Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 6bde9ec4129ed4b265c5b172cfe9b34d1e0247d2 / . / docs / LangRef.html
blob: 4de69b0e721d36c35b4a6e226af5e08f96ce20c6 [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>
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]29 <li><a href="#gc">Garbage Collector Names</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]30 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
31 <li><a href="#datalayout">Data Layout</a></li>
32 </ol>
33 </li>
34 <li><a href="#typesystem">Type System</a>
35 <ol>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]36 <li><a href="#t_classifications">Type Classifications</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]37 <li><a href="#t_primitive">Primitive Types</a>
38 <ol>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]39 <li><a href="#t_floating">Floating Point Types</a></li>
40 <li><a href="#t_void">Void Type</a></li>
41 <li><a href="#t_label">Label Type</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]42 </ol>
43 </li>
44 <li><a href="#t_derived">Derived Types</a>
45 <ol>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]46 <li><a href="#t_integer">Integer Type</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]47 <li><a href="#t_array">Array Type</a></li>
48 <li><a href="#t_function">Function Type</a></li>
49 <li><a href="#t_pointer">Pointer Type</a></li>
50 <li><a href="#t_struct">Structure Type</a></li>
51 <li><a href="#t_pstruct">Packed Structure Type</a></li>
52 <li><a href="#t_vector">Vector Type</a></li>
53 <li><a href="#t_opaque">Opaque Type</a></li>
54 </ol>
55 </li>
56 </ol>
57 </li>
58 <li><a href="#constants">Constants</a>
59 <ol>
60 <li><a href="#simpleconstants">Simple Constants</a>
61 <li><a href="#aggregateconstants">Aggregate Constants</a>
62 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
63 <li><a href="#undefvalues">Undefined Values</a>
64 <li><a href="#constantexprs">Constant Expressions</a>
65 </ol>
66 </li>
67 <li><a href="#othervalues">Other Values</a>
68 <ol>
69 <li><a href="#inlineasm">Inline Assembler Expressions</a>
70 </ol>
71 </li>
72 <li><a href="#instref">Instruction Reference</a>
73 <ol>
74 <li><a href="#terminators">Terminator Instructions</a>
75 <ol>
76 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
77 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
78 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
79 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
80 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
81 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
82 </ol>
83 </li>
84 <li><a href="#binaryops">Binary Operations</a>
85 <ol>
86 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
87 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
88 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
89 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
90 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
91 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
92 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
93 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
94 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
95 </ol>
96 </li>
97 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
98 <ol>
99 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
100 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
101 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
102 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
103 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
104 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
105 </ol>
106 </li>
107 <li><a href="#vectorops">Vector Operations</a>
108 <ol>
109 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
110 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
111 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
112 </ol>
113 </li>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]114 <li><a href="#aggregateops">Aggregate Operations</a>
115 <ol>
116 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
117 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
118 </ol>
119 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]120 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
121 <ol>
122 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
123 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
124 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
125 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
126 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
127 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
128 </ol>
129 </li>
130 <li><a href="#convertops">Conversion Operations</a>
131 <ol>
132 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
133 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
134 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
135 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
136 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
137 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
138 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
139 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
140 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
141 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
142 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
143 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
144 </ol>
145 <li><a href="#otherops">Other Operations</a>
146 <ol>
147 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
148 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]149 <li><a href="#i_vicmp">'<tt>vicmp</tt>' Instruction</a></li>
150 <li><a href="#i_vfcmp">'<tt>vfcmp</tt>' Instruction</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]151 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
152 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
153 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
154 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]155 <li><a href="#i_getresult">'<tt>getresult</tt>' Instruction</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]156 </ol>
157 </li>
158 </ol>
159 </li>
160 <li><a href="#intrinsics">Intrinsic Functions</a>
161 <ol>
162 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
163 <ol>
164 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
165 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
166 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
167 </ol>
168 </li>
169 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
170 <ol>
171 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
172 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
173 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
174 </ol>
175 </li>
176 <li><a href="#int_codegen">Code Generator Intrinsics</a>
177 <ol>
178 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
179 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
180 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
181 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
182 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
183 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
184 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
185 </ol>
186 </li>
187 <li><a href="#int_libc">Standard C Library Intrinsics</a>
188 <ol>
189 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
190 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
191 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
192 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
193 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]194 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
195 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
196 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]197 </ol>
198 </li>
199 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
200 <ol>
201 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
202 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
203 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
204 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
205 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
206 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
207 </ol>
208 </li>
209 <li><a href="#int_debugger">Debugger intrinsics</a></li>
210 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]211 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]212 <ol>
213 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]214 </ol>
215 </li>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]216 <li><a href="#int_atomics">Atomic intrinsics</a>
217 <ol>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]218 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]219 <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]220 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]221 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
222 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
223 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
224 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
225 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
226 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
227 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
228 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
229 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
230 <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]231 </ol>
232 </li>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]233 <li><a href="#int_general">General intrinsics</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]234 <ol>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]235 <li><a href="#int_var_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]236 <tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]237 <li><a href="#int_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]238 <tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]239 <li><a href="#int_trap">
240 <tt>llvm.trap</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]241 </ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]242 </li>
243 </ol>
244 </li>
245</ol>
246
247<div class="doc_author">
248 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
249 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
250</div>
251
252<!-- *********************************************************************** -->
253<div class="doc_section"> <a name="abstract">Abstract </a></div>
254<!-- *********************************************************************** -->
255
256<div class="doc_text">
257<p>This document is a reference manual for the LLVM assembly language.
258LLVM is an SSA based representation that provides type safety,
259low-level operations, flexibility, and the capability of representing
260'all' high-level languages cleanly. It is the common code
261representation used throughout all phases of the LLVM compilation
262strategy.</p>
263</div>
264
265<!-- *********************************************************************** -->
266<div class="doc_section"> <a name="introduction">Introduction</a> </div>
267<!-- *********************************************************************** -->
268
269<div class="doc_text">
270
271<p>The LLVM code representation is designed to be used in three
272different forms: as an in-memory compiler IR, as an on-disk bitcode
273representation (suitable for fast loading by a Just-In-Time compiler),
274and as a human readable assembly language representation. This allows
275LLVM to provide a powerful intermediate representation for efficient
276compiler transformations and analysis, while providing a natural means
277to debug and visualize the transformations. The three different forms
278of LLVM are all equivalent. This document describes the human readable
279representation and notation.</p>
280
281<p>The LLVM representation aims to be light-weight and low-level
282while being expressive, typed, and extensible at the same time. It
283aims to be a "universal IR" of sorts, by being at a low enough level
284that high-level ideas may be cleanly mapped to it (similar to how
285microprocessors are "universal IR's", allowing many source languages to
286be mapped to them). By providing type information, LLVM can be used as
287the target of optimizations: for example, through pointer analysis, it
288can be proven that a C automatic variable is never accessed outside of
289the current function... allowing it to be promoted to a simple SSA
290value instead of a memory location.</p>
291
292</div>
293
294<!-- _______________________________________________________________________ -->
295<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
296
297<div class="doc_text">
298
299<p>It is important to note that this document describes 'well formed'
300LLVM assembly language. There is a difference between what the parser
301accepts and what is considered 'well formed'. For example, the
302following instruction is syntactically okay, but not well formed:</p>
303
304<div class="doc_code">
305<pre>
306%x = <a href="#i_add">add</a> i32 1, %x
307</pre>
308</div>
309
310<p>...because the definition of <tt>%x</tt> does not dominate all of
311its uses. The LLVM infrastructure provides a verification pass that may
312be used to verify that an LLVM module is well formed. This pass is
313automatically run by the parser after parsing input assembly and by
314the optimizer before it outputs bitcode. The violations pointed out
315by the verifier pass indicate bugs in transformation passes or input to
316the parser.</p>
317</div>
318
Chris Lattnera83fdc02007-10-03 17:34:29 +0000[diff] [blame]319<!-- Describe the typesetting conventions here. -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]320
321<!-- *********************************************************************** -->
322<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
323<!-- *********************************************************************** -->
324
325<div class="doc_text">
326
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]327 <p>LLVM identifiers come in two basic types: global and local. Global
328 identifiers (functions, global variables) begin with the @ character. Local
329 identifiers (register names, types) begin with the % character. Additionally,
330 there are three different formats for identifiers, for different purposes:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]331
332<ol>
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]333 <li>Named values are represented as a string of characters with their prefix.
334 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
335 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]336 Identifiers which require other characters in their names can be surrounded
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]337 with quotes. In this way, anything except a <tt>&quot;</tt> character can
338 be used in a named value.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]339
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]340 <li>Unnamed values are represented as an unsigned numeric value with their
341 prefix. For example, %12, @2, %44.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]342
343 <li>Constants, which are described in a <a href="#constants">section about
344 constants</a>, below.</li>
345</ol>
346
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]347<p>LLVM requires that values start with a prefix for two reasons: Compilers
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]348don't need to worry about name clashes with reserved words, and the set of
349reserved words may be expanded in the future without penalty. Additionally,
350unnamed identifiers allow a compiler to quickly come up with a temporary
351variable without having to avoid symbol table conflicts.</p>
352
353<p>Reserved words in LLVM are very similar to reserved words in other
354languages. There are keywords for different opcodes
355('<tt><a href="#i_add">add</a></tt>',
356 '<tt><a href="#i_bitcast">bitcast</a></tt>',
357 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
358href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
359and others. These reserved words cannot conflict with variable names, because
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]360none of them start with a prefix character ('%' or '@').</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]361
362<p>Here is an example of LLVM code to multiply the integer variable
363'<tt>%X</tt>' by 8:</p>
364
365<p>The easy way:</p>
366
367<div class="doc_code">
368<pre>
369%result = <a href="#i_mul">mul</a> i32 %X, 8
370</pre>
371</div>
372
373<p>After strength reduction:</p>
374
375<div class="doc_code">
376<pre>
377%result = <a href="#i_shl">shl</a> i32 %X, i8 3
378</pre>
379</div>
380
381<p>And the hard way:</p>
382
383<div class="doc_code">
384<pre>
385<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
386<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
387%result = <a href="#i_add">add</a> i32 %1, %1
388</pre>
389</div>
390
391<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
392important lexical features of LLVM:</p>
393
394<ol>
395
396 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
397 line.</li>
398
399 <li>Unnamed temporaries are created when the result of a computation is not
400 assigned to a named value.</li>
401
402 <li>Unnamed temporaries are numbered sequentially</li>
403
404</ol>
405
406<p>...and it also shows a convention that we follow in this document. When
407demonstrating instructions, we will follow an instruction with a comment that
408defines the type and name of value produced. Comments are shown in italic
409text.</p>
410
411</div>
412
413<!-- *********************************************************************** -->
414<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
415<!-- *********************************************************************** -->
416
417<!-- ======================================================================= -->
418<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
419</div>
420
421<div class="doc_text">
422
423<p>LLVM programs are composed of "Module"s, each of which is a
424translation unit of the input programs. Each module consists of
425functions, global variables, and symbol table entries. Modules may be
426combined together with the LLVM linker, which merges function (and
427global variable) definitions, resolves forward declarations, and merges
428symbol table entries. Here is an example of the "hello world" module:</p>
429
430<div class="doc_code">
431<pre><i>; Declare the string constant as a global constant...</i>
432<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
433 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
434
435<i>; External declaration of the puts function</i>
436<a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
437
438<i>; Definition of main function</i>
439define i32 @main() { <i>; i32()* </i>
440 <i>; Convert [13x i8 ]* to i8 *...</i>
441 %cast210 = <a
442 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
443
444 <i>; Call puts function to write out the string to stdout...</i>
445 <a
446 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
447 <a
448 href="#i_ret">ret</a> i32 0<br>}<br>
449</pre>
450</div>
451
452<p>This example is made up of a <a href="#globalvars">global variable</a>
453named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
454function, and a <a href="#functionstructure">function definition</a>
455for "<tt>main</tt>".</p>
456
457<p>In general, a module is made up of a list of global values,
458where both functions and global variables are global values. Global values are
459represented by a pointer to a memory location (in this case, a pointer to an
460array of char, and a pointer to a function), and have one of the following <a
461href="#linkage">linkage types</a>.</p>
462
463</div>
464
465<!-- ======================================================================= -->
466<div class="doc_subsection">
467 <a name="linkage">Linkage Types</a>
468</div>
469
470<div class="doc_text">
471
472<p>
473All Global Variables and Functions have one of the following types of linkage:
474</p>
475
476<dl>
477
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]478 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]479
480 <dd>Global values with internal linkage are only directly accessible by
481 objects in the current module. In particular, linking code into a module with
482 an internal global value may cause the internal to be renamed as necessary to
483 avoid collisions. Because the symbol is internal to the module, all
484 references can be updated. This corresponds to the notion of the
485 '<tt>static</tt>' keyword in C.
486 </dd>
487
488 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
489
490 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
491 the same name when linkage occurs. This is typically used to implement
492 inline functions, templates, or other code which must be generated in each
493 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
494 allowed to be discarded.
495 </dd>
496
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]497 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
498
499 <dd>"<tt>common</tt>" linkage is exactly the same as <tt>linkonce</tt>
500 linkage, except that unreferenced <tt>common</tt> globals may not be
501 discarded. This is used for globals that may be emitted in multiple
502 translation units, but that are not guaranteed to be emitted into every
503 translation unit that uses them. One example of this is tentative
504 definitions in C, such as "<tt>int X;</tt>" at global scope.
505 </dd>
506
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]507 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
508
Dale Johannesen96e7e092008-05-23 23:13:41 +0000[diff] [blame]509 <dd>"<tt>weak</tt>" linkage is the same as <tt>common</tt> linkage, except
510 that some targets may choose to emit different assembly sequences for them
511 for target-dependent reasons. This is used for globals that are declared
512 "weak" in C source code.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]513 </dd>
514
515 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
516
517 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
518 pointer to array type. When two global variables with appending linkage are
519 linked together, the two global arrays are appended together. This is the
520 LLVM, typesafe, equivalent of having the system linker append together
521 "sections" with identical names when .o files are linked.
522 </dd>
523
524 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
525 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
526 until linked, if not linked, the symbol becomes null instead of being an
527 undefined reference.
528 </dd>
529
530 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
531
532 <dd>If none of the above identifiers are used, the global is externally
533 visible, meaning that it participates in linkage and can be used to resolve
534 external symbol references.
535 </dd>
536</dl>
537
538 <p>
539 The next two types of linkage are targeted for Microsoft Windows platform
540 only. They are designed to support importing (exporting) symbols from (to)
541 DLLs.
542 </p>
543
544 <dl>
545 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
546
547 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
548 or variable via a global pointer to a pointer that is set up by the DLL
549 exporting the symbol. On Microsoft Windows targets, the pointer name is
550 formed by combining <code>_imp__</code> and the function or variable name.
551 </dd>
552
553 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
554
555 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
556 pointer to a pointer in a DLL, so that it can be referenced with the
557 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
558 name is formed by combining <code>_imp__</code> and the function or variable
559 name.
560 </dd>
561
562</dl>
563
564<p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
565variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
566variable and was linked with this one, one of the two would be renamed,
567preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
568external (i.e., lacking any linkage declarations), they are accessible
569outside of the current module.</p>
570<p>It is illegal for a function <i>declaration</i>
571to have any linkage type other than "externally visible", <tt>dllimport</tt>,
572or <tt>extern_weak</tt>.</p>
573<p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
574linkages.
575</div>
576
577<!-- ======================================================================= -->
578<div class="doc_subsection">
579 <a name="callingconv">Calling Conventions</a>
580</div>
581
582<div class="doc_text">
583
584<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
585and <a href="#i_invoke">invokes</a> can all have an optional calling convention
586specified for the call. The calling convention of any pair of dynamic
587caller/callee must match, or the behavior of the program is undefined. The
588following calling conventions are supported by LLVM, and more may be added in
589the future:</p>
590
591<dl>
592 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
593
594 <dd>This calling convention (the default if no other calling convention is
595 specified) matches the target C calling conventions. This calling convention
596 supports varargs function calls and tolerates some mismatch in the declared
597 prototype and implemented declaration of the function (as does normal C).
598 </dd>
599
600 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
601
602 <dd>This calling convention attempts to make calls as fast as possible
603 (e.g. by passing things in registers). This calling convention allows the
604 target to use whatever tricks it wants to produce fast code for the target,
605 without having to conform to an externally specified ABI. Implementations of
Arnold Schwaighofer07444922008-05-14 09:17:12 +0000[diff] [blame]606 this convention should allow arbitrary
607 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> to be
608 supported. This calling convention does not support varargs and requires the
609 prototype of all callees to exactly match the prototype of the function
610 definition.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]611 </dd>
612
613 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
614
615 <dd>This calling convention attempts to make code in the caller as efficient
616 as possible under the assumption that the call is not commonly executed. As
617 such, these calls often preserve all registers so that the call does not break
618 any live ranges in the caller side. This calling convention does not support
619 varargs and requires the prototype of all callees to exactly match the
620 prototype of the function definition.
621 </dd>
622
623 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
624
625 <dd>Any calling convention may be specified by number, allowing
626 target-specific calling conventions to be used. Target specific calling
627 conventions start at 64.
628 </dd>
629</dl>
630
631<p>More calling conventions can be added/defined on an as-needed basis, to
632support pascal conventions or any other well-known target-independent
633convention.</p>
634
635</div>
636
637<!-- ======================================================================= -->
638<div class="doc_subsection">
639 <a name="visibility">Visibility Styles</a>
640</div>
641
642<div class="doc_text">
643
644<p>
645All Global Variables and Functions have one of the following visibility styles:
646</p>
647
648<dl>
649 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
650
651 <dd>On ELF, default visibility means that the declaration is visible to other
652 modules and, in shared libraries, means that the declared entity may be
653 overridden. On Darwin, default visibility means that the declaration is
654 visible to other modules. Default visibility corresponds to "external
655 linkage" in the language.
656 </dd>
657
658 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
659
660 <dd>Two declarations of an object with hidden visibility refer to the same
661 object if they are in the same shared object. Usually, hidden visibility
662 indicates that the symbol will not be placed into the dynamic symbol table,
663 so no other module (executable or shared library) can reference it
664 directly.
665 </dd>
666
667 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
668
669 <dd>On ELF, protected visibility indicates that the symbol will be placed in
670 the dynamic symbol table, but that references within the defining module will
671 bind to the local symbol. That is, the symbol cannot be overridden by another
672 module.
673 </dd>
674</dl>
675
676</div>
677
678<!-- ======================================================================= -->
679<div class="doc_subsection">
680 <a name="globalvars">Global Variables</a>
681</div>
682
683<div class="doc_text">
684
685<p>Global variables define regions of memory allocated at compilation time
686instead of run-time. Global variables may optionally be initialized, may have
687an explicit section to be placed in, and may have an optional explicit alignment
688specified. A variable may be defined as "thread_local", which means that it
689will not be shared by threads (each thread will have a separated copy of the
690variable). A variable may be defined as a global "constant," which indicates
691that the contents of the variable will <b>never</b> be modified (enabling better
692optimization, allowing the global data to be placed in the read-only section of
693an executable, etc). Note that variables that need runtime initialization
694cannot be marked "constant" as there is a store to the variable.</p>
695
696<p>
697LLVM explicitly allows <em>declarations</em> of global variables to be marked
698constant, even if the final definition of the global is not. This capability
699can be used to enable slightly better optimization of the program, but requires
700the language definition to guarantee that optimizations based on the
701'constantness' are valid for the translation units that do not include the
702definition.
703</p>
704
705<p>As SSA values, global variables define pointer values that are in
706scope (i.e. they dominate) all basic blocks in the program. Global
707variables always define a pointer to their "content" type because they
708describe a region of memory, and all memory objects in LLVM are
709accessed through pointers.</p>
710
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]711<p>A global variable may be declared to reside in a target-specifc numbered
712address space. For targets that support them, address spaces may affect how
713optimizations are performed and/or what target instructions are used to access
Christopher Lamb20a39e92007-12-12 08:44:39 +0000[diff] [blame]714the variable. The default address space is zero. The address space qualifier
715must precede any other attributes.</p>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]716
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]717<p>LLVM allows an explicit section to be specified for globals. If the target
718supports it, it will emit globals to the section specified.</p>
719
720<p>An explicit alignment may be specified for a global. If not present, or if
721the alignment is set to zero, the alignment of the global is set by the target
722to whatever it feels convenient. If an explicit alignment is specified, the
723global is forced to have at least that much alignment. All alignments must be
724a power of 2.</p>
725
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]726<p>For example, the following defines a global in a numbered address space with
727an initializer, section, and alignment:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]728
729<div class="doc_code">
730<pre>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]731@G = constant float 1.0 addrspace(5), section "foo", align 4
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]732</pre>
733</div>
734
735</div>
736
737
738<!-- ======================================================================= -->
739<div class="doc_subsection">
740 <a name="functionstructure">Functions</a>
741</div>
742
743<div class="doc_text">
744
745<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
746an optional <a href="#linkage">linkage type</a>, an optional
747<a href="#visibility">visibility style</a>, an optional
748<a href="#callingconv">calling convention</a>, a return type, an optional
749<a href="#paramattrs">parameter attribute</a> for the return type, a function
750name, a (possibly empty) argument list (each with optional
751<a href="#paramattrs">parameter attributes</a>), an optional section, an
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]752optional alignment, an optional <a href="#gc">garbage collector name</a>, an
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]753opening curly brace, a list of basic blocks, and a closing curly brace.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]754
755LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
756optional <a href="#linkage">linkage type</a>, an optional
757<a href="#visibility">visibility style</a>, an optional
758<a href="#callingconv">calling convention</a>, a return type, an optional
759<a href="#paramattrs">parameter attribute</a> for the return type, a function
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]760name, a possibly empty list of arguments, an optional alignment, and an optional
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]761<a href="#gc">garbage collector name</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]762
763<p>A function definition contains a list of basic blocks, forming the CFG for
764the function. Each basic block may optionally start with a label (giving the
765basic block a symbol table entry), contains a list of instructions, and ends
766with a <a href="#terminators">terminator</a> instruction (such as a branch or
767function return).</p>
768
769<p>The first basic block in a function is special in two ways: it is immediately
770executed on entrance to the function, and it is not allowed to have predecessor
771basic blocks (i.e. there can not be any branches to the entry block of a
772function). Because the block can have no predecessors, it also cannot have any
773<a href="#i_phi">PHI nodes</a>.</p>
774
775<p>LLVM allows an explicit section to be specified for functions. If the target
776supports it, it will emit functions to the section specified.</p>
777
778<p>An explicit alignment may be specified for a function. If not present, or if
779the alignment is set to zero, the alignment of the function is set by the target
780to whatever it feels convenient. If an explicit alignment is specified, the
781function is forced to have at least that much alignment. All alignments must be
782a power of 2.</p>
783
784</div>
785
786
787<!-- ======================================================================= -->
788<div class="doc_subsection">
789 <a name="aliasstructure">Aliases</a>
790</div>
791<div class="doc_text">
792 <p>Aliases act as "second name" for the aliasee value (which can be either
Anton Korobeynikov96822812008-03-22 08:36:14 +0000[diff] [blame]793 function, global variable, another alias or bitcast of global value). Aliases
794 may have an optional <a href="#linkage">linkage type</a>, and an
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]795 optional <a href="#visibility">visibility style</a>.</p>
796
797 <h5>Syntax:</h5>
798
799<div class="doc_code">
800<pre>
801@&lt;Name&gt; = [Linkage] [Visibility] alias &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
802</pre>
803</div>
804
805</div>
806
807
808
809<!-- ======================================================================= -->
810<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
811<div class="doc_text">
812 <p>The return type and each parameter of a function type may have a set of
813 <i>parameter attributes</i> associated with them. Parameter attributes are
814 used to communicate additional information about the result or parameters of
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]815 a function. Parameter attributes are considered to be part of the function,
816 not of the function type, so functions with different parameter attributes
817 can have the same function type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]818
819 <p>Parameter attributes are simple keywords that follow the type specified. If
820 multiple parameter attributes are needed, they are space separated. For
821 example:</p>
822
823<div class="doc_code">
824<pre>
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]825declare i32 @printf(i8* noalias , ...) nounwind
826declare i32 @atoi(i8*) nounwind readonly
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]827</pre>
828</div>
829
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]830 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
831 <tt>readonly</tt>) come immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]832
833 <p>Currently, only the following parameter attributes are defined:</p>
834 <dl>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]835 <dt><tt>zeroext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]836 <dd>This indicates that the parameter should be zero extended just before
837 a call to this function.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]838
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]839 <dt><tt>signext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]840 <dd>This indicates that the parameter should be sign extended just before
841 a call to this function.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]842
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]843 <dt><tt>inreg</tt></dt>
844 <dd>This indicates that the parameter should be placed in register (if
845 possible) during assembling function call. Support for this attribute is
846 target-specific</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]847
848 <dt><tt>byval</tt></dt>
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]849 <dd>This indicates that the pointer parameter should really be passed by
850 value to the function. The attribute implies that a hidden copy of the
851 pointee is made between the caller and the callee, so the callee is unable
852 to modify the value in the callee. This attribute is only valid on llvm
853 pointer arguments. It is generally used to pass structs and arrays by
854 value, but is also valid on scalars (even though this is silly).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]855
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]856 <dt><tt>sret</tt></dt>
Duncan Sands616cc032008-02-18 04:19:38 +0000[diff] [blame]857 <dd>This indicates that the pointer parameter specifies the address of a
858 structure that is the return value of the function in the source program.
Duncan Sandsef0e9e42008-03-17 12:17:41 +0000[diff] [blame]859 Loads and stores to the structure are assumed not to trap.
Duncan Sands616cc032008-02-18 04:19:38 +0000[diff] [blame]860 May only be applied to the first parameter.</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>noalias</tt></dt>
Owen Andersonc4fc4cd2008-02-18 04:09:01 +0000[diff] [blame]863 <dd>This indicates that the parameter does not alias any global or any other
864 parameter. The caller is responsible for ensuring that this is the case,
865 usually by placing the value in a stack allocation.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]866
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]867 <dt><tt>noreturn</tt></dt>
868 <dd>This function attribute indicates that the function never returns. This
869 indicates to LLVM that every call to this function should be treated as if
870 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]871
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]872 <dt><tt>nounwind</tt></dt>
Duncan Sandsef0e9e42008-03-17 12:17:41 +0000[diff] [blame]873 <dd>This function attribute indicates that no exceptions unwind out of the
874 function. Usually this is because the function makes no use of exceptions,
875 but it may also be that the function catches any exceptions thrown when
876 executing it.</dd>
877
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]878 <dt><tt>nest</tt></dt>
879 <dd>This indicates that the parameter can be excised using the
880 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
Duncan Sands13e13f82007-11-22 20:23:04 +0000[diff] [blame]881 <dt><tt>readonly</tt></dt>
Duncan Sandsd69c0e62007-11-14 21:14:02 +0000[diff] [blame]882 <dd>This function attribute indicates that the function has no side-effects
Duncan Sands13e13f82007-11-22 20:23:04 +0000[diff] [blame]883 except for producing a return value or throwing an exception. The value
884 returned must only depend on the function arguments and/or global variables.
885 It may use values obtained by dereferencing pointers.</dd>
886 <dt><tt>readnone</tt></dt>
887 <dd>A <tt>readnone</tt> function has the same restrictions as a <tt>readonly</tt>
Duncan Sandsd69c0e62007-11-14 21:14:02 +0000[diff] [blame]888 function, but in addition it is not allowed to dereference any pointer arguments
889 or global variables.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]890 </dl>
891
892</div>
893
894<!-- ======================================================================= -->
895<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]896 <a name="gc">Garbage Collector Names</a>
897</div>
898
899<div class="doc_text">
900<p>Each function may specify a garbage collector name, which is simply a
901string.</p>
902
903<div class="doc_code"><pre
904>define void @f() gc "name" { ...</pre></div>
905
906<p>The compiler declares the supported values of <i>name</i>. Specifying a
907collector which will cause the compiler to alter its output in order to support
908the named garbage collection algorithm.</p>
909</div>
910
911<!-- ======================================================================= -->
912<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]913 <a name="moduleasm">Module-Level Inline Assembly</a>
914</div>
915
916<div class="doc_text">
917<p>
918Modules may contain "module-level inline asm" blocks, which corresponds to the
919GCC "file scope inline asm" blocks. These blocks are internally concatenated by
920LLVM and treated as a single unit, but may be separated in the .ll file if
921desired. The syntax is very simple:
922</p>
923
924<div class="doc_code">
925<pre>
926module asm "inline asm code goes here"
927module asm "more can go here"
928</pre>
929</div>
930
931<p>The strings can contain any character by escaping non-printable characters.
932 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
933 for the number.
934</p>
935
936<p>
937 The inline asm code is simply printed to the machine code .s file when
938 assembly code is generated.
939</p>
940</div>
941
942<!-- ======================================================================= -->
943<div class="doc_subsection">
944 <a name="datalayout">Data Layout</a>
945</div>
946
947<div class="doc_text">
948<p>A module may specify a target specific data layout string that specifies how
949data is to be laid out in memory. The syntax for the data layout is simply:</p>
950<pre> target datalayout = "<i>layout specification</i>"</pre>
951<p>The <i>layout specification</i> consists of a list of specifications
952separated by the minus sign character ('-'). Each specification starts with a
953letter and may include other information after the letter to define some
954aspect of the data layout. The specifications accepted are as follows: </p>
955<dl>
956 <dt><tt>E</tt></dt>
957 <dd>Specifies that the target lays out data in big-endian form. That is, the
958 bits with the most significance have the lowest address location.</dd>
959 <dt><tt>e</tt></dt>
960 <dd>Specifies that hte target lays out data in little-endian form. That is,
961 the bits with the least significance have the lowest address location.</dd>
962 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
963 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
964 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
965 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
966 too.</dd>
967 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
968 <dd>This specifies the alignment for an integer type of a given bit
969 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
970 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
971 <dd>This specifies the alignment for a vector type of a given bit
972 <i>size</i>.</dd>
973 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
974 <dd>This specifies the alignment for a floating point type of a given bit
975 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
976 (double).</dd>
977 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
978 <dd>This specifies the alignment for an aggregate type of a given bit
979 <i>size</i>.</dd>
980</dl>
981<p>When constructing the data layout for a given target, LLVM starts with a
982default set of specifications which are then (possibly) overriden by the
983specifications in the <tt>datalayout</tt> keyword. The default specifications
984are given in this list:</p>
985<ul>
986 <li><tt>E</tt> - big endian</li>
987 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
988 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
989 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
990 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
991 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
992 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
993 alignment of 64-bits</li>
994 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
995 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
996 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
997 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
998 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
999</ul>
1000<p>When llvm is determining the alignment for a given type, it uses the
1001following rules:
1002<ol>
1003 <li>If the type sought is an exact match for one of the specifications, that
1004 specification is used.</li>
1005 <li>If no match is found, and the type sought is an integer type, then the
1006 smallest integer type that is larger than the bitwidth of the sought type is
1007 used. If none of the specifications are larger than the bitwidth then the the
1008 largest integer type is used. For example, given the default specifications
1009 above, the i7 type will use the alignment of i8 (next largest) while both
1010 i65 and i256 will use the alignment of i64 (largest specified).</li>
1011 <li>If no match is found, and the type sought is a vector type, then the
1012 largest vector type that is smaller than the sought vector type will be used
1013 as a fall back. This happens because <128 x double> can be implemented in
1014 terms of 64 <2 x double>, for example.</li>
1015</ol>
1016</div>
1017
1018<!-- *********************************************************************** -->
1019<div class="doc_section"> <a name="typesystem">Type System</a> </div>
1020<!-- *********************************************************************** -->
1021
1022<div class="doc_text">
1023
1024<p>The LLVM type system is one of the most important features of the
1025intermediate representation. Being typed enables a number of
1026optimizations to be performed on the IR directly, without having to do
1027extra analyses on the side before the transformation. A strong type
1028system makes it easier to read the generated code and enables novel
1029analyses and transformations that are not feasible to perform on normal
1030three address code representations.</p>
1031
1032</div>
1033
1034<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1035<div class="doc_subsection"> <a name="t_classifications">Type
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1036Classifications</a> </div>
1037<div class="doc_text">
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1038<p>The types fall into a few useful
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1039classifications:</p>
1040
1041<table border="1" cellspacing="0" cellpadding="4">
1042 <tbody>
1043 <tr><th>Classification</th><th>Types</th></tr>
1044 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1045 <td><a href="#t_integer">integer</a></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1046 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1047 </tr>
1048 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1049 <td><a href="#t_floating">floating point</a></td>
1050 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1051 </tr>
1052 <tr>
1053 <td><a name="t_firstclass">first class</a></td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1054 <td><a href="#t_integer">integer</a>,
1055 <a href="#t_floating">floating point</a>,
1056 <a href="#t_pointer">pointer</a>,
Dan Gohmanf6237db2008-06-18 18:42:13 +0000[diff] [blame]1057 <a href="#t_vector">vector</a>,
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]1058 <a href="#t_struct">structure</a>,
1059 <a href="#t_array">array</a>,
Dan Gohmand13951e2008-05-23 22:50:26 +0000[diff] [blame]1060 <a href="#t_label">label</a>.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1061 </td>
1062 </tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1063 <tr>
1064 <td><a href="#t_primitive">primitive</a></td>
1065 <td><a href="#t_label">label</a>,
1066 <a href="#t_void">void</a>,
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1067 <a href="#t_floating">floating point</a>.</td>
1068 </tr>
1069 <tr>
1070 <td><a href="#t_derived">derived</a></td>
1071 <td><a href="#t_integer">integer</a>,
1072 <a href="#t_array">array</a>,
1073 <a href="#t_function">function</a>,
1074 <a href="#t_pointer">pointer</a>,
1075 <a href="#t_struct">structure</a>,
1076 <a href="#t_pstruct">packed structure</a>,
1077 <a href="#t_vector">vector</a>,
1078 <a href="#t_opaque">opaque</a>.
1079 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1080 </tbody>
1081</table>
1082
1083<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1084most important. Values of these types are the only ones which can be
1085produced by instructions, passed as arguments, or used as operands to
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]1086instructions.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1087</div>
1088
1089<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1090<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1091
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1092<div class="doc_text">
1093<p>The primitive types are the fundamental building blocks of the LLVM
1094system.</p>
1095
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1096</div>
1097
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1098<!-- _______________________________________________________________________ -->
1099<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1100
1101<div class="doc_text">
1102 <table>
1103 <tbody>
1104 <tr><th>Type</th><th>Description</th></tr>
1105 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1106 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1107 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1108 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1109 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1110 </tbody>
1111 </table>
1112</div>
1113
1114<!-- _______________________________________________________________________ -->
1115<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1116
1117<div class="doc_text">
1118<h5>Overview:</h5>
1119<p>The void type does not represent any value and has no size.</p>
1120
1121<h5>Syntax:</h5>
1122
1123<pre>
1124 void
1125</pre>
1126</div>
1127
1128<!-- _______________________________________________________________________ -->
1129<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1130
1131<div class="doc_text">
1132<h5>Overview:</h5>
1133<p>The label type represents code labels.</p>
1134
1135<h5>Syntax:</h5>
1136
1137<pre>
1138 label
1139</pre>
1140</div>
1141
1142
1143<!-- ======================================================================= -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1144<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1145
1146<div class="doc_text">
1147
1148<p>The real power in LLVM comes from the derived types in the system.
1149This is what allows a programmer to represent arrays, functions,
1150pointers, and other useful types. Note that these derived types may be
1151recursive: For example, it is possible to have a two dimensional array.</p>
1152
1153</div>
1154
1155<!-- _______________________________________________________________________ -->
1156<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1157
1158<div class="doc_text">
1159
1160<h5>Overview:</h5>
1161<p>The integer type is a very simple derived type that simply specifies an
1162arbitrary bit width for the integer type desired. Any bit width from 1 bit to
11632^23-1 (about 8 million) can be specified.</p>
1164
1165<h5>Syntax:</h5>
1166
1167<pre>
1168 iN
1169</pre>
1170
1171<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1172value.</p>
1173
1174<h5>Examples:</h5>
1175<table class="layout">
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1176 <tbody>
1177 <tr>
1178 <td><tt>i1</tt></td>
1179 <td>a single-bit integer.</td>
1180 </tr><tr>
1181 <td><tt>i32</tt></td>
1182 <td>a 32-bit integer.</td>
1183 </tr><tr>
1184 <td><tt>i1942652</tt></td>
1185 <td>a really big integer of over 1 million bits.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1186 </tr>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1187 </tbody>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1188</table>
1189</div>
1190
1191<!-- _______________________________________________________________________ -->
1192<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1193
1194<div class="doc_text">
1195
1196<h5>Overview:</h5>
1197
1198<p>The array type is a very simple derived type that arranges elements
1199sequentially in memory. The array type requires a size (number of
1200elements) and an underlying data type.</p>
1201
1202<h5>Syntax:</h5>
1203
1204<pre>
1205 [&lt;# elements&gt; x &lt;elementtype&gt;]
1206</pre>
1207
1208<p>The number of elements is a constant integer value; elementtype may
1209be any type with a size.</p>
1210
1211<h5>Examples:</h5>
1212<table class="layout">
1213 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1214 <td class="left"><tt>[40 x i32]</tt></td>
1215 <td class="left">Array of 40 32-bit integer values.</td>
1216 </tr>
1217 <tr class="layout">
1218 <td class="left"><tt>[41 x i32]</tt></td>
1219 <td class="left">Array of 41 32-bit integer values.</td>
1220 </tr>
1221 <tr class="layout">
1222 <td class="left"><tt>[4 x i8]</tt></td>
1223 <td class="left">Array of 4 8-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1224 </tr>
1225</table>
1226<p>Here are some examples of multidimensional arrays:</p>
1227<table class="layout">
1228 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1229 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1230 <td class="left">3x4 array of 32-bit integer values.</td>
1231 </tr>
1232 <tr class="layout">
1233 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1234 <td class="left">12x10 array of single precision floating point values.</td>
1235 </tr>
1236 <tr class="layout">
1237 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1238 <td class="left">2x3x4 array of 16-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1239 </tr>
1240</table>
1241
1242<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1243length array. Normally, accesses past the end of an array are undefined in
1244LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1245As a special case, however, zero length arrays are recognized to be variable
1246length. This allows implementation of 'pascal style arrays' with the LLVM
1247type "{ i32, [0 x float]}", for example.</p>
1248
1249</div>
1250
1251<!-- _______________________________________________________________________ -->
1252<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1253<div class="doc_text">
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1254
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1255<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1256
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1257<p>The function type can be thought of as a function signature. It
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1258consists of a return type and a list of formal parameter types. The
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1259return 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]1260If 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]1261class types, and the struct must have at least one element.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1262
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1263<h5>Syntax:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1264
1265<pre>
1266 &lt;returntype list&gt; (&lt;parameter list&gt;)
1267</pre>
1268
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1269<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1270specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1271which indicates that the function takes a variable number of arguments.
1272Variable argument functions can access their arguments with the <a
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1273 href="#int_varargs">variable argument handling intrinsic</a> functions.
1274'<tt>&lt;returntype list&gt;</tt>' is a comma-separated list of
1275<a href="#t_firstclass">first class</a> type specifiers.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1276
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1277<h5>Examples:</h5>
1278<table class="layout">
1279 <tr class="layout">
1280 <td class="left"><tt>i32 (i32)</tt></td>
1281 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1282 </td>
1283 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1284 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1285 </tt></td>
1286 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1287 an <tt>i16</tt> that should be sign extended and a
1288 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1289 <tt>float</tt>.
1290 </td>
1291 </tr><tr class="layout">
1292 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1293 <td class="left">A vararg function that takes at least one
1294 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1295 which returns an integer. This is the signature for <tt>printf</tt> in
1296 LLVM.
1297 </td>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1298 </tr><tr class="layout">
1299 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Devang Patel6dddba22008-03-24 18:10:52 +0000[diff] [blame]1300 <td class="left">A function taking an <tt>i32></tt>, returning two
1301 <tt> i32 </tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1302 </td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1303 </tr>
1304</table>
1305
1306</div>
1307<!-- _______________________________________________________________________ -->
1308<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1309<div class="doc_text">
1310<h5>Overview:</h5>
1311<p>The structure type is used to represent a collection of data members
1312together in memory. The packing of the field types is defined to match
1313the ABI of the underlying processor. The elements of a structure may
1314be any type that has a size.</p>
1315<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1316and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1317field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1318instruction.</p>
1319<h5>Syntax:</h5>
1320<pre> { &lt;type list&gt; }<br></pre>
1321<h5>Examples:</h5>
1322<table class="layout">
1323 <tr class="layout">
1324 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1325 <td class="left">A triple of three <tt>i32</tt> values</td>
1326 </tr><tr class="layout">
1327 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1328 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1329 second element is a <a href="#t_pointer">pointer</a> to a
1330 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1331 an <tt>i32</tt>.</td>
1332 </tr>
1333</table>
1334</div>
1335
1336<!-- _______________________________________________________________________ -->
1337<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1338</div>
1339<div class="doc_text">
1340<h5>Overview:</h5>
1341<p>The packed structure type is used to represent a collection of data members
1342together in memory. There is no padding between fields. Further, the alignment
1343of a packed structure is 1 byte. The elements of a packed structure may
1344be any type that has a size.</p>
1345<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1346and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1347field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1348instruction.</p>
1349<h5>Syntax:</h5>
1350<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1351<h5>Examples:</h5>
1352<table class="layout">
1353 <tr class="layout">
1354 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1355 <td class="left">A triple of three <tt>i32</tt> values</td>
1356 </tr><tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1357 <td class="left"><tt>&lt; { float, i32 (i32)* } &gt;</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1358 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1359 second element is a <a href="#t_pointer">pointer</a> to a
1360 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1361 an <tt>i32</tt>.</td>
1362 </tr>
1363</table>
1364</div>
1365
1366<!-- _______________________________________________________________________ -->
1367<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1368<div class="doc_text">
1369<h5>Overview:</h5>
1370<p>As in many languages, the pointer type represents a pointer or
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1371reference to another object, which must live in memory. Pointer types may have
1372an optional address space attribute defining the target-specific numbered
1373address space where the pointed-to object resides. The default address space is
1374zero.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1375<h5>Syntax:</h5>
1376<pre> &lt;type&gt; *<br></pre>
1377<h5>Examples:</h5>
1378<table class="layout">
1379 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1380 <td class="left"><tt>[4x i32]*</tt></td>
1381 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1382 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1383 </tr>
1384 <tr class="layout">
1385 <td class="left"><tt>i32 (i32 *) *</tt></td>
1386 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1387 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1388 <tt>i32</tt>.</td>
1389 </tr>
1390 <tr class="layout">
1391 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1392 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1393 that resides in address space #5.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1394 </tr>
1395</table>
1396</div>
1397
1398<!-- _______________________________________________________________________ -->
1399<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1400<div class="doc_text">
1401
1402<h5>Overview:</h5>
1403
1404<p>A vector type is a simple derived type that represents a vector
1405of elements. Vector types are used when multiple primitive data
1406are operated in parallel using a single instruction (SIMD).
1407A vector type requires a size (number of
1408elements) and an underlying primitive data type. Vectors must have a power
1409of two length (1, 2, 4, 8, 16 ...). Vector types are
1410considered <a href="#t_firstclass">first class</a>.</p>
1411
1412<h5>Syntax:</h5>
1413
1414<pre>
1415 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1416</pre>
1417
1418<p>The number of elements is a constant integer value; elementtype may
1419be any integer or floating point type.</p>
1420
1421<h5>Examples:</h5>
1422
1423<table class="layout">
1424 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1425 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1426 <td class="left">Vector of 4 32-bit integer values.</td>
1427 </tr>
1428 <tr class="layout">
1429 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1430 <td class="left">Vector of 8 32-bit floating-point values.</td>
1431 </tr>
1432 <tr class="layout">
1433 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1434 <td class="left">Vector of 2 64-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1435 </tr>
1436</table>
1437</div>
1438
1439<!-- _______________________________________________________________________ -->
1440<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1441<div class="doc_text">
1442
1443<h5>Overview:</h5>
1444
1445<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1446corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1447In LLVM, opaque types can eventually be resolved to any type (not just a
1448structure type).</p>
1449
1450<h5>Syntax:</h5>
1451
1452<pre>
1453 opaque
1454</pre>
1455
1456<h5>Examples:</h5>
1457
1458<table class="layout">
1459 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1460 <td class="left"><tt>opaque</tt></td>
1461 <td class="left">An opaque type.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1462 </tr>
1463</table>
1464</div>
1465
1466
1467<!-- *********************************************************************** -->
1468<div class="doc_section"> <a name="constants">Constants</a> </div>
1469<!-- *********************************************************************** -->
1470
1471<div class="doc_text">
1472
1473<p>LLVM has several different basic types of constants. This section describes
1474them all and their syntax.</p>
1475
1476</div>
1477
1478<!-- ======================================================================= -->
1479<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1480
1481<div class="doc_text">
1482
1483<dl>
1484 <dt><b>Boolean constants</b></dt>
1485
1486 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1487 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1488 </dd>
1489
1490 <dt><b>Integer constants</b></dt>
1491
1492 <dd>Standard integers (such as '4') are constants of the <a
1493 href="#t_integer">integer</a> type. Negative numbers may be used with
1494 integer types.
1495 </dd>
1496
1497 <dt><b>Floating point constants</b></dt>
1498
1499 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1500 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]1501 notation (see below). The assembler requires the exact decimal value of
1502 a floating-point constant. For example, the assembler accepts 1.25 but
1503 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1504 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1505
1506 <dt><b>Null pointer constants</b></dt>
1507
1508 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1509 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1510
1511</dl>
1512
1513<p>The one non-intuitive notation for constants is the optional hexadecimal form
1514of floating point constants. For example, the form '<tt>double
15150x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
15164.5e+15</tt>'. The only time hexadecimal floating point constants are required
1517(and the only time that they are generated by the disassembler) is when a
1518floating point constant must be emitted but it cannot be represented as a
1519decimal floating point number. For example, NaN's, infinities, and other
1520special values are represented in their IEEE hexadecimal format so that
1521assembly and disassembly do not cause any bits to change in the constants.</p>
1522
1523</div>
1524
1525<!-- ======================================================================= -->
1526<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1527</div>
1528
1529<div class="doc_text">
1530<p>Aggregate constants arise from aggregation of simple constants
1531and smaller aggregate constants.</p>
1532
1533<dl>
1534 <dt><b>Structure constants</b></dt>
1535
1536 <dd>Structure constants are represented with notation similar to structure
1537 type definitions (a comma separated list of elements, surrounded by braces
Chris Lattner6c8de962007-12-25 20:34:52 +0000[diff] [blame]1538 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1539 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]1540 must have <a href="#t_struct">structure type</a>, and the number and
1541 types of elements must match those specified by the type.
1542 </dd>
1543
1544 <dt><b>Array constants</b></dt>
1545
1546 <dd>Array constants are represented with notation similar to array type
1547 definitions (a comma separated list of elements, surrounded by square brackets
1548 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1549 constants must have <a href="#t_array">array type</a>, and the number and
1550 types of elements must match those specified by the type.
1551 </dd>
1552
1553 <dt><b>Vector constants</b></dt>
1554
1555 <dd>Vector constants are represented with notation similar to vector type
1556 definitions (a comma separated list of elements, surrounded by
1557 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1558 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1559 href="#t_vector">vector type</a>, and the number and types of elements must
1560 match those specified by the type.
1561 </dd>
1562
1563 <dt><b>Zero initialization</b></dt>
1564
1565 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1566 value to zero of <em>any</em> type, including scalar and aggregate types.
1567 This is often used to avoid having to print large zero initializers (e.g. for
1568 large arrays) and is always exactly equivalent to using explicit zero
1569 initializers.
1570 </dd>
1571</dl>
1572
1573</div>
1574
1575<!-- ======================================================================= -->
1576<div class="doc_subsection">
1577 <a name="globalconstants">Global Variable and Function Addresses</a>
1578</div>
1579
1580<div class="doc_text">
1581
1582<p>The addresses of <a href="#globalvars">global variables</a> and <a
1583href="#functionstructure">functions</a> are always implicitly valid (link-time)
1584constants. These constants are explicitly referenced when the <a
1585href="#identifiers">identifier for the global</a> is used and always have <a
1586href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1587file:</p>
1588
1589<div class="doc_code">
1590<pre>
1591@X = global i32 17
1592@Y = global i32 42
1593@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1594</pre>
1595</div>
1596
1597</div>
1598
1599<!-- ======================================================================= -->
1600<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1601<div class="doc_text">
1602 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1603 no specific value. Undefined values may be of any type and be used anywhere
1604 a constant is permitted.</p>
1605
1606 <p>Undefined values indicate to the compiler that the program is well defined
1607 no matter what value is used, giving the compiler more freedom to optimize.
1608 </p>
1609</div>
1610
1611<!-- ======================================================================= -->
1612<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1613</div>
1614
1615<div class="doc_text">
1616
1617<p>Constant expressions are used to allow expressions involving other constants
1618to be used as constants. Constant expressions may be of any <a
1619href="#t_firstclass">first class</a> type and may involve any LLVM operation
1620that does not have side effects (e.g. load and call are not supported). The
1621following is the syntax for constant expressions:</p>
1622
1623<dl>
1624 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1625 <dd>Truncate a constant to another type. The bit size of CST must be larger
1626 than the bit size of TYPE. Both types must be integers.</dd>
1627
1628 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1629 <dd>Zero extend a constant to another type. The bit size of CST must be
1630 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1631
1632 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1633 <dd>Sign extend a constant to another type. The bit size of CST must be
1634 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1635
1636 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1637 <dd>Truncate a floating point constant to another floating point type. The
1638 size of CST must be larger than the size of TYPE. Both types must be
1639 floating point.</dd>
1640
1641 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1642 <dd>Floating point extend a constant to another type. The size of CST must be
1643 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1644
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1645 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1646 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1647 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1648 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1649 of the same number of elements. If the value won't fit in the integer type,
1650 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1651
1652 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1653 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1654 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1655 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1656 of the same number of elements. If the value won't fit in the integer type,
1657 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1658
1659 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1660 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1661 constant. TYPE must be a scalar or vector floating point type. CST must be of
1662 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1663 of the same number of elements. If the value won't fit in the floating point
1664 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1665
1666 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1667 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1668 constant. TYPE must be a scalar or vector floating point type. CST must be of
1669 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1670 of the same number of elements. If the value won't fit in the floating point
1671 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1672
1673 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1674 <dd>Convert a pointer typed constant to the corresponding integer constant
1675 TYPE must be an integer type. CST must be of pointer type. The CST value is
1676 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1677
1678 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1679 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1680 pointer type. CST must be of integer type. The CST value is zero extended,
1681 truncated, or unchanged to make it fit in a pointer size. This one is
1682 <i>really</i> dangerous!</dd>
1683
1684 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1685 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1686 identical (same number of bits). The conversion is done as if the CST value
1687 was stored to memory and read back as TYPE. In other words, no bits change
1688 with this operator, just the type. This can be used for conversion of
1689 vector types to any other type, as long as they have the same bit width. For
1690 pointers it is only valid to cast to another pointer type.
1691 </dd>
1692
1693 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1694
1695 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1696 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1697 instruction, the index list may have zero or more indexes, which are required
1698 to make sense for the type of "CSTPTR".</dd>
1699
1700 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1701
1702 <dd>Perform the <a href="#i_select">select operation</a> on
1703 constants.</dd>
1704
1705 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1706 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1707
1708 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1709 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1710
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]1711 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
1712 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
1713
1714 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
1715 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
1716
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1717 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1718
1719 <dd>Perform the <a href="#i_extractelement">extractelement
1720 operation</a> on constants.
1721
1722 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1723
1724 <dd>Perform the <a href="#i_insertelement">insertelement
1725 operation</a> on constants.</dd>
1726
1727
1728 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1729
1730 <dd>Perform the <a href="#i_shufflevector">shufflevector
1731 operation</a> on constants.</dd>
1732
1733 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1734
1735 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1736 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1737 binary</a> operations. The constraints on operands are the same as those for
1738 the corresponding instruction (e.g. no bitwise operations on floating point
1739 values are allowed).</dd>
1740</dl>
1741</div>
1742
1743<!-- *********************************************************************** -->
1744<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1745<!-- *********************************************************************** -->
1746
1747<!-- ======================================================================= -->
1748<div class="doc_subsection">
1749<a name="inlineasm">Inline Assembler Expressions</a>
1750</div>
1751
1752<div class="doc_text">
1753
1754<p>
1755LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1756Module-Level Inline Assembly</a>) through the use of a special value. This
1757value represents the inline assembler as a string (containing the instructions
1758to emit), a list of operand constraints (stored as a string), and a flag that
1759indicates whether or not the inline asm expression has side effects. An example
1760inline assembler expression is:
1761</p>
1762
1763<div class="doc_code">
1764<pre>
1765i32 (i32) asm "bswap $0", "=r,r"
1766</pre>
1767</div>
1768
1769<p>
1770Inline assembler expressions may <b>only</b> be used as the callee operand of
1771a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1772</p>
1773
1774<div class="doc_code">
1775<pre>
1776%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1777</pre>
1778</div>
1779
1780<p>
1781Inline asms with side effects not visible in the constraint list must be marked
1782as having side effects. This is done through the use of the
1783'<tt>sideeffect</tt>' keyword, like so:
1784</p>
1785
1786<div class="doc_code">
1787<pre>
1788call void asm sideeffect "eieio", ""()
1789</pre>
1790</div>
1791
1792<p>TODO: The format of the asm and constraints string still need to be
1793documented here. Constraints on what can be done (e.g. duplication, moving, etc
1794need to be documented).
1795</p>
1796
1797</div>
1798
1799<!-- *********************************************************************** -->
1800<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1801<!-- *********************************************************************** -->
1802
1803<div class="doc_text">
1804
1805<p>The LLVM instruction set consists of several different
1806classifications of instructions: <a href="#terminators">terminator
1807instructions</a>, <a href="#binaryops">binary instructions</a>,
1808<a href="#bitwiseops">bitwise binary instructions</a>, <a
1809 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1810instructions</a>.</p>
1811
1812</div>
1813
1814<!-- ======================================================================= -->
1815<div class="doc_subsection"> <a name="terminators">Terminator
1816Instructions</a> </div>
1817
1818<div class="doc_text">
1819
1820<p>As mentioned <a href="#functionstructure">previously</a>, every
1821basic block in a program ends with a "Terminator" instruction, which
1822indicates which block should be executed after the current block is
1823finished. These terminator instructions typically yield a '<tt>void</tt>'
1824value: they produce control flow, not values (the one exception being
1825the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1826<p>There are six different terminator instructions: the '<a
1827 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1828instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1829the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1830 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1831 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1832
1833</div>
1834
1835<!-- _______________________________________________________________________ -->
1836<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1837Instruction</a> </div>
1838<div class="doc_text">
1839<h5>Syntax:</h5>
1840<pre> ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
1841 ret void <i>; Return from void function</i>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1842 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]1843</pre>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1844
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1845<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1846
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1847<p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1848value) from a function back to the caller.</p>
1849<p>There are two forms of the '<tt>ret</tt>' instruction: one that
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1850returns value(s) and then causes control flow, and one that just causes
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1851control flow to occur.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1852
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1853<h5>Arguments:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1854
1855<p>The '<tt>ret</tt>' instruction may return zero, one or multiple values.
1856The type of each return value must be a '<a href="#t_firstclass">first
1857class</a>' type. Note that a function is not <a href="#wellformed">well
1858formed</a> if there exists a '<tt>ret</tt>' instruction inside of the
1859function that returns values that do not match the return type of the
1860function.</p>
1861
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1862<h5>Semantics:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1863
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1864<p>When the '<tt>ret</tt>' instruction is executed, control flow
1865returns back to the calling function's context. If the caller is a "<a
1866 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1867the instruction after the call. If the caller was an "<a
1868 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1869at the beginning of the "normal" destination block. If the instruction
1870returns a value, that value shall set the call or invoke instruction's
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1871return value. If the instruction returns multiple values then these
Devang Patelec8a5b02008-03-11 05:51:59 +0000[diff] [blame]1872values can only be accessed through a '<a href="#i_getresult"><tt>getresult</tt>
1873</a>' instruction.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1874
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1875<h5>Example:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1876
1877<pre>
1878 ret i32 5 <i>; Return an integer value of 5</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1879 ret void <i>; Return from a void function</i>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1880 ret i32 4, i8 2 <i>; Return two values 4 and 2 </i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1881</pre>
1882</div>
1883<!-- _______________________________________________________________________ -->
1884<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1885<div class="doc_text">
1886<h5>Syntax:</h5>
1887<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
1888</pre>
1889<h5>Overview:</h5>
1890<p>The '<tt>br</tt>' instruction is used to cause control flow to
1891transfer to a different basic block in the current function. There are
1892two forms of this instruction, corresponding to a conditional branch
1893and an unconditional branch.</p>
1894<h5>Arguments:</h5>
1895<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1896single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1897unconditional form of the '<tt>br</tt>' instruction takes a single
1898'<tt>label</tt>' value as a target.</p>
1899<h5>Semantics:</h5>
1900<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1901argument is evaluated. If the value is <tt>true</tt>, control flows
1902to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1903control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1904<h5>Example:</h5>
1905<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
1906 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1907</div>
1908<!-- _______________________________________________________________________ -->
1909<div class="doc_subsubsection">
1910 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1911</div>
1912
1913<div class="doc_text">
1914<h5>Syntax:</h5>
1915
1916<pre>
1917 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
1918</pre>
1919
1920<h5>Overview:</h5>
1921
1922<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1923several different places. It is a generalization of the '<tt>br</tt>'
1924instruction, allowing a branch to occur to one of many possible
1925destinations.</p>
1926
1927
1928<h5>Arguments:</h5>
1929
1930<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1931comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1932an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1933table is not allowed to contain duplicate constant entries.</p>
1934
1935<h5>Semantics:</h5>
1936
1937<p>The <tt>switch</tt> instruction specifies a table of values and
1938destinations. When the '<tt>switch</tt>' instruction is executed, this
1939table is searched for the given value. If the value is found, control flow is
1940transfered to the corresponding destination; otherwise, control flow is
1941transfered to the default destination.</p>
1942
1943<h5>Implementation:</h5>
1944
1945<p>Depending on properties of the target machine and the particular
1946<tt>switch</tt> instruction, this instruction may be code generated in different
1947ways. For example, it could be generated as a series of chained conditional
1948branches or with a lookup table.</p>
1949
1950<h5>Example:</h5>
1951
1952<pre>
1953 <i>; Emulate a conditional br instruction</i>
1954 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1955 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1956
1957 <i>; Emulate an unconditional br instruction</i>
1958 switch i32 0, label %dest [ ]
1959
1960 <i>; Implement a jump table:</i>
1961 switch i32 %val, label %otherwise [ i32 0, label %onzero
1962 i32 1, label %onone
1963 i32 2, label %ontwo ]
1964</pre>
1965</div>
1966
1967<!-- _______________________________________________________________________ -->
1968<div class="doc_subsubsection">
1969 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1970</div>
1971
1972<div class="doc_text">
1973
1974<h5>Syntax:</h5>
1975
1976<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]1977 &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]1978 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
1979</pre>
1980
1981<h5>Overview:</h5>
1982
1983<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1984function, with the possibility of control flow transfer to either the
1985'<tt>normal</tt>' label or the
1986'<tt>exception</tt>' label. If the callee function returns with the
1987"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1988"normal" label. If the callee (or any indirect callees) returns with the "<a
1989href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1990continued at the dynamically nearest "exception" label. If the callee function
Devang Patelec8a5b02008-03-11 05:51:59 +0000[diff] [blame]1991returns multiple values then individual return values are only accessible through
1992a '<tt><a href="#i_getresult">getresult</a></tt>' instruction.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1993
1994<h5>Arguments:</h5>
1995
1996<p>This instruction requires several arguments:</p>
1997
1998<ol>
1999 <li>
2000 The optional "cconv" marker indicates which <a href="#callingconv">calling
2001 convention</a> the call should use. If none is specified, the call defaults
2002 to using C calling conventions.
2003 </li>
2004 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2005 function value being invoked. In most cases, this is a direct function
2006 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2007 an arbitrary pointer to function value.
2008 </li>
2009
2010 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2011 function to be invoked. </li>
2012
2013 <li>'<tt>function args</tt>': argument list whose types match the function
2014 signature argument types. If the function signature indicates the function
2015 accepts a variable number of arguments, the extra arguments can be
2016 specified. </li>
2017
2018 <li>'<tt>normal label</tt>': the label reached when the called function
2019 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2020
2021 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2022 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2023
2024</ol>
2025
2026<h5>Semantics:</h5>
2027
2028<p>This instruction is designed to operate as a standard '<tt><a
2029href="#i_call">call</a></tt>' instruction in most regards. The primary
2030difference is that it establishes an association with a label, which is used by
2031the runtime library to unwind the stack.</p>
2032
2033<p>This instruction is used in languages with destructors to ensure that proper
2034cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2035exception. Additionally, this is important for implementation of
2036'<tt>catch</tt>' clauses in high-level languages that support them.</p>
2037
2038<h5>Example:</h5>
2039<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2040 %retval = invoke i32 @Test(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2041 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2042 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2043 unwind label %TestCleanup <i>; {i32}:retval set</i>
2044</pre>
2045</div>
2046
2047
2048<!-- _______________________________________________________________________ -->
2049
2050<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2051Instruction</a> </div>
2052
2053<div class="doc_text">
2054
2055<h5>Syntax:</h5>
2056<pre>
2057 unwind
2058</pre>
2059
2060<h5>Overview:</h5>
2061
2062<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2063at the first callee in the dynamic call stack which used an <a
2064href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2065primarily used to implement exception handling.</p>
2066
2067<h5>Semantics:</h5>
2068
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2069<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2070immediately halt. The dynamic call stack is then searched for the first <a
2071href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2072execution continues at the "exceptional" destination block specified by the
2073<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2074dynamic call chain, undefined behavior results.</p>
2075</div>
2076
2077<!-- _______________________________________________________________________ -->
2078
2079<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2080Instruction</a> </div>
2081
2082<div class="doc_text">
2083
2084<h5>Syntax:</h5>
2085<pre>
2086 unreachable
2087</pre>
2088
2089<h5>Overview:</h5>
2090
2091<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2092instruction is used to inform the optimizer that a particular portion of the
2093code is not reachable. This can be used to indicate that the code after a
2094no-return function cannot be reached, and other facts.</p>
2095
2096<h5>Semantics:</h5>
2097
2098<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2099</div>
2100
2101
2102
2103<!-- ======================================================================= -->
2104<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2105<div class="doc_text">
2106<p>Binary operators are used to do most of the computation in a
Chris Lattnerab596d92008-04-01 18:47:32 +0000[diff] [blame]2107program. They require two operands of the same type, execute an operation on them, and
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2108produce a single value. The operands might represent
2109multiple 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]2110The result value has the same type as its operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2111<p>There are several different binary operators:</p>
2112</div>
2113<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2114<div class="doc_subsubsection">
2115 <a name="i_add">'<tt>add</tt>' Instruction</a>
2116</div>
2117
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2118<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2119
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2120<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2121
2122<pre>
2123 &lt;result&gt; = add &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2124</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2125
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2126<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2127
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2128<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2129
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2130<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2131
2132<p>The two arguments to the '<tt>add</tt>' instruction must be <a
2133 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>, or
2134 <a href="#t_vector">vector</a> values. Both arguments must have identical
2135 types.</p>
2136
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2137<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2138
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2139<p>The value produced is the integer or floating point sum of the two
2140operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2141
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2142<p>If an integer sum has unsigned overflow, the result returned is the
2143mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2144the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2145
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2146<p>Because LLVM integers use a two's complement representation, this
2147instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2148
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2149<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2150
2151<pre>
2152 &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2153</pre>
2154</div>
2155<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2156<div class="doc_subsubsection">
2157 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2158</div>
2159
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2160<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2161
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2162<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2163
2164<pre>
2165 &lt;result&gt; = sub &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2166</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2167
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2168<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2169
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2170<p>The '<tt>sub</tt>' instruction returns the difference of its two
2171operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2172
2173<p>Note that the '<tt>sub</tt>' instruction is used to represent the
2174'<tt>neg</tt>' instruction present in most other intermediate
2175representations.</p>
2176
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2177<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2178
2179<p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2180 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2181 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2182 types.</p>
2183
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2184<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2185
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2186<p>The value produced is the integer or floating point difference of
2187the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2188
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2189<p>If an integer difference has unsigned overflow, the result returned is the
2190mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2191the result.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2192
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2193<p>Because LLVM integers use a two's complement representation, this
2194instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2195
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2196<h5>Example:</h5>
2197<pre>
2198 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2199 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2200</pre>
2201</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2202
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2203<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2204<div class="doc_subsubsection">
2205 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2206</div>
2207
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2208<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2209
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2210<h5>Syntax:</h5>
2211<pre> &lt;result&gt; = mul &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2212</pre>
2213<h5>Overview:</h5>
2214<p>The '<tt>mul</tt>' instruction returns the product of its two
2215operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2216
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2217<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2218
2219<p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2220href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2221or <a href="#t_vector">vector</a> values. Both arguments must have identical
2222types.</p>
2223
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2224<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2225
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2226<p>The value produced is the integer or floating point product of the
2227two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2228
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2229<p>If the result of an integer multiplication has unsigned overflow,
2230the result returned is the mathematical result modulo
22312<sup>n</sup>, where n is the bit width of the result.</p>
2232<p>Because LLVM integers use a two's complement representation, and the
2233result is the same width as the operands, this instruction returns the
2234correct result for both signed and unsigned integers. If a full product
2235(e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2236should be sign-extended or zero-extended as appropriate to the
2237width of the full product.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2238<h5>Example:</h5>
2239<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2240</pre>
2241</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2242
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2243<!-- _______________________________________________________________________ -->
2244<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2245</a></div>
2246<div class="doc_text">
2247<h5>Syntax:</h5>
2248<pre> &lt;result&gt; = udiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2249</pre>
2250<h5>Overview:</h5>
2251<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2252operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2253
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2254<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2255
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2256<p>The two arguments to the '<tt>udiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2257<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2258values. Both arguments must have identical types.</p>
2259
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2260<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2261
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2262<p>The value produced is the unsigned integer quotient of the two operands.</p>
2263<p>Note that unsigned integer division and signed integer division are distinct
2264operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2265<p>Division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2266<h5>Example:</h5>
2267<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2268</pre>
2269</div>
2270<!-- _______________________________________________________________________ -->
2271<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2272</a> </div>
2273<div class="doc_text">
2274<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2275<pre>
2276 &lt;result&gt; = sdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2277</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2278
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2279<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2280
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2281<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2282operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2283
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2284<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2285
2286<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2287<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2288values. Both arguments must have identical types.</p>
2289
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2290<h5>Semantics:</h5>
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2291<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]2292<p>Note that signed integer division and unsigned integer division are distinct
2293operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2294<p>Division by zero leads to undefined behavior. Overflow also leads to
2295undefined behavior; this is a rare case, but can occur, for example,
2296by doing a 32-bit division of -2147483648 by -1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2297<h5>Example:</h5>
2298<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2299</pre>
2300</div>
2301<!-- _______________________________________________________________________ -->
2302<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2303Instruction</a> </div>
2304<div class="doc_text">
2305<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2306<pre>
2307 &lt;result&gt; = fdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2308</pre>
2309<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2310
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2311<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2312operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2313
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2314<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2315
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2316<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2317<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2318of floating point values. Both arguments must have identical types.</p>
2319
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2320<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2321
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2322<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2323
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2324<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2325
2326<pre>
2327 &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]2328</pre>
2329</div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2330
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2331<!-- _______________________________________________________________________ -->
2332<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2333</div>
2334<div class="doc_text">
2335<h5>Syntax:</h5>
2336<pre> &lt;result&gt; = urem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2337</pre>
2338<h5>Overview:</h5>
2339<p>The '<tt>urem</tt>' instruction returns the remainder from the
2340unsigned division of its two arguments.</p>
2341<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2342<p>The two arguments to the '<tt>urem</tt>' instruction must be
2343<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2344values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2345<h5>Semantics:</h5>
2346<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2347This instruction always performs an unsigned division to get the remainder.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2348<p>Note that unsigned integer remainder and signed integer remainder are
2349distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2350<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2351<h5>Example:</h5>
2352<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2353</pre>
2354
2355</div>
2356<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2357<div class="doc_subsubsection">
2358 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2359</div>
2360
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2361<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2362
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2363<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2364
2365<pre>
2366 &lt;result&gt; = srem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2367</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2368
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2369<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2370
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2371<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2372signed division of its two operands. This instruction can also take
2373<a href="#t_vector">vector</a> versions of the values in which case
2374the elements must be integers.</p>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]2375
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2376<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2377
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2378<p>The two arguments to the '<tt>srem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2379<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2380values. Both arguments must have identical types.</p>
2381
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2382<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2383
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2384<p>This instruction returns the <i>remainder</i> of a division (where the result
2385has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2386operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2387a value. For more information about the difference, see <a
2388 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2389Math Forum</a>. For a table of how this is implemented in various languages,
2390please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2391Wikipedia: modulo operation</a>.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2392<p>Note that signed integer remainder and unsigned integer remainder are
2393distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2394<p>Taking the remainder of a division by zero leads to undefined behavior.
2395Overflow also leads to undefined behavior; this is a rare case, but can occur,
2396for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2397(The remainder doesn't actually overflow, but this rule lets srem be
2398implemented using instructions that return both the result of the division
2399and the remainder.)</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2400<h5>Example:</h5>
2401<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2402</pre>
2403
2404</div>
2405<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2406<div class="doc_subsubsection">
2407 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2408
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2409<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2410
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2411<h5>Syntax:</h5>
2412<pre> &lt;result&gt; = frem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2413</pre>
2414<h5>Overview:</h5>
2415<p>The '<tt>frem</tt>' instruction returns the remainder from the
2416division of its two operands.</p>
2417<h5>Arguments:</h5>
2418<p>The two arguments to the '<tt>frem</tt>' instruction must be
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2419<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2420of floating point values. Both arguments must have identical types.</p>
2421
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2422<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2423
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2424<p>This instruction returns the <i>remainder</i> of a division.
2425The remainder has the same sign as the dividend.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2426
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2427<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2428
2429<pre>
2430 &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]2431</pre>
2432</div>
2433
2434<!-- ======================================================================= -->
2435<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2436Operations</a> </div>
2437<div class="doc_text">
2438<p>Bitwise binary operators are used to do various forms of
2439bit-twiddling in a program. They are generally very efficient
2440instructions and can commonly be strength reduced from other
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2441instructions. They require two operands of the same type, execute an operation on them,
2442and 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]2443</div>
2444
2445<!-- _______________________________________________________________________ -->
2446<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2447Instruction</a> </div>
2448<div class="doc_text">
2449<h5>Syntax:</h5>
2450<pre> &lt;result&gt; = shl &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2451</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2452
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2453<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2454
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2455<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2456the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2457
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2458<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2459
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2460<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2461 href="#t_integer">integer</a> type. '<tt>var2</tt>' is treated as an
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2462unsigned value. This instruction does not support
2463<a href="#t_vector">vector</a> operands.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2464
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2465<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2466
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2467<p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup> mod 2<sup>n</sup>,
2468where n is the width of the result. If <tt>var2</tt> is (statically or dynamically) negative or
2469equal to or larger than the number of bits in <tt>var1</tt>, the result is undefined.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2470
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2471<h5>Example:</h5><pre>
2472 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2473 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2474 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2475 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2476</pre>
2477</div>
2478<!-- _______________________________________________________________________ -->
2479<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2480Instruction</a> </div>
2481<div class="doc_text">
2482<h5>Syntax:</h5>
2483<pre> &lt;result&gt; = lshr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2484</pre>
2485
2486<h5>Overview:</h5>
2487<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2488operand shifted to the right a specified number of bits with zero fill.</p>
2489
2490<h5>Arguments:</h5>
2491<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2492<a href="#t_integer">integer</a> type. '<tt>var2</tt>' is treated as an
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2493unsigned value. This instruction does not support
2494<a href="#t_vector">vector</a> operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2495
2496<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2497
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2498<p>This instruction always performs a logical shift right operation. The most
2499significant bits of the result will be filled with zero bits after the
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2500shift. If <tt>var2</tt> is (statically or dynamically) equal to or larger than
2501the number of bits in <tt>var1</tt>, the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2502
2503<h5>Example:</h5>
2504<pre>
2505 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2506 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2507 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2508 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2509 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2510</pre>
2511</div>
2512
2513<!-- _______________________________________________________________________ -->
2514<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2515Instruction</a> </div>
2516<div class="doc_text">
2517
2518<h5>Syntax:</h5>
2519<pre> &lt;result&gt; = ashr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2520</pre>
2521
2522<h5>Overview:</h5>
2523<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2524operand shifted to the right a specified number of bits with sign extension.</p>
2525
2526<h5>Arguments:</h5>
2527<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2528<a href="#t_integer">integer</a> type. '<tt>var2</tt>' is treated as an
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2529unsigned value. This instruction does not support
2530<a href="#t_vector">vector</a> operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2531
2532<h5>Semantics:</h5>
2533<p>This instruction always performs an arithmetic shift right operation,
2534The most significant bits of the result will be filled with the sign bit
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2535of <tt>var1</tt>. If <tt>var2</tt> is (statically or dynamically) equal to or
2536larger than the number of bits in <tt>var1</tt>, the result is undefined.
2537</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2538
2539<h5>Example:</h5>
2540<pre>
2541 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2542 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2543 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2544 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2545 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2546</pre>
2547</div>
2548
2549<!-- _______________________________________________________________________ -->
2550<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2551Instruction</a> </div>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2552
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2553<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2554
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2555<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2556
2557<pre>
2558 &lt;result&gt; = and &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2559</pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2560
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2561<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2562
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2563<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2564its two operands.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2565
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2566<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2567
2568<p>The two arguments to the '<tt>and</tt>' instruction must be
2569<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2570values. Both arguments must have identical types.</p>
2571
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2572<h5>Semantics:</h5>
2573<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2574<p> </p>
2575<div style="align: center">
2576<table border="1" cellspacing="0" cellpadding="4">
2577 <tbody>
2578 <tr>
2579 <td>In0</td>
2580 <td>In1</td>
2581 <td>Out</td>
2582 </tr>
2583 <tr>
2584 <td>0</td>
2585 <td>0</td>
2586 <td>0</td>
2587 </tr>
2588 <tr>
2589 <td>0</td>
2590 <td>1</td>
2591 <td>0</td>
2592 </tr>
2593 <tr>
2594 <td>1</td>
2595 <td>0</td>
2596 <td>0</td>
2597 </tr>
2598 <tr>
2599 <td>1</td>
2600 <td>1</td>
2601 <td>1</td>
2602 </tr>
2603 </tbody>
2604</table>
2605</div>
2606<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2607<pre>
2608 &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]2609 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2610 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2611</pre>
2612</div>
2613<!-- _______________________________________________________________________ -->
2614<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2615<div class="doc_text">
2616<h5>Syntax:</h5>
2617<pre> &lt;result&gt; = or &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2618</pre>
2619<h5>Overview:</h5>
2620<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2621or of its two operands.</p>
2622<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2623
2624<p>The two arguments to the '<tt>or</tt>' instruction must be
2625<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2626values. Both arguments must have identical types.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2627<h5>Semantics:</h5>
2628<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2629<p> </p>
2630<div style="align: center">
2631<table border="1" cellspacing="0" cellpadding="4">
2632 <tbody>
2633 <tr>
2634 <td>In0</td>
2635 <td>In1</td>
2636 <td>Out</td>
2637 </tr>
2638 <tr>
2639 <td>0</td>
2640 <td>0</td>
2641 <td>0</td>
2642 </tr>
2643 <tr>
2644 <td>0</td>
2645 <td>1</td>
2646 <td>1</td>
2647 </tr>
2648 <tr>
2649 <td>1</td>
2650 <td>0</td>
2651 <td>1</td>
2652 </tr>
2653 <tr>
2654 <td>1</td>
2655 <td>1</td>
2656 <td>1</td>
2657 </tr>
2658 </tbody>
2659</table>
2660</div>
2661<h5>Example:</h5>
2662<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2663 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2664 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2665</pre>
2666</div>
2667<!-- _______________________________________________________________________ -->
2668<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2669Instruction</a> </div>
2670<div class="doc_text">
2671<h5>Syntax:</h5>
2672<pre> &lt;result&gt; = xor &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2673</pre>
2674<h5>Overview:</h5>
2675<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2676or of its two operands. The <tt>xor</tt> is used to implement the
2677"one's complement" operation, which is the "~" operator in C.</p>
2678<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2679<p>The two arguments to the '<tt>xor</tt>' instruction must be
2680<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2681values. Both arguments must have identical types.</p>
2682
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2683<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]2684
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2685<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2686<p> </p>
2687<div style="align: center">
2688<table border="1" cellspacing="0" cellpadding="4">
2689 <tbody>
2690 <tr>
2691 <td>In0</td>
2692 <td>In1</td>
2693 <td>Out</td>
2694 </tr>
2695 <tr>
2696 <td>0</td>
2697 <td>0</td>
2698 <td>0</td>
2699 </tr>
2700 <tr>
2701 <td>0</td>
2702 <td>1</td>
2703 <td>1</td>
2704 </tr>
2705 <tr>
2706 <td>1</td>
2707 <td>0</td>
2708 <td>1</td>
2709 </tr>
2710 <tr>
2711 <td>1</td>
2712 <td>1</td>
2713 <td>0</td>
2714 </tr>
2715 </tbody>
2716</table>
2717</div>
2718<p> </p>
2719<h5>Example:</h5>
2720<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2721 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2722 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2723 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2724</pre>
2725</div>
2726
2727<!-- ======================================================================= -->
2728<div class="doc_subsection">
2729 <a name="vectorops">Vector Operations</a>
2730</div>
2731
2732<div class="doc_text">
2733
2734<p>LLVM supports several instructions to represent vector operations in a
2735target-independent manner. These instructions cover the element-access and
2736vector-specific operations needed to process vectors effectively. While LLVM
2737does directly support these vector operations, many sophisticated algorithms
2738will want to use target-specific intrinsics to take full advantage of a specific
2739target.</p>
2740
2741</div>
2742
2743<!-- _______________________________________________________________________ -->
2744<div class="doc_subsubsection">
2745 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2746</div>
2747
2748<div class="doc_text">
2749
2750<h5>Syntax:</h5>
2751
2752<pre>
2753 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
2754</pre>
2755
2756<h5>Overview:</h5>
2757
2758<p>
2759The '<tt>extractelement</tt>' instruction extracts a single scalar
2760element from a vector at a specified index.
2761</p>
2762
2763
2764<h5>Arguments:</h5>
2765
2766<p>
2767The first operand of an '<tt>extractelement</tt>' instruction is a
2768value of <a href="#t_vector">vector</a> type. The second operand is
2769an index indicating the position from which to extract the element.
2770The index may be a variable.</p>
2771
2772<h5>Semantics:</h5>
2773
2774<p>
2775The result is a scalar of the same type as the element type of
2776<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2777<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2778results are undefined.
2779</p>
2780
2781<h5>Example:</h5>
2782
2783<pre>
2784 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
2785</pre>
2786</div>
2787
2788
2789<!-- _______________________________________________________________________ -->
2790<div class="doc_subsubsection">
2791 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2792</div>
2793
2794<div class="doc_text">
2795
2796<h5>Syntax:</h5>
2797
2798<pre>
Dan Gohmanbcc3c502008-05-12 23:38:42 +0000[diff] [blame]2799 &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]2800</pre>
2801
2802<h5>Overview:</h5>
2803
2804<p>
2805The '<tt>insertelement</tt>' instruction inserts a scalar
2806element into a vector at a specified index.
2807</p>
2808
2809
2810<h5>Arguments:</h5>
2811
2812<p>
2813The first operand of an '<tt>insertelement</tt>' instruction is a
2814value of <a href="#t_vector">vector</a> type. The second operand is a
2815scalar value whose type must equal the element type of the first
2816operand. The third operand is an index indicating the position at
2817which to insert the value. The index may be a variable.</p>
2818
2819<h5>Semantics:</h5>
2820
2821<p>
2822The result is a vector of the same type as <tt>val</tt>. Its
2823element values are those of <tt>val</tt> except at position
2824<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2825exceeds the length of <tt>val</tt>, the results are undefined.
2826</p>
2827
2828<h5>Example:</h5>
2829
2830<pre>
2831 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
2832</pre>
2833</div>
2834
2835<!-- _______________________________________________________________________ -->
2836<div class="doc_subsubsection">
2837 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2838</div>
2839
2840<div class="doc_text">
2841
2842<h5>Syntax:</h5>
2843
2844<pre>
2845 &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>
2846</pre>
2847
2848<h5>Overview:</h5>
2849
2850<p>
2851The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2852from two input vectors, returning a vector of the same type.
2853</p>
2854
2855<h5>Arguments:</h5>
2856
2857<p>
2858The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2859with types that match each other and types that match the result of the
2860instruction. The third argument is a shuffle mask, which has the same number
2861of elements as the other vector type, but whose element type is always 'i32'.
2862</p>
2863
2864<p>
2865The shuffle mask operand is required to be a constant vector with either
2866constant integer or undef values.
2867</p>
2868
2869<h5>Semantics:</h5>
2870
2871<p>
2872The elements of the two input vectors are numbered from left to right across
2873both of the vectors. The shuffle mask operand specifies, for each element of
2874the result vector, which element of the two input registers the result element
2875gets. The element selector may be undef (meaning "don't care") and the second
2876operand may be undef if performing a shuffle from only one vector.
2877</p>
2878
2879<h5>Example:</h5>
2880
2881<pre>
2882 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
2883 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
2884 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
2885 &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.
2886</pre>
2887</div>
2888
2889
2890<!-- ======================================================================= -->
2891<div class="doc_subsection">
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2892 <a name="aggregateops">Aggregate Operations</a>
2893</div>
2894
2895<div class="doc_text">
2896
2897<p>LLVM supports several instructions for working with aggregate values.
2898</p>
2899
2900</div>
2901
2902<!-- _______________________________________________________________________ -->
2903<div class="doc_subsubsection">
2904 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
2905</div>
2906
2907<div class="doc_text">
2908
2909<h5>Syntax:</h5>
2910
2911<pre>
2912 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
2913</pre>
2914
2915<h5>Overview:</h5>
2916
2917<p>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2918The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
2919or array element from an aggregate value.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2920</p>
2921
2922
2923<h5>Arguments:</h5>
2924
2925<p>
2926The first operand of an '<tt>extractvalue</tt>' instruction is a
2927value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2928type. The operands are constant indices to specify which value to extract
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]2929in a similar manner as indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2930'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
2931</p>
2932
2933<h5>Semantics:</h5>
2934
2935<p>
2936The result is the value at the position in the aggregate specified by
2937the index operands.
2938</p>
2939
2940<h5>Example:</h5>
2941
2942<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]2943 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2944</pre>
2945</div>
2946
2947
2948<!-- _______________________________________________________________________ -->
2949<div class="doc_subsubsection">
2950 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
2951</div>
2952
2953<div class="doc_text">
2954
2955<h5>Syntax:</h5>
2956
2957<pre>
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]2958 &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]2959</pre>
2960
2961<h5>Overview:</h5>
2962
2963<p>
2964The '<tt>insertvalue</tt>' instruction inserts a value
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2965into a struct field or array element in an aggregate.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2966</p>
2967
2968
2969<h5>Arguments:</h5>
2970
2971<p>
2972The first operand of an '<tt>insertvalue</tt>' instruction is a
2973value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
2974The second operand is a first-class value to insert.
Dan Gohman4f29e422008-05-23 21:53:15 +0000[diff] [blame]2975The following operands are constant indices
Dan Gohmane5febe42008-05-31 00:58:22 +0000[diff] [blame]2976indicating the position at which to insert the value in a similar manner as
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2977indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2978'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
2979The value to insert must have the same type as the value identified
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2980by the indices.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2981
2982<h5>Semantics:</h5>
2983
2984<p>
2985The result is an aggregate of the same type as <tt>val</tt>. Its
2986value is that of <tt>val</tt> except that the value at the position
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2987specified by the indices is that of <tt>elt</tt>.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2988</p>
2989
2990<h5>Example:</h5>
2991
2992<pre>
Dan Gohmanb1aab4e2008-06-23 15:26:37 +0000[diff] [blame]2993 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2994</pre>
2995</div>
2996
2997
2998<!-- ======================================================================= -->
2999<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3000 <a name="memoryops">Memory Access and Addressing Operations</a>
3001</div>
3002
3003<div class="doc_text">
3004
3005<p>A key design point of an SSA-based representation is how it
3006represents memory. In LLVM, no memory locations are in SSA form, which
3007makes things very simple. This section describes how to read, write,
3008allocate, and free memory in LLVM.</p>
3009
3010</div>
3011
3012<!-- _______________________________________________________________________ -->
3013<div class="doc_subsubsection">
3014 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3015</div>
3016
3017<div class="doc_text">
3018
3019<h5>Syntax:</h5>
3020
3021<pre>
3022 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3023</pre>
3024
3025<h5>Overview:</h5>
3026
3027<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3028heap and returns a pointer to it. The object is always allocated in the generic
3029address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3030
3031<h5>Arguments:</h5>
3032
3033<p>The '<tt>malloc</tt>' instruction allocates
3034<tt>sizeof(&lt;type&gt;)*NumElements</tt>
3035bytes of memory from the operating system and returns a pointer of the
3036appropriate type to the program. If "NumElements" is specified, it is the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3037number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3038If a constant alignment is specified, the value result of the allocation is guaranteed to
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3039be aligned to at least that boundary. If not specified, or if zero, the target can
3040choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3041
3042<p>'<tt>type</tt>' must be a sized type.</p>
3043
3044<h5>Semantics:</h5>
3045
3046<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3047a pointer is returned. The result of a zero byte allocattion is undefined. The
3048result is null if there is insufficient memory available.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3049
3050<h5>Example:</h5>
3051
3052<pre>
3053 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
3054
3055 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3056 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3057 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3058 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3059 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3060</pre>
3061</div>
3062
3063<!-- _______________________________________________________________________ -->
3064<div class="doc_subsubsection">
3065 <a name="i_free">'<tt>free</tt>' Instruction</a>
3066</div>
3067
3068<div class="doc_text">
3069
3070<h5>Syntax:</h5>
3071
3072<pre>
3073 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
3074</pre>
3075
3076<h5>Overview:</h5>
3077
3078<p>The '<tt>free</tt>' instruction returns memory back to the unused
3079memory heap to be reallocated in the future.</p>
3080
3081<h5>Arguments:</h5>
3082
3083<p>'<tt>value</tt>' shall be a pointer value that points to a value
3084that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3085instruction.</p>
3086
3087<h5>Semantics:</h5>
3088
3089<p>Access to the memory pointed to by the pointer is no longer defined
Chris Lattner329d6532008-04-19 22:41:32 +0000[diff] [blame]3090after this instruction executes. If the pointer is null, the operation
3091is a noop.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3092
3093<h5>Example:</h5>
3094
3095<pre>
3096 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
3097 free [4 x i8]* %array
3098</pre>
3099</div>
3100
3101<!-- _______________________________________________________________________ -->
3102<div class="doc_subsubsection">
3103 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3104</div>
3105
3106<div class="doc_text">
3107
3108<h5>Syntax:</h5>
3109
3110<pre>
3111 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3112</pre>
3113
3114<h5>Overview:</h5>
3115
3116<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3117currently executing function, to be automatically released when this function
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3118returns to its caller. The object is always allocated in the generic address
3119space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3120
3121<h5>Arguments:</h5>
3122
3123<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
3124bytes of memory on the runtime stack, returning a pointer of the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3125appropriate type to the program. If "NumElements" is specified, it is the
3126number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3127If a constant alignment is specified, the value result of the allocation is guaranteed
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3128to be aligned to at least that boundary. If not specified, or if zero, the target
3129can choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3130
3131<p>'<tt>type</tt>' may be any sized type.</p>
3132
3133<h5>Semantics:</h5>
3134
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3135<p>Memory is allocated; a pointer is returned. The operation is undefiend if
3136there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3137memory is automatically released when the function returns. The '<tt>alloca</tt>'
3138instruction is commonly used to represent automatic variables that must
3139have an address available. When the function returns (either with the <tt><a
3140 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]3141instructions), the memory is reclaimed. Allocating zero bytes
3142is legal, but the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3143
3144<h5>Example:</h5>
3145
3146<pre>
3147 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3148 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3149 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3150 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3151</pre>
3152</div>
3153
3154<!-- _______________________________________________________________________ -->
3155<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3156Instruction</a> </div>
3157<div class="doc_text">
3158<h5>Syntax:</h5>
3159<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>
3160<h5>Overview:</h5>
3161<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3162<h5>Arguments:</h5>
3163<p>The argument to the '<tt>load</tt>' instruction specifies the memory
3164address from which to load. The pointer must point to a <a
3165 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3166marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3167the number or order of execution of this <tt>load</tt> with other
3168volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3169instructions. </p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3170<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3171The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3172(that is, the alignment of the memory address). A value of 0 or an
3173omitted "align" argument means that the operation has the preferential
3174alignment for the target. It is the responsibility of the code emitter
3175to ensure that the alignment information is correct. Overestimating
3176the alignment results in an undefined behavior. Underestimating the
3177alignment may produce less efficient code. An alignment of 1 is always
3178safe.
3179</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3180<h5>Semantics:</h5>
3181<p>The location of memory pointed to is loaded.</p>
3182<h5>Examples:</h5>
3183<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3184 <a
3185 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3186 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3187</pre>
3188</div>
3189<!-- _______________________________________________________________________ -->
3190<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3191Instruction</a> </div>
3192<div class="doc_text">
3193<h5>Syntax:</h5>
3194<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3195 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3196</pre>
3197<h5>Overview:</h5>
3198<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3199<h5>Arguments:</h5>
3200<p>There are two arguments to the '<tt>store</tt>' instruction: a value
3201to 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]3202operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3203of the '<tt>&lt;value&gt;</tt>'
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3204operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3205optimizer is not allowed to modify the number or order of execution of
3206this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3207 href="#i_store">store</a></tt> instructions.</p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3208<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3209The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3210(that is, the alignment of the memory address). A value of 0 or an
3211omitted "align" argument means that the operation has the preferential
3212alignment for the target. It is the responsibility of the code emitter
3213to ensure that the alignment information is correct. Overestimating
3214the alignment results in an undefined behavior. Underestimating the
3215alignment may produce less efficient code. An alignment of 1 is always
3216safe.
3217</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3218<h5>Semantics:</h5>
3219<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
3220at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
3221<h5>Example:</h5>
3222<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]3223 store i32 3, i32* %ptr <i>; yields {void}</i>
3224 %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]3225</pre>
3226</div>
3227
3228<!-- _______________________________________________________________________ -->
3229<div class="doc_subsubsection">
3230 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3231</div>
3232
3233<div class="doc_text">
3234<h5>Syntax:</h5>
3235<pre>
3236 &lt;result&gt; = getelementptr &lt;ty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
3237</pre>
3238
3239<h5>Overview:</h5>
3240
3241<p>
3242The '<tt>getelementptr</tt>' instruction is used to get the address of a
3243subelement of an aggregate data structure.</p>
3244
3245<h5>Arguments:</h5>
3246
3247<p>This instruction takes a list of integer operands that indicate what
3248elements of the aggregate object to index to. The actual types of the arguments
3249provided depend on the type of the first pointer argument. The
3250'<tt>getelementptr</tt>' instruction is used to index down through the type
3251levels of a structure or to a specific index in an array. When indexing into a
3252structure, only <tt>i32</tt> integer constants are allowed. When indexing
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]3253into an array or pointer, only integers of 32 or 64 bits are allowed; 32-bit
3254values will be sign extended to 64-bits if required.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3255
3256<p>For example, let's consider a C code fragment and how it gets
3257compiled to LLVM:</p>
3258
3259<div class="doc_code">
3260<pre>
3261struct RT {
3262 char A;
3263 int B[10][20];
3264 char C;
3265};
3266struct ST {
3267 int X;
3268 double Y;
3269 struct RT Z;
3270};
3271
3272int *foo(struct ST *s) {
3273 return &amp;s[1].Z.B[5][13];
3274}
3275</pre>
3276</div>
3277
3278<p>The LLVM code generated by the GCC frontend is:</p>
3279
3280<div class="doc_code">
3281<pre>
3282%RT = type { i8 , [10 x [20 x i32]], i8 }
3283%ST = type { i32, double, %RT }
3284
3285define i32* %foo(%ST* %s) {
3286entry:
3287 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3288 ret i32* %reg
3289}
3290</pre>
3291</div>
3292
3293<h5>Semantics:</h5>
3294
3295<p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
3296on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
3297and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
3298<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]3299to 64-bits. <a href="#t_struct">Structure</a> and <a href="#t_pstruct">packed
3300structure</a> types require <tt>i32</tt> <b>constants</b>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3301
3302<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3303type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3304}</tt>' type, a structure. The second index indexes into the third element of
3305the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3306i8 }</tt>' type, another structure. The third index indexes into the second
3307element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3308array. The two dimensions of the array are subscripted into, yielding an
3309'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3310to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3311
3312<p>Note that it is perfectly legal to index partially through a
3313structure, returning a pointer to an inner element. Because of this,
3314the LLVM code for the given testcase is equivalent to:</p>
3315
3316<pre>
3317 define i32* %foo(%ST* %s) {
3318 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3319 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3320 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3321 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3322 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3323 ret i32* %t5
3324 }
3325</pre>
3326
3327<p>Note that it is undefined to access an array out of bounds: array and
3328pointer indexes must always be within the defined bounds of the array type.
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]3329The one exception for this rule is zero length arrays. These arrays are
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3330defined to be accessible as variable length arrays, which requires access
3331beyond the zero'th element.</p>
3332
3333<p>The getelementptr instruction is often confusing. For some more insight
3334into how it works, see <a href="GetElementPtr.html">the getelementptr
3335FAQ</a>.</p>
3336
3337<h5>Example:</h5>
3338
3339<pre>
3340 <i>; yields [12 x i8]*:aptr</i>
3341 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
3342</pre>
3343</div>
3344
3345<!-- ======================================================================= -->
3346<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3347</div>
3348<div class="doc_text">
3349<p>The instructions in this category are the conversion instructions (casting)
3350which all take a single operand and a type. They perform various bit conversions
3351on the operand.</p>
3352</div>
3353
3354<!-- _______________________________________________________________________ -->
3355<div class="doc_subsubsection">
3356 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3357</div>
3358<div class="doc_text">
3359
3360<h5>Syntax:</h5>
3361<pre>
3362 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3363</pre>
3364
3365<h5>Overview:</h5>
3366<p>
3367The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3368</p>
3369
3370<h5>Arguments:</h5>
3371<p>
3372The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3373be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3374and type of the result, which must be an <a href="#t_integer">integer</a>
3375type. The bit size of <tt>value</tt> must be larger than the bit size of
3376<tt>ty2</tt>. Equal sized types are not allowed.</p>
3377
3378<h5>Semantics:</h5>
3379<p>
3380The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3381and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3382larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3383It will always truncate bits.</p>
3384
3385<h5>Example:</h5>
3386<pre>
3387 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3388 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3389 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3390</pre>
3391</div>
3392
3393<!-- _______________________________________________________________________ -->
3394<div class="doc_subsubsection">
3395 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3396</div>
3397<div class="doc_text">
3398
3399<h5>Syntax:</h5>
3400<pre>
3401 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3402</pre>
3403
3404<h5>Overview:</h5>
3405<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3406<tt>ty2</tt>.</p>
3407
3408
3409<h5>Arguments:</h5>
3410<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3411<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3412also be of <a href="#t_integer">integer</a> type. The bit size of the
3413<tt>value</tt> must be smaller than the bit size of the destination type,
3414<tt>ty2</tt>.</p>
3415
3416<h5>Semantics:</h5>
3417<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3418bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3419
3420<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3421
3422<h5>Example:</h5>
3423<pre>
3424 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3425 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3426</pre>
3427</div>
3428
3429<!-- _______________________________________________________________________ -->
3430<div class="doc_subsubsection">
3431 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3432</div>
3433<div class="doc_text">
3434
3435<h5>Syntax:</h5>
3436<pre>
3437 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3438</pre>
3439
3440<h5>Overview:</h5>
3441<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3442
3443<h5>Arguments:</h5>
3444<p>
3445The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3446<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3447also be of <a href="#t_integer">integer</a> type. The bit size of the
3448<tt>value</tt> must be smaller than the bit size of the destination type,
3449<tt>ty2</tt>.</p>
3450
3451<h5>Semantics:</h5>
3452<p>
3453The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3454bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3455the type <tt>ty2</tt>.</p>
3456
3457<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3458
3459<h5>Example:</h5>
3460<pre>
3461 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3462 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3463</pre>
3464</div>
3465
3466<!-- _______________________________________________________________________ -->
3467<div class="doc_subsubsection">
3468 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3469</div>
3470
3471<div class="doc_text">
3472
3473<h5>Syntax:</h5>
3474
3475<pre>
3476 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3477</pre>
3478
3479<h5>Overview:</h5>
3480<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3481<tt>ty2</tt>.</p>
3482
3483
3484<h5>Arguments:</h5>
3485<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3486 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3487cast it to. The size of <tt>value</tt> must be larger than the size of
3488<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3489<i>no-op cast</i>.</p>
3490
3491<h5>Semantics:</h5>
3492<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3493<a href="#t_floating">floating point</a> type to a smaller
3494<a href="#t_floating">floating point</a> type. If the value cannot fit within
3495the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3496
3497<h5>Example:</h5>
3498<pre>
3499 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3500 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3501</pre>
3502</div>
3503
3504<!-- _______________________________________________________________________ -->
3505<div class="doc_subsubsection">
3506 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3507</div>
3508<div class="doc_text">
3509
3510<h5>Syntax:</h5>
3511<pre>
3512 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3513</pre>
3514
3515<h5>Overview:</h5>
3516<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3517floating point value.</p>
3518
3519<h5>Arguments:</h5>
3520<p>The '<tt>fpext</tt>' instruction takes a
3521<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3522and a <a href="#t_floating">floating point</a> type to cast it to. The source
3523type must be smaller than the destination type.</p>
3524
3525<h5>Semantics:</h5>
3526<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3527<a href="#t_floating">floating point</a> type to a larger
3528<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3529used to make a <i>no-op cast</i> because it always changes bits. Use
3530<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3531
3532<h5>Example:</h5>
3533<pre>
3534 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3535 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3536</pre>
3537</div>
3538
3539<!-- _______________________________________________________________________ -->
3540<div class="doc_subsubsection">
3541 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3542</div>
3543<div class="doc_text">
3544
3545<h5>Syntax:</h5>
3546<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3547 &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]3548</pre>
3549
3550<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3551<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3552unsigned integer equivalent of type <tt>ty2</tt>.
3553</p>
3554
3555<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3556<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]3557scalar or vector <a href="#t_floating">floating point</a> value, and a type
3558to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3559type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3560vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3561
3562<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3563<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3564<a href="#t_floating">floating point</a> operand into the nearest (rounding
3565towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3566the results are undefined.</p>
3567
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3568<h5>Example:</h5>
3569<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3570 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3571 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3572 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3573</pre>
3574</div>
3575
3576<!-- _______________________________________________________________________ -->
3577<div class="doc_subsubsection">
3578 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3579</div>
3580<div class="doc_text">
3581
3582<h5>Syntax:</h5>
3583<pre>
3584 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3585</pre>
3586
3587<h5>Overview:</h5>
3588<p>The '<tt>fptosi</tt>' instruction converts
3589<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3590</p>
3591
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3592<h5>Arguments:</h5>
3593<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]3594scalar or vector <a href="#t_floating">floating point</a> value, and a type
3595to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3596type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3597vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3598
3599<h5>Semantics:</h5>
3600<p>The '<tt>fptosi</tt>' instruction converts its
3601<a href="#t_floating">floating point</a> operand into the nearest (rounding
3602towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3603the results are undefined.</p>
3604
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3605<h5>Example:</h5>
3606<pre>
3607 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3608 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3609 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3610</pre>
3611</div>
3612
3613<!-- _______________________________________________________________________ -->
3614<div class="doc_subsubsection">
3615 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3616</div>
3617<div class="doc_text">
3618
3619<h5>Syntax:</h5>
3620<pre>
3621 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3622</pre>
3623
3624<h5>Overview:</h5>
3625<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3626integer and converts that value to the <tt>ty2</tt> type.</p>
3627
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3628<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3629<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3630scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3631to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3632type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3633floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3634
3635<h5>Semantics:</h5>
3636<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3637integer quantity and converts it to the corresponding floating point value. If
3638the value cannot fit in the floating point value, the results are undefined.</p>
3639
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3640<h5>Example:</h5>
3641<pre>
3642 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3643 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3644</pre>
3645</div>
3646
3647<!-- _______________________________________________________________________ -->
3648<div class="doc_subsubsection">
3649 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3650</div>
3651<div class="doc_text">
3652
3653<h5>Syntax:</h5>
3654<pre>
3655 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3656</pre>
3657
3658<h5>Overview:</h5>
3659<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3660integer and converts that value to the <tt>ty2</tt> type.</p>
3661
3662<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3663<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3664scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3665to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3666type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3667floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3668
3669<h5>Semantics:</h5>
3670<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3671integer quantity and converts it to the corresponding floating point value. If
3672the value cannot fit in the floating point value, the results are undefined.</p>
3673
3674<h5>Example:</h5>
3675<pre>
3676 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3677 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3678</pre>
3679</div>
3680
3681<!-- _______________________________________________________________________ -->
3682<div class="doc_subsubsection">
3683 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3684</div>
3685<div class="doc_text">
3686
3687<h5>Syntax:</h5>
3688<pre>
3689 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3690</pre>
3691
3692<h5>Overview:</h5>
3693<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3694the integer type <tt>ty2</tt>.</p>
3695
3696<h5>Arguments:</h5>
3697<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3698must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3699<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3700
3701<h5>Semantics:</h5>
3702<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3703<tt>ty2</tt> by interpreting the pointer value as an integer and either
3704truncating or zero extending that value to the size of the integer type. If
3705<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3706<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3707are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3708change.</p>
3709
3710<h5>Example:</h5>
3711<pre>
3712 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3713 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3714</pre>
3715</div>
3716
3717<!-- _______________________________________________________________________ -->
3718<div class="doc_subsubsection">
3719 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3720</div>
3721<div class="doc_text">
3722
3723<h5>Syntax:</h5>
3724<pre>
3725 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3726</pre>
3727
3728<h5>Overview:</h5>
3729<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3730a pointer type, <tt>ty2</tt>.</p>
3731
3732<h5>Arguments:</h5>
3733<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3734value to cast, and a type to cast it to, which must be a
3735<a href="#t_pointer">pointer</a> type.
3736
3737<h5>Semantics:</h5>
3738<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3739<tt>ty2</tt> by applying either a zero extension or a truncation depending on
3740the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3741size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3742the size of a pointer then a zero extension is done. If they are the same size,
3743nothing is done (<i>no-op cast</i>).</p>
3744
3745<h5>Example:</h5>
3746<pre>
3747 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3748 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3749 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3750</pre>
3751</div>
3752
3753<!-- _______________________________________________________________________ -->
3754<div class="doc_subsubsection">
3755 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3756</div>
3757<div class="doc_text">
3758
3759<h5>Syntax:</h5>
3760<pre>
3761 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3762</pre>
3763
3764<h5>Overview:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3765
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3766<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3767<tt>ty2</tt> without changing any bits.</p>
3768
3769<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3770
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3771<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3772a first class value, and a type to cast it to, which must also be a <a
3773 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3774and the destination type, <tt>ty2</tt>, must be identical. If the source
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3775type is a pointer, the destination type must also be a pointer. This
3776instruction supports bitwise conversion of vectors to integers and to vectors
3777of other types (as long as they have the same size).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3778
3779<h5>Semantics:</h5>
3780<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3781<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3782this conversion. The conversion is done as if the <tt>value</tt> had been
3783stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3784converted to other pointer types with this instruction. To convert pointers to
3785other types, use the <a href="#i_inttoptr">inttoptr</a> or
3786<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3787
3788<h5>Example:</h5>
3789<pre>
3790 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3791 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3792 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3793</pre>
3794</div>
3795
3796<!-- ======================================================================= -->
3797<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3798<div class="doc_text">
3799<p>The instructions in this category are the "miscellaneous"
3800instructions, which defy better classification.</p>
3801</div>
3802
3803<!-- _______________________________________________________________________ -->
3804<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3805</div>
3806<div class="doc_text">
3807<h5>Syntax:</h5>
3808<pre> &lt;result&gt; = icmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3809</pre>
3810<h5>Overview:</h5>
3811<p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3812of its two integer or pointer operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3813<h5>Arguments:</h5>
3814<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3815the condition code indicating the kind of comparison to perform. It is not
3816a value, just a keyword. The possible condition code are:
3817<ol>
3818 <li><tt>eq</tt>: equal</li>
3819 <li><tt>ne</tt>: not equal </li>
3820 <li><tt>ugt</tt>: unsigned greater than</li>
3821 <li><tt>uge</tt>: unsigned greater or equal</li>
3822 <li><tt>ult</tt>: unsigned less than</li>
3823 <li><tt>ule</tt>: unsigned less or equal</li>
3824 <li><tt>sgt</tt>: signed greater than</li>
3825 <li><tt>sge</tt>: signed greater or equal</li>
3826 <li><tt>slt</tt>: signed less than</li>
3827 <li><tt>sle</tt>: signed less or equal</li>
3828</ol>
3829<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3830<a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3831<h5>Semantics:</h5>
3832<p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3833the condition code given as <tt>cond</tt>. The comparison performed always
3834yields a <a href="#t_primitive">i1</a> result, as follows:
3835<ol>
3836 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3837 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3838 </li>
3839 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3840 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3841 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3842 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3843 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3844 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3845 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3846 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3847 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3848 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3849 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3850 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3851 <li><tt>sge</tt>: interprets the operands as signed values and yields
3852 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3853 <li><tt>slt</tt>: interprets the operands as signed values and yields
3854 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3855 <li><tt>sle</tt>: interprets the operands as signed values and yields
3856 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3857</ol>
3858<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3859values are compared as if they were integers.</p>
3860
3861<h5>Example:</h5>
3862<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
3863 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
3864 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
3865 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3866 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3867 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
3868</pre>
3869</div>
3870
3871<!-- _______________________________________________________________________ -->
3872<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3873</div>
3874<div class="doc_text">
3875<h5>Syntax:</h5>
3876<pre> &lt;result&gt; = fcmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3877</pre>
3878<h5>Overview:</h5>
3879<p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3880of its floating point operands.</p>
3881<h5>Arguments:</h5>
3882<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3883the condition code indicating the kind of comparison to perform. It is not
3884a value, just a keyword. The possible condition code are:
3885<ol>
3886 <li><tt>false</tt>: no comparison, always returns false</li>
3887 <li><tt>oeq</tt>: ordered and equal</li>
3888 <li><tt>ogt</tt>: ordered and greater than </li>
3889 <li><tt>oge</tt>: ordered and greater than or equal</li>
3890 <li><tt>olt</tt>: ordered and less than </li>
3891 <li><tt>ole</tt>: ordered and less than or equal</li>
3892 <li><tt>one</tt>: ordered and not equal</li>
3893 <li><tt>ord</tt>: ordered (no nans)</li>
3894 <li><tt>ueq</tt>: unordered or equal</li>
3895 <li><tt>ugt</tt>: unordered or greater than </li>
3896 <li><tt>uge</tt>: unordered or greater than or equal</li>
3897 <li><tt>ult</tt>: unordered or less than </li>
3898 <li><tt>ule</tt>: unordered or less than or equal</li>
3899 <li><tt>une</tt>: unordered or not equal</li>
3900 <li><tt>uno</tt>: unordered (either nans)</li>
3901 <li><tt>true</tt>: no comparison, always returns true</li>
3902</ol>
3903<p><i>Ordered</i> means that neither operand is a QNAN while
3904<i>unordered</i> means that either operand may be a QNAN.</p>
3905<p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3906<a href="#t_floating">floating point</a> typed. They must have identical
3907types.</p>
3908<h5>Semantics:</h5>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]3909<p>The '<tt>fcmp</tt>' instruction compares <tt>var1</tt> and <tt>var2</tt>
3910according to the condition code given as <tt>cond</tt>. The comparison performed
3911always yields a <a href="#t_primitive">i1</a> result, as follows:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3912<ol>
3913 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3914 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3915 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3916 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3917 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3918 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3919 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3920 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3921 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3922 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3923 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3924 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3925 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3926 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3927 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3928 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3929 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3930 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3931 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3932 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3933 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3934 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3935 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3936 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3937 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3938 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3939 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3940 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3941</ol>
3942
3943<h5>Example:</h5>
3944<pre> &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3945 &lt;result&gt; = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3946 &lt;result&gt; = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3947 &lt;result&gt; = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3948</pre>
3949</div>
3950
3951<!-- _______________________________________________________________________ -->
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]3952<div class="doc_subsubsection">
3953 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
3954</div>
3955<div class="doc_text">
3956<h5>Syntax:</h5>
3957<pre> &lt;result&gt; = vicmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
3958</pre>
3959<h5>Overview:</h5>
3960<p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
3961element-wise comparison of its two integer vector operands.</p>
3962<h5>Arguments:</h5>
3963<p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
3964the condition code indicating the kind of comparison to perform. It is not
3965a value, just a keyword. The possible condition code are:
3966<ol>
3967 <li><tt>eq</tt>: equal</li>
3968 <li><tt>ne</tt>: not equal </li>
3969 <li><tt>ugt</tt>: unsigned greater than</li>
3970 <li><tt>uge</tt>: unsigned greater or equal</li>
3971 <li><tt>ult</tt>: unsigned less than</li>
3972 <li><tt>ule</tt>: unsigned less or equal</li>
3973 <li><tt>sgt</tt>: signed greater than</li>
3974 <li><tt>sge</tt>: signed greater or equal</li>
3975 <li><tt>slt</tt>: signed less than</li>
3976 <li><tt>sle</tt>: signed less or equal</li>
3977</ol>
3978<p>The remaining two arguments must be <a href="#t_vector">vector</a> of
3979<a href="#t_integer">integer</a> typed. They must also be identical types.</p>
3980<h5>Semantics:</h5>
3981<p>The '<tt>vicmp</tt>' instruction compares <tt>var1</tt> and <tt>var2</tt>
3982according to the condition code given as <tt>cond</tt>. The comparison yields a
3983<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
3984identical type as the values being compared. The most significant bit in each
3985element is 1 if the element-wise comparison evaluates to true, and is 0
3986otherwise. All other bits of the result are undefined. The condition codes
3987are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
3988instruction</a>.
3989
3990<h5>Example:</h5>
3991<pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]3992 &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>
3993 &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]3994</pre>
3995</div>
3996
3997<!-- _______________________________________________________________________ -->
3998<div class="doc_subsubsection">
3999 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4000</div>
4001<div class="doc_text">
4002<h5>Syntax:</h5>
4003<pre> &lt;result&gt; = vfcmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt;</pre>
4004<h5>Overview:</h5>
4005<p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4006element-wise comparison of its two floating point vector operands. The output
4007elements have the same width as the input elements.</p>
4008<h5>Arguments:</h5>
4009<p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4010the condition code indicating the kind of comparison to perform. It is not
4011a value, just a keyword. The possible condition code are:
4012<ol>
4013 <li><tt>false</tt>: no comparison, always returns false</li>
4014 <li><tt>oeq</tt>: ordered and equal</li>
4015 <li><tt>ogt</tt>: ordered and greater than </li>
4016 <li><tt>oge</tt>: ordered and greater than or equal</li>
4017 <li><tt>olt</tt>: ordered and less than </li>
4018 <li><tt>ole</tt>: ordered and less than or equal</li>
4019 <li><tt>one</tt>: ordered and not equal</li>
4020 <li><tt>ord</tt>: ordered (no nans)</li>
4021 <li><tt>ueq</tt>: unordered or equal</li>
4022 <li><tt>ugt</tt>: unordered or greater than </li>
4023 <li><tt>uge</tt>: unordered or greater than or equal</li>
4024 <li><tt>ult</tt>: unordered or less than </li>
4025 <li><tt>ule</tt>: unordered or less than or equal</li>
4026 <li><tt>une</tt>: unordered or not equal</li>
4027 <li><tt>uno</tt>: unordered (either nans)</li>
4028 <li><tt>true</tt>: no comparison, always returns true</li>
4029</ol>
4030<p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4031<a href="#t_floating">floating point</a> typed. They must also be identical
4032types.</p>
4033<h5>Semantics:</h5>
4034<p>The '<tt>vfcmp</tt>' instruction compares <tt>var1</tt> and <tt>var2</tt>
4035according to the condition code given as <tt>cond</tt>. The comparison yields a
4036<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4037an identical number of elements as the values being compared, and each element
4038having identical with to the width of the floating point elements. The most
4039significant bit in each element is 1 if the element-wise comparison evaluates to
4040true, and is 0 otherwise. All other bits of the result are undefined. The
4041condition codes are evaluated identically to the
4042<a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.
4043
4044<h5>Example:</h5>
4045<pre>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4046 &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>
4047 &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]4048</pre>
4049</div>
4050
4051<!-- _______________________________________________________________________ -->
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4052<div class="doc_subsubsection">
4053 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4054</div>
4055
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4056<div class="doc_text">
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4057
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4058<h5>Syntax:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4059
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4060<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
4061<h5>Overview:</h5>
4062<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
4063the SSA graph representing the function.</p>
4064<h5>Arguments:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4065
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4066<p>The type of the incoming values is specified with the first type
4067field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4068as arguments, with one pair for each predecessor basic block of the
4069current block. Only values of <a href="#t_firstclass">first class</a>
4070type may be used as the value arguments to the PHI node. Only labels
4071may be used as the label arguments.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4072
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4073<p>There must be no non-phi instructions between the start of a basic
4074block and the PHI instructions: i.e. PHI instructions must be first in
4075a basic block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4076
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4077<h5>Semantics:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4078
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4079<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4080specified by the pair corresponding to the predecessor basic block that executed
4081just prior to the current block.</p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4082
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4083<h5>Example:</h5>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4084<pre>
4085Loop: ; Infinite loop that counts from 0 on up...
4086 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4087 %nextindvar = add i32 %indvar, 1
4088 br label %Loop
4089</pre>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4090</div>
4091
4092<!-- _______________________________________________________________________ -->
4093<div class="doc_subsubsection">
4094 <a name="i_select">'<tt>select</tt>' Instruction</a>
4095</div>
4096
4097<div class="doc_text">
4098
4099<h5>Syntax:</h5>
4100
4101<pre>
4102 &lt;result&gt; = select i1 &lt;cond&gt;, &lt;ty&gt; &lt;val1&gt;, &lt;ty&gt; &lt;val2&gt; <i>; yields ty</i>
4103</pre>
4104
4105<h5>Overview:</h5>
4106
4107<p>
4108The '<tt>select</tt>' instruction is used to choose one value based on a
4109condition, without branching.
4110</p>
4111
4112
4113<h5>Arguments:</h5>
4114
4115<p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4116The '<tt>select</tt>' instruction requires an 'i1' value indicating the
4117condition, and two values of the same <a href="#t_firstclass">first class</a>
4118type. If the val1/val2 are vectors, the entire vectors are selected, not
4119individual elements.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4120</p>
4121
4122<h5>Semantics:</h5>
4123
4124<p>
Chris Lattner6704c212008-05-20 20:48:21 +0000[diff] [blame]4125If the i1 condition evaluates is 1, the instruction returns the first
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4126value argument; otherwise, it returns the second value argument.
4127</p>
4128
4129<h5>Example:</h5>
4130
4131<pre>
4132 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4133</pre>
4134</div>
4135
4136
4137<!-- _______________________________________________________________________ -->
4138<div class="doc_subsubsection">
4139 <a name="i_call">'<tt>call</tt>' Instruction</a>
4140</div>
4141
4142<div class="doc_text">
4143
4144<h5>Syntax:</h5>
4145<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4146 &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]4147</pre>
4148
4149<h5>Overview:</h5>
4150
4151<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4152
4153<h5>Arguments:</h5>
4154
4155<p>This instruction requires several arguments:</p>
4156
4157<ol>
4158 <li>
4159 <p>The optional "tail" marker indicates whether the callee function accesses
4160 any allocas or varargs in the caller. If the "tail" marker is present, the
4161 function call is eligible for tail call optimization. Note that calls may
4162 be marked "tail" even if they do not occur before a <a
4163 href="#i_ret"><tt>ret</tt></a> instruction.
4164 </li>
4165 <li>
4166 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4167 convention</a> the call should use. If none is specified, the call defaults
4168 to using C calling conventions.
4169 </li>
4170 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4171 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4172 the type of the return value. Functions that return no value are marked
4173 <tt><a href="#t_void">void</a></tt>.</p>
4174 </li>
4175 <li>
4176 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4177 value being invoked. The argument types must match the types implied by
4178 this signature. This type can be omitted if the function is not varargs
4179 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4180 </li>
4181 <li>
4182 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4183 be invoked. In most cases, this is a direct function invocation, but
4184 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4185 to function value.</p>
4186 </li>
4187 <li>
4188 <p>'<tt>function args</tt>': argument list whose types match the
4189 function signature argument types. All arguments must be of
4190 <a href="#t_firstclass">first class</a> type. If the function signature
4191 indicates the function accepts a variable number of arguments, the extra
4192 arguments can be specified.</p>
4193 </li>
4194</ol>
4195
4196<h5>Semantics:</h5>
4197
4198<p>The '<tt>call</tt>' instruction is used to cause control flow to
4199transfer to a specified function, with its incoming arguments bound to
4200the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4201instruction in the called function, control flow continues with the
4202instruction after the function call, and the return value of the
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4203function is bound to the result argument. If the callee returns multiple
4204values then the return values of the function are only accessible through
4205the '<tt><a href="#i_getresult">getresult</a></tt>' instruction.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4206
4207<h5>Example:</h5>
4208
4209<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4210 %retval = call i32 @test(i32 %argc)
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4211 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4212 %X = tail call i32 @foo() <i>; yields i32</i>
4213 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4214 call void %foo(i8 97 signext)
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4215
4216 %struct.A = type { i32, i8 }
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4217 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
4218 %gr = getresult %struct.A %r, 0 <i>; yields i32</i>
4219 %gr1 = getresult %struct.A %r, 1 <i>; yields i8</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4220</pre>
4221
4222</div>
4223
4224<!-- _______________________________________________________________________ -->
4225<div class="doc_subsubsection">
4226 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4227</div>
4228
4229<div class="doc_text">
4230
4231<h5>Syntax:</h5>
4232
4233<pre>
4234 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
4235</pre>
4236
4237<h5>Overview:</h5>
4238
4239<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4240the "variable argument" area of a function call. It is used to implement the
4241<tt>va_arg</tt> macro in C.</p>
4242
4243<h5>Arguments:</h5>
4244
4245<p>This instruction takes a <tt>va_list*</tt> value and the type of
4246the argument. It returns a value of the specified argument type and
4247increments the <tt>va_list</tt> to point to the next argument. The
4248actual type of <tt>va_list</tt> is target specific.</p>
4249
4250<h5>Semantics:</h5>
4251
4252<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4253type from the specified <tt>va_list</tt> and causes the
4254<tt>va_list</tt> to point to the next argument. For more information,
4255see the variable argument handling <a href="#int_varargs">Intrinsic
4256Functions</a>.</p>
4257
4258<p>It is legal for this instruction to be called in a function which does not
4259take a variable number of arguments, for example, the <tt>vfprintf</tt>
4260function.</p>
4261
4262<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4263href="#intrinsics">intrinsic function</a> because it takes a type as an
4264argument.</p>
4265
4266<h5>Example:</h5>
4267
4268<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4269
4270</div>
4271
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4272<!-- _______________________________________________________________________ -->
4273<div class="doc_subsubsection">
4274 <a name="i_getresult">'<tt>getresult</tt>' Instruction</a>
4275</div>
4276
4277<div class="doc_text">
4278
4279<h5>Syntax:</h5>
4280<pre>
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4281 &lt;resultval&gt; = getresult &lt;type&gt; &lt;retval&gt;, &lt;index&gt;
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4282</pre>
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4283
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4284<h5>Overview:</h5>
4285
4286<p> The '<tt>getresult</tt>' instruction is used to extract individual values
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4287from a '<tt><a href="#i_call">call</a></tt>'
4288or '<tt><a href="#i_invoke">invoke</a></tt>' instruction that returns multiple
4289results.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4290
4291<h5>Arguments:</h5>
4292
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4293<p>The '<tt>getresult</tt>' instruction takes a call or invoke value as its
Chris Lattnerd8dd3522008-04-23 04:06:52 +0000[diff] [blame]4294first argument, or an undef value. The value must have <a
4295href="#t_struct">structure type</a>. The second argument is a constant
4296unsigned index value which must be in range for the number of values returned
4297by the call.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4298
4299<h5>Semantics:</h5>
4300
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4301<p>The '<tt>getresult</tt>' instruction extracts the element identified by
4302'<tt>index</tt>' from the aggregate value.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4303
4304<h5>Example:</h5>
4305
4306<pre>
4307 %struct.A = type { i32, i8 }
4308
4309 %r = call %struct.A @foo()
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4310 %gr = getresult %struct.A %r, 0 <i>; yields i32:%gr</i>
4311 %gr1 = getresult %struct.A %r, 1 <i>; yields i8:%gr1</i>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4312 add i32 %gr, 42
4313 add i8 %gr1, 41
4314</pre>
4315
4316</div>
4317
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4318<!-- *********************************************************************** -->
4319<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4320<!-- *********************************************************************** -->
4321
4322<div class="doc_text">
4323
4324<p>LLVM supports the notion of an "intrinsic function". These functions have
4325well known names and semantics and are required to follow certain restrictions.
4326Overall, these intrinsics represent an extension mechanism for the LLVM
4327language that does not require changing all of the transformations in LLVM when
4328adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4329
4330<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4331prefix is reserved in LLVM for intrinsic names; thus, function names may not
4332begin with this prefix. Intrinsic functions must always be external functions:
4333you cannot define the body of intrinsic functions. Intrinsic functions may
4334only be used in call or invoke instructions: it is illegal to take the address
4335of an intrinsic function. Additionally, because intrinsic functions are part
4336of the LLVM language, it is required if any are added that they be documented
4337here.</p>
4338
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4339<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4340a family of functions that perform the same operation but on different data
4341types. Because LLVM can represent over 8 million different integer types,
4342overloading is used commonly to allow an intrinsic function to operate on any
4343integer type. One or more of the argument types or the result type can be
4344overloaded to accept any integer type. Argument types may also be defined as
4345exactly matching a previous argument's type or the result type. This allows an
4346intrinsic function which accepts multiple arguments, but needs all of them to
4347be of the same type, to only be overloaded with respect to a single argument or
4348the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4349
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4350<p>Overloaded intrinsics will have the names of its overloaded argument types
4351encoded into its function name, each preceded by a period. Only those types
4352which are overloaded result in a name suffix. Arguments whose type is matched
4353against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4354take an integer of any width and returns an integer of exactly the same integer
4355width. This leads to a family of functions such as
4356<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4357Only one type, the return type, is overloaded, and only one type suffix is
4358required. Because the argument's type is matched against the return type, it
4359does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4360
4361<p>To learn how to add an intrinsic function, please see the
4362<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4363</p>
4364
4365</div>
4366
4367<!-- ======================================================================= -->
4368<div class="doc_subsection">
4369 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4370</div>
4371
4372<div class="doc_text">
4373
4374<p>Variable argument support is defined in LLVM with the <a
4375 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4376intrinsic functions. These functions are related to the similarly
4377named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
4378
4379<p>All of these functions operate on arguments that use a
4380target-specific value type "<tt>va_list</tt>". The LLVM assembly
4381language reference manual does not define what this type is, so all
4382transformations should be prepared to handle these functions regardless of
4383the type used.</p>
4384
4385<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4386instruction and the variable argument handling intrinsic functions are
4387used.</p>
4388
4389<div class="doc_code">
4390<pre>
4391define i32 @test(i32 %X, ...) {
4392 ; Initialize variable argument processing
4393 %ap = alloca i8*
4394 %ap2 = bitcast i8** %ap to i8*
4395 call void @llvm.va_start(i8* %ap2)
4396
4397 ; Read a single integer argument
4398 %tmp = va_arg i8** %ap, i32
4399
4400 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4401 %aq = alloca i8*
4402 %aq2 = bitcast i8** %aq to i8*
4403 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4404 call void @llvm.va_end(i8* %aq2)
4405
4406 ; Stop processing of arguments.
4407 call void @llvm.va_end(i8* %ap2)
4408 ret i32 %tmp
4409}
4410
4411declare void @llvm.va_start(i8*)
4412declare void @llvm.va_copy(i8*, i8*)
4413declare void @llvm.va_end(i8*)
4414</pre>
4415</div>
4416
4417</div>
4418
4419<!-- _______________________________________________________________________ -->
4420<div class="doc_subsubsection">
4421 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4422</div>
4423
4424
4425<div class="doc_text">
4426<h5>Syntax:</h5>
4427<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
4428<h5>Overview:</h5>
4429<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
4430<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
4431href="#i_va_arg">va_arg</a></tt>.</p>
4432
4433<h5>Arguments:</h5>
4434
4435<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
4436
4437<h5>Semantics:</h5>
4438
4439<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
4440macro available in C. In a target-dependent way, it initializes the
4441<tt>va_list</tt> element to which the argument points, so that the next call to
4442<tt>va_arg</tt> will produce the first variable argument passed to the function.
4443Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4444last argument of the function as the compiler can figure that out.</p>
4445
4446</div>
4447
4448<!-- _______________________________________________________________________ -->
4449<div class="doc_subsubsection">
4450 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4451</div>
4452
4453<div class="doc_text">
4454<h5>Syntax:</h5>
4455<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
4456<h5>Overview:</h5>
4457
4458<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
4459which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4460or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4461
4462<h5>Arguments:</h5>
4463
4464<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4465
4466<h5>Semantics:</h5>
4467
4468<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4469macro available in C. In a target-dependent way, it destroys the
4470<tt>va_list</tt> element to which the argument points. Calls to <a
4471href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4472<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4473<tt>llvm.va_end</tt>.</p>
4474
4475</div>
4476
4477<!-- _______________________________________________________________________ -->
4478<div class="doc_subsubsection">
4479 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4480</div>
4481
4482<div class="doc_text">
4483
4484<h5>Syntax:</h5>
4485
4486<pre>
4487 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
4488</pre>
4489
4490<h5>Overview:</h5>
4491
4492<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4493from the source argument list to the destination argument list.</p>
4494
4495<h5>Arguments:</h5>
4496
4497<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4498The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4499
4500
4501<h5>Semantics:</h5>
4502
4503<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4504macro available in C. In a target-dependent way, it copies the source
4505<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4506intrinsic is necessary because the <tt><a href="#int_va_start">
4507llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4508example, memory allocation.</p>
4509
4510</div>
4511
4512<!-- ======================================================================= -->
4513<div class="doc_subsection">
4514 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4515</div>
4516
4517<div class="doc_text">
4518
4519<p>
4520LLVM support for <a href="GarbageCollection.html">Accurate Garbage
4521Collection</a> requires the implementation and generation of these intrinsics.
4522These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4523stack</a>, as well as garbage collector implementations that require <a
4524href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4525Front-ends for type-safe garbage collected languages should generate these
4526intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4527href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4528</p>
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]4529
4530<p>The garbage collection intrinsics only operate on objects in the generic
4531 address space (address space zero).</p>
4532
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4533</div>
4534
4535<!-- _______________________________________________________________________ -->
4536<div class="doc_subsubsection">
4537 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4538</div>
4539
4540<div class="doc_text">
4541
4542<h5>Syntax:</h5>
4543
4544<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4545 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4546</pre>
4547
4548<h5>Overview:</h5>
4549
4550<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4551the code generator, and allows some metadata to be associated with it.</p>
4552
4553<h5>Arguments:</h5>
4554
4555<p>The first argument specifies the address of a stack object that contains the
4556root pointer. The second pointer (which must be either a constant or a global
4557value address) contains the meta-data to be associated with the root.</p>
4558
4559<h5>Semantics:</h5>
4560
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]4561<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]4562location. At compile-time, the code generator generates information to allow
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4563the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4564intrinsic may only be used in a function which <a href="#gc">specifies a GC
4565algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4566
4567</div>
4568
4569
4570<!-- _______________________________________________________________________ -->
4571<div class="doc_subsubsection">
4572 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4573</div>
4574
4575<div class="doc_text">
4576
4577<h5>Syntax:</h5>
4578
4579<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4580 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4581</pre>
4582
4583<h5>Overview:</h5>
4584
4585<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4586locations, allowing garbage collector implementations that require read
4587barriers.</p>
4588
4589<h5>Arguments:</h5>
4590
4591<p>The second argument is the address to read from, which should be an address
4592allocated from the garbage collector. The first object is a pointer to the
4593start of the referenced object, if needed by the language runtime (otherwise
4594null).</p>
4595
4596<h5>Semantics:</h5>
4597
4598<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4599instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4600garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4601may only be used in a function which <a href="#gc">specifies a GC
4602algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4603
4604</div>
4605
4606
4607<!-- _______________________________________________________________________ -->
4608<div class="doc_subsubsection">
4609 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4610</div>
4611
4612<div class="doc_text">
4613
4614<h5>Syntax:</h5>
4615
4616<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4617 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4618</pre>
4619
4620<h5>Overview:</h5>
4621
4622<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4623locations, allowing garbage collector implementations that require write
4624barriers (such as generational or reference counting collectors).</p>
4625
4626<h5>Arguments:</h5>
4627
4628<p>The first argument is the reference to store, the second is the start of the
4629object to store it to, and the third is the address of the field of Obj to
4630store to. If the runtime does not require a pointer to the object, Obj may be
4631null.</p>
4632
4633<h5>Semantics:</h5>
4634
4635<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4636instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4637garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4638may only be used in a function which <a href="#gc">specifies a GC
4639algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4640
4641</div>
4642
4643
4644
4645<!-- ======================================================================= -->
4646<div class="doc_subsection">
4647 <a name="int_codegen">Code Generator Intrinsics</a>
4648</div>
4649
4650<div class="doc_text">
4651<p>
4652These intrinsics are provided by LLVM to expose special features that may only
4653be implemented with code generator support.
4654</p>
4655
4656</div>
4657
4658<!-- _______________________________________________________________________ -->
4659<div class="doc_subsubsection">
4660 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4661</div>
4662
4663<div class="doc_text">
4664
4665<h5>Syntax:</h5>
4666<pre>
4667 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
4668</pre>
4669
4670<h5>Overview:</h5>
4671
4672<p>
4673The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4674target-specific value indicating the return address of the current function
4675or one of its callers.
4676</p>
4677
4678<h5>Arguments:</h5>
4679
4680<p>
4681The argument to this intrinsic indicates which function to return the address
4682for. Zero indicates the calling function, one indicates its caller, etc. The
4683argument is <b>required</b> to be a constant integer value.
4684</p>
4685
4686<h5>Semantics:</h5>
4687
4688<p>
4689The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4690the return address of the specified call frame, or zero if it cannot be
4691identified. The value returned by this intrinsic is likely to be incorrect or 0
4692for arguments other than zero, so it should only be used for debugging purposes.
4693</p>
4694
4695<p>
4696Note that calling this intrinsic does not prevent function inlining or other
4697aggressive transformations, so the value returned may not be that of the obvious
4698source-language caller.
4699</p>
4700</div>
4701
4702
4703<!-- _______________________________________________________________________ -->
4704<div class="doc_subsubsection">
4705 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4706</div>
4707
4708<div class="doc_text">
4709
4710<h5>Syntax:</h5>
4711<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4712 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4713</pre>
4714
4715<h5>Overview:</h5>
4716
4717<p>
4718The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4719target-specific frame pointer value for the specified stack frame.
4720</p>
4721
4722<h5>Arguments:</h5>
4723
4724<p>
4725The argument to this intrinsic indicates which function to return the frame
4726pointer for. Zero indicates the calling function, one indicates its caller,
4727etc. The argument is <b>required</b> to be a constant integer value.
4728</p>
4729
4730<h5>Semantics:</h5>
4731
4732<p>
4733The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4734the frame address of the specified call frame, or zero if it cannot be
4735identified. The value returned by this intrinsic is likely to be incorrect or 0
4736for arguments other than zero, so it should only be used for debugging purposes.
4737</p>
4738
4739<p>
4740Note that calling this intrinsic does not prevent function inlining or other
4741aggressive transformations, so the value returned may not be that of the obvious
4742source-language caller.
4743</p>
4744</div>
4745
4746<!-- _______________________________________________________________________ -->
4747<div class="doc_subsubsection">
4748 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4749</div>
4750
4751<div class="doc_text">
4752
4753<h5>Syntax:</h5>
4754<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4755 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4756</pre>
4757
4758<h5>Overview:</h5>
4759
4760<p>
4761The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4762the function stack, for use with <a href="#int_stackrestore">
4763<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4764features like scoped automatic variable sized arrays in C99.
4765</p>
4766
4767<h5>Semantics:</h5>
4768
4769<p>
4770This intrinsic returns a opaque pointer value that can be passed to <a
4771href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4772<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4773<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4774state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4775practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4776that were allocated after the <tt>llvm.stacksave</tt> was executed.
4777</p>
4778
4779</div>
4780
4781<!-- _______________________________________________________________________ -->
4782<div class="doc_subsubsection">
4783 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4784</div>
4785
4786<div class="doc_text">
4787
4788<h5>Syntax:</h5>
4789<pre>
4790 declare void @llvm.stackrestore(i8 * %ptr)
4791</pre>
4792
4793<h5>Overview:</h5>
4794
4795<p>
4796The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4797the function stack to the state it was in when the corresponding <a
4798href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4799useful for implementing language features like scoped automatic variable sized
4800arrays in C99.
4801</p>
4802
4803<h5>Semantics:</h5>
4804
4805<p>
4806See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4807</p>
4808
4809</div>
4810
4811
4812<!-- _______________________________________________________________________ -->
4813<div class="doc_subsubsection">
4814 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4815</div>
4816
4817<div class="doc_text">
4818
4819<h5>Syntax:</h5>
4820<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4821 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]4822</pre>
4823
4824<h5>Overview:</h5>
4825
4826
4827<p>
4828The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4829a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4830no
4831effect on the behavior of the program but can change its performance
4832characteristics.
4833</p>
4834
4835<h5>Arguments:</h5>
4836
4837<p>
4838<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4839determining if the fetch should be for a read (0) or write (1), and
4840<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4841locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4842<tt>locality</tt> arguments must be constant integers.
4843</p>
4844
4845<h5>Semantics:</h5>
4846
4847<p>
4848This intrinsic does not modify the behavior of the program. In particular,
4849prefetches cannot trap and do not produce a value. On targets that support this
4850intrinsic, the prefetch can provide hints to the processor cache for better
4851performance.
4852</p>
4853
4854</div>
4855
4856<!-- _______________________________________________________________________ -->
4857<div class="doc_subsubsection">
4858 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4859</div>
4860
4861<div class="doc_text">
4862
4863<h5>Syntax:</h5>
4864<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4865 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4866</pre>
4867
4868<h5>Overview:</h5>
4869
4870
4871<p>
4872The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4873(PC) in a region of
4874code to simulators and other tools. The method is target specific, but it is
4875expected that the marker will use exported symbols to transmit the PC of the marker.
4876The marker makes no guarantees that it will remain with any specific instruction
4877after optimizations. It is possible that the presence of a marker will inhibit
4878optimizations. The intended use is to be inserted after optimizations to allow
4879correlations of simulation runs.
4880</p>
4881
4882<h5>Arguments:</h5>
4883
4884<p>
4885<tt>id</tt> is a numerical id identifying the marker.
4886</p>
4887
4888<h5>Semantics:</h5>
4889
4890<p>
4891This intrinsic does not modify the behavior of the program. Backends that do not
4892support this intrinisic may ignore it.
4893</p>
4894
4895</div>
4896
4897<!-- _______________________________________________________________________ -->
4898<div class="doc_subsubsection">
4899 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4900</div>
4901
4902<div class="doc_text">
4903
4904<h5>Syntax:</h5>
4905<pre>
4906 declare i64 @llvm.readcyclecounter( )
4907</pre>
4908
4909<h5>Overview:</h5>
4910
4911
4912<p>
4913The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4914counter register (or similar low latency, high accuracy clocks) on those targets
4915that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4916As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4917should only be used for small timings.
4918</p>
4919
4920<h5>Semantics:</h5>
4921
4922<p>
4923When directly supported, reading the cycle counter should not modify any memory.
4924Implementations are allowed to either return a application specific value or a
4925system wide value. On backends without support, this is lowered to a constant 0.
4926</p>
4927
4928</div>
4929
4930<!-- ======================================================================= -->
4931<div class="doc_subsection">
4932 <a name="int_libc">Standard C Library Intrinsics</a>
4933</div>
4934
4935<div class="doc_text">
4936<p>
4937LLVM provides intrinsics for a few important standard C library functions.
4938These intrinsics allow source-language front-ends to pass information about the
4939alignment of the pointer arguments to the code generator, providing opportunity
4940for more efficient code generation.
4941</p>
4942
4943</div>
4944
4945<!-- _______________________________________________________________________ -->
4946<div class="doc_subsubsection">
4947 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4948</div>
4949
4950<div class="doc_text">
4951
4952<h5>Syntax:</h5>
4953<pre>
4954 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4955 i32 &lt;len&gt;, i32 &lt;align&gt;)
4956 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4957 i64 &lt;len&gt;, i32 &lt;align&gt;)
4958</pre>
4959
4960<h5>Overview:</h5>
4961
4962<p>
4963The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4964location to the destination location.
4965</p>
4966
4967<p>
4968Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4969intrinsics do not return a value, and takes an extra alignment argument.
4970</p>
4971
4972<h5>Arguments:</h5>
4973
4974<p>
4975The first argument is a pointer to the destination, the second is a pointer to
4976the source. The third argument is an integer argument
4977specifying the number of bytes to copy, and the fourth argument is the alignment
4978of the source and destination locations.
4979</p>
4980
4981<p>
4982If the call to this intrinisic has an alignment value that is not 0 or 1, then
4983the caller guarantees that both the source and destination pointers are aligned
4984to that boundary.
4985</p>
4986
4987<h5>Semantics:</h5>
4988
4989<p>
4990The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4991location to the destination location, which are not allowed to overlap. It
4992copies "len" bytes of memory over. If the argument is known to be aligned to
4993some boundary, this can be specified as the fourth argument, otherwise it should
4994be set to 0 or 1.
4995</p>
4996</div>
4997
4998
4999<!-- _______________________________________________________________________ -->
5000<div class="doc_subsubsection">
5001 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5002</div>
5003
5004<div class="doc_text">
5005
5006<h5>Syntax:</h5>
5007<pre>
5008 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5009 i32 &lt;len&gt;, i32 &lt;align&gt;)
5010 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5011 i64 &lt;len&gt;, i32 &lt;align&gt;)
5012</pre>
5013
5014<h5>Overview:</h5>
5015
5016<p>
5017The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5018location to the destination location. It is similar to the
Chris Lattnerdba16ea2008-01-06 19:51:52 +0000[diff] [blame]5019'<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5020</p>
5021
5022<p>
5023Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5024intrinsics do not return a value, and takes an extra alignment argument.
5025</p>
5026
5027<h5>Arguments:</h5>
5028
5029<p>
5030The first argument is a pointer to the destination, the second is a pointer to
5031the source. The third argument is an integer argument
5032specifying the number of bytes to copy, and the fourth argument is the alignment
5033of the source and destination locations.
5034</p>
5035
5036<p>
5037If the call to this intrinisic has an alignment value that is not 0 or 1, then
5038the caller guarantees that the source and destination pointers are aligned to
5039that boundary.
5040</p>
5041
5042<h5>Semantics:</h5>
5043
5044<p>
5045The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5046location to the destination location, which may overlap. It
5047copies "len" bytes of memory over. If the argument is known to be aligned to
5048some boundary, this can be specified as the fourth argument, otherwise it should
5049be set to 0 or 1.
5050</p>
5051</div>
5052
5053
5054<!-- _______________________________________________________________________ -->
5055<div class="doc_subsubsection">
5056 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5057</div>
5058
5059<div class="doc_text">
5060
5061<h5>Syntax:</h5>
5062<pre>
5063 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5064 i32 &lt;len&gt;, i32 &lt;align&gt;)
5065 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5066 i64 &lt;len&gt;, i32 &lt;align&gt;)
5067</pre>
5068
5069<h5>Overview:</h5>
5070
5071<p>
5072The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5073byte value.
5074</p>
5075
5076<p>
5077Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5078does not return a value, and takes an extra alignment argument.
5079</p>
5080
5081<h5>Arguments:</h5>
5082
5083<p>
5084The first argument is a pointer to the destination to fill, the second is the
5085byte value to fill it with, the third argument is an integer
5086argument specifying the number of bytes to fill, and the fourth argument is the
5087known alignment of destination location.
5088</p>
5089
5090<p>
5091If the call to this intrinisic has an alignment value that is not 0 or 1, then
5092the caller guarantees that the destination pointer is aligned to that boundary.
5093</p>
5094
5095<h5>Semantics:</h5>
5096
5097<p>
5098The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5099the
5100destination location. If the argument is known to be aligned to some boundary,
5101this can be specified as the fourth argument, otherwise it should be set to 0 or
51021.
5103</p>
5104</div>
5105
5106
5107<!-- _______________________________________________________________________ -->
5108<div class="doc_subsubsection">
5109 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5110</div>
5111
5112<div class="doc_text">
5113
5114<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5115<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]5116floating point or vector of floating point type. Not all targets support all
5117types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5118<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5119 declare float @llvm.sqrt.f32(float %Val)
5120 declare double @llvm.sqrt.f64(double %Val)
5121 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5122 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5123 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5124</pre>
5125
5126<h5>Overview:</h5>
5127
5128<p>
5129The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5130returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5131<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
Chris Lattnerf914a002008-01-29 07:00:44 +0000[diff] [blame]5132negative numbers other than -0.0 (which allows for better optimization, because
5133there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5134defined to return -0.0 like IEEE sqrt.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5135</p>
5136
5137<h5>Arguments:</h5>
5138
5139<p>
5140The argument and return value are floating point numbers of the same type.
5141</p>
5142
5143<h5>Semantics:</h5>
5144
5145<p>
5146This function returns the sqrt of the specified operand if it is a nonnegative
5147floating point number.
5148</p>
5149</div>
5150
5151<!-- _______________________________________________________________________ -->
5152<div class="doc_subsubsection">
5153 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5154</div>
5155
5156<div class="doc_text">
5157
5158<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5159<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]5160floating point or vector of floating point type. Not all targets support all
5161types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5162<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5163 declare float @llvm.powi.f32(float %Val, i32 %power)
5164 declare double @llvm.powi.f64(double %Val, i32 %power)
5165 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5166 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5167 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5168</pre>
5169
5170<h5>Overview:</h5>
5171
5172<p>
5173The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5174specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5175multiplications is not defined. When a vector of floating point type is
5176used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5177</p>
5178
5179<h5>Arguments:</h5>
5180
5181<p>
5182The second argument is an integer power, and the first is a value to raise to
5183that power.
5184</p>
5185
5186<h5>Semantics:</h5>
5187
5188<p>
5189This function returns the first value raised to the second power with an
5190unspecified sequence of rounding operations.</p>
5191</div>
5192
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5193<!-- _______________________________________________________________________ -->
5194<div class="doc_subsubsection">
5195 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5196</div>
5197
5198<div class="doc_text">
5199
5200<h5>Syntax:</h5>
5201<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5202floating point or vector of floating point type. Not all targets support all
5203types however.
5204<pre>
5205 declare float @llvm.sin.f32(float %Val)
5206 declare double @llvm.sin.f64(double %Val)
5207 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5208 declare fp128 @llvm.sin.f128(fp128 %Val)
5209 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5210</pre>
5211
5212<h5>Overview:</h5>
5213
5214<p>
5215The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
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 sine of the specified operand, returning the
5228same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5229conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5230</div>
5231
5232<!-- _______________________________________________________________________ -->
5233<div class="doc_subsubsection">
5234 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5235</div>
5236
5237<div class="doc_text">
5238
5239<h5>Syntax:</h5>
5240<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5241floating point or vector of floating point type. Not all targets support all
5242types however.
5243<pre>
5244 declare float @llvm.cos.f32(float %Val)
5245 declare double @llvm.cos.f64(double %Val)
5246 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5247 declare fp128 @llvm.cos.f128(fp128 %Val)
5248 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5249</pre>
5250
5251<h5>Overview:</h5>
5252
5253<p>
5254The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5255</p>
5256
5257<h5>Arguments:</h5>
5258
5259<p>
5260The argument and return value are floating point numbers of the same type.
5261</p>
5262
5263<h5>Semantics:</h5>
5264
5265<p>
5266This function returns the cosine of the specified operand, returning the
5267same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5268conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5269</div>
5270
5271<!-- _______________________________________________________________________ -->
5272<div class="doc_subsubsection">
5273 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5274</div>
5275
5276<div class="doc_text">
5277
5278<h5>Syntax:</h5>
5279<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5280floating point or vector of floating point type. Not all targets support all
5281types however.
5282<pre>
5283 declare float @llvm.pow.f32(float %Val, float %Power)
5284 declare double @llvm.pow.f64(double %Val, double %Power)
5285 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5286 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5287 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5288</pre>
5289
5290<h5>Overview:</h5>
5291
5292<p>
5293The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5294specified (positive or negative) power.
5295</p>
5296
5297<h5>Arguments:</h5>
5298
5299<p>
5300The second argument is a floating point power, and the first is a value to
5301raise to that power.
5302</p>
5303
5304<h5>Semantics:</h5>
5305
5306<p>
5307This function returns the first value raised to the second power,
5308returning the
5309same values as the libm <tt>pow</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
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5313
5314<!-- ======================================================================= -->
5315<div class="doc_subsection">
5316 <a name="int_manip">Bit Manipulation Intrinsics</a>
5317</div>
5318
5319<div class="doc_text">
5320<p>
5321LLVM provides intrinsics for a few important bit manipulation operations.
5322These allow efficient code generation for some algorithms.
5323</p>
5324
5325</div>
5326
5327<!-- _______________________________________________________________________ -->
5328<div class="doc_subsubsection">
5329 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5330</div>
5331
5332<div class="doc_text">
5333
5334<h5>Syntax:</h5>
5335<p>This is an overloaded intrinsic function. You can use bswap on any integer
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5336type that is an even number of bytes (i.e. BitWidth % 16 == 0).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5337<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5338 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
5339 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
5340 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5341</pre>
5342
5343<h5>Overview:</h5>
5344
5345<p>
5346The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5347values with an even number of bytes (positive multiple of 16 bits). These are
5348useful for performing operations on data that is not in the target's native
5349byte order.
5350</p>
5351
5352<h5>Semantics:</h5>
5353
5354<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5355The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5356and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5357intrinsic returns an i32 value that has the four bytes of the input i32
5358swapped, 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]5359i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5360<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5361additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5362</p>
5363
5364</div>
5365
5366<!-- _______________________________________________________________________ -->
5367<div class="doc_subsubsection">
5368 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5369</div>
5370
5371<div class="doc_text">
5372
5373<h5>Syntax:</h5>
5374<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
5375width. Not all targets support all bit widths however.
5376<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5377 declare i8 @llvm.ctpop.i8 (i8 &lt;src&gt;)
5378 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5379 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5380 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
5381 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5382</pre>
5383
5384<h5>Overview:</h5>
5385
5386<p>
5387The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5388value.
5389</p>
5390
5391<h5>Arguments:</h5>
5392
5393<p>
5394The only argument is the value to be counted. The argument may be of any
5395integer type. The return type must match the argument type.
5396</p>
5397
5398<h5>Semantics:</h5>
5399
5400<p>
5401The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5402</p>
5403</div>
5404
5405<!-- _______________________________________________________________________ -->
5406<div class="doc_subsubsection">
5407 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5408</div>
5409
5410<div class="doc_text">
5411
5412<h5>Syntax:</h5>
5413<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5414integer bit width. Not all targets support all bit widths however.
5415<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5416 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
5417 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5418 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5419 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
5420 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5421</pre>
5422
5423<h5>Overview:</h5>
5424
5425<p>
5426The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5427leading zeros in a variable.
5428</p>
5429
5430<h5>Arguments:</h5>
5431
5432<p>
5433The only argument is the value to be counted. The argument may be of any
5434integer type. The return type must match the argument type.
5435</p>
5436
5437<h5>Semantics:</h5>
5438
5439<p>
5440The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5441in a variable. If the src == 0 then the result is the size in bits of the type
5442of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5443</p>
5444</div>
5445
5446
5447
5448<!-- _______________________________________________________________________ -->
5449<div class="doc_subsubsection">
5450 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5451</div>
5452
5453<div class="doc_text">
5454
5455<h5>Syntax:</h5>
5456<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5457integer bit width. Not all targets support all bit widths however.
5458<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5459 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
5460 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5461 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5462 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
5463 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5464</pre>
5465
5466<h5>Overview:</h5>
5467
5468<p>
5469The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5470trailing zeros.
5471</p>
5472
5473<h5>Arguments:</h5>
5474
5475<p>
5476The only argument is the value to be counted. The argument may be of any
5477integer type. The return type must match the argument type.
5478</p>
5479
5480<h5>Semantics:</h5>
5481
5482<p>
5483The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5484in a variable. If the src == 0 then the result is the size in bits of the type
5485of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5486</p>
5487</div>
5488
5489<!-- _______________________________________________________________________ -->
5490<div class="doc_subsubsection">
5491 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5492</div>
5493
5494<div class="doc_text">
5495
5496<h5>Syntax:</h5>
5497<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
5498on any integer bit width.
5499<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5500 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5501 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5502</pre>
5503
5504<h5>Overview:</h5>
5505<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5506range of bits from an integer value and returns them in the same bit width as
5507the original value.</p>
5508
5509<h5>Arguments:</h5>
5510<p>The first argument, <tt>%val</tt> and the result may be integer types of
5511any bit width but they must have the same bit width. The second and third
5512arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5513
5514<h5>Semantics:</h5>
5515<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5516of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5517<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5518operates in forward mode.</p>
5519<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5520right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5521only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5522<ol>
5523 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5524 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5525 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5526 to determine the number of bits to retain.</li>
5527 <li>A mask of the retained bits is created by shifting a -1 value.</li>
5528 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
5529</ol>
5530<p>In reverse mode, a similar computation is made except that the bits are
5531returned in the reverse order. So, for example, if <tt>X</tt> has the value
5532<tt>i16 0x0ACF (101011001111)</tt> and we apply
5533<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5534<tt>i16 0x0026 (000000100110)</tt>.</p>
5535</div>
5536
5537<div class="doc_subsubsection">
5538 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5539</div>
5540
5541<div class="doc_text">
5542
5543<h5>Syntax:</h5>
5544<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
5545on any integer bit width.
5546<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5547 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5548 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]5549</pre>
5550
5551<h5>Overview:</h5>
5552<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5553of bits in an integer value with another integer value. It returns the integer
5554with the replaced bits.</p>
5555
5556<h5>Arguments:</h5>
5557<p>The first argument, <tt>%val</tt> and the result may be integer types of
5558any bit width but they must have the same bit width. <tt>%val</tt> is the value
5559whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5560integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5561type since they specify only a bit index.</p>
5562
5563<h5>Semantics:</h5>
5564<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5565of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5566<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5567operates in forward mode.</p>
5568<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5569truncating it down to the size of the replacement area or zero extending it
5570up to that size.</p>
5571<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5572are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5573in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5574to the <tt>%hi</tt>th bit.
5575<p>In reverse mode, a similar computation is made except that the bits are
5576reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5577<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
5578<h5>Examples:</h5>
5579<pre>
5580 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
5581 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
5582 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
5583 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
5584 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
5585</pre>
5586</div>
5587
5588<!-- ======================================================================= -->
5589<div class="doc_subsection">
5590 <a name="int_debugger">Debugger Intrinsics</a>
5591</div>
5592
5593<div class="doc_text">
5594<p>
5595The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5596are described in the <a
5597href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5598Debugging</a> document.
5599</p>
5600</div>
5601
5602
5603<!-- ======================================================================= -->
5604<div class="doc_subsection">
5605 <a name="int_eh">Exception Handling Intrinsics</a>
5606</div>
5607
5608<div class="doc_text">
5609<p> The LLVM exception handling intrinsics (which all start with
5610<tt>llvm.eh.</tt> prefix), are described in the <a
5611href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5612Handling</a> document. </p>
5613</div>
5614
5615<!-- ======================================================================= -->
5616<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5617 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5618</div>
5619
5620<div class="doc_text">
5621<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5622 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5623 the <tt>nest</tt> attribute, from a function. The result is a callable
5624 function pointer lacking the nest parameter - the caller does not need
5625 to provide a value for it. Instead, the value to use is stored in
5626 advance in a "trampoline", a block of memory usually allocated
5627 on the stack, which also contains code to splice the nest value into the
5628 argument list. This is used to implement the GCC nested function address
5629 extension.
5630</p>
5631<p>
5632 For example, if the function is
5633 <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]5634 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]5635<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5636 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5637 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5638 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5639 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5640</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]5641 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5642 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5643</div>
5644
5645<!-- _______________________________________________________________________ -->
5646<div class="doc_subsubsection">
5647 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5648</div>
5649<div class="doc_text">
5650<h5>Syntax:</h5>
5651<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5652declare 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]5653</pre>
5654<h5>Overview:</h5>
5655<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5656 This fills the memory pointed to by <tt>tramp</tt> with code
5657 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5658</p>
5659<h5>Arguments:</h5>
5660<p>
5661 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5662 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5663 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]5664 intrinsic. Note that the size and the alignment are target-specific - LLVM
5665 currently provides no portable way of determining them, so a front-end that
5666 generates this intrinsic needs to have some target-specific knowledge.
5667 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]5668</p>
5669<h5>Semantics:</h5>
5670<p>
5671 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5672 dependent code, turning it into a function. A pointer to this function is
5673 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5674 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5675 before being called. The new function's signature is the same as that of
5676 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5677 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5678 of pointer type. Calling the new function is equivalent to calling
5679 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5680 missing <tt>nest</tt> argument. If, after calling
5681 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5682 modified, then the effect of any later call to the returned function pointer is
5683 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5684</p>
5685</div>
5686
5687<!-- ======================================================================= -->
5688<div class="doc_subsection">
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]5689 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
5690</div>
5691
5692<div class="doc_text">
5693<p>
5694 These intrinsic functions expand the "universal IR" of LLVM to represent
5695 hardware constructs for atomic operations and memory synchronization. This
5696 provides an interface to the hardware, not an interface to the programmer. It
5697 is aimed at a low enough level to allow any programming models or APIs which
5698 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
5699 hardware behavior. Just as hardware provides a "universal IR" for source
5700 languages, it also provides a starting point for developing a "universal"
5701 atomic operation and synchronization IR.
5702</p>
5703<p>
5704 These do <em>not</em> form an API such as high-level threading libraries,
5705 software transaction memory systems, atomic primitives, and intrinsic
5706 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
5707 application libraries. The hardware interface provided by LLVM should allow
5708 a clean implementation of all of these APIs and parallel programming models.
5709 No one model or paradigm should be selected above others unless the hardware
5710 itself ubiquitously does so.
5711
5712</p>
5713</div>
5714
5715<!-- _______________________________________________________________________ -->
5716<div class="doc_subsubsection">
5717 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
5718</div>
5719<div class="doc_text">
5720<h5>Syntax:</h5>
5721<pre>
5722declare void @llvm.memory.barrier( i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt;,
5723i1 &lt;device&gt; )
5724
5725</pre>
5726<h5>Overview:</h5>
5727<p>
5728 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
5729 specific pairs of memory access types.
5730</p>
5731<h5>Arguments:</h5>
5732<p>
5733 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
5734 The first four arguments enables a specific barrier as listed below. The fith
5735 argument specifies that the barrier applies to io or device or uncached memory.
5736
5737</p>
5738 <ul>
5739 <li><tt>ll</tt>: load-load barrier</li>
5740 <li><tt>ls</tt>: load-store barrier</li>
5741 <li><tt>sl</tt>: store-load barrier</li>
5742 <li><tt>ss</tt>: store-store barrier</li>
5743 <li><tt>device</tt>: barrier applies to device and uncached memory also.
5744 </ul>
5745<h5>Semantics:</h5>
5746<p>
5747 This intrinsic causes the system to enforce some ordering constraints upon
5748 the loads and stores of the program. This barrier does not indicate
5749 <em>when</em> any events will occur, it only enforces an <em>order</em> in
5750 which they occur. For any of the specified pairs of load and store operations
5751 (f.ex. load-load, or store-load), all of the first operations preceding the
5752 barrier will complete before any of the second operations succeeding the
5753 barrier begin. Specifically the semantics for each pairing is as follows:
5754</p>
5755 <ul>
5756 <li><tt>ll</tt>: All loads before the barrier must complete before any load
5757 after the barrier begins.</li>
5758
5759 <li><tt>ls</tt>: All loads before the barrier must complete before any
5760 store after the barrier begins.</li>
5761 <li><tt>ss</tt>: All stores before the barrier must complete before any
5762 store after the barrier begins.</li>
5763 <li><tt>sl</tt>: All stores before the barrier must complete before any
5764 load after the barrier begins.</li>
5765 </ul>
5766<p>
5767 These semantics are applied with a logical "and" behavior when more than one
5768 is enabled in a single memory barrier intrinsic.
5769</p>
5770<p>
5771 Backends may implement stronger barriers than those requested when they do not
5772 support as fine grained a barrier as requested. Some architectures do not
5773 need all types of barriers and on such architectures, these become noops.
5774</p>
5775<h5>Example:</h5>
5776<pre>
5777%ptr = malloc i32
5778 store i32 4, %ptr
5779
5780%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
5781 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
5782 <i>; guarantee the above finishes</i>
5783 store i32 8, %ptr <i>; before this begins</i>
5784</pre>
5785</div>
5786
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5787<!-- _______________________________________________________________________ -->
5788<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]5789 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5790</div>
5791<div class="doc_text">
5792<h5>Syntax:</h5>
5793<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]5794 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on any
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5795 integer bit width. Not all targets support all bit widths however.</p>
5796
5797<pre>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]5798declare i8 @llvm.atomic.cmp.swap.i8( i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt; )
5799declare i16 @llvm.atomic.cmp.swap.i16( i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt; )
5800declare i32 @llvm.atomic.cmp.swap.i32( i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt; )
5801declare i64 @llvm.atomic.cmp.swap.i64( i64* &lt;ptr&gt;, i64 &lt;cmp&gt;, i64 &lt;val&gt; )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5802
5803</pre>
5804<h5>Overview:</h5>
5805<p>
5806 This loads a value in memory and compares it to a given value. If they are
5807 equal, it stores a new value into the memory.
5808</p>
5809<h5>Arguments:</h5>
5810<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]5811 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5812 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
5813 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
5814 this integer type. While any bit width integer may be used, targets may only
5815 lower representations they support in hardware.
5816
5817</p>
5818<h5>Semantics:</h5>
5819<p>
5820 This entire intrinsic must be executed atomically. It first loads the value
5821 in memory pointed to by <tt>ptr</tt> and compares it with the value
5822 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
5823 loaded value is yielded in all cases. This provides the equivalent of an
5824 atomic compare-and-swap operation within the SSA framework.
5825</p>
5826<h5>Examples:</h5>
5827
5828<pre>
5829%ptr = malloc i32
5830 store i32 4, %ptr
5831
5832%val1 = add i32 4, 4
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]5833%result1 = call i32 @llvm.atomic.cmp.swap.i32( i32* %ptr, i32 4, %val1 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5834 <i>; yields {i32}:result1 = 4</i>
5835%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5836%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5837
5838%val2 = add i32 1, 1
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]5839%result2 = call i32 @llvm.atomic.cmp.swap.i32( i32* %ptr, i32 5, %val2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5840 <i>; yields {i32}:result2 = 8</i>
5841%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
5842
5843%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
5844</pre>
5845</div>
5846
5847<!-- _______________________________________________________________________ -->
5848<div class="doc_subsubsection">
5849 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
5850</div>
5851<div class="doc_text">
5852<h5>Syntax:</h5>
5853
5854<p>
5855 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
5856 integer bit width. Not all targets support all bit widths however.</p>
5857<pre>
5858declare i8 @llvm.atomic.swap.i8( i8* &lt;ptr&gt;, i8 &lt;val&gt; )
5859declare i16 @llvm.atomic.swap.i16( i16* &lt;ptr&gt;, i16 &lt;val&gt; )
5860declare i32 @llvm.atomic.swap.i32( i32* &lt;ptr&gt;, i32 &lt;val&gt; )
5861declare i64 @llvm.atomic.swap.i64( i64* &lt;ptr&gt;, i64 &lt;val&gt; )
5862
5863</pre>
5864<h5>Overview:</h5>
5865<p>
5866 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
5867 the value from memory. It then stores the value in <tt>val</tt> in the memory
5868 at <tt>ptr</tt>.
5869</p>
5870<h5>Arguments:</h5>
5871
5872<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]5873 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5874 <tt>val</tt> argument and the result must be integers of the same bit width.
5875 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
5876 integer type. The targets may only lower integer representations they
5877 support.
5878</p>
5879<h5>Semantics:</h5>
5880<p>
5881 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
5882 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
5883 equivalent of an atomic swap operation within the SSA framework.
5884
5885</p>
5886<h5>Examples:</h5>
5887<pre>
5888%ptr = malloc i32
5889 store i32 4, %ptr
5890
5891%val1 = add i32 4, 4
5892%result1 = call i32 @llvm.atomic.swap.i32( i32* %ptr, i32 %val1 )
5893 <i>; yields {i32}:result1 = 4</i>
5894%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5895%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5896
5897%val2 = add i32 1, 1
5898%result2 = call i32 @llvm.atomic.swap.i32( i32* %ptr, i32 %val2 )
5899 <i>; yields {i32}:result2 = 8</i>
5900
5901%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
5902%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
5903</pre>
5904</div>
5905
5906<!-- _______________________________________________________________________ -->
5907<div class="doc_subsubsection">
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]5908 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5909
5910</div>
5911<div class="doc_text">
5912<h5>Syntax:</h5>
5913<p>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]5914 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]5915 integer bit width. Not all targets support all bit widths however.</p>
5916<pre>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]5917declare i8 @llvm.atomic.load.add.i8.( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
5918declare i16 @llvm.atomic.load.add.i16.( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
5919declare i32 @llvm.atomic.load.add.i32.( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
5920declare i64 @llvm.atomic.load.add.i64.( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5921
5922</pre>
5923<h5>Overview:</h5>
5924<p>
5925 This intrinsic adds <tt>delta</tt> to the value stored in memory at
5926 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
5927</p>
5928<h5>Arguments:</h5>
5929<p>
5930
5931 The intrinsic takes two arguments, the first a pointer to an integer value
5932 and the second an integer value. The result is also an integer value. These
5933 integer types can have any bit width, but they must all have the same bit
5934 width. The targets may only lower integer representations they support.
5935</p>
5936<h5>Semantics:</h5>
5937<p>
5938 This intrinsic does a series of operations atomically. It first loads the
5939 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
5940 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
5941</p>
5942
5943<h5>Examples:</h5>
5944<pre>
5945%ptr = malloc i32
5946 store i32 4, %ptr
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]5947%result1 = call i32 @llvm.atomic.load.add.i32( i32* %ptr, i32 4 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5948 <i>; yields {i32}:result1 = 4</i>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]5949%result2 = call i32 @llvm.atomic.load.add.i32( i32* %ptr, i32 2 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5950 <i>; yields {i32}:result2 = 8</i>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]5951%result3 = call i32 @llvm.atomic.load.add.i32( i32* %ptr, i32 5 )
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5952 <i>; yields {i32}:result3 = 10</i>
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]5953%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5954</pre>
5955</div>
5956
Mon P Wang6bde9ec2008-06-25 08:15:39 +0000[diff] [blame^]5957<!-- _______________________________________________________________________ -->
5958<div class="doc_subsubsection">
5959 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
5960
5961</div>
5962<div class="doc_text">
5963<h5>Syntax:</h5>
5964<p>
5965 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
5966 any integer bit width. Not all targets support all bit widths however.</p>
5967<pre>
5968declare i8 @llvm.atomic.load.sub.i8.( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
5969declare i16 @llvm.atomic.load.sub.i16.( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
5970declare i32 @llvm.atomic.load.sub.i32.( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
5971declare i64 @llvm.atomic.load.sub.i64.( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
5972
5973</pre>
5974<h5>Overview:</h5>
5975<p>
5976 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
5977 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
5978</p>
5979<h5>Arguments:</h5>
5980<p>
5981
5982 The intrinsic takes two arguments, the first a pointer to an integer value
5983 and the second an integer value. The result is also an integer value. These
5984 integer types can have any bit width, but they must all have the same bit
5985 width. The targets may only lower integer representations they support.
5986</p>
5987<h5>Semantics:</h5>
5988<p>
5989 This intrinsic does a series of operations atomically. It first loads the
5990 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
5991 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
5992</p>
5993
5994<h5>Examples:</h5>
5995<pre>
5996%ptr = malloc i32
5997 store i32 8, %ptr
5998%result1 = call i32 @llvm.atomic.load.sub.i32( i32* %ptr, i32 4 )
5999 <i>; yields {i32}:result1 = 8</i>
6000%result2 = call i32 @llvm.atomic.load.sub.i32( i32* %ptr, i32 2 )
6001 <i>; yields {i32}:result2 = 4</i>
6002%result3 = call i32 @llvm.atomic.load.sub.i32( i32* %ptr, i32 5 )
6003 <i>; yields {i32}:result3 = 2</i>
6004%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6005</pre>
6006</div>
6007
6008<!-- _______________________________________________________________________ -->
6009<div class="doc_subsubsection">
6010 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6011 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6012 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6013 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6014
6015</div>
6016<div class="doc_text">
6017<h5>Syntax:</h5>
6018<p>
6019 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6020 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
6021 <tt>llvm.atomic.load_xor</tt> on any integer bit width. Not all targets
6022 support all bit widths however.</p>
6023<pre>
6024declare i8 @llvm.atomic.load.and.i8.( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6025declare i16 @llvm.atomic.load.and.i16.( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6026declare i32 @llvm.atomic.load.and.i32.( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6027declare i64 @llvm.atomic.load.and.i64.( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
6028
6029</pre>
6030
6031<pre>
6032declare i8 @llvm.atomic.load.or.i8.( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6033declare i16 @llvm.atomic.load.or.i16.( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6034declare i32 @llvm.atomic.load.or.i32.( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6035declare i64 @llvm.atomic.load.or.i64.( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
6036
6037</pre>
6038
6039<pre>
6040declare i8 @llvm.atomic.load.nand.i8.( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6041declare i16 @llvm.atomic.load.nand.i16.( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6042declare i32 @llvm.atomic.load.nand.i32.( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6043declare i64 @llvm.atomic.load.nand.i64.( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
6044
6045</pre>
6046
6047<pre>
6048declare i8 @llvm.atomic.load.xor.i8.( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6049declare i16 @llvm.atomic.load.xor.i16.( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6050declare i32 @llvm.atomic.load.xor.i32.( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6051declare i64 @llvm.atomic.load.xor.i64.( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
6052
6053</pre>
6054<h5>Overview:</h5>
6055<p>
6056 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6057 the value stored in memory at <tt>ptr</tt>. It yields the original value
6058 at <tt>ptr</tt>.
6059</p>
6060<h5>Arguments:</h5>
6061<p>
6062
6063 These intrinsics take two arguments, the first a pointer to an integer value
6064 and the second an integer value. The result is also an integer value. These
6065 integer types can have any bit width, but they must all have the same bit
6066 width. The targets may only lower integer representations they support.
6067</p>
6068<h5>Semantics:</h5>
6069<p>
6070 These intrinsics does a series of operations atomically. They first load the
6071 value stored at <tt>ptr</tt>. They then do the bitwise operation
6072 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6073 value stored at <tt>ptr</tt>.
6074</p>
6075
6076<h5>Examples:</h5>
6077<pre>
6078%ptr = malloc i32
6079 store i32 0x0F0F, %ptr
6080%result0 = call i32 @llvm.atomic.load.nand.i32( i32* %ptr, i32 0xFF )
6081 <i>; yields {i32}:result0 = 0x0F0F</i>
6082%result1 = call i32 @llvm.atomic.load.and.i32( i32* %ptr, i32 0xFF )
6083 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
6084%result2 = call i32 @llvm.atomic.load.or.i32( i32* %ptr, i32 0F )
6085 <i>; yields {i32}:result2 = 0xF0</i>
6086%result3 = call i32 @llvm.atomic.load.xor.i32( i32* %ptr, i32 0F )
6087 <i>; yields {i32}:result3 = FF</i>
6088%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6089</pre>
6090</div>
6091
6092
6093<!-- _______________________________________________________________________ -->
6094<div class="doc_subsubsection">
6095 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6096 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6097 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6098 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6099
6100</div>
6101<div class="doc_text">
6102<h5>Syntax:</h5>
6103<p>
6104 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6105 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
6106 <tt>llvm.atomic.load_umin</tt> on any integer bit width. Not all targets
6107 support all bit widths however.</p>
6108<pre>
6109declare i8 @llvm.atomic.load.max.i8.( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6110declare i16 @llvm.atomic.load.max.i16.( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6111declare i32 @llvm.atomic.load.max.i32.( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6112declare i64 @llvm.atomic.load.max.i64.( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
6113
6114</pre>
6115
6116<pre>
6117declare i8 @llvm.atomic.load.min.i8.( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6118declare i16 @llvm.atomic.load.min.i16.( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6119declare i32 @llvm.atomic.load.min.i32.( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6120declare i64 @llvm.atomic.load.min.i64.( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
6121
6122</pre>
6123
6124<pre>
6125declare i8 @llvm.atomic.load.umax.i8.( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6126declare i16 @llvm.atomic.load.umax.i16.( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6127declare i32 @llvm.atomic.load.umax.i32.( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6128declare i64 @llvm.atomic.load.umax.i64.( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
6129
6130</pre>
6131
6132<pre>
6133declare i8 @llvm.atomic.load.umin.i8.( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6134declare i16 @llvm.atomic.load.umin.i16.( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6135declare i32 @llvm.atomic.load.umin.i32.( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6136declare i64 @llvm.atomic.load.umin.i64.( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
6137
6138</pre>
6139<h5>Overview:</h5>
6140<p>
6141 These intrinsics takes the signed or unsigned minimum or maximum of
6142 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6143 original value at <tt>ptr</tt>.
6144</p>
6145<h5>Arguments:</h5>
6146<p>
6147
6148 These intrinsics take two arguments, the first a pointer to an integer value
6149 and the second an integer value. The result is also an integer value. These
6150 integer types can have any bit width, but they must all have the same bit
6151 width. The targets may only lower integer representations they support.
6152</p>
6153<h5>Semantics:</h5>
6154<p>
6155 These intrinsics does a series of operations atomically. They first load the
6156 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6157 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6158 the original value stored at <tt>ptr</tt>.
6159</p>
6160
6161<h5>Examples:</h5>
6162<pre>
6163%ptr = malloc i32
6164 store i32 7, %ptr
6165%result0 = call i32 @llvm.atomic.load.min.i32( i32* %ptr, i32 -2 )
6166 <i>; yields {i32}:result0 = 7</i>
6167%result1 = call i32 @llvm.atomic.load.max.i32( i32* %ptr, i32 8 )
6168 <i>; yields {i32}:result1 = -2</i>
6169%result2 = call i32 @llvm.atomic.load.umin.i32( i32* %ptr, i32 10 )
6170 <i>; yields {i32}:result2 = 8</i>
6171%result3 = call i32 @llvm.atomic.load.umax.i32( i32* %ptr, i32 30 )
6172 <i>; yields {i32}:result3 = 8</i>
6173%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6174</pre>
6175</div>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]6176
6177<!-- ======================================================================= -->
6178<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6179 <a name="int_general">General Intrinsics</a>
6180</div>
6181
6182<div class="doc_text">
6183<p> This class of intrinsics is designed to be generic and has
6184no specific purpose. </p>
6185</div>
6186
6187<!-- _______________________________________________________________________ -->
6188<div class="doc_subsubsection">
6189 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6190</div>
6191
6192<div class="doc_text">
6193
6194<h5>Syntax:</h5>
6195<pre>
6196 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6197</pre>
6198
6199<h5>Overview:</h5>
6200
6201<p>
6202The '<tt>llvm.var.annotation</tt>' intrinsic
6203</p>
6204
6205<h5>Arguments:</h5>
6206
6207<p>
6208The first argument is a pointer to a value, the second is a pointer to a
6209global string, the third is a pointer to a global string which is the source
6210file name, and the last argument is the line number.
6211</p>
6212
6213<h5>Semantics:</h5>
6214
6215<p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6216This intrinsic allows annotation of local variables with arbitrary strings.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6217This can be useful for special purpose optimizations that want to look for these
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6218annotations. These have no other defined use, they are ignored by code
6219generation and optimization.
6220</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6221</div>
6222
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6223<!-- _______________________________________________________________________ -->
6224<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]6225 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6226</div>
6227
6228<div class="doc_text">
6229
6230<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6231<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6232any integer bit width.
6233</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6234<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]6235 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6236 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6237 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6238 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6239 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]6240</pre>
6241
6242<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6243
6244<p>
6245The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6246</p>
6247
6248<h5>Arguments:</h5>
6249
6250<p>
6251The first argument is an integer value (result of some expression),
6252the second is a pointer to a global string, the third is a pointer to a global
6253string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]6254It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]6255</p>
6256
6257<h5>Semantics:</h5>
6258
6259<p>
6260This intrinsic allows annotations to be put on arbitrary expressions
6261with arbitrary strings. This can be useful for special purpose optimizations
6262that want to look for these annotations. These have no other defined use, they
6263are ignored by code generation and optimization.
6264</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6265
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]6266<!-- _______________________________________________________________________ -->
6267<div class="doc_subsubsection">
6268 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6269</div>
6270
6271<div class="doc_text">
6272
6273<h5>Syntax:</h5>
6274<pre>
6275 declare void @llvm.trap()
6276</pre>
6277
6278<h5>Overview:</h5>
6279
6280<p>
6281The '<tt>llvm.trap</tt>' intrinsic
6282</p>
6283
6284<h5>Arguments:</h5>
6285
6286<p>
6287None
6288</p>
6289
6290<h5>Semantics:</h5>
6291
6292<p>
6293This intrinsics is lowered to the target dependent trap instruction. If the
6294target does not have a trap instruction, this intrinsic will be lowered to the
6295call of the abort() function.
6296</p>
6297</div>
6298
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6299<!-- *********************************************************************** -->
6300<hr>
6301<address>
6302 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
6303 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
6304 <a href="http://validator.w3.org/check/referer"><img
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]6305 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6306
6307 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
6308 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
6309 Last modified: $Date$
6310</address>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]6311
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]6312</body>
6313</html>