Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 785610d18b401d3138edf13037d21d7ec89e748a / . / docs / LangRef.html
blob: 6267cf806cc5b9ca4f6b41a836b1bba5aeae8918 [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>
114 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
115 <ol>
116 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
117 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
118 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
119 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
120 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
121 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
122 </ol>
123 </li>
124 <li><a href="#convertops">Conversion Operations</a>
125 <ol>
126 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
127 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
128 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
129 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
130 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
131 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
132 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
133 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
134 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
135 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
136 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
137 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
138 </ol>
139 <li><a href="#otherops">Other Operations</a>
140 <ol>
141 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
142 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
143 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
144 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
145 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
146 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
147 </ol>
148 </li>
149 </ol>
150 </li>
151 <li><a href="#intrinsics">Intrinsic Functions</a>
152 <ol>
153 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
154 <ol>
155 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
156 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
157 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
158 </ol>
159 </li>
160 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
161 <ol>
162 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
163 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
164 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
165 </ol>
166 </li>
167 <li><a href="#int_codegen">Code Generator Intrinsics</a>
168 <ol>
169 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
170 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
171 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
172 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
173 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
174 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
175 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
176 </ol>
177 </li>
178 <li><a href="#int_libc">Standard C Library Intrinsics</a>
179 <ol>
180 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
181 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
182 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
183 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
184 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]185 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
186 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
187 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]188 </ol>
189 </li>
190 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
191 <ol>
192 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
193 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
194 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
195 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
196 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
197 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
198 </ol>
199 </li>
200 <li><a href="#int_debugger">Debugger intrinsics</a></li>
201 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]202 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]203 <ol>
204 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]205 </ol>
206 </li>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame^]207 <li><a href="#int_atomics">Atomic intrinsics</a>
208 <ol>
209 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></li>
210 </ol>
211 </li>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]212 <li><a href="#int_general">General intrinsics</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]213 <ol>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]214 <li><a href="#int_var_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]215 <tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]216 <li><a href="#int_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]217 <tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]218 <li><a href="#int_trap">
219 <tt>llvm.trap</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]220 </ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]221 </li>
222 </ol>
223 </li>
224</ol>
225
226<div class="doc_author">
227 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
228 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
229</div>
230
231<!-- *********************************************************************** -->
232<div class="doc_section"> <a name="abstract">Abstract </a></div>
233<!-- *********************************************************************** -->
234
235<div class="doc_text">
236<p>This document is a reference manual for the LLVM assembly language.
237LLVM is an SSA based representation that provides type safety,
238low-level operations, flexibility, and the capability of representing
239'all' high-level languages cleanly. It is the common code
240representation used throughout all phases of the LLVM compilation
241strategy.</p>
242</div>
243
244<!-- *********************************************************************** -->
245<div class="doc_section"> <a name="introduction">Introduction</a> </div>
246<!-- *********************************************************************** -->
247
248<div class="doc_text">
249
250<p>The LLVM code representation is designed to be used in three
251different forms: as an in-memory compiler IR, as an on-disk bitcode
252representation (suitable for fast loading by a Just-In-Time compiler),
253and as a human readable assembly language representation. This allows
254LLVM to provide a powerful intermediate representation for efficient
255compiler transformations and analysis, while providing a natural means
256to debug and visualize the transformations. The three different forms
257of LLVM are all equivalent. This document describes the human readable
258representation and notation.</p>
259
260<p>The LLVM representation aims to be light-weight and low-level
261while being expressive, typed, and extensible at the same time. It
262aims to be a "universal IR" of sorts, by being at a low enough level
263that high-level ideas may be cleanly mapped to it (similar to how
264microprocessors are "universal IR's", allowing many source languages to
265be mapped to them). By providing type information, LLVM can be used as
266the target of optimizations: for example, through pointer analysis, it
267can be proven that a C automatic variable is never accessed outside of
268the current function... allowing it to be promoted to a simple SSA
269value instead of a memory location.</p>
270
271</div>
272
273<!-- _______________________________________________________________________ -->
274<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
275
276<div class="doc_text">
277
278<p>It is important to note that this document describes 'well formed'
279LLVM assembly language. There is a difference between what the parser
280accepts and what is considered 'well formed'. For example, the
281following instruction is syntactically okay, but not well formed:</p>
282
283<div class="doc_code">
284<pre>
285%x = <a href="#i_add">add</a> i32 1, %x
286</pre>
287</div>
288
289<p>...because the definition of <tt>%x</tt> does not dominate all of
290its uses. The LLVM infrastructure provides a verification pass that may
291be used to verify that an LLVM module is well formed. This pass is
292automatically run by the parser after parsing input assembly and by
293the optimizer before it outputs bitcode. The violations pointed out
294by the verifier pass indicate bugs in transformation passes or input to
295the parser.</p>
296</div>
297
Chris Lattnera83fdc02007-10-03 17:34:29 +0000[diff] [blame]298<!-- Describe the typesetting conventions here. -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]299
300<!-- *********************************************************************** -->
301<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
302<!-- *********************************************************************** -->
303
304<div class="doc_text">
305
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]306 <p>LLVM identifiers come in two basic types: global and local. Global
307 identifiers (functions, global variables) begin with the @ character. Local
308 identifiers (register names, types) begin with the % character. Additionally,
309 there are three different formats for identifiers, for different purposes:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]310
311<ol>
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]312 <li>Named values are represented as a string of characters with their prefix.
313 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
314 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]315 Identifiers which require other characters in their names can be surrounded
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]316 with quotes. In this way, anything except a <tt>&quot;</tt> character can
317 be used in a named value.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]318
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]319 <li>Unnamed values are represented as an unsigned numeric value with their
320 prefix. For example, %12, @2, %44.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]321
322 <li>Constants, which are described in a <a href="#constants">section about
323 constants</a>, below.</li>
324</ol>
325
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]326<p>LLVM requires that values start with a prefix for two reasons: Compilers
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]327don't need to worry about name clashes with reserved words, and the set of
328reserved words may be expanded in the future without penalty. Additionally,
329unnamed identifiers allow a compiler to quickly come up with a temporary
330variable without having to avoid symbol table conflicts.</p>
331
332<p>Reserved words in LLVM are very similar to reserved words in other
333languages. There are keywords for different opcodes
334('<tt><a href="#i_add">add</a></tt>',
335 '<tt><a href="#i_bitcast">bitcast</a></tt>',
336 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
337href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
338and others. These reserved words cannot conflict with variable names, because
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]339none of them start with a prefix character ('%' or '@').</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]340
341<p>Here is an example of LLVM code to multiply the integer variable
342'<tt>%X</tt>' by 8:</p>
343
344<p>The easy way:</p>
345
346<div class="doc_code">
347<pre>
348%result = <a href="#i_mul">mul</a> i32 %X, 8
349</pre>
350</div>
351
352<p>After strength reduction:</p>
353
354<div class="doc_code">
355<pre>
356%result = <a href="#i_shl">shl</a> i32 %X, i8 3
357</pre>
358</div>
359
360<p>And the hard way:</p>
361
362<div class="doc_code">
363<pre>
364<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
365<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
366%result = <a href="#i_add">add</a> i32 %1, %1
367</pre>
368</div>
369
370<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
371important lexical features of LLVM:</p>
372
373<ol>
374
375 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
376 line.</li>
377
378 <li>Unnamed temporaries are created when the result of a computation is not
379 assigned to a named value.</li>
380
381 <li>Unnamed temporaries are numbered sequentially</li>
382
383</ol>
384
385<p>...and it also shows a convention that we follow in this document. When
386demonstrating instructions, we will follow an instruction with a comment that
387defines the type and name of value produced. Comments are shown in italic
388text.</p>
389
390</div>
391
392<!-- *********************************************************************** -->
393<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
394<!-- *********************************************************************** -->
395
396<!-- ======================================================================= -->
397<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
398</div>
399
400<div class="doc_text">
401
402<p>LLVM programs are composed of "Module"s, each of which is a
403translation unit of the input programs. Each module consists of
404functions, global variables, and symbol table entries. Modules may be
405combined together with the LLVM linker, which merges function (and
406global variable) definitions, resolves forward declarations, and merges
407symbol table entries. Here is an example of the "hello world" module:</p>
408
409<div class="doc_code">
410<pre><i>; Declare the string constant as a global constant...</i>
411<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
412 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
413
414<i>; External declaration of the puts function</i>
415<a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
416
417<i>; Definition of main function</i>
418define i32 @main() { <i>; i32()* </i>
419 <i>; Convert [13x i8 ]* to i8 *...</i>
420 %cast210 = <a
421 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
422
423 <i>; Call puts function to write out the string to stdout...</i>
424 <a
425 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
426 <a
427 href="#i_ret">ret</a> i32 0<br>}<br>
428</pre>
429</div>
430
431<p>This example is made up of a <a href="#globalvars">global variable</a>
432named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
433function, and a <a href="#functionstructure">function definition</a>
434for "<tt>main</tt>".</p>
435
436<p>In general, a module is made up of a list of global values,
437where both functions and global variables are global values. Global values are
438represented by a pointer to a memory location (in this case, a pointer to an
439array of char, and a pointer to a function), and have one of the following <a
440href="#linkage">linkage types</a>.</p>
441
442</div>
443
444<!-- ======================================================================= -->
445<div class="doc_subsection">
446 <a name="linkage">Linkage Types</a>
447</div>
448
449<div class="doc_text">
450
451<p>
452All Global Variables and Functions have one of the following types of linkage:
453</p>
454
455<dl>
456
457 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
458
459 <dd>Global values with internal linkage are only directly accessible by
460 objects in the current module. In particular, linking code into a module with
461 an internal global value may cause the internal to be renamed as necessary to
462 avoid collisions. Because the symbol is internal to the module, all
463 references can be updated. This corresponds to the notion of the
464 '<tt>static</tt>' keyword in C.
465 </dd>
466
467 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
468
469 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
470 the same name when linkage occurs. This is typically used to implement
471 inline functions, templates, or other code which must be generated in each
472 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
473 allowed to be discarded.
474 </dd>
475
476 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
477
478 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
479 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
480 used for globals that may be emitted in multiple translation units, but that
481 are not guaranteed to be emitted into every translation unit that uses them.
482 One example of this are common globals in C, such as "<tt>int X;</tt>" at
483 global scope.
484 </dd>
485
486 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
487
488 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
489 pointer to array type. When two global variables with appending linkage are
490 linked together, the two global arrays are appended together. This is the
491 LLVM, typesafe, equivalent of having the system linker append together
492 "sections" with identical names when .o files are linked.
493 </dd>
494
495 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
496 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
497 until linked, if not linked, the symbol becomes null instead of being an
498 undefined reference.
499 </dd>
500
501 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
502
503 <dd>If none of the above identifiers are used, the global is externally
504 visible, meaning that it participates in linkage and can be used to resolve
505 external symbol references.
506 </dd>
507</dl>
508
509 <p>
510 The next two types of linkage are targeted for Microsoft Windows platform
511 only. They are designed to support importing (exporting) symbols from (to)
512 DLLs.
513 </p>
514
515 <dl>
516 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
517
518 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
519 or variable via a global pointer to a pointer that is set up by the DLL
520 exporting the symbol. On Microsoft Windows targets, the pointer name is
521 formed by combining <code>_imp__</code> and the function or variable name.
522 </dd>
523
524 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
525
526 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
527 pointer to a pointer in a DLL, so that it can be referenced with the
528 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
529 name is formed by combining <code>_imp__</code> and the function or variable
530 name.
531 </dd>
532
533</dl>
534
535<p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
536variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
537variable and was linked with this one, one of the two would be renamed,
538preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
539external (i.e., lacking any linkage declarations), they are accessible
540outside of the current module.</p>
541<p>It is illegal for a function <i>declaration</i>
542to have any linkage type other than "externally visible", <tt>dllimport</tt>,
543or <tt>extern_weak</tt>.</p>
544<p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
545linkages.
546</div>
547
548<!-- ======================================================================= -->
549<div class="doc_subsection">
550 <a name="callingconv">Calling Conventions</a>
551</div>
552
553<div class="doc_text">
554
555<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
556and <a href="#i_invoke">invokes</a> can all have an optional calling convention
557specified for the call. The calling convention of any pair of dynamic
558caller/callee must match, or the behavior of the program is undefined. The
559following calling conventions are supported by LLVM, and more may be added in
560the future:</p>
561
562<dl>
563 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
564
565 <dd>This calling convention (the default if no other calling convention is
566 specified) matches the target C calling conventions. This calling convention
567 supports varargs function calls and tolerates some mismatch in the declared
568 prototype and implemented declaration of the function (as does normal C).
569 </dd>
570
571 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
572
573 <dd>This calling convention attempts to make calls as fast as possible
574 (e.g. by passing things in registers). This calling convention allows the
575 target to use whatever tricks it wants to produce fast code for the target,
576 without having to conform to an externally specified ABI. Implementations of
577 this convention should allow arbitrary tail call optimization to be supported.
578 This calling convention does not support varargs and requires the prototype of
579 all callees to exactly match the prototype of the function definition.
580 </dd>
581
582 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
583
584 <dd>This calling convention attempts to make code in the caller as efficient
585 as possible under the assumption that the call is not commonly executed. As
586 such, these calls often preserve all registers so that the call does not break
587 any live ranges in the caller side. This calling convention does not support
588 varargs and requires the prototype of all callees to exactly match the
589 prototype of the function definition.
590 </dd>
591
592 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
593
594 <dd>Any calling convention may be specified by number, allowing
595 target-specific calling conventions to be used. Target specific calling
596 conventions start at 64.
597 </dd>
598</dl>
599
600<p>More calling conventions can be added/defined on an as-needed basis, to
601support pascal conventions or any other well-known target-independent
602convention.</p>
603
604</div>
605
606<!-- ======================================================================= -->
607<div class="doc_subsection">
608 <a name="visibility">Visibility Styles</a>
609</div>
610
611<div class="doc_text">
612
613<p>
614All Global Variables and Functions have one of the following visibility styles:
615</p>
616
617<dl>
618 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
619
620 <dd>On ELF, default visibility means that the declaration is visible to other
621 modules and, in shared libraries, means that the declared entity may be
622 overridden. On Darwin, default visibility means that the declaration is
623 visible to other modules. Default visibility corresponds to "external
624 linkage" in the language.
625 </dd>
626
627 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
628
629 <dd>Two declarations of an object with hidden visibility refer to the same
630 object if they are in the same shared object. Usually, hidden visibility
631 indicates that the symbol will not be placed into the dynamic symbol table,
632 so no other module (executable or shared library) can reference it
633 directly.
634 </dd>
635
636 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
637
638 <dd>On ELF, protected visibility indicates that the symbol will be placed in
639 the dynamic symbol table, but that references within the defining module will
640 bind to the local symbol. That is, the symbol cannot be overridden by another
641 module.
642 </dd>
643</dl>
644
645</div>
646
647<!-- ======================================================================= -->
648<div class="doc_subsection">
649 <a name="globalvars">Global Variables</a>
650</div>
651
652<div class="doc_text">
653
654<p>Global variables define regions of memory allocated at compilation time
655instead of run-time. Global variables may optionally be initialized, may have
656an explicit section to be placed in, and may have an optional explicit alignment
657specified. A variable may be defined as "thread_local", which means that it
658will not be shared by threads (each thread will have a separated copy of the
659variable). A variable may be defined as a global "constant," which indicates
660that the contents of the variable will <b>never</b> be modified (enabling better
661optimization, allowing the global data to be placed in the read-only section of
662an executable, etc). Note that variables that need runtime initialization
663cannot be marked "constant" as there is a store to the variable.</p>
664
665<p>
666LLVM explicitly allows <em>declarations</em> of global variables to be marked
667constant, even if the final definition of the global is not. This capability
668can be used to enable slightly better optimization of the program, but requires
669the language definition to guarantee that optimizations based on the
670'constantness' are valid for the translation units that do not include the
671definition.
672</p>
673
674<p>As SSA values, global variables define pointer values that are in
675scope (i.e. they dominate) all basic blocks in the program. Global
676variables always define a pointer to their "content" type because they
677describe a region of memory, and all memory objects in LLVM are
678accessed through pointers.</p>
679
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]680<p>A global variable may be declared to reside in a target-specifc numbered
681address space. For targets that support them, address spaces may affect how
682optimizations are performed and/or what target instructions are used to access
Christopher Lamb20a39e92007-12-12 08:44:39 +0000[diff] [blame]683the variable. The default address space is zero. The address space qualifier
684must precede any other attributes.</p>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]685
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]686<p>LLVM allows an explicit section to be specified for globals. If the target
687supports it, it will emit globals to the section specified.</p>
688
689<p>An explicit alignment may be specified for a global. If not present, or if
690the alignment is set to zero, the alignment of the global is set by the target
691to whatever it feels convenient. If an explicit alignment is specified, the
692global is forced to have at least that much alignment. All alignments must be
693a power of 2.</p>
694
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]695<p>For example, the following defines a global in a numbered address space with
696an initializer, section, and alignment:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]697
698<div class="doc_code">
699<pre>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]700@G = constant float 1.0 addrspace(5), section "foo", align 4
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]701</pre>
702</div>
703
704</div>
705
706
707<!-- ======================================================================= -->
708<div class="doc_subsection">
709 <a name="functionstructure">Functions</a>
710</div>
711
712<div class="doc_text">
713
714<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
715an optional <a href="#linkage">linkage type</a>, an optional
716<a href="#visibility">visibility style</a>, an optional
717<a href="#callingconv">calling convention</a>, a return type, an optional
718<a href="#paramattrs">parameter attribute</a> for the return type, a function
719name, a (possibly empty) argument list (each with optional
720<a href="#paramattrs">parameter attributes</a>), an optional section, an
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]721optional alignment, an optional <a href="#gc">garbage collector name</a>, an
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]722opening curly brace, a list of basic blocks, and a closing curly brace.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]723
724LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
725optional <a href="#linkage">linkage type</a>, an optional
726<a href="#visibility">visibility style</a>, an optional
727<a href="#callingconv">calling convention</a>, a return type, an optional
728<a href="#paramattrs">parameter attribute</a> for the return type, a function
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]729name, a possibly empty list of arguments, an optional alignment, and an optional
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]730<a href="#gc">garbage collector name</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]731
732<p>A function definition contains a list of basic blocks, forming the CFG for
733the function. Each basic block may optionally start with a label (giving the
734basic block a symbol table entry), contains a list of instructions, and ends
735with a <a href="#terminators">terminator</a> instruction (such as a branch or
736function return).</p>
737
738<p>The first basic block in a function is special in two ways: it is immediately
739executed on entrance to the function, and it is not allowed to have predecessor
740basic blocks (i.e. there can not be any branches to the entry block of a
741function). Because the block can have no predecessors, it also cannot have any
742<a href="#i_phi">PHI nodes</a>.</p>
743
744<p>LLVM allows an explicit section to be specified for functions. If the target
745supports it, it will emit functions to the section specified.</p>
746
747<p>An explicit alignment may be specified for a function. If not present, or if
748the alignment is set to zero, the alignment of the function is set by the target
749to whatever it feels convenient. If an explicit alignment is specified, the
750function is forced to have at least that much alignment. All alignments must be
751a power of 2.</p>
752
753</div>
754
755
756<!-- ======================================================================= -->
757<div class="doc_subsection">
758 <a name="aliasstructure">Aliases</a>
759</div>
760<div class="doc_text">
761 <p>Aliases act as "second name" for the aliasee value (which can be either
762 function or global variable or bitcast of global value). Aliases may have an
763 optional <a href="#linkage">linkage type</a>, and an
764 optional <a href="#visibility">visibility style</a>.</p>
765
766 <h5>Syntax:</h5>
767
768<div class="doc_code">
769<pre>
770@&lt;Name&gt; = [Linkage] [Visibility] alias &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
771</pre>
772</div>
773
774</div>
775
776
777
778<!-- ======================================================================= -->
779<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
780<div class="doc_text">
781 <p>The return type and each parameter of a function type may have a set of
782 <i>parameter attributes</i> associated with them. Parameter attributes are
783 used to communicate additional information about the result or parameters of
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]784 a function. Parameter attributes are considered to be part of the function,
785 not of the function type, so functions with different parameter attributes
786 can have the same function type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]787
788 <p>Parameter attributes are simple keywords that follow the type specified. If
789 multiple parameter attributes are needed, they are space separated. For
790 example:</p>
791
792<div class="doc_code">
793<pre>
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]794declare i32 @printf(i8* noalias , ...) nounwind
795declare i32 @atoi(i8*) nounwind readonly
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]796</pre>
797</div>
798
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]799 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
800 <tt>readonly</tt>) come immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]801
802 <p>Currently, only the following parameter attributes are defined:</p>
803 <dl>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]804 <dt><tt>zeroext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]805 <dd>This indicates that the parameter should be zero extended just before
806 a call to this function.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]807
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]808 <dt><tt>signext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]809 <dd>This indicates that the parameter should be sign extended just before
810 a call to this function.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]811
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]812 <dt><tt>inreg</tt></dt>
813 <dd>This indicates that the parameter should be placed in register (if
814 possible) during assembling function call. Support for this attribute is
815 target-specific</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]816
817 <dt><tt>byval</tt></dt>
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]818 <dd>This indicates that the pointer parameter should really be passed by
819 value to the function. The attribute implies that a hidden copy of the
820 pointee is made between the caller and the callee, so the callee is unable
821 to modify the value in the callee. This attribute is only valid on llvm
822 pointer arguments. It is generally used to pass structs and arrays by
823 value, but is also valid on scalars (even though this is silly).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]824
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]825 <dt><tt>sret</tt></dt>
826 <dd>This indicates that the parameter specifies the address of a structure
827 that is the return value of the function in the source program.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]828
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]829 <dt><tt>noalias</tt></dt>
830 <dd>This indicates that the parameter not alias any other object or any
831 other "noalias" objects during the function call.
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]832
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]833 <dt><tt>noreturn</tt></dt>
834 <dd>This function attribute indicates that the function never returns. This
835 indicates to LLVM that every call to this function should be treated as if
836 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]837
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]838 <dt><tt>nounwind</tt></dt>
839 <dd>This function attribute indicates that the function type does not use
840 the unwind instruction and does not allow stack unwinding to propagate
841 through it.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]842
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]843 <dt><tt>nest</tt></dt>
844 <dd>This indicates that the parameter can be excised using the
845 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
Duncan Sands13e13f82007-11-22 20:23:04 +0000[diff] [blame]846 <dt><tt>readonly</tt></dt>
Duncan Sandsd69c0e62007-11-14 21:14:02 +0000[diff] [blame]847 <dd>This function attribute indicates that the function has no side-effects
Duncan Sands13e13f82007-11-22 20:23:04 +0000[diff] [blame]848 except for producing a return value or throwing an exception. The value
849 returned must only depend on the function arguments and/or global variables.
850 It may use values obtained by dereferencing pointers.</dd>
851 <dt><tt>readnone</tt></dt>
852 <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]853 function, but in addition it is not allowed to dereference any pointer arguments
854 or global variables.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]855 </dl>
856
857</div>
858
859<!-- ======================================================================= -->
860<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]861 <a name="gc">Garbage Collector Names</a>
862</div>
863
864<div class="doc_text">
865<p>Each function may specify a garbage collector name, which is simply a
866string.</p>
867
868<div class="doc_code"><pre
869>define void @f() gc "name" { ...</pre></div>
870
871<p>The compiler declares the supported values of <i>name</i>. Specifying a
872collector which will cause the compiler to alter its output in order to support
873the named garbage collection algorithm.</p>
874</div>
875
876<!-- ======================================================================= -->
877<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]878 <a name="moduleasm">Module-Level Inline Assembly</a>
879</div>
880
881<div class="doc_text">
882<p>
883Modules may contain "module-level inline asm" blocks, which corresponds to the
884GCC "file scope inline asm" blocks. These blocks are internally concatenated by
885LLVM and treated as a single unit, but may be separated in the .ll file if
886desired. The syntax is very simple:
887</p>
888
889<div class="doc_code">
890<pre>
891module asm "inline asm code goes here"
892module asm "more can go here"
893</pre>
894</div>
895
896<p>The strings can contain any character by escaping non-printable characters.
897 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
898 for the number.
899</p>
900
901<p>
902 The inline asm code is simply printed to the machine code .s file when
903 assembly code is generated.
904</p>
905</div>
906
907<!-- ======================================================================= -->
908<div class="doc_subsection">
909 <a name="datalayout">Data Layout</a>
910</div>
911
912<div class="doc_text">
913<p>A module may specify a target specific data layout string that specifies how
914data is to be laid out in memory. The syntax for the data layout is simply:</p>
915<pre> target datalayout = "<i>layout specification</i>"</pre>
916<p>The <i>layout specification</i> consists of a list of specifications
917separated by the minus sign character ('-'). Each specification starts with a
918letter and may include other information after the letter to define some
919aspect of the data layout. The specifications accepted are as follows: </p>
920<dl>
921 <dt><tt>E</tt></dt>
922 <dd>Specifies that the target lays out data in big-endian form. That is, the
923 bits with the most significance have the lowest address location.</dd>
924 <dt><tt>e</tt></dt>
925 <dd>Specifies that hte target lays out data in little-endian form. That is,
926 the bits with the least significance have the lowest address location.</dd>
927 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
928 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
929 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
930 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
931 too.</dd>
932 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
933 <dd>This specifies the alignment for an integer type of a given bit
934 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
935 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
936 <dd>This specifies the alignment for a vector type of a given bit
937 <i>size</i>.</dd>
938 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
939 <dd>This specifies the alignment for a floating point type of a given bit
940 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
941 (double).</dd>
942 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
943 <dd>This specifies the alignment for an aggregate type of a given bit
944 <i>size</i>.</dd>
945</dl>
946<p>When constructing the data layout for a given target, LLVM starts with a
947default set of specifications which are then (possibly) overriden by the
948specifications in the <tt>datalayout</tt> keyword. The default specifications
949are given in this list:</p>
950<ul>
951 <li><tt>E</tt> - big endian</li>
952 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
953 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
954 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
955 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
956 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
957 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
958 alignment of 64-bits</li>
959 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
960 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
961 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
962 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
963 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
964</ul>
965<p>When llvm is determining the alignment for a given type, it uses the
966following rules:
967<ol>
968 <li>If the type sought is an exact match for one of the specifications, that
969 specification is used.</li>
970 <li>If no match is found, and the type sought is an integer type, then the
971 smallest integer type that is larger than the bitwidth of the sought type is
972 used. If none of the specifications are larger than the bitwidth then the the
973 largest integer type is used. For example, given the default specifications
974 above, the i7 type will use the alignment of i8 (next largest) while both
975 i65 and i256 will use the alignment of i64 (largest specified).</li>
976 <li>If no match is found, and the type sought is a vector type, then the
977 largest vector type that is smaller than the sought vector type will be used
978 as a fall back. This happens because <128 x double> can be implemented in
979 terms of 64 <2 x double>, for example.</li>
980</ol>
981</div>
982
983<!-- *********************************************************************** -->
984<div class="doc_section"> <a name="typesystem">Type System</a> </div>
985<!-- *********************************************************************** -->
986
987<div class="doc_text">
988
989<p>The LLVM type system is one of the most important features of the
990intermediate representation. Being typed enables a number of
991optimizations to be performed on the IR directly, without having to do
992extra analyses on the side before the transformation. A strong type
993system makes it easier to read the generated code and enables novel
994analyses and transformations that are not feasible to perform on normal
995three address code representations.</p>
996
997</div>
998
999<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1000<div class="doc_subsection"> <a name="t_classifications">Type
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1001Classifications</a> </div>
1002<div class="doc_text">
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1003<p>The types fall into a few useful
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1004classifications:</p>
1005
1006<table border="1" cellspacing="0" cellpadding="4">
1007 <tbody>
1008 <tr><th>Classification</th><th>Types</th></tr>
1009 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1010 <td><a href="#t_integer">integer</a></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1011 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1012 </tr>
1013 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1014 <td><a href="#t_floating">floating point</a></td>
1015 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1016 </tr>
1017 <tr>
1018 <td><a name="t_firstclass">first class</a></td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1019 <td><a href="#t_integer">integer</a>,
1020 <a href="#t_floating">floating point</a>,
1021 <a href="#t_pointer">pointer</a>,
1022 <a href="#t_vector">vector</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1023 </td>
1024 </tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1025 <tr>
1026 <td><a href="#t_primitive">primitive</a></td>
1027 <td><a href="#t_label">label</a>,
1028 <a href="#t_void">void</a>,
1029 <a href="#t_integer">integer</a>,
1030 <a href="#t_floating">floating point</a>.</td>
1031 </tr>
1032 <tr>
1033 <td><a href="#t_derived">derived</a></td>
1034 <td><a href="#t_integer">integer</a>,
1035 <a href="#t_array">array</a>,
1036 <a href="#t_function">function</a>,
1037 <a href="#t_pointer">pointer</a>,
1038 <a href="#t_struct">structure</a>,
1039 <a href="#t_pstruct">packed structure</a>,
1040 <a href="#t_vector">vector</a>,
1041 <a href="#t_opaque">opaque</a>.
1042 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1043 </tbody>
1044</table>
1045
1046<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1047most important. Values of these types are the only ones which can be
1048produced by instructions, passed as arguments, or used as operands to
1049instructions. This means that all structures and arrays must be
1050manipulated either by pointer or by component.</p>
1051</div>
1052
1053<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1054<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1055
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1056<div class="doc_text">
1057<p>The primitive types are the fundamental building blocks of the LLVM
1058system.</p>
1059
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1060</div>
1061
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1062<!-- _______________________________________________________________________ -->
1063<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1064
1065<div class="doc_text">
1066 <table>
1067 <tbody>
1068 <tr><th>Type</th><th>Description</th></tr>
1069 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1070 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1071 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1072 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1073 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1074 </tbody>
1075 </table>
1076</div>
1077
1078<!-- _______________________________________________________________________ -->
1079<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1080
1081<div class="doc_text">
1082<h5>Overview:</h5>
1083<p>The void type does not represent any value and has no size.</p>
1084
1085<h5>Syntax:</h5>
1086
1087<pre>
1088 void
1089</pre>
1090</div>
1091
1092<!-- _______________________________________________________________________ -->
1093<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1094
1095<div class="doc_text">
1096<h5>Overview:</h5>
1097<p>The label type represents code labels.</p>
1098
1099<h5>Syntax:</h5>
1100
1101<pre>
1102 label
1103</pre>
1104</div>
1105
1106
1107<!-- ======================================================================= -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1108<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1109
1110<div class="doc_text">
1111
1112<p>The real power in LLVM comes from the derived types in the system.
1113This is what allows a programmer to represent arrays, functions,
1114pointers, and other useful types. Note that these derived types may be
1115recursive: For example, it is possible to have a two dimensional array.</p>
1116
1117</div>
1118
1119<!-- _______________________________________________________________________ -->
1120<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1121
1122<div class="doc_text">
1123
1124<h5>Overview:</h5>
1125<p>The integer type is a very simple derived type that simply specifies an
1126arbitrary bit width for the integer type desired. Any bit width from 1 bit to
11272^23-1 (about 8 million) can be specified.</p>
1128
1129<h5>Syntax:</h5>
1130
1131<pre>
1132 iN
1133</pre>
1134
1135<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1136value.</p>
1137
1138<h5>Examples:</h5>
1139<table class="layout">
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1140 <tbody>
1141 <tr>
1142 <td><tt>i1</tt></td>
1143 <td>a single-bit integer.</td>
1144 </tr><tr>
1145 <td><tt>i32</tt></td>
1146 <td>a 32-bit integer.</td>
1147 </tr><tr>
1148 <td><tt>i1942652</tt></td>
1149 <td>a really big integer of over 1 million bits.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1150 </tr>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1151 </tbody>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1152</table>
1153</div>
1154
1155<!-- _______________________________________________________________________ -->
1156<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1157
1158<div class="doc_text">
1159
1160<h5>Overview:</h5>
1161
1162<p>The array type is a very simple derived type that arranges elements
1163sequentially in memory. The array type requires a size (number of
1164elements) and an underlying data type.</p>
1165
1166<h5>Syntax:</h5>
1167
1168<pre>
1169 [&lt;# elements&gt; x &lt;elementtype&gt;]
1170</pre>
1171
1172<p>The number of elements is a constant integer value; elementtype may
1173be any type with a size.</p>
1174
1175<h5>Examples:</h5>
1176<table class="layout">
1177 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1178 <td class="left"><tt>[40 x i32]</tt></td>
1179 <td class="left">Array of 40 32-bit integer values.</td>
1180 </tr>
1181 <tr class="layout">
1182 <td class="left"><tt>[41 x i32]</tt></td>
1183 <td class="left">Array of 41 32-bit integer values.</td>
1184 </tr>
1185 <tr class="layout">
1186 <td class="left"><tt>[4 x i8]</tt></td>
1187 <td class="left">Array of 4 8-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1188 </tr>
1189</table>
1190<p>Here are some examples of multidimensional arrays:</p>
1191<table class="layout">
1192 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1193 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1194 <td class="left">3x4 array of 32-bit integer values.</td>
1195 </tr>
1196 <tr class="layout">
1197 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1198 <td class="left">12x10 array of single precision floating point values.</td>
1199 </tr>
1200 <tr class="layout">
1201 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1202 <td class="left">2x3x4 array of 16-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1203 </tr>
1204</table>
1205
1206<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1207length array. Normally, accesses past the end of an array are undefined in
1208LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1209As a special case, however, zero length arrays are recognized to be variable
1210length. This allows implementation of 'pascal style arrays' with the LLVM
1211type "{ i32, [0 x float]}", for example.</p>
1212
1213</div>
1214
1215<!-- _______________________________________________________________________ -->
1216<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1217<div class="doc_text">
1218<h5>Overview:</h5>
1219<p>The function type can be thought of as a function signature. It
1220consists of a return type and a list of formal parameter types.
1221Function types are usually used to build virtual function tables
1222(which are structures of pointers to functions), for indirect function
1223calls, and when defining a function.</p>
1224<p>
1225The return type of a function type cannot be an aggregate type.
1226</p>
1227<h5>Syntax:</h5>
1228<pre> &lt;returntype&gt; (&lt;parameter list&gt;)<br></pre>
1229<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1230specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1231which indicates that the function takes a variable number of arguments.
1232Variable argument functions can access their arguments with the <a
1233 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1234<h5>Examples:</h5>
1235<table class="layout">
1236 <tr class="layout">
1237 <td class="left"><tt>i32 (i32)</tt></td>
1238 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1239 </td>
1240 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1241 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1242 </tt></td>
1243 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1244 an <tt>i16</tt> that should be sign extended and a
1245 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1246 <tt>float</tt>.
1247 </td>
1248 </tr><tr class="layout">
1249 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1250 <td class="left">A vararg function that takes at least one
1251 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1252 which returns an integer. This is the signature for <tt>printf</tt> in
1253 LLVM.
1254 </td>
1255 </tr>
1256</table>
1257
1258</div>
1259<!-- _______________________________________________________________________ -->
1260<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1261<div class="doc_text">
1262<h5>Overview:</h5>
1263<p>The structure type is used to represent a collection of data members
1264together in memory. The packing of the field types is defined to match
1265the ABI of the underlying processor. The elements of a structure may
1266be any type that has a size.</p>
1267<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1268and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1269field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1270instruction.</p>
1271<h5>Syntax:</h5>
1272<pre> { &lt;type list&gt; }<br></pre>
1273<h5>Examples:</h5>
1274<table class="layout">
1275 <tr class="layout">
1276 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1277 <td class="left">A triple of three <tt>i32</tt> values</td>
1278 </tr><tr class="layout">
1279 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1280 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1281 second element is a <a href="#t_pointer">pointer</a> to a
1282 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1283 an <tt>i32</tt>.</td>
1284 </tr>
1285</table>
1286</div>
1287
1288<!-- _______________________________________________________________________ -->
1289<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1290</div>
1291<div class="doc_text">
1292<h5>Overview:</h5>
1293<p>The packed structure type is used to represent a collection of data members
1294together in memory. There is no padding between fields. Further, the alignment
1295of a packed structure is 1 byte. The elements of a packed structure may
1296be any type that has a size.</p>
1297<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1298and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1299field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1300instruction.</p>
1301<h5>Syntax:</h5>
1302<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1303<h5>Examples:</h5>
1304<table class="layout">
1305 <tr class="layout">
1306 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1307 <td class="left">A triple of three <tt>i32</tt> values</td>
1308 </tr><tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1309 <td class="left"><tt>&lt; { float, i32 (i32)* } &gt;</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1310 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1311 second element is a <a href="#t_pointer">pointer</a> to a
1312 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1313 an <tt>i32</tt>.</td>
1314 </tr>
1315</table>
1316</div>
1317
1318<!-- _______________________________________________________________________ -->
1319<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1320<div class="doc_text">
1321<h5>Overview:</h5>
1322<p>As in many languages, the pointer type represents a pointer or
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1323reference to another object, which must live in memory. Pointer types may have
1324an optional address space attribute defining the target-specific numbered
1325address space where the pointed-to object resides. The default address space is
1326zero.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1327<h5>Syntax:</h5>
1328<pre> &lt;type&gt; *<br></pre>
1329<h5>Examples:</h5>
1330<table class="layout">
1331 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1332 <td class="left"><tt>[4x i32]*</tt></td>
1333 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1334 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1335 </tr>
1336 <tr class="layout">
1337 <td class="left"><tt>i32 (i32 *) *</tt></td>
1338 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1339 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1340 <tt>i32</tt>.</td>
1341 </tr>
1342 <tr class="layout">
1343 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1344 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1345 that resides in address space #5.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1346 </tr>
1347</table>
1348</div>
1349
1350<!-- _______________________________________________________________________ -->
1351<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1352<div class="doc_text">
1353
1354<h5>Overview:</h5>
1355
1356<p>A vector type is a simple derived type that represents a vector
1357of elements. Vector types are used when multiple primitive data
1358are operated in parallel using a single instruction (SIMD).
1359A vector type requires a size (number of
1360elements) and an underlying primitive data type. Vectors must have a power
1361of two length (1, 2, 4, 8, 16 ...). Vector types are
1362considered <a href="#t_firstclass">first class</a>.</p>
1363
1364<h5>Syntax:</h5>
1365
1366<pre>
1367 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1368</pre>
1369
1370<p>The number of elements is a constant integer value; elementtype may
1371be any integer or floating point type.</p>
1372
1373<h5>Examples:</h5>
1374
1375<table class="layout">
1376 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1377 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1378 <td class="left">Vector of 4 32-bit integer values.</td>
1379 </tr>
1380 <tr class="layout">
1381 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1382 <td class="left">Vector of 8 32-bit floating-point values.</td>
1383 </tr>
1384 <tr class="layout">
1385 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1386 <td class="left">Vector of 2 64-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1387 </tr>
1388</table>
1389</div>
1390
1391<!-- _______________________________________________________________________ -->
1392<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1393<div class="doc_text">
1394
1395<h5>Overview:</h5>
1396
1397<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1398corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1399In LLVM, opaque types can eventually be resolved to any type (not just a
1400structure type).</p>
1401
1402<h5>Syntax:</h5>
1403
1404<pre>
1405 opaque
1406</pre>
1407
1408<h5>Examples:</h5>
1409
1410<table class="layout">
1411 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1412 <td class="left"><tt>opaque</tt></td>
1413 <td class="left">An opaque type.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1414 </tr>
1415</table>
1416</div>
1417
1418
1419<!-- *********************************************************************** -->
1420<div class="doc_section"> <a name="constants">Constants</a> </div>
1421<!-- *********************************************************************** -->
1422
1423<div class="doc_text">
1424
1425<p>LLVM has several different basic types of constants. This section describes
1426them all and their syntax.</p>
1427
1428</div>
1429
1430<!-- ======================================================================= -->
1431<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1432
1433<div class="doc_text">
1434
1435<dl>
1436 <dt><b>Boolean constants</b></dt>
1437
1438 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1439 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1440 </dd>
1441
1442 <dt><b>Integer constants</b></dt>
1443
1444 <dd>Standard integers (such as '4') are constants of the <a
1445 href="#t_integer">integer</a> type. Negative numbers may be used with
1446 integer types.
1447 </dd>
1448
1449 <dt><b>Floating point constants</b></dt>
1450
1451 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1452 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1453 notation (see below). Floating point constants must have a <a
1454 href="#t_floating">floating point</a> type. </dd>
1455
1456 <dt><b>Null pointer constants</b></dt>
1457
1458 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1459 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1460
1461</dl>
1462
1463<p>The one non-intuitive notation for constants is the optional hexadecimal form
1464of floating point constants. For example, the form '<tt>double
14650x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
14664.5e+15</tt>'. The only time hexadecimal floating point constants are required
1467(and the only time that they are generated by the disassembler) is when a
1468floating point constant must be emitted but it cannot be represented as a
1469decimal floating point number. For example, NaN's, infinities, and other
1470special values are represented in their IEEE hexadecimal format so that
1471assembly and disassembly do not cause any bits to change in the constants.</p>
1472
1473</div>
1474
1475<!-- ======================================================================= -->
1476<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1477</div>
1478
1479<div class="doc_text">
1480<p>Aggregate constants arise from aggregation of simple constants
1481and smaller aggregate constants.</p>
1482
1483<dl>
1484 <dt><b>Structure constants</b></dt>
1485
1486 <dd>Structure constants are represented with notation similar to structure
1487 type definitions (a comma separated list of elements, surrounded by braces
Chris Lattner6c8de962007-12-25 20:34:52 +0000[diff] [blame]1488 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1489 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]1490 must have <a href="#t_struct">structure type</a>, and the number and
1491 types of elements must match those specified by the type.
1492 </dd>
1493
1494 <dt><b>Array constants</b></dt>
1495
1496 <dd>Array constants are represented with notation similar to array type
1497 definitions (a comma separated list of elements, surrounded by square brackets
1498 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1499 constants must have <a href="#t_array">array type</a>, and the number and
1500 types of elements must match those specified by the type.
1501 </dd>
1502
1503 <dt><b>Vector constants</b></dt>
1504
1505 <dd>Vector constants are represented with notation similar to vector type
1506 definitions (a comma separated list of elements, surrounded by
1507 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1508 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1509 href="#t_vector">vector type</a>, and the number and types of elements must
1510 match those specified by the type.
1511 </dd>
1512
1513 <dt><b>Zero initialization</b></dt>
1514
1515 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1516 value to zero of <em>any</em> type, including scalar and aggregate types.
1517 This is often used to avoid having to print large zero initializers (e.g. for
1518 large arrays) and is always exactly equivalent to using explicit zero
1519 initializers.
1520 </dd>
1521</dl>
1522
1523</div>
1524
1525<!-- ======================================================================= -->
1526<div class="doc_subsection">
1527 <a name="globalconstants">Global Variable and Function Addresses</a>
1528</div>
1529
1530<div class="doc_text">
1531
1532<p>The addresses of <a href="#globalvars">global variables</a> and <a
1533href="#functionstructure">functions</a> are always implicitly valid (link-time)
1534constants. These constants are explicitly referenced when the <a
1535href="#identifiers">identifier for the global</a> is used and always have <a
1536href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1537file:</p>
1538
1539<div class="doc_code">
1540<pre>
1541@X = global i32 17
1542@Y = global i32 42
1543@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1544</pre>
1545</div>
1546
1547</div>
1548
1549<!-- ======================================================================= -->
1550<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1551<div class="doc_text">
1552 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1553 no specific value. Undefined values may be of any type and be used anywhere
1554 a constant is permitted.</p>
1555
1556 <p>Undefined values indicate to the compiler that the program is well defined
1557 no matter what value is used, giving the compiler more freedom to optimize.
1558 </p>
1559</div>
1560
1561<!-- ======================================================================= -->
1562<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1563</div>
1564
1565<div class="doc_text">
1566
1567<p>Constant expressions are used to allow expressions involving other constants
1568to be used as constants. Constant expressions may be of any <a
1569href="#t_firstclass">first class</a> type and may involve any LLVM operation
1570that does not have side effects (e.g. load and call are not supported). The
1571following is the syntax for constant expressions:</p>
1572
1573<dl>
1574 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1575 <dd>Truncate a constant to another type. The bit size of CST must be larger
1576 than the bit size of TYPE. Both types must be integers.</dd>
1577
1578 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1579 <dd>Zero extend a constant to another type. The bit size of CST must be
1580 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1581
1582 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1583 <dd>Sign extend a constant to another type. The bit size of CST must be
1584 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1585
1586 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1587 <dd>Truncate a floating point constant to another floating point type. The
1588 size of CST must be larger than the size of TYPE. Both types must be
1589 floating point.</dd>
1590
1591 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1592 <dd>Floating point extend a constant to another type. The size of CST must be
1593 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1594
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1595 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1596 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1597 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1598 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1599 of the same number of elements. If the value won't fit in the integer type,
1600 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1601
1602 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1603 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1604 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1605 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1606 of the same number of elements. If the value won't fit in the integer type,
1607 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1608
1609 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1610 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1611 constant. TYPE must be a scalar or vector floating point type. CST must be of
1612 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1613 of the same number of elements. If the value won't fit in the floating point
1614 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1615
1616 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1617 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1618 constant. TYPE must be a scalar or vector floating point type. CST must be of
1619 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1620 of the same number of elements. If the value won't fit in the floating point
1621 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1622
1623 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1624 <dd>Convert a pointer typed constant to the corresponding integer constant
1625 TYPE must be an integer type. CST must be of pointer type. The CST value is
1626 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1627
1628 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1629 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1630 pointer type. CST must be of integer type. The CST value is zero extended,
1631 truncated, or unchanged to make it fit in a pointer size. This one is
1632 <i>really</i> dangerous!</dd>
1633
1634 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1635 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1636 identical (same number of bits). The conversion is done as if the CST value
1637 was stored to memory and read back as TYPE. In other words, no bits change
1638 with this operator, just the type. This can be used for conversion of
1639 vector types to any other type, as long as they have the same bit width. For
1640 pointers it is only valid to cast to another pointer type.
1641 </dd>
1642
1643 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1644
1645 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1646 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1647 instruction, the index list may have zero or more indexes, which are required
1648 to make sense for the type of "CSTPTR".</dd>
1649
1650 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1651
1652 <dd>Perform the <a href="#i_select">select operation</a> on
1653 constants.</dd>
1654
1655 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1656 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1657
1658 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1659 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1660
1661 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1662
1663 <dd>Perform the <a href="#i_extractelement">extractelement
1664 operation</a> on constants.
1665
1666 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1667
1668 <dd>Perform the <a href="#i_insertelement">insertelement
1669 operation</a> on constants.</dd>
1670
1671
1672 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1673
1674 <dd>Perform the <a href="#i_shufflevector">shufflevector
1675 operation</a> on constants.</dd>
1676
1677 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1678
1679 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1680 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1681 binary</a> operations. The constraints on operands are the same as those for
1682 the corresponding instruction (e.g. no bitwise operations on floating point
1683 values are allowed).</dd>
1684</dl>
1685</div>
1686
1687<!-- *********************************************************************** -->
1688<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1689<!-- *********************************************************************** -->
1690
1691<!-- ======================================================================= -->
1692<div class="doc_subsection">
1693<a name="inlineasm">Inline Assembler Expressions</a>
1694</div>
1695
1696<div class="doc_text">
1697
1698<p>
1699LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1700Module-Level Inline Assembly</a>) through the use of a special value. This
1701value represents the inline assembler as a string (containing the instructions
1702to emit), a list of operand constraints (stored as a string), and a flag that
1703indicates whether or not the inline asm expression has side effects. An example
1704inline assembler expression is:
1705</p>
1706
1707<div class="doc_code">
1708<pre>
1709i32 (i32) asm "bswap $0", "=r,r"
1710</pre>
1711</div>
1712
1713<p>
1714Inline assembler expressions may <b>only</b> be used as the callee operand of
1715a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1716</p>
1717
1718<div class="doc_code">
1719<pre>
1720%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1721</pre>
1722</div>
1723
1724<p>
1725Inline asms with side effects not visible in the constraint list must be marked
1726as having side effects. This is done through the use of the
1727'<tt>sideeffect</tt>' keyword, like so:
1728</p>
1729
1730<div class="doc_code">
1731<pre>
1732call void asm sideeffect "eieio", ""()
1733</pre>
1734</div>
1735
1736<p>TODO: The format of the asm and constraints string still need to be
1737documented here. Constraints on what can be done (e.g. duplication, moving, etc
1738need to be documented).
1739</p>
1740
1741</div>
1742
1743<!-- *********************************************************************** -->
1744<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1745<!-- *********************************************************************** -->
1746
1747<div class="doc_text">
1748
1749<p>The LLVM instruction set consists of several different
1750classifications of instructions: <a href="#terminators">terminator
1751instructions</a>, <a href="#binaryops">binary instructions</a>,
1752<a href="#bitwiseops">bitwise binary instructions</a>, <a
1753 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1754instructions</a>.</p>
1755
1756</div>
1757
1758<!-- ======================================================================= -->
1759<div class="doc_subsection"> <a name="terminators">Terminator
1760Instructions</a> </div>
1761
1762<div class="doc_text">
1763
1764<p>As mentioned <a href="#functionstructure">previously</a>, every
1765basic block in a program ends with a "Terminator" instruction, which
1766indicates which block should be executed after the current block is
1767finished. These terminator instructions typically yield a '<tt>void</tt>'
1768value: they produce control flow, not values (the one exception being
1769the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1770<p>There are six different terminator instructions: the '<a
1771 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1772instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1773the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1774 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1775 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1776
1777</div>
1778
1779<!-- _______________________________________________________________________ -->
1780<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1781Instruction</a> </div>
1782<div class="doc_text">
1783<h5>Syntax:</h5>
1784<pre> ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
1785 ret void <i>; Return from void function</i>
1786</pre>
1787<h5>Overview:</h5>
1788<p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1789value) from a function back to the caller.</p>
1790<p>There are two forms of the '<tt>ret</tt>' instruction: one that
1791returns a value and then causes control flow, and one that just causes
1792control flow to occur.</p>
1793<h5>Arguments:</h5>
1794<p>The '<tt>ret</tt>' instruction may return any '<a
1795 href="#t_firstclass">first class</a>' type. Notice that a function is
1796not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1797instruction inside of the function that returns a value that does not
1798match the return type of the function.</p>
1799<h5>Semantics:</h5>
1800<p>When the '<tt>ret</tt>' instruction is executed, control flow
1801returns back to the calling function's context. If the caller is a "<a
1802 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1803the instruction after the call. If the caller was an "<a
1804 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1805at the beginning of the "normal" destination block. If the instruction
1806returns a value, that value shall set the call or invoke instruction's
1807return value.</p>
1808<h5>Example:</h5>
1809<pre> ret i32 5 <i>; Return an integer value of 5</i>
1810 ret void <i>; Return from a void function</i>
1811</pre>
1812</div>
1813<!-- _______________________________________________________________________ -->
1814<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1815<div class="doc_text">
1816<h5>Syntax:</h5>
1817<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
1818</pre>
1819<h5>Overview:</h5>
1820<p>The '<tt>br</tt>' instruction is used to cause control flow to
1821transfer to a different basic block in the current function. There are
1822two forms of this instruction, corresponding to a conditional branch
1823and an unconditional branch.</p>
1824<h5>Arguments:</h5>
1825<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1826single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1827unconditional form of the '<tt>br</tt>' instruction takes a single
1828'<tt>label</tt>' value as a target.</p>
1829<h5>Semantics:</h5>
1830<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1831argument is evaluated. If the value is <tt>true</tt>, control flows
1832to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1833control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1834<h5>Example:</h5>
1835<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
1836 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1837</div>
1838<!-- _______________________________________________________________________ -->
1839<div class="doc_subsubsection">
1840 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1841</div>
1842
1843<div class="doc_text">
1844<h5>Syntax:</h5>
1845
1846<pre>
1847 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
1848</pre>
1849
1850<h5>Overview:</h5>
1851
1852<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1853several different places. It is a generalization of the '<tt>br</tt>'
1854instruction, allowing a branch to occur to one of many possible
1855destinations.</p>
1856
1857
1858<h5>Arguments:</h5>
1859
1860<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1861comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1862an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1863table is not allowed to contain duplicate constant entries.</p>
1864
1865<h5>Semantics:</h5>
1866
1867<p>The <tt>switch</tt> instruction specifies a table of values and
1868destinations. When the '<tt>switch</tt>' instruction is executed, this
1869table is searched for the given value. If the value is found, control flow is
1870transfered to the corresponding destination; otherwise, control flow is
1871transfered to the default destination.</p>
1872
1873<h5>Implementation:</h5>
1874
1875<p>Depending on properties of the target machine and the particular
1876<tt>switch</tt> instruction, this instruction may be code generated in different
1877ways. For example, it could be generated as a series of chained conditional
1878branches or with a lookup table.</p>
1879
1880<h5>Example:</h5>
1881
1882<pre>
1883 <i>; Emulate a conditional br instruction</i>
1884 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1885 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1886
1887 <i>; Emulate an unconditional br instruction</i>
1888 switch i32 0, label %dest [ ]
1889
1890 <i>; Implement a jump table:</i>
1891 switch i32 %val, label %otherwise [ i32 0, label %onzero
1892 i32 1, label %onone
1893 i32 2, label %ontwo ]
1894</pre>
1895</div>
1896
1897<!-- _______________________________________________________________________ -->
1898<div class="doc_subsubsection">
1899 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1900</div>
1901
1902<div class="doc_text">
1903
1904<h5>Syntax:</h5>
1905
1906<pre>
1907 &lt;result&gt; = invoke [<a href="#callingconv">cconv</a>] &lt;ptr to function ty&gt; %&lt;function ptr val&gt;(&lt;function args&gt;)
1908 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
1909</pre>
1910
1911<h5>Overview:</h5>
1912
1913<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1914function, with the possibility of control flow transfer to either the
1915'<tt>normal</tt>' label or the
1916'<tt>exception</tt>' label. If the callee function returns with the
1917"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1918"normal" label. If the callee (or any indirect callees) returns with the "<a
1919href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1920continued at the dynamically nearest "exception" label.</p>
1921
1922<h5>Arguments:</h5>
1923
1924<p>This instruction requires several arguments:</p>
1925
1926<ol>
1927 <li>
1928 The optional "cconv" marker indicates which <a href="#callingconv">calling
1929 convention</a> the call should use. If none is specified, the call defaults
1930 to using C calling conventions.
1931 </li>
1932 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1933 function value being invoked. In most cases, this is a direct function
1934 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1935 an arbitrary pointer to function value.
1936 </li>
1937
1938 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1939 function to be invoked. </li>
1940
1941 <li>'<tt>function args</tt>': argument list whose types match the function
1942 signature argument types. If the function signature indicates the function
1943 accepts a variable number of arguments, the extra arguments can be
1944 specified. </li>
1945
1946 <li>'<tt>normal label</tt>': the label reached when the called function
1947 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1948
1949 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1950 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1951
1952</ol>
1953
1954<h5>Semantics:</h5>
1955
1956<p>This instruction is designed to operate as a standard '<tt><a
1957href="#i_call">call</a></tt>' instruction in most regards. The primary
1958difference is that it establishes an association with a label, which is used by
1959the runtime library to unwind the stack.</p>
1960
1961<p>This instruction is used in languages with destructors to ensure that proper
1962cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1963exception. Additionally, this is important for implementation of
1964'<tt>catch</tt>' clauses in high-level languages that support them.</p>
1965
1966<h5>Example:</h5>
1967<pre>
1968 %retval = invoke i32 %Test(i32 15) to label %Continue
1969 unwind label %TestCleanup <i>; {i32}:retval set</i>
1970 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1971 unwind label %TestCleanup <i>; {i32}:retval set</i>
1972</pre>
1973</div>
1974
1975
1976<!-- _______________________________________________________________________ -->
1977
1978<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1979Instruction</a> </div>
1980
1981<div class="doc_text">
1982
1983<h5>Syntax:</h5>
1984<pre>
1985 unwind
1986</pre>
1987
1988<h5>Overview:</h5>
1989
1990<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1991at the first callee in the dynamic call stack which used an <a
1992href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1993primarily used to implement exception handling.</p>
1994
1995<h5>Semantics:</h5>
1996
1997<p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1998immediately halt. The dynamic call stack is then searched for the first <a
1999href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2000execution continues at the "exceptional" destination block specified by the
2001<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2002dynamic call chain, undefined behavior results.</p>
2003</div>
2004
2005<!-- _______________________________________________________________________ -->
2006
2007<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2008Instruction</a> </div>
2009
2010<div class="doc_text">
2011
2012<h5>Syntax:</h5>
2013<pre>
2014 unreachable
2015</pre>
2016
2017<h5>Overview:</h5>
2018
2019<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2020instruction is used to inform the optimizer that a particular portion of the
2021code is not reachable. This can be used to indicate that the code after a
2022no-return function cannot be reached, and other facts.</p>
2023
2024<h5>Semantics:</h5>
2025
2026<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2027</div>
2028
2029
2030
2031<!-- ======================================================================= -->
2032<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2033<div class="doc_text">
2034<p>Binary operators are used to do most of the computation in a
2035program. They require two operands, execute an operation on them, and
2036produce a single value. The operands might represent
2037multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
2038The result value of a binary operator is not
2039necessarily the same type as its operands.</p>
2040<p>There are several different binary operators:</p>
2041</div>
2042<!-- _______________________________________________________________________ -->
2043<div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
2044Instruction</a> </div>
2045<div class="doc_text">
2046<h5>Syntax:</h5>
2047<pre> &lt;result&gt; = add &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2048</pre>
2049<h5>Overview:</h5>
2050<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2051<h5>Arguments:</h5>
2052<p>The two arguments to the '<tt>add</tt>' instruction must be either <a
2053 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
2054 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2055Both arguments must have identical types.</p>
2056<h5>Semantics:</h5>
2057<p>The value produced is the integer or floating point sum of the two
2058operands.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2059<p>If an integer sum has unsigned overflow, the result returned is the
2060mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2061the result.</p>
2062<p>Because LLVM integers use a two's complement representation, this
2063instruction is appropriate for both signed and unsigned integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2064<h5>Example:</h5>
2065<pre> &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2066</pre>
2067</div>
2068<!-- _______________________________________________________________________ -->
2069<div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
2070Instruction</a> </div>
2071<div class="doc_text">
2072<h5>Syntax:</h5>
2073<pre> &lt;result&gt; = sub &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2074</pre>
2075<h5>Overview:</h5>
2076<p>The '<tt>sub</tt>' instruction returns the difference of its two
2077operands.</p>
2078<p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
2079instruction present in most other intermediate representations.</p>
2080<h5>Arguments:</h5>
2081<p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
2082 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2083values.
2084This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2085Both arguments must have identical types.</p>
2086<h5>Semantics:</h5>
2087<p>The value produced is the integer or floating point difference of
2088the two operands.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2089<p>If an integer difference has unsigned overflow, the result returned is the
2090mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2091the result.</p>
2092<p>Because LLVM integers use a two's complement representation, this
2093instruction is appropriate for both signed and unsigned integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2094<h5>Example:</h5>
2095<pre>
2096 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2097 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2098</pre>
2099</div>
2100<!-- _______________________________________________________________________ -->
2101<div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
2102Instruction</a> </div>
2103<div class="doc_text">
2104<h5>Syntax:</h5>
2105<pre> &lt;result&gt; = mul &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2106</pre>
2107<h5>Overview:</h5>
2108<p>The '<tt>mul</tt>' instruction returns the product of its two
2109operands.</p>
2110<h5>Arguments:</h5>
2111<p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
2112 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2113values.
2114This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2115Both arguments must have identical types.</p>
2116<h5>Semantics:</h5>
2117<p>The value produced is the integer or floating point product of the
2118two operands.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2119<p>If the result of an integer multiplication has unsigned overflow,
2120the result returned is the mathematical result modulo
21212<sup>n</sup>, where n is the bit width of the result.</p>
2122<p>Because LLVM integers use a two's complement representation, and the
2123result is the same width as the operands, this instruction returns the
2124correct result for both signed and unsigned integers. If a full product
2125(e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2126should be sign-extended or zero-extended as appropriate to the
2127width of the full product.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2128<h5>Example:</h5>
2129<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2130</pre>
2131</div>
2132<!-- _______________________________________________________________________ -->
2133<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2134</a></div>
2135<div class="doc_text">
2136<h5>Syntax:</h5>
2137<pre> &lt;result&gt; = udiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2138</pre>
2139<h5>Overview:</h5>
2140<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2141operands.</p>
2142<h5>Arguments:</h5>
2143<p>The two arguments to the '<tt>udiv</tt>' instruction must be
2144<a href="#t_integer">integer</a> values. Both arguments must have identical
2145types. This instruction can also take <a href="#t_vector">vector</a> versions
2146of the values in which case the elements must be integers.</p>
2147<h5>Semantics:</h5>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2148<p>The value produced is the unsigned integer quotient of the two operands.</p>
2149<p>Note that unsigned integer division and signed integer division are distinct
2150operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2151<p>Division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2152<h5>Example:</h5>
2153<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2154</pre>
2155</div>
2156<!-- _______________________________________________________________________ -->
2157<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2158</a> </div>
2159<div class="doc_text">
2160<h5>Syntax:</h5>
2161<pre> &lt;result&gt; = sdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2162</pre>
2163<h5>Overview:</h5>
2164<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2165operands.</p>
2166<h5>Arguments:</h5>
2167<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2168<a href="#t_integer">integer</a> values. Both arguments must have identical
2169types. This instruction can also take <a href="#t_vector">vector</a> versions
2170of the values in which case the elements must be integers.</p>
2171<h5>Semantics:</h5>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2172<p>The value produced is the signed integer quotient of the two operands.</p>
2173<p>Note that signed integer division and unsigned integer division are distinct
2174operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2175<p>Division by zero leads to undefined behavior. Overflow also leads to
2176undefined behavior; this is a rare case, but can occur, for example,
2177by doing a 32-bit division of -2147483648 by -1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2178<h5>Example:</h5>
2179<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2180</pre>
2181</div>
2182<!-- _______________________________________________________________________ -->
2183<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2184Instruction</a> </div>
2185<div class="doc_text">
2186<h5>Syntax:</h5>
2187<pre> &lt;result&gt; = fdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2188</pre>
2189<h5>Overview:</h5>
2190<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2191operands.</p>
2192<h5>Arguments:</h5>
2193<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2194<a href="#t_floating">floating point</a> values. Both arguments must have
2195identical types. This instruction can also take <a href="#t_vector">vector</a>
2196versions of floating point values.</p>
2197<h5>Semantics:</h5>
2198<p>The value produced is the floating point quotient of the two operands.</p>
2199<h5>Example:</h5>
2200<pre> &lt;result&gt; = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2201</pre>
2202</div>
2203<!-- _______________________________________________________________________ -->
2204<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2205</div>
2206<div class="doc_text">
2207<h5>Syntax:</h5>
2208<pre> &lt;result&gt; = urem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2209</pre>
2210<h5>Overview:</h5>
2211<p>The '<tt>urem</tt>' instruction returns the remainder from the
2212unsigned division of its two arguments.</p>
2213<h5>Arguments:</h5>
2214<p>The two arguments to the '<tt>urem</tt>' instruction must be
2215<a href="#t_integer">integer</a> values. Both arguments must have identical
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2216types. This instruction can also take <a href="#t_vector">vector</a> versions
2217of the values in which case the elements must be integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2218<h5>Semantics:</h5>
2219<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2220This instruction always performs an unsigned division to get the remainder,
2221regardless of whether the arguments are unsigned or not.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2222<p>Note that unsigned integer remainder and signed integer remainder are
2223distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2224<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2225<h5>Example:</h5>
2226<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2227</pre>
2228
2229</div>
2230<!-- _______________________________________________________________________ -->
2231<div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2232Instruction</a> </div>
2233<div class="doc_text">
2234<h5>Syntax:</h5>
2235<pre> &lt;result&gt; = srem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2236</pre>
2237<h5>Overview:</h5>
2238<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2239signed division of its two operands. This instruction can also take
2240<a href="#t_vector">vector</a> versions of the values in which case
2241the elements must be integers.</p>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]2242
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2243<h5>Arguments:</h5>
2244<p>The two arguments to the '<tt>srem</tt>' instruction must be
2245<a href="#t_integer">integer</a> values. Both arguments must have identical
2246types.</p>
2247<h5>Semantics:</h5>
2248<p>This instruction returns the <i>remainder</i> of a division (where the result
2249has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2250operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2251a value. For more information about the difference, see <a
2252 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2253Math Forum</a>. For a table of how this is implemented in various languages,
2254please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2255Wikipedia: modulo operation</a>.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2256<p>Note that signed integer remainder and unsigned integer remainder are
2257distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2258<p>Taking the remainder of a division by zero leads to undefined behavior.
2259Overflow also leads to undefined behavior; this is a rare case, but can occur,
2260for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2261(The remainder doesn't actually overflow, but this rule lets srem be
2262implemented using instructions that return both the result of the division
2263and the remainder.)</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2264<h5>Example:</h5>
2265<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2266</pre>
2267
2268</div>
2269<!-- _______________________________________________________________________ -->
2270<div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2271Instruction</a> </div>
2272<div class="doc_text">
2273<h5>Syntax:</h5>
2274<pre> &lt;result&gt; = frem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2275</pre>
2276<h5>Overview:</h5>
2277<p>The '<tt>frem</tt>' instruction returns the remainder from the
2278division of its two operands.</p>
2279<h5>Arguments:</h5>
2280<p>The two arguments to the '<tt>frem</tt>' instruction must be
2281<a href="#t_floating">floating point</a> values. Both arguments must have
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2282identical types. This instruction can also take <a href="#t_vector">vector</a>
2283versions of floating point values.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2284<h5>Semantics:</h5>
2285<p>This instruction returns the <i>remainder</i> of a division.</p>
2286<h5>Example:</h5>
2287<pre> &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2288</pre>
2289</div>
2290
2291<!-- ======================================================================= -->
2292<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2293Operations</a> </div>
2294<div class="doc_text">
2295<p>Bitwise binary operators are used to do various forms of
2296bit-twiddling in a program. They are generally very efficient
2297instructions and can commonly be strength reduced from other
2298instructions. They require two operands, execute an operation on them,
2299and produce a single value. The resulting value of the bitwise binary
2300operators is always the same type as its first operand.</p>
2301</div>
2302
2303<!-- _______________________________________________________________________ -->
2304<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2305Instruction</a> </div>
2306<div class="doc_text">
2307<h5>Syntax:</h5>
2308<pre> &lt;result&gt; = shl &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2309</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2310
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2311<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2312
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2313<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2314the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2315
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2316<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2317
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2318<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2319 href="#t_integer">integer</a> type.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2320
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2321<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2322
2323<p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>. If
2324<tt>var2</tt> is (statically or dynamically) equal to or larger than the number
2325of bits in <tt>var1</tt>, the result is undefined.</p>
2326
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2327<h5>Example:</h5><pre>
2328 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2329 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2330 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2331 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2332</pre>
2333</div>
2334<!-- _______________________________________________________________________ -->
2335<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2336Instruction</a> </div>
2337<div class="doc_text">
2338<h5>Syntax:</h5>
2339<pre> &lt;result&gt; = lshr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2340</pre>
2341
2342<h5>Overview:</h5>
2343<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2344operand shifted to the right a specified number of bits with zero fill.</p>
2345
2346<h5>Arguments:</h5>
2347<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2348<a href="#t_integer">integer</a> type.</p>
2349
2350<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2351
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2352<p>This instruction always performs a logical shift right operation. The most
2353significant bits of the result will be filled with zero bits after the
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2354shift. If <tt>var2</tt> is (statically or dynamically) equal to or larger than
2355the number of bits in <tt>var1</tt>, the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2356
2357<h5>Example:</h5>
2358<pre>
2359 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2360 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2361 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2362 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2363 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2364</pre>
2365</div>
2366
2367<!-- _______________________________________________________________________ -->
2368<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2369Instruction</a> </div>
2370<div class="doc_text">
2371
2372<h5>Syntax:</h5>
2373<pre> &lt;result&gt; = ashr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2374</pre>
2375
2376<h5>Overview:</h5>
2377<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2378operand shifted to the right a specified number of bits with sign extension.</p>
2379
2380<h5>Arguments:</h5>
2381<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2382<a href="#t_integer">integer</a> type.</p>
2383
2384<h5>Semantics:</h5>
2385<p>This instruction always performs an arithmetic shift right operation,
2386The most significant bits of the result will be filled with the sign bit
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2387of <tt>var1</tt>. If <tt>var2</tt> is (statically or dynamically) equal to or
2388larger than the number of bits in <tt>var1</tt>, the result is undefined.
2389</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2390
2391<h5>Example:</h5>
2392<pre>
2393 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2394 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2395 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2396 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2397 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2398</pre>
2399</div>
2400
2401<!-- _______________________________________________________________________ -->
2402<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2403Instruction</a> </div>
2404<div class="doc_text">
2405<h5>Syntax:</h5>
2406<pre> &lt;result&gt; = and &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2407</pre>
2408<h5>Overview:</h5>
2409<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2410its two operands.</p>
2411<h5>Arguments:</h5>
2412<p>The two arguments to the '<tt>and</tt>' instruction must be <a
2413 href="#t_integer">integer</a> values. Both arguments must have
2414identical types.</p>
2415<h5>Semantics:</h5>
2416<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2417<p> </p>
2418<div style="align: center">
2419<table border="1" cellspacing="0" cellpadding="4">
2420 <tbody>
2421 <tr>
2422 <td>In0</td>
2423 <td>In1</td>
2424 <td>Out</td>
2425 </tr>
2426 <tr>
2427 <td>0</td>
2428 <td>0</td>
2429 <td>0</td>
2430 </tr>
2431 <tr>
2432 <td>0</td>
2433 <td>1</td>
2434 <td>0</td>
2435 </tr>
2436 <tr>
2437 <td>1</td>
2438 <td>0</td>
2439 <td>0</td>
2440 </tr>
2441 <tr>
2442 <td>1</td>
2443 <td>1</td>
2444 <td>1</td>
2445 </tr>
2446 </tbody>
2447</table>
2448</div>
2449<h5>Example:</h5>
2450<pre> &lt;result&gt; = and i32 4, %var <i>; yields {i32}:result = 4 &amp; %var</i>
2451 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2452 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2453</pre>
2454</div>
2455<!-- _______________________________________________________________________ -->
2456<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2457<div class="doc_text">
2458<h5>Syntax:</h5>
2459<pre> &lt;result&gt; = or &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2460</pre>
2461<h5>Overview:</h5>
2462<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2463or of its two operands.</p>
2464<h5>Arguments:</h5>
2465<p>The two arguments to the '<tt>or</tt>' instruction must be <a
2466 href="#t_integer">integer</a> values. Both arguments must have
2467identical types.</p>
2468<h5>Semantics:</h5>
2469<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2470<p> </p>
2471<div style="align: center">
2472<table border="1" cellspacing="0" cellpadding="4">
2473 <tbody>
2474 <tr>
2475 <td>In0</td>
2476 <td>In1</td>
2477 <td>Out</td>
2478 </tr>
2479 <tr>
2480 <td>0</td>
2481 <td>0</td>
2482 <td>0</td>
2483 </tr>
2484 <tr>
2485 <td>0</td>
2486 <td>1</td>
2487 <td>1</td>
2488 </tr>
2489 <tr>
2490 <td>1</td>
2491 <td>0</td>
2492 <td>1</td>
2493 </tr>
2494 <tr>
2495 <td>1</td>
2496 <td>1</td>
2497 <td>1</td>
2498 </tr>
2499 </tbody>
2500</table>
2501</div>
2502<h5>Example:</h5>
2503<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2504 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2505 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2506</pre>
2507</div>
2508<!-- _______________________________________________________________________ -->
2509<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2510Instruction</a> </div>
2511<div class="doc_text">
2512<h5>Syntax:</h5>
2513<pre> &lt;result&gt; = xor &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2514</pre>
2515<h5>Overview:</h5>
2516<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2517or of its two operands. The <tt>xor</tt> is used to implement the
2518"one's complement" operation, which is the "~" operator in C.</p>
2519<h5>Arguments:</h5>
2520<p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2521 href="#t_integer">integer</a> values. Both arguments must have
2522identical types.</p>
2523<h5>Semantics:</h5>
2524<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2525<p> </p>
2526<div style="align: center">
2527<table border="1" cellspacing="0" cellpadding="4">
2528 <tbody>
2529 <tr>
2530 <td>In0</td>
2531 <td>In1</td>
2532 <td>Out</td>
2533 </tr>
2534 <tr>
2535 <td>0</td>
2536 <td>0</td>
2537 <td>0</td>
2538 </tr>
2539 <tr>
2540 <td>0</td>
2541 <td>1</td>
2542 <td>1</td>
2543 </tr>
2544 <tr>
2545 <td>1</td>
2546 <td>0</td>
2547 <td>1</td>
2548 </tr>
2549 <tr>
2550 <td>1</td>
2551 <td>1</td>
2552 <td>0</td>
2553 </tr>
2554 </tbody>
2555</table>
2556</div>
2557<p> </p>
2558<h5>Example:</h5>
2559<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2560 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2561 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2562 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2563</pre>
2564</div>
2565
2566<!-- ======================================================================= -->
2567<div class="doc_subsection">
2568 <a name="vectorops">Vector Operations</a>
2569</div>
2570
2571<div class="doc_text">
2572
2573<p>LLVM supports several instructions to represent vector operations in a
2574target-independent manner. These instructions cover the element-access and
2575vector-specific operations needed to process vectors effectively. While LLVM
2576does directly support these vector operations, many sophisticated algorithms
2577will want to use target-specific intrinsics to take full advantage of a specific
2578target.</p>
2579
2580</div>
2581
2582<!-- _______________________________________________________________________ -->
2583<div class="doc_subsubsection">
2584 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2585</div>
2586
2587<div class="doc_text">
2588
2589<h5>Syntax:</h5>
2590
2591<pre>
2592 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
2593</pre>
2594
2595<h5>Overview:</h5>
2596
2597<p>
2598The '<tt>extractelement</tt>' instruction extracts a single scalar
2599element from a vector at a specified index.
2600</p>
2601
2602
2603<h5>Arguments:</h5>
2604
2605<p>
2606The first operand of an '<tt>extractelement</tt>' instruction is a
2607value of <a href="#t_vector">vector</a> type. The second operand is
2608an index indicating the position from which to extract the element.
2609The index may be a variable.</p>
2610
2611<h5>Semantics:</h5>
2612
2613<p>
2614The result is a scalar of the same type as the element type of
2615<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2616<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2617results are undefined.
2618</p>
2619
2620<h5>Example:</h5>
2621
2622<pre>
2623 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
2624</pre>
2625</div>
2626
2627
2628<!-- _______________________________________________________________________ -->
2629<div class="doc_subsubsection">
2630 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2631</div>
2632
2633<div class="doc_text">
2634
2635<h5>Syntax:</h5>
2636
2637<pre>
2638 &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>
2639</pre>
2640
2641<h5>Overview:</h5>
2642
2643<p>
2644The '<tt>insertelement</tt>' instruction inserts a scalar
2645element into a vector at a specified index.
2646</p>
2647
2648
2649<h5>Arguments:</h5>
2650
2651<p>
2652The first operand of an '<tt>insertelement</tt>' instruction is a
2653value of <a href="#t_vector">vector</a> type. The second operand is a
2654scalar value whose type must equal the element type of the first
2655operand. The third operand is an index indicating the position at
2656which to insert the value. The index may be a variable.</p>
2657
2658<h5>Semantics:</h5>
2659
2660<p>
2661The result is a vector of the same type as <tt>val</tt>. Its
2662element values are those of <tt>val</tt> except at position
2663<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2664exceeds the length of <tt>val</tt>, the results are undefined.
2665</p>
2666
2667<h5>Example:</h5>
2668
2669<pre>
2670 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
2671</pre>
2672</div>
2673
2674<!-- _______________________________________________________________________ -->
2675<div class="doc_subsubsection">
2676 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2677</div>
2678
2679<div class="doc_text">
2680
2681<h5>Syntax:</h5>
2682
2683<pre>
2684 &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>
2685</pre>
2686
2687<h5>Overview:</h5>
2688
2689<p>
2690The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2691from two input vectors, returning a vector of the same type.
2692</p>
2693
2694<h5>Arguments:</h5>
2695
2696<p>
2697The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2698with types that match each other and types that match the result of the
2699instruction. The third argument is a shuffle mask, which has the same number
2700of elements as the other vector type, but whose element type is always 'i32'.
2701</p>
2702
2703<p>
2704The shuffle mask operand is required to be a constant vector with either
2705constant integer or undef values.
2706</p>
2707
2708<h5>Semantics:</h5>
2709
2710<p>
2711The elements of the two input vectors are numbered from left to right across
2712both of the vectors. The shuffle mask operand specifies, for each element of
2713the result vector, which element of the two input registers the result element
2714gets. The element selector may be undef (meaning "don't care") and the second
2715operand may be undef if performing a shuffle from only one vector.
2716</p>
2717
2718<h5>Example:</h5>
2719
2720<pre>
2721 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
2722 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
2723 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
2724 &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.
2725</pre>
2726</div>
2727
2728
2729<!-- ======================================================================= -->
2730<div class="doc_subsection">
2731 <a name="memoryops">Memory Access and Addressing Operations</a>
2732</div>
2733
2734<div class="doc_text">
2735
2736<p>A key design point of an SSA-based representation is how it
2737represents memory. In LLVM, no memory locations are in SSA form, which
2738makes things very simple. This section describes how to read, write,
2739allocate, and free memory in LLVM.</p>
2740
2741</div>
2742
2743<!-- _______________________________________________________________________ -->
2744<div class="doc_subsubsection">
2745 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2746</div>
2747
2748<div class="doc_text">
2749
2750<h5>Syntax:</h5>
2751
2752<pre>
2753 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
2754</pre>
2755
2756<h5>Overview:</h5>
2757
2758<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]2759heap and returns a pointer to it. The object is always allocated in the generic
2760address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2761
2762<h5>Arguments:</h5>
2763
2764<p>The '<tt>malloc</tt>' instruction allocates
2765<tt>sizeof(&lt;type&gt;)*NumElements</tt>
2766bytes of memory from the operating system and returns a pointer of the
2767appropriate type to the program. If "NumElements" is specified, it is the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]2768number of elements allocated, otherwise "NumElements" is defaulted to be one.
2769If an alignment is specified, the value result of the allocation is guaranteed to
2770be aligned to at least that boundary. If not specified, or if zero, the target can
2771choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2772
2773<p>'<tt>type</tt>' must be a sized type.</p>
2774
2775<h5>Semantics:</h5>
2776
2777<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2778a pointer is returned.</p>
2779
2780<h5>Example:</h5>
2781
2782<pre>
2783 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2784
2785 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2786 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2787 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2788 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2789 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2790</pre>
2791</div>
2792
2793<!-- _______________________________________________________________________ -->
2794<div class="doc_subsubsection">
2795 <a name="i_free">'<tt>free</tt>' Instruction</a>
2796</div>
2797
2798<div class="doc_text">
2799
2800<h5>Syntax:</h5>
2801
2802<pre>
2803 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
2804</pre>
2805
2806<h5>Overview:</h5>
2807
2808<p>The '<tt>free</tt>' instruction returns memory back to the unused
2809memory heap to be reallocated in the future.</p>
2810
2811<h5>Arguments:</h5>
2812
2813<p>'<tt>value</tt>' shall be a pointer value that points to a value
2814that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2815instruction.</p>
2816
2817<h5>Semantics:</h5>
2818
2819<p>Access to the memory pointed to by the pointer is no longer defined
2820after this instruction executes.</p>
2821
2822<h5>Example:</h5>
2823
2824<pre>
2825 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2826 free [4 x i8]* %array
2827</pre>
2828</div>
2829
2830<!-- _______________________________________________________________________ -->
2831<div class="doc_subsubsection">
2832 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2833</div>
2834
2835<div class="doc_text">
2836
2837<h5>Syntax:</h5>
2838
2839<pre>
2840 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
2841</pre>
2842
2843<h5>Overview:</h5>
2844
2845<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
2846currently executing function, to be automatically released when this function
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]2847returns to its caller. The object is always allocated in the generic address
2848space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2849
2850<h5>Arguments:</h5>
2851
2852<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
2853bytes of memory on the runtime stack, returning a pointer of the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]2854appropriate type to the program. If "NumElements" is specified, it is the
2855number of elements allocated, otherwise "NumElements" is defaulted to be one.
2856If an alignment is specified, the value result of the allocation is guaranteed
2857to be aligned to at least that boundary. If not specified, or if zero, the target
2858can choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2859
2860<p>'<tt>type</tt>' may be any sized type.</p>
2861
2862<h5>Semantics:</h5>
2863
2864<p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2865memory is automatically released when the function returns. The '<tt>alloca</tt>'
2866instruction is commonly used to represent automatic variables that must
2867have an address available. When the function returns (either with the <tt><a
2868 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2869instructions), the memory is reclaimed.</p>
2870
2871<h5>Example:</h5>
2872
2873<pre>
2874 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2875 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2876 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2877 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2878</pre>
2879</div>
2880
2881<!-- _______________________________________________________________________ -->
2882<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2883Instruction</a> </div>
2884<div class="doc_text">
2885<h5>Syntax:</h5>
2886<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>
2887<h5>Overview:</h5>
2888<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2889<h5>Arguments:</h5>
2890<p>The argument to the '<tt>load</tt>' instruction specifies the memory
2891address from which to load. The pointer must point to a <a
2892 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2893marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2894the number or order of execution of this <tt>load</tt> with other
2895volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2896instructions. </p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]2897<p>
2898The optional "align" argument specifies the alignment of the operation
2899(that is, the alignment of the memory address). A value of 0 or an
2900omitted "align" argument means that the operation has the preferential
2901alignment for the target. It is the responsibility of the code emitter
2902to ensure that the alignment information is correct. Overestimating
2903the alignment results in an undefined behavior. Underestimating the
2904alignment may produce less efficient code. An alignment of 1 is always
2905safe.
2906</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2907<h5>Semantics:</h5>
2908<p>The location of memory pointed to is loaded.</p>
2909<h5>Examples:</h5>
2910<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2911 <a
2912 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2913 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2914</pre>
2915</div>
2916<!-- _______________________________________________________________________ -->
2917<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2918Instruction</a> </div>
2919<div class="doc_text">
2920<h5>Syntax:</h5>
2921<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
2922 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
2923</pre>
2924<h5>Overview:</h5>
2925<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2926<h5>Arguments:</h5>
2927<p>There are two arguments to the '<tt>store</tt>' instruction: a value
2928to store and an address at which to store it. The type of the '<tt>&lt;pointer&gt;</tt>'
2929operand must be a pointer to the type of the '<tt>&lt;value&gt;</tt>'
2930operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2931optimizer is not allowed to modify the number or order of execution of
2932this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2933 href="#i_store">store</a></tt> instructions.</p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]2934<p>
2935The optional "align" argument specifies the alignment of the operation
2936(that is, the alignment of the memory address). A value of 0 or an
2937omitted "align" argument means that the operation has the preferential
2938alignment for the target. It is the responsibility of the code emitter
2939to ensure that the alignment information is correct. Overestimating
2940the alignment results in an undefined behavior. Underestimating the
2941alignment may produce less efficient code. An alignment of 1 is always
2942safe.
2943</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2944<h5>Semantics:</h5>
2945<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
2946at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
2947<h5>Example:</h5>
2948<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]2949 store i32 3, i32* %ptr <i>; yields {void}</i>
2950 %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]2951</pre>
2952</div>
2953
2954<!-- _______________________________________________________________________ -->
2955<div class="doc_subsubsection">
2956 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2957</div>
2958
2959<div class="doc_text">
2960<h5>Syntax:</h5>
2961<pre>
2962 &lt;result&gt; = getelementptr &lt;ty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
2963</pre>
2964
2965<h5>Overview:</h5>
2966
2967<p>
2968The '<tt>getelementptr</tt>' instruction is used to get the address of a
2969subelement of an aggregate data structure.</p>
2970
2971<h5>Arguments:</h5>
2972
2973<p>This instruction takes a list of integer operands that indicate what
2974elements of the aggregate object to index to. The actual types of the arguments
2975provided depend on the type of the first pointer argument. The
2976'<tt>getelementptr</tt>' instruction is used to index down through the type
2977levels of a structure or to a specific index in an array. When indexing into a
2978structure, only <tt>i32</tt> integer constants are allowed. When indexing
2979into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2980be sign extended to 64-bit values.</p>
2981
2982<p>For example, let's consider a C code fragment and how it gets
2983compiled to LLVM:</p>
2984
2985<div class="doc_code">
2986<pre>
2987struct RT {
2988 char A;
2989 int B[10][20];
2990 char C;
2991};
2992struct ST {
2993 int X;
2994 double Y;
2995 struct RT Z;
2996};
2997
2998int *foo(struct ST *s) {
2999 return &amp;s[1].Z.B[5][13];
3000}
3001</pre>
3002</div>
3003
3004<p>The LLVM code generated by the GCC frontend is:</p>
3005
3006<div class="doc_code">
3007<pre>
3008%RT = type { i8 , [10 x [20 x i32]], i8 }
3009%ST = type { i32, double, %RT }
3010
3011define i32* %foo(%ST* %s) {
3012entry:
3013 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3014 ret i32* %reg
3015}
3016</pre>
3017</div>
3018
3019<h5>Semantics:</h5>
3020
3021<p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
3022on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
3023and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
3024<a href="#t_integer">integer</a> type but the value will always be sign extended
3025to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
3026<b>constants</b>.</p>
3027
3028<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3029type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3030}</tt>' type, a structure. The second index indexes into the third element of
3031the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3032i8 }</tt>' type, another structure. The third index indexes into the second
3033element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3034array. The two dimensions of the array are subscripted into, yielding an
3035'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3036to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3037
3038<p>Note that it is perfectly legal to index partially through a
3039structure, returning a pointer to an inner element. Because of this,
3040the LLVM code for the given testcase is equivalent to:</p>
3041
3042<pre>
3043 define i32* %foo(%ST* %s) {
3044 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3045 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3046 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3047 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3048 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3049 ret i32* %t5
3050 }
3051</pre>
3052
3053<p>Note that it is undefined to access an array out of bounds: array and
3054pointer indexes must always be within the defined bounds of the array type.
3055The one exception for this rules is zero length arrays. These arrays are
3056defined to be accessible as variable length arrays, which requires access
3057beyond the zero'th element.</p>
3058
3059<p>The getelementptr instruction is often confusing. For some more insight
3060into how it works, see <a href="GetElementPtr.html">the getelementptr
3061FAQ</a>.</p>
3062
3063<h5>Example:</h5>
3064
3065<pre>
3066 <i>; yields [12 x i8]*:aptr</i>
3067 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
3068</pre>
3069</div>
3070
3071<!-- ======================================================================= -->
3072<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3073</div>
3074<div class="doc_text">
3075<p>The instructions in this category are the conversion instructions (casting)
3076which all take a single operand and a type. They perform various bit conversions
3077on the operand.</p>
3078</div>
3079
3080<!-- _______________________________________________________________________ -->
3081<div class="doc_subsubsection">
3082 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3083</div>
3084<div class="doc_text">
3085
3086<h5>Syntax:</h5>
3087<pre>
3088 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3089</pre>
3090
3091<h5>Overview:</h5>
3092<p>
3093The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3094</p>
3095
3096<h5>Arguments:</h5>
3097<p>
3098The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3099be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3100and type of the result, which must be an <a href="#t_integer">integer</a>
3101type. The bit size of <tt>value</tt> must be larger than the bit size of
3102<tt>ty2</tt>. Equal sized types are not allowed.</p>
3103
3104<h5>Semantics:</h5>
3105<p>
3106The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3107and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3108larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3109It will always truncate bits.</p>
3110
3111<h5>Example:</h5>
3112<pre>
3113 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3114 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3115 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3116</pre>
3117</div>
3118
3119<!-- _______________________________________________________________________ -->
3120<div class="doc_subsubsection">
3121 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3122</div>
3123<div class="doc_text">
3124
3125<h5>Syntax:</h5>
3126<pre>
3127 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3128</pre>
3129
3130<h5>Overview:</h5>
3131<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3132<tt>ty2</tt>.</p>
3133
3134
3135<h5>Arguments:</h5>
3136<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3137<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3138also be of <a href="#t_integer">integer</a> type. The bit size of the
3139<tt>value</tt> must be smaller than the bit size of the destination type,
3140<tt>ty2</tt>.</p>
3141
3142<h5>Semantics:</h5>
3143<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3144bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3145
3146<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3147
3148<h5>Example:</h5>
3149<pre>
3150 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3151 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3152</pre>
3153</div>
3154
3155<!-- _______________________________________________________________________ -->
3156<div class="doc_subsubsection">
3157 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3158</div>
3159<div class="doc_text">
3160
3161<h5>Syntax:</h5>
3162<pre>
3163 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3164</pre>
3165
3166<h5>Overview:</h5>
3167<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3168
3169<h5>Arguments:</h5>
3170<p>
3171The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3172<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3173also be of <a href="#t_integer">integer</a> type. The bit size of the
3174<tt>value</tt> must be smaller than the bit size of the destination type,
3175<tt>ty2</tt>.</p>
3176
3177<h5>Semantics:</h5>
3178<p>
3179The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3180bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3181the type <tt>ty2</tt>.</p>
3182
3183<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3184
3185<h5>Example:</h5>
3186<pre>
3187 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3188 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3189</pre>
3190</div>
3191
3192<!-- _______________________________________________________________________ -->
3193<div class="doc_subsubsection">
3194 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3195</div>
3196
3197<div class="doc_text">
3198
3199<h5>Syntax:</h5>
3200
3201<pre>
3202 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3203</pre>
3204
3205<h5>Overview:</h5>
3206<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3207<tt>ty2</tt>.</p>
3208
3209
3210<h5>Arguments:</h5>
3211<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3212 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3213cast it to. The size of <tt>value</tt> must be larger than the size of
3214<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3215<i>no-op cast</i>.</p>
3216
3217<h5>Semantics:</h5>
3218<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3219<a href="#t_floating">floating point</a> type to a smaller
3220<a href="#t_floating">floating point</a> type. If the value cannot fit within
3221the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3222
3223<h5>Example:</h5>
3224<pre>
3225 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3226 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3227</pre>
3228</div>
3229
3230<!-- _______________________________________________________________________ -->
3231<div class="doc_subsubsection">
3232 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3233</div>
3234<div class="doc_text">
3235
3236<h5>Syntax:</h5>
3237<pre>
3238 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3239</pre>
3240
3241<h5>Overview:</h5>
3242<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3243floating point value.</p>
3244
3245<h5>Arguments:</h5>
3246<p>The '<tt>fpext</tt>' instruction takes a
3247<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3248and a <a href="#t_floating">floating point</a> type to cast it to. The source
3249type must be smaller than the destination type.</p>
3250
3251<h5>Semantics:</h5>
3252<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3253<a href="#t_floating">floating point</a> type to a larger
3254<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3255used to make a <i>no-op cast</i> because it always changes bits. Use
3256<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3257
3258<h5>Example:</h5>
3259<pre>
3260 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3261 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3262</pre>
3263</div>
3264
3265<!-- _______________________________________________________________________ -->
3266<div class="doc_subsubsection">
3267 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3268</div>
3269<div class="doc_text">
3270
3271<h5>Syntax:</h5>
3272<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3273 &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]3274</pre>
3275
3276<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3277<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3278unsigned integer equivalent of type <tt>ty2</tt>.
3279</p>
3280
3281<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3282<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]3283scalar or vector <a href="#t_floating">floating point</a> value, and a type
3284to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3285type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3286vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3287
3288<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3289<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3290<a href="#t_floating">floating point</a> operand into the nearest (rounding
3291towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3292the results are undefined.</p>
3293
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3294<h5>Example:</h5>
3295<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3296 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3297 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3298 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3299</pre>
3300</div>
3301
3302<!-- _______________________________________________________________________ -->
3303<div class="doc_subsubsection">
3304 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3305</div>
3306<div class="doc_text">
3307
3308<h5>Syntax:</h5>
3309<pre>
3310 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3311</pre>
3312
3313<h5>Overview:</h5>
3314<p>The '<tt>fptosi</tt>' instruction converts
3315<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3316</p>
3317
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3318<h5>Arguments:</h5>
3319<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]3320scalar or vector <a href="#t_floating">floating point</a> value, and a type
3321to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3322type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3323vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3324
3325<h5>Semantics:</h5>
3326<p>The '<tt>fptosi</tt>' instruction converts its
3327<a href="#t_floating">floating point</a> operand into the nearest (rounding
3328towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3329the results are undefined.</p>
3330
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3331<h5>Example:</h5>
3332<pre>
3333 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3334 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3335 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3336</pre>
3337</div>
3338
3339<!-- _______________________________________________________________________ -->
3340<div class="doc_subsubsection">
3341 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3342</div>
3343<div class="doc_text">
3344
3345<h5>Syntax:</h5>
3346<pre>
3347 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3348</pre>
3349
3350<h5>Overview:</h5>
3351<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3352integer and converts that value to the <tt>ty2</tt> type.</p>
3353
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3354<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3355<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3356scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3357to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3358type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3359floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3360
3361<h5>Semantics:</h5>
3362<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3363integer quantity and converts it to the corresponding floating point value. If
3364the value cannot fit in the floating point value, the results are undefined.</p>
3365
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3366<h5>Example:</h5>
3367<pre>
3368 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3369 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3370</pre>
3371</div>
3372
3373<!-- _______________________________________________________________________ -->
3374<div class="doc_subsubsection">
3375 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3376</div>
3377<div class="doc_text">
3378
3379<h5>Syntax:</h5>
3380<pre>
3381 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3382</pre>
3383
3384<h5>Overview:</h5>
3385<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3386integer and converts that value to the <tt>ty2</tt> type.</p>
3387
3388<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3389<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3390scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3391to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3392type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3393floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3394
3395<h5>Semantics:</h5>
3396<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3397integer quantity and converts it to the corresponding floating point value. If
3398the value cannot fit in the floating point value, the results are undefined.</p>
3399
3400<h5>Example:</h5>
3401<pre>
3402 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3403 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3404</pre>
3405</div>
3406
3407<!-- _______________________________________________________________________ -->
3408<div class="doc_subsubsection">
3409 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3410</div>
3411<div class="doc_text">
3412
3413<h5>Syntax:</h5>
3414<pre>
3415 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3416</pre>
3417
3418<h5>Overview:</h5>
3419<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3420the integer type <tt>ty2</tt>.</p>
3421
3422<h5>Arguments:</h5>
3423<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3424must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3425<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3426
3427<h5>Semantics:</h5>
3428<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3429<tt>ty2</tt> by interpreting the pointer value as an integer and either
3430truncating or zero extending that value to the size of the integer type. If
3431<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3432<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3433are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3434change.</p>
3435
3436<h5>Example:</h5>
3437<pre>
3438 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3439 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3440</pre>
3441</div>
3442
3443<!-- _______________________________________________________________________ -->
3444<div class="doc_subsubsection">
3445 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3446</div>
3447<div class="doc_text">
3448
3449<h5>Syntax:</h5>
3450<pre>
3451 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3452</pre>
3453
3454<h5>Overview:</h5>
3455<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3456a pointer type, <tt>ty2</tt>.</p>
3457
3458<h5>Arguments:</h5>
3459<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3460value to cast, and a type to cast it to, which must be a
3461<a href="#t_pointer">pointer</a> type.
3462
3463<h5>Semantics:</h5>
3464<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3465<tt>ty2</tt> by applying either a zero extension or a truncation depending on
3466the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3467size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3468the size of a pointer then a zero extension is done. If they are the same size,
3469nothing is done (<i>no-op cast</i>).</p>
3470
3471<h5>Example:</h5>
3472<pre>
3473 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3474 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3475 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3476</pre>
3477</div>
3478
3479<!-- _______________________________________________________________________ -->
3480<div class="doc_subsubsection">
3481 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3482</div>
3483<div class="doc_text">
3484
3485<h5>Syntax:</h5>
3486<pre>
3487 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3488</pre>
3489
3490<h5>Overview:</h5>
3491<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3492<tt>ty2</tt> without changing any bits.</p>
3493
3494<h5>Arguments:</h5>
3495<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3496a first class value, and a type to cast it to, which must also be a <a
3497 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3498and the destination type, <tt>ty2</tt>, must be identical. If the source
3499type is a pointer, the destination type must also be a pointer.</p>
3500
3501<h5>Semantics:</h5>
3502<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3503<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3504this conversion. The conversion is done as if the <tt>value</tt> had been
3505stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3506converted to other pointer types with this instruction. To convert pointers to
3507other types, use the <a href="#i_inttoptr">inttoptr</a> or
3508<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3509
3510<h5>Example:</h5>
3511<pre>
3512 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3513 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3514 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3515</pre>
3516</div>
3517
3518<!-- ======================================================================= -->
3519<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3520<div class="doc_text">
3521<p>The instructions in this category are the "miscellaneous"
3522instructions, which defy better classification.</p>
3523</div>
3524
3525<!-- _______________________________________________________________________ -->
3526<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3527</div>
3528<div class="doc_text">
3529<h5>Syntax:</h5>
3530<pre> &lt;result&gt; = icmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3531</pre>
3532<h5>Overview:</h5>
3533<p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3534of its two integer operands.</p>
3535<h5>Arguments:</h5>
3536<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3537the condition code indicating the kind of comparison to perform. It is not
3538a value, just a keyword. The possible condition code are:
3539<ol>
3540 <li><tt>eq</tt>: equal</li>
3541 <li><tt>ne</tt>: not equal </li>
3542 <li><tt>ugt</tt>: unsigned greater than</li>
3543 <li><tt>uge</tt>: unsigned greater or equal</li>
3544 <li><tt>ult</tt>: unsigned less than</li>
3545 <li><tt>ule</tt>: unsigned less or equal</li>
3546 <li><tt>sgt</tt>: signed greater than</li>
3547 <li><tt>sge</tt>: signed greater or equal</li>
3548 <li><tt>slt</tt>: signed less than</li>
3549 <li><tt>sle</tt>: signed less or equal</li>
3550</ol>
3551<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3552<a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3553<h5>Semantics:</h5>
3554<p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3555the condition code given as <tt>cond</tt>. The comparison performed always
3556yields a <a href="#t_primitive">i1</a> result, as follows:
3557<ol>
3558 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3559 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3560 </li>
3561 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3562 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3563 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3564 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3565 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3566 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3567 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3568 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3569 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3570 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3571 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3572 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3573 <li><tt>sge</tt>: interprets the operands as signed values and yields
3574 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3575 <li><tt>slt</tt>: interprets the operands as signed values and yields
3576 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3577 <li><tt>sle</tt>: interprets the operands as signed values and yields
3578 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3579</ol>
3580<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3581values are compared as if they were integers.</p>
3582
3583<h5>Example:</h5>
3584<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
3585 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
3586 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
3587 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3588 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3589 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
3590</pre>
3591</div>
3592
3593<!-- _______________________________________________________________________ -->
3594<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3595</div>
3596<div class="doc_text">
3597<h5>Syntax:</h5>
3598<pre> &lt;result&gt; = fcmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3599</pre>
3600<h5>Overview:</h5>
3601<p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3602of its floating point operands.</p>
3603<h5>Arguments:</h5>
3604<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3605the condition code indicating the kind of comparison to perform. It is not
3606a value, just a keyword. The possible condition code are:
3607<ol>
3608 <li><tt>false</tt>: no comparison, always returns false</li>
3609 <li><tt>oeq</tt>: ordered and equal</li>
3610 <li><tt>ogt</tt>: ordered and greater than </li>
3611 <li><tt>oge</tt>: ordered and greater than or equal</li>
3612 <li><tt>olt</tt>: ordered and less than </li>
3613 <li><tt>ole</tt>: ordered and less than or equal</li>
3614 <li><tt>one</tt>: ordered and not equal</li>
3615 <li><tt>ord</tt>: ordered (no nans)</li>
3616 <li><tt>ueq</tt>: unordered or equal</li>
3617 <li><tt>ugt</tt>: unordered or greater than </li>
3618 <li><tt>uge</tt>: unordered or greater than or equal</li>
3619 <li><tt>ult</tt>: unordered or less than </li>
3620 <li><tt>ule</tt>: unordered or less than or equal</li>
3621 <li><tt>une</tt>: unordered or not equal</li>
3622 <li><tt>uno</tt>: unordered (either nans)</li>
3623 <li><tt>true</tt>: no comparison, always returns true</li>
3624</ol>
3625<p><i>Ordered</i> means that neither operand is a QNAN while
3626<i>unordered</i> means that either operand may be a QNAN.</p>
3627<p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3628<a href="#t_floating">floating point</a> typed. They must have identical
3629types.</p>
3630<h5>Semantics:</h5>
3631<p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3632the condition code given as <tt>cond</tt>. The comparison performed always
3633yields a <a href="#t_primitive">i1</a> result, as follows:
3634<ol>
3635 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3636 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3637 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3638 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3639 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3640 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3641 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3642 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3643 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3644 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3645 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3646 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3647 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3648 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3649 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3650 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3651 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3652 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3653 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3654 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3655 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3656 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3657 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3658 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3659 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3660 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3661 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3662 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3663</ol>
3664
3665<h5>Example:</h5>
3666<pre> &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3667 &lt;result&gt; = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3668 &lt;result&gt; = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3669 &lt;result&gt; = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3670</pre>
3671</div>
3672
3673<!-- _______________________________________________________________________ -->
3674<div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3675Instruction</a> </div>
3676<div class="doc_text">
3677<h5>Syntax:</h5>
3678<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
3679<h5>Overview:</h5>
3680<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
3681the SSA graph representing the function.</p>
3682<h5>Arguments:</h5>
3683<p>The type of the incoming values is specified with the first type
3684field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3685as arguments, with one pair for each predecessor basic block of the
3686current block. Only values of <a href="#t_firstclass">first class</a>
3687type may be used as the value arguments to the PHI node. Only labels
3688may be used as the label arguments.</p>
3689<p>There must be no non-phi instructions between the start of a basic
3690block and the PHI instructions: i.e. PHI instructions must be first in
3691a basic block.</p>
3692<h5>Semantics:</h5>
3693<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3694specified by the pair corresponding to the predecessor basic block that executed
3695just prior to the current block.</p>
3696<h5>Example:</h5>
3697<pre>Loop: ; Infinite loop that counts from 0 on up...<br> %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]<br> %nextindvar = add i32 %indvar, 1<br> br label %Loop<br></pre>
3698</div>
3699
3700<!-- _______________________________________________________________________ -->
3701<div class="doc_subsubsection">
3702 <a name="i_select">'<tt>select</tt>' Instruction</a>
3703</div>
3704
3705<div class="doc_text">
3706
3707<h5>Syntax:</h5>
3708
3709<pre>
3710 &lt;result&gt; = select i1 &lt;cond&gt;, &lt;ty&gt; &lt;val1&gt;, &lt;ty&gt; &lt;val2&gt; <i>; yields ty</i>
3711</pre>
3712
3713<h5>Overview:</h5>
3714
3715<p>
3716The '<tt>select</tt>' instruction is used to choose one value based on a
3717condition, without branching.
3718</p>
3719
3720
3721<h5>Arguments:</h5>
3722
3723<p>
3724The '<tt>select</tt>' instruction requires a boolean value indicating the condition, and two values of the same <a href="#t_firstclass">first class</a> type.
3725</p>
3726
3727<h5>Semantics:</h5>
3728
3729<p>
3730If the boolean condition evaluates to true, the instruction returns the first
3731value argument; otherwise, it returns the second value argument.
3732</p>
3733
3734<h5>Example:</h5>
3735
3736<pre>
3737 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3738</pre>
3739</div>
3740
3741
3742<!-- _______________________________________________________________________ -->
3743<div class="doc_subsubsection">
3744 <a name="i_call">'<tt>call</tt>' Instruction</a>
3745</div>
3746
3747<div class="doc_text">
3748
3749<h5>Syntax:</h5>
3750<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3751 &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]3752</pre>
3753
3754<h5>Overview:</h5>
3755
3756<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3757
3758<h5>Arguments:</h5>
3759
3760<p>This instruction requires several arguments:</p>
3761
3762<ol>
3763 <li>
3764 <p>The optional "tail" marker indicates whether the callee function accesses
3765 any allocas or varargs in the caller. If the "tail" marker is present, the
3766 function call is eligible for tail call optimization. Note that calls may
3767 be marked "tail" even if they do not occur before a <a
3768 href="#i_ret"><tt>ret</tt></a> instruction.
3769 </li>
3770 <li>
3771 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3772 convention</a> the call should use. If none is specified, the call defaults
3773 to using C calling conventions.
3774 </li>
3775 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3776 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
3777 the type of the return value. Functions that return no value are marked
3778 <tt><a href="#t_void">void</a></tt>.</p>
3779 </li>
3780 <li>
3781 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
3782 value being invoked. The argument types must match the types implied by
3783 this signature. This type can be omitted if the function is not varargs
3784 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3785 </li>
3786 <li>
3787 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3788 be invoked. In most cases, this is a direct function invocation, but
3789 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3790 to function value.</p>
3791 </li>
3792 <li>
3793 <p>'<tt>function args</tt>': argument list whose types match the
3794 function signature argument types. All arguments must be of
3795 <a href="#t_firstclass">first class</a> type. If the function signature
3796 indicates the function accepts a variable number of arguments, the extra
3797 arguments can be specified.</p>
3798 </li>
3799</ol>
3800
3801<h5>Semantics:</h5>
3802
3803<p>The '<tt>call</tt>' instruction is used to cause control flow to
3804transfer to a specified function, with its incoming arguments bound to
3805the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3806instruction in the called function, control flow continues with the
3807instruction after the function call, and the return value of the
3808function is bound to the result argument. This is a simpler case of
3809the <a href="#i_invoke">invoke</a> instruction.</p>
3810
3811<h5>Example:</h5>
3812
3813<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3814 %retval = call i32 @test(i32 %argc)
3815 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42);
3816 %X = tail call i32 @foo()
3817 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo()
3818 %Z = call void %foo(i8 97 signext)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3819</pre>
3820
3821</div>
3822
3823<!-- _______________________________________________________________________ -->
3824<div class="doc_subsubsection">
3825 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3826</div>
3827
3828<div class="doc_text">
3829
3830<h5>Syntax:</h5>
3831
3832<pre>
3833 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
3834</pre>
3835
3836<h5>Overview:</h5>
3837
3838<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3839the "variable argument" area of a function call. It is used to implement the
3840<tt>va_arg</tt> macro in C.</p>
3841
3842<h5>Arguments:</h5>
3843
3844<p>This instruction takes a <tt>va_list*</tt> value and the type of
3845the argument. It returns a value of the specified argument type and
3846increments the <tt>va_list</tt> to point to the next argument. The
3847actual type of <tt>va_list</tt> is target specific.</p>
3848
3849<h5>Semantics:</h5>
3850
3851<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3852type from the specified <tt>va_list</tt> and causes the
3853<tt>va_list</tt> to point to the next argument. For more information,
3854see the variable argument handling <a href="#int_varargs">Intrinsic
3855Functions</a>.</p>
3856
3857<p>It is legal for this instruction to be called in a function which does not
3858take a variable number of arguments, for example, the <tt>vfprintf</tt>
3859function.</p>
3860
3861<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3862href="#intrinsics">intrinsic function</a> because it takes a type as an
3863argument.</p>
3864
3865<h5>Example:</h5>
3866
3867<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3868
3869</div>
3870
3871<!-- *********************************************************************** -->
3872<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3873<!-- *********************************************************************** -->
3874
3875<div class="doc_text">
3876
3877<p>LLVM supports the notion of an "intrinsic function". These functions have
3878well known names and semantics and are required to follow certain restrictions.
3879Overall, these intrinsics represent an extension mechanism for the LLVM
3880language that does not require changing all of the transformations in LLVM when
3881adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
3882
3883<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3884prefix is reserved in LLVM for intrinsic names; thus, function names may not
3885begin with this prefix. Intrinsic functions must always be external functions:
3886you cannot define the body of intrinsic functions. Intrinsic functions may
3887only be used in call or invoke instructions: it is illegal to take the address
3888of an intrinsic function. Additionally, because intrinsic functions are part
3889of the LLVM language, it is required if any are added that they be documented
3890here.</p>
3891
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]3892<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
3893a family of functions that perform the same operation but on different data
3894types. Because LLVM can represent over 8 million different integer types,
3895overloading is used commonly to allow an intrinsic function to operate on any
3896integer type. One or more of the argument types or the result type can be
3897overloaded to accept any integer type. Argument types may also be defined as
3898exactly matching a previous argument's type or the result type. This allows an
3899intrinsic function which accepts multiple arguments, but needs all of them to
3900be of the same type, to only be overloaded with respect to a single argument or
3901the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3902
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]3903<p>Overloaded intrinsics will have the names of its overloaded argument types
3904encoded into its function name, each preceded by a period. Only those types
3905which are overloaded result in a name suffix. Arguments whose type is matched
3906against another type do not. For example, the <tt>llvm.ctpop</tt> function can
3907take an integer of any width and returns an integer of exactly the same integer
3908width. This leads to a family of functions such as
3909<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
3910Only one type, the return type, is overloaded, and only one type suffix is
3911required. Because the argument's type is matched against the return type, it
3912does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3913
3914<p>To learn how to add an intrinsic function, please see the
3915<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3916</p>
3917
3918</div>
3919
3920<!-- ======================================================================= -->
3921<div class="doc_subsection">
3922 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3923</div>
3924
3925<div class="doc_text">
3926
3927<p>Variable argument support is defined in LLVM with the <a
3928 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3929intrinsic functions. These functions are related to the similarly
3930named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
3931
3932<p>All of these functions operate on arguments that use a
3933target-specific value type "<tt>va_list</tt>". The LLVM assembly
3934language reference manual does not define what this type is, so all
3935transformations should be prepared to handle these functions regardless of
3936the type used.</p>
3937
3938<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3939instruction and the variable argument handling intrinsic functions are
3940used.</p>
3941
3942<div class="doc_code">
3943<pre>
3944define i32 @test(i32 %X, ...) {
3945 ; Initialize variable argument processing
3946 %ap = alloca i8*
3947 %ap2 = bitcast i8** %ap to i8*
3948 call void @llvm.va_start(i8* %ap2)
3949
3950 ; Read a single integer argument
3951 %tmp = va_arg i8** %ap, i32
3952
3953 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3954 %aq = alloca i8*
3955 %aq2 = bitcast i8** %aq to i8*
3956 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
3957 call void @llvm.va_end(i8* %aq2)
3958
3959 ; Stop processing of arguments.
3960 call void @llvm.va_end(i8* %ap2)
3961 ret i32 %tmp
3962}
3963
3964declare void @llvm.va_start(i8*)
3965declare void @llvm.va_copy(i8*, i8*)
3966declare void @llvm.va_end(i8*)
3967</pre>
3968</div>
3969
3970</div>
3971
3972<!-- _______________________________________________________________________ -->
3973<div class="doc_subsubsection">
3974 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3975</div>
3976
3977
3978<div class="doc_text">
3979<h5>Syntax:</h5>
3980<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
3981<h5>Overview:</h5>
3982<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3983<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
3984href="#i_va_arg">va_arg</a></tt>.</p>
3985
3986<h5>Arguments:</h5>
3987
3988<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3989
3990<h5>Semantics:</h5>
3991
3992<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3993macro available in C. In a target-dependent way, it initializes the
3994<tt>va_list</tt> element to which the argument points, so that the next call to
3995<tt>va_arg</tt> will produce the first variable argument passed to the function.
3996Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3997last argument of the function as the compiler can figure that out.</p>
3998
3999</div>
4000
4001<!-- _______________________________________________________________________ -->
4002<div class="doc_subsubsection">
4003 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4004</div>
4005
4006<div class="doc_text">
4007<h5>Syntax:</h5>
4008<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
4009<h5>Overview:</h5>
4010
4011<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
4012which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4013or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4014
4015<h5>Arguments:</h5>
4016
4017<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4018
4019<h5>Semantics:</h5>
4020
4021<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4022macro available in C. In a target-dependent way, it destroys the
4023<tt>va_list</tt> element to which the argument points. Calls to <a
4024href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4025<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4026<tt>llvm.va_end</tt>.</p>
4027
4028</div>
4029
4030<!-- _______________________________________________________________________ -->
4031<div class="doc_subsubsection">
4032 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4033</div>
4034
4035<div class="doc_text">
4036
4037<h5>Syntax:</h5>
4038
4039<pre>
4040 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
4041</pre>
4042
4043<h5>Overview:</h5>
4044
4045<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4046from the source argument list to the destination argument list.</p>
4047
4048<h5>Arguments:</h5>
4049
4050<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4051The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4052
4053
4054<h5>Semantics:</h5>
4055
4056<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4057macro available in C. In a target-dependent way, it copies the source
4058<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4059intrinsic is necessary because the <tt><a href="#int_va_start">
4060llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4061example, memory allocation.</p>
4062
4063</div>
4064
4065<!-- ======================================================================= -->
4066<div class="doc_subsection">
4067 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4068</div>
4069
4070<div class="doc_text">
4071
4072<p>
4073LLVM support for <a href="GarbageCollection.html">Accurate Garbage
4074Collection</a> requires the implementation and generation of these intrinsics.
4075These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4076stack</a>, as well as garbage collector implementations that require <a
4077href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4078Front-ends for type-safe garbage collected languages should generate these
4079intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4080href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4081</p>
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]4082
4083<p>The garbage collection intrinsics only operate on objects in the generic
4084 address space (address space zero).</p>
4085
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4086</div>
4087
4088<!-- _______________________________________________________________________ -->
4089<div class="doc_subsubsection">
4090 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4091</div>
4092
4093<div class="doc_text">
4094
4095<h5>Syntax:</h5>
4096
4097<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4098 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4099</pre>
4100
4101<h5>Overview:</h5>
4102
4103<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4104the code generator, and allows some metadata to be associated with it.</p>
4105
4106<h5>Arguments:</h5>
4107
4108<p>The first argument specifies the address of a stack object that contains the
4109root pointer. The second pointer (which must be either a constant or a global
4110value address) contains the meta-data to be associated with the root.</p>
4111
4112<h5>Semantics:</h5>
4113
4114<p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
4115location. At compile-time, the code generator generates information to allow
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4116the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4117intrinsic may only be used in a function which <a href="#gc">specifies a GC
4118algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4119
4120</div>
4121
4122
4123<!-- _______________________________________________________________________ -->
4124<div class="doc_subsubsection">
4125 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4126</div>
4127
4128<div class="doc_text">
4129
4130<h5>Syntax:</h5>
4131
4132<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4133 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4134</pre>
4135
4136<h5>Overview:</h5>
4137
4138<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4139locations, allowing garbage collector implementations that require read
4140barriers.</p>
4141
4142<h5>Arguments:</h5>
4143
4144<p>The second argument is the address to read from, which should be an address
4145allocated from the garbage collector. The first object is a pointer to the
4146start of the referenced object, if needed by the language runtime (otherwise
4147null).</p>
4148
4149<h5>Semantics:</h5>
4150
4151<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4152instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4153garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4154may only be used in a function which <a href="#gc">specifies a GC
4155algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4156
4157</div>
4158
4159
4160<!-- _______________________________________________________________________ -->
4161<div class="doc_subsubsection">
4162 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4163</div>
4164
4165<div class="doc_text">
4166
4167<h5>Syntax:</h5>
4168
4169<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4170 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4171</pre>
4172
4173<h5>Overview:</h5>
4174
4175<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4176locations, allowing garbage collector implementations that require write
4177barriers (such as generational or reference counting collectors).</p>
4178
4179<h5>Arguments:</h5>
4180
4181<p>The first argument is the reference to store, the second is the start of the
4182object to store it to, and the third is the address of the field of Obj to
4183store to. If the runtime does not require a pointer to the object, Obj may be
4184null.</p>
4185
4186<h5>Semantics:</h5>
4187
4188<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4189instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4190garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4191may only be used in a function which <a href="#gc">specifies a GC
4192algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4193
4194</div>
4195
4196
4197
4198<!-- ======================================================================= -->
4199<div class="doc_subsection">
4200 <a name="int_codegen">Code Generator Intrinsics</a>
4201</div>
4202
4203<div class="doc_text">
4204<p>
4205These intrinsics are provided by LLVM to expose special features that may only
4206be implemented with code generator support.
4207</p>
4208
4209</div>
4210
4211<!-- _______________________________________________________________________ -->
4212<div class="doc_subsubsection">
4213 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4214</div>
4215
4216<div class="doc_text">
4217
4218<h5>Syntax:</h5>
4219<pre>
4220 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
4221</pre>
4222
4223<h5>Overview:</h5>
4224
4225<p>
4226The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4227target-specific value indicating the return address of the current function
4228or one of its callers.
4229</p>
4230
4231<h5>Arguments:</h5>
4232
4233<p>
4234The argument to this intrinsic indicates which function to return the address
4235for. Zero indicates the calling function, one indicates its caller, etc. The
4236argument is <b>required</b> to be a constant integer value.
4237</p>
4238
4239<h5>Semantics:</h5>
4240
4241<p>
4242The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4243the return address of the specified call frame, or zero if it cannot be
4244identified. The value returned by this intrinsic is likely to be incorrect or 0
4245for arguments other than zero, so it should only be used for debugging purposes.
4246</p>
4247
4248<p>
4249Note that calling this intrinsic does not prevent function inlining or other
4250aggressive transformations, so the value returned may not be that of the obvious
4251source-language caller.
4252</p>
4253</div>
4254
4255
4256<!-- _______________________________________________________________________ -->
4257<div class="doc_subsubsection">
4258 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4259</div>
4260
4261<div class="doc_text">
4262
4263<h5>Syntax:</h5>
4264<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4265 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4266</pre>
4267
4268<h5>Overview:</h5>
4269
4270<p>
4271The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4272target-specific frame pointer value for the specified stack frame.
4273</p>
4274
4275<h5>Arguments:</h5>
4276
4277<p>
4278The argument to this intrinsic indicates which function to return the frame
4279pointer for. Zero indicates the calling function, one indicates its caller,
4280etc. The argument is <b>required</b> to be a constant integer value.
4281</p>
4282
4283<h5>Semantics:</h5>
4284
4285<p>
4286The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4287the frame address of the specified call frame, or zero if it cannot be
4288identified. The value returned by this intrinsic is likely to be incorrect or 0
4289for arguments other than zero, so it should only be used for debugging purposes.
4290</p>
4291
4292<p>
4293Note that calling this intrinsic does not prevent function inlining or other
4294aggressive transformations, so the value returned may not be that of the obvious
4295source-language caller.
4296</p>
4297</div>
4298
4299<!-- _______________________________________________________________________ -->
4300<div class="doc_subsubsection">
4301 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4302</div>
4303
4304<div class="doc_text">
4305
4306<h5>Syntax:</h5>
4307<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4308 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4309</pre>
4310
4311<h5>Overview:</h5>
4312
4313<p>
4314The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4315the function stack, for use with <a href="#int_stackrestore">
4316<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4317features like scoped automatic variable sized arrays in C99.
4318</p>
4319
4320<h5>Semantics:</h5>
4321
4322<p>
4323This intrinsic returns a opaque pointer value that can be passed to <a
4324href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4325<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4326<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4327state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4328practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4329that were allocated after the <tt>llvm.stacksave</tt> was executed.
4330</p>
4331
4332</div>
4333
4334<!-- _______________________________________________________________________ -->
4335<div class="doc_subsubsection">
4336 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4337</div>
4338
4339<div class="doc_text">
4340
4341<h5>Syntax:</h5>
4342<pre>
4343 declare void @llvm.stackrestore(i8 * %ptr)
4344</pre>
4345
4346<h5>Overview:</h5>
4347
4348<p>
4349The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4350the function stack to the state it was in when the corresponding <a
4351href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4352useful for implementing language features like scoped automatic variable sized
4353arrays in C99.
4354</p>
4355
4356<h5>Semantics:</h5>
4357
4358<p>
4359See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4360</p>
4361
4362</div>
4363
4364
4365<!-- _______________________________________________________________________ -->
4366<div class="doc_subsubsection">
4367 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4368</div>
4369
4370<div class="doc_text">
4371
4372<h5>Syntax:</h5>
4373<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4374 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]4375</pre>
4376
4377<h5>Overview:</h5>
4378
4379
4380<p>
4381The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4382a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4383no
4384effect on the behavior of the program but can change its performance
4385characteristics.
4386</p>
4387
4388<h5>Arguments:</h5>
4389
4390<p>
4391<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4392determining if the fetch should be for a read (0) or write (1), and
4393<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4394locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4395<tt>locality</tt> arguments must be constant integers.
4396</p>
4397
4398<h5>Semantics:</h5>
4399
4400<p>
4401This intrinsic does not modify the behavior of the program. In particular,
4402prefetches cannot trap and do not produce a value. On targets that support this
4403intrinsic, the prefetch can provide hints to the processor cache for better
4404performance.
4405</p>
4406
4407</div>
4408
4409<!-- _______________________________________________________________________ -->
4410<div class="doc_subsubsection">
4411 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4412</div>
4413
4414<div class="doc_text">
4415
4416<h5>Syntax:</h5>
4417<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4418 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4419</pre>
4420
4421<h5>Overview:</h5>
4422
4423
4424<p>
4425The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4426(PC) in a region of
4427code to simulators and other tools. The method is target specific, but it is
4428expected that the marker will use exported symbols to transmit the PC of the marker.
4429The marker makes no guarantees that it will remain with any specific instruction
4430after optimizations. It is possible that the presence of a marker will inhibit
4431optimizations. The intended use is to be inserted after optimizations to allow
4432correlations of simulation runs.
4433</p>
4434
4435<h5>Arguments:</h5>
4436
4437<p>
4438<tt>id</tt> is a numerical id identifying the marker.
4439</p>
4440
4441<h5>Semantics:</h5>
4442
4443<p>
4444This intrinsic does not modify the behavior of the program. Backends that do not
4445support this intrinisic may ignore it.
4446</p>
4447
4448</div>
4449
4450<!-- _______________________________________________________________________ -->
4451<div class="doc_subsubsection">
4452 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4453</div>
4454
4455<div class="doc_text">
4456
4457<h5>Syntax:</h5>
4458<pre>
4459 declare i64 @llvm.readcyclecounter( )
4460</pre>
4461
4462<h5>Overview:</h5>
4463
4464
4465<p>
4466The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4467counter register (or similar low latency, high accuracy clocks) on those targets
4468that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4469As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4470should only be used for small timings.
4471</p>
4472
4473<h5>Semantics:</h5>
4474
4475<p>
4476When directly supported, reading the cycle counter should not modify any memory.
4477Implementations are allowed to either return a application specific value or a
4478system wide value. On backends without support, this is lowered to a constant 0.
4479</p>
4480
4481</div>
4482
4483<!-- ======================================================================= -->
4484<div class="doc_subsection">
4485 <a name="int_libc">Standard C Library Intrinsics</a>
4486</div>
4487
4488<div class="doc_text">
4489<p>
4490LLVM provides intrinsics for a few important standard C library functions.
4491These intrinsics allow source-language front-ends to pass information about the
4492alignment of the pointer arguments to the code generator, providing opportunity
4493for more efficient code generation.
4494</p>
4495
4496</div>
4497
4498<!-- _______________________________________________________________________ -->
4499<div class="doc_subsubsection">
4500 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4501</div>
4502
4503<div class="doc_text">
4504
4505<h5>Syntax:</h5>
4506<pre>
4507 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4508 i32 &lt;len&gt;, i32 &lt;align&gt;)
4509 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4510 i64 &lt;len&gt;, i32 &lt;align&gt;)
4511</pre>
4512
4513<h5>Overview:</h5>
4514
4515<p>
4516The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4517location to the destination location.
4518</p>
4519
4520<p>
4521Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4522intrinsics do not return a value, and takes an extra alignment argument.
4523</p>
4524
4525<h5>Arguments:</h5>
4526
4527<p>
4528The first argument is a pointer to the destination, the second is a pointer to
4529the source. The third argument is an integer argument
4530specifying the number of bytes to copy, and the fourth argument is the alignment
4531of the source and destination locations.
4532</p>
4533
4534<p>
4535If the call to this intrinisic has an alignment value that is not 0 or 1, then
4536the caller guarantees that both the source and destination pointers are aligned
4537to that boundary.
4538</p>
4539
4540<h5>Semantics:</h5>
4541
4542<p>
4543The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4544location to the destination location, which are not allowed to overlap. It
4545copies "len" bytes of memory over. If the argument is known to be aligned to
4546some boundary, this can be specified as the fourth argument, otherwise it should
4547be set to 0 or 1.
4548</p>
4549</div>
4550
4551
4552<!-- _______________________________________________________________________ -->
4553<div class="doc_subsubsection">
4554 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4555</div>
4556
4557<div class="doc_text">
4558
4559<h5>Syntax:</h5>
4560<pre>
4561 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4562 i32 &lt;len&gt;, i32 &lt;align&gt;)
4563 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4564 i64 &lt;len&gt;, i32 &lt;align&gt;)
4565</pre>
4566
4567<h5>Overview:</h5>
4568
4569<p>
4570The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4571location to the destination location. It is similar to the
Chris Lattnerdba16ea2008-01-06 19:51:52 +0000[diff] [blame]4572'<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4573</p>
4574
4575<p>
4576Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4577intrinsics do not return a value, and takes an extra alignment argument.
4578</p>
4579
4580<h5>Arguments:</h5>
4581
4582<p>
4583The first argument is a pointer to the destination, the second is a pointer to
4584the source. The third argument is an integer argument
4585specifying the number of bytes to copy, and the fourth argument is the alignment
4586of the source and destination locations.
4587</p>
4588
4589<p>
4590If the call to this intrinisic has an alignment value that is not 0 or 1, then
4591the caller guarantees that the source and destination pointers are aligned to
4592that boundary.
4593</p>
4594
4595<h5>Semantics:</h5>
4596
4597<p>
4598The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4599location to the destination location, which may overlap. It
4600copies "len" bytes of memory over. If the argument is known to be aligned to
4601some boundary, this can be specified as the fourth argument, otherwise it should
4602be set to 0 or 1.
4603</p>
4604</div>
4605
4606
4607<!-- _______________________________________________________________________ -->
4608<div class="doc_subsubsection">
4609 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4610</div>
4611
4612<div class="doc_text">
4613
4614<h5>Syntax:</h5>
4615<pre>
4616 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
4617 i32 &lt;len&gt;, i32 &lt;align&gt;)
4618 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
4619 i64 &lt;len&gt;, i32 &lt;align&gt;)
4620</pre>
4621
4622<h5>Overview:</h5>
4623
4624<p>
4625The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4626byte value.
4627</p>
4628
4629<p>
4630Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4631does not return a value, and takes an extra alignment argument.
4632</p>
4633
4634<h5>Arguments:</h5>
4635
4636<p>
4637The first argument is a pointer to the destination to fill, the second is the
4638byte value to fill it with, the third argument is an integer
4639argument specifying the number of bytes to fill, and the fourth argument is the
4640known alignment of destination location.
4641</p>
4642
4643<p>
4644If the call to this intrinisic has an alignment value that is not 0 or 1, then
4645the caller guarantees that the destination pointer is aligned to that boundary.
4646</p>
4647
4648<h5>Semantics:</h5>
4649
4650<p>
4651The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4652the
4653destination location. If the argument is known to be aligned to some boundary,
4654this can be specified as the fourth argument, otherwise it should be set to 0 or
46551.
4656</p>
4657</div>
4658
4659
4660<!-- _______________________________________________________________________ -->
4661<div class="doc_subsubsection">
4662 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4663</div>
4664
4665<div class="doc_text">
4666
4667<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4668<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]4669floating point or vector of floating point type. Not all targets support all
4670types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4671<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4672 declare float @llvm.sqrt.f32(float %Val)
4673 declare double @llvm.sqrt.f64(double %Val)
4674 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
4675 declare fp128 @llvm.sqrt.f128(fp128 %Val)
4676 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4677</pre>
4678
4679<h5>Overview:</h5>
4680
4681<p>
4682The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4683returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4684<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
Chris Lattnerf914a002008-01-29 07:00:44 +0000[diff] [blame]4685negative numbers other than -0.0 (which allows for better optimization, because
4686there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
4687defined to return -0.0 like IEEE sqrt.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4688</p>
4689
4690<h5>Arguments:</h5>
4691
4692<p>
4693The argument and return value are floating point numbers of the same type.
4694</p>
4695
4696<h5>Semantics:</h5>
4697
4698<p>
4699This function returns the sqrt of the specified operand if it is a nonnegative
4700floating point number.
4701</p>
4702</div>
4703
4704<!-- _______________________________________________________________________ -->
4705<div class="doc_subsubsection">
4706 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4707</div>
4708
4709<div class="doc_text">
4710
4711<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4712<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]4713floating point or vector of floating point type. Not all targets support all
4714types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4715<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4716 declare float @llvm.powi.f32(float %Val, i32 %power)
4717 declare double @llvm.powi.f64(double %Val, i32 %power)
4718 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
4719 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
4720 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4721</pre>
4722
4723<h5>Overview:</h5>
4724
4725<p>
4726The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4727specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4728multiplications is not defined. When a vector of floating point type is
4729used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4730</p>
4731
4732<h5>Arguments:</h5>
4733
4734<p>
4735The second argument is an integer power, and the first is a value to raise to
4736that power.
4737</p>
4738
4739<h5>Semantics:</h5>
4740
4741<p>
4742This function returns the first value raised to the second power with an
4743unspecified sequence of rounding operations.</p>
4744</div>
4745
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4746<!-- _______________________________________________________________________ -->
4747<div class="doc_subsubsection">
4748 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
4749</div>
4750
4751<div class="doc_text">
4752
4753<h5>Syntax:</h5>
4754<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
4755floating point or vector of floating point type. Not all targets support all
4756types however.
4757<pre>
4758 declare float @llvm.sin.f32(float %Val)
4759 declare double @llvm.sin.f64(double %Val)
4760 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
4761 declare fp128 @llvm.sin.f128(fp128 %Val)
4762 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
4763</pre>
4764
4765<h5>Overview:</h5>
4766
4767<p>
4768The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
4769</p>
4770
4771<h5>Arguments:</h5>
4772
4773<p>
4774The argument and return value are floating point numbers of the same type.
4775</p>
4776
4777<h5>Semantics:</h5>
4778
4779<p>
4780This function returns the sine of the specified operand, returning the
4781same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4782conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4783</div>
4784
4785<!-- _______________________________________________________________________ -->
4786<div class="doc_subsubsection">
4787 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
4788</div>
4789
4790<div class="doc_text">
4791
4792<h5>Syntax:</h5>
4793<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
4794floating point or vector of floating point type. Not all targets support all
4795types however.
4796<pre>
4797 declare float @llvm.cos.f32(float %Val)
4798 declare double @llvm.cos.f64(double %Val)
4799 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
4800 declare fp128 @llvm.cos.f128(fp128 %Val)
4801 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
4802</pre>
4803
4804<h5>Overview:</h5>
4805
4806<p>
4807The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
4808</p>
4809
4810<h5>Arguments:</h5>
4811
4812<p>
4813The argument and return value are floating point numbers of the same type.
4814</p>
4815
4816<h5>Semantics:</h5>
4817
4818<p>
4819This function returns the cosine of the specified operand, returning the
4820same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4821conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4822</div>
4823
4824<!-- _______________________________________________________________________ -->
4825<div class="doc_subsubsection">
4826 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
4827</div>
4828
4829<div class="doc_text">
4830
4831<h5>Syntax:</h5>
4832<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
4833floating point or vector of floating point type. Not all targets support all
4834types however.
4835<pre>
4836 declare float @llvm.pow.f32(float %Val, float %Power)
4837 declare double @llvm.pow.f64(double %Val, double %Power)
4838 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
4839 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
4840 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
4841</pre>
4842
4843<h5>Overview:</h5>
4844
4845<p>
4846The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
4847specified (positive or negative) power.
4848</p>
4849
4850<h5>Arguments:</h5>
4851
4852<p>
4853The second argument is a floating point power, and the first is a value to
4854raise to that power.
4855</p>
4856
4857<h5>Semantics:</h5>
4858
4859<p>
4860This function returns the first value raised to the second power,
4861returning the
4862same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4863conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4864</div>
4865
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4866
4867<!-- ======================================================================= -->
4868<div class="doc_subsection">
4869 <a name="int_manip">Bit Manipulation Intrinsics</a>
4870</div>
4871
4872<div class="doc_text">
4873<p>
4874LLVM provides intrinsics for a few important bit manipulation operations.
4875These allow efficient code generation for some algorithms.
4876</p>
4877
4878</div>
4879
4880<!-- _______________________________________________________________________ -->
4881<div class="doc_subsubsection">
4882 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4883</div>
4884
4885<div class="doc_text">
4886
4887<h5>Syntax:</h5>
4888<p>This is an overloaded intrinsic function. You can use bswap on any integer
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4889type that is an even number of bytes (i.e. BitWidth % 16 == 0).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4890<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4891 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
4892 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
4893 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4894</pre>
4895
4896<h5>Overview:</h5>
4897
4898<p>
4899The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
4900values with an even number of bytes (positive multiple of 16 bits). These are
4901useful for performing operations on data that is not in the target's native
4902byte order.
4903</p>
4904
4905<h5>Semantics:</h5>
4906
4907<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4908The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4909and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4910intrinsic returns an i32 value that has the four bytes of the input i32
4911swapped, 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]4912i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
4913<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4914additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4915</p>
4916
4917</div>
4918
4919<!-- _______________________________________________________________________ -->
4920<div class="doc_subsubsection">
4921 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4922</div>
4923
4924<div class="doc_text">
4925
4926<h5>Syntax:</h5>
4927<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4928width. Not all targets support all bit widths however.
4929<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4930 declare i8 @llvm.ctpop.i8 (i8 &lt;src&gt;)
4931 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4932 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4933 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
4934 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4935</pre>
4936
4937<h5>Overview:</h5>
4938
4939<p>
4940The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4941value.
4942</p>
4943
4944<h5>Arguments:</h5>
4945
4946<p>
4947The only argument is the value to be counted. The argument may be of any
4948integer type. The return type must match the argument type.
4949</p>
4950
4951<h5>Semantics:</h5>
4952
4953<p>
4954The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4955</p>
4956</div>
4957
4958<!-- _______________________________________________________________________ -->
4959<div class="doc_subsubsection">
4960 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4961</div>
4962
4963<div class="doc_text">
4964
4965<h5>Syntax:</h5>
4966<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4967integer bit width. Not all targets support all bit widths however.
4968<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4969 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
4970 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4971 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4972 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
4973 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4974</pre>
4975
4976<h5>Overview:</h5>
4977
4978<p>
4979The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4980leading zeros in a variable.
4981</p>
4982
4983<h5>Arguments:</h5>
4984
4985<p>
4986The only argument is the value to be counted. The argument may be of any
4987integer type. The return type must match the argument type.
4988</p>
4989
4990<h5>Semantics:</h5>
4991
4992<p>
4993The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4994in a variable. If the src == 0 then the result is the size in bits of the type
4995of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4996</p>
4997</div>
4998
4999
5000
5001<!-- _______________________________________________________________________ -->
5002<div class="doc_subsubsection">
5003 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5004</div>
5005
5006<div class="doc_text">
5007
5008<h5>Syntax:</h5>
5009<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5010integer bit width. Not all targets support all bit widths however.
5011<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5012 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
5013 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5014 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5015 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
5016 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5017</pre>
5018
5019<h5>Overview:</h5>
5020
5021<p>
5022The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5023trailing zeros.
5024</p>
5025
5026<h5>Arguments:</h5>
5027
5028<p>
5029The only argument is the value to be counted. The argument may be of any
5030integer type. The return type must match the argument type.
5031</p>
5032
5033<h5>Semantics:</h5>
5034
5035<p>
5036The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5037in a variable. If the src == 0 then the result is the size in bits of the type
5038of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5039</p>
5040</div>
5041
5042<!-- _______________________________________________________________________ -->
5043<div class="doc_subsubsection">
5044 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5045</div>
5046
5047<div class="doc_text">
5048
5049<h5>Syntax:</h5>
5050<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
5051on any integer bit width.
5052<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5053 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5054 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5055</pre>
5056
5057<h5>Overview:</h5>
5058<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5059range of bits from an integer value and returns them in the same bit width as
5060the original value.</p>
5061
5062<h5>Arguments:</h5>
5063<p>The first argument, <tt>%val</tt> and the result may be integer types of
5064any bit width but they must have the same bit width. The second and third
5065arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5066
5067<h5>Semantics:</h5>
5068<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5069of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5070<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5071operates in forward mode.</p>
5072<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5073right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5074only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5075<ol>
5076 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5077 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5078 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5079 to determine the number of bits to retain.</li>
5080 <li>A mask of the retained bits is created by shifting a -1 value.</li>
5081 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
5082</ol>
5083<p>In reverse mode, a similar computation is made except that the bits are
5084returned in the reverse order. So, for example, if <tt>X</tt> has the value
5085<tt>i16 0x0ACF (101011001111)</tt> and we apply
5086<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5087<tt>i16 0x0026 (000000100110)</tt>.</p>
5088</div>
5089
5090<div class="doc_subsubsection">
5091 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5092</div>
5093
5094<div class="doc_text">
5095
5096<h5>Syntax:</h5>
5097<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
5098on any integer bit width.
5099<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5100 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5101 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]5102</pre>
5103
5104<h5>Overview:</h5>
5105<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5106of bits in an integer value with another integer value. It returns the integer
5107with the replaced bits.</p>
5108
5109<h5>Arguments:</h5>
5110<p>The first argument, <tt>%val</tt> and the result may be integer types of
5111any bit width but they must have the same bit width. <tt>%val</tt> is the value
5112whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5113integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5114type since they specify only a bit index.</p>
5115
5116<h5>Semantics:</h5>
5117<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5118of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5119<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5120operates in forward mode.</p>
5121<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5122truncating it down to the size of the replacement area or zero extending it
5123up to that size.</p>
5124<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5125are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5126in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5127to the <tt>%hi</tt>th bit.
5128<p>In reverse mode, a similar computation is made except that the bits are
5129reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5130<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
5131<h5>Examples:</h5>
5132<pre>
5133 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
5134 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
5135 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
5136 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
5137 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
5138</pre>
5139</div>
5140
5141<!-- ======================================================================= -->
5142<div class="doc_subsection">
5143 <a name="int_debugger">Debugger Intrinsics</a>
5144</div>
5145
5146<div class="doc_text">
5147<p>
5148The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5149are described in the <a
5150href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5151Debugging</a> document.
5152</p>
5153</div>
5154
5155
5156<!-- ======================================================================= -->
5157<div class="doc_subsection">
5158 <a name="int_eh">Exception Handling Intrinsics</a>
5159</div>
5160
5161<div class="doc_text">
5162<p> The LLVM exception handling intrinsics (which all start with
5163<tt>llvm.eh.</tt> prefix), are described in the <a
5164href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5165Handling</a> document. </p>
5166</div>
5167
5168<!-- ======================================================================= -->
5169<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5170 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5171</div>
5172
5173<div class="doc_text">
5174<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5175 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5176 the <tt>nest</tt> attribute, from a function. The result is a callable
5177 function pointer lacking the nest parameter - the caller does not need
5178 to provide a value for it. Instead, the value to use is stored in
5179 advance in a "trampoline", a block of memory usually allocated
5180 on the stack, which also contains code to splice the nest value into the
5181 argument list. This is used to implement the GCC nested function address
5182 extension.
5183</p>
5184<p>
5185 For example, if the function is
5186 <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]5187 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]5188<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5189 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5190 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5191 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5192 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5193</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]5194 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5195 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5196</div>
5197
5198<!-- _______________________________________________________________________ -->
5199<div class="doc_subsubsection">
5200 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5201</div>
5202<div class="doc_text">
5203<h5>Syntax:</h5>
5204<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5205declare 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]5206</pre>
5207<h5>Overview:</h5>
5208<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5209 This fills the memory pointed to by <tt>tramp</tt> with code
5210 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5211</p>
5212<h5>Arguments:</h5>
5213<p>
5214 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5215 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5216 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]5217 intrinsic. Note that the size and the alignment are target-specific - LLVM
5218 currently provides no portable way of determining them, so a front-end that
5219 generates this intrinsic needs to have some target-specific knowledge.
5220 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]5221</p>
5222<h5>Semantics:</h5>
5223<p>
5224 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5225 dependent code, turning it into a function. A pointer to this function is
5226 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5227 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5228 before being called. The new function's signature is the same as that of
5229 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5230 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5231 of pointer type. Calling the new function is equivalent to calling
5232 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5233 missing <tt>nest</tt> argument. If, after calling
5234 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5235 modified, then the effect of any later call to the returned function pointer is
5236 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5237</p>
5238</div>
5239
5240<!-- ======================================================================= -->
5241<div class="doc_subsection">
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame^]5242 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
5243</div>
5244
5245<div class="doc_text">
5246<p>
5247 These intrinsic functions expand the "universal IR" of LLVM to represent
5248 hardware constructs for atomic operations and memory synchronization. This
5249 provides an interface to the hardware, not an interface to the programmer. It
5250 is aimed at a low enough level to allow any programming models or APIs which
5251 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
5252 hardware behavior. Just as hardware provides a "universal IR" for source
5253 languages, it also provides a starting point for developing a "universal"
5254 atomic operation and synchronization IR.
5255</p>
5256<p>
5257 These do <em>not</em> form an API such as high-level threading libraries,
5258 software transaction memory systems, atomic primitives, and intrinsic
5259 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
5260 application libraries. The hardware interface provided by LLVM should allow
5261 a clean implementation of all of these APIs and parallel programming models.
5262 No one model or paradigm should be selected above others unless the hardware
5263 itself ubiquitously does so.
5264
5265</p>
5266</div>
5267
5268<!-- _______________________________________________________________________ -->
5269<div class="doc_subsubsection">
5270 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
5271</div>
5272<div class="doc_text">
5273<h5>Syntax:</h5>
5274<pre>
5275declare void @llvm.memory.barrier( i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt;,
5276i1 &lt;device&gt; )
5277
5278</pre>
5279<h5>Overview:</h5>
5280<p>
5281 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
5282 specific pairs of memory access types.
5283</p>
5284<h5>Arguments:</h5>
5285<p>
5286 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
5287 The first four arguments enables a specific barrier as listed below. The fith
5288 argument specifies that the barrier applies to io or device or uncached memory.
5289
5290</p>
5291 <ul>
5292 <li><tt>ll</tt>: load-load barrier</li>
5293 <li><tt>ls</tt>: load-store barrier</li>
5294 <li><tt>sl</tt>: store-load barrier</li>
5295 <li><tt>ss</tt>: store-store barrier</li>
5296 <li><tt>device</tt>: barrier applies to device and uncached memory also.
5297 </ul>
5298<h5>Semantics:</h5>
5299<p>
5300 This intrinsic causes the system to enforce some ordering constraints upon
5301 the loads and stores of the program. This barrier does not indicate
5302 <em>when</em> any events will occur, it only enforces an <em>order</em> in
5303 which they occur. For any of the specified pairs of load and store operations
5304 (f.ex. load-load, or store-load), all of the first operations preceding the
5305 barrier will complete before any of the second operations succeeding the
5306 barrier begin. Specifically the semantics for each pairing is as follows:
5307</p>
5308 <ul>
5309 <li><tt>ll</tt>: All loads before the barrier must complete before any load
5310 after the barrier begins.</li>
5311
5312 <li><tt>ls</tt>: All loads before the barrier must complete before any
5313 store after the barrier begins.</li>
5314 <li><tt>ss</tt>: All stores before the barrier must complete before any
5315 store after the barrier begins.</li>
5316 <li><tt>sl</tt>: All stores before the barrier must complete before any
5317 load after the barrier begins.</li>
5318 </ul>
5319<p>
5320 These semantics are applied with a logical "and" behavior when more than one
5321 is enabled in a single memory barrier intrinsic.
5322</p>
5323<p>
5324 Backends may implement stronger barriers than those requested when they do not
5325 support as fine grained a barrier as requested. Some architectures do not
5326 need all types of barriers and on such architectures, these become noops.
5327</p>
5328<h5>Example:</h5>
5329<pre>
5330%ptr = malloc i32
5331 store i32 4, %ptr
5332
5333%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
5334 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
5335 <i>; guarantee the above finishes</i>
5336 store i32 8, %ptr <i>; before this begins</i>
5337</pre>
5338</div>
5339
5340
5341<!-- ======================================================================= -->
5342<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5343 <a name="int_general">General Intrinsics</a>
5344</div>
5345
5346<div class="doc_text">
5347<p> This class of intrinsics is designed to be generic and has
5348no specific purpose. </p>
5349</div>
5350
5351<!-- _______________________________________________________________________ -->
5352<div class="doc_subsubsection">
5353 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
5354</div>
5355
5356<div class="doc_text">
5357
5358<h5>Syntax:</h5>
5359<pre>
5360 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5361</pre>
5362
5363<h5>Overview:</h5>
5364
5365<p>
5366The '<tt>llvm.var.annotation</tt>' intrinsic
5367</p>
5368
5369<h5>Arguments:</h5>
5370
5371<p>
5372The first argument is a pointer to a value, the second is a pointer to a
5373global string, the third is a pointer to a global string which is the source
5374file name, and the last argument is the line number.
5375</p>
5376
5377<h5>Semantics:</h5>
5378
5379<p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]5380This intrinsic allows annotation of local variables with arbitrary strings.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5381This can be useful for special purpose optimizations that want to look for these
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]5382annotations. These have no other defined use, they are ignored by code
5383generation and optimization.
5384</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5385</div>
5386
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5387<!-- _______________________________________________________________________ -->
5388<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]5389 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5390</div>
5391
5392<div class="doc_text">
5393
5394<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5395<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
5396any integer bit width.
5397</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5398<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]5399 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5400 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5401 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5402 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5403 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]5404</pre>
5405
5406<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5407
5408<p>
5409The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5410</p>
5411
5412<h5>Arguments:</h5>
5413
5414<p>
5415The first argument is an integer value (result of some expression),
5416the second is a pointer to a global string, the third is a pointer to a global
5417string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5418It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5419</p>
5420
5421<h5>Semantics:</h5>
5422
5423<p>
5424This intrinsic allows annotations to be put on arbitrary expressions
5425with arbitrary strings. This can be useful for special purpose optimizations
5426that want to look for these annotations. These have no other defined use, they
5427are ignored by code generation and optimization.
5428</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5429
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]5430<!-- _______________________________________________________________________ -->
5431<div class="doc_subsubsection">
5432 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
5433</div>
5434
5435<div class="doc_text">
5436
5437<h5>Syntax:</h5>
5438<pre>
5439 declare void @llvm.trap()
5440</pre>
5441
5442<h5>Overview:</h5>
5443
5444<p>
5445The '<tt>llvm.trap</tt>' intrinsic
5446</p>
5447
5448<h5>Arguments:</h5>
5449
5450<p>
5451None
5452</p>
5453
5454<h5>Semantics:</h5>
5455
5456<p>
5457This intrinsics is lowered to the target dependent trap instruction. If the
5458target does not have a trap instruction, this intrinsic will be lowered to the
5459call of the abort() function.
5460</p>
5461</div>
5462
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5463<!-- *********************************************************************** -->
5464<hr>
5465<address>
5466 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
5467 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
5468 <a href="http://validator.w3.org/check/referer"><img
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]5469 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5470
5471 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
5472 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
5473 Last modified: $Date$
5474</address>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]5475
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5476</body>
5477</html>