of a given bit
1018 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1019 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1020 <dd>This specifies the alignment for a vector type of a given bit
1021 <i>size</i>.</dd>
1022 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1023 <dd>This specifies the alignment for a floating point type of a given bit
1024 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1025 (double).</dd>
1026 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1027 <dd>This specifies the alignment for an aggregate type of a given bit
1028 <i>size</i>.</dd>
1029</dl>
1030<p>When constructing the data layout for a given target, LLVM starts with a
1031default set of specifications which are then (possibly) overriden by the
1032specifications in the <tt>datalayout</tt> keyword. The default specifications
1033are given in this list:</p>
1034<ul>
1035 <li><tt>E</tt> - big endian</li>
1036 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1037 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1038 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1039 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1040 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1041 <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]1042 alignment of 64-bits</li>
1043 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1044 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1045 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1046 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1047 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1048</ul>
Chris Lattner6a9f3c42008-08-05 18:21:08 +0000[diff] [blame]1049<p>When LLVM is determining the alignment for a given type, it uses the
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1050following rules:
1051<ol>
1052 <li>If the type sought is an exact match for one of the specifications, that
1053 specification is used.</li>
1054 <li>If no match is found, and the type sought is an integer type, then the
1055 smallest integer type that is larger than the bitwidth of the sought type is
1056 used. If none of the specifications are larger than the bitwidth then the the
1057 largest integer type is used. For example, given the default specifications
1058 above, the i7 type will use the alignment of i8 (next largest) while both
1059 i65 and i256 will use the alignment of i64 (largest specified).</li>
1060 <li>If no match is found, and the type sought is a vector type, then the
1061 largest vector type that is smaller than the sought vector type will be used
1062 as a fall back. This happens because <128 x double> can be implemented in
1063 terms of 64 <2 x double>, for example.</li>
1064</ol>
1065</div>
1066
1067<!-- *********************************************************************** -->
1068<div class="doc_section"> <a name="typesystem">Type System</a> </div>
1069<!-- *********************************************************************** -->
1070
1071<div class="doc_text">
1072
1073<p>The LLVM type system is one of the most important features of the
1074intermediate representation. Being typed enables a number of
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]1075optimizations to be performed on the intermediate representation directly,
1076without having to do
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1077extra analyses on the side before the transformation. A strong type
1078system makes it easier to read the generated code and enables novel
1079analyses and transformations that are not feasible to perform on normal
1080three address code representations.</p>
1081
1082</div>
1083
1084<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1085<div class="doc_subsection"> <a name="t_classifications">Type
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1086Classifications</a> </div>
1087<div class="doc_text">
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1088<p>The types fall into a few useful
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1089classifications:</p>
1090
1091<table border="1" cellspacing="0" cellpadding="4">
1092 <tbody>
1093 <tr><th>Classification</th><th>Types</th></tr>
1094 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1095 <td><a href="#t_integer">integer</a></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1096 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1097 </tr>
1098 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1099 <td><a href="#t_floating">floating point</a></td>
1100 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1101 </tr>
1102 <tr>
1103 <td><a name="t_firstclass">first class</a></td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1104 <td><a href="#t_integer">integer</a>,
1105 <a href="#t_floating">floating point</a>,
1106 <a href="#t_pointer">pointer</a>,
Dan Gohmanf6237db2008-06-18 18:42:13 +0000[diff] [blame]1107 <a href="#t_vector">vector</a>,
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]1108 <a href="#t_struct">structure</a>,
1109 <a href="#t_array">array</a>,
Dan Gohmand13951e2008-05-23 22:50:26 +0000[diff] [blame]1110 <a href="#t_label">label</a>.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1111 </td>
1112 </tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1113 <tr>
1114 <td><a href="#t_primitive">primitive</a></td>
1115 <td><a href="#t_label">label</a>,
1116 <a href="#t_void">void</a>,
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1117 <a href="#t_floating">floating point</a>.</td>
1118 </tr>
1119 <tr>
1120 <td><a href="#t_derived">derived</a></td>
1121 <td><a href="#t_integer">integer</a>,
1122 <a href="#t_array">array</a>,
1123 <a href="#t_function">function</a>,
1124 <a href="#t_pointer">pointer</a>,
1125 <a href="#t_struct">structure</a>,
1126 <a href="#t_pstruct">packed structure</a>,
1127 <a href="#t_vector">vector</a>,
1128 <a href="#t_opaque">opaque</a>.
1129 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1130 </tbody>
1131</table>
1132
1133<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1134most important. Values of these types are the only ones which can be
1135produced by instructions, passed as arguments, or used as operands to
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]1136instructions.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1137</div>
1138
1139<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1140<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1141
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1142<div class="doc_text">
1143<p>The primitive types are the fundamental building blocks of the LLVM
1144system.</p>
1145
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1146</div>
1147
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1148<!-- _______________________________________________________________________ -->
1149<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1150
1151<div class="doc_text">
1152 <table>
1153 <tbody>
1154 <tr><th>Type</th><th>Description</th></tr>
1155 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1156 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1157 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1158 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1159 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1160 </tbody>
1161 </table>
1162</div>
1163
1164<!-- _______________________________________________________________________ -->
1165<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1166
1167<div class="doc_text">
1168<h5>Overview:</h5>
1169<p>The void type does not represent any value and has no size.</p>
1170
1171<h5>Syntax:</h5>
1172
1173<pre>
1174 void
1175</pre>
1176</div>
1177
1178<!-- _______________________________________________________________________ -->
1179<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1180
1181<div class="doc_text">
1182<h5>Overview:</h5>
1183<p>The label type represents code labels.</p>
1184
1185<h5>Syntax:</h5>
1186
1187<pre>
1188 label
1189</pre>
1190</div>
1191
1192
1193<!-- ======================================================================= -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1194<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1195
1196<div class="doc_text">
1197
1198<p>The real power in LLVM comes from the derived types in the system.
1199This is what allows a programmer to represent arrays, functions,
1200pointers, and other useful types. Note that these derived types may be
1201recursive: For example, it is possible to have a two dimensional array.</p>
1202
1203</div>
1204
1205<!-- _______________________________________________________________________ -->
1206<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1207
1208<div class="doc_text">
1209
1210<h5>Overview:</h5>
1211<p>The integer type is a very simple derived type that simply specifies an
1212arbitrary bit width for the integer type desired. Any bit width from 1 bit to
12132^23-1 (about 8 million) can be specified.</p>
1214
1215<h5>Syntax:</h5>
1216
1217<pre>
1218 iN
1219</pre>
1220
1221<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1222value.</p>
1223
1224<h5>Examples:</h5>
1225<table class="layout">
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1226 <tbody>
1227 <tr>
1228 <td><tt>i1</tt></td>
1229 <td>a single-bit integer.</td>
1230 </tr><tr>
1231 <td><tt>i32</tt></td>
1232 <td>a 32-bit integer.</td>
1233 </tr><tr>
1234 <td><tt>i1942652</tt></td>
1235 <td>a really big integer of over 1 million bits.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1236 </tr>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1237 </tbody>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1238</table>
1239</div>
1240
1241<!-- _______________________________________________________________________ -->
1242<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1243
1244<div class="doc_text">
1245
1246<h5>Overview:</h5>
1247
1248<p>The array type is a very simple derived type that arranges elements
1249sequentially in memory. The array type requires a size (number of
1250elements) and an underlying data type.</p>
1251
1252<h5>Syntax:</h5>
1253
1254<pre>
1255 [&lt;# elements&gt; x &lt;elementtype&gt;]
1256</pre>
1257
1258<p>The number of elements is a constant integer value; elementtype may
1259be any type with a size.</p>
1260
1261<h5>Examples:</h5>
1262<table class="layout">
1263 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1264 <td class="left"><tt>[40 x i32]</tt></td>
1265 <td class="left">Array of 40 32-bit integer values.</td>
1266 </tr>
1267 <tr class="layout">
1268 <td class="left"><tt>[41 x i32]</tt></td>
1269 <td class="left">Array of 41 32-bit integer values.</td>
1270 </tr>
1271 <tr class="layout">
1272 <td class="left"><tt>[4 x i8]</tt></td>
1273 <td class="left">Array of 4 8-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1274 </tr>
1275</table>
1276<p>Here are some examples of multidimensional arrays:</p>
1277<table class="layout">
1278 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1279 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1280 <td class="left">3x4 array of 32-bit integer values.</td>
1281 </tr>
1282 <tr class="layout">
1283 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1284 <td class="left">12x10 array of single precision floating point values.</td>
1285 </tr>
1286 <tr class="layout">
1287 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1288 <td class="left">2x3x4 array of 16-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1289 </tr>
1290</table>
1291
1292<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1293length array. Normally, accesses past the end of an array are undefined in
1294LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1295As a special case, however, zero length arrays are recognized to be variable
1296length. This allows implementation of 'pascal style arrays' with the LLVM
1297type "{ i32, [0 x float]}", for example.</p>
1298
1299</div>
1300
1301<!-- _______________________________________________________________________ -->
1302<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1303<div class="doc_text">
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1304
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1305<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1306
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1307<p>The function type can be thought of as a function signature. It
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1308consists of a return type and a list of formal parameter types. The
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1309return 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]1310If 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]1311class types, and the struct must have at least one element.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1312
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1313<h5>Syntax:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1314
1315<pre>
1316 &lt;returntype list&gt; (&lt;parameter list&gt;)
1317</pre>
1318
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1319<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1320specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1321which indicates that the function takes a variable number of arguments.
1322Variable argument functions can access their arguments with the <a
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1323 href="#int_varargs">variable argument handling intrinsic</a> functions.
1324'<tt>&lt;returntype list&gt;</tt>' is a comma-separated list of
1325<a href="#t_firstclass">first class</a> type specifiers.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1326
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1327<h5>Examples:</h5>
1328<table class="layout">
1329 <tr class="layout">
1330 <td class="left"><tt>i32 (i32)</tt></td>
1331 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1332 </td>
1333 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1334 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1335 </tt></td>
1336 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1337 an <tt>i16</tt> that should be sign extended and a
1338 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1339 <tt>float</tt>.
1340 </td>
1341 </tr><tr class="layout">
1342 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1343 <td class="left">A vararg function that takes at least one
1344 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1345 which returns an integer. This is the signature for <tt>printf</tt> in
1346 LLVM.
1347 </td>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1348 </tr><tr class="layout">
1349 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Devang Patel6dddba22008-03-24 18:10:52 +0000[diff] [blame]1350 <td class="left">A function taking an <tt>i32></tt>, returning two
1351 <tt> i32 </tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1352 </td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1353 </tr>
1354</table>
1355
1356</div>
1357<!-- _______________________________________________________________________ -->
1358<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1359<div class="doc_text">
1360<h5>Overview:</h5>
1361<p>The structure type is used to represent a collection of data members
1362together in memory. The packing of the field types is defined to match
1363the ABI of the underlying processor. The elements of a structure may
1364be any type that has a size.</p>
1365<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1366and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1367field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1368instruction.</p>
1369<h5>Syntax:</h5>
1370<pre> { &lt;type list&gt; }<br></pre>
1371<h5>Examples:</h5>
1372<table class="layout">
1373 <tr class="layout">
1374 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1375 <td class="left">A triple of three <tt>i32</tt> values</td>
1376 </tr><tr class="layout">
1377 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1378 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1379 second element is a <a href="#t_pointer">pointer</a> to a
1380 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1381 an <tt>i32</tt>.</td>
1382 </tr>
1383</table>
1384</div>
1385
1386<!-- _______________________________________________________________________ -->
1387<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1388</div>
1389<div class="doc_text">
1390<h5>Overview:</h5>
1391<p>The packed structure type is used to represent a collection of data members
1392together in memory. There is no padding between fields. Further, the alignment
1393of a packed structure is 1 byte. The elements of a packed structure may
1394be any type that has a size.</p>
1395<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1396and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1397field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1398instruction.</p>
1399<h5>Syntax:</h5>
1400<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1401<h5>Examples:</h5>
1402<table class="layout">
1403 <tr class="layout">
1404 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1405 <td class="left">A triple of three <tt>i32</tt> values</td>
1406 </tr><tr class="layout">
Bill Wendling74d3eac2008-09-07 10:26:33 +0000[diff] [blame]1407 <td class="left">
1408<tt>&lt;&nbsp;{&nbsp;float,&nbsp;i32&nbsp;(i32)*&nbsp;}&nbsp;&gt;</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1409 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1410 second element is a <a href="#t_pointer">pointer</a> to a
1411 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1412 an <tt>i32</tt>.</td>
1413 </tr>
1414</table>
1415</div>
1416
1417<!-- _______________________________________________________________________ -->
1418<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1419<div class="doc_text">
1420<h5>Overview:</h5>
1421<p>As in many languages, the pointer type represents a pointer or
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1422reference to another object, which must live in memory. Pointer types may have
1423an optional address space attribute defining the target-specific numbered
1424address space where the pointed-to object resides. The default address space is
1425zero.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1426<h5>Syntax:</h5>
1427<pre> &lt;type&gt; *<br></pre>
1428<h5>Examples:</h5>
1429<table class="layout">
1430 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1431 <td class="left"><tt>[4x i32]*</tt></td>
1432 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1433 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1434 </tr>
1435 <tr class="layout">
1436 <td class="left"><tt>i32 (i32 *) *</tt></td>
1437 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1438 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1439 <tt>i32</tt>.</td>
1440 </tr>
1441 <tr class="layout">
1442 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1443 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1444 that resides in address space #5.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1445 </tr>
1446</table>
1447</div>
1448
1449<!-- _______________________________________________________________________ -->
1450<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1451<div class="doc_text">
1452
1453<h5>Overview:</h5>
1454
1455<p>A vector type is a simple derived type that represents a vector
1456of elements. Vector types are used when multiple primitive data
1457are operated in parallel using a single instruction (SIMD).
1458A vector type requires a size (number of
1459elements) and an underlying primitive data type. Vectors must have a power
1460of two length (1, 2, 4, 8, 16 ...). Vector types are
1461considered <a href="#t_firstclass">first class</a>.</p>
1462
1463<h5>Syntax:</h5>
1464
1465<pre>
1466 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1467</pre>
1468
1469<p>The number of elements is a constant integer value; elementtype may
1470be any integer or floating point type.</p>
1471
1472<h5>Examples:</h5>
1473
1474<table class="layout">
1475 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1476 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1477 <td class="left">Vector of 4 32-bit integer values.</td>
1478 </tr>
1479 <tr class="layout">
1480 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1481 <td class="left">Vector of 8 32-bit floating-point values.</td>
1482 </tr>
1483 <tr class="layout">
1484 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1485 <td class="left">Vector of 2 64-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1486 </tr>
1487</table>
1488</div>
1489
1490<!-- _______________________________________________________________________ -->
1491<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1492<div class="doc_text">
1493
1494<h5>Overview:</h5>
1495
1496<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1497corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1498In LLVM, opaque types can eventually be resolved to any type (not just a
1499structure type).</p>
1500
1501<h5>Syntax:</h5>
1502
1503<pre>
1504 opaque
1505</pre>
1506
1507<h5>Examples:</h5>
1508
1509<table class="layout">
1510 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1511 <td class="left"><tt>opaque</tt></td>
1512 <td class="left">An opaque type.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1513 </tr>
1514</table>
1515</div>
1516
1517
1518<!-- *********************************************************************** -->
1519<div class="doc_section"> <a name="constants">Constants</a> </div>
1520<!-- *********************************************************************** -->
1521
1522<div class="doc_text">
1523
1524<p>LLVM has several different basic types of constants. This section describes
1525them all and their syntax.</p>
1526
1527</div>
1528
1529<!-- ======================================================================= -->
1530<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1531
1532<div class="doc_text">
1533
1534<dl>
1535 <dt><b>Boolean constants</b></dt>
1536
1537 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1538 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1539 </dd>
1540
1541 <dt><b>Integer constants</b></dt>
1542
1543 <dd>Standard integers (such as '4') are constants of the <a
1544 href="#t_integer">integer</a> type. Negative numbers may be used with
1545 integer types.
1546 </dd>
1547
1548 <dt><b>Floating point constants</b></dt>
1549
1550 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1551 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]1552 notation (see below). The assembler requires the exact decimal value of
1553 a floating-point constant. For example, the assembler accepts 1.25 but
1554 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1555 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1556
1557 <dt><b>Null pointer constants</b></dt>
1558
1559 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1560 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1561
1562</dl>
1563
1564<p>The one non-intuitive notation for constants is the optional hexadecimal form
1565of floating point constants. For example, the form '<tt>double
15660x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
15674.5e+15</tt>'. The only time hexadecimal floating point constants are required
1568(and the only time that they are generated by the disassembler) is when a
1569floating point constant must be emitted but it cannot be represented as a
1570decimal floating point number. For example, NaN's, infinities, and other
1571special values are represented in their IEEE hexadecimal format so that
1572assembly and disassembly do not cause any bits to change in the constants.</p>
1573
1574</div>
1575
1576<!-- ======================================================================= -->
1577<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1578</div>
1579
1580<div class="doc_text">
1581<p>Aggregate constants arise from aggregation of simple constants
1582and smaller aggregate constants.</p>
1583
1584<dl>
1585 <dt><b>Structure constants</b></dt>
1586
1587 <dd>Structure constants are represented with notation similar to structure
1588 type definitions (a comma separated list of elements, surrounded by braces
Chris Lattner6c8de962007-12-25 20:34:52 +0000[diff] [blame]1589 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1590 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]1591 must have <a href="#t_struct">structure type</a>, and the number and
1592 types of elements must match those specified by the type.
1593 </dd>
1594
1595 <dt><b>Array constants</b></dt>
1596
1597 <dd>Array constants are represented with notation similar to array type
1598 definitions (a comma separated list of elements, surrounded by square brackets
1599 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1600 constants must have <a href="#t_array">array type</a>, and the number and
1601 types of elements must match those specified by the type.
1602 </dd>
1603
1604 <dt><b>Vector constants</b></dt>
1605
1606 <dd>Vector constants are represented with notation similar to vector type
1607 definitions (a comma separated list of elements, surrounded by
1608 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1609 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1610 href="#t_vector">vector type</a>, and the number and types of elements must
1611 match those specified by the type.
1612 </dd>
1613
1614 <dt><b>Zero initialization</b></dt>
1615
1616 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1617 value to zero of <em>any</em> type, including scalar and aggregate types.
1618 This is often used to avoid having to print large zero initializers (e.g. for
1619 large arrays) and is always exactly equivalent to using explicit zero
1620 initializers.
1621 </dd>
1622</dl>
1623
1624</div>
1625
1626<!-- ======================================================================= -->
1627<div class="doc_subsection">
1628 <a name="globalconstants">Global Variable and Function Addresses</a>
1629</div>
1630
1631<div class="doc_text">
1632
1633<p>The addresses of <a href="#globalvars">global variables</a> and <a
1634href="#functionstructure">functions</a> are always implicitly valid (link-time)
1635constants. These constants are explicitly referenced when the <a
1636href="#identifiers">identifier for the global</a> is used and always have <a
1637href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1638file:</p>
1639
1640<div class="doc_code">
1641<pre>
1642@X = global i32 17
1643@Y = global i32 42
1644@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1645</pre>
1646</div>
1647
1648</div>
1649
1650<!-- ======================================================================= -->
1651<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1652<div class="doc_text">
1653 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1654 no specific value. Undefined values may be of any type and be used anywhere
1655 a constant is permitted.</p>
1656
1657 <p>Undefined values indicate to the compiler that the program is well defined
1658 no matter what value is used, giving the compiler more freedom to optimize.
1659 </p>
1660</div>
1661
1662<!-- ======================================================================= -->
1663<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1664</div>
1665
1666<div class="doc_text">
1667
1668<p>Constant expressions are used to allow expressions involving other constants
1669to be used as constants. Constant expressions may be of any <a
1670href="#t_firstclass">first class</a> type and may involve any LLVM operation
1671that does not have side effects (e.g. load and call are not supported). The
1672following is the syntax for constant expressions:</p>
1673
1674<dl>
1675 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1676 <dd>Truncate a constant to another type. The bit size of CST must be larger
1677 than the bit size of TYPE. Both types must be integers.</dd>
1678
1679 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1680 <dd>Zero extend a constant to another type. The bit size of CST must be
1681 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1682
1683 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1684 <dd>Sign extend a constant to another type. The bit size of CST must be
1685 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1686
1687 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1688 <dd>Truncate a floating point constant to another floating point type. The
1689 size of CST must be larger than the size of TYPE. Both types must be
1690 floating point.</dd>
1691
1692 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1693 <dd>Floating point extend a constant to another type. The size of CST must be
1694 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1695
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1696 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1697 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1698 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1699 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1700 of the same number of elements. If the value won't fit in the integer type,
1701 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1702
1703 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1704 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1705 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1706 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1707 of the same number of elements. If the value won't fit in the integer type,
1708 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1709
1710 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1711 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1712 constant. TYPE must be a scalar or vector floating point type. CST must be of
1713 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1714 of the same number of elements. If the value won't fit in the floating point
1715 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1716
1717 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1718 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1719 constant. TYPE must be a scalar or vector floating point type. CST must be of
1720 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1721 of the same number of elements. If the value won't fit in the floating point
1722 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1723
1724 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1725 <dd>Convert a pointer typed constant to the corresponding integer constant
1726 TYPE must be an integer type. CST must be of pointer type. The CST value is
1727 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1728
1729 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1730 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1731 pointer type. CST must be of integer type. The CST value is zero extended,
1732 truncated, or unchanged to make it fit in a pointer size. This one is
1733 <i>really</i> dangerous!</dd>
1734
1735 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1736 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1737 identical (same number of bits). The conversion is done as if the CST value
1738 was stored to memory and read back as TYPE. In other words, no bits change
1739 with this operator, just the type. This can be used for conversion of
1740 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]1741 pointers it is only valid to cast to another pointer type. It is not valid
1742 to bitcast to or from an aggregate type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1743 </dd>
1744
1745 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1746
1747 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1748 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1749 instruction, the index list may have zero or more indexes, which are required
1750 to make sense for the type of "CSTPTR".</dd>
1751
1752 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1753
1754 <dd>Perform the <a href="#i_select">select operation</a> on
1755 constants.</dd>
1756
1757 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1758 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1759
1760 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1761 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1762
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]1763 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
1764 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
1765
1766 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
1767 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
1768
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1769 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1770
1771 <dd>Perform the <a href="#i_extractelement">extractelement
1772 operation</a> on constants.
1773
1774 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1775
1776 <dd>Perform the <a href="#i_insertelement">insertelement
1777 operation</a> on constants.</dd>
1778
1779
1780 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1781
1782 <dd>Perform the <a href="#i_shufflevector">shufflevector
1783 operation</a> on constants.</dd>
1784
1785 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1786
1787 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1788 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1789 binary</a> operations. The constraints on operands are the same as those for
1790 the corresponding instruction (e.g. no bitwise operations on floating point
1791 values are allowed).</dd>
1792</dl>
1793</div>
1794
1795<!-- *********************************************************************** -->
1796<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1797<!-- *********************************************************************** -->
1798
1799<!-- ======================================================================= -->
1800<div class="doc_subsection">
1801<a name="inlineasm">Inline Assembler Expressions</a>
1802</div>
1803
1804<div class="doc_text">
1805
1806<p>
1807LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1808Module-Level Inline Assembly</a>) through the use of a special value. This
1809value represents the inline assembler as a string (containing the instructions
1810to emit), a list of operand constraints (stored as a string), and a flag that
1811indicates whether or not the inline asm expression has side effects. An example
1812inline assembler expression is:
1813</p>
1814
1815<div class="doc_code">
1816<pre>
1817i32 (i32) asm "bswap $0", "=r,r"
1818</pre>
1819</div>
1820
1821<p>
1822Inline assembler expressions may <b>only</b> be used as the callee operand of
1823a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1824</p>
1825
1826<div class="doc_code">
1827<pre>
1828%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1829</pre>
1830</div>
1831
1832<p>
1833Inline asms with side effects not visible in the constraint list must be marked
1834as having side effects. This is done through the use of the
1835'<tt>sideeffect</tt>' keyword, like so:
1836</p>
1837
1838<div class="doc_code">
1839<pre>
1840call void asm sideeffect "eieio", ""()
1841</pre>
1842</div>
1843
1844<p>TODO: The format of the asm and constraints string still need to be
1845documented here. Constraints on what can be done (e.g. duplication, moving, etc
1846need to be documented).
1847</p>
1848
1849</div>
1850
1851<!-- *********************************************************************** -->
1852<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1853<!-- *********************************************************************** -->
1854
1855<div class="doc_text">
1856
1857<p>The LLVM instruction set consists of several different
1858classifications of instructions: <a href="#terminators">terminator
1859instructions</a>, <a href="#binaryops">binary instructions</a>,
1860<a href="#bitwiseops">bitwise binary instructions</a>, <a
1861 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1862instructions</a>.</p>
1863
1864</div>
1865
1866<!-- ======================================================================= -->
1867<div class="doc_subsection"> <a name="terminators">Terminator
1868Instructions</a> </div>
1869
1870<div class="doc_text">
1871
1872<p>As mentioned <a href="#functionstructure">previously</a>, every
1873basic block in a program ends with a "Terminator" instruction, which
1874indicates which block should be executed after the current block is
1875finished. These terminator instructions typically yield a '<tt>void</tt>'
1876value: they produce control flow, not values (the one exception being
1877the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1878<p>There are six different terminator instructions: the '<a
1879 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1880instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1881the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1882 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1883 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1884
1885</div>
1886
1887<!-- _______________________________________________________________________ -->
1888<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1889Instruction</a> </div>
1890<div class="doc_text">
1891<h5>Syntax:</h5>
1892<pre> ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
1893 ret void <i>; Return from void function</i>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1894 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]1895</pre>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1896
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1897<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1898
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1899<p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1900value) from a function back to the caller.</p>
1901<p>There are two forms of the '<tt>ret</tt>' instruction: one that
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1902returns value(s) and then causes control flow, and one that just causes
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1903control flow to occur.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1904
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1905<h5>Arguments:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1906
1907<p>The '<tt>ret</tt>' instruction may return zero, one or multiple values.
1908The type of each return value must be a '<a href="#t_firstclass">first
1909class</a>' type. Note that a function is not <a href="#wellformed">well
1910formed</a> if there exists a '<tt>ret</tt>' instruction inside of the
1911function that returns values that do not match the return type of the
1912function.</p>
1913
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1914<h5>Semantics:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1915
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1916<p>When the '<tt>ret</tt>' instruction is executed, control flow
1917returns back to the calling function's context. If the caller is a "<a
1918 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1919the instruction after the call. If the caller was an "<a
1920 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1921at the beginning of the "normal" destination block. If the instruction
1922returns a value, that value shall set the call or invoke instruction's
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1923return value. If the instruction returns multiple values then these
Devang Patelec8a5b02008-03-11 05:51:59 +0000[diff] [blame]1924values can only be accessed through a '<a href="#i_getresult"><tt>getresult</tt>
1925</a>' instruction.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1926
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1927<h5>Example:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1928
1929<pre>
1930 ret i32 5 <i>; Return an integer value of 5</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1931 ret void <i>; Return from a void function</i>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1932 ret i32 4, i8 2 <i>; Return two values 4 and 2 </i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1933</pre>
1934</div>
1935<!-- _______________________________________________________________________ -->
1936<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1937<div class="doc_text">
1938<h5>Syntax:</h5>
1939<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
1940</pre>
1941<h5>Overview:</h5>
1942<p>The '<tt>br</tt>' instruction is used to cause control flow to
1943transfer to a different basic block in the current function. There are
1944two forms of this instruction, corresponding to a conditional branch
1945and an unconditional branch.</p>
1946<h5>Arguments:</h5>
1947<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1948single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1949unconditional form of the '<tt>br</tt>' instruction takes a single
1950'<tt>label</tt>' value as a target.</p>
1951<h5>Semantics:</h5>
1952<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1953argument is evaluated. If the value is <tt>true</tt>, control flows
1954to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1955control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1956<h5>Example:</h5>
1957<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
1958 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1959</div>
1960<!-- _______________________________________________________________________ -->
1961<div class="doc_subsubsection">
1962 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1963</div>
1964
1965<div class="doc_text">
1966<h5>Syntax:</h5>
1967
1968<pre>
1969 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
1970</pre>
1971
1972<h5>Overview:</h5>
1973
1974<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1975several different places. It is a generalization of the '<tt>br</tt>'
1976instruction, allowing a branch to occur to one of many possible
1977destinations.</p>
1978
1979
1980<h5>Arguments:</h5>
1981
1982<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1983comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1984an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1985table is not allowed to contain duplicate constant entries.</p>
1986
1987<h5>Semantics:</h5>
1988
1989<p>The <tt>switch</tt> instruction specifies a table of values and
1990destinations. When the '<tt>switch</tt>' instruction is executed, this
1991table is searched for the given value. If the value is found, control flow is
1992transfered to the corresponding destination; otherwise, control flow is
1993transfered to the default destination.</p>
1994
1995<h5>Implementation:</h5>
1996
1997<p>Depending on properties of the target machine and the particular
1998<tt>switch</tt> instruction, this instruction may be code generated in different
1999ways. For example, it could be generated as a series of chained conditional
2000branches or with a lookup table.</p>
2001
2002<h5>Example:</h5>
2003
2004<pre>
2005 <i>; Emulate a conditional br instruction</i>
2006 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2007 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
2008
2009 <i>; Emulate an unconditional br instruction</i>
2010 switch i32 0, label %dest [ ]
2011
2012 <i>; Implement a jump table:</i>
2013 switch i32 %val, label %otherwise [ i32 0, label %onzero
2014 i32 1, label %onone
2015 i32 2, label %ontwo ]
2016</pre>
2017</div>
2018
2019<!-- _______________________________________________________________________ -->
2020<div class="doc_subsubsection">
2021 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2022</div>
2023
2024<div class="doc_text">
2025
2026<h5>Syntax:</h5>
2027
2028<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2029 &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]2030 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
2031</pre>
2032
2033<h5>Overview:</h5>
2034
2035<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2036function, with the possibility of control flow transfer to either the
2037'<tt>normal</tt>' label or the
2038'<tt>exception</tt>' label. If the callee function returns with the
2039"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2040"normal" label. If the callee (or any indirect callees) returns with the "<a
2041href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]2042continued at the dynamically nearest "exception" label. If the callee function
Devang Patelec8a5b02008-03-11 05:51:59 +0000[diff] [blame]2043returns multiple values then individual return values are only accessible through
2044a '<tt><a href="#i_getresult">getresult</a></tt>' instruction.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2045
2046<h5>Arguments:</h5>
2047
2048<p>This instruction requires several arguments:</p>
2049
2050<ol>
2051 <li>
2052 The optional "cconv" marker indicates which <a href="#callingconv">calling
2053 convention</a> the call should use. If none is specified, the call defaults
2054 to using C calling conventions.
2055 </li>
2056 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2057 function value being invoked. In most cases, this is a direct function
2058 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2059 an arbitrary pointer to function value.
2060 </li>
2061
2062 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2063 function to be invoked. </li>
2064
2065 <li>'<tt>function args</tt>': argument list whose types match the function
2066 signature argument types. If the function signature indicates the function
2067 accepts a variable number of arguments, the extra arguments can be
2068 specified. </li>
2069
2070 <li>'<tt>normal label</tt>': the label reached when the called function
2071 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2072
2073 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2074 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2075
2076</ol>
2077
2078<h5>Semantics:</h5>
2079
2080<p>This instruction is designed to operate as a standard '<tt><a
2081href="#i_call">call</a></tt>' instruction in most regards. The primary
2082difference is that it establishes an association with a label, which is used by
2083the runtime library to unwind the stack.</p>
2084
2085<p>This instruction is used in languages with destructors to ensure that proper
2086cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2087exception. Additionally, this is important for implementation of
2088'<tt>catch</tt>' clauses in high-level languages that support them.</p>
2089
2090<h5>Example:</h5>
2091<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2092 %retval = invoke i32 @Test(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2093 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2094 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2095 unwind label %TestCleanup <i>; {i32}:retval set</i>
2096</pre>
2097</div>
2098
2099
2100<!-- _______________________________________________________________________ -->
2101
2102<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2103Instruction</a> </div>
2104
2105<div class="doc_text">
2106
2107<h5>Syntax:</h5>
2108<pre>
2109 unwind
2110</pre>
2111
2112<h5>Overview:</h5>
2113
2114<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2115at the first callee in the dynamic call stack which used an <a
2116href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2117primarily used to implement exception handling.</p>
2118
2119<h5>Semantics:</h5>
2120
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2121<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2122immediately halt. The dynamic call stack is then searched for the first <a
2123href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2124execution continues at the "exceptional" destination block specified by the
2125<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2126dynamic call chain, undefined behavior results.</p>
2127</div>
2128
2129<!-- _______________________________________________________________________ -->
2130
2131<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2132Instruction</a> </div>
2133
2134<div class="doc_text">
2135
2136<h5>Syntax:</h5>
2137<pre>
2138 unreachable
2139</pre>
2140
2141<h5>Overview:</h5>
2142
2143<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2144instruction is used to inform the optimizer that a particular portion of the
2145code is not reachable. This can be used to indicate that the code after a
2146no-return function cannot be reached, and other facts.</p>
2147
2148<h5>Semantics:</h5>
2149
2150<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2151</div>
2152
2153
2154
2155<!-- ======================================================================= -->
2156<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2157<div class="doc_text">
2158<p>Binary operators are used to do most of the computation in a
Chris Lattnerab596d92008-04-01 18:47:32 +0000[diff] [blame]2159program. They require two operands of the same type, execute an operation on them, and
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2160produce a single value. The operands might represent
2161multiple 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]2162The result value has the same type as its operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2163<p>There are several different binary operators:</p>
2164</div>
2165<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2166<div class="doc_subsubsection">
2167 <a name="i_add">'<tt>add</tt>' Instruction</a>
2168</div>
2169
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2170<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2171
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2172<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2173
2174<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2175 &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]2176</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2177
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2178<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2179
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2180<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2181
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2182<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2183
2184<p>The two arguments to the '<tt>add</tt>' instruction must be <a
2185 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>, or
2186 <a href="#t_vector">vector</a> values. Both arguments must have identical
2187 types.</p>
2188
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2189<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2190
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2191<p>The value produced is the integer or floating point sum of the two
2192operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2193
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2194<p>If an integer sum has unsigned overflow, the result returned is the
2195mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2196the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2197
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2198<p>Because LLVM integers use a two's complement representation, this
2199instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2200
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2201<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2202
2203<pre>
2204 &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2205</pre>
2206</div>
2207<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2208<div class="doc_subsubsection">
2209 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2210</div>
2211
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2212<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2213
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2214<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2215
2216<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2217 &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]2218</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2219
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2220<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2221
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2222<p>The '<tt>sub</tt>' instruction returns the difference of its two
2223operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2224
2225<p>Note that the '<tt>sub</tt>' instruction is used to represent the
2226'<tt>neg</tt>' instruction present in most other intermediate
2227representations.</p>
2228
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2229<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2230
2231<p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2232 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2233 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2234 types.</p>
2235
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2236<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2237
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2238<p>The value produced is the integer or floating point difference of
2239the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2240
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2241<p>If an integer difference has unsigned overflow, the result returned is the
2242mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2243the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2244
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2245<p>Because LLVM integers use a two's complement representation, this
2246instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2247
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2248<h5>Example:</h5>
2249<pre>
2250 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2251 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2252</pre>
2253</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2254
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2255<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2256<div class="doc_subsubsection">
2257 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2258</div>
2259
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2260<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2261
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2262<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2263<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]2264</pre>
2265<h5>Overview:</h5>
2266<p>The '<tt>mul</tt>' instruction returns the product of its two
2267operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2268
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2269<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2270
2271<p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2272href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2273or <a href="#t_vector">vector</a> values. Both arguments must have identical
2274types.</p>
2275
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2276<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2277
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2278<p>The value produced is the integer or floating point product of the
2279two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2280
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2281<p>If the result of an integer multiplication has unsigned overflow,
2282the result returned is the mathematical result modulo
22832<sup>n</sup>, where n is the bit width of the result.</p>
2284<p>Because LLVM integers use a two's complement representation, and the
2285result is the same width as the operands, this instruction returns the
2286correct result for both signed and unsigned integers. If a full product
2287(e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2288should be sign-extended or zero-extended as appropriate to the
2289width of the full product.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2290<h5>Example:</h5>
2291<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2292</pre>
2293</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2294
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2295<!-- _______________________________________________________________________ -->
2296<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2297</a></div>
2298<div class="doc_text">
2299<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2300<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]2301</pre>
2302<h5>Overview:</h5>
2303<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2304operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2305
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2306<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2307
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2308<p>The two arguments to the '<tt>udiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2309<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2310values. Both arguments must have identical types.</p>
2311
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2312<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2313
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2314<p>The value produced is the unsigned integer quotient of the two operands.</p>
2315<p>Note that unsigned integer division and signed integer division are distinct
2316operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2317<p>Division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2318<h5>Example:</h5>
2319<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2320</pre>
2321</div>
2322<!-- _______________________________________________________________________ -->
2323<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2324</a> </div>
2325<div class="doc_text">
2326<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2327<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2328 &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]2329</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2330
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2331<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2332
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2333<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2334operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2335
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2336<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2337
2338<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2339<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2340values. Both arguments must have identical types.</p>
2341
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2342<h5>Semantics:</h5>
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2343<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]2344<p>Note that signed integer division and unsigned integer division are distinct
2345operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2346<p>Division by zero leads to undefined behavior. Overflow also leads to
2347undefined behavior; this is a rare case, but can occur, for example,
2348by doing a 32-bit division of -2147483648 by -1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2349<h5>Example:</h5>
2350<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2351</pre>
2352</div>
2353<!-- _______________________________________________________________________ -->
2354<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2355Instruction</a> </div>
2356<div class="doc_text">
2357<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2358<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2359 &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]2360</pre>
2361<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2362
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2363<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2364operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2365
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2366<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2367
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2368<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2369<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2370of floating point values. Both arguments must have identical types.</p>
2371
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2372<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2373
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2374<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2375
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2376<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2377
2378<pre>
2379 &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]2380</pre>
2381</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2382
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2383<!-- _______________________________________________________________________ -->
2384<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2385</div>
2386<div class="doc_text">
2387<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2388<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]2389</pre>
2390<h5>Overview:</h5>
2391<p>The '<tt>urem</tt>' instruction returns the remainder from the
2392unsigned division of its two arguments.</p>
2393<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2394<p>The two arguments to the '<tt>urem</tt>' instruction must be
2395<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2396values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2397<h5>Semantics:</h5>
2398<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2399This instruction always performs an unsigned division to get the remainder.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2400<p>Note that unsigned integer remainder and signed integer remainder are
2401distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2402<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2403<h5>Example:</h5>
2404<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2405</pre>
2406
2407</div>
2408<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2409<div class="doc_subsubsection">
2410 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2411</div>
2412
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2413<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2414
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2415<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2416
2417<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2418 &lt;result&gt; = 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]2419</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2420
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2421<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2422
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2423<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2424signed division of its two operands. This instruction can also take
2425<a href="#t_vector">vector</a> versions of the values in which case
2426the elements must be integers.</p>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]2427
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2428<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2429
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2430<p>The two arguments to the '<tt>srem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2431<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2432values. Both arguments must have identical types.</p>
2433
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2434<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2435
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2436<p>This instruction returns the <i>remainder</i> of a division (where the result
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2437has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2438operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2439a value. For more information about the difference, see <a
2440 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2441Math Forum</a>. For a table of how this is implemented in various languages,
2442please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2443Wikipedia: modulo operation</a>.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2444<p>Note that signed integer remainder and unsigned integer remainder are
2445distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2446<p>Taking the remainder of a division by zero leads to undefined behavior.
2447Overflow also leads to undefined behavior; this is a rare case, but can occur,
2448for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2449(The remainder doesn't actually overflow, but this rule lets srem be
2450implemented using instructions that return both the result of the division
2451and the remainder.)</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2452<h5>Example:</h5>
2453<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2454</pre>
2455
2456</div>
2457<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2458<div class="doc_subsubsection">
2459 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2460
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2461<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2462
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2463<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2464<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]2465</pre>
2466<h5>Overview:</h5>
2467<p>The '<tt>frem</tt>' instruction returns the remainder from the
2468division of its two operands.</p>
2469<h5>Arguments:</h5>
2470<p>The two arguments to the '<tt>frem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2471<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2472of floating point values. Both arguments must have identical types.</p>
2473
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2474<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2475
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2476<p>This instruction returns the <i>remainder</i> of a division.
2477The remainder has the same sign as the dividend.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2478
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2479<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2480
2481<pre>
2482 &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2483</pre>
2484</div>
2485
2486<!-- ======================================================================= -->
2487<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2488Operations</a> </div>
2489<div class="doc_text">
2490<p>Bitwise binary operators are used to do various forms of
2491bit-twiddling in a program. They are generally very efficient
2492instructions and can commonly be strength reduced from other
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2493instructions. They require two operands of the same type, execute an operation on them,
2494and 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]2495</div>
2496
2497<!-- _______________________________________________________________________ -->
2498<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2499Instruction</a> </div>
2500<div class="doc_text">
2501<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2502<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]2503</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2504
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2505<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2506
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2507<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2508the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2509
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2510<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2511
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2512<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2513 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2514type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2515
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2516<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2517
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2518<p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2519where n is the width of the result. If <tt>op2</tt> is (statically or dynamically) negative or
2520equal 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]2521
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2522<h5>Example:</h5><pre>
2523 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2524 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2525 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2526 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2527</pre>
2528</div>
2529<!-- _______________________________________________________________________ -->
2530<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2531Instruction</a> </div>
2532<div class="doc_text">
2533<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2534<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]2535</pre>
2536
2537<h5>Overview:</h5>
2538<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2539operand shifted to the right a specified number of bits with zero fill.</p>
2540
2541<h5>Arguments:</h5>
2542<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2543<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]2544type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2545
2546<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2547
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2548<p>This instruction always performs a logical shift right operation. The most
2549significant bits of the result will be filled with zero bits after the
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2550shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
2551the number of bits in <tt>op1</tt>, the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2552
2553<h5>Example:</h5>
2554<pre>
2555 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2556 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2557 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2558 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2559 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2560</pre>
2561</div>
2562
2563<!-- _______________________________________________________________________ -->
2564<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2565Instruction</a> </div>
2566<div class="doc_text">
2567
2568<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2569<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]2570</pre>
2571
2572<h5>Overview:</h5>
2573<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2574operand shifted to the right a specified number of bits with sign extension.</p>
2575
2576<h5>Arguments:</h5>
2577<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Nate Begemanbb1ce942008-07-29 15:49:41 +0000[diff] [blame]2578<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]2579type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2580
2581<h5>Semantics:</h5>
2582<p>This instruction always performs an arithmetic shift right operation,
2583The most significant bits of the result will be filled with the sign bit
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2584of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
2585larger than the number of bits in <tt>op1</tt>, the result is undefined.
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2586</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2587
2588<h5>Example:</h5>
2589<pre>
2590 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2591 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2592 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2593 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2594 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2595</pre>
2596</div>
2597
2598<!-- _______________________________________________________________________ -->
2599<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2600Instruction</a> </div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2601
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2602<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2603
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2604<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2605
2606<pre>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2607 &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]2608</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2609
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2610<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2611
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2612<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2613its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2614
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2615<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2616
2617<p>The two arguments to the '<tt>and</tt>' instruction must be
2618<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2619values. Both arguments must have identical types.</p>
2620
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2621<h5>Semantics:</h5>
2622<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2623<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2624<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2625<table border="1" cellspacing="0" cellpadding="4">
2626 <tbody>
2627 <tr>
2628 <td>In0</td>
2629 <td>In1</td>
2630 <td>Out</td>
2631 </tr>
2632 <tr>
2633 <td>0</td>
2634 <td>0</td>
2635 <td>0</td>
2636 </tr>
2637 <tr>
2638 <td>0</td>
2639 <td>1</td>
2640 <td>0</td>
2641 </tr>
2642 <tr>
2643 <td>1</td>
2644 <td>0</td>
2645 <td>0</td>
2646 </tr>
2647 <tr>
2648 <td>1</td>
2649 <td>1</td>
2650 <td>1</td>
2651 </tr>
2652 </tbody>
2653</table>
2654</div>
2655<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2656<pre>
2657 &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]2658 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2659 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2660</pre>
2661</div>
2662<!-- _______________________________________________________________________ -->
2663<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2664<div class="doc_text">
2665<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2666<pre> &lt;result&gt; = 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]2667</pre>
2668<h5>Overview:</h5>
2669<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2670or of its two operands.</p>
2671<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2672
2673<p>The two arguments to the '<tt>or</tt>' instruction must be
2674<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2675values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2676<h5>Semantics:</h5>
2677<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2678<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2679<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2680<table border="1" cellspacing="0" cellpadding="4">
2681 <tbody>
2682 <tr>
2683 <td>In0</td>
2684 <td>In1</td>
2685 <td>Out</td>
2686 </tr>
2687 <tr>
2688 <td>0</td>
2689 <td>0</td>
2690 <td>0</td>
2691 </tr>
2692 <tr>
2693 <td>0</td>
2694 <td>1</td>
2695 <td>1</td>
2696 </tr>
2697 <tr>
2698 <td>1</td>
2699 <td>0</td>
2700 <td>1</td>
2701 </tr>
2702 <tr>
2703 <td>1</td>
2704 <td>1</td>
2705 <td>1</td>
2706 </tr>
2707 </tbody>
2708</table>
2709</div>
2710<h5>Example:</h5>
2711<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2712 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2713 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2714</pre>
2715</div>
2716<!-- _______________________________________________________________________ -->
2717<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2718Instruction</a> </div>
2719<div class="doc_text">
2720<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]2721<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]2722</pre>
2723<h5>Overview:</h5>
2724<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2725or of its two operands. The <tt>xor</tt> is used to implement the
2726"one's complement" operation, which is the "~" operator in C.</p>
2727<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2728<p>The two arguments to the '<tt>xor</tt>' instruction must be
2729<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2730values. Both arguments must have identical types.</p>
2731
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2732<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2733
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2734<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2735<p> </p>
Bill Wendling1e19eac2008-09-07 10:29:20 +0000[diff] [blame]2736<div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2737<table border="1" cellspacing="0" cellpadding="4">
2738 <tbody>
2739 <tr>
2740 <td>In0</td>
2741 <td>In1</td>
2742 <td>Out</td>
2743 </tr>
2744 <tr>
2745 <td>0</td>
2746 <td>0</td>
2747 <td>0</td>
2748 </tr>
2749 <tr>
2750 <td>0</td>
2751 <td>1</td>
2752 <td>1</td>
2753 </tr>
2754 <tr>
2755 <td>1</td>
2756 <td>0</td>
2757 <td>1</td>
2758 </tr>
2759 <tr>
2760 <td>1</td>
2761 <td>1</td>
2762 <td>0</td>
2763 </tr>
2764 </tbody>
2765</table>
2766</div>
2767<p> </p>
2768<h5>Example:</h5>
2769<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2770 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2771 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2772 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2773</pre>
2774</div>
2775
2776<!-- ======================================================================= -->
2777<div class="doc_subsection">
2778 <a name="vectorops">Vector Operations</a>
2779</div>
2780
2781<div class="doc_text">
2782
2783<p>LLVM supports several instructions to represent vector operations in a
2784target-independent manner. These instructions cover the element-access and
2785vector-specific operations needed to process vectors effectively. While LLVM
2786does directly support these vector operations, many sophisticated algorithms
2787will want to use target-specific intrinsics to take full advantage of a specific
2788target.</p>
2789
2790</div>
2791
2792<!-- _______________________________________________________________________ -->
2793<div class="doc_subsubsection">
2794 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2795</div>
2796
2797<div class="doc_text">
2798
2799<h5>Syntax:</h5>
2800
2801<pre>
2802 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
2803</pre>
2804
2805<h5>Overview:</h5>
2806
2807<p>
2808The '<tt>extractelement</tt>' instruction extracts a single scalar
2809element from a vector at a specified index.
2810</p>
2811
2812
2813<h5>Arguments:</h5>
2814
2815<p>
2816The first operand of an '<tt>extractelement</tt>' instruction is a
2817value of <a href="#t_vector">vector</a> type. The second operand is
2818an index indicating the position from which to extract the element.
2819The index may be a variable.</p>
2820
2821<h5>Semantics:</h5>
2822
2823<p>
2824The result is a scalar of the same type as the element type of
2825<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2826<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2827results are undefined.
2828</p>
2829
2830<h5>Example:</h5>
2831
2832<pre>
2833 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
2834</pre>
2835</div>
2836
2837
2838<!-- _______________________________________________________________________ -->
2839<div class="doc_subsubsection">
2840 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2841</div>
2842
2843<div class="doc_text">
2844
2845<h5>Syntax:</h5>
2846
2847<pre>
Dan Gohmanbcc3c502008-05-12 23:38:42 +0000[diff] [blame]2848 &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]2849</pre>
2850
2851<h5>Overview:</h5>
2852
2853<p>
2854The '<tt>insertelement</tt>' instruction inserts a scalar
2855element into a vector at a specified index.
2856</p>
2857
2858
2859<h5>Arguments:</h5>
2860
2861<p>
2862The first operand of an '<tt>insertelement</tt>' instruction is a
2863value of <a href="#t_vector">vector</a> type. The second operand is a
2864scalar value whose type must equal the element type of the first
2865operand. The third operand is an index indicating the position at
2866which to insert the value. The index may be a variable.</p>
2867
2868<h5>Semantics:</h5>
2869
2870<p>
2871The result is a vector of the same type as <tt>val</tt>. Its
2872element values are those of <tt>val</tt> except at position
2873<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2874exceeds the length of <tt>val</tt>, the results are undefined.
2875</p>
2876
2877<h5>Example:</h5>
2878
2879<pre>
2880 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
2881</pre>
2882</div>
2883
2884<!-- _______________________________________________________________________ -->
2885<div class="doc_subsubsection">
2886 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2887</div>
2888
2889<div class="doc_text">
2890
2891<h5>Syntax:</h5>
2892
2893<pre>
2894 &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>
2895</pre>
2896
2897<h5>Overview:</h5>
2898
2899<p>
2900The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2901from two input vectors, returning a vector of the same type.
2902</p>
2903
2904<h5>Arguments:</h5>
2905
2906<p>
2907The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2908with types that match each other and types that match the result of the
2909instruction. The third argument is a shuffle mask, which has the same number
2910of elements as the other vector type, but whose element type is always 'i32'.
2911</p>
2912
2913<p>
2914The shuffle mask operand is required to be a constant vector with either
2915constant integer or undef values.
2916</p>
2917
2918<h5>Semantics:</h5>
2919
2920<p>
2921The elements of the two input vectors are numbered from left to right across
2922both of the vectors. The shuffle mask operand specifies, for each element of
2923the result vector, which element of the two input registers the result element
2924gets. The element selector may be undef (meaning "don't care") and the second
2925operand may be undef if performing a shuffle from only one vector.
2926</p>
2927
2928<h5>Example:</h5>
2929
2930<pre>
2931 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
2932 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
2933 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
2934 &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.
2935</pre>
2936</div>
2937
2938
2939<!-- ======================================================================= -->
2940<div class="doc_subsection">
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2941 <a name="aggregateops">Aggregate Operations</a>
2942</div>
2943
2944<div class="doc_text">
2945
2946<p>LLVM supports several instructions for working with aggregate values.
2947</p>
2948
2949</div>
2950
2951<!-- _______________________________________________________________________ -->
2952<div class="doc_subsubsection">
2953 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
2954</div>
2955
2956<div class="doc_text">
2957
2958<h5>Syntax:</h5>
2959
2960<pre>
2961 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
2962</pre>
2963
2964<h5>Overview:</h5>
2965
2966<p>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2967The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
2968or array element from an aggregate value.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2969</p>
2970
2971
2972<h5>Arguments:</h5>
2973
2974<p>
2975The first operand of an '<tt>extractvalue</tt>' instruction is a
2976value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2977type. The operands are constant indices to specify which value to extract
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]2978in a similar manner as indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2979'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
2980</p>
2981
2982<h5>Semantics:</h5>
2983
2984<p>
2985The result is the value at the position in the aggregate specified by
2986the index operands.
2987</p>
2988
2989<h5>Example:</h5>
2990
2991<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]2992 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2993</pre>
2994</div>
2995
2996
2997<!-- _______________________________________________________________________ -->
2998<div class="doc_subsubsection">
2999 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3000</div>
3001
3002<div class="doc_text">
3003
3004<h5>Syntax:</h5>
3005
3006<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3007 &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]3008</pre>
3009
3010<h5>Overview:</h5>
3011
3012<p>
3013The '<tt>insertvalue</tt>' instruction inserts a value
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3014into a struct field or array element in an aggregate.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3015</p>
3016
3017
3018<h5>Arguments:</h5>
3019
3020<p>
3021The first operand of an '<tt>insertvalue</tt>' instruction is a
3022value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3023The second operand is a first-class value to insert.
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]3024The following operands are constant indices
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]3025indicating the position at which to insert the value in a similar manner as
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3026indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3027'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3028The value to insert must have the same type as the value identified
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3029by the indices.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3030
3031<h5>Semantics:</h5>
3032
3033<p>
3034The result is an aggregate of the same type as <tt>val</tt>. Its
3035value is that of <tt>val</tt> except that the value at the position
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]3036specified by the indices is that of <tt>elt</tt>.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3037</p>
3038
3039<h5>Example:</h5>
3040
3041<pre>
Dan Gohmanb1aab4e2008-06-23 15:26:37 +0000[diff] [blame]3042 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]3043</pre>
3044</div>
3045
3046
3047<!-- ======================================================================= -->
3048<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3049 <a name="memoryops">Memory Access and Addressing Operations</a>
3050</div>
3051
3052<div class="doc_text">
3053
3054<p>A key design point of an SSA-based representation is how it
3055represents memory. In LLVM, no memory locations are in SSA form, which
3056makes things very simple. This section describes how to read, write,
3057allocate, and free memory in LLVM.</p>
3058
3059</div>
3060
3061<!-- _______________________________________________________________________ -->
3062<div class="doc_subsubsection">
3063 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3064</div>
3065
3066<div class="doc_text">
3067
3068<h5>Syntax:</h5>
3069
3070<pre>
3071 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3072</pre>
3073
3074<h5>Overview:</h5>
3075
3076<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3077heap and returns a pointer to it. The object is always allocated in the generic
3078address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3079
3080<h5>Arguments:</h5>
3081
3082<p>The '<tt>malloc</tt>' instruction allocates
3083<tt>sizeof(&lt;type&gt;)*NumElements</tt>
3084bytes of memory from the operating system and returns a pointer of the
3085appropriate type to the program. If "NumElements" is specified, it is the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3086number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3087If a constant alignment is specified, the value result of the allocation is guaranteed to
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3088be aligned to at least that boundary. If not specified, or if zero, the target can
3089choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3090
3091<p>'<tt>type</tt>' must be a sized type.</p>
3092
3093<h5>Semantics:</h5>
3094
3095<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3096a pointer is returned. The result of a zero byte allocattion is undefined. The
3097result is null if there is insufficient memory available.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3098
3099<h5>Example:</h5>
3100
3101<pre>
3102 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
3103
3104 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3105 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3106 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3107 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3108 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3109</pre>
3110</div>
3111
3112<!-- _______________________________________________________________________ -->
3113<div class="doc_subsubsection">
3114 <a name="i_free">'<tt>free</tt>' Instruction</a>
3115</div>
3116
3117<div class="doc_text">
3118
3119<h5>Syntax:</h5>
3120
3121<pre>
3122 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
3123</pre>
3124
3125<h5>Overview:</h5>
3126
3127<p>The '<tt>free</tt>' instruction returns memory back to the unused
3128memory heap to be reallocated in the future.</p>
3129
3130<h5>Arguments:</h5>
3131
3132<p>'<tt>value</tt>' shall be a pointer value that points to a value
3133that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3134instruction.</p>
3135
3136<h5>Semantics:</h5>
3137
3138<p>Access to the memory pointed to by the pointer is no longer defined
Chris Lattner329d6532008-04-19 22:41:32 +0000[diff] [blame]3139after this instruction executes. If the pointer is null, the operation
3140is a noop.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3141
3142<h5>Example:</h5>
3143
3144<pre>
3145 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
3146 free [4 x i8]* %array
3147</pre>
3148</div>
3149
3150<!-- _______________________________________________________________________ -->
3151<div class="doc_subsubsection">
3152 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3153</div>
3154
3155<div class="doc_text">
3156
3157<h5>Syntax:</h5>
3158
3159<pre>
3160 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3161</pre>
3162
3163<h5>Overview:</h5>
3164
3165<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3166currently executing function, to be automatically released when this function
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3167returns to its caller. The object is always allocated in the generic address
3168space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3169
3170<h5>Arguments:</h5>
3171
3172<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
3173bytes of memory on the runtime stack, returning a pointer of the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3174appropriate type to the program. If "NumElements" is specified, it is the
3175number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3176If a constant alignment is specified, the value result of the allocation is guaranteed
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3177to be aligned to at least that boundary. If not specified, or if zero, the target
3178can choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3179
3180<p>'<tt>type</tt>' may be any sized type.</p>
3181
3182<h5>Semantics:</h5>
3183
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3184<p>Memory is allocated; a pointer is returned. The operation is undefiend if
3185there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3186memory is automatically released when the function returns. The '<tt>alloca</tt>'
3187instruction is commonly used to represent automatic variables that must
3188have an address available. When the function returns (either with the <tt><a
3189 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]3190instructions), the memory is reclaimed. Allocating zero bytes
3191is legal, but the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3192
3193<h5>Example:</h5>
3194
3195<pre>
3196 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3197 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3198 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3199 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3200</pre>
3201</div>
3202
3203<!-- _______________________________________________________________________ -->
3204<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3205Instruction</a> </div>
3206<div class="doc_text">
3207<h5>Syntax:</h5>
3208<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>
3209<h5>Overview:</h5>
3210<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3211<h5>Arguments:</h5>
3212<p>The argument to the '<tt>load</tt>' instruction specifies the memory
3213address from which to load. The pointer must point to a <a
3214 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3215marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3216the number or order of execution of this <tt>load</tt> with other
3217volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3218instructions. </p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3219<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3220The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3221(that is, the alignment of the memory address). A value of 0 or an
3222omitted "align" argument means that the operation has the preferential
3223alignment for the target. It is the responsibility of the code emitter
3224to ensure that the alignment information is correct. Overestimating
3225the alignment results in an undefined behavior. Underestimating the
3226alignment may produce less efficient code. An alignment of 1 is always
3227safe.
3228</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3229<h5>Semantics:</h5>
3230<p>The location of memory pointed to is loaded.</p>
3231<h5>Examples:</h5>
3232<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3233 <a
3234 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3235 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3236</pre>
3237</div>
3238<!-- _______________________________________________________________________ -->
3239<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3240Instruction</a> </div>
3241<div class="doc_text">
3242<h5>Syntax:</h5>
3243<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3244 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3245</pre>
3246<h5>Overview:</h5>
3247<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3248<h5>Arguments:</h5>
3249<p>There are two arguments to the '<tt>store</tt>' instruction: a value
3250to 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]3251operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3252of the '<tt>&lt;value&gt;</tt>'
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3253operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3254optimizer is not allowed to modify the number or order of execution of
3255this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3256 href="#i_store">store</a></tt> instructions.</p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3257<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3258The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3259(that is, the alignment of the memory address). A value of 0 or an
3260omitted "align" argument means that the operation has the preferential
3261alignment for the target. It is the responsibility of the code emitter
3262to ensure that the alignment information is correct. Overestimating
3263the alignment results in an undefined behavior. Underestimating the
3264alignment may produce less efficient code. An alignment of 1 is always
3265safe.
3266</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3267<h5>Semantics:</h5>
3268<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
3269at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
3270<h5>Example:</h5>
3271<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]3272 store i32 3, i32* %ptr <i>; yields {void}</i>
3273 %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]3274</pre>
3275</div>
3276
3277<!-- _______________________________________________________________________ -->
3278<div class="doc_subsubsection">
3279 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3280</div>
3281
3282<div class="doc_text">
3283<h5>Syntax:</h5>
3284<pre>
3285 &lt;result&gt; = getelementptr &lt;ty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
3286</pre>
3287
3288<h5>Overview:</h5>
3289
3290<p>
3291The '<tt>getelementptr</tt>' instruction is used to get the address of a
3292subelement of an aggregate data structure.</p>
3293
3294<h5>Arguments:</h5>
3295
3296<p>This instruction takes a list of integer operands that indicate what
3297elements of the aggregate object to index to. The actual types of the arguments
3298provided depend on the type of the first pointer argument. The
3299'<tt>getelementptr</tt>' instruction is used to index down through the type
3300levels of a structure or to a specific index in an array. When indexing into a
3301structure, only <tt>i32</tt> integer constants are allowed. When indexing
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]3302into an array or pointer, only integers of 32 or 64 bits are allowed; 32-bit
3303values will be sign extended to 64-bits if required.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3304
3305<p>For example, let's consider a C code fragment and how it gets
3306compiled to LLVM:</p>
3307
3308<div class="doc_code">
3309<pre>
3310struct RT {
3311 char A;
3312 int B[10][20];
3313 char C;
3314};
3315struct ST {
3316 int X;
3317 double Y;
3318 struct RT Z;
3319};
3320
3321int *foo(struct ST *s) {
3322 return &amp;s[1].Z.B[5][13];
3323}
3324</pre>
3325</div>
3326
3327<p>The LLVM code generated by the GCC frontend is:</p>
3328
3329<div class="doc_code">
3330<pre>
3331%RT = type { i8 , [10 x [20 x i32]], i8 }
3332%ST = type { i32, double, %RT }
3333
3334define i32* %foo(%ST* %s) {
3335entry:
3336 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3337 ret i32* %reg
3338}
3339</pre>
3340</div>
3341
3342<h5>Semantics:</h5>
3343
3344<p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
3345on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
3346and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
3347<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]3348to 64-bits. <a href="#t_struct">Structure</a> and <a href="#t_pstruct">packed
3349structure</a> types require <tt>i32</tt> <b>constants</b>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3350
3351<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3352type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3353}</tt>' type, a structure. The second index indexes into the third element of
3354the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3355i8 }</tt>' type, another structure. The third index indexes into the second
3356element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3357array. The two dimensions of the array are subscripted into, yielding an
3358'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3359to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3360
3361<p>Note that it is perfectly legal to index partially through a
3362structure, returning a pointer to an inner element. Because of this,
3363the LLVM code for the given testcase is equivalent to:</p>
3364
3365<pre>
3366 define i32* %foo(%ST* %s) {
3367 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3368 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3369 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3370 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3371 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3372 ret i32* %t5
3373 }
3374</pre>
3375
3376<p>Note that it is undefined to access an array out of bounds: array and
3377pointer indexes must always be within the defined bounds of the array type.
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]3378The one exception for this rule is zero length arrays. These arrays are
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3379defined to be accessible as variable length arrays, which requires access
3380beyond the zero'th element.</p>
3381
3382<p>The getelementptr instruction is often confusing. For some more insight
3383into how it works, see <a href="GetElementPtr.html">the getelementptr
3384FAQ</a>.</p>
3385
3386<h5>Example:</h5>
3387
3388<pre>
3389 <i>; yields [12 x i8]*:aptr</i>
3390 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
3391</pre>
3392</div>
3393
3394<!-- ======================================================================= -->
3395<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3396</div>
3397<div class="doc_text">
3398<p>The instructions in this category are the conversion instructions (casting)
3399which all take a single operand and a type. They perform various bit conversions
3400on the operand.</p>
3401</div>
3402
3403<!-- _______________________________________________________________________ -->
3404<div class="doc_subsubsection">
3405 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3406</div>
3407<div class="doc_text">
3408
3409<h5>Syntax:</h5>
3410<pre>
3411 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3412</pre>
3413
3414<h5>Overview:</h5>
3415<p>
3416The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3417</p>
3418
3419<h5>Arguments:</h5>
3420<p>
3421The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3422be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3423and type of the result, which must be an <a href="#t_integer">integer</a>
3424type. The bit size of <tt>value</tt> must be larger than the bit size of
3425<tt>ty2</tt>. Equal sized types are not allowed.</p>
3426
3427<h5>Semantics:</h5>
3428<p>
3429The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3430and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3431larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3432It will always truncate bits.</p>
3433
3434<h5>Example:</h5>
3435<pre>
3436 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3437 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3438 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3439</pre>
3440</div>
3441
3442<!-- _______________________________________________________________________ -->
3443<div class="doc_subsubsection">
3444 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3445</div>
3446<div class="doc_text">
3447
3448<h5>Syntax:</h5>
3449<pre>
3450 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3451</pre>
3452
3453<h5>Overview:</h5>
3454<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3455<tt>ty2</tt>.</p>
3456
3457
3458<h5>Arguments:</h5>
3459<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3460<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3461also be of <a href="#t_integer">integer</a> type. The bit size of the
3462<tt>value</tt> must be smaller than the bit size of the destination type,
3463<tt>ty2</tt>.</p>
3464
3465<h5>Semantics:</h5>
3466<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3467bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3468
3469<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3470
3471<h5>Example:</h5>
3472<pre>
3473 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3474 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3475</pre>
3476</div>
3477
3478<!-- _______________________________________________________________________ -->
3479<div class="doc_subsubsection">
3480 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3481</div>
3482<div class="doc_text">
3483
3484<h5>Syntax:</h5>
3485<pre>
3486 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3487</pre>
3488
3489<h5>Overview:</h5>
3490<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3491
3492<h5>Arguments:</h5>
3493<p>
3494The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3495<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3496also be of <a href="#t_integer">integer</a> type. The bit size of the
3497<tt>value</tt> must be smaller than the bit size of the destination type,
3498<tt>ty2</tt>.</p>
3499
3500<h5>Semantics:</h5>
3501<p>
3502The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3503bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3504the type <tt>ty2</tt>.</p>
3505
3506<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3507
3508<h5>Example:</h5>
3509<pre>
3510 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3511 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3512</pre>
3513</div>
3514
3515<!-- _______________________________________________________________________ -->
3516<div class="doc_subsubsection">
3517 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3518</div>
3519
3520<div class="doc_text">
3521
3522<h5>Syntax:</h5>
3523
3524<pre>
3525 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3526</pre>
3527
3528<h5>Overview:</h5>
3529<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3530<tt>ty2</tt>.</p>
3531
3532
3533<h5>Arguments:</h5>
3534<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3535 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3536cast it to. The size of <tt>value</tt> must be larger than the size of
3537<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3538<i>no-op cast</i>.</p>
3539
3540<h5>Semantics:</h5>
3541<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3542<a href="#t_floating">floating point</a> type to a smaller
3543<a href="#t_floating">floating point</a> type. If the value cannot fit within
3544the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3545
3546<h5>Example:</h5>
3547<pre>
3548 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3549 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3550</pre>
3551</div>
3552
3553<!-- _______________________________________________________________________ -->
3554<div class="doc_subsubsection">
3555 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3556</div>
3557<div class="doc_text">
3558
3559<h5>Syntax:</h5>
3560<pre>
3561 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3562</pre>
3563
3564<h5>Overview:</h5>
3565<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3566floating point value.</p>
3567
3568<h5>Arguments:</h5>
3569<p>The '<tt>fpext</tt>' instruction takes a
3570<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3571and a <a href="#t_floating">floating point</a> type to cast it to. The source
3572type must be smaller than the destination type.</p>
3573
3574<h5>Semantics:</h5>
3575<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3576<a href="#t_floating">floating point</a> type to a larger
3577<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3578used to make a <i>no-op cast</i> because it always changes bits. Use
3579<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3580
3581<h5>Example:</h5>
3582<pre>
3583 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3584 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3585</pre>
3586</div>
3587
3588<!-- _______________________________________________________________________ -->
3589<div class="doc_subsubsection">
3590 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3591</div>
3592<div class="doc_text">
3593
3594<h5>Syntax:</h5>
3595<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3596 &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]3597</pre>
3598
3599<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3600<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3601unsigned integer equivalent of type <tt>ty2</tt>.
3602</p>
3603
3604<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3605<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]3606scalar or vector <a href="#t_floating">floating point</a> value, and a type
3607to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3608type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3609vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3610
3611<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3612<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3613<a href="#t_floating">floating point</a> operand into the nearest (rounding
3614towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3615the results are undefined.</p>
3616
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3617<h5>Example:</h5>
3618<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3619 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3620 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3621 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3622</pre>
3623</div>
3624
3625<!-- _______________________________________________________________________ -->
3626<div class="doc_subsubsection">
3627 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3628</div>
3629<div class="doc_text">
3630
3631<h5>Syntax:</h5>
3632<pre>
3633 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3634</pre>
3635
3636<h5>Overview:</h5>
3637<p>The '<tt>fptosi</tt>' instruction converts
3638<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3639</p>
3640
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3641<h5>Arguments:</h5>
3642<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]3643scalar or vector <a href="#t_floating">floating point</a> value, and a type
3644to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3645type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3646vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3647
3648<h5>Semantics:</h5>
3649<p>The '<tt>fptosi</tt>' instruction converts its
3650<a href="#t_floating">floating point</a> operand into the nearest (rounding
3651towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3652the results are undefined.</p>
3653
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3654<h5>Example:</h5>
3655<pre>
3656 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3657 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3658 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3659</pre>
3660</div>
3661
3662<!-- _______________________________________________________________________ -->
3663<div class="doc_subsubsection">
3664 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3665</div>
3666<div class="doc_text">
3667
3668<h5>Syntax:</h5>
3669<pre>
3670 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3671</pre>
3672
3673<h5>Overview:</h5>
3674<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3675integer and converts that value to the <tt>ty2</tt> type.</p>
3676
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3677<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3678<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3679scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3680to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3681type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3682floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3683
3684<h5>Semantics:</h5>
3685<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3686integer quantity and converts it to the corresponding floating point value. If
3687the value cannot fit in the floating point value, the results are undefined.</p>
3688
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3689<h5>Example:</h5>
3690<pre>
3691 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3692 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3693</pre>
3694</div>
3695
3696<!-- _______________________________________________________________________ -->
3697<div class="doc_subsubsection">
3698 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3699</div>
3700<div class="doc_text">
3701
3702<h5>Syntax:</h5>
3703<pre>
3704 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3705</pre>
3706
3707<h5>Overview:</h5>
3708<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3709integer and converts that value to the <tt>ty2</tt> type.</p>
3710
3711<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3712<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3713scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3714to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3715type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3716floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3717
3718<h5>Semantics:</h5>
3719<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3720integer quantity and converts it to the corresponding floating point value. If
3721the value cannot fit in the floating point value, the results are undefined.</p>
3722
3723<h5>Example:</h5>
3724<pre>
3725 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3726 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3727</pre>
3728</div>
3729
3730<!-- _______________________________________________________________________ -->
3731<div class="doc_subsubsection">
3732 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3733</div>
3734<div class="doc_text">
3735
3736<h5>Syntax:</h5>
3737<pre>
3738 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3739</pre>
3740
3741<h5>Overview:</h5>
3742<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3743the integer type <tt>ty2</tt>.</p>
3744
3745<h5>Arguments:</h5>
3746<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3747must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3748<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3749
3750<h5>Semantics:</h5>
3751<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3752<tt>ty2</tt> by interpreting the pointer value as an integer and either
3753truncating or zero extending that value to the size of the integer type. If
3754<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3755<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3756are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3757change.</p>
3758
3759<h5>Example:</h5>
3760<pre>
3761 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3762 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3763</pre>
3764</div>
3765
3766<!-- _______________________________________________________________________ -->
3767<div class="doc_subsubsection">
3768 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3769</div>
3770<div class="doc_text">
3771
3772<h5>Syntax:</h5>
3773<pre>
3774 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3775</pre>
3776
3777<h5>Overview:</h5>
3778<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3779a pointer type, <tt>ty2</tt>.</p>
3780
3781<h5>Arguments:</h5>
3782<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3783value to cast, and a type to cast it to, which must be a
3784<a href="#t_pointer">pointer</a> type.
3785
3786<h5>Semantics:</h5>
3787<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3788<tt>ty2</tt> by applying either a zero extension or a truncation depending on
3789the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3790size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3791the size of a pointer then a zero extension is done. If they are the same size,
3792nothing is done (<i>no-op cast</i>).</p>
3793
3794<h5>Example:</h5>
3795<pre>
3796 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3797 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3798 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3799</pre>
3800</div>
3801
3802<!-- _______________________________________________________________________ -->
3803<div class="doc_subsubsection">
3804 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3805</div>
3806<div class="doc_text">
3807
3808<h5>Syntax:</h5>
3809<pre>
3810 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3811</pre>
3812
3813<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3814
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3815<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3816<tt>ty2</tt> without changing any bits.</p>
3817
3818<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3819
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3820<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
Dan Gohman7305fa02008-09-08 16:45:59 +0000[diff] [blame]3821a non-aggregate first class value, and a type to cast it to, which must also be
3822a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
3823<tt>value</tt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3824and the destination type, <tt>ty2</tt>, must be identical. If the source
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3825type is a pointer, the destination type must also be a pointer. This
3826instruction supports bitwise conversion of vectors to integers and to vectors
3827of other types (as long as they have the same size).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3828
3829<h5>Semantics:</h5>
3830<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3831<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3832this conversion. The conversion is done as if the <tt>value</tt> had been
3833stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3834converted to other pointer types with this instruction. To convert pointers to
3835other types, use the <a href="#i_inttoptr">inttoptr</a> or
3836<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3837
3838<h5>Example:</h5>
3839<pre>
3840 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3841 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3842 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3843</pre>
3844</div>
3845
3846<!-- ======================================================================= -->
3847<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3848<div class="doc_text">
3849<p>The instructions in this category are the "miscellaneous"
3850instructions, which defy better classification.</p>
3851</div>
3852
3853<!-- _______________________________________________________________________ -->
3854<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3855</div>
3856<div class="doc_text">
3857<h5>Syntax:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3858<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]3859</pre>
3860<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3861<p>The '<tt>icmp</tt>' instruction returns a boolean value or
3862a vector of boolean values based on comparison
3863of its two integer, integer vector, or pointer operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3864<h5>Arguments:</h5>
3865<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3866the condition code indicating the kind of comparison to perform. It is not
3867a value, just a keyword. The possible condition code are:
3868<ol>
3869 <li><tt>eq</tt>: equal</li>
3870 <li><tt>ne</tt>: not equal </li>
3871 <li><tt>ugt</tt>: unsigned greater than</li>
3872 <li><tt>uge</tt>: unsigned greater or equal</li>
3873 <li><tt>ult</tt>: unsigned less than</li>
3874 <li><tt>ule</tt>: unsigned less or equal</li>
3875 <li><tt>sgt</tt>: signed greater than</li>
3876 <li><tt>sge</tt>: signed greater or equal</li>
3877 <li><tt>slt</tt>: signed less than</li>
3878 <li><tt>sle</tt>: signed less or equal</li>
3879</ol>
3880<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3881<a href="#t_pointer">pointer</a>
3882or integer <a href="#t_vector">vector</a> typed.
3883They must also be identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3884<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3885<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]3886the condition code given as <tt>cond</tt>. The comparison performed always
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3887yields 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]3888<ol>
3889 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3890 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3891 </li>
3892 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3893 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3894 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3895 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3896 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3897 <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]3898 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3899 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3900 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3901 <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]3902 <li><tt>sgt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3903 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3904 <li><tt>sge</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3905 <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]3906 <li><tt>slt</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3907 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3908 <li><tt>sle</tt>: interprets the operands as signed values and yields
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3909 <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]3910</ol>
3911<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3912values are compared as if they were integers.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3913<p>If the operands are integer vectors, then they are compared
3914element by element. The result is an <tt>i1</tt> vector with
3915the same number of elements as the values being compared.
3916Otherwise, the result is an <tt>i1</tt>.
3917</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3918
3919<h5>Example:</h5>
3920<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
3921 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
3922 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
3923 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3924 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3925 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
3926</pre>
3927</div>
3928
3929<!-- _______________________________________________________________________ -->
3930<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3931</div>
3932<div class="doc_text">
3933<h5>Syntax:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3934<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]3935</pre>
3936<h5>Overview:</h5>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3937<p>The '<tt>fcmp</tt>' instruction returns a boolean value
3938or vector of boolean values based on comparison
3939of its operands.
3940<p>
3941If the operands are floating point scalars, then the result
3942type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
3943</p>
3944<p>If the operands are floating point vectors, then the result type
3945is a vector of boolean with the same number of elements as the
3946operands being compared.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3947<h5>Arguments:</h5>
3948<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3949the condition code indicating the kind of comparison to perform. It is not
3950a value, just a keyword. The possible condition code are:
3951<ol>
3952 <li><tt>false</tt>: no comparison, always returns false</li>
3953 <li><tt>oeq</tt>: ordered and equal</li>
3954 <li><tt>ogt</tt>: ordered and greater than </li>
3955 <li><tt>oge</tt>: ordered and greater than or equal</li>
3956 <li><tt>olt</tt>: ordered and less than </li>
3957 <li><tt>ole</tt>: ordered and less than or equal</li>
3958 <li><tt>one</tt>: ordered and not equal</li>
3959 <li><tt>ord</tt>: ordered (no nans)</li>
3960 <li><tt>ueq</tt>: unordered or equal</li>
3961 <li><tt>ugt</tt>: unordered or greater than </li>
3962 <li><tt>uge</tt>: unordered or greater than or equal</li>
3963 <li><tt>ult</tt>: unordered or less than </li>
3964 <li><tt>ule</tt>: unordered or less than or equal</li>
3965 <li><tt>une</tt>: unordered or not equal</li>
3966 <li><tt>uno</tt>: unordered (either nans)</li>
3967 <li><tt>true</tt>: no comparison, always returns true</li>
3968</ol>
3969<p><i>Ordered</i> means that neither operand is a QNAN while
3970<i>unordered</i> means that either operand may be a QNAN.</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3971<p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
3972either a <a href="#t_floating">floating point</a> type
3973or a <a href="#t_vector">vector</a> of floating point type.
3974They must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3975<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]3976<p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]3977according to the condition code given as <tt>cond</tt>.
3978If the operands are vectors, then the vectors are compared
3979element by element.
3980Each comparison performed
3981always yields an <a href="#t_primitive">i1</a> result, as follows:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3982<ol>
3983 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3984 <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]3985 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3986 <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]3987 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3988 <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]3989 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3990 <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]3991 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3992 <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]3993 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3994 <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]3995 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3996 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3997 <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]3998 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3999 <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]4000 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4001 <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]4002 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4003 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4004 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4005 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4006 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4007 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4008 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4009 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4010 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4011</ol>
4012
4013<h5>Example:</h5>
4014<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]4015 &lt;result&gt; = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4016 &lt;result&gt; = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4017 &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]4018</pre>
4019</div>
4020
4021<!-- _______________________________________________________________________ -->
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4022<div class="doc_subsubsection">
4023 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
4024</div>
4025<div class="doc_text">
4026<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4027<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]4028</pre>
4029<h5>Overview:</h5>
4030<p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
4031element-wise comparison of its two integer vector operands.</p>
4032<h5>Arguments:</h5>
4033<p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
4034the condition code indicating the kind of comparison to perform. It is not
4035a value, just a keyword. The possible condition code are:
4036<ol>
4037 <li><tt>eq</tt>: equal</li>
4038 <li><tt>ne</tt>: not equal </li>
4039 <li><tt>ugt</tt>: unsigned greater than</li>
4040 <li><tt>uge</tt>: unsigned greater or equal</li>
4041 <li><tt>ult</tt>: unsigned less than</li>
4042 <li><tt>ule</tt>: unsigned less or equal</li>
4043 <li><tt>sgt</tt>: signed greater than</li>
4044 <li><tt>sge</tt>: signed greater or equal</li>
4045 <li><tt>slt</tt>: signed less than</li>
4046 <li><tt>sle</tt>: signed less or equal</li>
4047</ol>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4048<p>The remaining two arguments must be <a href="#t_vector">vector</a> or
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4049<a href="#t_integer">integer</a> typed. They must also be identical types.</p>
4050<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4051<p>The '<tt>vicmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4052according to the condition code given as <tt>cond</tt>. The comparison yields a
4053<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
4054identical type as the values being compared. The most significant bit in each
4055element is 1 if the element-wise comparison evaluates to true, and is 0
4056otherwise. All other bits of the result are undefined. The condition codes
4057are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
4058instruction</a>.
4059
4060<h5>Example:</h5>
4061<pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4062 &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>
4063 &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]4064</pre>
4065</div>
4066
4067<!-- _______________________________________________________________________ -->
4068<div class="doc_subsubsection">
4069 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4070</div>
4071<div class="doc_text">
4072<h5>Syntax:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4073<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]4074<h5>Overview:</h5>
4075<p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4076element-wise comparison of its two floating point vector operands. The output
4077elements have the same width as the input elements.</p>
4078<h5>Arguments:</h5>
4079<p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4080the condition code indicating the kind of comparison to perform. It is not
4081a value, just a keyword. The possible condition code are:
4082<ol>
4083 <li><tt>false</tt>: no comparison, always returns false</li>
4084 <li><tt>oeq</tt>: ordered and equal</li>
4085 <li><tt>ogt</tt>: ordered and greater than </li>
4086 <li><tt>oge</tt>: ordered and greater than or equal</li>
4087 <li><tt>olt</tt>: ordered and less than </li>
4088 <li><tt>ole</tt>: ordered and less than or equal</li>
4089 <li><tt>one</tt>: ordered and not equal</li>
4090 <li><tt>ord</tt>: ordered (no nans)</li>
4091 <li><tt>ueq</tt>: unordered or equal</li>
4092 <li><tt>ugt</tt>: unordered or greater than </li>
4093 <li><tt>uge</tt>: unordered or greater than or equal</li>
4094 <li><tt>ult</tt>: unordered or less than </li>
4095 <li><tt>ule</tt>: unordered or less than or equal</li>
4096 <li><tt>une</tt>: unordered or not equal</li>
4097 <li><tt>uno</tt>: unordered (either nans)</li>
4098 <li><tt>true</tt>: no comparison, always returns true</li>
4099</ol>
4100<p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4101<a href="#t_floating">floating point</a> typed. They must also be identical
4102types.</p>
4103<h5>Semantics:</h5>
Gabor Greifd9068fe2008-08-07 21:46:00 +0000[diff] [blame]4104<p>The '<tt>vfcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]4105according to the condition code given as <tt>cond</tt>. The comparison yields a
4106<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4107an identical number of elements as the values being compared, and each element
4108having identical with to the width of the floating point elements. The most
4109significant bit in each element is 1 if the element-wise comparison evaluates to
4110true, and is 0 otherwise. All other bits of the result are undefined. The
4111condition codes are evaluated identically to the
4112<a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.
4113
4114<h5>Example:</h5>
4115<pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4116 &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>
4117 &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]4118</pre>
4119</div>
4120
4121<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4122<div class="doc_subsubsection">
4123 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4124</div>
4125
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4126<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4127
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4128<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4129
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4130<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
4131<h5>Overview:</h5>
4132<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
4133the SSA graph representing the function.</p>
4134<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4135
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4136<p>The type of the incoming values is specified with the first type
4137field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4138as arguments, with one pair for each predecessor basic block of the
4139current block. Only values of <a href="#t_firstclass">first class</a>
4140type may be used as the value arguments to the PHI node. Only labels
4141may be used as the label arguments.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4142
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4143<p>There must be no non-phi instructions between the start of a basic
4144block and the PHI instructions: i.e. PHI instructions must be first in
4145a basic block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4146
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4147<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4148
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4149<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4150specified by the pair corresponding to the predecessor basic block that executed
4151just prior to the current block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4152
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4153<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4154<pre>
4155Loop: ; Infinite loop that counts from 0 on up...
4156 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4157 %nextindvar = add i32 %indvar, 1
4158 br label %Loop
4159</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4160</div>
4161
4162<!-- _______________________________________________________________________ -->
4163<div class="doc_subsubsection">
4164 <a name="i_select">'<tt>select</tt>' Instruction</a>
4165</div>
4166
4167<div class="doc_text">
4168
4169<h5>Syntax:</h5>
4170
4171<pre>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4172 &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>
4173
4174 <i>selty</i> is either i1 or {&lt;N x i1&gt}
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4175</pre>
4176
4177<h5>Overview:</h5>
4178
4179<p>
4180The '<tt>select</tt>' instruction is used to choose one value based on a
4181condition, without branching.
4182</p>
4183
4184
4185<h5>Arguments:</h5>
4186
4187<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4188The '<tt>select</tt>' instruction requires an 'i1' value or
4189a vector of 'i1' values indicating the
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4190condition, and two values of the same <a href="#t_firstclass">first class</a>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4191type. If the val1/val2 are vectors and
4192the condition is a scalar, then entire vectors are selected, not
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4193individual elements.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4194</p>
4195
4196<h5>Semantics:</h5>
4197
4198<p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4199If 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]4200value argument; otherwise, it returns the second value argument.
4201</p>
Dan Gohmanb60ca3c2008-09-09 01:02:47 +0000[diff] [blame]4202<p>
4203If the condition is a vector of i1, then the value arguments must
4204be vectors of the same size, and the selection is done element
4205by element.
4206</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4207
4208<h5>Example:</h5>
4209
4210<pre>
4211 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4212</pre>
4213</div>
4214
4215
4216<!-- _______________________________________________________________________ -->
4217<div class="doc_subsubsection">
4218 <a name="i_call">'<tt>call</tt>' Instruction</a>
4219</div>
4220
4221<div class="doc_text">
4222
4223<h5>Syntax:</h5>
4224<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4225 &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]4226</pre>
4227
4228<h5>Overview:</h5>
4229
4230<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4231
4232<h5>Arguments:</h5>
4233
4234<p>This instruction requires several arguments:</p>
4235
4236<ol>
4237 <li>
4238 <p>The optional "tail" marker indicates whether the callee function accesses
4239 any allocas or varargs in the caller. If the "tail" marker is present, the
4240 function call is eligible for tail call optimization. Note that calls may
4241 be marked "tail" even if they do not occur before a <a
4242 href="#i_ret"><tt>ret</tt></a> instruction.
4243 </li>
4244 <li>
4245 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4246 convention</a> the call should use. If none is specified, the call defaults
4247 to using C calling conventions.
4248 </li>
4249 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4250 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4251 the type of the return value. Functions that return no value are marked
4252 <tt><a href="#t_void">void</a></tt>.</p>
4253 </li>
4254 <li>
4255 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4256 value being invoked. The argument types must match the types implied by
4257 this signature. This type can be omitted if the function is not varargs
4258 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4259 </li>
4260 <li>
4261 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4262 be invoked. In most cases, this is a direct function invocation, but
4263 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4264 to function value.</p>
4265 </li>
4266 <li>
4267 <p>'<tt>function args</tt>': argument list whose types match the
4268 function signature argument types. All arguments must be of
4269 <a href="#t_firstclass">first class</a> type. If the function signature
4270 indicates the function accepts a variable number of arguments, the extra
4271 arguments can be specified.</p>
4272 </li>
4273</ol>
4274
4275<h5>Semantics:</h5>
4276
4277<p>The '<tt>call</tt>' instruction is used to cause control flow to
4278transfer to a specified function, with its incoming arguments bound to
4279the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4280instruction in the called function, control flow continues with the
4281instruction after the function call, and the return value of the
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4282function is bound to the result argument. If the callee returns multiple
4283values then the return values of the function are only accessible through
4284the '<tt><a href="#i_getresult">getresult</a></tt>' instruction.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4285
4286<h5>Example:</h5>
4287
4288<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4289 %retval = call i32 @test(i32 %argc)
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4290 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4291 %X = tail call i32 @foo() <i>; yields i32</i>
4292 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4293 call void %foo(i8 97 signext)
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4294
4295 %struct.A = type { i32, i8 }
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4296 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
4297 %gr = getresult %struct.A %r, 0 <i>; yields i32</i>
4298 %gr1 = getresult %struct.A %r, 1 <i>; yields i8</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4299</pre>
4300
4301</div>
4302
4303<!-- _______________________________________________________________________ -->
4304<div class="doc_subsubsection">
4305 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4306</div>
4307
4308<div class="doc_text">
4309
4310<h5>Syntax:</h5>
4311
4312<pre>
4313 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
4314</pre>
4315
4316<h5>Overview:</h5>
4317
4318<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4319the "variable argument" area of a function call. It is used to implement the
4320<tt>va_arg</tt> macro in C.</p>
4321
4322<h5>Arguments:</h5>
4323
4324<p>This instruction takes a <tt>va_list*</tt> value and the type of
4325the argument. It returns a value of the specified argument type and
4326increments the <tt>va_list</tt> to point to the next argument. The
4327actual type of <tt>va_list</tt> is target specific.</p>
4328
4329<h5>Semantics:</h5>
4330
4331<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4332type from the specified <tt>va_list</tt> and causes the
4333<tt>va_list</tt> to point to the next argument. For more information,
4334see the variable argument handling <a href="#int_varargs">Intrinsic
4335Functions</a>.</p>
4336
4337<p>It is legal for this instruction to be called in a function which does not
4338take a variable number of arguments, for example, the <tt>vfprintf</tt>
4339function.</p>
4340
4341<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4342href="#intrinsics">intrinsic function</a> because it takes a type as an
4343argument.</p>
4344
4345<h5>Example:</h5>
4346
4347<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4348
4349</div>
4350
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4351<!-- _______________________________________________________________________ -->
4352<div class="doc_subsubsection">
4353 <a name="i_getresult">'<tt>getresult</tt>' Instruction</a>
4354</div>
4355
4356<div class="doc_text">
4357
4358<h5>Syntax:</h5>
4359<pre>
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4360 &lt;resultval&gt; = getresult &lt;type&gt; &lt;retval&gt;, &lt;index&gt;
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4361</pre>
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4362
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4363<h5>Overview:</h5>
4364
4365<p> The '<tt>getresult</tt>' instruction is used to extract individual values
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4366from a '<tt><a href="#i_call">call</a></tt>'
4367or '<tt><a href="#i_invoke">invoke</a></tt>' instruction that returns multiple
4368results.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4369
4370<h5>Arguments:</h5>
4371
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4372<p>The '<tt>getresult</tt>' instruction takes a call or invoke value as its
Chris Lattnerd8dd3522008-04-23 04:06:52 +0000[diff] [blame]4373first argument, or an undef value. The value must have <a
4374href="#t_struct">structure type</a>. The second argument is a constant
4375unsigned index value which must be in range for the number of values returned
4376by the call.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4377
4378<h5>Semantics:</h5>
4379
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4380<p>The '<tt>getresult</tt>' instruction extracts the element identified by
4381'<tt>index</tt>' from the aggregate value.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4382
4383<h5>Example:</h5>
4384
4385<pre>
4386 %struct.A = type { i32, i8 }
4387
4388 %r = call %struct.A @foo()
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4389 %gr = getresult %struct.A %r, 0 <i>; yields i32:%gr</i>
4390 %gr1 = getresult %struct.A %r, 1 <i>; yields i8:%gr1</i>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4391 add i32 %gr, 42
4392 add i8 %gr1, 41
4393</pre>
4394
4395</div>
4396
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4397<!-- *********************************************************************** -->
4398<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4399<!-- *********************************************************************** -->
4400
4401<div class="doc_text">
4402
4403<p>LLVM supports the notion of an "intrinsic function". These functions have
4404well known names and semantics and are required to follow certain restrictions.
4405Overall, these intrinsics represent an extension mechanism for the LLVM
4406language that does not require changing all of the transformations in LLVM when
4407adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4408
4409<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4410prefix is reserved in LLVM for intrinsic names; thus, function names may not
4411begin with this prefix. Intrinsic functions must always be external functions:
4412you cannot define the body of intrinsic functions. Intrinsic functions may
4413only be used in call or invoke instructions: it is illegal to take the address
4414of an intrinsic function. Additionally, because intrinsic functions are part
4415of the LLVM language, it is required if any are added that they be documented
4416here.</p>
4417
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4418<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4419a family of functions that perform the same operation but on different data
4420types. Because LLVM can represent over 8 million different integer types,
4421overloading is used commonly to allow an intrinsic function to operate on any
4422integer type. One or more of the argument types or the result type can be
4423overloaded to accept any integer type. Argument types may also be defined as
4424exactly matching a previous argument's type or the result type. This allows an
4425intrinsic function which accepts multiple arguments, but needs all of them to
4426be of the same type, to only be overloaded with respect to a single argument or
4427the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4428
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4429<p>Overloaded intrinsics will have the names of its overloaded argument types
4430encoded into its function name, each preceded by a period. Only those types
4431which are overloaded result in a name suffix. Arguments whose type is matched
4432against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4433take an integer of any width and returns an integer of exactly the same integer
4434width. This leads to a family of functions such as
4435<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4436Only one type, the return type, is overloaded, and only one type suffix is
4437required. Because the argument's type is matched against the return type, it
4438does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4439
4440<p>To learn how to add an intrinsic function, please see the
4441<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4442</p>
4443
4444</div>
4445
4446<!-- ======================================================================= -->
4447<div class="doc_subsection">
4448 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4449</div>
4450
4451<div class="doc_text">
4452
4453<p>Variable argument support is defined in LLVM with the <a
4454 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4455intrinsic functions. These functions are related to the similarly
4456named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
4457
4458<p>All of these functions operate on arguments that use a
4459target-specific value type "<tt>va_list</tt>". The LLVM assembly
4460language reference manual does not define what this type is, so all
4461transformations should be prepared to handle these functions regardless of
4462the type used.</p>
4463
4464<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4465instruction and the variable argument handling intrinsic functions are
4466used.</p>
4467
4468<div class="doc_code">
4469<pre>
4470define i32 @test(i32 %X, ...) {
4471 ; Initialize variable argument processing
4472 %ap = alloca i8*
4473 %ap2 = bitcast i8** %ap to i8*
4474 call void @llvm.va_start(i8* %ap2)
4475
4476 ; Read a single integer argument
4477 %tmp = va_arg i8** %ap, i32
4478
4479 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4480 %aq = alloca i8*
4481 %aq2 = bitcast i8** %aq to i8*
4482 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4483 call void @llvm.va_end(i8* %aq2)
4484
4485 ; Stop processing of arguments.
4486 call void @llvm.va_end(i8* %ap2)
4487 ret i32 %tmp
4488}
4489
4490declare void @llvm.va_start(i8*)
4491declare void @llvm.va_copy(i8*, i8*)
4492declare void @llvm.va_end(i8*)
4493</pre>
4494</div>
4495
4496</div>
4497
4498<!-- _______________________________________________________________________ -->
4499<div class="doc_subsubsection">
4500 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4501</div>
4502
4503
4504<div class="doc_text">
4505<h5>Syntax:</h5>
4506<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
4507<h5>Overview:</h5>
4508<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
4509<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
4510href="#i_va_arg">va_arg</a></tt>.</p>
4511
4512<h5>Arguments:</h5>
4513
4514<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
4515
4516<h5>Semantics:</h5>
4517
4518<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
4519macro available in C. In a target-dependent way, it initializes the
4520<tt>va_list</tt> element to which the argument points, so that the next call to
4521<tt>va_arg</tt> will produce the first variable argument passed to the function.
4522Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4523last argument of the function as the compiler can figure that out.</p>
4524
4525</div>
4526
4527<!-- _______________________________________________________________________ -->
4528<div class="doc_subsubsection">
4529 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4530</div>
4531
4532<div class="doc_text">
4533<h5>Syntax:</h5>
4534<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
4535<h5>Overview:</h5>
4536
4537<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
4538which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4539or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4540
4541<h5>Arguments:</h5>
4542
4543<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4544
4545<h5>Semantics:</h5>
4546
4547<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4548macro available in C. In a target-dependent way, it destroys the
4549<tt>va_list</tt> element to which the argument points. Calls to <a
4550href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4551<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4552<tt>llvm.va_end</tt>.</p>
4553
4554</div>
4555
4556<!-- _______________________________________________________________________ -->
4557<div class="doc_subsubsection">
4558 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4559</div>
4560
4561<div class="doc_text">
4562
4563<h5>Syntax:</h5>
4564
4565<pre>
4566 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
4567</pre>
4568
4569<h5>Overview:</h5>
4570
4571<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4572from the source argument list to the destination argument list.</p>
4573
4574<h5>Arguments:</h5>
4575
4576<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4577The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4578
4579
4580<h5>Semantics:</h5>
4581
4582<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4583macro available in C. In a target-dependent way, it copies the source
4584<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4585intrinsic is necessary because the <tt><a href="#int_va_start">
4586llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4587example, memory allocation.</p>
4588
4589</div>
4590
4591<!-- ======================================================================= -->
4592<div class="doc_subsection">
4593 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4594</div>
4595
4596<div class="doc_text">
4597
4598<p>
4599LLVM support for <a href="GarbageCollection.html">Accurate Garbage
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]4600Collection</a> (GC) requires the implementation and generation of these
4601intrinsics.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4602These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4603stack</a>, as well as garbage collector implementations that require <a
4604href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4605Front-ends for type-safe garbage collected languages should generate these
4606intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4607href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4608</p>
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]4609
4610<p>The garbage collection intrinsics only operate on objects in the generic
4611 address space (address space zero).</p>
4612
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4613</div>
4614
4615<!-- _______________________________________________________________________ -->
4616<div class="doc_subsubsection">
4617 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4618</div>
4619
4620<div class="doc_text">
4621
4622<h5>Syntax:</h5>
4623
4624<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4625 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4626</pre>
4627
4628<h5>Overview:</h5>
4629
4630<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4631the code generator, and allows some metadata to be associated with it.</p>
4632
4633<h5>Arguments:</h5>
4634
4635<p>The first argument specifies the address of a stack object that contains the
4636root pointer. The second pointer (which must be either a constant or a global
4637value address) contains the meta-data to be associated with the root.</p>
4638
4639<h5>Semantics:</h5>
4640
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]4641<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]4642location. At compile-time, the code generator generates information to allow
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4643the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4644intrinsic may only be used in a function which <a href="#gc">specifies a GC
4645algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4646
4647</div>
4648
4649
4650<!-- _______________________________________________________________________ -->
4651<div class="doc_subsubsection">
4652 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4653</div>
4654
4655<div class="doc_text">
4656
4657<h5>Syntax:</h5>
4658
4659<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4660 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4661</pre>
4662
4663<h5>Overview:</h5>
4664
4665<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4666locations, allowing garbage collector implementations that require read
4667barriers.</p>
4668
4669<h5>Arguments:</h5>
4670
4671<p>The second argument is the address to read from, which should be an address
4672allocated from the garbage collector. The first object is a pointer to the
4673start of the referenced object, if needed by the language runtime (otherwise
4674null).</p>
4675
4676<h5>Semantics:</h5>
4677
4678<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4679instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4680garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4681may only be used in a function which <a href="#gc">specifies a GC
4682algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4683
4684</div>
4685
4686
4687<!-- _______________________________________________________________________ -->
4688<div class="doc_subsubsection">
4689 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4690</div>
4691
4692<div class="doc_text">
4693
4694<h5>Syntax:</h5>
4695
4696<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4697 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4698</pre>
4699
4700<h5>Overview:</h5>
4701
4702<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4703locations, allowing garbage collector implementations that require write
4704barriers (such as generational or reference counting collectors).</p>
4705
4706<h5>Arguments:</h5>
4707
4708<p>The first argument is the reference to store, the second is the start of the
4709object to store it to, and the third is the address of the field of Obj to
4710store to. If the runtime does not require a pointer to the object, Obj may be
4711null.</p>
4712
4713<h5>Semantics:</h5>
4714
4715<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4716instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4717garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4718may only be used in a function which <a href="#gc">specifies a GC
4719algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4720
4721</div>
4722
4723
4724
4725<!-- ======================================================================= -->
4726<div class="doc_subsection">
4727 <a name="int_codegen">Code Generator Intrinsics</a>
4728</div>
4729
4730<div class="doc_text">
4731<p>
4732These intrinsics are provided by LLVM to expose special features that may only
4733be implemented with code generator support.
4734</p>
4735
4736</div>
4737
4738<!-- _______________________________________________________________________ -->
4739<div class="doc_subsubsection">
4740 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4741</div>
4742
4743<div class="doc_text">
4744
4745<h5>Syntax:</h5>
4746<pre>
4747 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
4748</pre>
4749
4750<h5>Overview:</h5>
4751
4752<p>
4753The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4754target-specific value indicating the return address of the current function
4755or one of its callers.
4756</p>
4757
4758<h5>Arguments:</h5>
4759
4760<p>
4761The argument to this intrinsic indicates which function to return the address
4762for. Zero indicates the calling function, one indicates its caller, etc. The
4763argument is <b>required</b> to be a constant integer value.
4764</p>
4765
4766<h5>Semantics:</h5>
4767
4768<p>
4769The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4770the return address of the specified call frame, or zero if it cannot be
4771identified. The value returned by this intrinsic is likely to be incorrect or 0
4772for arguments other than zero, so it should only be used for debugging purposes.
4773</p>
4774
4775<p>
4776Note that calling this intrinsic does not prevent function inlining or other
4777aggressive transformations, so the value returned may not be that of the obvious
4778source-language caller.
4779</p>
4780</div>
4781
4782
4783<!-- _______________________________________________________________________ -->
4784<div class="doc_subsubsection">
4785 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4786</div>
4787
4788<div class="doc_text">
4789
4790<h5>Syntax:</h5>
4791<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4792 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4793</pre>
4794
4795<h5>Overview:</h5>
4796
4797<p>
4798The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4799target-specific frame pointer value for the specified stack frame.
4800</p>
4801
4802<h5>Arguments:</h5>
4803
4804<p>
4805The argument to this intrinsic indicates which function to return the frame
4806pointer for. Zero indicates the calling function, one indicates its caller,
4807etc. The argument is <b>required</b> to be a constant integer value.
4808</p>
4809
4810<h5>Semantics:</h5>
4811
4812<p>
4813The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4814the frame address of the specified call frame, or zero if it cannot be
4815identified. The value returned by this intrinsic is likely to be incorrect or 0
4816for arguments other than zero, so it should only be used for debugging purposes.
4817</p>
4818
4819<p>
4820Note that calling this intrinsic does not prevent function inlining or other
4821aggressive transformations, so the value returned may not be that of the obvious
4822source-language caller.
4823</p>
4824</div>
4825
4826<!-- _______________________________________________________________________ -->
4827<div class="doc_subsubsection">
4828 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4829</div>
4830
4831<div class="doc_text">
4832
4833<h5>Syntax:</h5>
4834<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4835 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4836</pre>
4837
4838<h5>Overview:</h5>
4839
4840<p>
4841The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4842the function stack, for use with <a href="#int_stackrestore">
4843<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4844features like scoped automatic variable sized arrays in C99.
4845</p>
4846
4847<h5>Semantics:</h5>
4848
4849<p>
4850This intrinsic returns a opaque pointer value that can be passed to <a
4851href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4852<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4853<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4854state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4855practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4856that were allocated after the <tt>llvm.stacksave</tt> was executed.
4857</p>
4858
4859</div>
4860
4861<!-- _______________________________________________________________________ -->
4862<div class="doc_subsubsection">
4863 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4864</div>
4865
4866<div class="doc_text">
4867
4868<h5>Syntax:</h5>
4869<pre>
4870 declare void @llvm.stackrestore(i8 * %ptr)
4871</pre>
4872
4873<h5>Overview:</h5>
4874
4875<p>
4876The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4877the function stack to the state it was in when the corresponding <a
4878href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4879useful for implementing language features like scoped automatic variable sized
4880arrays in C99.
4881</p>
4882
4883<h5>Semantics:</h5>
4884
4885<p>
4886See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4887</p>
4888
4889</div>
4890
4891
4892<!-- _______________________________________________________________________ -->
4893<div class="doc_subsubsection">
4894 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4895</div>
4896
4897<div class="doc_text">
4898
4899<h5>Syntax:</h5>
4900<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4901 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]4902</pre>
4903
4904<h5>Overview:</h5>
4905
4906
4907<p>
4908The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4909a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4910no
4911effect on the behavior of the program but can change its performance
4912characteristics.
4913</p>
4914
4915<h5>Arguments:</h5>
4916
4917<p>
4918<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4919determining if the fetch should be for a read (0) or write (1), and
4920<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4921locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4922<tt>locality</tt> arguments must be constant integers.
4923</p>
4924
4925<h5>Semantics:</h5>
4926
4927<p>
4928This intrinsic does not modify the behavior of the program. In particular,
4929prefetches cannot trap and do not produce a value. On targets that support this
4930intrinsic, the prefetch can provide hints to the processor cache for better
4931performance.
4932</p>
4933
4934</div>
4935
4936<!-- _______________________________________________________________________ -->
4937<div class="doc_subsubsection">
4938 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4939</div>
4940
4941<div class="doc_text">
4942
4943<h5>Syntax:</h5>
4944<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4945 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4946</pre>
4947
4948<h5>Overview:</h5>
4949
4950
4951<p>
4952The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]4953(PC) in a region of
4954code to simulators and other tools. The method is target specific, but it is
4955expected that the marker will use exported symbols to transmit the PC of the
4956marker.
4957The marker makes no guarantees that it will remain with any specific instruction
4958after optimizations. It is possible that the presence of a marker will inhibit
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4959optimizations. The intended use is to be inserted after optimizations to allow
4960correlations of simulation runs.
4961</p>
4962
4963<h5>Arguments:</h5>
4964
4965<p>
4966<tt>id</tt> is a numerical id identifying the marker.
4967</p>
4968
4969<h5>Semantics:</h5>
4970
4971<p>
4972This intrinsic does not modify the behavior of the program. Backends that do not
4973support this intrinisic may ignore it.
4974</p>
4975
4976</div>
4977
4978<!-- _______________________________________________________________________ -->
4979<div class="doc_subsubsection">
4980 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4981</div>
4982
4983<div class="doc_text">
4984
4985<h5>Syntax:</h5>
4986<pre>
4987 declare i64 @llvm.readcyclecounter( )
4988</pre>
4989
4990<h5>Overview:</h5>
4991
4992
4993<p>
4994The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4995counter register (or similar low latency, high accuracy clocks) on those targets
4996that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4997As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4998should only be used for small timings.
4999</p>
5000
5001<h5>Semantics:</h5>
5002
5003<p>
5004When directly supported, reading the cycle counter should not modify any memory.
5005Implementations are allowed to either return a application specific value or a
5006system wide value. On backends without support, this is lowered to a constant 0.
5007</p>
5008
5009</div>
5010
5011<!-- ======================================================================= -->
5012<div class="doc_subsection">
5013 <a name="int_libc">Standard C Library Intrinsics</a>
5014</div>
5015
5016<div class="doc_text">
5017<p>
5018LLVM provides intrinsics for a few important standard C library functions.
5019These intrinsics allow source-language front-ends to pass information about the
5020alignment of the pointer arguments to the code generator, providing opportunity
5021for more efficient code generation.
5022</p>
5023
5024</div>
5025
5026<!-- _______________________________________________________________________ -->
5027<div class="doc_subsubsection">
5028 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5029</div>
5030
5031<div class="doc_text">
5032
5033<h5>Syntax:</h5>
5034<pre>
5035 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5036 i32 &lt;len&gt;, i32 &lt;align&gt;)
5037 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5038 i64 &lt;len&gt;, i32 &lt;align&gt;)
5039</pre>
5040
5041<h5>Overview:</h5>
5042
5043<p>
5044The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5045location to the destination location.
5046</p>
5047
5048<p>
5049Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5050intrinsics do not return a value, and takes an extra alignment argument.
5051</p>
5052
5053<h5>Arguments:</h5>
5054
5055<p>
5056The first argument is a pointer to the destination, the second is a pointer to
5057the source. The third argument is an integer argument
5058specifying the number of bytes to copy, and the fourth argument is the alignment
5059of the source and destination locations.
5060</p>
5061
5062<p>
5063If the call to this intrinisic has an alignment value that is not 0 or 1, then
5064the caller guarantees that both the source and destination pointers are aligned
5065to that boundary.
5066</p>
5067
5068<h5>Semantics:</h5>
5069
5070<p>
5071The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5072location to the destination location, which are not allowed to overlap. It
5073copies "len" bytes of memory over. If the argument is known to be aligned to
5074some boundary, this can be specified as the fourth argument, otherwise it should
5075be set to 0 or 1.
5076</p>
5077</div>
5078
5079
5080<!-- _______________________________________________________________________ -->
5081<div class="doc_subsubsection">
5082 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5083</div>
5084
5085<div class="doc_text">
5086
5087<h5>Syntax:</h5>
5088<pre>
5089 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5090 i32 &lt;len&gt;, i32 &lt;align&gt;)
5091 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5092 i64 &lt;len&gt;, i32 &lt;align&gt;)
5093</pre>
5094
5095<h5>Overview:</h5>
5096
5097<p>
5098The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5099location to the destination location. It is similar to the
Chris Lattnerdba16ea2008-01-06 19:51:52 +0000[diff] [blame]5100'<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5101</p>
5102
5103<p>
5104Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5105intrinsics do not return a value, and takes an extra alignment argument.
5106</p>
5107
5108<h5>Arguments:</h5>
5109
5110<p>
5111The first argument is a pointer to the destination, the second is a pointer to
5112the source. The third argument is an integer argument
5113specifying the number of bytes to copy, and the fourth argument is the alignment
5114of the source and destination locations.
5115</p>
5116
5117<p>
5118If the call to this intrinisic has an alignment value that is not 0 or 1, then
5119the caller guarantees that the source and destination pointers are aligned to
5120that boundary.
5121</p>
5122
5123<h5>Semantics:</h5>
5124
5125<p>
5126The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5127location to the destination location, which may overlap. It
5128copies "len" bytes of memory over. If the argument is known to be aligned to
5129some boundary, this can be specified as the fourth argument, otherwise it should
5130be set to 0 or 1.
5131</p>
5132</div>
5133
5134
5135<!-- _______________________________________________________________________ -->
5136<div class="doc_subsubsection">
5137 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5138</div>
5139
5140<div class="doc_text">
5141
5142<h5>Syntax:</h5>
5143<pre>
5144 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5145 i32 &lt;len&gt;, i32 &lt;align&gt;)
5146 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5147 i64 &lt;len&gt;, i32 &lt;align&gt;)
5148</pre>
5149
5150<h5>Overview:</h5>
5151
5152<p>
5153The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5154byte value.
5155</p>
5156
5157<p>
5158Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5159does not return a value, and takes an extra alignment argument.
5160</p>
5161
5162<h5>Arguments:</h5>
5163
5164<p>
5165The first argument is a pointer to the destination to fill, the second is the
5166byte value to fill it with, the third argument is an integer
5167argument specifying the number of bytes to fill, and the fourth argument is the
5168known alignment of destination location.
5169</p>
5170
5171<p>
5172If the call to this intrinisic has an alignment value that is not 0 or 1, then
5173the caller guarantees that the destination pointer is aligned to that boundary.
5174</p>
5175
5176<h5>Semantics:</h5>
5177
5178<p>
5179The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5180the
5181destination location. If the argument is known to be aligned to some boundary,
5182this can be specified as the fourth argument, otherwise it should be set to 0 or
51831.
5184</p>
5185</div>
5186
5187
5188<!-- _______________________________________________________________________ -->
5189<div class="doc_subsubsection">
5190 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5191</div>
5192
5193<div class="doc_text">
5194
5195<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5196<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]5197floating point or vector of floating point type. Not all targets support all
5198types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5199<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5200 declare float @llvm.sqrt.f32(float %Val)
5201 declare double @llvm.sqrt.f64(double %Val)
5202 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5203 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5204 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5205</pre>
5206
5207<h5>Overview:</h5>
5208
5209<p>
5210The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5211returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5212<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
Chris Lattnerf914a002008-01-29 07:00:44 +0000[diff] [blame]5213negative numbers other than -0.0 (which allows for better optimization, because
5214there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5215defined to return -0.0 like IEEE sqrt.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5216</p>
5217
5218<h5>Arguments:</h5>
5219
5220<p>
5221The argument and return value are floating point numbers of the same type.
5222</p>
5223
5224<h5>Semantics:</h5>
5225
5226<p>
5227This function returns the sqrt of the specified operand if it is a nonnegative
5228floating point number.
5229</p>
5230</div>
5231
5232<!-- _______________________________________________________________________ -->
5233<div class="doc_subsubsection">
5234 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5235</div>
5236
5237<div class="doc_text">
5238
5239<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5240<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]5241floating point or vector of floating point type. Not all targets support all
5242types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5243<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5244 declare float @llvm.powi.f32(float %Val, i32 %power)
5245 declare double @llvm.powi.f64(double %Val, i32 %power)
5246 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5247 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5248 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5249</pre>
5250
5251<h5>Overview:</h5>
5252
5253<p>
5254The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5255specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5256multiplications is not defined. When a vector of floating point type is
5257used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5258</p>
5259
5260<h5>Arguments:</h5>
5261
5262<p>
5263The second argument is an integer power, and the first is a value to raise to
5264that power.
5265</p>
5266
5267<h5>Semantics:</h5>
5268
5269<p>
5270This function returns the first value raised to the second power with an
5271unspecified sequence of rounding operations.</p>
5272</div>
5273
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5274<!-- _______________________________________________________________________ -->
5275<div class="doc_subsubsection">
5276 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5277</div>
5278
5279<div class="doc_text">
5280
5281<h5>Syntax:</h5>
5282<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5283floating point or vector of floating point type. Not all targets support all
5284types however.
5285<pre>
5286 declare float @llvm.sin.f32(float %Val)
5287 declare double @llvm.sin.f64(double %Val)
5288 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5289 declare fp128 @llvm.sin.f128(fp128 %Val)
5290 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5291</pre>
5292
5293<h5>Overview:</h5>
5294
5295<p>
5296The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5297</p>
5298
5299<h5>Arguments:</h5>
5300
5301<p>
5302The argument and return value are floating point numbers of the same type.
5303</p>
5304
5305<h5>Semantics:</h5>
5306
5307<p>
5308This function returns the sine of the specified operand, returning the
5309same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5310conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5311</div>
5312
5313<!-- _______________________________________________________________________ -->
5314<div class="doc_subsubsection">
5315 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5316</div>
5317
5318<div class="doc_text">
5319
5320<h5>Syntax:</h5>
5321<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5322floating point or vector of floating point type. Not all targets support all
5323types however.
5324<pre>
5325 declare float @llvm.cos.f32(float %Val)
5326 declare double @llvm.cos.f64(double %Val)
5327 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5328 declare fp128 @llvm.cos.f128(fp128 %Val)
5329 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5330</pre>
5331
5332<h5>Overview:</h5>
5333
5334<p>
5335The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5336</p>
5337
5338<h5>Arguments:</h5>
5339
5340<p>
5341The argument and return value are floating point numbers of the same type.
5342</p>
5343
5344<h5>Semantics:</h5>
5345
5346<p>
5347This function returns the cosine of the specified operand, returning the
5348same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5349conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5350</div>
5351
5352<!-- _______________________________________________________________________ -->
5353<div class="doc_subsubsection">
5354 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5355</div>
5356
5357<div class="doc_text">
5358
5359<h5>Syntax:</h5>
5360<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5361floating point or vector of floating point type. Not all targets support all
5362types however.
5363<pre>
5364 declare float @llvm.pow.f32(float %Val, float %Power)
5365 declare double @llvm.pow.f64(double %Val, double %Power)
5366 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5367 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5368 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5369</pre>
5370
5371<h5>Overview:</h5>
5372
5373<p>
5374The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5375specified (positive or negative) power.
5376</p>
5377
5378<h5>Arguments:</h5>
5379
5380<p>
5381The second argument is a floating point power, and the first is a value to
5382raise to that power.
5383</p>
5384
5385<h5>Semantics:</h5>
5386
5387<p>
5388This function returns the first value raised to the second power,
5389returning the
5390same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5391conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5392</div>
5393
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5394
5395<!-- ======================================================================= -->
5396<div class="doc_subsection">
5397 <a name="int_manip">Bit Manipulation Intrinsics</a>
5398</div>
5399
5400<div class="doc_text">
5401<p>
5402LLVM provides intrinsics for a few important bit manipulation operations.
5403These allow efficient code generation for some algorithms.
5404</p>
5405
5406</div>
5407
5408<!-- _______________________________________________________________________ -->
5409<div class="doc_subsubsection">
5410 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5411</div>
5412
5413<div class="doc_text">
5414
5415<h5>Syntax:</h5>
5416<p>This is an overloaded intrinsic function. You can use bswap on any integer
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5417type that is an even number of bytes (i.e. BitWidth % 16 == 0).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5418<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5419 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
5420 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
5421 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5422</pre>
5423
5424<h5>Overview:</h5>
5425
5426<p>
5427The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5428values with an even number of bytes (positive multiple of 16 bits). These are
5429useful for performing operations on data that is not in the target's native
5430byte order.
5431</p>
5432
5433<h5>Semantics:</h5>
5434
5435<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5436The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5437and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5438intrinsic returns an i32 value that has the four bytes of the input i32
5439swapped, 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]5440i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5441<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5442additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5443</p>
5444
5445</div>
5446
5447<!-- _______________________________________________________________________ -->
5448<div class="doc_subsubsection">
5449 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5450</div>
5451
5452<div class="doc_text">
5453
5454<h5>Syntax:</h5>
5455<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
5456width. Not all targets support all bit widths however.
5457<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5458 declare i8 @llvm.ctpop.i8 (i8 &lt;src&gt;)
5459 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5460 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5461 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
5462 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5463</pre>
5464
5465<h5>Overview:</h5>
5466
5467<p>
5468The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5469value.
5470</p>
5471
5472<h5>Arguments:</h5>
5473
5474<p>
5475The only argument is the value to be counted. The argument may be of any
5476integer type. The return type must match the argument type.
5477</p>
5478
5479<h5>Semantics:</h5>
5480
5481<p>
5482The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5483</p>
5484</div>
5485
5486<!-- _______________________________________________________________________ -->
5487<div class="doc_subsubsection">
5488 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5489</div>
5490
5491<div class="doc_text">
5492
5493<h5>Syntax:</h5>
5494<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5495integer bit width. Not all targets support all bit widths however.
5496<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5497 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
5498 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5499 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5500 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
5501 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5502</pre>
5503
5504<h5>Overview:</h5>
5505
5506<p>
5507The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5508leading zeros in a variable.
5509</p>
5510
5511<h5>Arguments:</h5>
5512
5513<p>
5514The only argument is the value to be counted. The argument may be of any
5515integer type. The return type must match the argument type.
5516</p>
5517
5518<h5>Semantics:</h5>
5519
5520<p>
5521The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5522in a variable. If the src == 0 then the result is the size in bits of the type
5523of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5524</p>
5525</div>
5526
5527
5528
5529<!-- _______________________________________________________________________ -->
5530<div class="doc_subsubsection">
5531 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5532</div>
5533
5534<div class="doc_text">
5535
5536<h5>Syntax:</h5>
5537<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5538integer bit width. Not all targets support all bit widths however.
5539<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5540 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
5541 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5542 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5543 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
5544 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5545</pre>
5546
5547<h5>Overview:</h5>
5548
5549<p>
5550The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5551trailing zeros.
5552</p>
5553
5554<h5>Arguments:</h5>
5555
5556<p>
5557The only argument is the value to be counted. The argument may be of any
5558integer type. The return type must match the argument type.
5559</p>
5560
5561<h5>Semantics:</h5>
5562
5563<p>
5564The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5565in a variable. If the src == 0 then the result is the size in bits of the type
5566of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5567</p>
5568</div>
5569
5570<!-- _______________________________________________________________________ -->
5571<div class="doc_subsubsection">
5572 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5573</div>
5574
5575<div class="doc_text">
5576
5577<h5>Syntax:</h5>
5578<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
5579on any integer bit width.
5580<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5581 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5582 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5583</pre>
5584
5585<h5>Overview:</h5>
5586<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5587range of bits from an integer value and returns them in the same bit width as
5588the original value.</p>
5589
5590<h5>Arguments:</h5>
5591<p>The first argument, <tt>%val</tt> and the result may be integer types of
5592any bit width but they must have the same bit width. The second and third
5593arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5594
5595<h5>Semantics:</h5>
5596<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5597of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5598<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5599operates in forward mode.</p>
5600<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5601right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5602only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5603<ol>
5604 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5605 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5606 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5607 to determine the number of bits to retain.</li>
5608 <li>A mask of the retained bits is created by shifting a -1 value.</li>
5609 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
5610</ol>
5611<p>In reverse mode, a similar computation is made except that the bits are
5612returned in the reverse order. So, for example, if <tt>X</tt> has the value
5613<tt>i16 0x0ACF (101011001111)</tt> and we apply
5614<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5615<tt>i16 0x0026 (000000100110)</tt>.</p>
5616</div>
5617
5618<div class="doc_subsubsection">
5619 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5620</div>
5621
5622<div class="doc_text">
5623
5624<h5>Syntax:</h5>
5625<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
5626on any integer bit width.
5627<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5628 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5629 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]5630</pre>
5631
5632<h5>Overview:</h5>
5633<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5634of bits in an integer value with another integer value. It returns the integer
5635with the replaced bits.</p>
5636
5637<h5>Arguments:</h5>
5638<p>The first argument, <tt>%val</tt> and the result may be integer types of
5639any bit width but they must have the same bit width. <tt>%val</tt> is the value
5640whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5641integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5642type since they specify only a bit index.</p>
5643
5644<h5>Semantics:</h5>
5645<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5646of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5647<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5648operates in forward mode.</p>
5649<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5650truncating it down to the size of the replacement area or zero extending it
5651up to that size.</p>
5652<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5653are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5654in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5655to the <tt>%hi</tt>th bit.
5656<p>In reverse mode, a similar computation is made except that the bits are
5657reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5658<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
5659<h5>Examples:</h5>
5660<pre>
5661 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
5662 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
5663 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
5664 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
5665 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
5666</pre>
5667</div>
5668
5669<!-- ======================================================================= -->
5670<div class="doc_subsection">
5671 <a name="int_debugger">Debugger Intrinsics</a>
5672</div>
5673
5674<div class="doc_text">
5675<p>
5676The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5677are described in the <a
5678href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5679Debugging</a> document.
5680</p>
5681</div>
5682
5683
5684<!-- ======================================================================= -->
5685<div class="doc_subsection">
5686 <a name="int_eh">Exception Handling Intrinsics</a>
5687</div>
5688
5689<div class="doc_text">
5690<p> The LLVM exception handling intrinsics (which all start with
5691<tt>llvm.eh.</tt> prefix), are described in the <a
5692href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5693Handling</a> document. </p>
5694</div>
5695
5696<!-- ======================================================================= -->
5697<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5698 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5699</div>
5700
5701<div class="doc_text">
5702<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5703 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5704 the <tt>nest</tt> attribute, from a function. The result is a callable
5705 function pointer lacking the nest parameter - the caller does not need
5706 to provide a value for it. Instead, the value to use is stored in
5707 advance in a "trampoline", a block of memory usually allocated
5708 on the stack, which also contains code to splice the nest value into the
5709 argument list. This is used to implement the GCC nested function address
5710 extension.
5711</p>
5712<p>
5713 For example, if the function is
5714 <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]5715 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]5716<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5717 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5718 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5719 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5720 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5721</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]5722 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5723 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5724</div>
5725
5726<!-- _______________________________________________________________________ -->
5727<div class="doc_subsubsection">
5728 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5729</div>
5730<div class="doc_text">
5731<h5>Syntax:</h5>
5732<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5733declare 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]5734</pre>
5735<h5>Overview:</h5>
5736<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5737 This fills the memory pointed to by <tt>tramp</tt> with code
5738 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5739</p>
5740<h5>Arguments:</h5>
5741<p>
5742 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5743 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5744 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]5745 intrinsic. Note that the size and the alignment are target-specific - LLVM
5746 currently provides no portable way of determining them, so a front-end that
5747 generates this intrinsic needs to have some target-specific knowledge.
5748 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]5749</p>
5750<h5>Semantics:</h5>
5751<p>
5752 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5753 dependent code, turning it into a function. A pointer to this function is
5754 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5755 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5756 before being called. The new function's signature is the same as that of
5757 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5758 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5759 of pointer type. Calling the new function is equivalent to calling
5760 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5761 missing <tt>nest</tt> argument. If, after calling
5762 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5763 modified, then the effect of any later call to the returned function pointer is
5764 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5765</p>
5766</div>
5767
5768<!-- ======================================================================= -->
5769<div class="doc_subsection">
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]5770 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
5771</div>
5772
5773<div class="doc_text">
5774<p>
5775 These intrinsic functions expand the "universal IR" of LLVM to represent
5776 hardware constructs for atomic operations and memory synchronization. This
5777 provides an interface to the hardware, not an interface to the programmer. It
Chris Lattner96451482008-08-05 18:29:16 +0000[diff] [blame]5778 is aimed at a low enough level to allow any programming models or APIs
5779 (Application Programming Interfaces) which
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]5780 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
5781 hardware behavior. Just as hardware provides a "universal IR" for source
5782 languages, it also provides a starting point for developing a "universal"
5783 atomic operation and synchronization IR.
5784</p>
5785<p>
5786 These do <em>not</em> form an API such as high-level threading libraries,
5787 software transaction memory systems, atomic primitives, and intrinsic
5788 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
5789 application libraries. The hardware interface provided by LLVM should allow
5790 a clean implementation of all of these APIs and parallel programming models.
5791 No one model or paradigm should be selected above others unless the hardware
5792 itself ubiquitously does so.
5793
5794</p>
5795</div>
5796
5797<!-- _______________________________________________________________________ -->
5798<div class="doc_subsubsection">
5799 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
5800</div>
5801<div class="doc_text">
5802<h5>Syntax:</h5>
5803<pre>
5804declare void @llvm.memory.barrier( i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt;,
5805i1 &lt;device&gt; )
5806
5807</pre>
5808<h5>Overview:</h5>
5809<p>
5810 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
5811 specific pairs of memory access types.
5812</p>
5813<h5>Arguments:</h5>
5814<p>
5815 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
5816 The first four arguments enables a specific barrier as listed below. The fith
5817 argument specifies that the barrier applies to io or device or uncached memory.
5818
5819</p>
5820 <ul>
5821 <li><tt>ll</tt>: load-load barrier</li>
5822 <li><tt>ls</tt>: load-store barrier</li>
5823 <li><tt>sl</tt>: store-load barrier</li>
5824 <li><tt>ss</tt>: store-store barrier</li>
5825 <li><tt>device</tt>: barrier applies to device and uncached memory also.
5826 </ul>
5827<h5>Semantics:</h5>
5828<p>
5829 This intrinsic causes the system to enforce some ordering constraints upon
5830 the loads and stores of the program. This barrier does not indicate
5831 <em>when</em> any events will occur, it only enforces an <em>order</em> in
5832 which they occur. For any of the specified pairs of load and store operations
5833 (f.ex. load-load, or store-load), all of the first operations preceding the
5834 barrier will complete before any of the second operations succeeding the
5835 barrier begin. Specifically the semantics for each pairing is as follows:
5836</p>
5837 <ul>
5838 <li><tt>ll</tt>: All loads before the barrier must complete before any load
5839 after the barrier begins.</li>
5840
5841 <li><tt>ls</tt>: All loads before the barrier must complete before any
5842 store after the barrier begins.</li>
5843 <li><tt>ss</tt>: All stores before the barrier must complete before any
5844 store after the barrier begins.</li>
5845 <li><tt>sl</tt>: All stores before the barrier must complete before any
5846 load after the barrier begins.</li>
5847 </ul>
5848<p>
5849 These semantics are applied with a logical "and" behavior when more than one
5850 is enabled in a single memory barrier intrinsic.
5851</p>
5852<p>
5853 Backends may implement stronger barriers than those requested when they do not
5854 support as fine grained a barrier as requested. Some architectures do not
5855 need all types of barriers and on such architectures, these become noops.
5856</p>
5857<h5>Example:</h5>
5858<pre>
5859%ptr = malloc i32
5860 store i32 4, %ptr
5861
5862%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
5863 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
5864 <i>; guarantee the above finishes</i>
5865 store i32 8, %ptr <i>; before this begins</i>
5866</pre>
5867</div>
5868
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5869<!-- _______________________________________________________________________ -->
5870<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5871 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5872</div>
5873<div class="doc_text">
5874<h5>Syntax:</h5>
5875<p>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5876 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
5877 any integer bit width and for different address spaces. Not all targets
5878 support all bit widths however.</p>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5879
5880<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5881declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt; )
5882declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt; )
5883declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt; )
5884declare 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]5885
5886</pre>
5887<h5>Overview:</h5>
5888<p>
5889 This loads a value in memory and compares it to a given value. If they are
5890 equal, it stores a new value into the memory.
5891</p>
5892<h5>Arguments:</h5>
5893<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5894 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5895 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
5896 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
5897 this integer type. While any bit width integer may be used, targets may only
5898 lower representations they support in hardware.
5899
5900</p>
5901<h5>Semantics:</h5>
5902<p>
5903 This entire intrinsic must be executed atomically. It first loads the value
5904 in memory pointed to by <tt>ptr</tt> and compares it with the value
5905 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
5906 loaded value is yielded in all cases. This provides the equivalent of an
5907 atomic compare-and-swap operation within the SSA framework.
5908</p>
5909<h5>Examples:</h5>
5910
5911<pre>
5912%ptr = malloc i32
5913 store i32 4, %ptr
5914
5915%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5916%result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5917 <i>; yields {i32}:result1 = 4</i>
5918%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5919%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5920
5921%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5922%result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5923 <i>; yields {i32}:result2 = 8</i>
5924%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
5925
5926%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
5927</pre>
5928</div>
5929
5930<!-- _______________________________________________________________________ -->
5931<div class="doc_subsubsection">
5932 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
5933</div>
5934<div class="doc_text">
5935<h5>Syntax:</h5>
5936
5937<p>
5938 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
5939 integer bit width. Not all targets support all bit widths however.</p>
5940<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5941declare i8 @llvm.atomic.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;val&gt; )
5942declare i16 @llvm.atomic.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;val&gt; )
5943declare i32 @llvm.atomic.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;val&gt; )
5944declare i64 @llvm.atomic.swap.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;val&gt; )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5945
5946</pre>
5947<h5>Overview:</h5>
5948<p>
5949 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
5950 the value from memory. It then stores the value in <tt>val</tt> in the memory
5951 at <tt>ptr</tt>.
5952</p>
5953<h5>Arguments:</h5>
5954
5955<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5956 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5957 <tt>val</tt> argument and the result must be integers of the same bit width.
5958 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
5959 integer type. The targets may only lower integer representations they
5960 support.
5961</p>
5962<h5>Semantics:</h5>
5963<p>
5964 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
5965 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
5966 equivalent of an atomic swap operation within the SSA framework.
5967
5968</p>
5969<h5>Examples:</h5>
5970<pre>
5971%ptr = malloc i32
5972 store i32 4, %ptr
5973
5974%val1 = add i32 4, 4
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5975%result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5976 <i>; yields {i32}:result1 = 4</i>
5977%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5978%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5979
5980%val2 = add i32 1, 1
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]5981%result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5982 <i>; yields {i32}:result2 = 8</i>
5983
5984%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
5985%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
5986</pre>
5987</div>
5988
5989<!-- _______________________________________________________________________ -->
5990<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5991 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5992
5993</div>
5994<div class="doc_text">
5995<h5>Syntax:</h5>
5996<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]5997 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]5998 integer bit width. Not all targets support all bit widths however.</p>
5999<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6000declare i8 @llvm.atomic.load.add.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6001declare i16 @llvm.atomic.load.add.i16..p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6002declare i32 @llvm.atomic.load.add.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6003declare 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]6004
6005</pre>
6006<h5>Overview:</h5>
6007<p>
6008 This intrinsic adds <tt>delta</tt> to the value stored in memory at
6009 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6010</p>
6011<h5>Arguments:</h5>
6012<p>
6013
6014 The intrinsic takes two arguments, the first a pointer to an integer value
6015 and the second an integer value. The result is also an integer value. These
6016 integer types can have any bit width, but they must all have the same bit
6017 width. The targets may only lower integer representations they support.
6018</p>
6019<h5>Semantics:</h5>
6020<p>
6021 This intrinsic does a series of operations atomically. It first loads the
6022 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6023 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6024</p>
6025
6026<h5>Examples:</h5>
6027<pre>
6028%ptr = malloc i32
6029 store i32 4, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6030%result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6031 <i>; yields {i32}:result1 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6032%result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6033 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6034%result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6035 <i>; yields {i32}:result3 = 10</i>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6036%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]6037</pre>
6038</div>
6039
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6040<!-- _______________________________________________________________________ -->
6041<div class="doc_subsubsection">
6042 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6043
6044</div>
6045<div class="doc_text">
6046<h5>Syntax:</h5>
6047<p>
6048 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]6049 any integer bit width and for different address spaces. Not all targets
6050 support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6051<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6052declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6053declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6054declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6055declare 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]6056
6057</pre>
6058<h5>Overview:</h5>
6059<p>
6060 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6061 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6062</p>
6063<h5>Arguments:</h5>
6064<p>
6065
6066 The intrinsic takes two arguments, the first a pointer to an integer value
6067 and the second an integer value. The result is also an integer value. These
6068 integer types can have any bit width, but they must all have the same bit
6069 width. The targets may only lower integer representations they support.
6070</p>
6071<h5>Semantics:</h5>
6072<p>
6073 This intrinsic does a series of operations atomically. It first loads the
6074 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6075 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6076</p>
6077
6078<h5>Examples:</h5>
6079<pre>
6080%ptr = malloc i32
6081 store i32 8, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6082%result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6083 <i>; yields {i32}:result1 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6084%result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6085 <i>; yields {i32}:result2 = 4</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6086%result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6087 <i>; yields {i32}:result3 = 2</i>
6088%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6089</pre>
6090</div>
6091
6092<!-- _______________________________________________________________________ -->
6093<div class="doc_subsubsection">
6094 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6095 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6096 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6097 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6098
6099</div>
6100<div class="doc_text">
6101<h5>Syntax:</h5>
6102<p>
6103 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6104 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6105 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6106 address spaces. Not all targets support all bit widths however.</p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6107<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6108declare i8 @llvm.atomic.load.and.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6109declare i16 @llvm.atomic.load.and.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6110declare i32 @llvm.atomic.load.and.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6111declare 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]6112
6113</pre>
6114
6115<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6116declare i8 @llvm.atomic.load.or.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6117declare i16 @llvm.atomic.load.or.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6118declare i32 @llvm.atomic.load.or.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6119declare 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]6120
6121</pre>
6122
6123<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6124declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6125declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6126declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6127declare 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]6128
6129</pre>
6130
6131<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6132declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6133declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6134declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6135declare 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]6136
6137</pre>
6138<h5>Overview:</h5>
6139<p>
6140 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6141 the value stored in memory at <tt>ptr</tt>. It yields the original value
6142 at <tt>ptr</tt>.
6143</p>
6144<h5>Arguments:</h5>
6145<p>
6146
6147 These intrinsics take two arguments, the first a pointer to an integer value
6148 and the second an integer value. The result is also an integer value. These
6149 integer types can have any bit width, but they must all have the same bit
6150 width. The targets may only lower integer representations they support.
6151</p>
6152<h5>Semantics:</h5>
6153<p>
6154 These intrinsics does a series of operations atomically. They first load the
6155 value stored at <tt>ptr</tt>. They then do the bitwise operation
6156 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6157 value stored at <tt>ptr</tt>.
6158</p>
6159
6160<h5>Examples:</h5>
6161<pre>
6162%ptr = malloc i32
6163 store i32 0x0F0F, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6164%result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6165 <i>; yields {i32}:result0 = 0x0F0F</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6166%result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6167 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6168%result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6169 <i>; yields {i32}:result2 = 0xF0</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6170%result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6171 <i>; yields {i32}:result3 = FF</i>
6172%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6173</pre>
6174</div>
6175
6176
6177<!-- _______________________________________________________________________ -->
6178<div class="doc_subsubsection">
6179 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6180 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6181 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6182 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6183
6184</div>
6185<div class="doc_text">
6186<h5>Syntax:</h5>
6187<p>
6188 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6189 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6190 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6191 address spaces. Not all targets
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6192 support all bit widths however.</p>
6193<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6194declare i8 @llvm.atomic.load.max.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6195declare i16 @llvm.atomic.load.max.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6196declare i32 @llvm.atomic.load.max.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6197declare 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]6198
6199</pre>
6200
6201<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6202declare i8 @llvm.atomic.load.min.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6203declare i16 @llvm.atomic.load.min.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6204declare i32 @llvm.atomic.load.min.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6205declare 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]6206
6207</pre>
6208
6209<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6210declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6211declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6212declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6213declare 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]6214
6215</pre>
6216
6217<pre>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6218declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6219declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6220declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6221declare 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]6222
6223</pre>
6224<h5>Overview:</h5>
6225<p>
6226 These intrinsics takes the signed or unsigned minimum or maximum of
6227 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6228 original value at <tt>ptr</tt>.
6229</p>
6230<h5>Arguments:</h5>
6231<p>
6232
6233 These intrinsics take two arguments, the first a pointer to an integer value
6234 and the second an integer value. The result is also an integer value. These
6235 integer types can have any bit width, but they must all have the same bit
6236 width. The targets may only lower integer representations they support.
6237</p>
6238<h5>Semantics:</h5>
6239<p>
6240 These intrinsics does a series of operations atomically. They first load the
6241 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6242 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6243 the original value stored at <tt>ptr</tt>.
6244</p>
6245
6246<h5>Examples:</h5>
6247<pre>
6248%ptr = malloc i32
6249 store i32 7, %ptr
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6250%result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6251 <i>; yields {i32}:result0 = 7</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6252%result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6253 <i>; yields {i32}:result1 = -2</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6254%result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6255 <i>; yields {i32}:result2 = 8</i>
Mon P Wangce3ac892008-07-30 04:36:53 +0000[diff] [blame]6256%result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame]6257 <i>; yields {i32}:result3 = 8</i>
6258%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6259</pre>
6260</div>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6261
6262<!-- ======================================================================= -->
6263<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6264 <a name="int_general">General Intrinsics</a>
6265</div>
6266
6267<div class="doc_text">
6268<p> This class of intrinsics is designed to be generic and has
6269no specific purpose. </p>
6270</div>
6271
6272<!-- _______________________________________________________________________ -->
6273<div class="doc_subsubsection">
6274 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6275</div>
6276
6277<div class="doc_text">
6278
6279<h5>Syntax:</h5>
6280<pre>
6281 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6282</pre>
6283
6284<h5>Overview:</h5>
6285
6286<p>
6287The '<tt>llvm.var.annotation</tt>' intrinsic
6288</p>
6289
6290<h5>Arguments:</h5>
6291
6292<p>
6293The first argument is a pointer to a value, the second is a pointer to a
6294global string, the third is a pointer to a global string which is the source
6295file name, and the last argument is the line number.
6296</p>
6297
6298<h5>Semantics:</h5>
6299
6300<p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6301This intrinsic allows annotation of local variables with arbitrary strings.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6302This can be useful for special purpose optimizations that want to look for these
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6303annotations. These have no other defined use, they are ignored by code
6304generation and optimization.
6305</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6306</div>
6307
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6308<!-- _______________________________________________________________________ -->
6309<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]6310 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6311</div>
6312
6313<div class="doc_text">
6314
6315<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6316<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6317any integer bit width.
6318</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6319<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]6320 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6321 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6322 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6323 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6324 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]6325</pre>
6326
6327<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6328
6329<p>
6330The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6331</p>
6332
6333<h5>Arguments:</h5>
6334
6335<p>
6336The first argument is an integer value (result of some expression),
6337the second is a pointer to a global string, the third is a pointer to a global
6338string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6339It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6340</p>
6341
6342<h5>Semantics:</h5>
6343
6344<p>
6345This intrinsic allows annotations to be put on arbitrary expressions
6346with arbitrary strings. This can be useful for special purpose optimizations
6347that want to look for these annotations. These have no other defined use, they
6348are ignored by code generation and optimization.
6349</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6350
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6351<!-- _______________________________________________________________________ -->
6352<div class="doc_subsubsection">
6353 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6354</div>
6355
6356<div class="doc_text">
6357
6358<h5>Syntax:</h5>
6359<pre>
6360 declare void @llvm.trap()
6361</pre>
6362
6363<h5>Overview:</h5>
6364
6365<p>
6366The '<tt>llvm.trap</tt>' intrinsic
6367</p>
6368
6369<h5>Arguments:</h5>
6370
6371<p>
6372None
6373</p>
6374
6375<h5>Semantics:</h5>
6376
6377<p>
6378This intrinsics is lowered to the target dependent trap instruction. If the
6379target does not have a trap instruction, this intrinsic will be lowered to the
6380call of the abort() function.
6381</p>
6382</div>
6383
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6384<!-- *********************************************************************** -->
6385<hr>
6386<address>
6387 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
6388 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
6389 <a href="http://validator.w3.org/check/referer"><img
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]6390 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6391
6392 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
6393 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
6394 Last modified: $Date$
6395</address>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]6396
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6397</body>
6398</html>