Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / llvm / 9aba1e2863182794132023bc6e6bc911fafe0800 / . / docs / LangRef.html
blob: a2ee339bc0bac83a8b56e332ee21581b867096ec [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>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]207 <li><a href="#int_general">General intrinsics</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]208 <ol>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]209 <li><a href="#int_var_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]210 <tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]211 <li><a href="#int_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]212 <tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]213 <li><a href="#int_trap">
214 <tt>llvm.trap</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]215 </ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]216 </li>
217 </ol>
218 </li>
219</ol>
220
221<div class="doc_author">
222 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
223 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
224</div>
225
226<!-- *********************************************************************** -->
227<div class="doc_section"> <a name="abstract">Abstract </a></div>
228<!-- *********************************************************************** -->
229
230<div class="doc_text">
231<p>This document is a reference manual for the LLVM assembly language.
232LLVM is an SSA based representation that provides type safety,
233low-level operations, flexibility, and the capability of representing
234'all' high-level languages cleanly. It is the common code
235representation used throughout all phases of the LLVM compilation
236strategy.</p>
237</div>
238
239<!-- *********************************************************************** -->
240<div class="doc_section"> <a name="introduction">Introduction</a> </div>
241<!-- *********************************************************************** -->
242
243<div class="doc_text">
244
245<p>The LLVM code representation is designed to be used in three
246different forms: as an in-memory compiler IR, as an on-disk bitcode
247representation (suitable for fast loading by a Just-In-Time compiler),
248and as a human readable assembly language representation. This allows
249LLVM to provide a powerful intermediate representation for efficient
250compiler transformations and analysis, while providing a natural means
251to debug and visualize the transformations. The three different forms
252of LLVM are all equivalent. This document describes the human readable
253representation and notation.</p>
254
255<p>The LLVM representation aims to be light-weight and low-level
256while being expressive, typed, and extensible at the same time. It
257aims to be a "universal IR" of sorts, by being at a low enough level
258that high-level ideas may be cleanly mapped to it (similar to how
259microprocessors are "universal IR's", allowing many source languages to
260be mapped to them). By providing type information, LLVM can be used as
261the target of optimizations: for example, through pointer analysis, it
262can be proven that a C automatic variable is never accessed outside of
263the current function... allowing it to be promoted to a simple SSA
264value instead of a memory location.</p>
265
266</div>
267
268<!-- _______________________________________________________________________ -->
269<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
270
271<div class="doc_text">
272
273<p>It is important to note that this document describes 'well formed'
274LLVM assembly language. There is a difference between what the parser
275accepts and what is considered 'well formed'. For example, the
276following instruction is syntactically okay, but not well formed:</p>
277
278<div class="doc_code">
279<pre>
280%x = <a href="#i_add">add</a> i32 1, %x
281</pre>
282</div>
283
284<p>...because the definition of <tt>%x</tt> does not dominate all of
285its uses. The LLVM infrastructure provides a verification pass that may
286be used to verify that an LLVM module is well formed. This pass is
287automatically run by the parser after parsing input assembly and by
288the optimizer before it outputs bitcode. The violations pointed out
289by the verifier pass indicate bugs in transformation passes or input to
290the parser.</p>
291</div>
292
Chris Lattnera83fdc02007-10-03 17:34:29 +0000[diff] [blame]293<!-- Describe the typesetting conventions here. -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]294
295<!-- *********************************************************************** -->
296<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
297<!-- *********************************************************************** -->
298
299<div class="doc_text">
300
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]301 <p>LLVM identifiers come in two basic types: global and local. Global
302 identifiers (functions, global variables) begin with the @ character. Local
303 identifiers (register names, types) begin with the % character. Additionally,
304 there are three different formats for identifiers, for different purposes:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]305
306<ol>
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]307 <li>Named values are represented as a string of characters with their prefix.
308 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
309 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]310 Identifiers which require other characters in their names can be surrounded
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]311 with quotes. In this way, anything except a <tt>&quot;</tt> character can
312 be used in a named value.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]313
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]314 <li>Unnamed values are represented as an unsigned numeric value with their
315 prefix. For example, %12, @2, %44.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]316
317 <li>Constants, which are described in a <a href="#constants">section about
318 constants</a>, below.</li>
319</ol>
320
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]321<p>LLVM requires that values start with a prefix for two reasons: Compilers
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]322don't need to worry about name clashes with reserved words, and the set of
323reserved words may be expanded in the future without penalty. Additionally,
324unnamed identifiers allow a compiler to quickly come up with a temporary
325variable without having to avoid symbol table conflicts.</p>
326
327<p>Reserved words in LLVM are very similar to reserved words in other
328languages. There are keywords for different opcodes
329('<tt><a href="#i_add">add</a></tt>',
330 '<tt><a href="#i_bitcast">bitcast</a></tt>',
331 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
332href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
333and others. These reserved words cannot conflict with variable names, because
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]334none of them start with a prefix character ('%' or '@').</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]335
336<p>Here is an example of LLVM code to multiply the integer variable
337'<tt>%X</tt>' by 8:</p>
338
339<p>The easy way:</p>
340
341<div class="doc_code">
342<pre>
343%result = <a href="#i_mul">mul</a> i32 %X, 8
344</pre>
345</div>
346
347<p>After strength reduction:</p>
348
349<div class="doc_code">
350<pre>
351%result = <a href="#i_shl">shl</a> i32 %X, i8 3
352</pre>
353</div>
354
355<p>And the hard way:</p>
356
357<div class="doc_code">
358<pre>
359<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
360<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
361%result = <a href="#i_add">add</a> i32 %1, %1
362</pre>
363</div>
364
365<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
366important lexical features of LLVM:</p>
367
368<ol>
369
370 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
371 line.</li>
372
373 <li>Unnamed temporaries are created when the result of a computation is not
374 assigned to a named value.</li>
375
376 <li>Unnamed temporaries are numbered sequentially</li>
377
378</ol>
379
380<p>...and it also shows a convention that we follow in this document. When
381demonstrating instructions, we will follow an instruction with a comment that
382defines the type and name of value produced. Comments are shown in italic
383text.</p>
384
385</div>
386
387<!-- *********************************************************************** -->
388<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
389<!-- *********************************************************************** -->
390
391<!-- ======================================================================= -->
392<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
393</div>
394
395<div class="doc_text">
396
397<p>LLVM programs are composed of "Module"s, each of which is a
398translation unit of the input programs. Each module consists of
399functions, global variables, and symbol table entries. Modules may be
400combined together with the LLVM linker, which merges function (and
401global variable) definitions, resolves forward declarations, and merges
402symbol table entries. Here is an example of the "hello world" module:</p>
403
404<div class="doc_code">
405<pre><i>; Declare the string constant as a global constant...</i>
406<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
407 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
408
409<i>; External declaration of the puts function</i>
410<a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
411
412<i>; Definition of main function</i>
413define i32 @main() { <i>; i32()* </i>
414 <i>; Convert [13x i8 ]* to i8 *...</i>
415 %cast210 = <a
416 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
417
418 <i>; Call puts function to write out the string to stdout...</i>
419 <a
420 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
421 <a
422 href="#i_ret">ret</a> i32 0<br>}<br>
423</pre>
424</div>
425
426<p>This example is made up of a <a href="#globalvars">global variable</a>
427named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
428function, and a <a href="#functionstructure">function definition</a>
429for "<tt>main</tt>".</p>
430
431<p>In general, a module is made up of a list of global values,
432where both functions and global variables are global values. Global values are
433represented by a pointer to a memory location (in this case, a pointer to an
434array of char, and a pointer to a function), and have one of the following <a
435href="#linkage">linkage types</a>.</p>
436
437</div>
438
439<!-- ======================================================================= -->
440<div class="doc_subsection">
441 <a name="linkage">Linkage Types</a>
442</div>
443
444<div class="doc_text">
445
446<p>
447All Global Variables and Functions have one of the following types of linkage:
448</p>
449
450<dl>
451
452 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
453
454 <dd>Global values with internal linkage are only directly accessible by
455 objects in the current module. In particular, linking code into a module with
456 an internal global value may cause the internal to be renamed as necessary to
457 avoid collisions. Because the symbol is internal to the module, all
458 references can be updated. This corresponds to the notion of the
459 '<tt>static</tt>' keyword in C.
460 </dd>
461
462 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
463
464 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
465 the same name when linkage occurs. This is typically used to implement
466 inline functions, templates, or other code which must be generated in each
467 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
468 allowed to be discarded.
469 </dd>
470
471 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
472
473 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
474 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
475 used for globals that may be emitted in multiple translation units, but that
476 are not guaranteed to be emitted into every translation unit that uses them.
477 One example of this are common globals in C, such as "<tt>int X;</tt>" at
478 global scope.
479 </dd>
480
481 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
482
483 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
484 pointer to array type. When two global variables with appending linkage are
485 linked together, the two global arrays are appended together. This is the
486 LLVM, typesafe, equivalent of having the system linker append together
487 "sections" with identical names when .o files are linked.
488 </dd>
489
490 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
491 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
492 until linked, if not linked, the symbol becomes null instead of being an
493 undefined reference.
494 </dd>
495
496 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
497
498 <dd>If none of the above identifiers are used, the global is externally
499 visible, meaning that it participates in linkage and can be used to resolve
500 external symbol references.
501 </dd>
502</dl>
503
504 <p>
505 The next two types of linkage are targeted for Microsoft Windows platform
506 only. They are designed to support importing (exporting) symbols from (to)
507 DLLs.
508 </p>
509
510 <dl>
511 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
512
513 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
514 or variable via a global pointer to a pointer that is set up by the DLL
515 exporting the symbol. On Microsoft Windows targets, the pointer name is
516 formed by combining <code>_imp__</code> and the function or variable name.
517 </dd>
518
519 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
520
521 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
522 pointer to a pointer in a DLL, so that it can be referenced with the
523 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
524 name is formed by combining <code>_imp__</code> and the function or variable
525 name.
526 </dd>
527
528</dl>
529
530<p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
531variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
532variable and was linked with this one, one of the two would be renamed,
533preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
534external (i.e., lacking any linkage declarations), they are accessible
535outside of the current module.</p>
536<p>It is illegal for a function <i>declaration</i>
537to have any linkage type other than "externally visible", <tt>dllimport</tt>,
538or <tt>extern_weak</tt>.</p>
539<p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
540linkages.
541</div>
542
543<!-- ======================================================================= -->
544<div class="doc_subsection">
545 <a name="callingconv">Calling Conventions</a>
546</div>
547
548<div class="doc_text">
549
550<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
551and <a href="#i_invoke">invokes</a> can all have an optional calling convention
552specified for the call. The calling convention of any pair of dynamic
553caller/callee must match, or the behavior of the program is undefined. The
554following calling conventions are supported by LLVM, and more may be added in
555the future:</p>
556
557<dl>
558 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
559
560 <dd>This calling convention (the default if no other calling convention is
561 specified) matches the target C calling conventions. This calling convention
562 supports varargs function calls and tolerates some mismatch in the declared
563 prototype and implemented declaration of the function (as does normal C).
564 </dd>
565
566 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
567
568 <dd>This calling convention attempts to make calls as fast as possible
569 (e.g. by passing things in registers). This calling convention allows the
570 target to use whatever tricks it wants to produce fast code for the target,
571 without having to conform to an externally specified ABI. Implementations of
572 this convention should allow arbitrary tail call optimization to be supported.
573 This calling convention does not support varargs and requires the prototype of
574 all callees to exactly match the prototype of the function definition.
575 </dd>
576
577 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
578
579 <dd>This calling convention attempts to make code in the caller as efficient
580 as possible under the assumption that the call is not commonly executed. As
581 such, these calls often preserve all registers so that the call does not break
582 any live ranges in the caller side. This calling convention does not support
583 varargs and requires the prototype of all callees to exactly match the
584 prototype of the function definition.
585 </dd>
586
587 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
588
589 <dd>Any calling convention may be specified by number, allowing
590 target-specific calling conventions to be used. Target specific calling
591 conventions start at 64.
592 </dd>
593</dl>
594
595<p>More calling conventions can be added/defined on an as-needed basis, to
596support pascal conventions or any other well-known target-independent
597convention.</p>
598
599</div>
600
601<!-- ======================================================================= -->
602<div class="doc_subsection">
603 <a name="visibility">Visibility Styles</a>
604</div>
605
606<div class="doc_text">
607
608<p>
609All Global Variables and Functions have one of the following visibility styles:
610</p>
611
612<dl>
613 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
614
615 <dd>On ELF, default visibility means that the declaration is visible to other
616 modules and, in shared libraries, means that the declared entity may be
617 overridden. On Darwin, default visibility means that the declaration is
618 visible to other modules. Default visibility corresponds to "external
619 linkage" in the language.
620 </dd>
621
622 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
623
624 <dd>Two declarations of an object with hidden visibility refer to the same
625 object if they are in the same shared object. Usually, hidden visibility
626 indicates that the symbol will not be placed into the dynamic symbol table,
627 so no other module (executable or shared library) can reference it
628 directly.
629 </dd>
630
631 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
632
633 <dd>On ELF, protected visibility indicates that the symbol will be placed in
634 the dynamic symbol table, but that references within the defining module will
635 bind to the local symbol. That is, the symbol cannot be overridden by another
636 module.
637 </dd>
638</dl>
639
640</div>
641
642<!-- ======================================================================= -->
643<div class="doc_subsection">
644 <a name="globalvars">Global Variables</a>
645</div>
646
647<div class="doc_text">
648
649<p>Global variables define regions of memory allocated at compilation time
650instead of run-time. Global variables may optionally be initialized, may have
651an explicit section to be placed in, and may have an optional explicit alignment
652specified. A variable may be defined as "thread_local", which means that it
653will not be shared by threads (each thread will have a separated copy of the
654variable). A variable may be defined as a global "constant," which indicates
655that the contents of the variable will <b>never</b> be modified (enabling better
656optimization, allowing the global data to be placed in the read-only section of
657an executable, etc). Note that variables that need runtime initialization
658cannot be marked "constant" as there is a store to the variable.</p>
659
660<p>
661LLVM explicitly allows <em>declarations</em> of global variables to be marked
662constant, even if the final definition of the global is not. This capability
663can be used to enable slightly better optimization of the program, but requires
664the language definition to guarantee that optimizations based on the
665'constantness' are valid for the translation units that do not include the
666definition.
667</p>
668
669<p>As SSA values, global variables define pointer values that are in
670scope (i.e. they dominate) all basic blocks in the program. Global
671variables always define a pointer to their "content" type because they
672describe a region of memory, and all memory objects in LLVM are
673accessed through pointers.</p>
674
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]675<p>A global variable may be declared to reside in a target-specifc numbered
676address space. For targets that support them, address spaces may affect how
677optimizations are performed and/or what target instructions are used to access
Christopher Lamb20a39e92007-12-12 08:44:39 +0000[diff] [blame]678the variable. The default address space is zero. The address space qualifier
679must precede any other attributes.</p>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]680
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]681<p>LLVM allows an explicit section to be specified for globals. If the target
682supports it, it will emit globals to the section specified.</p>
683
684<p>An explicit alignment may be specified for a global. If not present, or if
685the alignment is set to zero, the alignment of the global is set by the target
686to whatever it feels convenient. If an explicit alignment is specified, the
687global is forced to have at least that much alignment. All alignments must be
688a power of 2.</p>
689
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]690<p>For example, the following defines a global in a numbered address space with
691an initializer, section, and alignment:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]692
693<div class="doc_code">
694<pre>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]695@G = constant float 1.0 addrspace(5), section "foo", align 4
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]696</pre>
697</div>
698
699</div>
700
701
702<!-- ======================================================================= -->
703<div class="doc_subsection">
704 <a name="functionstructure">Functions</a>
705</div>
706
707<div class="doc_text">
708
709<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
710an optional <a href="#linkage">linkage type</a>, an optional
711<a href="#visibility">visibility style</a>, an optional
712<a href="#callingconv">calling convention</a>, a return type, an optional
713<a href="#paramattrs">parameter attribute</a> for the return type, a function
714name, a (possibly empty) argument list (each with optional
715<a href="#paramattrs">parameter attributes</a>), an optional section, an
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]716optional alignment, an optional <a href="#gc">garbage collector name</a>, an
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]717opening curly brace, a list of basic blocks, and a closing curly brace.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]718
719LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
720optional <a href="#linkage">linkage type</a>, an optional
721<a href="#visibility">visibility style</a>, an optional
722<a href="#callingconv">calling convention</a>, a return type, an optional
723<a href="#paramattrs">parameter attribute</a> for the return type, a function
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]724name, a possibly empty list of arguments, an optional alignment, and an optional
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]725<a href="#gc">garbage collector name</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]726
727<p>A function definition contains a list of basic blocks, forming the CFG for
728the function. Each basic block may optionally start with a label (giving the
729basic block a symbol table entry), contains a list of instructions, and ends
730with a <a href="#terminators">terminator</a> instruction (such as a branch or
731function return).</p>
732
733<p>The first basic block in a function is special in two ways: it is immediately
734executed on entrance to the function, and it is not allowed to have predecessor
735basic blocks (i.e. there can not be any branches to the entry block of a
736function). Because the block can have no predecessors, it also cannot have any
737<a href="#i_phi">PHI nodes</a>.</p>
738
739<p>LLVM allows an explicit section to be specified for functions. If the target
740supports it, it will emit functions to the section specified.</p>
741
742<p>An explicit alignment may be specified for a function. If not present, or if
743the alignment is set to zero, the alignment of the function is set by the target
744to whatever it feels convenient. If an explicit alignment is specified, the
745function is forced to have at least that much alignment. All alignments must be
746a power of 2.</p>
747
748</div>
749
750
751<!-- ======================================================================= -->
752<div class="doc_subsection">
753 <a name="aliasstructure">Aliases</a>
754</div>
755<div class="doc_text">
756 <p>Aliases act as "second name" for the aliasee value (which can be either
757 function or global variable or bitcast of global value). Aliases may have an
758 optional <a href="#linkage">linkage type</a>, and an
759 optional <a href="#visibility">visibility style</a>.</p>
760
761 <h5>Syntax:</h5>
762
763<div class="doc_code">
764<pre>
765@&lt;Name&gt; = [Linkage] [Visibility] alias &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
766</pre>
767</div>
768
769</div>
770
771
772
773<!-- ======================================================================= -->
774<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
775<div class="doc_text">
776 <p>The return type and each parameter of a function type may have a set of
777 <i>parameter attributes</i> associated with them. Parameter attributes are
778 used to communicate additional information about the result or parameters of
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]779 a function. Parameter attributes are considered to be part of the function,
780 not of the function type, so functions with different parameter attributes
781 can have the same function type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]782
783 <p>Parameter attributes are simple keywords that follow the type specified. If
784 multiple parameter attributes are needed, they are space separated. For
785 example:</p>
786
787<div class="doc_code">
788<pre>
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]789declare i32 @printf(i8* noalias , ...) nounwind
790declare i32 @atoi(i8*) nounwind readonly
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]791</pre>
792</div>
793
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]794 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
795 <tt>readonly</tt>) come immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]796
797 <p>Currently, only the following parameter attributes are defined:</p>
798 <dl>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]799 <dt><tt>zeroext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]800 <dd>This indicates that the parameter should be zero extended just before
801 a call to this function.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]802
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]803 <dt><tt>signext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]804 <dd>This indicates that the parameter should be sign extended just before
805 a call to this function.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]806
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]807 <dt><tt>inreg</tt></dt>
808 <dd>This indicates that the parameter should be placed in register (if
809 possible) during assembling function call. Support for this attribute is
810 target-specific</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]811
812 <dt><tt>byval</tt></dt>
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]813 <dd>This indicates that the pointer parameter should really be passed by
814 value to the function. The attribute implies that a hidden copy of the
815 pointee is made between the caller and the callee, so the callee is unable
816 to modify the value in the callee. This attribute is only valid on llvm
817 pointer arguments. It is generally used to pass structs and arrays by
818 value, but is also valid on scalars (even though this is silly).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]819
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]820 <dt><tt>sret</tt></dt>
821 <dd>This indicates that the parameter specifies the address of a structure
822 that is the return value of the function in the source program.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]823
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]824 <dt><tt>noalias</tt></dt>
825 <dd>This indicates that the parameter not alias any other object or any
826 other "noalias" objects during the function call.
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]827
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]828 <dt><tt>noreturn</tt></dt>
829 <dd>This function attribute indicates that the function never returns. This
830 indicates to LLVM that every call to this function should be treated as if
831 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]832
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]833 <dt><tt>nounwind</tt></dt>
834 <dd>This function attribute indicates that the function type does not use
835 the unwind instruction and does not allow stack unwinding to propagate
836 through it.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]837
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]838 <dt><tt>nest</tt></dt>
839 <dd>This indicates that the parameter can be excised using the
840 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
Duncan Sands13e13f82007-11-22 20:23:04 +0000[diff] [blame]841 <dt><tt>readonly</tt></dt>
Duncan Sandsd69c0e62007-11-14 21:14:02 +0000[diff] [blame]842 <dd>This function attribute indicates that the function has no side-effects
Duncan Sands13e13f82007-11-22 20:23:04 +0000[diff] [blame]843 except for producing a return value or throwing an exception. The value
844 returned must only depend on the function arguments and/or global variables.
845 It may use values obtained by dereferencing pointers.</dd>
846 <dt><tt>readnone</tt></dt>
847 <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]848 function, but in addition it is not allowed to dereference any pointer arguments
849 or global variables.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]850 </dl>
851
852</div>
853
854<!-- ======================================================================= -->
855<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]856 <a name="gc">Garbage Collector Names</a>
857</div>
858
859<div class="doc_text">
860<p>Each function may specify a garbage collector name, which is simply a
861string.</p>
862
863<div class="doc_code"><pre
864>define void @f() gc "name" { ...</pre></div>
865
866<p>The compiler declares the supported values of <i>name</i>. Specifying a
867collector which will cause the compiler to alter its output in order to support
868the named garbage collection algorithm.</p>
869</div>
870
871<!-- ======================================================================= -->
872<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]873 <a name="moduleasm">Module-Level Inline Assembly</a>
874</div>
875
876<div class="doc_text">
877<p>
878Modules may contain "module-level inline asm" blocks, which corresponds to the
879GCC "file scope inline asm" blocks. These blocks are internally concatenated by
880LLVM and treated as a single unit, but may be separated in the .ll file if
881desired. The syntax is very simple:
882</p>
883
884<div class="doc_code">
885<pre>
886module asm "inline asm code goes here"
887module asm "more can go here"
888</pre>
889</div>
890
891<p>The strings can contain any character by escaping non-printable characters.
892 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
893 for the number.
894</p>
895
896<p>
897 The inline asm code is simply printed to the machine code .s file when
898 assembly code is generated.
899</p>
900</div>
901
902<!-- ======================================================================= -->
903<div class="doc_subsection">
904 <a name="datalayout">Data Layout</a>
905</div>
906
907<div class="doc_text">
908<p>A module may specify a target specific data layout string that specifies how
909data is to be laid out in memory. The syntax for the data layout is simply:</p>
910<pre> target datalayout = "<i>layout specification</i>"</pre>
911<p>The <i>layout specification</i> consists of a list of specifications
912separated by the minus sign character ('-'). Each specification starts with a
913letter and may include other information after the letter to define some
914aspect of the data layout. The specifications accepted are as follows: </p>
915<dl>
916 <dt><tt>E</tt></dt>
917 <dd>Specifies that the target lays out data in big-endian form. That is, the
918 bits with the most significance have the lowest address location.</dd>
919 <dt><tt>e</tt></dt>
920 <dd>Specifies that hte target lays out data in little-endian form. That is,
921 the bits with the least significance have the lowest address location.</dd>
922 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
923 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
924 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
925 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
926 too.</dd>
927 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
928 <dd>This specifies the alignment for an integer type of a given bit
929 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
930 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
931 <dd>This specifies the alignment for a vector type of a given bit
932 <i>size</i>.</dd>
933 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
934 <dd>This specifies the alignment for a floating point type of a given bit
935 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
936 (double).</dd>
937 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
938 <dd>This specifies the alignment for an aggregate type of a given bit
939 <i>size</i>.</dd>
940</dl>
941<p>When constructing the data layout for a given target, LLVM starts with a
942default set of specifications which are then (possibly) overriden by the
943specifications in the <tt>datalayout</tt> keyword. The default specifications
944are given in this list:</p>
945<ul>
946 <li><tt>E</tt> - big endian</li>
947 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
948 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
949 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
950 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
951 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
952 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
953 alignment of 64-bits</li>
954 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
955 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
956 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
957 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
958 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
959</ul>
960<p>When llvm is determining the alignment for a given type, it uses the
961following rules:
962<ol>
963 <li>If the type sought is an exact match for one of the specifications, that
964 specification is used.</li>
965 <li>If no match is found, and the type sought is an integer type, then the
966 smallest integer type that is larger than the bitwidth of the sought type is
967 used. If none of the specifications are larger than the bitwidth then the the
968 largest integer type is used. For example, given the default specifications
969 above, the i7 type will use the alignment of i8 (next largest) while both
970 i65 and i256 will use the alignment of i64 (largest specified).</li>
971 <li>If no match is found, and the type sought is a vector type, then the
972 largest vector type that is smaller than the sought vector type will be used
973 as a fall back. This happens because <128 x double> can be implemented in
974 terms of 64 <2 x double>, for example.</li>
975</ol>
976</div>
977
978<!-- *********************************************************************** -->
979<div class="doc_section"> <a name="typesystem">Type System</a> </div>
980<!-- *********************************************************************** -->
981
982<div class="doc_text">
983
984<p>The LLVM type system is one of the most important features of the
985intermediate representation. Being typed enables a number of
986optimizations to be performed on the IR directly, without having to do
987extra analyses on the side before the transformation. A strong type
988system makes it easier to read the generated code and enables novel
989analyses and transformations that are not feasible to perform on normal
990three address code representations.</p>
991
992</div>
993
994<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]995<div class="doc_subsection"> <a name="t_classifications">Type
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]996Classifications</a> </div>
997<div class="doc_text">
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]998<p>The types fall into a few useful
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]999classifications:</p>
1000
1001<table border="1" cellspacing="0" cellpadding="4">
1002 <tbody>
1003 <tr><th>Classification</th><th>Types</th></tr>
1004 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1005 <td><a href="#t_integer">integer</a></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1006 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1007 </tr>
1008 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1009 <td><a href="#t_floating">floating point</a></td>
1010 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1011 </tr>
1012 <tr>
1013 <td><a name="t_firstclass">first class</a></td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1014 <td><a href="#t_integer">integer</a>,
1015 <a href="#t_floating">floating point</a>,
1016 <a href="#t_pointer">pointer</a>,
1017 <a href="#t_vector">vector</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1018 </td>
1019 </tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1020 <tr>
1021 <td><a href="#t_primitive">primitive</a></td>
1022 <td><a href="#t_label">label</a>,
1023 <a href="#t_void">void</a>,
1024 <a href="#t_integer">integer</a>,
1025 <a href="#t_floating">floating point</a>.</td>
1026 </tr>
1027 <tr>
1028 <td><a href="#t_derived">derived</a></td>
1029 <td><a href="#t_integer">integer</a>,
1030 <a href="#t_array">array</a>,
1031 <a href="#t_function">function</a>,
1032 <a href="#t_pointer">pointer</a>,
1033 <a href="#t_struct">structure</a>,
1034 <a href="#t_pstruct">packed structure</a>,
1035 <a href="#t_vector">vector</a>,
1036 <a href="#t_opaque">opaque</a>.
1037 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1038 </tbody>
1039</table>
1040
1041<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1042most important. Values of these types are the only ones which can be
1043produced by instructions, passed as arguments, or used as operands to
1044instructions. This means that all structures and arrays must be
1045manipulated either by pointer or by component.</p>
1046</div>
1047
1048<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1049<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1050
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1051<div class="doc_text">
1052<p>The primitive types are the fundamental building blocks of the LLVM
1053system.</p>
1054
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1055</div>
1056
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1057<!-- _______________________________________________________________________ -->
1058<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1059
1060<div class="doc_text">
1061 <table>
1062 <tbody>
1063 <tr><th>Type</th><th>Description</th></tr>
1064 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1065 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1066 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1067 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1068 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1069 </tbody>
1070 </table>
1071</div>
1072
1073<!-- _______________________________________________________________________ -->
1074<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1075
1076<div class="doc_text">
1077<h5>Overview:</h5>
1078<p>The void type does not represent any value and has no size.</p>
1079
1080<h5>Syntax:</h5>
1081
1082<pre>
1083 void
1084</pre>
1085</div>
1086
1087<!-- _______________________________________________________________________ -->
1088<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1089
1090<div class="doc_text">
1091<h5>Overview:</h5>
1092<p>The label type represents code labels.</p>
1093
1094<h5>Syntax:</h5>
1095
1096<pre>
1097 label
1098</pre>
1099</div>
1100
1101
1102<!-- ======================================================================= -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1103<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1104
1105<div class="doc_text">
1106
1107<p>The real power in LLVM comes from the derived types in the system.
1108This is what allows a programmer to represent arrays, functions,
1109pointers, and other useful types. Note that these derived types may be
1110recursive: For example, it is possible to have a two dimensional array.</p>
1111
1112</div>
1113
1114<!-- _______________________________________________________________________ -->
1115<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1116
1117<div class="doc_text">
1118
1119<h5>Overview:</h5>
1120<p>The integer type is a very simple derived type that simply specifies an
1121arbitrary bit width for the integer type desired. Any bit width from 1 bit to
11222^23-1 (about 8 million) can be specified.</p>
1123
1124<h5>Syntax:</h5>
1125
1126<pre>
1127 iN
1128</pre>
1129
1130<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1131value.</p>
1132
1133<h5>Examples:</h5>
1134<table class="layout">
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1135 <tbody>
1136 <tr>
1137 <td><tt>i1</tt></td>
1138 <td>a single-bit integer.</td>
1139 </tr><tr>
1140 <td><tt>i32</tt></td>
1141 <td>a 32-bit integer.</td>
1142 </tr><tr>
1143 <td><tt>i1942652</tt></td>
1144 <td>a really big integer of over 1 million bits.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1145 </tr>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1146 </tbody>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1147</table>
1148</div>
1149
1150<!-- _______________________________________________________________________ -->
1151<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1152
1153<div class="doc_text">
1154
1155<h5>Overview:</h5>
1156
1157<p>The array type is a very simple derived type that arranges elements
1158sequentially in memory. The array type requires a size (number of
1159elements) and an underlying data type.</p>
1160
1161<h5>Syntax:</h5>
1162
1163<pre>
1164 [&lt;# elements&gt; x &lt;elementtype&gt;]
1165</pre>
1166
1167<p>The number of elements is a constant integer value; elementtype may
1168be any type with a size.</p>
1169
1170<h5>Examples:</h5>
1171<table class="layout">
1172 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1173 <td class="left"><tt>[40 x i32]</tt></td>
1174 <td class="left">Array of 40 32-bit integer values.</td>
1175 </tr>
1176 <tr class="layout">
1177 <td class="left"><tt>[41 x i32]</tt></td>
1178 <td class="left">Array of 41 32-bit integer values.</td>
1179 </tr>
1180 <tr class="layout">
1181 <td class="left"><tt>[4 x i8]</tt></td>
1182 <td class="left">Array of 4 8-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1183 </tr>
1184</table>
1185<p>Here are some examples of multidimensional arrays:</p>
1186<table class="layout">
1187 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1188 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1189 <td class="left">3x4 array of 32-bit integer values.</td>
1190 </tr>
1191 <tr class="layout">
1192 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1193 <td class="left">12x10 array of single precision floating point values.</td>
1194 </tr>
1195 <tr class="layout">
1196 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1197 <td class="left">2x3x4 array of 16-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1198 </tr>
1199</table>
1200
1201<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1202length array. Normally, accesses past the end of an array are undefined in
1203LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1204As a special case, however, zero length arrays are recognized to be variable
1205length. This allows implementation of 'pascal style arrays' with the LLVM
1206type "{ i32, [0 x float]}", for example.</p>
1207
1208</div>
1209
1210<!-- _______________________________________________________________________ -->
1211<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1212<div class="doc_text">
1213<h5>Overview:</h5>
1214<p>The function type can be thought of as a function signature. It
1215consists of a return type and a list of formal parameter types.
1216Function types are usually used to build virtual function tables
1217(which are structures of pointers to functions), for indirect function
1218calls, and when defining a function.</p>
1219<p>
1220The return type of a function type cannot be an aggregate type.
1221</p>
1222<h5>Syntax:</h5>
1223<pre> &lt;returntype&gt; (&lt;parameter list&gt;)<br></pre>
1224<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1225specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1226which indicates that the function takes a variable number of arguments.
1227Variable argument functions can access their arguments with the <a
1228 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1229<h5>Examples:</h5>
1230<table class="layout">
1231 <tr class="layout">
1232 <td class="left"><tt>i32 (i32)</tt></td>
1233 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1234 </td>
1235 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1236 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1237 </tt></td>
1238 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1239 an <tt>i16</tt> that should be sign extended and a
1240 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1241 <tt>float</tt>.
1242 </td>
1243 </tr><tr class="layout">
1244 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1245 <td class="left">A vararg function that takes at least one
1246 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1247 which returns an integer. This is the signature for <tt>printf</tt> in
1248 LLVM.
1249 </td>
1250 </tr>
1251</table>
1252
1253</div>
1254<!-- _______________________________________________________________________ -->
1255<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1256<div class="doc_text">
1257<h5>Overview:</h5>
1258<p>The structure type is used to represent a collection of data members
1259together in memory. The packing of the field types is defined to match
1260the ABI of the underlying processor. The elements of a structure may
1261be any type that has a size.</p>
1262<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1263and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1264field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1265instruction.</p>
1266<h5>Syntax:</h5>
1267<pre> { &lt;type list&gt; }<br></pre>
1268<h5>Examples:</h5>
1269<table class="layout">
1270 <tr class="layout">
1271 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1272 <td class="left">A triple of three <tt>i32</tt> values</td>
1273 </tr><tr class="layout">
1274 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1275 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1276 second element is a <a href="#t_pointer">pointer</a> to a
1277 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1278 an <tt>i32</tt>.</td>
1279 </tr>
1280</table>
1281</div>
1282
1283<!-- _______________________________________________________________________ -->
1284<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1285</div>
1286<div class="doc_text">
1287<h5>Overview:</h5>
1288<p>The packed structure type is used to represent a collection of data members
1289together in memory. There is no padding between fields. Further, the alignment
1290of a packed structure is 1 byte. The elements of a packed structure may
1291be any type that has a size.</p>
1292<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1293and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1294field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1295instruction.</p>
1296<h5>Syntax:</h5>
1297<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1298<h5>Examples:</h5>
1299<table class="layout">
1300 <tr class="layout">
1301 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1302 <td class="left">A triple of three <tt>i32</tt> values</td>
1303 </tr><tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1304 <td class="left"><tt>&lt; { float, i32 (i32)* } &gt;</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1305 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1306 second element is a <a href="#t_pointer">pointer</a> to a
1307 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1308 an <tt>i32</tt>.</td>
1309 </tr>
1310</table>
1311</div>
1312
1313<!-- _______________________________________________________________________ -->
1314<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1315<div class="doc_text">
1316<h5>Overview:</h5>
1317<p>As in many languages, the pointer type represents a pointer or
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1318reference to another object, which must live in memory. Pointer types may have
1319an optional address space attribute defining the target-specific numbered
1320address space where the pointed-to object resides. The default address space is
1321zero.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1322<h5>Syntax:</h5>
1323<pre> &lt;type&gt; *<br></pre>
1324<h5>Examples:</h5>
1325<table class="layout">
1326 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1327 <td class="left"><tt>[4x i32]*</tt></td>
1328 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1329 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1330 </tr>
1331 <tr class="layout">
1332 <td class="left"><tt>i32 (i32 *) *</tt></td>
1333 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1334 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1335 <tt>i32</tt>.</td>
1336 </tr>
1337 <tr class="layout">
1338 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1339 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1340 that resides in address space #5.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1341 </tr>
1342</table>
1343</div>
1344
1345<!-- _______________________________________________________________________ -->
1346<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1347<div class="doc_text">
1348
1349<h5>Overview:</h5>
1350
1351<p>A vector type is a simple derived type that represents a vector
1352of elements. Vector types are used when multiple primitive data
1353are operated in parallel using a single instruction (SIMD).
1354A vector type requires a size (number of
1355elements) and an underlying primitive data type. Vectors must have a power
1356of two length (1, 2, 4, 8, 16 ...). Vector types are
1357considered <a href="#t_firstclass">first class</a>.</p>
1358
1359<h5>Syntax:</h5>
1360
1361<pre>
1362 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1363</pre>
1364
1365<p>The number of elements is a constant integer value; elementtype may
1366be any integer or floating point type.</p>
1367
1368<h5>Examples:</h5>
1369
1370<table class="layout">
1371 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1372 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1373 <td class="left">Vector of 4 32-bit integer values.</td>
1374 </tr>
1375 <tr class="layout">
1376 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1377 <td class="left">Vector of 8 32-bit floating-point values.</td>
1378 </tr>
1379 <tr class="layout">
1380 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1381 <td class="left">Vector of 2 64-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1382 </tr>
1383</table>
1384</div>
1385
1386<!-- _______________________________________________________________________ -->
1387<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1388<div class="doc_text">
1389
1390<h5>Overview:</h5>
1391
1392<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1393corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1394In LLVM, opaque types can eventually be resolved to any type (not just a
1395structure type).</p>
1396
1397<h5>Syntax:</h5>
1398
1399<pre>
1400 opaque
1401</pre>
1402
1403<h5>Examples:</h5>
1404
1405<table class="layout">
1406 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1407 <td class="left"><tt>opaque</tt></td>
1408 <td class="left">An opaque type.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1409 </tr>
1410</table>
1411</div>
1412
1413
1414<!-- *********************************************************************** -->
1415<div class="doc_section"> <a name="constants">Constants</a> </div>
1416<!-- *********************************************************************** -->
1417
1418<div class="doc_text">
1419
1420<p>LLVM has several different basic types of constants. This section describes
1421them all and their syntax.</p>
1422
1423</div>
1424
1425<!-- ======================================================================= -->
1426<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1427
1428<div class="doc_text">
1429
1430<dl>
1431 <dt><b>Boolean constants</b></dt>
1432
1433 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1434 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1435 </dd>
1436
1437 <dt><b>Integer constants</b></dt>
1438
1439 <dd>Standard integers (such as '4') are constants of the <a
1440 href="#t_integer">integer</a> type. Negative numbers may be used with
1441 integer types.
1442 </dd>
1443
1444 <dt><b>Floating point constants</b></dt>
1445
1446 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1447 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1448 notation (see below). Floating point constants must have a <a
1449 href="#t_floating">floating point</a> type. </dd>
1450
1451 <dt><b>Null pointer constants</b></dt>
1452
1453 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1454 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1455
1456</dl>
1457
1458<p>The one non-intuitive notation for constants is the optional hexadecimal form
1459of floating point constants. For example, the form '<tt>double
14600x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
14614.5e+15</tt>'. The only time hexadecimal floating point constants are required
1462(and the only time that they are generated by the disassembler) is when a
1463floating point constant must be emitted but it cannot be represented as a
1464decimal floating point number. For example, NaN's, infinities, and other
1465special values are represented in their IEEE hexadecimal format so that
1466assembly and disassembly do not cause any bits to change in the constants.</p>
1467
1468</div>
1469
1470<!-- ======================================================================= -->
1471<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1472</div>
1473
1474<div class="doc_text">
1475<p>Aggregate constants arise from aggregation of simple constants
1476and smaller aggregate constants.</p>
1477
1478<dl>
1479 <dt><b>Structure constants</b></dt>
1480
1481 <dd>Structure constants are represented with notation similar to structure
1482 type definitions (a comma separated list of elements, surrounded by braces
Chris Lattner6c8de962007-12-25 20:34:52 +0000[diff] [blame]1483 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1484 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]1485 must have <a href="#t_struct">structure type</a>, and the number and
1486 types of elements must match those specified by the type.
1487 </dd>
1488
1489 <dt><b>Array constants</b></dt>
1490
1491 <dd>Array constants are represented with notation similar to array type
1492 definitions (a comma separated list of elements, surrounded by square brackets
1493 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1494 constants must have <a href="#t_array">array type</a>, and the number and
1495 types of elements must match those specified by the type.
1496 </dd>
1497
1498 <dt><b>Vector constants</b></dt>
1499
1500 <dd>Vector constants are represented with notation similar to vector type
1501 definitions (a comma separated list of elements, surrounded by
1502 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1503 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1504 href="#t_vector">vector type</a>, and the number and types of elements must
1505 match those specified by the type.
1506 </dd>
1507
1508 <dt><b>Zero initialization</b></dt>
1509
1510 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1511 value to zero of <em>any</em> type, including scalar and aggregate types.
1512 This is often used to avoid having to print large zero initializers (e.g. for
1513 large arrays) and is always exactly equivalent to using explicit zero
1514 initializers.
1515 </dd>
1516</dl>
1517
1518</div>
1519
1520<!-- ======================================================================= -->
1521<div class="doc_subsection">
1522 <a name="globalconstants">Global Variable and Function Addresses</a>
1523</div>
1524
1525<div class="doc_text">
1526
1527<p>The addresses of <a href="#globalvars">global variables</a> and <a
1528href="#functionstructure">functions</a> are always implicitly valid (link-time)
1529constants. These constants are explicitly referenced when the <a
1530href="#identifiers">identifier for the global</a> is used and always have <a
1531href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1532file:</p>
1533
1534<div class="doc_code">
1535<pre>
1536@X = global i32 17
1537@Y = global i32 42
1538@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1539</pre>
1540</div>
1541
1542</div>
1543
1544<!-- ======================================================================= -->
1545<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1546<div class="doc_text">
1547 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1548 no specific value. Undefined values may be of any type and be used anywhere
1549 a constant is permitted.</p>
1550
1551 <p>Undefined values indicate to the compiler that the program is well defined
1552 no matter what value is used, giving the compiler more freedom to optimize.
1553 </p>
1554</div>
1555
1556<!-- ======================================================================= -->
1557<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1558</div>
1559
1560<div class="doc_text">
1561
1562<p>Constant expressions are used to allow expressions involving other constants
1563to be used as constants. Constant expressions may be of any <a
1564href="#t_firstclass">first class</a> type and may involve any LLVM operation
1565that does not have side effects (e.g. load and call are not supported). The
1566following is the syntax for constant expressions:</p>
1567
1568<dl>
1569 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1570 <dd>Truncate a constant to another type. The bit size of CST must be larger
1571 than the bit size of TYPE. Both types must be integers.</dd>
1572
1573 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1574 <dd>Zero extend a constant to another type. The bit size of CST must be
1575 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1576
1577 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1578 <dd>Sign extend a constant to another type. The bit size of CST must be
1579 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1580
1581 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1582 <dd>Truncate a floating point constant to another floating point type. The
1583 size of CST must be larger than the size of TYPE. Both types must be
1584 floating point.</dd>
1585
1586 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1587 <dd>Floating point extend a constant to another type. The size of CST must be
1588 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1589
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1590 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1591 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1592 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1593 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1594 of the same number of elements. If the value won't fit in the integer type,
1595 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1596
1597 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1598 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1599 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1600 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1601 of the same number of elements. If the value won't fit in the integer type,
1602 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1603
1604 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1605 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1606 constant. TYPE must be a scalar or vector floating point type. CST must be of
1607 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1608 of the same number of elements. If the value won't fit in the floating point
1609 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1610
1611 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1612 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1613 constant. TYPE must be a scalar or vector floating point type. CST must be of
1614 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1615 of the same number of elements. If the value won't fit in the floating point
1616 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1617
1618 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1619 <dd>Convert a pointer typed constant to the corresponding integer constant
1620 TYPE must be an integer type. CST must be of pointer type. The CST value is
1621 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1622
1623 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1624 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1625 pointer type. CST must be of integer type. The CST value is zero extended,
1626 truncated, or unchanged to make it fit in a pointer size. This one is
1627 <i>really</i> dangerous!</dd>
1628
1629 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1630 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1631 identical (same number of bits). The conversion is done as if the CST value
1632 was stored to memory and read back as TYPE. In other words, no bits change
1633 with this operator, just the type. This can be used for conversion of
1634 vector types to any other type, as long as they have the same bit width. For
1635 pointers it is only valid to cast to another pointer type.
1636 </dd>
1637
1638 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1639
1640 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1641 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1642 instruction, the index list may have zero or more indexes, which are required
1643 to make sense for the type of "CSTPTR".</dd>
1644
1645 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1646
1647 <dd>Perform the <a href="#i_select">select operation</a> on
1648 constants.</dd>
1649
1650 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1651 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1652
1653 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1654 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1655
1656 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1657
1658 <dd>Perform the <a href="#i_extractelement">extractelement
1659 operation</a> on constants.
1660
1661 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1662
1663 <dd>Perform the <a href="#i_insertelement">insertelement
1664 operation</a> on constants.</dd>
1665
1666
1667 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1668
1669 <dd>Perform the <a href="#i_shufflevector">shufflevector
1670 operation</a> on constants.</dd>
1671
1672 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1673
1674 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1675 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1676 binary</a> operations. The constraints on operands are the same as those for
1677 the corresponding instruction (e.g. no bitwise operations on floating point
1678 values are allowed).</dd>
1679</dl>
1680</div>
1681
1682<!-- *********************************************************************** -->
1683<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1684<!-- *********************************************************************** -->
1685
1686<!-- ======================================================================= -->
1687<div class="doc_subsection">
1688<a name="inlineasm">Inline Assembler Expressions</a>
1689</div>
1690
1691<div class="doc_text">
1692
1693<p>
1694LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1695Module-Level Inline Assembly</a>) through the use of a special value. This
1696value represents the inline assembler as a string (containing the instructions
1697to emit), a list of operand constraints (stored as a string), and a flag that
1698indicates whether or not the inline asm expression has side effects. An example
1699inline assembler expression is:
1700</p>
1701
1702<div class="doc_code">
1703<pre>
1704i32 (i32) asm "bswap $0", "=r,r"
1705</pre>
1706</div>
1707
1708<p>
1709Inline assembler expressions may <b>only</b> be used as the callee operand of
1710a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1711</p>
1712
1713<div class="doc_code">
1714<pre>
1715%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1716</pre>
1717</div>
1718
1719<p>
1720Inline asms with side effects not visible in the constraint list must be marked
1721as having side effects. This is done through the use of the
1722'<tt>sideeffect</tt>' keyword, like so:
1723</p>
1724
1725<div class="doc_code">
1726<pre>
1727call void asm sideeffect "eieio", ""()
1728</pre>
1729</div>
1730
1731<p>TODO: The format of the asm and constraints string still need to be
1732documented here. Constraints on what can be done (e.g. duplication, moving, etc
1733need to be documented).
1734</p>
1735
1736</div>
1737
1738<!-- *********************************************************************** -->
1739<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1740<!-- *********************************************************************** -->
1741
1742<div class="doc_text">
1743
1744<p>The LLVM instruction set consists of several different
1745classifications of instructions: <a href="#terminators">terminator
1746instructions</a>, <a href="#binaryops">binary instructions</a>,
1747<a href="#bitwiseops">bitwise binary instructions</a>, <a
1748 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1749instructions</a>.</p>
1750
1751</div>
1752
1753<!-- ======================================================================= -->
1754<div class="doc_subsection"> <a name="terminators">Terminator
1755Instructions</a> </div>
1756
1757<div class="doc_text">
1758
1759<p>As mentioned <a href="#functionstructure">previously</a>, every
1760basic block in a program ends with a "Terminator" instruction, which
1761indicates which block should be executed after the current block is
1762finished. These terminator instructions typically yield a '<tt>void</tt>'
1763value: they produce control flow, not values (the one exception being
1764the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1765<p>There are six different terminator instructions: the '<a
1766 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1767instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1768the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1769 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1770 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1771
1772</div>
1773
1774<!-- _______________________________________________________________________ -->
1775<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1776Instruction</a> </div>
1777<div class="doc_text">
1778<h5>Syntax:</h5>
1779<pre> ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
1780 ret void <i>; Return from void function</i>
1781</pre>
1782<h5>Overview:</h5>
1783<p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1784value) from a function back to the caller.</p>
1785<p>There are two forms of the '<tt>ret</tt>' instruction: one that
1786returns a value and then causes control flow, and one that just causes
1787control flow to occur.</p>
1788<h5>Arguments:</h5>
1789<p>The '<tt>ret</tt>' instruction may return any '<a
1790 href="#t_firstclass">first class</a>' type. Notice that a function is
1791not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1792instruction inside of the function that returns a value that does not
1793match the return type of the function.</p>
1794<h5>Semantics:</h5>
1795<p>When the '<tt>ret</tt>' instruction is executed, control flow
1796returns back to the calling function's context. If the caller is a "<a
1797 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1798the instruction after the call. If the caller was an "<a
1799 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1800at the beginning of the "normal" destination block. If the instruction
1801returns a value, that value shall set the call or invoke instruction's
1802return value.</p>
1803<h5>Example:</h5>
1804<pre> ret i32 5 <i>; Return an integer value of 5</i>
1805 ret void <i>; Return from a void function</i>
1806</pre>
1807</div>
1808<!-- _______________________________________________________________________ -->
1809<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1810<div class="doc_text">
1811<h5>Syntax:</h5>
1812<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
1813</pre>
1814<h5>Overview:</h5>
1815<p>The '<tt>br</tt>' instruction is used to cause control flow to
1816transfer to a different basic block in the current function. There are
1817two forms of this instruction, corresponding to a conditional branch
1818and an unconditional branch.</p>
1819<h5>Arguments:</h5>
1820<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1821single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1822unconditional form of the '<tt>br</tt>' instruction takes a single
1823'<tt>label</tt>' value as a target.</p>
1824<h5>Semantics:</h5>
1825<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1826argument is evaluated. If the value is <tt>true</tt>, control flows
1827to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1828control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1829<h5>Example:</h5>
1830<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
1831 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1832</div>
1833<!-- _______________________________________________________________________ -->
1834<div class="doc_subsubsection">
1835 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1836</div>
1837
1838<div class="doc_text">
1839<h5>Syntax:</h5>
1840
1841<pre>
1842 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
1843</pre>
1844
1845<h5>Overview:</h5>
1846
1847<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1848several different places. It is a generalization of the '<tt>br</tt>'
1849instruction, allowing a branch to occur to one of many possible
1850destinations.</p>
1851
1852
1853<h5>Arguments:</h5>
1854
1855<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1856comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1857an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1858table is not allowed to contain duplicate constant entries.</p>
1859
1860<h5>Semantics:</h5>
1861
1862<p>The <tt>switch</tt> instruction specifies a table of values and
1863destinations. When the '<tt>switch</tt>' instruction is executed, this
1864table is searched for the given value. If the value is found, control flow is
1865transfered to the corresponding destination; otherwise, control flow is
1866transfered to the default destination.</p>
1867
1868<h5>Implementation:</h5>
1869
1870<p>Depending on properties of the target machine and the particular
1871<tt>switch</tt> instruction, this instruction may be code generated in different
1872ways. For example, it could be generated as a series of chained conditional
1873branches or with a lookup table.</p>
1874
1875<h5>Example:</h5>
1876
1877<pre>
1878 <i>; Emulate a conditional br instruction</i>
1879 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1880 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1881
1882 <i>; Emulate an unconditional br instruction</i>
1883 switch i32 0, label %dest [ ]
1884
1885 <i>; Implement a jump table:</i>
1886 switch i32 %val, label %otherwise [ i32 0, label %onzero
1887 i32 1, label %onone
1888 i32 2, label %ontwo ]
1889</pre>
1890</div>
1891
1892<!-- _______________________________________________________________________ -->
1893<div class="doc_subsubsection">
1894 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1895</div>
1896
1897<div class="doc_text">
1898
1899<h5>Syntax:</h5>
1900
1901<pre>
1902 &lt;result&gt; = invoke [<a href="#callingconv">cconv</a>] &lt;ptr to function ty&gt; %&lt;function ptr val&gt;(&lt;function args&gt;)
1903 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
1904</pre>
1905
1906<h5>Overview:</h5>
1907
1908<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1909function, with the possibility of control flow transfer to either the
1910'<tt>normal</tt>' label or the
1911'<tt>exception</tt>' label. If the callee function returns with the
1912"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1913"normal" label. If the callee (or any indirect callees) returns with the "<a
1914href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1915continued at the dynamically nearest "exception" label.</p>
1916
1917<h5>Arguments:</h5>
1918
1919<p>This instruction requires several arguments:</p>
1920
1921<ol>
1922 <li>
1923 The optional "cconv" marker indicates which <a href="#callingconv">calling
1924 convention</a> the call should use. If none is specified, the call defaults
1925 to using C calling conventions.
1926 </li>
1927 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1928 function value being invoked. In most cases, this is a direct function
1929 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1930 an arbitrary pointer to function value.
1931 </li>
1932
1933 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1934 function to be invoked. </li>
1935
1936 <li>'<tt>function args</tt>': argument list whose types match the function
1937 signature argument types. If the function signature indicates the function
1938 accepts a variable number of arguments, the extra arguments can be
1939 specified. </li>
1940
1941 <li>'<tt>normal label</tt>': the label reached when the called function
1942 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1943
1944 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1945 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1946
1947</ol>
1948
1949<h5>Semantics:</h5>
1950
1951<p>This instruction is designed to operate as a standard '<tt><a
1952href="#i_call">call</a></tt>' instruction in most regards. The primary
1953difference is that it establishes an association with a label, which is used by
1954the runtime library to unwind the stack.</p>
1955
1956<p>This instruction is used in languages with destructors to ensure that proper
1957cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1958exception. Additionally, this is important for implementation of
1959'<tt>catch</tt>' clauses in high-level languages that support them.</p>
1960
1961<h5>Example:</h5>
1962<pre>
1963 %retval = invoke i32 %Test(i32 15) to label %Continue
1964 unwind label %TestCleanup <i>; {i32}:retval set</i>
1965 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1966 unwind label %TestCleanup <i>; {i32}:retval set</i>
1967</pre>
1968</div>
1969
1970
1971<!-- _______________________________________________________________________ -->
1972
1973<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1974Instruction</a> </div>
1975
1976<div class="doc_text">
1977
1978<h5>Syntax:</h5>
1979<pre>
1980 unwind
1981</pre>
1982
1983<h5>Overview:</h5>
1984
1985<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1986at the first callee in the dynamic call stack which used an <a
1987href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1988primarily used to implement exception handling.</p>
1989
1990<h5>Semantics:</h5>
1991
1992<p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1993immediately halt. The dynamic call stack is then searched for the first <a
1994href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1995execution continues at the "exceptional" destination block specified by the
1996<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1997dynamic call chain, undefined behavior results.</p>
1998</div>
1999
2000<!-- _______________________________________________________________________ -->
2001
2002<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2003Instruction</a> </div>
2004
2005<div class="doc_text">
2006
2007<h5>Syntax:</h5>
2008<pre>
2009 unreachable
2010</pre>
2011
2012<h5>Overview:</h5>
2013
2014<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2015instruction is used to inform the optimizer that a particular portion of the
2016code is not reachable. This can be used to indicate that the code after a
2017no-return function cannot be reached, and other facts.</p>
2018
2019<h5>Semantics:</h5>
2020
2021<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2022</div>
2023
2024
2025
2026<!-- ======================================================================= -->
2027<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2028<div class="doc_text">
2029<p>Binary operators are used to do most of the computation in a
2030program. They require two operands, execute an operation on them, and
2031produce a single value. The operands might represent
2032multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
2033The result value of a binary operator is not
2034necessarily the same type as its operands.</p>
2035<p>There are several different binary operators:</p>
2036</div>
2037<!-- _______________________________________________________________________ -->
2038<div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
2039Instruction</a> </div>
2040<div class="doc_text">
2041<h5>Syntax:</h5>
2042<pre> &lt;result&gt; = add &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2043</pre>
2044<h5>Overview:</h5>
2045<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2046<h5>Arguments:</h5>
2047<p>The two arguments to the '<tt>add</tt>' instruction must be either <a
2048 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
2049 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2050Both arguments must have identical types.</p>
2051<h5>Semantics:</h5>
2052<p>The value produced is the integer or floating point sum of the two
2053operands.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame^]2054<p>If an integer sum has unsigned overflow, the result returned is the
2055mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2056the result.</p>
2057<p>Because LLVM integers use a two's complement representation, this
2058instruction is appropriate for both signed and unsigned integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2059<h5>Example:</h5>
2060<pre> &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2061</pre>
2062</div>
2063<!-- _______________________________________________________________________ -->
2064<div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
2065Instruction</a> </div>
2066<div class="doc_text">
2067<h5>Syntax:</h5>
2068<pre> &lt;result&gt; = sub &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2069</pre>
2070<h5>Overview:</h5>
2071<p>The '<tt>sub</tt>' instruction returns the difference of its two
2072operands.</p>
2073<p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
2074instruction present in most other intermediate representations.</p>
2075<h5>Arguments:</h5>
2076<p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
2077 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2078values.
2079This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2080Both arguments must have identical types.</p>
2081<h5>Semantics:</h5>
2082<p>The value produced is the integer or floating point difference of
2083the two operands.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame^]2084<p>If an integer difference has unsigned overflow, the result returned is the
2085mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2086the result.</p>
2087<p>Because LLVM integers use a two's complement representation, this
2088instruction is appropriate for both signed and unsigned integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2089<h5>Example:</h5>
2090<pre>
2091 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2092 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2093</pre>
2094</div>
2095<!-- _______________________________________________________________________ -->
2096<div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
2097Instruction</a> </div>
2098<div class="doc_text">
2099<h5>Syntax:</h5>
2100<pre> &lt;result&gt; = mul &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2101</pre>
2102<h5>Overview:</h5>
2103<p>The '<tt>mul</tt>' instruction returns the product of its two
2104operands.</p>
2105<h5>Arguments:</h5>
2106<p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
2107 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2108values.
2109This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2110Both arguments must have identical types.</p>
2111<h5>Semantics:</h5>
2112<p>The value produced is the integer or floating point product of the
2113two operands.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame^]2114<p>If the result of an integer multiplication has unsigned overflow,
2115the result returned is the mathematical result modulo
21162<sup>n</sup>, where n is the bit width of the result.</p>
2117<p>Because LLVM integers use a two's complement representation, and the
2118result is the same width as the operands, this instruction returns the
2119correct result for both signed and unsigned integers. If a full product
2120(e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2121should be sign-extended or zero-extended as appropriate to the
2122width of the full product.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2123<h5>Example:</h5>
2124<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2125</pre>
2126</div>
2127<!-- _______________________________________________________________________ -->
2128<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2129</a></div>
2130<div class="doc_text">
2131<h5>Syntax:</h5>
2132<pre> &lt;result&gt; = udiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2133</pre>
2134<h5>Overview:</h5>
2135<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2136operands.</p>
2137<h5>Arguments:</h5>
2138<p>The two arguments to the '<tt>udiv</tt>' instruction must be
2139<a href="#t_integer">integer</a> values. Both arguments must have identical
2140types. This instruction can also take <a href="#t_vector">vector</a> versions
2141of the values in which case the elements must be integers.</p>
2142<h5>Semantics:</h5>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame^]2143<p>The value produced is the unsigned integer quotient of the two operands.</p>
2144<p>Note that unsigned integer division and signed integer division are distinct
2145operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2146<p>Division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2147<h5>Example:</h5>
2148<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2149</pre>
2150</div>
2151<!-- _______________________________________________________________________ -->
2152<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2153</a> </div>
2154<div class="doc_text">
2155<h5>Syntax:</h5>
2156<pre> &lt;result&gt; = sdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2157</pre>
2158<h5>Overview:</h5>
2159<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2160operands.</p>
2161<h5>Arguments:</h5>
2162<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2163<a href="#t_integer">integer</a> values. Both arguments must have identical
2164types. This instruction can also take <a href="#t_vector">vector</a> versions
2165of the values in which case the elements must be integers.</p>
2166<h5>Semantics:</h5>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame^]2167<p>The value produced is the signed integer quotient of the two operands.</p>
2168<p>Note that signed integer division and unsigned integer division are distinct
2169operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2170<p>Division by zero leads to undefined behavior. Overflow also leads to
2171undefined behavior; this is a rare case, but can occur, for example,
2172by doing a 32-bit division of -2147483648 by -1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2173<h5>Example:</h5>
2174<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2175</pre>
2176</div>
2177<!-- _______________________________________________________________________ -->
2178<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2179Instruction</a> </div>
2180<div class="doc_text">
2181<h5>Syntax:</h5>
2182<pre> &lt;result&gt; = fdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2183</pre>
2184<h5>Overview:</h5>
2185<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2186operands.</p>
2187<h5>Arguments:</h5>
2188<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2189<a href="#t_floating">floating point</a> values. Both arguments must have
2190identical types. This instruction can also take <a href="#t_vector">vector</a>
2191versions of floating point values.</p>
2192<h5>Semantics:</h5>
2193<p>The value produced is the floating point quotient of the two operands.</p>
2194<h5>Example:</h5>
2195<pre> &lt;result&gt; = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2196</pre>
2197</div>
2198<!-- _______________________________________________________________________ -->
2199<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2200</div>
2201<div class="doc_text">
2202<h5>Syntax:</h5>
2203<pre> &lt;result&gt; = urem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2204</pre>
2205<h5>Overview:</h5>
2206<p>The '<tt>urem</tt>' instruction returns the remainder from the
2207unsigned division of its two arguments.</p>
2208<h5>Arguments:</h5>
2209<p>The two arguments to the '<tt>urem</tt>' instruction must be
2210<a href="#t_integer">integer</a> values. Both arguments must have identical
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2211types. This instruction can also take <a href="#t_vector">vector</a> versions
2212of the values in which case the elements must be integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2213<h5>Semantics:</h5>
2214<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2215This instruction always performs an unsigned division to get the remainder,
2216regardless of whether the arguments are unsigned or not.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame^]2217<p>Note that unsigned integer remainder and signed integer remainder are
2218distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2219<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2220<h5>Example:</h5>
2221<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2222</pre>
2223
2224</div>
2225<!-- _______________________________________________________________________ -->
2226<div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2227Instruction</a> </div>
2228<div class="doc_text">
2229<h5>Syntax:</h5>
2230<pre> &lt;result&gt; = srem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2231</pre>
2232<h5>Overview:</h5>
2233<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2234signed division of its two operands. This instruction can also take
2235<a href="#t_vector">vector</a> versions of the values in which case
2236the elements must be integers.</p>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]2237
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2238<h5>Arguments:</h5>
2239<p>The two arguments to the '<tt>srem</tt>' instruction must be
2240<a href="#t_integer">integer</a> values. Both arguments must have identical
2241types.</p>
2242<h5>Semantics:</h5>
2243<p>This instruction returns the <i>remainder</i> of a division (where the result
2244has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2245operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2246a value. For more information about the difference, see <a
2247 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2248Math Forum</a>. For a table of how this is implemented in various languages,
2249please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2250Wikipedia: modulo operation</a>.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame^]2251<p>Note that signed integer remainder and unsigned integer remainder are
2252distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2253<p>Taking the remainder of a division by zero leads to undefined behavior.
2254Overflow also leads to undefined behavior; this is a rare case, but can occur,
2255for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2256(The remainder doesn't actually overflow, but this rule lets srem be
2257implemented using instructions that return both the result of the division
2258and the remainder.)</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2259<h5>Example:</h5>
2260<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2261</pre>
2262
2263</div>
2264<!-- _______________________________________________________________________ -->
2265<div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2266Instruction</a> </div>
2267<div class="doc_text">
2268<h5>Syntax:</h5>
2269<pre> &lt;result&gt; = frem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2270</pre>
2271<h5>Overview:</h5>
2272<p>The '<tt>frem</tt>' instruction returns the remainder from the
2273division of its two operands.</p>
2274<h5>Arguments:</h5>
2275<p>The two arguments to the '<tt>frem</tt>' instruction must be
2276<a href="#t_floating">floating point</a> values. Both arguments must have
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2277identical types. This instruction can also take <a href="#t_vector">vector</a>
2278versions of floating point values.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2279<h5>Semantics:</h5>
2280<p>This instruction returns the <i>remainder</i> of a division.</p>
2281<h5>Example:</h5>
2282<pre> &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2283</pre>
2284</div>
2285
2286<!-- ======================================================================= -->
2287<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2288Operations</a> </div>
2289<div class="doc_text">
2290<p>Bitwise binary operators are used to do various forms of
2291bit-twiddling in a program. They are generally very efficient
2292instructions and can commonly be strength reduced from other
2293instructions. They require two operands, execute an operation on them,
2294and produce a single value. The resulting value of the bitwise binary
2295operators is always the same type as its first operand.</p>
2296</div>
2297
2298<!-- _______________________________________________________________________ -->
2299<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2300Instruction</a> </div>
2301<div class="doc_text">
2302<h5>Syntax:</h5>
2303<pre> &lt;result&gt; = shl &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2304</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2305
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2306<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2307
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2308<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2309the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2310
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2311<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2312
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2313<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2314 href="#t_integer">integer</a> type.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2315
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2316<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2317
2318<p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>. If
2319<tt>var2</tt> is (statically or dynamically) equal to or larger than the number
2320of bits in <tt>var1</tt>, the result is undefined.</p>
2321
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2322<h5>Example:</h5><pre>
2323 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2324 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2325 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2326 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2327</pre>
2328</div>
2329<!-- _______________________________________________________________________ -->
2330<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2331Instruction</a> </div>
2332<div class="doc_text">
2333<h5>Syntax:</h5>
2334<pre> &lt;result&gt; = lshr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2335</pre>
2336
2337<h5>Overview:</h5>
2338<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2339operand shifted to the right a specified number of bits with zero fill.</p>
2340
2341<h5>Arguments:</h5>
2342<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2343<a href="#t_integer">integer</a> type.</p>
2344
2345<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2346
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2347<p>This instruction always performs a logical shift right operation. The most
2348significant bits of the result will be filled with zero bits after the
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2349shift. If <tt>var2</tt> is (statically or dynamically) equal to or larger than
2350the number of bits in <tt>var1</tt>, the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2351
2352<h5>Example:</h5>
2353<pre>
2354 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2355 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2356 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2357 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2358 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2359</pre>
2360</div>
2361
2362<!-- _______________________________________________________________________ -->
2363<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2364Instruction</a> </div>
2365<div class="doc_text">
2366
2367<h5>Syntax:</h5>
2368<pre> &lt;result&gt; = ashr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2369</pre>
2370
2371<h5>Overview:</h5>
2372<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2373operand shifted to the right a specified number of bits with sign extension.</p>
2374
2375<h5>Arguments:</h5>
2376<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2377<a href="#t_integer">integer</a> type.</p>
2378
2379<h5>Semantics:</h5>
2380<p>This instruction always performs an arithmetic shift right operation,
2381The most significant bits of the result will be filled with the sign bit
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2382of <tt>var1</tt>. If <tt>var2</tt> is (statically or dynamically) equal to or
2383larger than the number of bits in <tt>var1</tt>, the result is undefined.
2384</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2385
2386<h5>Example:</h5>
2387<pre>
2388 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2389 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2390 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2391 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2392 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2393</pre>
2394</div>
2395
2396<!-- _______________________________________________________________________ -->
2397<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2398Instruction</a> </div>
2399<div class="doc_text">
2400<h5>Syntax:</h5>
2401<pre> &lt;result&gt; = and &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2402</pre>
2403<h5>Overview:</h5>
2404<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2405its two operands.</p>
2406<h5>Arguments:</h5>
2407<p>The two arguments to the '<tt>and</tt>' instruction must be <a
2408 href="#t_integer">integer</a> values. Both arguments must have
2409identical types.</p>
2410<h5>Semantics:</h5>
2411<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2412<p> </p>
2413<div style="align: center">
2414<table border="1" cellspacing="0" cellpadding="4">
2415 <tbody>
2416 <tr>
2417 <td>In0</td>
2418 <td>In1</td>
2419 <td>Out</td>
2420 </tr>
2421 <tr>
2422 <td>0</td>
2423 <td>0</td>
2424 <td>0</td>
2425 </tr>
2426 <tr>
2427 <td>0</td>
2428 <td>1</td>
2429 <td>0</td>
2430 </tr>
2431 <tr>
2432 <td>1</td>
2433 <td>0</td>
2434 <td>0</td>
2435 </tr>
2436 <tr>
2437 <td>1</td>
2438 <td>1</td>
2439 <td>1</td>
2440 </tr>
2441 </tbody>
2442</table>
2443</div>
2444<h5>Example:</h5>
2445<pre> &lt;result&gt; = and i32 4, %var <i>; yields {i32}:result = 4 &amp; %var</i>
2446 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2447 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2448</pre>
2449</div>
2450<!-- _______________________________________________________________________ -->
2451<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2452<div class="doc_text">
2453<h5>Syntax:</h5>
2454<pre> &lt;result&gt; = or &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2455</pre>
2456<h5>Overview:</h5>
2457<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2458or of its two operands.</p>
2459<h5>Arguments:</h5>
2460<p>The two arguments to the '<tt>or</tt>' instruction must be <a
2461 href="#t_integer">integer</a> values. Both arguments must have
2462identical types.</p>
2463<h5>Semantics:</h5>
2464<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2465<p> </p>
2466<div style="align: center">
2467<table border="1" cellspacing="0" cellpadding="4">
2468 <tbody>
2469 <tr>
2470 <td>In0</td>
2471 <td>In1</td>
2472 <td>Out</td>
2473 </tr>
2474 <tr>
2475 <td>0</td>
2476 <td>0</td>
2477 <td>0</td>
2478 </tr>
2479 <tr>
2480 <td>0</td>
2481 <td>1</td>
2482 <td>1</td>
2483 </tr>
2484 <tr>
2485 <td>1</td>
2486 <td>0</td>
2487 <td>1</td>
2488 </tr>
2489 <tr>
2490 <td>1</td>
2491 <td>1</td>
2492 <td>1</td>
2493 </tr>
2494 </tbody>
2495</table>
2496</div>
2497<h5>Example:</h5>
2498<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2499 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2500 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2501</pre>
2502</div>
2503<!-- _______________________________________________________________________ -->
2504<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2505Instruction</a> </div>
2506<div class="doc_text">
2507<h5>Syntax:</h5>
2508<pre> &lt;result&gt; = xor &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2509</pre>
2510<h5>Overview:</h5>
2511<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2512or of its two operands. The <tt>xor</tt> is used to implement the
2513"one's complement" operation, which is the "~" operator in C.</p>
2514<h5>Arguments:</h5>
2515<p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2516 href="#t_integer">integer</a> values. Both arguments must have
2517identical types.</p>
2518<h5>Semantics:</h5>
2519<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2520<p> </p>
2521<div style="align: center">
2522<table border="1" cellspacing="0" cellpadding="4">
2523 <tbody>
2524 <tr>
2525 <td>In0</td>
2526 <td>In1</td>
2527 <td>Out</td>
2528 </tr>
2529 <tr>
2530 <td>0</td>
2531 <td>0</td>
2532 <td>0</td>
2533 </tr>
2534 <tr>
2535 <td>0</td>
2536 <td>1</td>
2537 <td>1</td>
2538 </tr>
2539 <tr>
2540 <td>1</td>
2541 <td>0</td>
2542 <td>1</td>
2543 </tr>
2544 <tr>
2545 <td>1</td>
2546 <td>1</td>
2547 <td>0</td>
2548 </tr>
2549 </tbody>
2550</table>
2551</div>
2552<p> </p>
2553<h5>Example:</h5>
2554<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2555 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2556 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2557 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2558</pre>
2559</div>
2560
2561<!-- ======================================================================= -->
2562<div class="doc_subsection">
2563 <a name="vectorops">Vector Operations</a>
2564</div>
2565
2566<div class="doc_text">
2567
2568<p>LLVM supports several instructions to represent vector operations in a
2569target-independent manner. These instructions cover the element-access and
2570vector-specific operations needed to process vectors effectively. While LLVM
2571does directly support these vector operations, many sophisticated algorithms
2572will want to use target-specific intrinsics to take full advantage of a specific
2573target.</p>
2574
2575</div>
2576
2577<!-- _______________________________________________________________________ -->
2578<div class="doc_subsubsection">
2579 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2580</div>
2581
2582<div class="doc_text">
2583
2584<h5>Syntax:</h5>
2585
2586<pre>
2587 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
2588</pre>
2589
2590<h5>Overview:</h5>
2591
2592<p>
2593The '<tt>extractelement</tt>' instruction extracts a single scalar
2594element from a vector at a specified index.
2595</p>
2596
2597
2598<h5>Arguments:</h5>
2599
2600<p>
2601The first operand of an '<tt>extractelement</tt>' instruction is a
2602value of <a href="#t_vector">vector</a> type. The second operand is
2603an index indicating the position from which to extract the element.
2604The index may be a variable.</p>
2605
2606<h5>Semantics:</h5>
2607
2608<p>
2609The result is a scalar of the same type as the element type of
2610<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2611<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2612results are undefined.
2613</p>
2614
2615<h5>Example:</h5>
2616
2617<pre>
2618 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
2619</pre>
2620</div>
2621
2622
2623<!-- _______________________________________________________________________ -->
2624<div class="doc_subsubsection">
2625 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2626</div>
2627
2628<div class="doc_text">
2629
2630<h5>Syntax:</h5>
2631
2632<pre>
2633 &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>
2634</pre>
2635
2636<h5>Overview:</h5>
2637
2638<p>
2639The '<tt>insertelement</tt>' instruction inserts a scalar
2640element into a vector at a specified index.
2641</p>
2642
2643
2644<h5>Arguments:</h5>
2645
2646<p>
2647The first operand of an '<tt>insertelement</tt>' instruction is a
2648value of <a href="#t_vector">vector</a> type. The second operand is a
2649scalar value whose type must equal the element type of the first
2650operand. The third operand is an index indicating the position at
2651which to insert the value. The index may be a variable.</p>
2652
2653<h5>Semantics:</h5>
2654
2655<p>
2656The result is a vector of the same type as <tt>val</tt>. Its
2657element values are those of <tt>val</tt> except at position
2658<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2659exceeds the length of <tt>val</tt>, the results are undefined.
2660</p>
2661
2662<h5>Example:</h5>
2663
2664<pre>
2665 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
2666</pre>
2667</div>
2668
2669<!-- _______________________________________________________________________ -->
2670<div class="doc_subsubsection">
2671 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2672</div>
2673
2674<div class="doc_text">
2675
2676<h5>Syntax:</h5>
2677
2678<pre>
2679 &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>
2680</pre>
2681
2682<h5>Overview:</h5>
2683
2684<p>
2685The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2686from two input vectors, returning a vector of the same type.
2687</p>
2688
2689<h5>Arguments:</h5>
2690
2691<p>
2692The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2693with types that match each other and types that match the result of the
2694instruction. The third argument is a shuffle mask, which has the same number
2695of elements as the other vector type, but whose element type is always 'i32'.
2696</p>
2697
2698<p>
2699The shuffle mask operand is required to be a constant vector with either
2700constant integer or undef values.
2701</p>
2702
2703<h5>Semantics:</h5>
2704
2705<p>
2706The elements of the two input vectors are numbered from left to right across
2707both of the vectors. The shuffle mask operand specifies, for each element of
2708the result vector, which element of the two input registers the result element
2709gets. The element selector may be undef (meaning "don't care") and the second
2710operand may be undef if performing a shuffle from only one vector.
2711</p>
2712
2713<h5>Example:</h5>
2714
2715<pre>
2716 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
2717 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
2718 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
2719 &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.
2720</pre>
2721</div>
2722
2723
2724<!-- ======================================================================= -->
2725<div class="doc_subsection">
2726 <a name="memoryops">Memory Access and Addressing Operations</a>
2727</div>
2728
2729<div class="doc_text">
2730
2731<p>A key design point of an SSA-based representation is how it
2732represents memory. In LLVM, no memory locations are in SSA form, which
2733makes things very simple. This section describes how to read, write,
2734allocate, and free memory in LLVM.</p>
2735
2736</div>
2737
2738<!-- _______________________________________________________________________ -->
2739<div class="doc_subsubsection">
2740 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2741</div>
2742
2743<div class="doc_text">
2744
2745<h5>Syntax:</h5>
2746
2747<pre>
2748 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
2749</pre>
2750
2751<h5>Overview:</h5>
2752
2753<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]2754heap and returns a pointer to it. The object is always allocated in the generic
2755address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2756
2757<h5>Arguments:</h5>
2758
2759<p>The '<tt>malloc</tt>' instruction allocates
2760<tt>sizeof(&lt;type&gt;)*NumElements</tt>
2761bytes of memory from the operating system and returns a pointer of the
2762appropriate type to the program. If "NumElements" is specified, it is the
2763number of elements allocated. If an alignment is specified, the value result
2764of the allocation is guaranteed to be aligned to at least that boundary. If
2765not specified, or if zero, the target can choose to align the allocation on any
2766convenient boundary.</p>
2767
2768<p>'<tt>type</tt>' must be a sized type.</p>
2769
2770<h5>Semantics:</h5>
2771
2772<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2773a pointer is returned.</p>
2774
2775<h5>Example:</h5>
2776
2777<pre>
2778 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2779
2780 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2781 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2782 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2783 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2784 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2785</pre>
2786</div>
2787
2788<!-- _______________________________________________________________________ -->
2789<div class="doc_subsubsection">
2790 <a name="i_free">'<tt>free</tt>' Instruction</a>
2791</div>
2792
2793<div class="doc_text">
2794
2795<h5>Syntax:</h5>
2796
2797<pre>
2798 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
2799</pre>
2800
2801<h5>Overview:</h5>
2802
2803<p>The '<tt>free</tt>' instruction returns memory back to the unused
2804memory heap to be reallocated in the future.</p>
2805
2806<h5>Arguments:</h5>
2807
2808<p>'<tt>value</tt>' shall be a pointer value that points to a value
2809that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2810instruction.</p>
2811
2812<h5>Semantics:</h5>
2813
2814<p>Access to the memory pointed to by the pointer is no longer defined
2815after this instruction executes.</p>
2816
2817<h5>Example:</h5>
2818
2819<pre>
2820 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2821 free [4 x i8]* %array
2822</pre>
2823</div>
2824
2825<!-- _______________________________________________________________________ -->
2826<div class="doc_subsubsection">
2827 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2828</div>
2829
2830<div class="doc_text">
2831
2832<h5>Syntax:</h5>
2833
2834<pre>
2835 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
2836</pre>
2837
2838<h5>Overview:</h5>
2839
2840<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
2841currently executing function, to be automatically released when this function
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]2842returns to its caller. The object is always allocated in the generic address
2843space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2844
2845<h5>Arguments:</h5>
2846
2847<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
2848bytes of memory on the runtime stack, returning a pointer of the
2849appropriate type to the program. If "NumElements" is specified, it is the
2850number of elements allocated. If an alignment is specified, the value result
2851of the allocation is guaranteed to be aligned to at least that boundary. If
2852not specified, or if zero, the target can choose to align the allocation on any
2853convenient boundary.</p>
2854
2855<p>'<tt>type</tt>' may be any sized type.</p>
2856
2857<h5>Semantics:</h5>
2858
2859<p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2860memory is automatically released when the function returns. The '<tt>alloca</tt>'
2861instruction is commonly used to represent automatic variables that must
2862have an address available. When the function returns (either with the <tt><a
2863 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2864instructions), the memory is reclaimed.</p>
2865
2866<h5>Example:</h5>
2867
2868<pre>
2869 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2870 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2871 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2872 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2873</pre>
2874</div>
2875
2876<!-- _______________________________________________________________________ -->
2877<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2878Instruction</a> </div>
2879<div class="doc_text">
2880<h5>Syntax:</h5>
2881<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>
2882<h5>Overview:</h5>
2883<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2884<h5>Arguments:</h5>
2885<p>The argument to the '<tt>load</tt>' instruction specifies the memory
2886address from which to load. The pointer must point to a <a
2887 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2888marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2889the number or order of execution of this <tt>load</tt> with other
2890volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2891instructions. </p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]2892<p>
2893The optional "align" argument specifies the alignment of the operation
2894(that is, the alignment of the memory address). A value of 0 or an
2895omitted "align" argument means that the operation has the preferential
2896alignment for the target. It is the responsibility of the code emitter
2897to ensure that the alignment information is correct. Overestimating
2898the alignment results in an undefined behavior. Underestimating the
2899alignment may produce less efficient code. An alignment of 1 is always
2900safe.
2901</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2902<h5>Semantics:</h5>
2903<p>The location of memory pointed to is loaded.</p>
2904<h5>Examples:</h5>
2905<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2906 <a
2907 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2908 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2909</pre>
2910</div>
2911<!-- _______________________________________________________________________ -->
2912<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2913Instruction</a> </div>
2914<div class="doc_text">
2915<h5>Syntax:</h5>
2916<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
2917 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
2918</pre>
2919<h5>Overview:</h5>
2920<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2921<h5>Arguments:</h5>
2922<p>There are two arguments to the '<tt>store</tt>' instruction: a value
2923to store and an address at which to store it. The type of the '<tt>&lt;pointer&gt;</tt>'
2924operand must be a pointer to the type of the '<tt>&lt;value&gt;</tt>'
2925operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2926optimizer is not allowed to modify the number or order of execution of
2927this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2928 href="#i_store">store</a></tt> instructions.</p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]2929<p>
2930The optional "align" argument specifies the alignment of the operation
2931(that is, the alignment of the memory address). A value of 0 or an
2932omitted "align" argument means that the operation has the preferential
2933alignment for the target. It is the responsibility of the code emitter
2934to ensure that the alignment information is correct. Overestimating
2935the alignment results in an undefined behavior. Underestimating the
2936alignment may produce less efficient code. An alignment of 1 is always
2937safe.
2938</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2939<h5>Semantics:</h5>
2940<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
2941at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
2942<h5>Example:</h5>
2943<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]2944 store i32 3, i32* %ptr <i>; yields {void}</i>
2945 %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]2946</pre>
2947</div>
2948
2949<!-- _______________________________________________________________________ -->
2950<div class="doc_subsubsection">
2951 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2952</div>
2953
2954<div class="doc_text">
2955<h5>Syntax:</h5>
2956<pre>
2957 &lt;result&gt; = getelementptr &lt;ty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
2958</pre>
2959
2960<h5>Overview:</h5>
2961
2962<p>
2963The '<tt>getelementptr</tt>' instruction is used to get the address of a
2964subelement of an aggregate data structure.</p>
2965
2966<h5>Arguments:</h5>
2967
2968<p>This instruction takes a list of integer operands that indicate what
2969elements of the aggregate object to index to. The actual types of the arguments
2970provided depend on the type of the first pointer argument. The
2971'<tt>getelementptr</tt>' instruction is used to index down through the type
2972levels of a structure or to a specific index in an array. When indexing into a
2973structure, only <tt>i32</tt> integer constants are allowed. When indexing
2974into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2975be sign extended to 64-bit values.</p>
2976
2977<p>For example, let's consider a C code fragment and how it gets
2978compiled to LLVM:</p>
2979
2980<div class="doc_code">
2981<pre>
2982struct RT {
2983 char A;
2984 int B[10][20];
2985 char C;
2986};
2987struct ST {
2988 int X;
2989 double Y;
2990 struct RT Z;
2991};
2992
2993int *foo(struct ST *s) {
2994 return &amp;s[1].Z.B[5][13];
2995}
2996</pre>
2997</div>
2998
2999<p>The LLVM code generated by the GCC frontend is:</p>
3000
3001<div class="doc_code">
3002<pre>
3003%RT = type { i8 , [10 x [20 x i32]], i8 }
3004%ST = type { i32, double, %RT }
3005
3006define i32* %foo(%ST* %s) {
3007entry:
3008 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3009 ret i32* %reg
3010}
3011</pre>
3012</div>
3013
3014<h5>Semantics:</h5>
3015
3016<p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
3017on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
3018and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
3019<a href="#t_integer">integer</a> type but the value will always be sign extended
3020to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
3021<b>constants</b>.</p>
3022
3023<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3024type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3025}</tt>' type, a structure. The second index indexes into the third element of
3026the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3027i8 }</tt>' type, another structure. The third index indexes into the second
3028element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3029array. The two dimensions of the array are subscripted into, yielding an
3030'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3031to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3032
3033<p>Note that it is perfectly legal to index partially through a
3034structure, returning a pointer to an inner element. Because of this,
3035the LLVM code for the given testcase is equivalent to:</p>
3036
3037<pre>
3038 define i32* %foo(%ST* %s) {
3039 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3040 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3041 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3042 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3043 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3044 ret i32* %t5
3045 }
3046</pre>
3047
3048<p>Note that it is undefined to access an array out of bounds: array and
3049pointer indexes must always be within the defined bounds of the array type.
3050The one exception for this rules is zero length arrays. These arrays are
3051defined to be accessible as variable length arrays, which requires access
3052beyond the zero'th element.</p>
3053
3054<p>The getelementptr instruction is often confusing. For some more insight
3055into how it works, see <a href="GetElementPtr.html">the getelementptr
3056FAQ</a>.</p>
3057
3058<h5>Example:</h5>
3059
3060<pre>
3061 <i>; yields [12 x i8]*:aptr</i>
3062 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
3063</pre>
3064</div>
3065
3066<!-- ======================================================================= -->
3067<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3068</div>
3069<div class="doc_text">
3070<p>The instructions in this category are the conversion instructions (casting)
3071which all take a single operand and a type. They perform various bit conversions
3072on the operand.</p>
3073</div>
3074
3075<!-- _______________________________________________________________________ -->
3076<div class="doc_subsubsection">
3077 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3078</div>
3079<div class="doc_text">
3080
3081<h5>Syntax:</h5>
3082<pre>
3083 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3084</pre>
3085
3086<h5>Overview:</h5>
3087<p>
3088The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3089</p>
3090
3091<h5>Arguments:</h5>
3092<p>
3093The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3094be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3095and type of the result, which must be an <a href="#t_integer">integer</a>
3096type. The bit size of <tt>value</tt> must be larger than the bit size of
3097<tt>ty2</tt>. Equal sized types are not allowed.</p>
3098
3099<h5>Semantics:</h5>
3100<p>
3101The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3102and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3103larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3104It will always truncate bits.</p>
3105
3106<h5>Example:</h5>
3107<pre>
3108 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3109 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3110 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3111</pre>
3112</div>
3113
3114<!-- _______________________________________________________________________ -->
3115<div class="doc_subsubsection">
3116 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3117</div>
3118<div class="doc_text">
3119
3120<h5>Syntax:</h5>
3121<pre>
3122 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3123</pre>
3124
3125<h5>Overview:</h5>
3126<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3127<tt>ty2</tt>.</p>
3128
3129
3130<h5>Arguments:</h5>
3131<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3132<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3133also be of <a href="#t_integer">integer</a> type. The bit size of the
3134<tt>value</tt> must be smaller than the bit size of the destination type,
3135<tt>ty2</tt>.</p>
3136
3137<h5>Semantics:</h5>
3138<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3139bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3140
3141<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3142
3143<h5>Example:</h5>
3144<pre>
3145 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3146 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3147</pre>
3148</div>
3149
3150<!-- _______________________________________________________________________ -->
3151<div class="doc_subsubsection">
3152 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3153</div>
3154<div class="doc_text">
3155
3156<h5>Syntax:</h5>
3157<pre>
3158 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3159</pre>
3160
3161<h5>Overview:</h5>
3162<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3163
3164<h5>Arguments:</h5>
3165<p>
3166The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3167<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3168also be of <a href="#t_integer">integer</a> type. The bit size of the
3169<tt>value</tt> must be smaller than the bit size of the destination type,
3170<tt>ty2</tt>.</p>
3171
3172<h5>Semantics:</h5>
3173<p>
3174The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3175bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3176the type <tt>ty2</tt>.</p>
3177
3178<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3179
3180<h5>Example:</h5>
3181<pre>
3182 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3183 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3184</pre>
3185</div>
3186
3187<!-- _______________________________________________________________________ -->
3188<div class="doc_subsubsection">
3189 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3190</div>
3191
3192<div class="doc_text">
3193
3194<h5>Syntax:</h5>
3195
3196<pre>
3197 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3198</pre>
3199
3200<h5>Overview:</h5>
3201<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3202<tt>ty2</tt>.</p>
3203
3204
3205<h5>Arguments:</h5>
3206<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3207 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3208cast it to. The size of <tt>value</tt> must be larger than the size of
3209<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3210<i>no-op cast</i>.</p>
3211
3212<h5>Semantics:</h5>
3213<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3214<a href="#t_floating">floating point</a> type to a smaller
3215<a href="#t_floating">floating point</a> type. If the value cannot fit within
3216the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3217
3218<h5>Example:</h5>
3219<pre>
3220 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3221 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3222</pre>
3223</div>
3224
3225<!-- _______________________________________________________________________ -->
3226<div class="doc_subsubsection">
3227 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3228</div>
3229<div class="doc_text">
3230
3231<h5>Syntax:</h5>
3232<pre>
3233 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3234</pre>
3235
3236<h5>Overview:</h5>
3237<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3238floating point value.</p>
3239
3240<h5>Arguments:</h5>
3241<p>The '<tt>fpext</tt>' instruction takes a
3242<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3243and a <a href="#t_floating">floating point</a> type to cast it to. The source
3244type must be smaller than the destination type.</p>
3245
3246<h5>Semantics:</h5>
3247<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3248<a href="#t_floating">floating point</a> type to a larger
3249<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3250used to make a <i>no-op cast</i> because it always changes bits. Use
3251<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3252
3253<h5>Example:</h5>
3254<pre>
3255 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3256 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3257</pre>
3258</div>
3259
3260<!-- _______________________________________________________________________ -->
3261<div class="doc_subsubsection">
3262 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3263</div>
3264<div class="doc_text">
3265
3266<h5>Syntax:</h5>
3267<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3268 &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]3269</pre>
3270
3271<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3272<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3273unsigned integer equivalent of type <tt>ty2</tt>.
3274</p>
3275
3276<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3277<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]3278scalar or vector <a href="#t_floating">floating point</a> value, and a type
3279to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3280type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3281vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3282
3283<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3284<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3285<a href="#t_floating">floating point</a> operand into the nearest (rounding
3286towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3287the results are undefined.</p>
3288
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3289<h5>Example:</h5>
3290<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3291 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3292 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3293 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3294</pre>
3295</div>
3296
3297<!-- _______________________________________________________________________ -->
3298<div class="doc_subsubsection">
3299 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3300</div>
3301<div class="doc_text">
3302
3303<h5>Syntax:</h5>
3304<pre>
3305 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3306</pre>
3307
3308<h5>Overview:</h5>
3309<p>The '<tt>fptosi</tt>' instruction converts
3310<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3311</p>
3312
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3313<h5>Arguments:</h5>
3314<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]3315scalar or vector <a href="#t_floating">floating point</a> value, and a type
3316to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3317type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3318vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3319
3320<h5>Semantics:</h5>
3321<p>The '<tt>fptosi</tt>' instruction converts its
3322<a href="#t_floating">floating point</a> operand into the nearest (rounding
3323towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3324the results are undefined.</p>
3325
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3326<h5>Example:</h5>
3327<pre>
3328 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3329 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3330 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3331</pre>
3332</div>
3333
3334<!-- _______________________________________________________________________ -->
3335<div class="doc_subsubsection">
3336 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3337</div>
3338<div class="doc_text">
3339
3340<h5>Syntax:</h5>
3341<pre>
3342 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3343</pre>
3344
3345<h5>Overview:</h5>
3346<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3347integer and converts that value to the <tt>ty2</tt> type.</p>
3348
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3349<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3350<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3351scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3352to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3353type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3354floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3355
3356<h5>Semantics:</h5>
3357<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3358integer quantity and converts it to the corresponding floating point value. If
3359the value cannot fit in the floating point value, the results are undefined.</p>
3360
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3361<h5>Example:</h5>
3362<pre>
3363 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3364 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3365</pre>
3366</div>
3367
3368<!-- _______________________________________________________________________ -->
3369<div class="doc_subsubsection">
3370 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3371</div>
3372<div class="doc_text">
3373
3374<h5>Syntax:</h5>
3375<pre>
3376 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3377</pre>
3378
3379<h5>Overview:</h5>
3380<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3381integer and converts that value to the <tt>ty2</tt> type.</p>
3382
3383<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3384<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3385scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3386to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3387type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3388floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3389
3390<h5>Semantics:</h5>
3391<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3392integer quantity and converts it to the corresponding floating point value. If
3393the value cannot fit in the floating point value, the results are undefined.</p>
3394
3395<h5>Example:</h5>
3396<pre>
3397 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3398 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3399</pre>
3400</div>
3401
3402<!-- _______________________________________________________________________ -->
3403<div class="doc_subsubsection">
3404 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3405</div>
3406<div class="doc_text">
3407
3408<h5>Syntax:</h5>
3409<pre>
3410 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3411</pre>
3412
3413<h5>Overview:</h5>
3414<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3415the integer type <tt>ty2</tt>.</p>
3416
3417<h5>Arguments:</h5>
3418<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3419must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3420<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3421
3422<h5>Semantics:</h5>
3423<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3424<tt>ty2</tt> by interpreting the pointer value as an integer and either
3425truncating or zero extending that value to the size of the integer type. If
3426<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3427<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3428are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3429change.</p>
3430
3431<h5>Example:</h5>
3432<pre>
3433 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3434 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3435</pre>
3436</div>
3437
3438<!-- _______________________________________________________________________ -->
3439<div class="doc_subsubsection">
3440 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3441</div>
3442<div class="doc_text">
3443
3444<h5>Syntax:</h5>
3445<pre>
3446 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3447</pre>
3448
3449<h5>Overview:</h5>
3450<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3451a pointer type, <tt>ty2</tt>.</p>
3452
3453<h5>Arguments:</h5>
3454<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3455value to cast, and a type to cast it to, which must be a
3456<a href="#t_pointer">pointer</a> type.
3457
3458<h5>Semantics:</h5>
3459<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3460<tt>ty2</tt> by applying either a zero extension or a truncation depending on
3461the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3462size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3463the size of a pointer then a zero extension is done. If they are the same size,
3464nothing is done (<i>no-op cast</i>).</p>
3465
3466<h5>Example:</h5>
3467<pre>
3468 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3469 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3470 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3471</pre>
3472</div>
3473
3474<!-- _______________________________________________________________________ -->
3475<div class="doc_subsubsection">
3476 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3477</div>
3478<div class="doc_text">
3479
3480<h5>Syntax:</h5>
3481<pre>
3482 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3483</pre>
3484
3485<h5>Overview:</h5>
3486<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3487<tt>ty2</tt> without changing any bits.</p>
3488
3489<h5>Arguments:</h5>
3490<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3491a first class value, and a type to cast it to, which must also be a <a
3492 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3493and the destination type, <tt>ty2</tt>, must be identical. If the source
3494type is a pointer, the destination type must also be a pointer.</p>
3495
3496<h5>Semantics:</h5>
3497<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3498<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3499this conversion. The conversion is done as if the <tt>value</tt> had been
3500stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3501converted to other pointer types with this instruction. To convert pointers to
3502other types, use the <a href="#i_inttoptr">inttoptr</a> or
3503<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3504
3505<h5>Example:</h5>
3506<pre>
3507 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3508 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3509 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3510</pre>
3511</div>
3512
3513<!-- ======================================================================= -->
3514<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3515<div class="doc_text">
3516<p>The instructions in this category are the "miscellaneous"
3517instructions, which defy better classification.</p>
3518</div>
3519
3520<!-- _______________________________________________________________________ -->
3521<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3522</div>
3523<div class="doc_text">
3524<h5>Syntax:</h5>
3525<pre> &lt;result&gt; = icmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3526</pre>
3527<h5>Overview:</h5>
3528<p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3529of its two integer operands.</p>
3530<h5>Arguments:</h5>
3531<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3532the condition code indicating the kind of comparison to perform. It is not
3533a value, just a keyword. The possible condition code are:
3534<ol>
3535 <li><tt>eq</tt>: equal</li>
3536 <li><tt>ne</tt>: not equal </li>
3537 <li><tt>ugt</tt>: unsigned greater than</li>
3538 <li><tt>uge</tt>: unsigned greater or equal</li>
3539 <li><tt>ult</tt>: unsigned less than</li>
3540 <li><tt>ule</tt>: unsigned less or equal</li>
3541 <li><tt>sgt</tt>: signed greater than</li>
3542 <li><tt>sge</tt>: signed greater or equal</li>
3543 <li><tt>slt</tt>: signed less than</li>
3544 <li><tt>sle</tt>: signed less or equal</li>
3545</ol>
3546<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3547<a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3548<h5>Semantics:</h5>
3549<p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3550the condition code given as <tt>cond</tt>. The comparison performed always
3551yields a <a href="#t_primitive">i1</a> result, as follows:
3552<ol>
3553 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3554 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3555 </li>
3556 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3557 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3558 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3559 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3560 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3561 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3562 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3563 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3564 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3565 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3566 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3567 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3568 <li><tt>sge</tt>: interprets the operands as signed values and yields
3569 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3570 <li><tt>slt</tt>: interprets the operands as signed values and yields
3571 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3572 <li><tt>sle</tt>: interprets the operands as signed values and yields
3573 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3574</ol>
3575<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3576values are compared as if they were integers.</p>
3577
3578<h5>Example:</h5>
3579<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
3580 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
3581 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
3582 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3583 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3584 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
3585</pre>
3586</div>
3587
3588<!-- _______________________________________________________________________ -->
3589<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3590</div>
3591<div class="doc_text">
3592<h5>Syntax:</h5>
3593<pre> &lt;result&gt; = fcmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3594</pre>
3595<h5>Overview:</h5>
3596<p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3597of its floating point operands.</p>
3598<h5>Arguments:</h5>
3599<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3600the condition code indicating the kind of comparison to perform. It is not
3601a value, just a keyword. The possible condition code are:
3602<ol>
3603 <li><tt>false</tt>: no comparison, always returns false</li>
3604 <li><tt>oeq</tt>: ordered and equal</li>
3605 <li><tt>ogt</tt>: ordered and greater than </li>
3606 <li><tt>oge</tt>: ordered and greater than or equal</li>
3607 <li><tt>olt</tt>: ordered and less than </li>
3608 <li><tt>ole</tt>: ordered and less than or equal</li>
3609 <li><tt>one</tt>: ordered and not equal</li>
3610 <li><tt>ord</tt>: ordered (no nans)</li>
3611 <li><tt>ueq</tt>: unordered or equal</li>
3612 <li><tt>ugt</tt>: unordered or greater than </li>
3613 <li><tt>uge</tt>: unordered or greater than or equal</li>
3614 <li><tt>ult</tt>: unordered or less than </li>
3615 <li><tt>ule</tt>: unordered or less than or equal</li>
3616 <li><tt>une</tt>: unordered or not equal</li>
3617 <li><tt>uno</tt>: unordered (either nans)</li>
3618 <li><tt>true</tt>: no comparison, always returns true</li>
3619</ol>
3620<p><i>Ordered</i> means that neither operand is a QNAN while
3621<i>unordered</i> means that either operand may be a QNAN.</p>
3622<p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3623<a href="#t_floating">floating point</a> typed. They must have identical
3624types.</p>
3625<h5>Semantics:</h5>
3626<p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3627the condition code given as <tt>cond</tt>. The comparison performed always
3628yields a <a href="#t_primitive">i1</a> result, as follows:
3629<ol>
3630 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3631 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3632 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3633 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3634 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3635 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3636 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3637 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3638 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3639 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3640 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3641 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3642 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3643 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3644 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3645 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3646 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3647 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3648 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3649 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3650 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3651 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3652 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3653 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3654 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3655 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3656 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3657 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3658</ol>
3659
3660<h5>Example:</h5>
3661<pre> &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3662 &lt;result&gt; = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3663 &lt;result&gt; = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3664 &lt;result&gt; = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3665</pre>
3666</div>
3667
3668<!-- _______________________________________________________________________ -->
3669<div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3670Instruction</a> </div>
3671<div class="doc_text">
3672<h5>Syntax:</h5>
3673<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
3674<h5>Overview:</h5>
3675<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
3676the SSA graph representing the function.</p>
3677<h5>Arguments:</h5>
3678<p>The type of the incoming values is specified with the first type
3679field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3680as arguments, with one pair for each predecessor basic block of the
3681current block. Only values of <a href="#t_firstclass">first class</a>
3682type may be used as the value arguments to the PHI node. Only labels
3683may be used as the label arguments.</p>
3684<p>There must be no non-phi instructions between the start of a basic
3685block and the PHI instructions: i.e. PHI instructions must be first in
3686a basic block.</p>
3687<h5>Semantics:</h5>
3688<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3689specified by the pair corresponding to the predecessor basic block that executed
3690just prior to the current block.</p>
3691<h5>Example:</h5>
3692<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>
3693</div>
3694
3695<!-- _______________________________________________________________________ -->
3696<div class="doc_subsubsection">
3697 <a name="i_select">'<tt>select</tt>' Instruction</a>
3698</div>
3699
3700<div class="doc_text">
3701
3702<h5>Syntax:</h5>
3703
3704<pre>
3705 &lt;result&gt; = select i1 &lt;cond&gt;, &lt;ty&gt; &lt;val1&gt;, &lt;ty&gt; &lt;val2&gt; <i>; yields ty</i>
3706</pre>
3707
3708<h5>Overview:</h5>
3709
3710<p>
3711The '<tt>select</tt>' instruction is used to choose one value based on a
3712condition, without branching.
3713</p>
3714
3715
3716<h5>Arguments:</h5>
3717
3718<p>
3719The '<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.
3720</p>
3721
3722<h5>Semantics:</h5>
3723
3724<p>
3725If the boolean condition evaluates to true, the instruction returns the first
3726value argument; otherwise, it returns the second value argument.
3727</p>
3728
3729<h5>Example:</h5>
3730
3731<pre>
3732 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3733</pre>
3734</div>
3735
3736
3737<!-- _______________________________________________________________________ -->
3738<div class="doc_subsubsection">
3739 <a name="i_call">'<tt>call</tt>' Instruction</a>
3740</div>
3741
3742<div class="doc_text">
3743
3744<h5>Syntax:</h5>
3745<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3746 &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]3747</pre>
3748
3749<h5>Overview:</h5>
3750
3751<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3752
3753<h5>Arguments:</h5>
3754
3755<p>This instruction requires several arguments:</p>
3756
3757<ol>
3758 <li>
3759 <p>The optional "tail" marker indicates whether the callee function accesses
3760 any allocas or varargs in the caller. If the "tail" marker is present, the
3761 function call is eligible for tail call optimization. Note that calls may
3762 be marked "tail" even if they do not occur before a <a
3763 href="#i_ret"><tt>ret</tt></a> instruction.
3764 </li>
3765 <li>
3766 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3767 convention</a> the call should use. If none is specified, the call defaults
3768 to using C calling conventions.
3769 </li>
3770 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3771 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
3772 the type of the return value. Functions that return no value are marked
3773 <tt><a href="#t_void">void</a></tt>.</p>
3774 </li>
3775 <li>
3776 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
3777 value being invoked. The argument types must match the types implied by
3778 this signature. This type can be omitted if the function is not varargs
3779 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3780 </li>
3781 <li>
3782 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3783 be invoked. In most cases, this is a direct function invocation, but
3784 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3785 to function value.</p>
3786 </li>
3787 <li>
3788 <p>'<tt>function args</tt>': argument list whose types match the
3789 function signature argument types. All arguments must be of
3790 <a href="#t_firstclass">first class</a> type. If the function signature
3791 indicates the function accepts a variable number of arguments, the extra
3792 arguments can be specified.</p>
3793 </li>
3794</ol>
3795
3796<h5>Semantics:</h5>
3797
3798<p>The '<tt>call</tt>' instruction is used to cause control flow to
3799transfer to a specified function, with its incoming arguments bound to
3800the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3801instruction in the called function, control flow continues with the
3802instruction after the function call, and the return value of the
3803function is bound to the result argument. This is a simpler case of
3804the <a href="#i_invoke">invoke</a> instruction.</p>
3805
3806<h5>Example:</h5>
3807
3808<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3809 %retval = call i32 @test(i32 %argc)
3810 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42);
3811 %X = tail call i32 @foo()
3812 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo()
3813 %Z = call void %foo(i8 97 signext)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3814</pre>
3815
3816</div>
3817
3818<!-- _______________________________________________________________________ -->
3819<div class="doc_subsubsection">
3820 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3821</div>
3822
3823<div class="doc_text">
3824
3825<h5>Syntax:</h5>
3826
3827<pre>
3828 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
3829</pre>
3830
3831<h5>Overview:</h5>
3832
3833<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3834the "variable argument" area of a function call. It is used to implement the
3835<tt>va_arg</tt> macro in C.</p>
3836
3837<h5>Arguments:</h5>
3838
3839<p>This instruction takes a <tt>va_list*</tt> value and the type of
3840the argument. It returns a value of the specified argument type and
3841increments the <tt>va_list</tt> to point to the next argument. The
3842actual type of <tt>va_list</tt> is target specific.</p>
3843
3844<h5>Semantics:</h5>
3845
3846<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3847type from the specified <tt>va_list</tt> and causes the
3848<tt>va_list</tt> to point to the next argument. For more information,
3849see the variable argument handling <a href="#int_varargs">Intrinsic
3850Functions</a>.</p>
3851
3852<p>It is legal for this instruction to be called in a function which does not
3853take a variable number of arguments, for example, the <tt>vfprintf</tt>
3854function.</p>
3855
3856<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3857href="#intrinsics">intrinsic function</a> because it takes a type as an
3858argument.</p>
3859
3860<h5>Example:</h5>
3861
3862<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3863
3864</div>
3865
3866<!-- *********************************************************************** -->
3867<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3868<!-- *********************************************************************** -->
3869
3870<div class="doc_text">
3871
3872<p>LLVM supports the notion of an "intrinsic function". These functions have
3873well known names and semantics and are required to follow certain restrictions.
3874Overall, these intrinsics represent an extension mechanism for the LLVM
3875language that does not require changing all of the transformations in LLVM when
3876adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
3877
3878<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3879prefix is reserved in LLVM for intrinsic names; thus, function names may not
3880begin with this prefix. Intrinsic functions must always be external functions:
3881you cannot define the body of intrinsic functions. Intrinsic functions may
3882only be used in call or invoke instructions: it is illegal to take the address
3883of an intrinsic function. Additionally, because intrinsic functions are part
3884of the LLVM language, it is required if any are added that they be documented
3885here.</p>
3886
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]3887<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
3888a family of functions that perform the same operation but on different data
3889types. Because LLVM can represent over 8 million different integer types,
3890overloading is used commonly to allow an intrinsic function to operate on any
3891integer type. One or more of the argument types or the result type can be
3892overloaded to accept any integer type. Argument types may also be defined as
3893exactly matching a previous argument's type or the result type. This allows an
3894intrinsic function which accepts multiple arguments, but needs all of them to
3895be of the same type, to only be overloaded with respect to a single argument or
3896the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3897
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]3898<p>Overloaded intrinsics will have the names of its overloaded argument types
3899encoded into its function name, each preceded by a period. Only those types
3900which are overloaded result in a name suffix. Arguments whose type is matched
3901against another type do not. For example, the <tt>llvm.ctpop</tt> function can
3902take an integer of any width and returns an integer of exactly the same integer
3903width. This leads to a family of functions such as
3904<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
3905Only one type, the return type, is overloaded, and only one type suffix is
3906required. Because the argument's type is matched against the return type, it
3907does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3908
3909<p>To learn how to add an intrinsic function, please see the
3910<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3911</p>
3912
3913</div>
3914
3915<!-- ======================================================================= -->
3916<div class="doc_subsection">
3917 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3918</div>
3919
3920<div class="doc_text">
3921
3922<p>Variable argument support is defined in LLVM with the <a
3923 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3924intrinsic functions. These functions are related to the similarly
3925named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
3926
3927<p>All of these functions operate on arguments that use a
3928target-specific value type "<tt>va_list</tt>". The LLVM assembly
3929language reference manual does not define what this type is, so all
3930transformations should be prepared to handle these functions regardless of
3931the type used.</p>
3932
3933<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3934instruction and the variable argument handling intrinsic functions are
3935used.</p>
3936
3937<div class="doc_code">
3938<pre>
3939define i32 @test(i32 %X, ...) {
3940 ; Initialize variable argument processing
3941 %ap = alloca i8*
3942 %ap2 = bitcast i8** %ap to i8*
3943 call void @llvm.va_start(i8* %ap2)
3944
3945 ; Read a single integer argument
3946 %tmp = va_arg i8** %ap, i32
3947
3948 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3949 %aq = alloca i8*
3950 %aq2 = bitcast i8** %aq to i8*
3951 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
3952 call void @llvm.va_end(i8* %aq2)
3953
3954 ; Stop processing of arguments.
3955 call void @llvm.va_end(i8* %ap2)
3956 ret i32 %tmp
3957}
3958
3959declare void @llvm.va_start(i8*)
3960declare void @llvm.va_copy(i8*, i8*)
3961declare void @llvm.va_end(i8*)
3962</pre>
3963</div>
3964
3965</div>
3966
3967<!-- _______________________________________________________________________ -->
3968<div class="doc_subsubsection">
3969 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3970</div>
3971
3972
3973<div class="doc_text">
3974<h5>Syntax:</h5>
3975<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
3976<h5>Overview:</h5>
3977<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3978<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
3979href="#i_va_arg">va_arg</a></tt>.</p>
3980
3981<h5>Arguments:</h5>
3982
3983<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3984
3985<h5>Semantics:</h5>
3986
3987<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3988macro available in C. In a target-dependent way, it initializes the
3989<tt>va_list</tt> element to which the argument points, so that the next call to
3990<tt>va_arg</tt> will produce the first variable argument passed to the function.
3991Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3992last argument of the function as the compiler can figure that out.</p>
3993
3994</div>
3995
3996<!-- _______________________________________________________________________ -->
3997<div class="doc_subsubsection">
3998 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3999</div>
4000
4001<div class="doc_text">
4002<h5>Syntax:</h5>
4003<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
4004<h5>Overview:</h5>
4005
4006<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
4007which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4008or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4009
4010<h5>Arguments:</h5>
4011
4012<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4013
4014<h5>Semantics:</h5>
4015
4016<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4017macro available in C. In a target-dependent way, it destroys the
4018<tt>va_list</tt> element to which the argument points. Calls to <a
4019href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4020<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4021<tt>llvm.va_end</tt>.</p>
4022
4023</div>
4024
4025<!-- _______________________________________________________________________ -->
4026<div class="doc_subsubsection">
4027 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4028</div>
4029
4030<div class="doc_text">
4031
4032<h5>Syntax:</h5>
4033
4034<pre>
4035 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
4036</pre>
4037
4038<h5>Overview:</h5>
4039
4040<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4041from the source argument list to the destination argument list.</p>
4042
4043<h5>Arguments:</h5>
4044
4045<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4046The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4047
4048
4049<h5>Semantics:</h5>
4050
4051<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4052macro available in C. In a target-dependent way, it copies the source
4053<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4054intrinsic is necessary because the <tt><a href="#int_va_start">
4055llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4056example, memory allocation.</p>
4057
4058</div>
4059
4060<!-- ======================================================================= -->
4061<div class="doc_subsection">
4062 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4063</div>
4064
4065<div class="doc_text">
4066
4067<p>
4068LLVM support for <a href="GarbageCollection.html">Accurate Garbage
4069Collection</a> requires the implementation and generation of these intrinsics.
4070These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4071stack</a>, as well as garbage collector implementations that require <a
4072href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4073Front-ends for type-safe garbage collected languages should generate these
4074intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4075href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4076</p>
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]4077
4078<p>The garbage collection intrinsics only operate on objects in the generic
4079 address space (address space zero).</p>
4080
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4081</div>
4082
4083<!-- _______________________________________________________________________ -->
4084<div class="doc_subsubsection">
4085 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4086</div>
4087
4088<div class="doc_text">
4089
4090<h5>Syntax:</h5>
4091
4092<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4093 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4094</pre>
4095
4096<h5>Overview:</h5>
4097
4098<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4099the code generator, and allows some metadata to be associated with it.</p>
4100
4101<h5>Arguments:</h5>
4102
4103<p>The first argument specifies the address of a stack object that contains the
4104root pointer. The second pointer (which must be either a constant or a global
4105value address) contains the meta-data to be associated with the root.</p>
4106
4107<h5>Semantics:</h5>
4108
4109<p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
4110location. At compile-time, the code generator generates information to allow
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4111the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4112intrinsic may only be used in a function which <a href="#gc">specifies a GC
4113algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4114
4115</div>
4116
4117
4118<!-- _______________________________________________________________________ -->
4119<div class="doc_subsubsection">
4120 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4121</div>
4122
4123<div class="doc_text">
4124
4125<h5>Syntax:</h5>
4126
4127<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4128 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4129</pre>
4130
4131<h5>Overview:</h5>
4132
4133<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4134locations, allowing garbage collector implementations that require read
4135barriers.</p>
4136
4137<h5>Arguments:</h5>
4138
4139<p>The second argument is the address to read from, which should be an address
4140allocated from the garbage collector. The first object is a pointer to the
4141start of the referenced object, if needed by the language runtime (otherwise
4142null).</p>
4143
4144<h5>Semantics:</h5>
4145
4146<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4147instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4148garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4149may only be used in a function which <a href="#gc">specifies a GC
4150algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4151
4152</div>
4153
4154
4155<!-- _______________________________________________________________________ -->
4156<div class="doc_subsubsection">
4157 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4158</div>
4159
4160<div class="doc_text">
4161
4162<h5>Syntax:</h5>
4163
4164<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4165 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4166</pre>
4167
4168<h5>Overview:</h5>
4169
4170<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4171locations, allowing garbage collector implementations that require write
4172barriers (such as generational or reference counting collectors).</p>
4173
4174<h5>Arguments:</h5>
4175
4176<p>The first argument is the reference to store, the second is the start of the
4177object to store it to, and the third is the address of the field of Obj to
4178store to. If the runtime does not require a pointer to the object, Obj may be
4179null.</p>
4180
4181<h5>Semantics:</h5>
4182
4183<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4184instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4185garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4186may only be used in a function which <a href="#gc">specifies a GC
4187algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4188
4189</div>
4190
4191
4192
4193<!-- ======================================================================= -->
4194<div class="doc_subsection">
4195 <a name="int_codegen">Code Generator Intrinsics</a>
4196</div>
4197
4198<div class="doc_text">
4199<p>
4200These intrinsics are provided by LLVM to expose special features that may only
4201be implemented with code generator support.
4202</p>
4203
4204</div>
4205
4206<!-- _______________________________________________________________________ -->
4207<div class="doc_subsubsection">
4208 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4209</div>
4210
4211<div class="doc_text">
4212
4213<h5>Syntax:</h5>
4214<pre>
4215 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
4216</pre>
4217
4218<h5>Overview:</h5>
4219
4220<p>
4221The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4222target-specific value indicating the return address of the current function
4223or one of its callers.
4224</p>
4225
4226<h5>Arguments:</h5>
4227
4228<p>
4229The argument to this intrinsic indicates which function to return the address
4230for. Zero indicates the calling function, one indicates its caller, etc. The
4231argument is <b>required</b> to be a constant integer value.
4232</p>
4233
4234<h5>Semantics:</h5>
4235
4236<p>
4237The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4238the return address of the specified call frame, or zero if it cannot be
4239identified. The value returned by this intrinsic is likely to be incorrect or 0
4240for arguments other than zero, so it should only be used for debugging purposes.
4241</p>
4242
4243<p>
4244Note that calling this intrinsic does not prevent function inlining or other
4245aggressive transformations, so the value returned may not be that of the obvious
4246source-language caller.
4247</p>
4248</div>
4249
4250
4251<!-- _______________________________________________________________________ -->
4252<div class="doc_subsubsection">
4253 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4254</div>
4255
4256<div class="doc_text">
4257
4258<h5>Syntax:</h5>
4259<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4260 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4261</pre>
4262
4263<h5>Overview:</h5>
4264
4265<p>
4266The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4267target-specific frame pointer value for the specified stack frame.
4268</p>
4269
4270<h5>Arguments:</h5>
4271
4272<p>
4273The argument to this intrinsic indicates which function to return the frame
4274pointer for. Zero indicates the calling function, one indicates its caller,
4275etc. The argument is <b>required</b> to be a constant integer value.
4276</p>
4277
4278<h5>Semantics:</h5>
4279
4280<p>
4281The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4282the frame address of the specified call frame, or zero if it cannot be
4283identified. The value returned by this intrinsic is likely to be incorrect or 0
4284for arguments other than zero, so it should only be used for debugging purposes.
4285</p>
4286
4287<p>
4288Note that calling this intrinsic does not prevent function inlining or other
4289aggressive transformations, so the value returned may not be that of the obvious
4290source-language caller.
4291</p>
4292</div>
4293
4294<!-- _______________________________________________________________________ -->
4295<div class="doc_subsubsection">
4296 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4297</div>
4298
4299<div class="doc_text">
4300
4301<h5>Syntax:</h5>
4302<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4303 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4304</pre>
4305
4306<h5>Overview:</h5>
4307
4308<p>
4309The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4310the function stack, for use with <a href="#int_stackrestore">
4311<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4312features like scoped automatic variable sized arrays in C99.
4313</p>
4314
4315<h5>Semantics:</h5>
4316
4317<p>
4318This intrinsic returns a opaque pointer value that can be passed to <a
4319href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4320<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4321<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4322state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4323practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4324that were allocated after the <tt>llvm.stacksave</tt> was executed.
4325</p>
4326
4327</div>
4328
4329<!-- _______________________________________________________________________ -->
4330<div class="doc_subsubsection">
4331 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4332</div>
4333
4334<div class="doc_text">
4335
4336<h5>Syntax:</h5>
4337<pre>
4338 declare void @llvm.stackrestore(i8 * %ptr)
4339</pre>
4340
4341<h5>Overview:</h5>
4342
4343<p>
4344The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4345the function stack to the state it was in when the corresponding <a
4346href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4347useful for implementing language features like scoped automatic variable sized
4348arrays in C99.
4349</p>
4350
4351<h5>Semantics:</h5>
4352
4353<p>
4354See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4355</p>
4356
4357</div>
4358
4359
4360<!-- _______________________________________________________________________ -->
4361<div class="doc_subsubsection">
4362 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4363</div>
4364
4365<div class="doc_text">
4366
4367<h5>Syntax:</h5>
4368<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4369 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]4370</pre>
4371
4372<h5>Overview:</h5>
4373
4374
4375<p>
4376The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4377a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4378no
4379effect on the behavior of the program but can change its performance
4380characteristics.
4381</p>
4382
4383<h5>Arguments:</h5>
4384
4385<p>
4386<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4387determining if the fetch should be for a read (0) or write (1), and
4388<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4389locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4390<tt>locality</tt> arguments must be constant integers.
4391</p>
4392
4393<h5>Semantics:</h5>
4394
4395<p>
4396This intrinsic does not modify the behavior of the program. In particular,
4397prefetches cannot trap and do not produce a value. On targets that support this
4398intrinsic, the prefetch can provide hints to the processor cache for better
4399performance.
4400</p>
4401
4402</div>
4403
4404<!-- _______________________________________________________________________ -->
4405<div class="doc_subsubsection">
4406 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4407</div>
4408
4409<div class="doc_text">
4410
4411<h5>Syntax:</h5>
4412<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4413 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4414</pre>
4415
4416<h5>Overview:</h5>
4417
4418
4419<p>
4420The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4421(PC) in a region of
4422code to simulators and other tools. The method is target specific, but it is
4423expected that the marker will use exported symbols to transmit the PC of the marker.
4424The marker makes no guarantees that it will remain with any specific instruction
4425after optimizations. It is possible that the presence of a marker will inhibit
4426optimizations. The intended use is to be inserted after optimizations to allow
4427correlations of simulation runs.
4428</p>
4429
4430<h5>Arguments:</h5>
4431
4432<p>
4433<tt>id</tt> is a numerical id identifying the marker.
4434</p>
4435
4436<h5>Semantics:</h5>
4437
4438<p>
4439This intrinsic does not modify the behavior of the program. Backends that do not
4440support this intrinisic may ignore it.
4441</p>
4442
4443</div>
4444
4445<!-- _______________________________________________________________________ -->
4446<div class="doc_subsubsection">
4447 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4448</div>
4449
4450<div class="doc_text">
4451
4452<h5>Syntax:</h5>
4453<pre>
4454 declare i64 @llvm.readcyclecounter( )
4455</pre>
4456
4457<h5>Overview:</h5>
4458
4459
4460<p>
4461The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4462counter register (or similar low latency, high accuracy clocks) on those targets
4463that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4464As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4465should only be used for small timings.
4466</p>
4467
4468<h5>Semantics:</h5>
4469
4470<p>
4471When directly supported, reading the cycle counter should not modify any memory.
4472Implementations are allowed to either return a application specific value or a
4473system wide value. On backends without support, this is lowered to a constant 0.
4474</p>
4475
4476</div>
4477
4478<!-- ======================================================================= -->
4479<div class="doc_subsection">
4480 <a name="int_libc">Standard C Library Intrinsics</a>
4481</div>
4482
4483<div class="doc_text">
4484<p>
4485LLVM provides intrinsics for a few important standard C library functions.
4486These intrinsics allow source-language front-ends to pass information about the
4487alignment of the pointer arguments to the code generator, providing opportunity
4488for more efficient code generation.
4489</p>
4490
4491</div>
4492
4493<!-- _______________________________________________________________________ -->
4494<div class="doc_subsubsection">
4495 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4496</div>
4497
4498<div class="doc_text">
4499
4500<h5>Syntax:</h5>
4501<pre>
4502 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4503 i32 &lt;len&gt;, i32 &lt;align&gt;)
4504 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4505 i64 &lt;len&gt;, i32 &lt;align&gt;)
4506</pre>
4507
4508<h5>Overview:</h5>
4509
4510<p>
4511The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4512location to the destination location.
4513</p>
4514
4515<p>
4516Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4517intrinsics do not return a value, and takes an extra alignment argument.
4518</p>
4519
4520<h5>Arguments:</h5>
4521
4522<p>
4523The first argument is a pointer to the destination, the second is a pointer to
4524the source. The third argument is an integer argument
4525specifying the number of bytes to copy, and the fourth argument is the alignment
4526of the source and destination locations.
4527</p>
4528
4529<p>
4530If the call to this intrinisic has an alignment value that is not 0 or 1, then
4531the caller guarantees that both the source and destination pointers are aligned
4532to that boundary.
4533</p>
4534
4535<h5>Semantics:</h5>
4536
4537<p>
4538The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4539location to the destination location, which are not allowed to overlap. It
4540copies "len" bytes of memory over. If the argument is known to be aligned to
4541some boundary, this can be specified as the fourth argument, otherwise it should
4542be set to 0 or 1.
4543</p>
4544</div>
4545
4546
4547<!-- _______________________________________________________________________ -->
4548<div class="doc_subsubsection">
4549 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4550</div>
4551
4552<div class="doc_text">
4553
4554<h5>Syntax:</h5>
4555<pre>
4556 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4557 i32 &lt;len&gt;, i32 &lt;align&gt;)
4558 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4559 i64 &lt;len&gt;, i32 &lt;align&gt;)
4560</pre>
4561
4562<h5>Overview:</h5>
4563
4564<p>
4565The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4566location to the destination location. It is similar to the
Chris Lattnerdba16ea2008-01-06 19:51:52 +0000[diff] [blame]4567'<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4568</p>
4569
4570<p>
4571Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4572intrinsics do not return a value, and takes an extra alignment argument.
4573</p>
4574
4575<h5>Arguments:</h5>
4576
4577<p>
4578The first argument is a pointer to the destination, the second is a pointer to
4579the source. The third argument is an integer argument
4580specifying the number of bytes to copy, and the fourth argument is the alignment
4581of the source and destination locations.
4582</p>
4583
4584<p>
4585If the call to this intrinisic has an alignment value that is not 0 or 1, then
4586the caller guarantees that the source and destination pointers are aligned to
4587that boundary.
4588</p>
4589
4590<h5>Semantics:</h5>
4591
4592<p>
4593The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4594location to the destination location, which may overlap. It
4595copies "len" bytes of memory over. If the argument is known to be aligned to
4596some boundary, this can be specified as the fourth argument, otherwise it should
4597be set to 0 or 1.
4598</p>
4599</div>
4600
4601
4602<!-- _______________________________________________________________________ -->
4603<div class="doc_subsubsection">
4604 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4605</div>
4606
4607<div class="doc_text">
4608
4609<h5>Syntax:</h5>
4610<pre>
4611 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
4612 i32 &lt;len&gt;, i32 &lt;align&gt;)
4613 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
4614 i64 &lt;len&gt;, i32 &lt;align&gt;)
4615</pre>
4616
4617<h5>Overview:</h5>
4618
4619<p>
4620The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4621byte value.
4622</p>
4623
4624<p>
4625Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4626does not return a value, and takes an extra alignment argument.
4627</p>
4628
4629<h5>Arguments:</h5>
4630
4631<p>
4632The first argument is a pointer to the destination to fill, the second is the
4633byte value to fill it with, the third argument is an integer
4634argument specifying the number of bytes to fill, and the fourth argument is the
4635known alignment of destination location.
4636</p>
4637
4638<p>
4639If the call to this intrinisic has an alignment value that is not 0 or 1, then
4640the caller guarantees that the destination pointer is aligned to that boundary.
4641</p>
4642
4643<h5>Semantics:</h5>
4644
4645<p>
4646The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4647the
4648destination location. If the argument is known to be aligned to some boundary,
4649this can be specified as the fourth argument, otherwise it should be set to 0 or
46501.
4651</p>
4652</div>
4653
4654
4655<!-- _______________________________________________________________________ -->
4656<div class="doc_subsubsection">
4657 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4658</div>
4659
4660<div class="doc_text">
4661
4662<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4663<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]4664floating point or vector of floating point type. Not all targets support all
4665types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4666<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4667 declare float @llvm.sqrt.f32(float %Val)
4668 declare double @llvm.sqrt.f64(double %Val)
4669 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
4670 declare fp128 @llvm.sqrt.f128(fp128 %Val)
4671 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4672</pre>
4673
4674<h5>Overview:</h5>
4675
4676<p>
4677The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4678returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4679<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4680negative numbers (which allows for better optimization).
4681</p>
4682
4683<h5>Arguments:</h5>
4684
4685<p>
4686The argument and return value are floating point numbers of the same type.
4687</p>
4688
4689<h5>Semantics:</h5>
4690
4691<p>
4692This function returns the sqrt of the specified operand if it is a nonnegative
4693floating point number.
4694</p>
4695</div>
4696
4697<!-- _______________________________________________________________________ -->
4698<div class="doc_subsubsection">
4699 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4700</div>
4701
4702<div class="doc_text">
4703
4704<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4705<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]4706floating point or vector of floating point type. Not all targets support all
4707types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4708<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4709 declare float @llvm.powi.f32(float %Val, i32 %power)
4710 declare double @llvm.powi.f64(double %Val, i32 %power)
4711 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
4712 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
4713 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4714</pre>
4715
4716<h5>Overview:</h5>
4717
4718<p>
4719The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4720specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4721multiplications is not defined. When a vector of floating point type is
4722used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4723</p>
4724
4725<h5>Arguments:</h5>
4726
4727<p>
4728The second argument is an integer power, and the first is a value to raise to
4729that power.
4730</p>
4731
4732<h5>Semantics:</h5>
4733
4734<p>
4735This function returns the first value raised to the second power with an
4736unspecified sequence of rounding operations.</p>
4737</div>
4738
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4739<!-- _______________________________________________________________________ -->
4740<div class="doc_subsubsection">
4741 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
4742</div>
4743
4744<div class="doc_text">
4745
4746<h5>Syntax:</h5>
4747<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
4748floating point or vector of floating point type. Not all targets support all
4749types however.
4750<pre>
4751 declare float @llvm.sin.f32(float %Val)
4752 declare double @llvm.sin.f64(double %Val)
4753 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
4754 declare fp128 @llvm.sin.f128(fp128 %Val)
4755 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
4756</pre>
4757
4758<h5>Overview:</h5>
4759
4760<p>
4761The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
4762</p>
4763
4764<h5>Arguments:</h5>
4765
4766<p>
4767The argument and return value are floating point numbers of the same type.
4768</p>
4769
4770<h5>Semantics:</h5>
4771
4772<p>
4773This function returns the sine of the specified operand, returning the
4774same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4775conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4776</div>
4777
4778<!-- _______________________________________________________________________ -->
4779<div class="doc_subsubsection">
4780 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
4781</div>
4782
4783<div class="doc_text">
4784
4785<h5>Syntax:</h5>
4786<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
4787floating point or vector of floating point type. Not all targets support all
4788types however.
4789<pre>
4790 declare float @llvm.cos.f32(float %Val)
4791 declare double @llvm.cos.f64(double %Val)
4792 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
4793 declare fp128 @llvm.cos.f128(fp128 %Val)
4794 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
4795</pre>
4796
4797<h5>Overview:</h5>
4798
4799<p>
4800The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
4801</p>
4802
4803<h5>Arguments:</h5>
4804
4805<p>
4806The argument and return value are floating point numbers of the same type.
4807</p>
4808
4809<h5>Semantics:</h5>
4810
4811<p>
4812This function returns the cosine of the specified operand, returning the
4813same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4814conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4815</div>
4816
4817<!-- _______________________________________________________________________ -->
4818<div class="doc_subsubsection">
4819 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
4820</div>
4821
4822<div class="doc_text">
4823
4824<h5>Syntax:</h5>
4825<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
4826floating point or vector of floating point type. Not all targets support all
4827types however.
4828<pre>
4829 declare float @llvm.pow.f32(float %Val, float %Power)
4830 declare double @llvm.pow.f64(double %Val, double %Power)
4831 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
4832 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
4833 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
4834</pre>
4835
4836<h5>Overview:</h5>
4837
4838<p>
4839The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
4840specified (positive or negative) power.
4841</p>
4842
4843<h5>Arguments:</h5>
4844
4845<p>
4846The second argument is a floating point power, and the first is a value to
4847raise to that power.
4848</p>
4849
4850<h5>Semantics:</h5>
4851
4852<p>
4853This function returns the first value raised to the second power,
4854returning the
4855same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4856conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4857</div>
4858
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4859
4860<!-- ======================================================================= -->
4861<div class="doc_subsection">
4862 <a name="int_manip">Bit Manipulation Intrinsics</a>
4863</div>
4864
4865<div class="doc_text">
4866<p>
4867LLVM provides intrinsics for a few important bit manipulation operations.
4868These allow efficient code generation for some algorithms.
4869</p>
4870
4871</div>
4872
4873<!-- _______________________________________________________________________ -->
4874<div class="doc_subsubsection">
4875 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4876</div>
4877
4878<div class="doc_text">
4879
4880<h5>Syntax:</h5>
4881<p>This is an overloaded intrinsic function. You can use bswap on any integer
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4882type that is an even number of bytes (i.e. BitWidth % 16 == 0).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4883<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4884 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
4885 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
4886 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4887</pre>
4888
4889<h5>Overview:</h5>
4890
4891<p>
4892The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
4893values with an even number of bytes (positive multiple of 16 bits). These are
4894useful for performing operations on data that is not in the target's native
4895byte order.
4896</p>
4897
4898<h5>Semantics:</h5>
4899
4900<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4901The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4902and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4903intrinsic returns an i32 value that has the four bytes of the input i32
4904swapped, 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]4905i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
4906<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4907additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4908</p>
4909
4910</div>
4911
4912<!-- _______________________________________________________________________ -->
4913<div class="doc_subsubsection">
4914 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4915</div>
4916
4917<div class="doc_text">
4918
4919<h5>Syntax:</h5>
4920<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4921width. Not all targets support all bit widths however.
4922<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4923 declare i8 @llvm.ctpop.i8 (i8 &lt;src&gt;)
4924 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4925 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4926 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
4927 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4928</pre>
4929
4930<h5>Overview:</h5>
4931
4932<p>
4933The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4934value.
4935</p>
4936
4937<h5>Arguments:</h5>
4938
4939<p>
4940The only argument is the value to be counted. The argument may be of any
4941integer type. The return type must match the argument type.
4942</p>
4943
4944<h5>Semantics:</h5>
4945
4946<p>
4947The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4948</p>
4949</div>
4950
4951<!-- _______________________________________________________________________ -->
4952<div class="doc_subsubsection">
4953 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4954</div>
4955
4956<div class="doc_text">
4957
4958<h5>Syntax:</h5>
4959<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4960integer bit width. Not all targets support all bit widths however.
4961<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4962 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
4963 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4964 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4965 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
4966 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4967</pre>
4968
4969<h5>Overview:</h5>
4970
4971<p>
4972The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4973leading zeros in a variable.
4974</p>
4975
4976<h5>Arguments:</h5>
4977
4978<p>
4979The only argument is the value to be counted. The argument may be of any
4980integer type. The return type must match the argument type.
4981</p>
4982
4983<h5>Semantics:</h5>
4984
4985<p>
4986The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4987in a variable. If the src == 0 then the result is the size in bits of the type
4988of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4989</p>
4990</div>
4991
4992
4993
4994<!-- _______________________________________________________________________ -->
4995<div class="doc_subsubsection">
4996 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4997</div>
4998
4999<div class="doc_text">
5000
5001<h5>Syntax:</h5>
5002<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5003integer bit width. Not all targets support all bit widths however.
5004<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5005 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
5006 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5007 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5008 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
5009 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5010</pre>
5011
5012<h5>Overview:</h5>
5013
5014<p>
5015The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5016trailing zeros.
5017</p>
5018
5019<h5>Arguments:</h5>
5020
5021<p>
5022The only argument is the value to be counted. The argument may be of any
5023integer type. The return type must match the argument type.
5024</p>
5025
5026<h5>Semantics:</h5>
5027
5028<p>
5029The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5030in a variable. If the src == 0 then the result is the size in bits of the type
5031of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5032</p>
5033</div>
5034
5035<!-- _______________________________________________________________________ -->
5036<div class="doc_subsubsection">
5037 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5038</div>
5039
5040<div class="doc_text">
5041
5042<h5>Syntax:</h5>
5043<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
5044on any integer bit width.
5045<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5046 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5047 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5048</pre>
5049
5050<h5>Overview:</h5>
5051<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5052range of bits from an integer value and returns them in the same bit width as
5053the original value.</p>
5054
5055<h5>Arguments:</h5>
5056<p>The first argument, <tt>%val</tt> and the result may be integer types of
5057any bit width but they must have the same bit width. The second and third
5058arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5059
5060<h5>Semantics:</h5>
5061<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5062of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5063<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5064operates in forward mode.</p>
5065<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5066right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5067only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5068<ol>
5069 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5070 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5071 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5072 to determine the number of bits to retain.</li>
5073 <li>A mask of the retained bits is created by shifting a -1 value.</li>
5074 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
5075</ol>
5076<p>In reverse mode, a similar computation is made except that the bits are
5077returned in the reverse order. So, for example, if <tt>X</tt> has the value
5078<tt>i16 0x0ACF (101011001111)</tt> and we apply
5079<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5080<tt>i16 0x0026 (000000100110)</tt>.</p>
5081</div>
5082
5083<div class="doc_subsubsection">
5084 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5085</div>
5086
5087<div class="doc_text">
5088
5089<h5>Syntax:</h5>
5090<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
5091on any integer bit width.
5092<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5093 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5094 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]5095</pre>
5096
5097<h5>Overview:</h5>
5098<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5099of bits in an integer value with another integer value. It returns the integer
5100with the replaced bits.</p>
5101
5102<h5>Arguments:</h5>
5103<p>The first argument, <tt>%val</tt> and the result may be integer types of
5104any bit width but they must have the same bit width. <tt>%val</tt> is the value
5105whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5106integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5107type since they specify only a bit index.</p>
5108
5109<h5>Semantics:</h5>
5110<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5111of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5112<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5113operates in forward mode.</p>
5114<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5115truncating it down to the size of the replacement area or zero extending it
5116up to that size.</p>
5117<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5118are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5119in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5120to the <tt>%hi</tt>th bit.
5121<p>In reverse mode, a similar computation is made except that the bits are
5122reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5123<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
5124<h5>Examples:</h5>
5125<pre>
5126 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
5127 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
5128 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
5129 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
5130 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
5131</pre>
5132</div>
5133
5134<!-- ======================================================================= -->
5135<div class="doc_subsection">
5136 <a name="int_debugger">Debugger Intrinsics</a>
5137</div>
5138
5139<div class="doc_text">
5140<p>
5141The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5142are described in the <a
5143href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5144Debugging</a> document.
5145</p>
5146</div>
5147
5148
5149<!-- ======================================================================= -->
5150<div class="doc_subsection">
5151 <a name="int_eh">Exception Handling Intrinsics</a>
5152</div>
5153
5154<div class="doc_text">
5155<p> The LLVM exception handling intrinsics (which all start with
5156<tt>llvm.eh.</tt> prefix), are described in the <a
5157href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5158Handling</a> document. </p>
5159</div>
5160
5161<!-- ======================================================================= -->
5162<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5163 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5164</div>
5165
5166<div class="doc_text">
5167<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5168 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5169 the <tt>nest</tt> attribute, from a function. The result is a callable
5170 function pointer lacking the nest parameter - the caller does not need
5171 to provide a value for it. Instead, the value to use is stored in
5172 advance in a "trampoline", a block of memory usually allocated
5173 on the stack, which also contains code to splice the nest value into the
5174 argument list. This is used to implement the GCC nested function address
5175 extension.
5176</p>
5177<p>
5178 For example, if the function is
5179 <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]5180 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]5181<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5182 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5183 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5184 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5185 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5186</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]5187 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5188 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5189</div>
5190
5191<!-- _______________________________________________________________________ -->
5192<div class="doc_subsubsection">
5193 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5194</div>
5195<div class="doc_text">
5196<h5>Syntax:</h5>
5197<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5198declare 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]5199</pre>
5200<h5>Overview:</h5>
5201<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5202 This fills the memory pointed to by <tt>tramp</tt> with code
5203 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5204</p>
5205<h5>Arguments:</h5>
5206<p>
5207 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5208 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5209 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]5210 intrinsic. Note that the size and the alignment are target-specific - LLVM
5211 currently provides no portable way of determining them, so a front-end that
5212 generates this intrinsic needs to have some target-specific knowledge.
5213 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]5214</p>
5215<h5>Semantics:</h5>
5216<p>
5217 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5218 dependent code, turning it into a function. A pointer to this function is
5219 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5220 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5221 before being called. The new function's signature is the same as that of
5222 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5223 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5224 of pointer type. Calling the new function is equivalent to calling
5225 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5226 missing <tt>nest</tt> argument. If, after calling
5227 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5228 modified, then the effect of any later call to the returned function pointer is
5229 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5230</p>
5231</div>
5232
5233<!-- ======================================================================= -->
5234<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5235 <a name="int_general">General Intrinsics</a>
5236</div>
5237
5238<div class="doc_text">
5239<p> This class of intrinsics is designed to be generic and has
5240no specific purpose. </p>
5241</div>
5242
5243<!-- _______________________________________________________________________ -->
5244<div class="doc_subsubsection">
5245 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
5246</div>
5247
5248<div class="doc_text">
5249
5250<h5>Syntax:</h5>
5251<pre>
5252 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5253</pre>
5254
5255<h5>Overview:</h5>
5256
5257<p>
5258The '<tt>llvm.var.annotation</tt>' intrinsic
5259</p>
5260
5261<h5>Arguments:</h5>
5262
5263<p>
5264The first argument is a pointer to a value, the second is a pointer to a
5265global string, the third is a pointer to a global string which is the source
5266file name, and the last argument is the line number.
5267</p>
5268
5269<h5>Semantics:</h5>
5270
5271<p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]5272This intrinsic allows annotation of local variables with arbitrary strings.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5273This can be useful for special purpose optimizations that want to look for these
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]5274annotations. These have no other defined use, they are ignored by code
5275generation and optimization.
5276</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5277</div>
5278
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5279<!-- _______________________________________________________________________ -->
5280<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]5281 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5282</div>
5283
5284<div class="doc_text">
5285
5286<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5287<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
5288any integer bit width.
5289</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5290<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]5291 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5292 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5293 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5294 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5295 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]5296</pre>
5297
5298<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5299
5300<p>
5301The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5302</p>
5303
5304<h5>Arguments:</h5>
5305
5306<p>
5307The first argument is an integer value (result of some expression),
5308the second is a pointer to a global string, the third is a pointer to a global
5309string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5310It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5311</p>
5312
5313<h5>Semantics:</h5>
5314
5315<p>
5316This intrinsic allows annotations to be put on arbitrary expressions
5317with arbitrary strings. This can be useful for special purpose optimizations
5318that want to look for these annotations. These have no other defined use, they
5319are ignored by code generation and optimization.
5320</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5321
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]5322<!-- _______________________________________________________________________ -->
5323<div class="doc_subsubsection">
5324 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
5325</div>
5326
5327<div class="doc_text">
5328
5329<h5>Syntax:</h5>
5330<pre>
5331 declare void @llvm.trap()
5332</pre>
5333
5334<h5>Overview:</h5>
5335
5336<p>
5337The '<tt>llvm.trap</tt>' intrinsic
5338</p>
5339
5340<h5>Arguments:</h5>
5341
5342<p>
5343None
5344</p>
5345
5346<h5>Semantics:</h5>
5347
5348<p>
5349This intrinsics is lowered to the target dependent trap instruction. If the
5350target does not have a trap instruction, this intrinsic will be lowered to the
5351call of the abort() function.
5352</p>
5353</div>
5354
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5355<!-- *********************************************************************** -->
5356<hr>
5357<address>
5358 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
5359 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
5360 <a href="http://validator.w3.org/check/referer"><img
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]5361 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5362
5363 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
5364 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
5365 Last modified: $Date$
5366</address>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]5367
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5368</body>
5369</html>