Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 7311d224c027af846bc5bcea44895152a81e2a04 / . / docs / LangRef.html
blob: 86f1ee3b51bba0657d23657816d7f6d72e52d846 [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>
36 <li><a href="#t_primitive">Primitive Types</a>
37 <ol>
38 <li><a href="#t_classifications">Type Classifications</a></li>
39 </ol>
40 </li>
41 <li><a href="#t_derived">Derived Types</a>
42 <ol>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]43 <li><a href="#t_integer">Integer Type</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]44 <li><a href="#t_array">Array Type</a></li>
45 <li><a href="#t_function">Function Type</a></li>
46 <li><a href="#t_pointer">Pointer Type</a></li>
47 <li><a href="#t_struct">Structure Type</a></li>
48 <li><a href="#t_pstruct">Packed Structure Type</a></li>
49 <li><a href="#t_vector">Vector Type</a></li>
50 <li><a href="#t_opaque">Opaque Type</a></li>
51 </ol>
52 </li>
53 </ol>
54 </li>
55 <li><a href="#constants">Constants</a>
56 <ol>
57 <li><a href="#simpleconstants">Simple Constants</a>
58 <li><a href="#aggregateconstants">Aggregate Constants</a>
59 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
60 <li><a href="#undefvalues">Undefined Values</a>
61 <li><a href="#constantexprs">Constant Expressions</a>
62 </ol>
63 </li>
64 <li><a href="#othervalues">Other Values</a>
65 <ol>
66 <li><a href="#inlineasm">Inline Assembler Expressions</a>
67 </ol>
68 </li>
69 <li><a href="#instref">Instruction Reference</a>
70 <ol>
71 <li><a href="#terminators">Terminator Instructions</a>
72 <ol>
73 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
74 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
75 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
76 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
77 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
78 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
79 </ol>
80 </li>
81 <li><a href="#binaryops">Binary Operations</a>
82 <ol>
83 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
84 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
85 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
86 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
87 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
88 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
89 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
90 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
91 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
92 </ol>
93 </li>
94 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
95 <ol>
96 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
97 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
98 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
99 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
100 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
101 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
102 </ol>
103 </li>
104 <li><a href="#vectorops">Vector Operations</a>
105 <ol>
106 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
107 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
108 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
109 </ol>
110 </li>
111 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
112 <ol>
113 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
114 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
115 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
116 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
117 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
118 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
119 </ol>
120 </li>
121 <li><a href="#convertops">Conversion Operations</a>
122 <ol>
123 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
124 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
125 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
126 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
127 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
128 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
129 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
130 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
131 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
132 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
133 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
134 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
135 </ol>
136 <li><a href="#otherops">Other Operations</a>
137 <ol>
138 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
139 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
140 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
141 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
142 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
143 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
144 </ol>
145 </li>
146 </ol>
147 </li>
148 <li><a href="#intrinsics">Intrinsic Functions</a>
149 <ol>
150 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
151 <ol>
152 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
153 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
154 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
155 </ol>
156 </li>
157 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
158 <ol>
159 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
160 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
161 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
162 </ol>
163 </li>
164 <li><a href="#int_codegen">Code Generator Intrinsics</a>
165 <ol>
166 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
167 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
168 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
169 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
170 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
171 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
172 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
173 </ol>
174 </li>
175 <li><a href="#int_libc">Standard C Library Intrinsics</a>
176 <ol>
177 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
178 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
179 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
180 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
181 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]182 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
183 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
184 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]185 </ol>
186 </li>
187 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
188 <ol>
189 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
190 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
191 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
192 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
193 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
194 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
195 </ol>
196 </li>
197 <li><a href="#int_debugger">Debugger intrinsics</a></li>
198 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]199 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]200 <ol>
201 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]202 </ol>
203 </li>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]204 <li><a href="#int_general">General intrinsics</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]205 <ol>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]206 <li><a href="#int_var_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]207 <tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]208 </ol>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]209 <ol>
210 <li><a href="#int_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]211 <tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]212 </ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]213 </li>
214 </ol>
215 </li>
216</ol>
217
218<div class="doc_author">
219 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
220 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
221</div>
222
223<!-- *********************************************************************** -->
224<div class="doc_section"> <a name="abstract">Abstract </a></div>
225<!-- *********************************************************************** -->
226
227<div class="doc_text">
228<p>This document is a reference manual for the LLVM assembly language.
229LLVM is an SSA based representation that provides type safety,
230low-level operations, flexibility, and the capability of representing
231'all' high-level languages cleanly. It is the common code
232representation used throughout all phases of the LLVM compilation
233strategy.</p>
234</div>
235
236<!-- *********************************************************************** -->
237<div class="doc_section"> <a name="introduction">Introduction</a> </div>
238<!-- *********************************************************************** -->
239
240<div class="doc_text">
241
242<p>The LLVM code representation is designed to be used in three
243different forms: as an in-memory compiler IR, as an on-disk bitcode
244representation (suitable for fast loading by a Just-In-Time compiler),
245and as a human readable assembly language representation. This allows
246LLVM to provide a powerful intermediate representation for efficient
247compiler transformations and analysis, while providing a natural means
248to debug and visualize the transformations. The three different forms
249of LLVM are all equivalent. This document describes the human readable
250representation and notation.</p>
251
252<p>The LLVM representation aims to be light-weight and low-level
253while being expressive, typed, and extensible at the same time. It
254aims to be a "universal IR" of sorts, by being at a low enough level
255that high-level ideas may be cleanly mapped to it (similar to how
256microprocessors are "universal IR's", allowing many source languages to
257be mapped to them). By providing type information, LLVM can be used as
258the target of optimizations: for example, through pointer analysis, it
259can be proven that a C automatic variable is never accessed outside of
260the current function... allowing it to be promoted to a simple SSA
261value instead of a memory location.</p>
262
263</div>
264
265<!-- _______________________________________________________________________ -->
266<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
267
268<div class="doc_text">
269
270<p>It is important to note that this document describes 'well formed'
271LLVM assembly language. There is a difference between what the parser
272accepts and what is considered 'well formed'. For example, the
273following instruction is syntactically okay, but not well formed:</p>
274
275<div class="doc_code">
276<pre>
277%x = <a href="#i_add">add</a> i32 1, %x
278</pre>
279</div>
280
281<p>...because the definition of <tt>%x</tt> does not dominate all of
282its uses. The LLVM infrastructure provides a verification pass that may
283be used to verify that an LLVM module is well formed. This pass is
284automatically run by the parser after parsing input assembly and by
285the optimizer before it outputs bitcode. The violations pointed out
286by the verifier pass indicate bugs in transformation passes or input to
287the parser.</p>
288</div>
289
Chris Lattnera83fdc02007-10-03 17:34:29 +0000[diff] [blame]290<!-- Describe the typesetting conventions here. -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]291
292<!-- *********************************************************************** -->
293<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
294<!-- *********************************************************************** -->
295
296<div class="doc_text">
297
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]298 <p>LLVM identifiers come in two basic types: global and local. Global
299 identifiers (functions, global variables) begin with the @ character. Local
300 identifiers (register names, types) begin with the % character. Additionally,
301 there are three different formats for identifiers, for different purposes:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]302
303<ol>
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]304 <li>Named values are represented as a string of characters with their prefix.
305 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
306 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]307 Identifiers which require other characters in their names can be surrounded
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]308 with quotes. In this way, anything except a <tt>&quot;</tt> character can
309 be used in a named value.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]310
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]311 <li>Unnamed values are represented as an unsigned numeric value with their
312 prefix. For example, %12, @2, %44.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]313
314 <li>Constants, which are described in a <a href="#constants">section about
315 constants</a>, below.</li>
316</ol>
317
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]318<p>LLVM requires that values start with a prefix for two reasons: Compilers
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]319don't need to worry about name clashes with reserved words, and the set of
320reserved words may be expanded in the future without penalty. Additionally,
321unnamed identifiers allow a compiler to quickly come up with a temporary
322variable without having to avoid symbol table conflicts.</p>
323
324<p>Reserved words in LLVM are very similar to reserved words in other
325languages. There are keywords for different opcodes
326('<tt><a href="#i_add">add</a></tt>',
327 '<tt><a href="#i_bitcast">bitcast</a></tt>',
328 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
329href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
330and others. These reserved words cannot conflict with variable names, because
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]331none of them start with a prefix character ('%' or '@').</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]332
333<p>Here is an example of LLVM code to multiply the integer variable
334'<tt>%X</tt>' by 8:</p>
335
336<p>The easy way:</p>
337
338<div class="doc_code">
339<pre>
340%result = <a href="#i_mul">mul</a> i32 %X, 8
341</pre>
342</div>
343
344<p>After strength reduction:</p>
345
346<div class="doc_code">
347<pre>
348%result = <a href="#i_shl">shl</a> i32 %X, i8 3
349</pre>
350</div>
351
352<p>And the hard way:</p>
353
354<div class="doc_code">
355<pre>
356<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
357<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
358%result = <a href="#i_add">add</a> i32 %1, %1
359</pre>
360</div>
361
362<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
363important lexical features of LLVM:</p>
364
365<ol>
366
367 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
368 line.</li>
369
370 <li>Unnamed temporaries are created when the result of a computation is not
371 assigned to a named value.</li>
372
373 <li>Unnamed temporaries are numbered sequentially</li>
374
375</ol>
376
377<p>...and it also shows a convention that we follow in this document. When
378demonstrating instructions, we will follow an instruction with a comment that
379defines the type and name of value produced. Comments are shown in italic
380text.</p>
381
382</div>
383
384<!-- *********************************************************************** -->
385<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
386<!-- *********************************************************************** -->
387
388<!-- ======================================================================= -->
389<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
390</div>
391
392<div class="doc_text">
393
394<p>LLVM programs are composed of "Module"s, each of which is a
395translation unit of the input programs. Each module consists of
396functions, global variables, and symbol table entries. Modules may be
397combined together with the LLVM linker, which merges function (and
398global variable) definitions, resolves forward declarations, and merges
399symbol table entries. Here is an example of the "hello world" module:</p>
400
401<div class="doc_code">
402<pre><i>; Declare the string constant as a global constant...</i>
403<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
404 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
405
406<i>; External declaration of the puts function</i>
407<a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
408
409<i>; Definition of main function</i>
410define i32 @main() { <i>; i32()* </i>
411 <i>; Convert [13x i8 ]* to i8 *...</i>
412 %cast210 = <a
413 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
414
415 <i>; Call puts function to write out the string to stdout...</i>
416 <a
417 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
418 <a
419 href="#i_ret">ret</a> i32 0<br>}<br>
420</pre>
421</div>
422
423<p>This example is made up of a <a href="#globalvars">global variable</a>
424named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
425function, and a <a href="#functionstructure">function definition</a>
426for "<tt>main</tt>".</p>
427
428<p>In general, a module is made up of a list of global values,
429where both functions and global variables are global values. Global values are
430represented by a pointer to a memory location (in this case, a pointer to an
431array of char, and a pointer to a function), and have one of the following <a
432href="#linkage">linkage types</a>.</p>
433
434</div>
435
436<!-- ======================================================================= -->
437<div class="doc_subsection">
438 <a name="linkage">Linkage Types</a>
439</div>
440
441<div class="doc_text">
442
443<p>
444All Global Variables and Functions have one of the following types of linkage:
445</p>
446
447<dl>
448
449 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
450
451 <dd>Global values with internal linkage are only directly accessible by
452 objects in the current module. In particular, linking code into a module with
453 an internal global value may cause the internal to be renamed as necessary to
454 avoid collisions. Because the symbol is internal to the module, all
455 references can be updated. This corresponds to the notion of the
456 '<tt>static</tt>' keyword in C.
457 </dd>
458
459 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
460
461 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
462 the same name when linkage occurs. This is typically used to implement
463 inline functions, templates, or other code which must be generated in each
464 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
465 allowed to be discarded.
466 </dd>
467
468 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
469
470 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
471 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
472 used for globals that may be emitted in multiple translation units, but that
473 are not guaranteed to be emitted into every translation unit that uses them.
474 One example of this are common globals in C, such as "<tt>int X;</tt>" at
475 global scope.
476 </dd>
477
478 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
479
480 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
481 pointer to array type. When two global variables with appending linkage are
482 linked together, the two global arrays are appended together. This is the
483 LLVM, typesafe, equivalent of having the system linker append together
484 "sections" with identical names when .o files are linked.
485 </dd>
486
487 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
488 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
489 until linked, if not linked, the symbol becomes null instead of being an
490 undefined reference.
491 </dd>
492
493 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
494
495 <dd>If none of the above identifiers are used, the global is externally
496 visible, meaning that it participates in linkage and can be used to resolve
497 external symbol references.
498 </dd>
499</dl>
500
501 <p>
502 The next two types of linkage are targeted for Microsoft Windows platform
503 only. They are designed to support importing (exporting) symbols from (to)
504 DLLs.
505 </p>
506
507 <dl>
508 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
509
510 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
511 or variable via a global pointer to a pointer that is set up by the DLL
512 exporting the symbol. On Microsoft Windows targets, the pointer name is
513 formed by combining <code>_imp__</code> and the function or variable name.
514 </dd>
515
516 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
517
518 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
519 pointer to a pointer in a DLL, so that it can be referenced with the
520 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
521 name is formed by combining <code>_imp__</code> and the function or variable
522 name.
523 </dd>
524
525</dl>
526
527<p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
528variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
529variable and was linked with this one, one of the two would be renamed,
530preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
531external (i.e., lacking any linkage declarations), they are accessible
532outside of the current module.</p>
533<p>It is illegal for a function <i>declaration</i>
534to have any linkage type other than "externally visible", <tt>dllimport</tt>,
535or <tt>extern_weak</tt>.</p>
536<p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
537linkages.
538</div>
539
540<!-- ======================================================================= -->
541<div class="doc_subsection">
542 <a name="callingconv">Calling Conventions</a>
543</div>
544
545<div class="doc_text">
546
547<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
548and <a href="#i_invoke">invokes</a> can all have an optional calling convention
549specified for the call. The calling convention of any pair of dynamic
550caller/callee must match, or the behavior of the program is undefined. The
551following calling conventions are supported by LLVM, and more may be added in
552the future:</p>
553
554<dl>
555 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
556
557 <dd>This calling convention (the default if no other calling convention is
558 specified) matches the target C calling conventions. This calling convention
559 supports varargs function calls and tolerates some mismatch in the declared
560 prototype and implemented declaration of the function (as does normal C).
561 </dd>
562
563 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
564
565 <dd>This calling convention attempts to make calls as fast as possible
566 (e.g. by passing things in registers). This calling convention allows the
567 target to use whatever tricks it wants to produce fast code for the target,
568 without having to conform to an externally specified ABI. Implementations of
569 this convention should allow arbitrary tail call optimization to be supported.
570 This calling convention does not support varargs and requires the prototype of
571 all callees to exactly match the prototype of the function definition.
572 </dd>
573
574 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
575
576 <dd>This calling convention attempts to make code in the caller as efficient
577 as possible under the assumption that the call is not commonly executed. As
578 such, these calls often preserve all registers so that the call does not break
579 any live ranges in the caller side. This calling convention does not support
580 varargs and requires the prototype of all callees to exactly match the
581 prototype of the function definition.
582 </dd>
583
584 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
585
586 <dd>Any calling convention may be specified by number, allowing
587 target-specific calling conventions to be used. Target specific calling
588 conventions start at 64.
589 </dd>
590</dl>
591
592<p>More calling conventions can be added/defined on an as-needed basis, to
593support pascal conventions or any other well-known target-independent
594convention.</p>
595
596</div>
597
598<!-- ======================================================================= -->
599<div class="doc_subsection">
600 <a name="visibility">Visibility Styles</a>
601</div>
602
603<div class="doc_text">
604
605<p>
606All Global Variables and Functions have one of the following visibility styles:
607</p>
608
609<dl>
610 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
611
612 <dd>On ELF, default visibility means that the declaration is visible to other
613 modules and, in shared libraries, means that the declared entity may be
614 overridden. On Darwin, default visibility means that the declaration is
615 visible to other modules. Default visibility corresponds to "external
616 linkage" in the language.
617 </dd>
618
619 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
620
621 <dd>Two declarations of an object with hidden visibility refer to the same
622 object if they are in the same shared object. Usually, hidden visibility
623 indicates that the symbol will not be placed into the dynamic symbol table,
624 so no other module (executable or shared library) can reference it
625 directly.
626 </dd>
627
628 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
629
630 <dd>On ELF, protected visibility indicates that the symbol will be placed in
631 the dynamic symbol table, but that references within the defining module will
632 bind to the local symbol. That is, the symbol cannot be overridden by another
633 module.
634 </dd>
635</dl>
636
637</div>
638
639<!-- ======================================================================= -->
640<div class="doc_subsection">
641 <a name="globalvars">Global Variables</a>
642</div>
643
644<div class="doc_text">
645
646<p>Global variables define regions of memory allocated at compilation time
647instead of run-time. Global variables may optionally be initialized, may have
648an explicit section to be placed in, and may have an optional explicit alignment
649specified. A variable may be defined as "thread_local", which means that it
650will not be shared by threads (each thread will have a separated copy of the
651variable). A variable may be defined as a global "constant," which indicates
652that the contents of the variable will <b>never</b> be modified (enabling better
653optimization, allowing the global data to be placed in the read-only section of
654an executable, etc). Note that variables that need runtime initialization
655cannot be marked "constant" as there is a store to the variable.</p>
656
657<p>
658LLVM explicitly allows <em>declarations</em> of global variables to be marked
659constant, even if the final definition of the global is not. This capability
660can be used to enable slightly better optimization of the program, but requires
661the language definition to guarantee that optimizations based on the
662'constantness' are valid for the translation units that do not include the
663definition.
664</p>
665
666<p>As SSA values, global variables define pointer values that are in
667scope (i.e. they dominate) all basic blocks in the program. Global
668variables always define a pointer to their "content" type because they
669describe a region of memory, and all memory objects in LLVM are
670accessed through pointers.</p>
671
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]672<p>A global variable may be declared to reside in a target-specifc numbered
673address space. For targets that support them, address spaces may affect how
674optimizations are performed and/or what target instructions are used to access
Christopher Lamb20a39e92007-12-12 08:44:39 +0000[diff] [blame]675the variable. The default address space is zero. The address space qualifier
676must precede any other attributes.</p>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]677
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]678<p>LLVM allows an explicit section to be specified for globals. If the target
679supports it, it will emit globals to the section specified.</p>
680
681<p>An explicit alignment may be specified for a global. If not present, or if
682the alignment is set to zero, the alignment of the global is set by the target
683to whatever it feels convenient. If an explicit alignment is specified, the
684global is forced to have at least that much alignment. All alignments must be
685a power of 2.</p>
686
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]687<p>For example, the following defines a global in a numbered address space with
688an initializer, section, and alignment:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]689
690<div class="doc_code">
691<pre>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]692@G = constant float 1.0 addrspace(5), section "foo", align 4
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]693</pre>
694</div>
695
696</div>
697
698
699<!-- ======================================================================= -->
700<div class="doc_subsection">
701 <a name="functionstructure">Functions</a>
702</div>
703
704<div class="doc_text">
705
706<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
707an optional <a href="#linkage">linkage type</a>, an optional
708<a href="#visibility">visibility style</a>, an optional
709<a href="#callingconv">calling convention</a>, a return type, an optional
710<a href="#paramattrs">parameter attribute</a> for the return type, a function
711name, a (possibly empty) argument list (each with optional
712<a href="#paramattrs">parameter attributes</a>), an optional section, an
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]713optional alignment, an optional <a href="#gc">garbage collector name</a>, an
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]714opening curly brace, a list of basic blocks, and a closing curly brace.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]715
716LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
717optional <a href="#linkage">linkage type</a>, an optional
718<a href="#visibility">visibility style</a>, an optional
719<a href="#callingconv">calling convention</a>, a return type, an optional
720<a href="#paramattrs">parameter attribute</a> for the return type, a function
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]721name, a possibly empty list of arguments, an optional alignment, and an optional
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]722<a href="#gc">garbage collector name</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]723
724<p>A function definition contains a list of basic blocks, forming the CFG for
725the function. Each basic block may optionally start with a label (giving the
726basic block a symbol table entry), contains a list of instructions, and ends
727with a <a href="#terminators">terminator</a> instruction (such as a branch or
728function return).</p>
729
730<p>The first basic block in a function is special in two ways: it is immediately
731executed on entrance to the function, and it is not allowed to have predecessor
732basic blocks (i.e. there can not be any branches to the entry block of a
733function). Because the block can have no predecessors, it also cannot have any
734<a href="#i_phi">PHI nodes</a>.</p>
735
736<p>LLVM allows an explicit section to be specified for functions. If the target
737supports it, it will emit functions to the section specified.</p>
738
739<p>An explicit alignment may be specified for a function. If not present, or if
740the alignment is set to zero, the alignment of the function is set by the target
741to whatever it feels convenient. If an explicit alignment is specified, the
742function is forced to have at least that much alignment. All alignments must be
743a power of 2.</p>
744
745</div>
746
747
748<!-- ======================================================================= -->
749<div class="doc_subsection">
750 <a name="aliasstructure">Aliases</a>
751</div>
752<div class="doc_text">
753 <p>Aliases act as "second name" for the aliasee value (which can be either
754 function or global variable or bitcast of global value). Aliases may have an
755 optional <a href="#linkage">linkage type</a>, and an
756 optional <a href="#visibility">visibility style</a>.</p>
757
758 <h5>Syntax:</h5>
759
760<div class="doc_code">
761<pre>
762@&lt;Name&gt; = [Linkage] [Visibility] alias &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
763</pre>
764</div>
765
766</div>
767
768
769
770<!-- ======================================================================= -->
771<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
772<div class="doc_text">
773 <p>The return type and each parameter of a function type may have a set of
774 <i>parameter attributes</i> associated with them. Parameter attributes are
775 used to communicate additional information about the result or parameters of
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]776 a function. Parameter attributes are considered to be part of the function,
777 not of the function type, so functions with different parameter attributes
778 can have the same function type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]779
780 <p>Parameter attributes are simple keywords that follow the type specified. If
781 multiple parameter attributes are needed, they are space separated. For
782 example:</p>
783
784<div class="doc_code">
785<pre>
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]786declare i32 @printf(i8* noalias , ...) nounwind
787declare i32 @atoi(i8*) nounwind readonly
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]788</pre>
789</div>
790
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]791 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
792 <tt>readonly</tt>) come immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]793
794 <p>Currently, only the following parameter attributes are defined:</p>
795 <dl>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]796 <dt><tt>zeroext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]797 <dd>This indicates that the parameter should be zero extended just before
798 a call to this function.</dd>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]799 <dt><tt>signext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]800 <dd>This indicates that the parameter should be sign extended just before
801 a call to this function.</dd>
802 <dt><tt>inreg</tt></dt>
803 <dd>This indicates that the parameter should be placed in register (if
804 possible) during assembling function call. Support for this attribute is
805 target-specific</dd>
806 <dt><tt>sret</tt></dt>
807 <dd>This indicates that the parameter specifies the address of a structure
808 that is the return value of the function in the source program.</dd>
809 <dt><tt>noalias</tt></dt>
810 <dd>This indicates that the parameter not alias any other object or any
811 other "noalias" objects during the function call.
812 <dt><tt>noreturn</tt></dt>
813 <dd>This function attribute indicates that the function never returns. This
814 indicates to LLVM that every call to this function should be treated as if
815 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
816 <dt><tt>nounwind</tt></dt>
817 <dd>This function attribute indicates that the function type does not use
818 the unwind instruction and does not allow stack unwinding to propagate
819 through it.</dd>
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]820 <dt><tt>nest</tt></dt>
821 <dd>This indicates that the parameter can be excised using the
822 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
Duncan Sands13e13f82007-11-22 20:23:04 +0000[diff] [blame]823 <dt><tt>readonly</tt></dt>
Duncan Sandsd69c0e62007-11-14 21:14:02 +0000[diff] [blame]824 <dd>This function attribute indicates that the function has no side-effects
Duncan Sands13e13f82007-11-22 20:23:04 +0000[diff] [blame]825 except for producing a return value or throwing an exception. The value
826 returned must only depend on the function arguments and/or global variables.
827 It may use values obtained by dereferencing pointers.</dd>
828 <dt><tt>readnone</tt></dt>
829 <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]830 function, but in addition it is not allowed to dereference any pointer arguments
831 or global variables.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]832 </dl>
833
834</div>
835
836<!-- ======================================================================= -->
837<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]838 <a name="gc">Garbage Collector Names</a>
839</div>
840
841<div class="doc_text">
842<p>Each function may specify a garbage collector name, which is simply a
843string.</p>
844
845<div class="doc_code"><pre
846>define void @f() gc "name" { ...</pre></div>
847
848<p>The compiler declares the supported values of <i>name</i>. Specifying a
849collector which will cause the compiler to alter its output in order to support
850the named garbage collection algorithm.</p>
851</div>
852
853<!-- ======================================================================= -->
854<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]855 <a name="moduleasm">Module-Level Inline Assembly</a>
856</div>
857
858<div class="doc_text">
859<p>
860Modules may contain "module-level inline asm" blocks, which corresponds to the
861GCC "file scope inline asm" blocks. These blocks are internally concatenated by
862LLVM and treated as a single unit, but may be separated in the .ll file if
863desired. The syntax is very simple:
864</p>
865
866<div class="doc_code">
867<pre>
868module asm "inline asm code goes here"
869module asm "more can go here"
870</pre>
871</div>
872
873<p>The strings can contain any character by escaping non-printable characters.
874 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
875 for the number.
876</p>
877
878<p>
879 The inline asm code is simply printed to the machine code .s file when
880 assembly code is generated.
881</p>
882</div>
883
884<!-- ======================================================================= -->
885<div class="doc_subsection">
886 <a name="datalayout">Data Layout</a>
887</div>
888
889<div class="doc_text">
890<p>A module may specify a target specific data layout string that specifies how
891data is to be laid out in memory. The syntax for the data layout is simply:</p>
892<pre> target datalayout = "<i>layout specification</i>"</pre>
893<p>The <i>layout specification</i> consists of a list of specifications
894separated by the minus sign character ('-'). Each specification starts with a
895letter and may include other information after the letter to define some
896aspect of the data layout. The specifications accepted are as follows: </p>
897<dl>
898 <dt><tt>E</tt></dt>
899 <dd>Specifies that the target lays out data in big-endian form. That is, the
900 bits with the most significance have the lowest address location.</dd>
901 <dt><tt>e</tt></dt>
902 <dd>Specifies that hte target lays out data in little-endian form. That is,
903 the bits with the least significance have the lowest address location.</dd>
904 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
905 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
906 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
907 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
908 too.</dd>
909 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
910 <dd>This specifies the alignment for an integer type of a given bit
911 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
912 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
913 <dd>This specifies the alignment for a vector type of a given bit
914 <i>size</i>.</dd>
915 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
916 <dd>This specifies the alignment for a floating point type of a given bit
917 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
918 (double).</dd>
919 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
920 <dd>This specifies the alignment for an aggregate type of a given bit
921 <i>size</i>.</dd>
922</dl>
923<p>When constructing the data layout for a given target, LLVM starts with a
924default set of specifications which are then (possibly) overriden by the
925specifications in the <tt>datalayout</tt> keyword. The default specifications
926are given in this list:</p>
927<ul>
928 <li><tt>E</tt> - big endian</li>
929 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
930 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
931 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
932 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
933 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
934 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
935 alignment of 64-bits</li>
936 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
937 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
938 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
939 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
940 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
941</ul>
942<p>When llvm is determining the alignment for a given type, it uses the
943following rules:
944<ol>
945 <li>If the type sought is an exact match for one of the specifications, that
946 specification is used.</li>
947 <li>If no match is found, and the type sought is an integer type, then the
948 smallest integer type that is larger than the bitwidth of the sought type is
949 used. If none of the specifications are larger than the bitwidth then the the
950 largest integer type is used. For example, given the default specifications
951 above, the i7 type will use the alignment of i8 (next largest) while both
952 i65 and i256 will use the alignment of i64 (largest specified).</li>
953 <li>If no match is found, and the type sought is a vector type, then the
954 largest vector type that is smaller than the sought vector type will be used
955 as a fall back. This happens because <128 x double> can be implemented in
956 terms of 64 <2 x double>, for example.</li>
957</ol>
958</div>
959
960<!-- *********************************************************************** -->
961<div class="doc_section"> <a name="typesystem">Type System</a> </div>
962<!-- *********************************************************************** -->
963
964<div class="doc_text">
965
966<p>The LLVM type system is one of the most important features of the
967intermediate representation. Being typed enables a number of
968optimizations to be performed on the IR directly, without having to do
969extra analyses on the side before the transformation. A strong type
970system makes it easier to read the generated code and enables novel
971analyses and transformations that are not feasible to perform on normal
972three address code representations.</p>
973
974</div>
975
976<!-- ======================================================================= -->
977<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
978<div class="doc_text">
979<p>The primitive types are the fundamental building blocks of the LLVM
980system. The current set of primitive types is as follows:</p>
981
982<table class="layout">
983 <tr class="layout">
984 <td class="left">
985 <table>
986 <tbody>
987 <tr><th>Type</th><th>Description</th></tr>
988 <tr><td><tt><a name="t_void">void</a></tt></td><td>No value</td></tr>
989 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
990 </tbody>
991 </table>
992 </td>
993 <td class="right">
994 <table>
995 <tbody>
996 <tr><th>Type</th><th>Description</th></tr>
997 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
998 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
999 </tbody>
1000 </table>
1001 </td>
1002 </tr>
1003</table>
1004</div>
1005
1006<!-- _______________________________________________________________________ -->
1007<div class="doc_subsubsection"> <a name="t_classifications">Type
1008Classifications</a> </div>
1009<div class="doc_text">
1010<p>These different primitive types fall into a few useful
1011classifications:</p>
1012
1013<table border="1" cellspacing="0" cellpadding="4">
1014 <tbody>
1015 <tr><th>Classification</th><th>Types</th></tr>
1016 <tr>
1017 <td><a name="t_integer">integer</a></td>
1018 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1019 </tr>
1020 <tr>
1021 <td><a name="t_floating">floating point</a></td>
1022 <td><tt>float, double</tt></td>
1023 </tr>
1024 <tr>
1025 <td><a name="t_firstclass">first class</a></td>
1026 <td><tt>i1, ..., float, double, <br/>
1027 <a href="#t_pointer">pointer</a>,<a href="#t_vector">vector</a></tt>
1028 </td>
1029 </tr>
1030 </tbody>
1031</table>
1032
1033<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1034most important. Values of these types are the only ones which can be
1035produced by instructions, passed as arguments, or used as operands to
1036instructions. This means that all structures and arrays must be
1037manipulated either by pointer or by component.</p>
1038</div>
1039
1040<!-- ======================================================================= -->
1041<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1042
1043<div class="doc_text">
1044
1045<p>The real power in LLVM comes from the derived types in the system.
1046This is what allows a programmer to represent arrays, functions,
1047pointers, and other useful types. Note that these derived types may be
1048recursive: For example, it is possible to have a two dimensional array.</p>
1049
1050</div>
1051
1052<!-- _______________________________________________________________________ -->
1053<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1054
1055<div class="doc_text">
1056
1057<h5>Overview:</h5>
1058<p>The integer type is a very simple derived type that simply specifies an
1059arbitrary bit width for the integer type desired. Any bit width from 1 bit to
10602^23-1 (about 8 million) can be specified.</p>
1061
1062<h5>Syntax:</h5>
1063
1064<pre>
1065 iN
1066</pre>
1067
1068<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1069value.</p>
1070
1071<h5>Examples:</h5>
1072<table class="layout">
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1073 <tbody>
1074 <tr>
1075 <td><tt>i1</tt></td>
1076 <td>a single-bit integer.</td>
1077 </tr><tr>
1078 <td><tt>i32</tt></td>
1079 <td>a 32-bit integer.</td>
1080 </tr><tr>
1081 <td><tt>i1942652</tt></td>
1082 <td>a really big integer of over 1 million bits.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1083 </tr>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1084 </tbody>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1085</table>
1086</div>
1087
1088<!-- _______________________________________________________________________ -->
1089<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1090
1091<div class="doc_text">
1092
1093<h5>Overview:</h5>
1094
1095<p>The array type is a very simple derived type that arranges elements
1096sequentially in memory. The array type requires a size (number of
1097elements) and an underlying data type.</p>
1098
1099<h5>Syntax:</h5>
1100
1101<pre>
1102 [&lt;# elements&gt; x &lt;elementtype&gt;]
1103</pre>
1104
1105<p>The number of elements is a constant integer value; elementtype may
1106be any type with a size.</p>
1107
1108<h5>Examples:</h5>
1109<table class="layout">
1110 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame^]1111 <td class="left"><tt>[40 x i32]</tt></td>
1112 <td class="left">Array of 40 32-bit integer values.</td>
1113 </tr>
1114 <tr class="layout">
1115 <td class="left"><tt>[41 x i32]</tt></td>
1116 <td class="left">Array of 41 32-bit integer values.</td>
1117 </tr>
1118 <tr class="layout">
1119 <td class="left"><tt>[4 x i8]</tt></td>
1120 <td class="left">Array of 4 8-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1121 </tr>
1122</table>
1123<p>Here are some examples of multidimensional arrays:</p>
1124<table class="layout">
1125 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame^]1126 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1127 <td class="left">3x4 array of 32-bit integer values.</td>
1128 </tr>
1129 <tr class="layout">
1130 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1131 <td class="left">12x10 array of single precision floating point values.</td>
1132 </tr>
1133 <tr class="layout">
1134 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1135 <td class="left">2x3x4 array of 16-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1136 </tr>
1137</table>
1138
1139<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1140length array. Normally, accesses past the end of an array are undefined in
1141LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1142As a special case, however, zero length arrays are recognized to be variable
1143length. This allows implementation of 'pascal style arrays' with the LLVM
1144type "{ i32, [0 x float]}", for example.</p>
1145
1146</div>
1147
1148<!-- _______________________________________________________________________ -->
1149<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1150<div class="doc_text">
1151<h5>Overview:</h5>
1152<p>The function type can be thought of as a function signature. It
1153consists of a return type and a list of formal parameter types.
1154Function types are usually used to build virtual function tables
1155(which are structures of pointers to functions), for indirect function
1156calls, and when defining a function.</p>
1157<p>
1158The return type of a function type cannot be an aggregate type.
1159</p>
1160<h5>Syntax:</h5>
1161<pre> &lt;returntype&gt; (&lt;parameter list&gt;)<br></pre>
1162<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1163specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1164which indicates that the function takes a variable number of arguments.
1165Variable argument functions can access their arguments with the <a
1166 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1167<h5>Examples:</h5>
1168<table class="layout">
1169 <tr class="layout">
1170 <td class="left"><tt>i32 (i32)</tt></td>
1171 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1172 </td>
1173 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1174 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1175 </tt></td>
1176 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1177 an <tt>i16</tt> that should be sign extended and a
1178 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1179 <tt>float</tt>.
1180 </td>
1181 </tr><tr class="layout">
1182 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1183 <td class="left">A vararg function that takes at least one
1184 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1185 which returns an integer. This is the signature for <tt>printf</tt> in
1186 LLVM.
1187 </td>
1188 </tr>
1189</table>
1190
1191</div>
1192<!-- _______________________________________________________________________ -->
1193<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1194<div class="doc_text">
1195<h5>Overview:</h5>
1196<p>The structure type is used to represent a collection of data members
1197together in memory. The packing of the field types is defined to match
1198the ABI of the underlying processor. The elements of a structure may
1199be any type that has a size.</p>
1200<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1201and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1202field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1203instruction.</p>
1204<h5>Syntax:</h5>
1205<pre> { &lt;type list&gt; }<br></pre>
1206<h5>Examples:</h5>
1207<table class="layout">
1208 <tr class="layout">
1209 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1210 <td class="left">A triple of three <tt>i32</tt> values</td>
1211 </tr><tr class="layout">
1212 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1213 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1214 second element is a <a href="#t_pointer">pointer</a> to a
1215 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1216 an <tt>i32</tt>.</td>
1217 </tr>
1218</table>
1219</div>
1220
1221<!-- _______________________________________________________________________ -->
1222<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1223</div>
1224<div class="doc_text">
1225<h5>Overview:</h5>
1226<p>The packed structure type is used to represent a collection of data members
1227together in memory. There is no padding between fields. Further, the alignment
1228of a packed structure is 1 byte. The elements of a packed structure may
1229be any type that has a size.</p>
1230<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1231and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1232field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1233instruction.</p>
1234<h5>Syntax:</h5>
1235<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1236<h5>Examples:</h5>
1237<table class="layout">
1238 <tr class="layout">
1239 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1240 <td class="left">A triple of three <tt>i32</tt> values</td>
1241 </tr><tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame^]1242 <td class="left"><tt>&lt; { float, i32 (i32)* } &gt;</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1243 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1244 second element is a <a href="#t_pointer">pointer</a> to a
1245 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1246 an <tt>i32</tt>.</td>
1247 </tr>
1248</table>
1249</div>
1250
1251<!-- _______________________________________________________________________ -->
1252<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1253<div class="doc_text">
1254<h5>Overview:</h5>
1255<p>As in many languages, the pointer type represents a pointer or
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1256reference to another object, which must live in memory. Pointer types may have
1257an optional address space attribute defining the target-specific numbered
1258address space where the pointed-to object resides. The default address space is
1259zero.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1260<h5>Syntax:</h5>
1261<pre> &lt;type&gt; *<br></pre>
1262<h5>Examples:</h5>
1263<table class="layout">
1264 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame^]1265 <td class="left"><tt>[4x i32]*</tt></td>
1266 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1267 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1268 </tr>
1269 <tr class="layout">
1270 <td class="left"><tt>i32 (i32 *) *</tt></td>
1271 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1272 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame^]1273 <tt>i32</tt>.</td>
1274 </tr>
1275 <tr class="layout">
1276 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1277 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1278 that resides in address space #5.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1279 </tr>
1280</table>
1281</div>
1282
1283<!-- _______________________________________________________________________ -->
1284<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1285<div class="doc_text">
1286
1287<h5>Overview:</h5>
1288
1289<p>A vector type is a simple derived type that represents a vector
1290of elements. Vector types are used when multiple primitive data
1291are operated in parallel using a single instruction (SIMD).
1292A vector type requires a size (number of
1293elements) and an underlying primitive data type. Vectors must have a power
1294of two length (1, 2, 4, 8, 16 ...). Vector types are
1295considered <a href="#t_firstclass">first class</a>.</p>
1296
1297<h5>Syntax:</h5>
1298
1299<pre>
1300 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1301</pre>
1302
1303<p>The number of elements is a constant integer value; elementtype may
1304be any integer or floating point type.</p>
1305
1306<h5>Examples:</h5>
1307
1308<table class="layout">
1309 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame^]1310 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1311 <td class="left">Vector of 4 32-bit integer values.</td>
1312 </tr>
1313 <tr class="layout">
1314 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1315 <td class="left">Vector of 8 32-bit floating-point values.</td>
1316 </tr>
1317 <tr class="layout">
1318 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1319 <td class="left">Vector of 2 64-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1320 </tr>
1321</table>
1322</div>
1323
1324<!-- _______________________________________________________________________ -->
1325<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1326<div class="doc_text">
1327
1328<h5>Overview:</h5>
1329
1330<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1331corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1332In LLVM, opaque types can eventually be resolved to any type (not just a
1333structure type).</p>
1334
1335<h5>Syntax:</h5>
1336
1337<pre>
1338 opaque
1339</pre>
1340
1341<h5>Examples:</h5>
1342
1343<table class="layout">
1344 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame^]1345 <td class="left"><tt>opaque</tt></td>
1346 <td class="left">An opaque type.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1347 </tr>
1348</table>
1349</div>
1350
1351
1352<!-- *********************************************************************** -->
1353<div class="doc_section"> <a name="constants">Constants</a> </div>
1354<!-- *********************************************************************** -->
1355
1356<div class="doc_text">
1357
1358<p>LLVM has several different basic types of constants. This section describes
1359them all and their syntax.</p>
1360
1361</div>
1362
1363<!-- ======================================================================= -->
1364<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1365
1366<div class="doc_text">
1367
1368<dl>
1369 <dt><b>Boolean constants</b></dt>
1370
1371 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1372 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1373 </dd>
1374
1375 <dt><b>Integer constants</b></dt>
1376
1377 <dd>Standard integers (such as '4') are constants of the <a
1378 href="#t_integer">integer</a> type. Negative numbers may be used with
1379 integer types.
1380 </dd>
1381
1382 <dt><b>Floating point constants</b></dt>
1383
1384 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1385 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1386 notation (see below). Floating point constants must have a <a
1387 href="#t_floating">floating point</a> type. </dd>
1388
1389 <dt><b>Null pointer constants</b></dt>
1390
1391 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1392 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1393
1394</dl>
1395
1396<p>The one non-intuitive notation for constants is the optional hexadecimal form
1397of floating point constants. For example, the form '<tt>double
13980x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
13994.5e+15</tt>'. The only time hexadecimal floating point constants are required
1400(and the only time that they are generated by the disassembler) is when a
1401floating point constant must be emitted but it cannot be represented as a
1402decimal floating point number. For example, NaN's, infinities, and other
1403special values are represented in their IEEE hexadecimal format so that
1404assembly and disassembly do not cause any bits to change in the constants.</p>
1405
1406</div>
1407
1408<!-- ======================================================================= -->
1409<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1410</div>
1411
1412<div class="doc_text">
1413<p>Aggregate constants arise from aggregation of simple constants
1414and smaller aggregate constants.</p>
1415
1416<dl>
1417 <dt><b>Structure constants</b></dt>
1418
1419 <dd>Structure constants are represented with notation similar to structure
1420 type definitions (a comma separated list of elements, surrounded by braces
1421 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* %G }</tt>",
1422 where "<tt>%G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1423 must have <a href="#t_struct">structure type</a>, and the number and
1424 types of elements must match those specified by the type.
1425 </dd>
1426
1427 <dt><b>Array constants</b></dt>
1428
1429 <dd>Array constants are represented with notation similar to array type
1430 definitions (a comma separated list of elements, surrounded by square brackets
1431 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1432 constants must have <a href="#t_array">array type</a>, and the number and
1433 types of elements must match those specified by the type.
1434 </dd>
1435
1436 <dt><b>Vector constants</b></dt>
1437
1438 <dd>Vector constants are represented with notation similar to vector type
1439 definitions (a comma separated list of elements, surrounded by
1440 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1441 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1442 href="#t_vector">vector type</a>, and the number and types of elements must
1443 match those specified by the type.
1444 </dd>
1445
1446 <dt><b>Zero initialization</b></dt>
1447
1448 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1449 value to zero of <em>any</em> type, including scalar and aggregate types.
1450 This is often used to avoid having to print large zero initializers (e.g. for
1451 large arrays) and is always exactly equivalent to using explicit zero
1452 initializers.
1453 </dd>
1454</dl>
1455
1456</div>
1457
1458<!-- ======================================================================= -->
1459<div class="doc_subsection">
1460 <a name="globalconstants">Global Variable and Function Addresses</a>
1461</div>
1462
1463<div class="doc_text">
1464
1465<p>The addresses of <a href="#globalvars">global variables</a> and <a
1466href="#functionstructure">functions</a> are always implicitly valid (link-time)
1467constants. These constants are explicitly referenced when the <a
1468href="#identifiers">identifier for the global</a> is used and always have <a
1469href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1470file:</p>
1471
1472<div class="doc_code">
1473<pre>
1474@X = global i32 17
1475@Y = global i32 42
1476@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1477</pre>
1478</div>
1479
1480</div>
1481
1482<!-- ======================================================================= -->
1483<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1484<div class="doc_text">
1485 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1486 no specific value. Undefined values may be of any type and be used anywhere
1487 a constant is permitted.</p>
1488
1489 <p>Undefined values indicate to the compiler that the program is well defined
1490 no matter what value is used, giving the compiler more freedom to optimize.
1491 </p>
1492</div>
1493
1494<!-- ======================================================================= -->
1495<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1496</div>
1497
1498<div class="doc_text">
1499
1500<p>Constant expressions are used to allow expressions involving other constants
1501to be used as constants. Constant expressions may be of any <a
1502href="#t_firstclass">first class</a> type and may involve any LLVM operation
1503that does not have side effects (e.g. load and call are not supported). The
1504following is the syntax for constant expressions:</p>
1505
1506<dl>
1507 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1508 <dd>Truncate a constant to another type. The bit size of CST must be larger
1509 than the bit size of TYPE. Both types must be integers.</dd>
1510
1511 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1512 <dd>Zero extend a constant to another type. The bit size of CST must be
1513 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1514
1515 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1516 <dd>Sign extend a constant to another type. The bit size of CST must be
1517 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1518
1519 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1520 <dd>Truncate a floating point constant to another floating point type. The
1521 size of CST must be larger than the size of TYPE. Both types must be
1522 floating point.</dd>
1523
1524 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1525 <dd>Floating point extend a constant to another type. The size of CST must be
1526 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1527
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1528 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1529 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1530 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1531 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1532 of the same number of elements. If the value won't fit in the integer type,
1533 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1534
1535 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1536 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1537 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1538 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1539 of the same number of elements. If the value won't fit in the integer type,
1540 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1541
1542 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1543 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1544 constant. TYPE must be a scalar or vector floating point type. CST must be of
1545 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1546 of the same number of elements. If the value won't fit in the floating point
1547 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1548
1549 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1550 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1551 constant. TYPE must be a scalar or vector floating point type. CST must be of
1552 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1553 of the same number of elements. If the value won't fit in the floating point
1554 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1555
1556 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1557 <dd>Convert a pointer typed constant to the corresponding integer constant
1558 TYPE must be an integer type. CST must be of pointer type. The CST value is
1559 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1560
1561 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1562 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1563 pointer type. CST must be of integer type. The CST value is zero extended,
1564 truncated, or unchanged to make it fit in a pointer size. This one is
1565 <i>really</i> dangerous!</dd>
1566
1567 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1568 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1569 identical (same number of bits). The conversion is done as if the CST value
1570 was stored to memory and read back as TYPE. In other words, no bits change
1571 with this operator, just the type. This can be used for conversion of
1572 vector types to any other type, as long as they have the same bit width. For
1573 pointers it is only valid to cast to another pointer type.
1574 </dd>
1575
1576 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1577
1578 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1579 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1580 instruction, the index list may have zero or more indexes, which are required
1581 to make sense for the type of "CSTPTR".</dd>
1582
1583 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1584
1585 <dd>Perform the <a href="#i_select">select operation</a> on
1586 constants.</dd>
1587
1588 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1589 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1590
1591 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1592 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1593
1594 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1595
1596 <dd>Perform the <a href="#i_extractelement">extractelement
1597 operation</a> on constants.
1598
1599 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1600
1601 <dd>Perform the <a href="#i_insertelement">insertelement
1602 operation</a> on constants.</dd>
1603
1604
1605 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1606
1607 <dd>Perform the <a href="#i_shufflevector">shufflevector
1608 operation</a> on constants.</dd>
1609
1610 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1611
1612 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1613 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1614 binary</a> operations. The constraints on operands are the same as those for
1615 the corresponding instruction (e.g. no bitwise operations on floating point
1616 values are allowed).</dd>
1617</dl>
1618</div>
1619
1620<!-- *********************************************************************** -->
1621<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1622<!-- *********************************************************************** -->
1623
1624<!-- ======================================================================= -->
1625<div class="doc_subsection">
1626<a name="inlineasm">Inline Assembler Expressions</a>
1627</div>
1628
1629<div class="doc_text">
1630
1631<p>
1632LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1633Module-Level Inline Assembly</a>) through the use of a special value. This
1634value represents the inline assembler as a string (containing the instructions
1635to emit), a list of operand constraints (stored as a string), and a flag that
1636indicates whether or not the inline asm expression has side effects. An example
1637inline assembler expression is:
1638</p>
1639
1640<div class="doc_code">
1641<pre>
1642i32 (i32) asm "bswap $0", "=r,r"
1643</pre>
1644</div>
1645
1646<p>
1647Inline assembler expressions may <b>only</b> be used as the callee operand of
1648a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1649</p>
1650
1651<div class="doc_code">
1652<pre>
1653%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1654</pre>
1655</div>
1656
1657<p>
1658Inline asms with side effects not visible in the constraint list must be marked
1659as having side effects. This is done through the use of the
1660'<tt>sideeffect</tt>' keyword, like so:
1661</p>
1662
1663<div class="doc_code">
1664<pre>
1665call void asm sideeffect "eieio", ""()
1666</pre>
1667</div>
1668
1669<p>TODO: The format of the asm and constraints string still need to be
1670documented here. Constraints on what can be done (e.g. duplication, moving, etc
1671need to be documented).
1672</p>
1673
1674</div>
1675
1676<!-- *********************************************************************** -->
1677<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1678<!-- *********************************************************************** -->
1679
1680<div class="doc_text">
1681
1682<p>The LLVM instruction set consists of several different
1683classifications of instructions: <a href="#terminators">terminator
1684instructions</a>, <a href="#binaryops">binary instructions</a>,
1685<a href="#bitwiseops">bitwise binary instructions</a>, <a
1686 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1687instructions</a>.</p>
1688
1689</div>
1690
1691<!-- ======================================================================= -->
1692<div class="doc_subsection"> <a name="terminators">Terminator
1693Instructions</a> </div>
1694
1695<div class="doc_text">
1696
1697<p>As mentioned <a href="#functionstructure">previously</a>, every
1698basic block in a program ends with a "Terminator" instruction, which
1699indicates which block should be executed after the current block is
1700finished. These terminator instructions typically yield a '<tt>void</tt>'
1701value: they produce control flow, not values (the one exception being
1702the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1703<p>There are six different terminator instructions: the '<a
1704 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1705instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1706the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1707 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1708 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1709
1710</div>
1711
1712<!-- _______________________________________________________________________ -->
1713<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1714Instruction</a> </div>
1715<div class="doc_text">
1716<h5>Syntax:</h5>
1717<pre> ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
1718 ret void <i>; Return from void function</i>
1719</pre>
1720<h5>Overview:</h5>
1721<p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1722value) from a function back to the caller.</p>
1723<p>There are two forms of the '<tt>ret</tt>' instruction: one that
1724returns a value and then causes control flow, and one that just causes
1725control flow to occur.</p>
1726<h5>Arguments:</h5>
1727<p>The '<tt>ret</tt>' instruction may return any '<a
1728 href="#t_firstclass">first class</a>' type. Notice that a function is
1729not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1730instruction inside of the function that returns a value that does not
1731match the return type of the function.</p>
1732<h5>Semantics:</h5>
1733<p>When the '<tt>ret</tt>' instruction is executed, control flow
1734returns back to the calling function's context. If the caller is a "<a
1735 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1736the instruction after the call. If the caller was an "<a
1737 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1738at the beginning of the "normal" destination block. If the instruction
1739returns a value, that value shall set the call or invoke instruction's
1740return value.</p>
1741<h5>Example:</h5>
1742<pre> ret i32 5 <i>; Return an integer value of 5</i>
1743 ret void <i>; Return from a void function</i>
1744</pre>
1745</div>
1746<!-- _______________________________________________________________________ -->
1747<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1748<div class="doc_text">
1749<h5>Syntax:</h5>
1750<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
1751</pre>
1752<h5>Overview:</h5>
1753<p>The '<tt>br</tt>' instruction is used to cause control flow to
1754transfer to a different basic block in the current function. There are
1755two forms of this instruction, corresponding to a conditional branch
1756and an unconditional branch.</p>
1757<h5>Arguments:</h5>
1758<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1759single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1760unconditional form of the '<tt>br</tt>' instruction takes a single
1761'<tt>label</tt>' value as a target.</p>
1762<h5>Semantics:</h5>
1763<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1764argument is evaluated. If the value is <tt>true</tt>, control flows
1765to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1766control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1767<h5>Example:</h5>
1768<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
1769 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1770</div>
1771<!-- _______________________________________________________________________ -->
1772<div class="doc_subsubsection">
1773 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1774</div>
1775
1776<div class="doc_text">
1777<h5>Syntax:</h5>
1778
1779<pre>
1780 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
1781</pre>
1782
1783<h5>Overview:</h5>
1784
1785<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1786several different places. It is a generalization of the '<tt>br</tt>'
1787instruction, allowing a branch to occur to one of many possible
1788destinations.</p>
1789
1790
1791<h5>Arguments:</h5>
1792
1793<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1794comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1795an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1796table is not allowed to contain duplicate constant entries.</p>
1797
1798<h5>Semantics:</h5>
1799
1800<p>The <tt>switch</tt> instruction specifies a table of values and
1801destinations. When the '<tt>switch</tt>' instruction is executed, this
1802table is searched for the given value. If the value is found, control flow is
1803transfered to the corresponding destination; otherwise, control flow is
1804transfered to the default destination.</p>
1805
1806<h5>Implementation:</h5>
1807
1808<p>Depending on properties of the target machine and the particular
1809<tt>switch</tt> instruction, this instruction may be code generated in different
1810ways. For example, it could be generated as a series of chained conditional
1811branches or with a lookup table.</p>
1812
1813<h5>Example:</h5>
1814
1815<pre>
1816 <i>; Emulate a conditional br instruction</i>
1817 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1818 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1819
1820 <i>; Emulate an unconditional br instruction</i>
1821 switch i32 0, label %dest [ ]
1822
1823 <i>; Implement a jump table:</i>
1824 switch i32 %val, label %otherwise [ i32 0, label %onzero
1825 i32 1, label %onone
1826 i32 2, label %ontwo ]
1827</pre>
1828</div>
1829
1830<!-- _______________________________________________________________________ -->
1831<div class="doc_subsubsection">
1832 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1833</div>
1834
1835<div class="doc_text">
1836
1837<h5>Syntax:</h5>
1838
1839<pre>
1840 &lt;result&gt; = invoke [<a href="#callingconv">cconv</a>] &lt;ptr to function ty&gt; %&lt;function ptr val&gt;(&lt;function args&gt;)
1841 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
1842</pre>
1843
1844<h5>Overview:</h5>
1845
1846<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1847function, with the possibility of control flow transfer to either the
1848'<tt>normal</tt>' label or the
1849'<tt>exception</tt>' label. If the callee function returns with the
1850"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1851"normal" label. If the callee (or any indirect callees) returns with the "<a
1852href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1853continued at the dynamically nearest "exception" label.</p>
1854
1855<h5>Arguments:</h5>
1856
1857<p>This instruction requires several arguments:</p>
1858
1859<ol>
1860 <li>
1861 The optional "cconv" marker indicates which <a href="#callingconv">calling
1862 convention</a> the call should use. If none is specified, the call defaults
1863 to using C calling conventions.
1864 </li>
1865 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1866 function value being invoked. In most cases, this is a direct function
1867 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1868 an arbitrary pointer to function value.
1869 </li>
1870
1871 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1872 function to be invoked. </li>
1873
1874 <li>'<tt>function args</tt>': argument list whose types match the function
1875 signature argument types. If the function signature indicates the function
1876 accepts a variable number of arguments, the extra arguments can be
1877 specified. </li>
1878
1879 <li>'<tt>normal label</tt>': the label reached when the called function
1880 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1881
1882 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1883 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1884
1885</ol>
1886
1887<h5>Semantics:</h5>
1888
1889<p>This instruction is designed to operate as a standard '<tt><a
1890href="#i_call">call</a></tt>' instruction in most regards. The primary
1891difference is that it establishes an association with a label, which is used by
1892the runtime library to unwind the stack.</p>
1893
1894<p>This instruction is used in languages with destructors to ensure that proper
1895cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1896exception. Additionally, this is important for implementation of
1897'<tt>catch</tt>' clauses in high-level languages that support them.</p>
1898
1899<h5>Example:</h5>
1900<pre>
1901 %retval = invoke i32 %Test(i32 15) to label %Continue
1902 unwind label %TestCleanup <i>; {i32}:retval set</i>
1903 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1904 unwind label %TestCleanup <i>; {i32}:retval set</i>
1905</pre>
1906</div>
1907
1908
1909<!-- _______________________________________________________________________ -->
1910
1911<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1912Instruction</a> </div>
1913
1914<div class="doc_text">
1915
1916<h5>Syntax:</h5>
1917<pre>
1918 unwind
1919</pre>
1920
1921<h5>Overview:</h5>
1922
1923<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1924at the first callee in the dynamic call stack which used an <a
1925href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1926primarily used to implement exception handling.</p>
1927
1928<h5>Semantics:</h5>
1929
1930<p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1931immediately halt. The dynamic call stack is then searched for the first <a
1932href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1933execution continues at the "exceptional" destination block specified by the
1934<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1935dynamic call chain, undefined behavior results.</p>
1936</div>
1937
1938<!-- _______________________________________________________________________ -->
1939
1940<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1941Instruction</a> </div>
1942
1943<div class="doc_text">
1944
1945<h5>Syntax:</h5>
1946<pre>
1947 unreachable
1948</pre>
1949
1950<h5>Overview:</h5>
1951
1952<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1953instruction is used to inform the optimizer that a particular portion of the
1954code is not reachable. This can be used to indicate that the code after a
1955no-return function cannot be reached, and other facts.</p>
1956
1957<h5>Semantics:</h5>
1958
1959<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1960</div>
1961
1962
1963
1964<!-- ======================================================================= -->
1965<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1966<div class="doc_text">
1967<p>Binary operators are used to do most of the computation in a
1968program. They require two operands, execute an operation on them, and
1969produce a single value. The operands might represent
1970multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
1971The result value of a binary operator is not
1972necessarily the same type as its operands.</p>
1973<p>There are several different binary operators:</p>
1974</div>
1975<!-- _______________________________________________________________________ -->
1976<div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1977Instruction</a> </div>
1978<div class="doc_text">
1979<h5>Syntax:</h5>
1980<pre> &lt;result&gt; = add &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
1981</pre>
1982<h5>Overview:</h5>
1983<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1984<h5>Arguments:</h5>
1985<p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1986 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1987 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1988Both arguments must have identical types.</p>
1989<h5>Semantics:</h5>
1990<p>The value produced is the integer or floating point sum of the two
1991operands.</p>
1992<h5>Example:</h5>
1993<pre> &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
1994</pre>
1995</div>
1996<!-- _______________________________________________________________________ -->
1997<div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1998Instruction</a> </div>
1999<div class="doc_text">
2000<h5>Syntax:</h5>
2001<pre> &lt;result&gt; = sub &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2002</pre>
2003<h5>Overview:</h5>
2004<p>The '<tt>sub</tt>' instruction returns the difference of its two
2005operands.</p>
2006<p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
2007instruction present in most other intermediate representations.</p>
2008<h5>Arguments:</h5>
2009<p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
2010 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2011values.
2012This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2013Both arguments must have identical types.</p>
2014<h5>Semantics:</h5>
2015<p>The value produced is the integer or floating point difference of
2016the two operands.</p>
2017<h5>Example:</h5>
2018<pre>
2019 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2020 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2021</pre>
2022</div>
2023<!-- _______________________________________________________________________ -->
2024<div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
2025Instruction</a> </div>
2026<div class="doc_text">
2027<h5>Syntax:</h5>
2028<pre> &lt;result&gt; = mul &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2029</pre>
2030<h5>Overview:</h5>
2031<p>The '<tt>mul</tt>' instruction returns the product of its two
2032operands.</p>
2033<h5>Arguments:</h5>
2034<p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
2035 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2036values.
2037This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2038Both arguments must have identical types.</p>
2039<h5>Semantics:</h5>
2040<p>The value produced is the integer or floating point product of the
2041two operands.</p>
2042<p>Because the operands are the same width, the result of an integer
2043multiplication is the same whether the operands should be deemed unsigned or
2044signed.</p>
2045<h5>Example:</h5>
2046<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2047</pre>
2048</div>
2049<!-- _______________________________________________________________________ -->
2050<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2051</a></div>
2052<div class="doc_text">
2053<h5>Syntax:</h5>
2054<pre> &lt;result&gt; = udiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2055</pre>
2056<h5>Overview:</h5>
2057<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2058operands.</p>
2059<h5>Arguments:</h5>
2060<p>The two arguments to the '<tt>udiv</tt>' instruction must be
2061<a href="#t_integer">integer</a> values. Both arguments must have identical
2062types. This instruction can also take <a href="#t_vector">vector</a> versions
2063of the values in which case the elements must be integers.</p>
2064<h5>Semantics:</h5>
2065<p>The value produced is the unsigned integer quotient of the two operands. This
2066instruction always performs an unsigned division operation, regardless of
2067whether the arguments are unsigned or not.</p>
2068<h5>Example:</h5>
2069<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2070</pre>
2071</div>
2072<!-- _______________________________________________________________________ -->
2073<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2074</a> </div>
2075<div class="doc_text">
2076<h5>Syntax:</h5>
2077<pre> &lt;result&gt; = sdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2078</pre>
2079<h5>Overview:</h5>
2080<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2081operands.</p>
2082<h5>Arguments:</h5>
2083<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2084<a href="#t_integer">integer</a> values. Both arguments must have identical
2085types. This instruction can also take <a href="#t_vector">vector</a> versions
2086of the values in which case the elements must be integers.</p>
2087<h5>Semantics:</h5>
2088<p>The value produced is the signed integer quotient of the two operands. This
2089instruction always performs a signed division operation, regardless of whether
2090the arguments are signed or not.</p>
2091<h5>Example:</h5>
2092<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2093</pre>
2094</div>
2095<!-- _______________________________________________________________________ -->
2096<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2097Instruction</a> </div>
2098<div class="doc_text">
2099<h5>Syntax:</h5>
2100<pre> &lt;result&gt; = fdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2101</pre>
2102<h5>Overview:</h5>
2103<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2104operands.</p>
2105<h5>Arguments:</h5>
2106<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2107<a href="#t_floating">floating point</a> values. Both arguments must have
2108identical types. This instruction can also take <a href="#t_vector">vector</a>
2109versions of floating point values.</p>
2110<h5>Semantics:</h5>
2111<p>The value produced is the floating point quotient of the two operands.</p>
2112<h5>Example:</h5>
2113<pre> &lt;result&gt; = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2114</pre>
2115</div>
2116<!-- _______________________________________________________________________ -->
2117<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2118</div>
2119<div class="doc_text">
2120<h5>Syntax:</h5>
2121<pre> &lt;result&gt; = urem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2122</pre>
2123<h5>Overview:</h5>
2124<p>The '<tt>urem</tt>' instruction returns the remainder from the
2125unsigned division of its two arguments.</p>
2126<h5>Arguments:</h5>
2127<p>The two arguments to the '<tt>urem</tt>' instruction must be
2128<a href="#t_integer">integer</a> values. Both arguments must have identical
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2129types. This instruction can also take <a href="#t_vector">vector</a> versions
2130of the values in which case the elements must be integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2131<h5>Semantics:</h5>
2132<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2133This instruction always performs an unsigned division to get the remainder,
2134regardless of whether the arguments are unsigned or not.</p>
2135<h5>Example:</h5>
2136<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2137</pre>
2138
2139</div>
2140<!-- _______________________________________________________________________ -->
2141<div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2142Instruction</a> </div>
2143<div class="doc_text">
2144<h5>Syntax:</h5>
2145<pre> &lt;result&gt; = srem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2146</pre>
2147<h5>Overview:</h5>
2148<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2149signed division of its two operands. This instruction can also take
2150<a href="#t_vector">vector</a> versions of the values in which case
2151the elements must be integers.</p>
2152</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2153<h5>Arguments:</h5>
2154<p>The two arguments to the '<tt>srem</tt>' instruction must be
2155<a href="#t_integer">integer</a> values. Both arguments must have identical
2156types.</p>
2157<h5>Semantics:</h5>
2158<p>This instruction returns the <i>remainder</i> of a division (where the result
2159has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2160operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2161a value. For more information about the difference, see <a
2162 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2163Math Forum</a>. For a table of how this is implemented in various languages,
2164please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2165Wikipedia: modulo operation</a>.</p>
2166<h5>Example:</h5>
2167<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2168</pre>
2169
2170</div>
2171<!-- _______________________________________________________________________ -->
2172<div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2173Instruction</a> </div>
2174<div class="doc_text">
2175<h5>Syntax:</h5>
2176<pre> &lt;result&gt; = frem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2177</pre>
2178<h5>Overview:</h5>
2179<p>The '<tt>frem</tt>' instruction returns the remainder from the
2180division of its two operands.</p>
2181<h5>Arguments:</h5>
2182<p>The two arguments to the '<tt>frem</tt>' instruction must be
2183<a href="#t_floating">floating point</a> values. Both arguments must have
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2184identical types. This instruction can also take <a href="#t_vector">vector</a>
2185versions of floating point values.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2186<h5>Semantics:</h5>
2187<p>This instruction returns the <i>remainder</i> of a division.</p>
2188<h5>Example:</h5>
2189<pre> &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2190</pre>
2191</div>
2192
2193<!-- ======================================================================= -->
2194<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2195Operations</a> </div>
2196<div class="doc_text">
2197<p>Bitwise binary operators are used to do various forms of
2198bit-twiddling in a program. They are generally very efficient
2199instructions and can commonly be strength reduced from other
2200instructions. They require two operands, execute an operation on them,
2201and produce a single value. The resulting value of the bitwise binary
2202operators is always the same type as its first operand.</p>
2203</div>
2204
2205<!-- _______________________________________________________________________ -->
2206<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2207Instruction</a> </div>
2208<div class="doc_text">
2209<h5>Syntax:</h5>
2210<pre> &lt;result&gt; = shl &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2211</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2212
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2213<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2214
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2215<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2216the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2217
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2218<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2219
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2220<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2221 href="#t_integer">integer</a> type.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2222
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2223<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2224
2225<p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>. If
2226<tt>var2</tt> is (statically or dynamically) equal to or larger than the number
2227of bits in <tt>var1</tt>, the result is undefined.</p>
2228
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2229<h5>Example:</h5><pre>
2230 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2231 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2232 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2233 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2234</pre>
2235</div>
2236<!-- _______________________________________________________________________ -->
2237<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2238Instruction</a> </div>
2239<div class="doc_text">
2240<h5>Syntax:</h5>
2241<pre> &lt;result&gt; = lshr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2242</pre>
2243
2244<h5>Overview:</h5>
2245<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2246operand shifted to the right a specified number of bits with zero fill.</p>
2247
2248<h5>Arguments:</h5>
2249<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2250<a href="#t_integer">integer</a> type.</p>
2251
2252<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2253
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2254<p>This instruction always performs a logical shift right operation. The most
2255significant bits of the result will be filled with zero bits after the
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2256shift. If <tt>var2</tt> is (statically or dynamically) equal to or larger than
2257the number of bits in <tt>var1</tt>, the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2258
2259<h5>Example:</h5>
2260<pre>
2261 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2262 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2263 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2264 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2265 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2266</pre>
2267</div>
2268
2269<!-- _______________________________________________________________________ -->
2270<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2271Instruction</a> </div>
2272<div class="doc_text">
2273
2274<h5>Syntax:</h5>
2275<pre> &lt;result&gt; = ashr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2276</pre>
2277
2278<h5>Overview:</h5>
2279<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2280operand shifted to the right a specified number of bits with sign extension.</p>
2281
2282<h5>Arguments:</h5>
2283<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2284<a href="#t_integer">integer</a> type.</p>
2285
2286<h5>Semantics:</h5>
2287<p>This instruction always performs an arithmetic shift right operation,
2288The most significant bits of the result will be filled with the sign bit
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2289of <tt>var1</tt>. If <tt>var2</tt> is (statically or dynamically) equal to or
2290larger than the number of bits in <tt>var1</tt>, the result is undefined.
2291</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2292
2293<h5>Example:</h5>
2294<pre>
2295 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2296 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2297 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2298 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2299 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2300</pre>
2301</div>
2302
2303<!-- _______________________________________________________________________ -->
2304<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2305Instruction</a> </div>
2306<div class="doc_text">
2307<h5>Syntax:</h5>
2308<pre> &lt;result&gt; = and &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2309</pre>
2310<h5>Overview:</h5>
2311<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2312its two operands.</p>
2313<h5>Arguments:</h5>
2314<p>The two arguments to the '<tt>and</tt>' instruction must be <a
2315 href="#t_integer">integer</a> values. Both arguments must have
2316identical types.</p>
2317<h5>Semantics:</h5>
2318<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2319<p> </p>
2320<div style="align: center">
2321<table border="1" cellspacing="0" cellpadding="4">
2322 <tbody>
2323 <tr>
2324 <td>In0</td>
2325 <td>In1</td>
2326 <td>Out</td>
2327 </tr>
2328 <tr>
2329 <td>0</td>
2330 <td>0</td>
2331 <td>0</td>
2332 </tr>
2333 <tr>
2334 <td>0</td>
2335 <td>1</td>
2336 <td>0</td>
2337 </tr>
2338 <tr>
2339 <td>1</td>
2340 <td>0</td>
2341 <td>0</td>
2342 </tr>
2343 <tr>
2344 <td>1</td>
2345 <td>1</td>
2346 <td>1</td>
2347 </tr>
2348 </tbody>
2349</table>
2350</div>
2351<h5>Example:</h5>
2352<pre> &lt;result&gt; = and i32 4, %var <i>; yields {i32}:result = 4 &amp; %var</i>
2353 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2354 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2355</pre>
2356</div>
2357<!-- _______________________________________________________________________ -->
2358<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2359<div class="doc_text">
2360<h5>Syntax:</h5>
2361<pre> &lt;result&gt; = or &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2362</pre>
2363<h5>Overview:</h5>
2364<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2365or of its two operands.</p>
2366<h5>Arguments:</h5>
2367<p>The two arguments to the '<tt>or</tt>' instruction must be <a
2368 href="#t_integer">integer</a> values. Both arguments must have
2369identical types.</p>
2370<h5>Semantics:</h5>
2371<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2372<p> </p>
2373<div style="align: center">
2374<table border="1" cellspacing="0" cellpadding="4">
2375 <tbody>
2376 <tr>
2377 <td>In0</td>
2378 <td>In1</td>
2379 <td>Out</td>
2380 </tr>
2381 <tr>
2382 <td>0</td>
2383 <td>0</td>
2384 <td>0</td>
2385 </tr>
2386 <tr>
2387 <td>0</td>
2388 <td>1</td>
2389 <td>1</td>
2390 </tr>
2391 <tr>
2392 <td>1</td>
2393 <td>0</td>
2394 <td>1</td>
2395 </tr>
2396 <tr>
2397 <td>1</td>
2398 <td>1</td>
2399 <td>1</td>
2400 </tr>
2401 </tbody>
2402</table>
2403</div>
2404<h5>Example:</h5>
2405<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2406 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2407 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2408</pre>
2409</div>
2410<!-- _______________________________________________________________________ -->
2411<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2412Instruction</a> </div>
2413<div class="doc_text">
2414<h5>Syntax:</h5>
2415<pre> &lt;result&gt; = xor &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2416</pre>
2417<h5>Overview:</h5>
2418<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2419or of its two operands. The <tt>xor</tt> is used to implement the
2420"one's complement" operation, which is the "~" operator in C.</p>
2421<h5>Arguments:</h5>
2422<p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2423 href="#t_integer">integer</a> values. Both arguments must have
2424identical types.</p>
2425<h5>Semantics:</h5>
2426<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2427<p> </p>
2428<div style="align: center">
2429<table border="1" cellspacing="0" cellpadding="4">
2430 <tbody>
2431 <tr>
2432 <td>In0</td>
2433 <td>In1</td>
2434 <td>Out</td>
2435 </tr>
2436 <tr>
2437 <td>0</td>
2438 <td>0</td>
2439 <td>0</td>
2440 </tr>
2441 <tr>
2442 <td>0</td>
2443 <td>1</td>
2444 <td>1</td>
2445 </tr>
2446 <tr>
2447 <td>1</td>
2448 <td>0</td>
2449 <td>1</td>
2450 </tr>
2451 <tr>
2452 <td>1</td>
2453 <td>1</td>
2454 <td>0</td>
2455 </tr>
2456 </tbody>
2457</table>
2458</div>
2459<p> </p>
2460<h5>Example:</h5>
2461<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2462 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2463 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2464 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2465</pre>
2466</div>
2467
2468<!-- ======================================================================= -->
2469<div class="doc_subsection">
2470 <a name="vectorops">Vector Operations</a>
2471</div>
2472
2473<div class="doc_text">
2474
2475<p>LLVM supports several instructions to represent vector operations in a
2476target-independent manner. These instructions cover the element-access and
2477vector-specific operations needed to process vectors effectively. While LLVM
2478does directly support these vector operations, many sophisticated algorithms
2479will want to use target-specific intrinsics to take full advantage of a specific
2480target.</p>
2481
2482</div>
2483
2484<!-- _______________________________________________________________________ -->
2485<div class="doc_subsubsection">
2486 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2487</div>
2488
2489<div class="doc_text">
2490
2491<h5>Syntax:</h5>
2492
2493<pre>
2494 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
2495</pre>
2496
2497<h5>Overview:</h5>
2498
2499<p>
2500The '<tt>extractelement</tt>' instruction extracts a single scalar
2501element from a vector at a specified index.
2502</p>
2503
2504
2505<h5>Arguments:</h5>
2506
2507<p>
2508The first operand of an '<tt>extractelement</tt>' instruction is a
2509value of <a href="#t_vector">vector</a> type. The second operand is
2510an index indicating the position from which to extract the element.
2511The index may be a variable.</p>
2512
2513<h5>Semantics:</h5>
2514
2515<p>
2516The result is a scalar of the same type as the element type of
2517<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2518<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2519results are undefined.
2520</p>
2521
2522<h5>Example:</h5>
2523
2524<pre>
2525 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
2526</pre>
2527</div>
2528
2529
2530<!-- _______________________________________________________________________ -->
2531<div class="doc_subsubsection">
2532 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2533</div>
2534
2535<div class="doc_text">
2536
2537<h5>Syntax:</h5>
2538
2539<pre>
2540 &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>
2541</pre>
2542
2543<h5>Overview:</h5>
2544
2545<p>
2546The '<tt>insertelement</tt>' instruction inserts a scalar
2547element into a vector at a specified index.
2548</p>
2549
2550
2551<h5>Arguments:</h5>
2552
2553<p>
2554The first operand of an '<tt>insertelement</tt>' instruction is a
2555value of <a href="#t_vector">vector</a> type. The second operand is a
2556scalar value whose type must equal the element type of the first
2557operand. The third operand is an index indicating the position at
2558which to insert the value. The index may be a variable.</p>
2559
2560<h5>Semantics:</h5>
2561
2562<p>
2563The result is a vector of the same type as <tt>val</tt>. Its
2564element values are those of <tt>val</tt> except at position
2565<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2566exceeds the length of <tt>val</tt>, the results are undefined.
2567</p>
2568
2569<h5>Example:</h5>
2570
2571<pre>
2572 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
2573</pre>
2574</div>
2575
2576<!-- _______________________________________________________________________ -->
2577<div class="doc_subsubsection">
2578 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2579</div>
2580
2581<div class="doc_text">
2582
2583<h5>Syntax:</h5>
2584
2585<pre>
2586 &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>
2587</pre>
2588
2589<h5>Overview:</h5>
2590
2591<p>
2592The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2593from two input vectors, returning a vector of the same type.
2594</p>
2595
2596<h5>Arguments:</h5>
2597
2598<p>
2599The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2600with types that match each other and types that match the result of the
2601instruction. The third argument is a shuffle mask, which has the same number
2602of elements as the other vector type, but whose element type is always 'i32'.
2603</p>
2604
2605<p>
2606The shuffle mask operand is required to be a constant vector with either
2607constant integer or undef values.
2608</p>
2609
2610<h5>Semantics:</h5>
2611
2612<p>
2613The elements of the two input vectors are numbered from left to right across
2614both of the vectors. The shuffle mask operand specifies, for each element of
2615the result vector, which element of the two input registers the result element
2616gets. The element selector may be undef (meaning "don't care") and the second
2617operand may be undef if performing a shuffle from only one vector.
2618</p>
2619
2620<h5>Example:</h5>
2621
2622<pre>
2623 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
2624 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
2625 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
2626 &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.
2627</pre>
2628</div>
2629
2630
2631<!-- ======================================================================= -->
2632<div class="doc_subsection">
2633 <a name="memoryops">Memory Access and Addressing Operations</a>
2634</div>
2635
2636<div class="doc_text">
2637
2638<p>A key design point of an SSA-based representation is how it
2639represents memory. In LLVM, no memory locations are in SSA form, which
2640makes things very simple. This section describes how to read, write,
2641allocate, and free memory in LLVM.</p>
2642
2643</div>
2644
2645<!-- _______________________________________________________________________ -->
2646<div class="doc_subsubsection">
2647 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2648</div>
2649
2650<div class="doc_text">
2651
2652<h5>Syntax:</h5>
2653
2654<pre>
2655 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
2656</pre>
2657
2658<h5>Overview:</h5>
2659
2660<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]2661heap and returns a pointer to it. The object is always allocated in the generic
2662address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2663
2664<h5>Arguments:</h5>
2665
2666<p>The '<tt>malloc</tt>' instruction allocates
2667<tt>sizeof(&lt;type&gt;)*NumElements</tt>
2668bytes of memory from the operating system and returns a pointer of the
2669appropriate type to the program. If "NumElements" is specified, it is the
2670number of elements allocated. If an alignment is specified, the value result
2671of the allocation is guaranteed to be aligned to at least that boundary. If
2672not specified, or if zero, the target can choose to align the allocation on any
2673convenient boundary.</p>
2674
2675<p>'<tt>type</tt>' must be a sized type.</p>
2676
2677<h5>Semantics:</h5>
2678
2679<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2680a pointer is returned.</p>
2681
2682<h5>Example:</h5>
2683
2684<pre>
2685 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2686
2687 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2688 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2689 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2690 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2691 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2692</pre>
2693</div>
2694
2695<!-- _______________________________________________________________________ -->
2696<div class="doc_subsubsection">
2697 <a name="i_free">'<tt>free</tt>' Instruction</a>
2698</div>
2699
2700<div class="doc_text">
2701
2702<h5>Syntax:</h5>
2703
2704<pre>
2705 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
2706</pre>
2707
2708<h5>Overview:</h5>
2709
2710<p>The '<tt>free</tt>' instruction returns memory back to the unused
2711memory heap to be reallocated in the future.</p>
2712
2713<h5>Arguments:</h5>
2714
2715<p>'<tt>value</tt>' shall be a pointer value that points to a value
2716that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2717instruction.</p>
2718
2719<h5>Semantics:</h5>
2720
2721<p>Access to the memory pointed to by the pointer is no longer defined
2722after this instruction executes.</p>
2723
2724<h5>Example:</h5>
2725
2726<pre>
2727 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2728 free [4 x i8]* %array
2729</pre>
2730</div>
2731
2732<!-- _______________________________________________________________________ -->
2733<div class="doc_subsubsection">
2734 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2735</div>
2736
2737<div class="doc_text">
2738
2739<h5>Syntax:</h5>
2740
2741<pre>
2742 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
2743</pre>
2744
2745<h5>Overview:</h5>
2746
2747<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
2748currently executing function, to be automatically released when this function
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]2749returns to its caller. The object is always allocated in the generic address
2750space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2751
2752<h5>Arguments:</h5>
2753
2754<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
2755bytes of memory on the runtime stack, returning a pointer of the
2756appropriate type to the program. If "NumElements" is specified, it is the
2757number of elements allocated. If an alignment is specified, the value result
2758of the allocation is guaranteed to be aligned to at least that boundary. If
2759not specified, or if zero, the target can choose to align the allocation on any
2760convenient boundary.</p>
2761
2762<p>'<tt>type</tt>' may be any sized type.</p>
2763
2764<h5>Semantics:</h5>
2765
2766<p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2767memory is automatically released when the function returns. The '<tt>alloca</tt>'
2768instruction is commonly used to represent automatic variables that must
2769have an address available. When the function returns (either with the <tt><a
2770 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2771instructions), the memory is reclaimed.</p>
2772
2773<h5>Example:</h5>
2774
2775<pre>
2776 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2777 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2778 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2779 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2780</pre>
2781</div>
2782
2783<!-- _______________________________________________________________________ -->
2784<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2785Instruction</a> </div>
2786<div class="doc_text">
2787<h5>Syntax:</h5>
2788<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>
2789<h5>Overview:</h5>
2790<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2791<h5>Arguments:</h5>
2792<p>The argument to the '<tt>load</tt>' instruction specifies the memory
2793address from which to load. The pointer must point to a <a
2794 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2795marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2796the number or order of execution of this <tt>load</tt> with other
2797volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2798instructions. </p>
2799<h5>Semantics:</h5>
2800<p>The location of memory pointed to is loaded.</p>
2801<h5>Examples:</h5>
2802<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2803 <a
2804 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2805 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2806</pre>
2807</div>
2808<!-- _______________________________________________________________________ -->
2809<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2810Instruction</a> </div>
2811<div class="doc_text">
2812<h5>Syntax:</h5>
2813<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
2814 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
2815</pre>
2816<h5>Overview:</h5>
2817<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2818<h5>Arguments:</h5>
2819<p>There are two arguments to the '<tt>store</tt>' instruction: a value
2820to store and an address at which to store it. The type of the '<tt>&lt;pointer&gt;</tt>'
2821operand must be a pointer to the type of the '<tt>&lt;value&gt;</tt>'
2822operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2823optimizer is not allowed to modify the number or order of execution of
2824this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2825 href="#i_store">store</a></tt> instructions.</p>
2826<h5>Semantics:</h5>
2827<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
2828at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
2829<h5>Example:</h5>
2830<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]2831 store i32 3, i32* %ptr <i>; yields {void}</i>
2832 %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]2833</pre>
2834</div>
2835
2836<!-- _______________________________________________________________________ -->
2837<div class="doc_subsubsection">
2838 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2839</div>
2840
2841<div class="doc_text">
2842<h5>Syntax:</h5>
2843<pre>
2844 &lt;result&gt; = getelementptr &lt;ty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
2845</pre>
2846
2847<h5>Overview:</h5>
2848
2849<p>
2850The '<tt>getelementptr</tt>' instruction is used to get the address of a
2851subelement of an aggregate data structure.</p>
2852
2853<h5>Arguments:</h5>
2854
2855<p>This instruction takes a list of integer operands that indicate what
2856elements of the aggregate object to index to. The actual types of the arguments
2857provided depend on the type of the first pointer argument. The
2858'<tt>getelementptr</tt>' instruction is used to index down through the type
2859levels of a structure or to a specific index in an array. When indexing into a
2860structure, only <tt>i32</tt> integer constants are allowed. When indexing
2861into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2862be sign extended to 64-bit values.</p>
2863
2864<p>For example, let's consider a C code fragment and how it gets
2865compiled to LLVM:</p>
2866
2867<div class="doc_code">
2868<pre>
2869struct RT {
2870 char A;
2871 int B[10][20];
2872 char C;
2873};
2874struct ST {
2875 int X;
2876 double Y;
2877 struct RT Z;
2878};
2879
2880int *foo(struct ST *s) {
2881 return &amp;s[1].Z.B[5][13];
2882}
2883</pre>
2884</div>
2885
2886<p>The LLVM code generated by the GCC frontend is:</p>
2887
2888<div class="doc_code">
2889<pre>
2890%RT = type { i8 , [10 x [20 x i32]], i8 }
2891%ST = type { i32, double, %RT }
2892
2893define i32* %foo(%ST* %s) {
2894entry:
2895 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
2896 ret i32* %reg
2897}
2898</pre>
2899</div>
2900
2901<h5>Semantics:</h5>
2902
2903<p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2904on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2905and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
2906<a href="#t_integer">integer</a> type but the value will always be sign extended
2907to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
2908<b>constants</b>.</p>
2909
2910<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2911type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
2912}</tt>' type, a structure. The second index indexes into the third element of
2913the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
2914i8 }</tt>' type, another structure. The third index indexes into the second
2915element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
2916array. The two dimensions of the array are subscripted into, yielding an
2917'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2918to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
2919
2920<p>Note that it is perfectly legal to index partially through a
2921structure, returning a pointer to an inner element. Because of this,
2922the LLVM code for the given testcase is equivalent to:</p>
2923
2924<pre>
2925 define i32* %foo(%ST* %s) {
2926 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
2927 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
2928 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
2929 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
2930 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
2931 ret i32* %t5
2932 }
2933</pre>
2934
2935<p>Note that it is undefined to access an array out of bounds: array and
2936pointer indexes must always be within the defined bounds of the array type.
2937The one exception for this rules is zero length arrays. These arrays are
2938defined to be accessible as variable length arrays, which requires access
2939beyond the zero'th element.</p>
2940
2941<p>The getelementptr instruction is often confusing. For some more insight
2942into how it works, see <a href="GetElementPtr.html">the getelementptr
2943FAQ</a>.</p>
2944
2945<h5>Example:</h5>
2946
2947<pre>
2948 <i>; yields [12 x i8]*:aptr</i>
2949 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
2950</pre>
2951</div>
2952
2953<!-- ======================================================================= -->
2954<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2955</div>
2956<div class="doc_text">
2957<p>The instructions in this category are the conversion instructions (casting)
2958which all take a single operand and a type. They perform various bit conversions
2959on the operand.</p>
2960</div>
2961
2962<!-- _______________________________________________________________________ -->
2963<div class="doc_subsubsection">
2964 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2965</div>
2966<div class="doc_text">
2967
2968<h5>Syntax:</h5>
2969<pre>
2970 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
2971</pre>
2972
2973<h5>Overview:</h5>
2974<p>
2975The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2976</p>
2977
2978<h5>Arguments:</h5>
2979<p>
2980The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2981be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2982and type of the result, which must be an <a href="#t_integer">integer</a>
2983type. The bit size of <tt>value</tt> must be larger than the bit size of
2984<tt>ty2</tt>. Equal sized types are not allowed.</p>
2985
2986<h5>Semantics:</h5>
2987<p>
2988The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
2989and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
2990larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
2991It will always truncate bits.</p>
2992
2993<h5>Example:</h5>
2994<pre>
2995 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
2996 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
2997 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
2998</pre>
2999</div>
3000
3001<!-- _______________________________________________________________________ -->
3002<div class="doc_subsubsection">
3003 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3004</div>
3005<div class="doc_text">
3006
3007<h5>Syntax:</h5>
3008<pre>
3009 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3010</pre>
3011
3012<h5>Overview:</h5>
3013<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3014<tt>ty2</tt>.</p>
3015
3016
3017<h5>Arguments:</h5>
3018<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3019<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3020also be of <a href="#t_integer">integer</a> type. The bit size of the
3021<tt>value</tt> must be smaller than the bit size of the destination type,
3022<tt>ty2</tt>.</p>
3023
3024<h5>Semantics:</h5>
3025<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3026bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3027
3028<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3029
3030<h5>Example:</h5>
3031<pre>
3032 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3033 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3034</pre>
3035</div>
3036
3037<!-- _______________________________________________________________________ -->
3038<div class="doc_subsubsection">
3039 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3040</div>
3041<div class="doc_text">
3042
3043<h5>Syntax:</h5>
3044<pre>
3045 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3046</pre>
3047
3048<h5>Overview:</h5>
3049<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3050
3051<h5>Arguments:</h5>
3052<p>
3053The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3054<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3055also be of <a href="#t_integer">integer</a> type. The bit size of the
3056<tt>value</tt> must be smaller than the bit size of the destination type,
3057<tt>ty2</tt>.</p>
3058
3059<h5>Semantics:</h5>
3060<p>
3061The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3062bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3063the type <tt>ty2</tt>.</p>
3064
3065<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3066
3067<h5>Example:</h5>
3068<pre>
3069 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3070 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3071</pre>
3072</div>
3073
3074<!-- _______________________________________________________________________ -->
3075<div class="doc_subsubsection">
3076 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3077</div>
3078
3079<div class="doc_text">
3080
3081<h5>Syntax:</h5>
3082
3083<pre>
3084 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3085</pre>
3086
3087<h5>Overview:</h5>
3088<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3089<tt>ty2</tt>.</p>
3090
3091
3092<h5>Arguments:</h5>
3093<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3094 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3095cast it to. The size of <tt>value</tt> must be larger than the size of
3096<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3097<i>no-op cast</i>.</p>
3098
3099<h5>Semantics:</h5>
3100<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3101<a href="#t_floating">floating point</a> type to a smaller
3102<a href="#t_floating">floating point</a> type. If the value cannot fit within
3103the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3104
3105<h5>Example:</h5>
3106<pre>
3107 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3108 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3109</pre>
3110</div>
3111
3112<!-- _______________________________________________________________________ -->
3113<div class="doc_subsubsection">
3114 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3115</div>
3116<div class="doc_text">
3117
3118<h5>Syntax:</h5>
3119<pre>
3120 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3121</pre>
3122
3123<h5>Overview:</h5>
3124<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3125floating point value.</p>
3126
3127<h5>Arguments:</h5>
3128<p>The '<tt>fpext</tt>' instruction takes a
3129<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3130and a <a href="#t_floating">floating point</a> type to cast it to. The source
3131type must be smaller than the destination type.</p>
3132
3133<h5>Semantics:</h5>
3134<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3135<a href="#t_floating">floating point</a> type to a larger
3136<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3137used to make a <i>no-op cast</i> because it always changes bits. Use
3138<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3139
3140<h5>Example:</h5>
3141<pre>
3142 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3143 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3144</pre>
3145</div>
3146
3147<!-- _______________________________________________________________________ -->
3148<div class="doc_subsubsection">
3149 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3150</div>
3151<div class="doc_text">
3152
3153<h5>Syntax:</h5>
3154<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3155 &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]3156</pre>
3157
3158<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3159<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3160unsigned integer equivalent of type <tt>ty2</tt>.
3161</p>
3162
3163<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3164<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]3165scalar or vector <a href="#t_floating">floating point</a> value, and a type
3166to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3167type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3168vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3169
3170<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3171<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3172<a href="#t_floating">floating point</a> operand into the nearest (rounding
3173towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3174the results are undefined.</p>
3175
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3176<h5>Example:</h5>
3177<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3178 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3179 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3180 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3181</pre>
3182</div>
3183
3184<!-- _______________________________________________________________________ -->
3185<div class="doc_subsubsection">
3186 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3187</div>
3188<div class="doc_text">
3189
3190<h5>Syntax:</h5>
3191<pre>
3192 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3193</pre>
3194
3195<h5>Overview:</h5>
3196<p>The '<tt>fptosi</tt>' instruction converts
3197<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3198</p>
3199
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3200<h5>Arguments:</h5>
3201<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]3202scalar or vector <a href="#t_floating">floating point</a> value, and a type
3203to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3204type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3205vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3206
3207<h5>Semantics:</h5>
3208<p>The '<tt>fptosi</tt>' instruction converts its
3209<a href="#t_floating">floating point</a> operand into the nearest (rounding
3210towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3211the results are undefined.</p>
3212
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3213<h5>Example:</h5>
3214<pre>
3215 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3216 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3217 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3218</pre>
3219</div>
3220
3221<!-- _______________________________________________________________________ -->
3222<div class="doc_subsubsection">
3223 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3224</div>
3225<div class="doc_text">
3226
3227<h5>Syntax:</h5>
3228<pre>
3229 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3230</pre>
3231
3232<h5>Overview:</h5>
3233<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3234integer and converts that value to the <tt>ty2</tt> type.</p>
3235
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3236<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3237<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3238scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3239to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3240type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3241floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3242
3243<h5>Semantics:</h5>
3244<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3245integer quantity and converts it to the corresponding floating point value. If
3246the value cannot fit in the floating point value, the results are undefined.</p>
3247
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3248<h5>Example:</h5>
3249<pre>
3250 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3251 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3252</pre>
3253</div>
3254
3255<!-- _______________________________________________________________________ -->
3256<div class="doc_subsubsection">
3257 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3258</div>
3259<div class="doc_text">
3260
3261<h5>Syntax:</h5>
3262<pre>
3263 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3264</pre>
3265
3266<h5>Overview:</h5>
3267<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3268integer and converts that value to the <tt>ty2</tt> type.</p>
3269
3270<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3271<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3272scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3273to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3274type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3275floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3276
3277<h5>Semantics:</h5>
3278<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3279integer quantity and converts it to the corresponding floating point value. If
3280the value cannot fit in the floating point value, the results are undefined.</p>
3281
3282<h5>Example:</h5>
3283<pre>
3284 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3285 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3286</pre>
3287</div>
3288
3289<!-- _______________________________________________________________________ -->
3290<div class="doc_subsubsection">
3291 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3292</div>
3293<div class="doc_text">
3294
3295<h5>Syntax:</h5>
3296<pre>
3297 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3298</pre>
3299
3300<h5>Overview:</h5>
3301<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3302the integer type <tt>ty2</tt>.</p>
3303
3304<h5>Arguments:</h5>
3305<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3306must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3307<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3308
3309<h5>Semantics:</h5>
3310<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3311<tt>ty2</tt> by interpreting the pointer value as an integer and either
3312truncating or zero extending that value to the size of the integer type. If
3313<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3314<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3315are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3316change.</p>
3317
3318<h5>Example:</h5>
3319<pre>
3320 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3321 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3322</pre>
3323</div>
3324
3325<!-- _______________________________________________________________________ -->
3326<div class="doc_subsubsection">
3327 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3328</div>
3329<div class="doc_text">
3330
3331<h5>Syntax:</h5>
3332<pre>
3333 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3334</pre>
3335
3336<h5>Overview:</h5>
3337<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3338a pointer type, <tt>ty2</tt>.</p>
3339
3340<h5>Arguments:</h5>
3341<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3342value to cast, and a type to cast it to, which must be a
3343<a href="#t_pointer">pointer</a> type.
3344
3345<h5>Semantics:</h5>
3346<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3347<tt>ty2</tt> by applying either a zero extension or a truncation depending on
3348the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3349size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3350the size of a pointer then a zero extension is done. If they are the same size,
3351nothing is done (<i>no-op cast</i>).</p>
3352
3353<h5>Example:</h5>
3354<pre>
3355 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3356 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3357 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3358</pre>
3359</div>
3360
3361<!-- _______________________________________________________________________ -->
3362<div class="doc_subsubsection">
3363 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3364</div>
3365<div class="doc_text">
3366
3367<h5>Syntax:</h5>
3368<pre>
3369 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3370</pre>
3371
3372<h5>Overview:</h5>
3373<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3374<tt>ty2</tt> without changing any bits.</p>
3375
3376<h5>Arguments:</h5>
3377<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3378a first class value, and a type to cast it to, which must also be a <a
3379 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3380and the destination type, <tt>ty2</tt>, must be identical. If the source
3381type is a pointer, the destination type must also be a pointer.</p>
3382
3383<h5>Semantics:</h5>
3384<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3385<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3386this conversion. The conversion is done as if the <tt>value</tt> had been
3387stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3388converted to other pointer types with this instruction. To convert pointers to
3389other types, use the <a href="#i_inttoptr">inttoptr</a> or
3390<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3391
3392<h5>Example:</h5>
3393<pre>
3394 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3395 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3396 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3397</pre>
3398</div>
3399
3400<!-- ======================================================================= -->
3401<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3402<div class="doc_text">
3403<p>The instructions in this category are the "miscellaneous"
3404instructions, which defy better classification.</p>
3405</div>
3406
3407<!-- _______________________________________________________________________ -->
3408<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3409</div>
3410<div class="doc_text">
3411<h5>Syntax:</h5>
3412<pre> &lt;result&gt; = icmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3413</pre>
3414<h5>Overview:</h5>
3415<p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3416of its two integer operands.</p>
3417<h5>Arguments:</h5>
3418<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3419the condition code indicating the kind of comparison to perform. It is not
3420a value, just a keyword. The possible condition code are:
3421<ol>
3422 <li><tt>eq</tt>: equal</li>
3423 <li><tt>ne</tt>: not equal </li>
3424 <li><tt>ugt</tt>: unsigned greater than</li>
3425 <li><tt>uge</tt>: unsigned greater or equal</li>
3426 <li><tt>ult</tt>: unsigned less than</li>
3427 <li><tt>ule</tt>: unsigned less or equal</li>
3428 <li><tt>sgt</tt>: signed greater than</li>
3429 <li><tt>sge</tt>: signed greater or equal</li>
3430 <li><tt>slt</tt>: signed less than</li>
3431 <li><tt>sle</tt>: signed less or equal</li>
3432</ol>
3433<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3434<a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3435<h5>Semantics:</h5>
3436<p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3437the condition code given as <tt>cond</tt>. The comparison performed always
3438yields a <a href="#t_primitive">i1</a> result, as follows:
3439<ol>
3440 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3441 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3442 </li>
3443 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3444 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3445 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3446 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3447 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3448 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3449 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3450 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3451 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3452 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3453 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3454 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3455 <li><tt>sge</tt>: interprets the operands as signed values and yields
3456 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3457 <li><tt>slt</tt>: interprets the operands as signed values and yields
3458 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3459 <li><tt>sle</tt>: interprets the operands as signed values and yields
3460 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3461</ol>
3462<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3463values are compared as if they were integers.</p>
3464
3465<h5>Example:</h5>
3466<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
3467 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
3468 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
3469 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3470 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3471 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
3472</pre>
3473</div>
3474
3475<!-- _______________________________________________________________________ -->
3476<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3477</div>
3478<div class="doc_text">
3479<h5>Syntax:</h5>
3480<pre> &lt;result&gt; = fcmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3481</pre>
3482<h5>Overview:</h5>
3483<p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3484of its floating point operands.</p>
3485<h5>Arguments:</h5>
3486<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3487the condition code indicating the kind of comparison to perform. It is not
3488a value, just a keyword. The possible condition code are:
3489<ol>
3490 <li><tt>false</tt>: no comparison, always returns false</li>
3491 <li><tt>oeq</tt>: ordered and equal</li>
3492 <li><tt>ogt</tt>: ordered and greater than </li>
3493 <li><tt>oge</tt>: ordered and greater than or equal</li>
3494 <li><tt>olt</tt>: ordered and less than </li>
3495 <li><tt>ole</tt>: ordered and less than or equal</li>
3496 <li><tt>one</tt>: ordered and not equal</li>
3497 <li><tt>ord</tt>: ordered (no nans)</li>
3498 <li><tt>ueq</tt>: unordered or equal</li>
3499 <li><tt>ugt</tt>: unordered or greater than </li>
3500 <li><tt>uge</tt>: unordered or greater than or equal</li>
3501 <li><tt>ult</tt>: unordered or less than </li>
3502 <li><tt>ule</tt>: unordered or less than or equal</li>
3503 <li><tt>une</tt>: unordered or not equal</li>
3504 <li><tt>uno</tt>: unordered (either nans)</li>
3505 <li><tt>true</tt>: no comparison, always returns true</li>
3506</ol>
3507<p><i>Ordered</i> means that neither operand is a QNAN while
3508<i>unordered</i> means that either operand may be a QNAN.</p>
3509<p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3510<a href="#t_floating">floating point</a> typed. They must have identical
3511types.</p>
3512<h5>Semantics:</h5>
3513<p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3514the condition code given as <tt>cond</tt>. The comparison performed always
3515yields a <a href="#t_primitive">i1</a> result, as follows:
3516<ol>
3517 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3518 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3519 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3520 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3521 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3522 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3523 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3524 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3525 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3526 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3527 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3528 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3529 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3530 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3531 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3532 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3533 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3534 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3535 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3536 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3537 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3538 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3539 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3540 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3541 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3542 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3543 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3544 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3545</ol>
3546
3547<h5>Example:</h5>
3548<pre> &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3549 &lt;result&gt; = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3550 &lt;result&gt; = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3551 &lt;result&gt; = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3552</pre>
3553</div>
3554
3555<!-- _______________________________________________________________________ -->
3556<div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3557Instruction</a> </div>
3558<div class="doc_text">
3559<h5>Syntax:</h5>
3560<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
3561<h5>Overview:</h5>
3562<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
3563the SSA graph representing the function.</p>
3564<h5>Arguments:</h5>
3565<p>The type of the incoming values is specified with the first type
3566field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3567as arguments, with one pair for each predecessor basic block of the
3568current block. Only values of <a href="#t_firstclass">first class</a>
3569type may be used as the value arguments to the PHI node. Only labels
3570may be used as the label arguments.</p>
3571<p>There must be no non-phi instructions between the start of a basic
3572block and the PHI instructions: i.e. PHI instructions must be first in
3573a basic block.</p>
3574<h5>Semantics:</h5>
3575<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3576specified by the pair corresponding to the predecessor basic block that executed
3577just prior to the current block.</p>
3578<h5>Example:</h5>
3579<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>
3580</div>
3581
3582<!-- _______________________________________________________________________ -->
3583<div class="doc_subsubsection">
3584 <a name="i_select">'<tt>select</tt>' Instruction</a>
3585</div>
3586
3587<div class="doc_text">
3588
3589<h5>Syntax:</h5>
3590
3591<pre>
3592 &lt;result&gt; = select i1 &lt;cond&gt;, &lt;ty&gt; &lt;val1&gt;, &lt;ty&gt; &lt;val2&gt; <i>; yields ty</i>
3593</pre>
3594
3595<h5>Overview:</h5>
3596
3597<p>
3598The '<tt>select</tt>' instruction is used to choose one value based on a
3599condition, without branching.
3600</p>
3601
3602
3603<h5>Arguments:</h5>
3604
3605<p>
3606The '<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.
3607</p>
3608
3609<h5>Semantics:</h5>
3610
3611<p>
3612If the boolean condition evaluates to true, the instruction returns the first
3613value argument; otherwise, it returns the second value argument.
3614</p>
3615
3616<h5>Example:</h5>
3617
3618<pre>
3619 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3620</pre>
3621</div>
3622
3623
3624<!-- _______________________________________________________________________ -->
3625<div class="doc_subsubsection">
3626 <a name="i_call">'<tt>call</tt>' Instruction</a>
3627</div>
3628
3629<div class="doc_text">
3630
3631<h5>Syntax:</h5>
3632<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3633 &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]3634</pre>
3635
3636<h5>Overview:</h5>
3637
3638<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3639
3640<h5>Arguments:</h5>
3641
3642<p>This instruction requires several arguments:</p>
3643
3644<ol>
3645 <li>
3646 <p>The optional "tail" marker indicates whether the callee function accesses
3647 any allocas or varargs in the caller. If the "tail" marker is present, the
3648 function call is eligible for tail call optimization. Note that calls may
3649 be marked "tail" even if they do not occur before a <a
3650 href="#i_ret"><tt>ret</tt></a> instruction.
3651 </li>
3652 <li>
3653 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3654 convention</a> the call should use. If none is specified, the call defaults
3655 to using C calling conventions.
3656 </li>
3657 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3658 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
3659 the type of the return value. Functions that return no value are marked
3660 <tt><a href="#t_void">void</a></tt>.</p>
3661 </li>
3662 <li>
3663 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
3664 value being invoked. The argument types must match the types implied by
3665 this signature. This type can be omitted if the function is not varargs
3666 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3667 </li>
3668 <li>
3669 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3670 be invoked. In most cases, this is a direct function invocation, but
3671 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3672 to function value.</p>
3673 </li>
3674 <li>
3675 <p>'<tt>function args</tt>': argument list whose types match the
3676 function signature argument types. All arguments must be of
3677 <a href="#t_firstclass">first class</a> type. If the function signature
3678 indicates the function accepts a variable number of arguments, the extra
3679 arguments can be specified.</p>
3680 </li>
3681</ol>
3682
3683<h5>Semantics:</h5>
3684
3685<p>The '<tt>call</tt>' instruction is used to cause control flow to
3686transfer to a specified function, with its incoming arguments bound to
3687the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3688instruction in the called function, control flow continues with the
3689instruction after the function call, and the return value of the
3690function is bound to the result argument. This is a simpler case of
3691the <a href="#i_invoke">invoke</a> instruction.</p>
3692
3693<h5>Example:</h5>
3694
3695<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3696 %retval = call i32 @test(i32 %argc)
3697 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42);
3698 %X = tail call i32 @foo()
3699 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo()
3700 %Z = call void %foo(i8 97 signext)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3701</pre>
3702
3703</div>
3704
3705<!-- _______________________________________________________________________ -->
3706<div class="doc_subsubsection">
3707 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3708</div>
3709
3710<div class="doc_text">
3711
3712<h5>Syntax:</h5>
3713
3714<pre>
3715 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
3716</pre>
3717
3718<h5>Overview:</h5>
3719
3720<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3721the "variable argument" area of a function call. It is used to implement the
3722<tt>va_arg</tt> macro in C.</p>
3723
3724<h5>Arguments:</h5>
3725
3726<p>This instruction takes a <tt>va_list*</tt> value and the type of
3727the argument. It returns a value of the specified argument type and
3728increments the <tt>va_list</tt> to point to the next argument. The
3729actual type of <tt>va_list</tt> is target specific.</p>
3730
3731<h5>Semantics:</h5>
3732
3733<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3734type from the specified <tt>va_list</tt> and causes the
3735<tt>va_list</tt> to point to the next argument. For more information,
3736see the variable argument handling <a href="#int_varargs">Intrinsic
3737Functions</a>.</p>
3738
3739<p>It is legal for this instruction to be called in a function which does not
3740take a variable number of arguments, for example, the <tt>vfprintf</tt>
3741function.</p>
3742
3743<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3744href="#intrinsics">intrinsic function</a> because it takes a type as an
3745argument.</p>
3746
3747<h5>Example:</h5>
3748
3749<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3750
3751</div>
3752
3753<!-- *********************************************************************** -->
3754<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3755<!-- *********************************************************************** -->
3756
3757<div class="doc_text">
3758
3759<p>LLVM supports the notion of an "intrinsic function". These functions have
3760well known names and semantics and are required to follow certain restrictions.
3761Overall, these intrinsics represent an extension mechanism for the LLVM
3762language that does not require changing all of the transformations in LLVM when
3763adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
3764
3765<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3766prefix is reserved in LLVM for intrinsic names; thus, function names may not
3767begin with this prefix. Intrinsic functions must always be external functions:
3768you cannot define the body of intrinsic functions. Intrinsic functions may
3769only be used in call or invoke instructions: it is illegal to take the address
3770of an intrinsic function. Additionally, because intrinsic functions are part
3771of the LLVM language, it is required if any are added that they be documented
3772here.</p>
3773
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]3774<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
3775a family of functions that perform the same operation but on different data
3776types. Because LLVM can represent over 8 million different integer types,
3777overloading is used commonly to allow an intrinsic function to operate on any
3778integer type. One or more of the argument types or the result type can be
3779overloaded to accept any integer type. Argument types may also be defined as
3780exactly matching a previous argument's type or the result type. This allows an
3781intrinsic function which accepts multiple arguments, but needs all of them to
3782be of the same type, to only be overloaded with respect to a single argument or
3783the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3784
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]3785<p>Overloaded intrinsics will have the names of its overloaded argument types
3786encoded into its function name, each preceded by a period. Only those types
3787which are overloaded result in a name suffix. Arguments whose type is matched
3788against another type do not. For example, the <tt>llvm.ctpop</tt> function can
3789take an integer of any width and returns an integer of exactly the same integer
3790width. This leads to a family of functions such as
3791<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
3792Only one type, the return type, is overloaded, and only one type suffix is
3793required. Because the argument's type is matched against the return type, it
3794does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3795
3796<p>To learn how to add an intrinsic function, please see the
3797<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3798</p>
3799
3800</div>
3801
3802<!-- ======================================================================= -->
3803<div class="doc_subsection">
3804 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3805</div>
3806
3807<div class="doc_text">
3808
3809<p>Variable argument support is defined in LLVM with the <a
3810 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3811intrinsic functions. These functions are related to the similarly
3812named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
3813
3814<p>All of these functions operate on arguments that use a
3815target-specific value type "<tt>va_list</tt>". The LLVM assembly
3816language reference manual does not define what this type is, so all
3817transformations should be prepared to handle these functions regardless of
3818the type used.</p>
3819
3820<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3821instruction and the variable argument handling intrinsic functions are
3822used.</p>
3823
3824<div class="doc_code">
3825<pre>
3826define i32 @test(i32 %X, ...) {
3827 ; Initialize variable argument processing
3828 %ap = alloca i8*
3829 %ap2 = bitcast i8** %ap to i8*
3830 call void @llvm.va_start(i8* %ap2)
3831
3832 ; Read a single integer argument
3833 %tmp = va_arg i8** %ap, i32
3834
3835 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3836 %aq = alloca i8*
3837 %aq2 = bitcast i8** %aq to i8*
3838 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
3839 call void @llvm.va_end(i8* %aq2)
3840
3841 ; Stop processing of arguments.
3842 call void @llvm.va_end(i8* %ap2)
3843 ret i32 %tmp
3844}
3845
3846declare void @llvm.va_start(i8*)
3847declare void @llvm.va_copy(i8*, i8*)
3848declare void @llvm.va_end(i8*)
3849</pre>
3850</div>
3851
3852</div>
3853
3854<!-- _______________________________________________________________________ -->
3855<div class="doc_subsubsection">
3856 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3857</div>
3858
3859
3860<div class="doc_text">
3861<h5>Syntax:</h5>
3862<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
3863<h5>Overview:</h5>
3864<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3865<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
3866href="#i_va_arg">va_arg</a></tt>.</p>
3867
3868<h5>Arguments:</h5>
3869
3870<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3871
3872<h5>Semantics:</h5>
3873
3874<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3875macro available in C. In a target-dependent way, it initializes the
3876<tt>va_list</tt> element to which the argument points, so that the next call to
3877<tt>va_arg</tt> will produce the first variable argument passed to the function.
3878Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3879last argument of the function as the compiler can figure that out.</p>
3880
3881</div>
3882
3883<!-- _______________________________________________________________________ -->
3884<div class="doc_subsubsection">
3885 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3886</div>
3887
3888<div class="doc_text">
3889<h5>Syntax:</h5>
3890<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
3891<h5>Overview:</h5>
3892
3893<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
3894which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
3895or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3896
3897<h5>Arguments:</h5>
3898
3899<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
3900
3901<h5>Semantics:</h5>
3902
3903<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3904macro available in C. In a target-dependent way, it destroys the
3905<tt>va_list</tt> element to which the argument points. Calls to <a
3906href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
3907<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
3908<tt>llvm.va_end</tt>.</p>
3909
3910</div>
3911
3912<!-- _______________________________________________________________________ -->
3913<div class="doc_subsubsection">
3914 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3915</div>
3916
3917<div class="doc_text">
3918
3919<h5>Syntax:</h5>
3920
3921<pre>
3922 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
3923</pre>
3924
3925<h5>Overview:</h5>
3926
3927<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
3928from the source argument list to the destination argument list.</p>
3929
3930<h5>Arguments:</h5>
3931
3932<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3933The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3934
3935
3936<h5>Semantics:</h5>
3937
3938<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
3939macro available in C. In a target-dependent way, it copies the source
3940<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
3941intrinsic is necessary because the <tt><a href="#int_va_start">
3942llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
3943example, memory allocation.</p>
3944
3945</div>
3946
3947<!-- ======================================================================= -->
3948<div class="doc_subsection">
3949 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3950</div>
3951
3952<div class="doc_text">
3953
3954<p>
3955LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3956Collection</a> requires the implementation and generation of these intrinsics.
3957These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
3958stack</a>, as well as garbage collector implementations that require <a
3959href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
3960Front-ends for type-safe garbage collected languages should generate these
3961intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3962href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3963</p>
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3964
3965<p>The garbage collection intrinsics only operate on objects in the generic
3966 address space (address space zero).</p>
3967
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3968</div>
3969
3970<!-- _______________________________________________________________________ -->
3971<div class="doc_subsubsection">
3972 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3973</div>
3974
3975<div class="doc_text">
3976
3977<h5>Syntax:</h5>
3978
3979<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]3980 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3981</pre>
3982
3983<h5>Overview:</h5>
3984
3985<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3986the code generator, and allows some metadata to be associated with it.</p>
3987
3988<h5>Arguments:</h5>
3989
3990<p>The first argument specifies the address of a stack object that contains the
3991root pointer. The second pointer (which must be either a constant or a global
3992value address) contains the meta-data to be associated with the root.</p>
3993
3994<h5>Semantics:</h5>
3995
3996<p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3997location. At compile-time, the code generator generates information to allow
3998the runtime to find the pointer at GC safe points.
3999</p>
4000
4001</div>
4002
4003
4004<!-- _______________________________________________________________________ -->
4005<div class="doc_subsubsection">
4006 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4007</div>
4008
4009<div class="doc_text">
4010
4011<h5>Syntax:</h5>
4012
4013<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4014 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4015</pre>
4016
4017<h5>Overview:</h5>
4018
4019<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4020locations, allowing garbage collector implementations that require read
4021barriers.</p>
4022
4023<h5>Arguments:</h5>
4024
4025<p>The second argument is the address to read from, which should be an address
4026allocated from the garbage collector. The first object is a pointer to the
4027start of the referenced object, if needed by the language runtime (otherwise
4028null).</p>
4029
4030<h5>Semantics:</h5>
4031
4032<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4033instruction, but may be replaced with substantially more complex code by the
4034garbage collector runtime, as needed.</p>
4035
4036</div>
4037
4038
4039<!-- _______________________________________________________________________ -->
4040<div class="doc_subsubsection">
4041 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4042</div>
4043
4044<div class="doc_text">
4045
4046<h5>Syntax:</h5>
4047
4048<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4049 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4050</pre>
4051
4052<h5>Overview:</h5>
4053
4054<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4055locations, allowing garbage collector implementations that require write
4056barriers (such as generational or reference counting collectors).</p>
4057
4058<h5>Arguments:</h5>
4059
4060<p>The first argument is the reference to store, the second is the start of the
4061object to store it to, and the third is the address of the field of Obj to
4062store to. If the runtime does not require a pointer to the object, Obj may be
4063null.</p>
4064
4065<h5>Semantics:</h5>
4066
4067<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4068instruction, but may be replaced with substantially more complex code by the
4069garbage collector runtime, as needed.</p>
4070
4071</div>
4072
4073
4074
4075<!-- ======================================================================= -->
4076<div class="doc_subsection">
4077 <a name="int_codegen">Code Generator Intrinsics</a>
4078</div>
4079
4080<div class="doc_text">
4081<p>
4082These intrinsics are provided by LLVM to expose special features that may only
4083be implemented with code generator support.
4084</p>
4085
4086</div>
4087
4088<!-- _______________________________________________________________________ -->
4089<div class="doc_subsubsection">
4090 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4091</div>
4092
4093<div class="doc_text">
4094
4095<h5>Syntax:</h5>
4096<pre>
4097 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
4098</pre>
4099
4100<h5>Overview:</h5>
4101
4102<p>
4103The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4104target-specific value indicating the return address of the current function
4105or one of its callers.
4106</p>
4107
4108<h5>Arguments:</h5>
4109
4110<p>
4111The argument to this intrinsic indicates which function to return the address
4112for. Zero indicates the calling function, one indicates its caller, etc. The
4113argument is <b>required</b> to be a constant integer value.
4114</p>
4115
4116<h5>Semantics:</h5>
4117
4118<p>
4119The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4120the return address of the specified call frame, or zero if it cannot be
4121identified. The value returned by this intrinsic is likely to be incorrect or 0
4122for arguments other than zero, so it should only be used for debugging purposes.
4123</p>
4124
4125<p>
4126Note that calling this intrinsic does not prevent function inlining or other
4127aggressive transformations, so the value returned may not be that of the obvious
4128source-language caller.
4129</p>
4130</div>
4131
4132
4133<!-- _______________________________________________________________________ -->
4134<div class="doc_subsubsection">
4135 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4136</div>
4137
4138<div class="doc_text">
4139
4140<h5>Syntax:</h5>
4141<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4142 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4143</pre>
4144
4145<h5>Overview:</h5>
4146
4147<p>
4148The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4149target-specific frame pointer value for the specified stack frame.
4150</p>
4151
4152<h5>Arguments:</h5>
4153
4154<p>
4155The argument to this intrinsic indicates which function to return the frame
4156pointer for. Zero indicates the calling function, one indicates its caller,
4157etc. The argument is <b>required</b> to be a constant integer value.
4158</p>
4159
4160<h5>Semantics:</h5>
4161
4162<p>
4163The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4164the frame address of the specified call frame, or zero if it cannot be
4165identified. The value returned by this intrinsic is likely to be incorrect or 0
4166for arguments other than zero, so it should only be used for debugging purposes.
4167</p>
4168
4169<p>
4170Note that calling this intrinsic does not prevent function inlining or other
4171aggressive transformations, so the value returned may not be that of the obvious
4172source-language caller.
4173</p>
4174</div>
4175
4176<!-- _______________________________________________________________________ -->
4177<div class="doc_subsubsection">
4178 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4179</div>
4180
4181<div class="doc_text">
4182
4183<h5>Syntax:</h5>
4184<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4185 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4186</pre>
4187
4188<h5>Overview:</h5>
4189
4190<p>
4191The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4192the function stack, for use with <a href="#int_stackrestore">
4193<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4194features like scoped automatic variable sized arrays in C99.
4195</p>
4196
4197<h5>Semantics:</h5>
4198
4199<p>
4200This intrinsic returns a opaque pointer value that can be passed to <a
4201href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4202<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4203<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4204state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4205practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4206that were allocated after the <tt>llvm.stacksave</tt> was executed.
4207</p>
4208
4209</div>
4210
4211<!-- _______________________________________________________________________ -->
4212<div class="doc_subsubsection">
4213 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4214</div>
4215
4216<div class="doc_text">
4217
4218<h5>Syntax:</h5>
4219<pre>
4220 declare void @llvm.stackrestore(i8 * %ptr)
4221</pre>
4222
4223<h5>Overview:</h5>
4224
4225<p>
4226The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4227the function stack to the state it was in when the corresponding <a
4228href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4229useful for implementing language features like scoped automatic variable sized
4230arrays in C99.
4231</p>
4232
4233<h5>Semantics:</h5>
4234
4235<p>
4236See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4237</p>
4238
4239</div>
4240
4241
4242<!-- _______________________________________________________________________ -->
4243<div class="doc_subsubsection">
4244 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4245</div>
4246
4247<div class="doc_text">
4248
4249<h5>Syntax:</h5>
4250<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4251 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]4252</pre>
4253
4254<h5>Overview:</h5>
4255
4256
4257<p>
4258The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4259a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4260no
4261effect on the behavior of the program but can change its performance
4262characteristics.
4263</p>
4264
4265<h5>Arguments:</h5>
4266
4267<p>
4268<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4269determining if the fetch should be for a read (0) or write (1), and
4270<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4271locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4272<tt>locality</tt> arguments must be constant integers.
4273</p>
4274
4275<h5>Semantics:</h5>
4276
4277<p>
4278This intrinsic does not modify the behavior of the program. In particular,
4279prefetches cannot trap and do not produce a value. On targets that support this
4280intrinsic, the prefetch can provide hints to the processor cache for better
4281performance.
4282</p>
4283
4284</div>
4285
4286<!-- _______________________________________________________________________ -->
4287<div class="doc_subsubsection">
4288 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4289</div>
4290
4291<div class="doc_text">
4292
4293<h5>Syntax:</h5>
4294<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4295 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4296</pre>
4297
4298<h5>Overview:</h5>
4299
4300
4301<p>
4302The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4303(PC) in a region of
4304code to simulators and other tools. The method is target specific, but it is
4305expected that the marker will use exported symbols to transmit the PC of the marker.
4306The marker makes no guarantees that it will remain with any specific instruction
4307after optimizations. It is possible that the presence of a marker will inhibit
4308optimizations. The intended use is to be inserted after optimizations to allow
4309correlations of simulation runs.
4310</p>
4311
4312<h5>Arguments:</h5>
4313
4314<p>
4315<tt>id</tt> is a numerical id identifying the marker.
4316</p>
4317
4318<h5>Semantics:</h5>
4319
4320<p>
4321This intrinsic does not modify the behavior of the program. Backends that do not
4322support this intrinisic may ignore it.
4323</p>
4324
4325</div>
4326
4327<!-- _______________________________________________________________________ -->
4328<div class="doc_subsubsection">
4329 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4330</div>
4331
4332<div class="doc_text">
4333
4334<h5>Syntax:</h5>
4335<pre>
4336 declare i64 @llvm.readcyclecounter( )
4337</pre>
4338
4339<h5>Overview:</h5>
4340
4341
4342<p>
4343The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4344counter register (or similar low latency, high accuracy clocks) on those targets
4345that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4346As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4347should only be used for small timings.
4348</p>
4349
4350<h5>Semantics:</h5>
4351
4352<p>
4353When directly supported, reading the cycle counter should not modify any memory.
4354Implementations are allowed to either return a application specific value or a
4355system wide value. On backends without support, this is lowered to a constant 0.
4356</p>
4357
4358</div>
4359
4360<!-- ======================================================================= -->
4361<div class="doc_subsection">
4362 <a name="int_libc">Standard C Library Intrinsics</a>
4363</div>
4364
4365<div class="doc_text">
4366<p>
4367LLVM provides intrinsics for a few important standard C library functions.
4368These intrinsics allow source-language front-ends to pass information about the
4369alignment of the pointer arguments to the code generator, providing opportunity
4370for more efficient code generation.
4371</p>
4372
4373</div>
4374
4375<!-- _______________________________________________________________________ -->
4376<div class="doc_subsubsection">
4377 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4378</div>
4379
4380<div class="doc_text">
4381
4382<h5>Syntax:</h5>
4383<pre>
4384 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4385 i32 &lt;len&gt;, i32 &lt;align&gt;)
4386 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4387 i64 &lt;len&gt;, i32 &lt;align&gt;)
4388</pre>
4389
4390<h5>Overview:</h5>
4391
4392<p>
4393The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4394location to the destination location.
4395</p>
4396
4397<p>
4398Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4399intrinsics do not return a value, and takes an extra alignment argument.
4400</p>
4401
4402<h5>Arguments:</h5>
4403
4404<p>
4405The first argument is a pointer to the destination, the second is a pointer to
4406the source. The third argument is an integer argument
4407specifying the number of bytes to copy, and the fourth argument is the alignment
4408of the source and destination locations.
4409</p>
4410
4411<p>
4412If the call to this intrinisic has an alignment value that is not 0 or 1, then
4413the caller guarantees that both the source and destination pointers are aligned
4414to that boundary.
4415</p>
4416
4417<h5>Semantics:</h5>
4418
4419<p>
4420The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4421location to the destination location, which are not allowed to overlap. It
4422copies "len" bytes of memory over. If the argument is known to be aligned to
4423some boundary, this can be specified as the fourth argument, otherwise it should
4424be set to 0 or 1.
4425</p>
4426</div>
4427
4428
4429<!-- _______________________________________________________________________ -->
4430<div class="doc_subsubsection">
4431 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4432</div>
4433
4434<div class="doc_text">
4435
4436<h5>Syntax:</h5>
4437<pre>
4438 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4439 i32 &lt;len&gt;, i32 &lt;align&gt;)
4440 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4441 i64 &lt;len&gt;, i32 &lt;align&gt;)
4442</pre>
4443
4444<h5>Overview:</h5>
4445
4446<p>
4447The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4448location to the destination location. It is similar to the
4449'<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4450</p>
4451
4452<p>
4453Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4454intrinsics do not return a value, and takes an extra alignment argument.
4455</p>
4456
4457<h5>Arguments:</h5>
4458
4459<p>
4460The first argument is a pointer to the destination, the second is a pointer to
4461the source. The third argument is an integer argument
4462specifying the number of bytes to copy, and the fourth argument is the alignment
4463of the source and destination locations.
4464</p>
4465
4466<p>
4467If the call to this intrinisic has an alignment value that is not 0 or 1, then
4468the caller guarantees that the source and destination pointers are aligned to
4469that boundary.
4470</p>
4471
4472<h5>Semantics:</h5>
4473
4474<p>
4475The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4476location to the destination location, which may overlap. It
4477copies "len" bytes of memory over. If the argument is known to be aligned to
4478some boundary, this can be specified as the fourth argument, otherwise it should
4479be set to 0 or 1.
4480</p>
4481</div>
4482
4483
4484<!-- _______________________________________________________________________ -->
4485<div class="doc_subsubsection">
4486 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4487</div>
4488
4489<div class="doc_text">
4490
4491<h5>Syntax:</h5>
4492<pre>
4493 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
4494 i32 &lt;len&gt;, i32 &lt;align&gt;)
4495 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
4496 i64 &lt;len&gt;, i32 &lt;align&gt;)
4497</pre>
4498
4499<h5>Overview:</h5>
4500
4501<p>
4502The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4503byte value.
4504</p>
4505
4506<p>
4507Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4508does not return a value, and takes an extra alignment argument.
4509</p>
4510
4511<h5>Arguments:</h5>
4512
4513<p>
4514The first argument is a pointer to the destination to fill, the second is the
4515byte value to fill it with, the third argument is an integer
4516argument specifying the number of bytes to fill, and the fourth argument is the
4517known alignment of destination location.
4518</p>
4519
4520<p>
4521If the call to this intrinisic has an alignment value that is not 0 or 1, then
4522the caller guarantees that the destination pointer is aligned to that boundary.
4523</p>
4524
4525<h5>Semantics:</h5>
4526
4527<p>
4528The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4529the
4530destination location. If the argument is known to be aligned to some boundary,
4531this can be specified as the fourth argument, otherwise it should be set to 0 or
45321.
4533</p>
4534</div>
4535
4536
4537<!-- _______________________________________________________________________ -->
4538<div class="doc_subsubsection">
4539 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4540</div>
4541
4542<div class="doc_text">
4543
4544<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4545<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]4546floating point or vector of floating point type. Not all targets support all
4547types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4548<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4549 declare float @llvm.sqrt.f32(float %Val)
4550 declare double @llvm.sqrt.f64(double %Val)
4551 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
4552 declare fp128 @llvm.sqrt.f128(fp128 %Val)
4553 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4554</pre>
4555
4556<h5>Overview:</h5>
4557
4558<p>
4559The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4560returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4561<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4562negative numbers (which allows for better optimization).
4563</p>
4564
4565<h5>Arguments:</h5>
4566
4567<p>
4568The argument and return value are floating point numbers of the same type.
4569</p>
4570
4571<h5>Semantics:</h5>
4572
4573<p>
4574This function returns the sqrt of the specified operand if it is a nonnegative
4575floating point number.
4576</p>
4577</div>
4578
4579<!-- _______________________________________________________________________ -->
4580<div class="doc_subsubsection">
4581 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4582</div>
4583
4584<div class="doc_text">
4585
4586<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4587<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]4588floating point or vector of floating point type. Not all targets support all
4589types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4590<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4591 declare float @llvm.powi.f32(float %Val, i32 %power)
4592 declare double @llvm.powi.f64(double %Val, i32 %power)
4593 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
4594 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
4595 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4596</pre>
4597
4598<h5>Overview:</h5>
4599
4600<p>
4601The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4602specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4603multiplications is not defined. When a vector of floating point type is
4604used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4605</p>
4606
4607<h5>Arguments:</h5>
4608
4609<p>
4610The second argument is an integer power, and the first is a value to raise to
4611that power.
4612</p>
4613
4614<h5>Semantics:</h5>
4615
4616<p>
4617This function returns the first value raised to the second power with an
4618unspecified sequence of rounding operations.</p>
4619</div>
4620
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4621<!-- _______________________________________________________________________ -->
4622<div class="doc_subsubsection">
4623 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
4624</div>
4625
4626<div class="doc_text">
4627
4628<h5>Syntax:</h5>
4629<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
4630floating point or vector of floating point type. Not all targets support all
4631types however.
4632<pre>
4633 declare float @llvm.sin.f32(float %Val)
4634 declare double @llvm.sin.f64(double %Val)
4635 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
4636 declare fp128 @llvm.sin.f128(fp128 %Val)
4637 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
4638</pre>
4639
4640<h5>Overview:</h5>
4641
4642<p>
4643The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
4644</p>
4645
4646<h5>Arguments:</h5>
4647
4648<p>
4649The argument and return value are floating point numbers of the same type.
4650</p>
4651
4652<h5>Semantics:</h5>
4653
4654<p>
4655This function returns the sine of the specified operand, returning the
4656same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4657conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4658</div>
4659
4660<!-- _______________________________________________________________________ -->
4661<div class="doc_subsubsection">
4662 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
4663</div>
4664
4665<div class="doc_text">
4666
4667<h5>Syntax:</h5>
4668<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
4669floating point or vector of floating point type. Not all targets support all
4670types however.
4671<pre>
4672 declare float @llvm.cos.f32(float %Val)
4673 declare double @llvm.cos.f64(double %Val)
4674 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
4675 declare fp128 @llvm.cos.f128(fp128 %Val)
4676 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
4677</pre>
4678
4679<h5>Overview:</h5>
4680
4681<p>
4682The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
4683</p>
4684
4685<h5>Arguments:</h5>
4686
4687<p>
4688The argument and return value are floating point numbers of the same type.
4689</p>
4690
4691<h5>Semantics:</h5>
4692
4693<p>
4694This function returns the cosine of the specified operand, returning the
4695same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4696conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4697</div>
4698
4699<!-- _______________________________________________________________________ -->
4700<div class="doc_subsubsection">
4701 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
4702</div>
4703
4704<div class="doc_text">
4705
4706<h5>Syntax:</h5>
4707<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
4708floating point or vector of floating point type. Not all targets support all
4709types however.
4710<pre>
4711 declare float @llvm.pow.f32(float %Val, float %Power)
4712 declare double @llvm.pow.f64(double %Val, double %Power)
4713 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
4714 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
4715 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
4716</pre>
4717
4718<h5>Overview:</h5>
4719
4720<p>
4721The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
4722specified (positive or negative) power.
4723</p>
4724
4725<h5>Arguments:</h5>
4726
4727<p>
4728The second argument is a floating point power, and the first is a value to
4729raise to that power.
4730</p>
4731
4732<h5>Semantics:</h5>
4733
4734<p>
4735This function returns the first value raised to the second power,
4736returning the
4737same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4738conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4739</div>
4740
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4741
4742<!-- ======================================================================= -->
4743<div class="doc_subsection">
4744 <a name="int_manip">Bit Manipulation Intrinsics</a>
4745</div>
4746
4747<div class="doc_text">
4748<p>
4749LLVM provides intrinsics for a few important bit manipulation operations.
4750These allow efficient code generation for some algorithms.
4751</p>
4752
4753</div>
4754
4755<!-- _______________________________________________________________________ -->
4756<div class="doc_subsubsection">
4757 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4758</div>
4759
4760<div class="doc_text">
4761
4762<h5>Syntax:</h5>
4763<p>This is an overloaded intrinsic function. You can use bswap on any integer
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4764type that is an even number of bytes (i.e. BitWidth % 16 == 0).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4765<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4766 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
4767 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
4768 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4769</pre>
4770
4771<h5>Overview:</h5>
4772
4773<p>
4774The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
4775values with an even number of bytes (positive multiple of 16 bits). These are
4776useful for performing operations on data that is not in the target's native
4777byte order.
4778</p>
4779
4780<h5>Semantics:</h5>
4781
4782<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4783The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4784and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4785intrinsic returns an i32 value that has the four bytes of the input i32
4786swapped, 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]4787i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
4788<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4789additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4790</p>
4791
4792</div>
4793
4794<!-- _______________________________________________________________________ -->
4795<div class="doc_subsubsection">
4796 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4797</div>
4798
4799<div class="doc_text">
4800
4801<h5>Syntax:</h5>
4802<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4803width. Not all targets support all bit widths however.
4804<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4805 declare i8 @llvm.ctpop.i8 (i8 &lt;src&gt;)
4806 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4807 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4808 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
4809 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4810</pre>
4811
4812<h5>Overview:</h5>
4813
4814<p>
4815The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4816value.
4817</p>
4818
4819<h5>Arguments:</h5>
4820
4821<p>
4822The only argument is the value to be counted. The argument may be of any
4823integer type. The return type must match the argument type.
4824</p>
4825
4826<h5>Semantics:</h5>
4827
4828<p>
4829The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4830</p>
4831</div>
4832
4833<!-- _______________________________________________________________________ -->
4834<div class="doc_subsubsection">
4835 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4836</div>
4837
4838<div class="doc_text">
4839
4840<h5>Syntax:</h5>
4841<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4842integer bit width. Not all targets support all bit widths however.
4843<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4844 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
4845 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4846 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4847 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
4848 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4849</pre>
4850
4851<h5>Overview:</h5>
4852
4853<p>
4854The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4855leading zeros in a variable.
4856</p>
4857
4858<h5>Arguments:</h5>
4859
4860<p>
4861The only argument is the value to be counted. The argument may be of any
4862integer type. The return type must match the argument type.
4863</p>
4864
4865<h5>Semantics:</h5>
4866
4867<p>
4868The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4869in a variable. If the src == 0 then the result is the size in bits of the type
4870of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4871</p>
4872</div>
4873
4874
4875
4876<!-- _______________________________________________________________________ -->
4877<div class="doc_subsubsection">
4878 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4879</div>
4880
4881<div class="doc_text">
4882
4883<h5>Syntax:</h5>
4884<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
4885integer bit width. Not all targets support all bit widths however.
4886<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4887 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
4888 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4889 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4890 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
4891 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4892</pre>
4893
4894<h5>Overview:</h5>
4895
4896<p>
4897The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4898trailing zeros.
4899</p>
4900
4901<h5>Arguments:</h5>
4902
4903<p>
4904The only argument is the value to be counted. The argument may be of any
4905integer type. The return type must match the argument type.
4906</p>
4907
4908<h5>Semantics:</h5>
4909
4910<p>
4911The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4912in a variable. If the src == 0 then the result is the size in bits of the type
4913of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4914</p>
4915</div>
4916
4917<!-- _______________________________________________________________________ -->
4918<div class="doc_subsubsection">
4919 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
4920</div>
4921
4922<div class="doc_text">
4923
4924<h5>Syntax:</h5>
4925<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
4926on any integer bit width.
4927<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4928 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
4929 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4930</pre>
4931
4932<h5>Overview:</h5>
4933<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
4934range of bits from an integer value and returns them in the same bit width as
4935the original value.</p>
4936
4937<h5>Arguments:</h5>
4938<p>The first argument, <tt>%val</tt> and the result may be integer types of
4939any bit width but they must have the same bit width. The second and third
4940arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
4941
4942<h5>Semantics:</h5>
4943<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
4944of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
4945<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
4946operates in forward mode.</p>
4947<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
4948right by <tt>%loBit</tt> bits and then ANDing it with a mask with
4949only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
4950<ol>
4951 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
4952 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
4953 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
4954 to determine the number of bits to retain.</li>
4955 <li>A mask of the retained bits is created by shifting a -1 value.</li>
4956 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
4957</ol>
4958<p>In reverse mode, a similar computation is made except that the bits are
4959returned in the reverse order. So, for example, if <tt>X</tt> has the value
4960<tt>i16 0x0ACF (101011001111)</tt> and we apply
4961<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
4962<tt>i16 0x0026 (000000100110)</tt>.</p>
4963</div>
4964
4965<div class="doc_subsubsection">
4966 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
4967</div>
4968
4969<div class="doc_text">
4970
4971<h5>Syntax:</h5>
4972<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
4973on any integer bit width.
4974<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4975 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
4976 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]4977</pre>
4978
4979<h5>Overview:</h5>
4980<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
4981of bits in an integer value with another integer value. It returns the integer
4982with the replaced bits.</p>
4983
4984<h5>Arguments:</h5>
4985<p>The first argument, <tt>%val</tt> and the result may be integer types of
4986any bit width but they must have the same bit width. <tt>%val</tt> is the value
4987whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
4988integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
4989type since they specify only a bit index.</p>
4990
4991<h5>Semantics:</h5>
4992<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
4993of operation: forwards and reverse. If <tt>%lo</tt> is greater than
4994<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
4995operates in forward mode.</p>
4996<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
4997truncating it down to the size of the replacement area or zero extending it
4998up to that size.</p>
4999<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5000are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5001in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5002to the <tt>%hi</tt>th bit.
5003<p>In reverse mode, a similar computation is made except that the bits are
5004reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5005<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
5006<h5>Examples:</h5>
5007<pre>
5008 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
5009 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
5010 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
5011 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
5012 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
5013</pre>
5014</div>
5015
5016<!-- ======================================================================= -->
5017<div class="doc_subsection">
5018 <a name="int_debugger">Debugger Intrinsics</a>
5019</div>
5020
5021<div class="doc_text">
5022<p>
5023The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5024are described in the <a
5025href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5026Debugging</a> document.
5027</p>
5028</div>
5029
5030
5031<!-- ======================================================================= -->
5032<div class="doc_subsection">
5033 <a name="int_eh">Exception Handling Intrinsics</a>
5034</div>
5035
5036<div class="doc_text">
5037<p> The LLVM exception handling intrinsics (which all start with
5038<tt>llvm.eh.</tt> prefix), are described in the <a
5039href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5040Handling</a> document. </p>
5041</div>
5042
5043<!-- ======================================================================= -->
5044<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5045 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5046</div>
5047
5048<div class="doc_text">
5049<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5050 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5051 the <tt>nest</tt> attribute, from a function. The result is a callable
5052 function pointer lacking the nest parameter - the caller does not need
5053 to provide a value for it. Instead, the value to use is stored in
5054 advance in a "trampoline", a block of memory usually allocated
5055 on the stack, which also contains code to splice the nest value into the
5056 argument list. This is used to implement the GCC nested function address
5057 extension.
5058</p>
5059<p>
5060 For example, if the function is
5061 <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]5062 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]5063<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5064 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5065 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5066 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5067 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5068</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]5069 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5070 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5071</div>
5072
5073<!-- _______________________________________________________________________ -->
5074<div class="doc_subsubsection">
5075 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5076</div>
5077<div class="doc_text">
5078<h5>Syntax:</h5>
5079<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5080declare 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]5081</pre>
5082<h5>Overview:</h5>
5083<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5084 This fills the memory pointed to by <tt>tramp</tt> with code
5085 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5086</p>
5087<h5>Arguments:</h5>
5088<p>
5089 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5090 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5091 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]5092 intrinsic. Note that the size and the alignment are target-specific - LLVM
5093 currently provides no portable way of determining them, so a front-end that
5094 generates this intrinsic needs to have some target-specific knowledge.
5095 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]5096</p>
5097<h5>Semantics:</h5>
5098<p>
5099 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5100 dependent code, turning it into a function. A pointer to this function is
5101 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5102 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5103 before being called. The new function's signature is the same as that of
5104 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5105 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5106 of pointer type. Calling the new function is equivalent to calling
5107 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5108 missing <tt>nest</tt> argument. If, after calling
5109 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5110 modified, then the effect of any later call to the returned function pointer is
5111 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5112</p>
5113</div>
5114
5115<!-- ======================================================================= -->
5116<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5117 <a name="int_general">General Intrinsics</a>
5118</div>
5119
5120<div class="doc_text">
5121<p> This class of intrinsics is designed to be generic and has
5122no specific purpose. </p>
5123</div>
5124
5125<!-- _______________________________________________________________________ -->
5126<div class="doc_subsubsection">
5127 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
5128</div>
5129
5130<div class="doc_text">
5131
5132<h5>Syntax:</h5>
5133<pre>
5134 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5135</pre>
5136
5137<h5>Overview:</h5>
5138
5139<p>
5140The '<tt>llvm.var.annotation</tt>' intrinsic
5141</p>
5142
5143<h5>Arguments:</h5>
5144
5145<p>
5146The first argument is a pointer to a value, the second is a pointer to a
5147global string, the third is a pointer to a global string which is the source
5148file name, and the last argument is the line number.
5149</p>
5150
5151<h5>Semantics:</h5>
5152
5153<p>
5154This intrinsic allows annotation of local variables with arbitrary strings.
5155This can be useful for special purpose optimizations that want to look for these
5156 annotations. These have no other defined use, they are ignored by code
5157 generation and optimization.
5158</div>
5159
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5160<!-- _______________________________________________________________________ -->
5161<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]5162 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5163</div>
5164
5165<div class="doc_text">
5166
5167<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5168<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
5169any integer bit width.
5170</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5171<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]5172 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5173 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5174 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5175 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5176 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]5177</pre>
5178
5179<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5180
5181<p>
5182The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5183</p>
5184
5185<h5>Arguments:</h5>
5186
5187<p>
5188The first argument is an integer value (result of some expression),
5189the second is a pointer to a global string, the third is a pointer to a global
5190string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5191It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5192</p>
5193
5194<h5>Semantics:</h5>
5195
5196<p>
5197This intrinsic allows annotations to be put on arbitrary expressions
5198with arbitrary strings. This can be useful for special purpose optimizations
5199that want to look for these annotations. These have no other defined use, they
5200are ignored by code generation and optimization.
5201</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5202
5203<!-- *********************************************************************** -->
5204<hr>
5205<address>
5206 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
5207 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
5208 <a href="http://validator.w3.org/check/referer"><img
5209 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
5210
5211 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
5212 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
5213 Last modified: $Date$
5214</address>
5215</body>
5216</html>