Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 0cc4b11c173716cef30e157b2f67fe7027fe67c1 / . / docs / LangRef.html
blob: f9e0824260a3fec6158f200df1aced855e4532d3 [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>
43 <li><a href="#t_array">Array Type</a></li>
44 <li><a href="#t_function">Function Type</a></li>
45 <li><a href="#t_pointer">Pointer Type</a></li>
46 <li><a href="#t_struct">Structure Type</a></li>
47 <li><a href="#t_pstruct">Packed Structure Type</a></li>
48 <li><a href="#t_vector">Vector Type</a></li>
49 <li><a href="#t_opaque">Opaque Type</a></li>
50 </ol>
51 </li>
52 </ol>
53 </li>
54 <li><a href="#constants">Constants</a>
55 <ol>
56 <li><a href="#simpleconstants">Simple Constants</a>
57 <li><a href="#aggregateconstants">Aggregate Constants</a>
58 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
59 <li><a href="#undefvalues">Undefined Values</a>
60 <li><a href="#constantexprs">Constant Expressions</a>
61 </ol>
62 </li>
63 <li><a href="#othervalues">Other Values</a>
64 <ol>
65 <li><a href="#inlineasm">Inline Assembler Expressions</a>
66 </ol>
67 </li>
68 <li><a href="#instref">Instruction Reference</a>
69 <ol>
70 <li><a href="#terminators">Terminator Instructions</a>
71 <ol>
72 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
73 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
74 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
75 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
76 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
77 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
78 </ol>
79 </li>
80 <li><a href="#binaryops">Binary Operations</a>
81 <ol>
82 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
83 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
84 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
85 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
86 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
87 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
88 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
89 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
90 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
91 </ol>
92 </li>
93 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
94 <ol>
95 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
96 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
97 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
98 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
99 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
100 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
101 </ol>
102 </li>
103 <li><a href="#vectorops">Vector Operations</a>
104 <ol>
105 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
106 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
107 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
108 </ol>
109 </li>
110 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
111 <ol>
112 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
113 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
114 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
115 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
116 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
117 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
118 </ol>
119 </li>
120 <li><a href="#convertops">Conversion Operations</a>
121 <ol>
122 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
123 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
124 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
125 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
126 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
127 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
128 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
129 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
130 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
131 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
132 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
133 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
134 </ol>
135 <li><a href="#otherops">Other Operations</a>
136 <ol>
137 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
138 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
139 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
140 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
141 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
142 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
143 </ol>
144 </li>
145 </ol>
146 </li>
147 <li><a href="#intrinsics">Intrinsic Functions</a>
148 <ol>
149 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
150 <ol>
151 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
152 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
153 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
154 </ol>
155 </li>
156 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
157 <ol>
158 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
159 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
160 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
161 </ol>
162 </li>
163 <li><a href="#int_codegen">Code Generator Intrinsics</a>
164 <ol>
165 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
166 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
167 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
168 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
169 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
170 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
171 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
172 </ol>
173 </li>
174 <li><a href="#int_libc">Standard C Library Intrinsics</a>
175 <ol>
176 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
177 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
178 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
179 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
180 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]181 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
182 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
183 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]184 </ol>
185 </li>
186 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
187 <ol>
188 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
189 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
190 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
191 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
192 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
193 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
194 </ol>
195 </li>
196 <li><a href="#int_debugger">Debugger intrinsics</a></li>
197 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]198 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]199 <ol>
200 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]201 </ol>
202 </li>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]203 <li><a href="#int_general">General intrinsics</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]204 <ol>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]205 <li><a href="#int_var_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]206 <tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]207 </ol>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]208 <ol>
209 <li><a href="#int_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]210 <tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]211 </ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]212 </li>
213 </ol>
214 </li>
215</ol>
216
217<div class="doc_author">
218 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
219 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
220</div>
221
222<!-- *********************************************************************** -->
223<div class="doc_section"> <a name="abstract">Abstract </a></div>
224<!-- *********************************************************************** -->
225
226<div class="doc_text">
227<p>This document is a reference manual for the LLVM assembly language.
228LLVM is an SSA based representation that provides type safety,
229low-level operations, flexibility, and the capability of representing
230'all' high-level languages cleanly. It is the common code
231representation used throughout all phases of the LLVM compilation
232strategy.</p>
233</div>
234
235<!-- *********************************************************************** -->
236<div class="doc_section"> <a name="introduction">Introduction</a> </div>
237<!-- *********************************************************************** -->
238
239<div class="doc_text">
240
241<p>The LLVM code representation is designed to be used in three
242different forms: as an in-memory compiler IR, as an on-disk bitcode
243representation (suitable for fast loading by a Just-In-Time compiler),
244and as a human readable assembly language representation. This allows
245LLVM to provide a powerful intermediate representation for efficient
246compiler transformations and analysis, while providing a natural means
247to debug and visualize the transformations. The three different forms
248of LLVM are all equivalent. This document describes the human readable
249representation and notation.</p>
250
251<p>The LLVM representation aims to be light-weight and low-level
252while being expressive, typed, and extensible at the same time. It
253aims to be a "universal IR" of sorts, by being at a low enough level
254that high-level ideas may be cleanly mapped to it (similar to how
255microprocessors are "universal IR's", allowing many source languages to
256be mapped to them). By providing type information, LLVM can be used as
257the target of optimizations: for example, through pointer analysis, it
258can be proven that a C automatic variable is never accessed outside of
259the current function... allowing it to be promoted to a simple SSA
260value instead of a memory location.</p>
261
262</div>
263
264<!-- _______________________________________________________________________ -->
265<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
266
267<div class="doc_text">
268
269<p>It is important to note that this document describes 'well formed'
270LLVM assembly language. There is a difference between what the parser
271accepts and what is considered 'well formed'. For example, the
272following instruction is syntactically okay, but not well formed:</p>
273
274<div class="doc_code">
275<pre>
276%x = <a href="#i_add">add</a> i32 1, %x
277</pre>
278</div>
279
280<p>...because the definition of <tt>%x</tt> does not dominate all of
281its uses. The LLVM infrastructure provides a verification pass that may
282be used to verify that an LLVM module is well formed. This pass is
283automatically run by the parser after parsing input assembly and by
284the optimizer before it outputs bitcode. The violations pointed out
285by the verifier pass indicate bugs in transformation passes or input to
286the parser.</p>
287</div>
288
Chris Lattnera83fdc02007-10-03 17:34:29 +0000[diff] [blame]289<!-- Describe the typesetting conventions here. -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]290
291<!-- *********************************************************************** -->
292<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
293<!-- *********************************************************************** -->
294
295<div class="doc_text">
296
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]297 <p>LLVM identifiers come in two basic types: global and local. Global
298 identifiers (functions, global variables) begin with the @ character. Local
299 identifiers (register names, types) begin with the % character. Additionally,
300 there are three different formats for identifiers, for different purposes:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]301
302<ol>
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]303 <li>Named values are represented as a string of characters with their prefix.
304 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
305 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]306 Identifiers which require other characters in their names can be surrounded
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]307 with quotes. In this way, anything except a <tt>&quot;</tt> character can
308 be used in a named value.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]309
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]310 <li>Unnamed values are represented as an unsigned numeric value with their
311 prefix. For example, %12, @2, %44.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]312
313 <li>Constants, which are described in a <a href="#constants">section about
314 constants</a>, below.</li>
315</ol>
316
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]317<p>LLVM requires that values start with a prefix for two reasons: Compilers
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]318don't need to worry about name clashes with reserved words, and the set of
319reserved words may be expanded in the future without penalty. Additionally,
320unnamed identifiers allow a compiler to quickly come up with a temporary
321variable without having to avoid symbol table conflicts.</p>
322
323<p>Reserved words in LLVM are very similar to reserved words in other
324languages. There are keywords for different opcodes
325('<tt><a href="#i_add">add</a></tt>',
326 '<tt><a href="#i_bitcast">bitcast</a></tt>',
327 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
328href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
329and others. These reserved words cannot conflict with variable names, because
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]330none of them start with a prefix character ('%' or '@').</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]331
332<p>Here is an example of LLVM code to multiply the integer variable
333'<tt>%X</tt>' by 8:</p>
334
335<p>The easy way:</p>
336
337<div class="doc_code">
338<pre>
339%result = <a href="#i_mul">mul</a> i32 %X, 8
340</pre>
341</div>
342
343<p>After strength reduction:</p>
344
345<div class="doc_code">
346<pre>
347%result = <a href="#i_shl">shl</a> i32 %X, i8 3
348</pre>
349</div>
350
351<p>And the hard way:</p>
352
353<div class="doc_code">
354<pre>
355<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
356<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
357%result = <a href="#i_add">add</a> i32 %1, %1
358</pre>
359</div>
360
361<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
362important lexical features of LLVM:</p>
363
364<ol>
365
366 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
367 line.</li>
368
369 <li>Unnamed temporaries are created when the result of a computation is not
370 assigned to a named value.</li>
371
372 <li>Unnamed temporaries are numbered sequentially</li>
373
374</ol>
375
376<p>...and it also shows a convention that we follow in this document. When
377demonstrating instructions, we will follow an instruction with a comment that
378defines the type and name of value produced. Comments are shown in italic
379text.</p>
380
381</div>
382
383<!-- *********************************************************************** -->
384<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
385<!-- *********************************************************************** -->
386
387<!-- ======================================================================= -->
388<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
389</div>
390
391<div class="doc_text">
392
393<p>LLVM programs are composed of "Module"s, each of which is a
394translation unit of the input programs. Each module consists of
395functions, global variables, and symbol table entries. Modules may be
396combined together with the LLVM linker, which merges function (and
397global variable) definitions, resolves forward declarations, and merges
398symbol table entries. Here is an example of the "hello world" module:</p>
399
400<div class="doc_code">
401<pre><i>; Declare the string constant as a global constant...</i>
402<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
403 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
404
405<i>; External declaration of the puts function</i>
406<a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
407
408<i>; Definition of main function</i>
409define i32 @main() { <i>; i32()* </i>
410 <i>; Convert [13x i8 ]* to i8 *...</i>
411 %cast210 = <a
412 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
413
414 <i>; Call puts function to write out the string to stdout...</i>
415 <a
416 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
417 <a
418 href="#i_ret">ret</a> i32 0<br>}<br>
419</pre>
420</div>
421
422<p>This example is made up of a <a href="#globalvars">global variable</a>
423named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
424function, and a <a href="#functionstructure">function definition</a>
425for "<tt>main</tt>".</p>
426
427<p>In general, a module is made up of a list of global values,
428where both functions and global variables are global values. Global values are
429represented by a pointer to a memory location (in this case, a pointer to an
430array of char, and a pointer to a function), and have one of the following <a
431href="#linkage">linkage types</a>.</p>
432
433</div>
434
435<!-- ======================================================================= -->
436<div class="doc_subsection">
437 <a name="linkage">Linkage Types</a>
438</div>
439
440<div class="doc_text">
441
442<p>
443All Global Variables and Functions have one of the following types of linkage:
444</p>
445
446<dl>
447
448 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
449
450 <dd>Global values with internal linkage are only directly accessible by
451 objects in the current module. In particular, linking code into a module with
452 an internal global value may cause the internal to be renamed as necessary to
453 avoid collisions. Because the symbol is internal to the module, all
454 references can be updated. This corresponds to the notion of the
455 '<tt>static</tt>' keyword in C.
456 </dd>
457
458 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
459
460 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
461 the same name when linkage occurs. This is typically used to implement
462 inline functions, templates, or other code which must be generated in each
463 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
464 allowed to be discarded.
465 </dd>
466
467 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
468
469 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
470 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
471 used for globals that may be emitted in multiple translation units, but that
472 are not guaranteed to be emitted into every translation unit that uses them.
473 One example of this are common globals in C, such as "<tt>int X;</tt>" at
474 global scope.
475 </dd>
476
477 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
478
479 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
480 pointer to array type. When two global variables with appending linkage are
481 linked together, the two global arrays are appended together. This is the
482 LLVM, typesafe, equivalent of having the system linker append together
483 "sections" with identical names when .o files are linked.
484 </dd>
485
486 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
487 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
488 until linked, if not linked, the symbol becomes null instead of being an
489 undefined reference.
490 </dd>
491
492 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
493
494 <dd>If none of the above identifiers are used, the global is externally
495 visible, meaning that it participates in linkage and can be used to resolve
496 external symbol references.
497 </dd>
498</dl>
499
500 <p>
501 The next two types of linkage are targeted for Microsoft Windows platform
502 only. They are designed to support importing (exporting) symbols from (to)
503 DLLs.
504 </p>
505
506 <dl>
507 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
508
509 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
510 or variable via a global pointer to a pointer that is set up by the DLL
511 exporting the symbol. On Microsoft Windows targets, the pointer name is
512 formed by combining <code>_imp__</code> and the function or variable name.
513 </dd>
514
515 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
516
517 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
518 pointer to a pointer in a DLL, so that it can be referenced with the
519 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
520 name is formed by combining <code>_imp__</code> and the function or variable
521 name.
522 </dd>
523
524</dl>
525
526<p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
527variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
528variable and was linked with this one, one of the two would be renamed,
529preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
530external (i.e., lacking any linkage declarations), they are accessible
531outside of the current module.</p>
532<p>It is illegal for a function <i>declaration</i>
533to have any linkage type other than "externally visible", <tt>dllimport</tt>,
534or <tt>extern_weak</tt>.</p>
535<p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
536linkages.
537</div>
538
539<!-- ======================================================================= -->
540<div class="doc_subsection">
541 <a name="callingconv">Calling Conventions</a>
542</div>
543
544<div class="doc_text">
545
546<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
547and <a href="#i_invoke">invokes</a> can all have an optional calling convention
548specified for the call. The calling convention of any pair of dynamic
549caller/callee must match, or the behavior of the program is undefined. The
550following calling conventions are supported by LLVM, and more may be added in
551the future:</p>
552
553<dl>
554 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
555
556 <dd>This calling convention (the default if no other calling convention is
557 specified) matches the target C calling conventions. This calling convention
558 supports varargs function calls and tolerates some mismatch in the declared
559 prototype and implemented declaration of the function (as does normal C).
560 </dd>
561
562 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
563
564 <dd>This calling convention attempts to make calls as fast as possible
565 (e.g. by passing things in registers). This calling convention allows the
566 target to use whatever tricks it wants to produce fast code for the target,
567 without having to conform to an externally specified ABI. Implementations of
568 this convention should allow arbitrary tail call optimization to be supported.
569 This calling convention does not support varargs and requires the prototype of
570 all callees to exactly match the prototype of the function definition.
571 </dd>
572
573 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
574
575 <dd>This calling convention attempts to make code in the caller as efficient
576 as possible under the assumption that the call is not commonly executed. As
577 such, these calls often preserve all registers so that the call does not break
578 any live ranges in the caller side. This calling convention does not support
579 varargs and requires the prototype of all callees to exactly match the
580 prototype of the function definition.
581 </dd>
582
583 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
584
585 <dd>Any calling convention may be specified by number, allowing
586 target-specific calling conventions to be used. Target specific calling
587 conventions start at 64.
588 </dd>
589</dl>
590
591<p>More calling conventions can be added/defined on an as-needed basis, to
592support pascal conventions or any other well-known target-independent
593convention.</p>
594
595</div>
596
597<!-- ======================================================================= -->
598<div class="doc_subsection">
599 <a name="visibility">Visibility Styles</a>
600</div>
601
602<div class="doc_text">
603
604<p>
605All Global Variables and Functions have one of the following visibility styles:
606</p>
607
608<dl>
609 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
610
611 <dd>On ELF, default visibility means that the declaration is visible to other
612 modules and, in shared libraries, means that the declared entity may be
613 overridden. On Darwin, default visibility means that the declaration is
614 visible to other modules. Default visibility corresponds to "external
615 linkage" in the language.
616 </dd>
617
618 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
619
620 <dd>Two declarations of an object with hidden visibility refer to the same
621 object if they are in the same shared object. Usually, hidden visibility
622 indicates that the symbol will not be placed into the dynamic symbol table,
623 so no other module (executable or shared library) can reference it
624 directly.
625 </dd>
626
627 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
628
629 <dd>On ELF, protected visibility indicates that the symbol will be placed in
630 the dynamic symbol table, but that references within the defining module will
631 bind to the local symbol. That is, the symbol cannot be overridden by another
632 module.
633 </dd>
634</dl>
635
636</div>
637
638<!-- ======================================================================= -->
639<div class="doc_subsection">
640 <a name="globalvars">Global Variables</a>
641</div>
642
643<div class="doc_text">
644
645<p>Global variables define regions of memory allocated at compilation time
646instead of run-time. Global variables may optionally be initialized, may have
647an explicit section to be placed in, and may have an optional explicit alignment
648specified. A variable may be defined as "thread_local", which means that it
649will not be shared by threads (each thread will have a separated copy of the
650variable). A variable may be defined as a global "constant," which indicates
651that the contents of the variable will <b>never</b> be modified (enabling better
652optimization, allowing the global data to be placed in the read-only section of
653an executable, etc). Note that variables that need runtime initialization
654cannot be marked "constant" as there is a store to the variable.</p>
655
656<p>
657LLVM explicitly allows <em>declarations</em> of global variables to be marked
658constant, even if the final definition of the global is not. This capability
659can be used to enable slightly better optimization of the program, but requires
660the language definition to guarantee that optimizations based on the
661'constantness' are valid for the translation units that do not include the
662definition.
663</p>
664
665<p>As SSA values, global variables define pointer values that are in
666scope (i.e. they dominate) all basic blocks in the program. Global
667variables always define a pointer to their "content" type because they
668describe a region of memory, and all memory objects in LLVM are
669accessed through pointers.</p>
670
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]671<p>A global variable may be declared to reside in a target-specifc numbered
672address space. For targets that support them, address spaces may affect how
673optimizations are performed and/or what target instructions are used to access
Christopher Lamb20a39e92007-12-12 08:44:39 +0000[diff] [blame]674the variable. The default address space is zero. The address space qualifier
675must precede any other attributes.</p>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]676
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]677<p>LLVM allows an explicit section to be specified for globals. If the target
678supports it, it will emit globals to the section specified.</p>
679
680<p>An explicit alignment may be specified for a global. If not present, or if
681the alignment is set to zero, the alignment of the global is set by the target
682to whatever it feels convenient. If an explicit alignment is specified, the
683global is forced to have at least that much alignment. All alignments must be
684a power of 2.</p>
685
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]686<p>For example, the following defines a global in a numbered address space with
687an initializer, section, and alignment:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]688
689<div class="doc_code">
690<pre>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]691@G = constant float 1.0 addrspace(5), section "foo", align 4
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]692</pre>
693</div>
694
695</div>
696
697
698<!-- ======================================================================= -->
699<div class="doc_subsection">
700 <a name="functionstructure">Functions</a>
701</div>
702
703<div class="doc_text">
704
705<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
706an optional <a href="#linkage">linkage type</a>, an optional
707<a href="#visibility">visibility style</a>, an optional
708<a href="#callingconv">calling convention</a>, a return type, an optional
709<a href="#paramattrs">parameter attribute</a> for the return type, a function
710name, a (possibly empty) argument list (each with optional
711<a href="#paramattrs">parameter attributes</a>), an optional section, an
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]712optional alignment, an optional <a href="#gc">garbage collector name</a>, an
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]713opening curly brace, a list of basic blocks, and a closing curly brace.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]714
715LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
716optional <a href="#linkage">linkage type</a>, an optional
717<a href="#visibility">visibility style</a>, an optional
718<a href="#callingconv">calling convention</a>, a return type, an optional
719<a href="#paramattrs">parameter attribute</a> for the return type, a function
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]720name, a possibly empty list of arguments, an optional alignment, and an optional
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]721<a href="#gc">garbage collector name</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]722
723<p>A function definition contains a list of basic blocks, forming the CFG for
724the function. Each basic block may optionally start with a label (giving the
725basic block a symbol table entry), contains a list of instructions, and ends
726with a <a href="#terminators">terminator</a> instruction (such as a branch or
727function return).</p>
728
729<p>The first basic block in a function is special in two ways: it is immediately
730executed on entrance to the function, and it is not allowed to have predecessor
731basic blocks (i.e. there can not be any branches to the entry block of a
732function). Because the block can have no predecessors, it also cannot have any
733<a href="#i_phi">PHI nodes</a>.</p>
734
735<p>LLVM allows an explicit section to be specified for functions. If the target
736supports it, it will emit functions to the section specified.</p>
737
738<p>An explicit alignment may be specified for a function. If not present, or if
739the alignment is set to zero, the alignment of the function is set by the target
740to whatever it feels convenient. If an explicit alignment is specified, the
741function is forced to have at least that much alignment. All alignments must be
742a power of 2.</p>
743
744</div>
745
746
747<!-- ======================================================================= -->
748<div class="doc_subsection">
749 <a name="aliasstructure">Aliases</a>
750</div>
751<div class="doc_text">
752 <p>Aliases act as "second name" for the aliasee value (which can be either
753 function or global variable or bitcast of global value). Aliases may have an
754 optional <a href="#linkage">linkage type</a>, and an
755 optional <a href="#visibility">visibility style</a>.</p>
756
757 <h5>Syntax:</h5>
758
759<div class="doc_code">
760<pre>
761@&lt;Name&gt; = [Linkage] [Visibility] alias &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
762</pre>
763</div>
764
765</div>
766
767
768
769<!-- ======================================================================= -->
770<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
771<div class="doc_text">
772 <p>The return type and each parameter of a function type may have a set of
773 <i>parameter attributes</i> associated with them. Parameter attributes are
774 used to communicate additional information about the result or parameters of
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]775 a function. Parameter attributes are considered to be part of the function,
776 not of the function type, so functions with different parameter attributes
777 can have the same function type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]778
779 <p>Parameter attributes are simple keywords that follow the type specified. If
780 multiple parameter attributes are needed, they are space separated. For
781 example:</p>
782
783<div class="doc_code">
784<pre>
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]785declare i32 @printf(i8* noalias , ...) nounwind
786declare i32 @atoi(i8*) nounwind readonly
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]787</pre>
788</div>
789
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]790 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
791 <tt>readonly</tt>) come immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]792
793 <p>Currently, only the following parameter attributes are defined:</p>
794 <dl>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]795 <dt><tt>zeroext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]796 <dd>This indicates that the parameter should be zero extended just before
797 a call to this function.</dd>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]798 <dt><tt>signext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]799 <dd>This indicates that the parameter should be sign extended just before
800 a call to this function.</dd>
801 <dt><tt>inreg</tt></dt>
802 <dd>This indicates that the parameter should be placed in register (if
803 possible) during assembling function call. Support for this attribute is
804 target-specific</dd>
805 <dt><tt>sret</tt></dt>
806 <dd>This indicates that the parameter specifies the address of a structure
807 that is the return value of the function in the source program.</dd>
808 <dt><tt>noalias</tt></dt>
809 <dd>This indicates that the parameter not alias any other object or any
810 other "noalias" objects during the function call.
811 <dt><tt>noreturn</tt></dt>
812 <dd>This function attribute indicates that the function never returns. This
813 indicates to LLVM that every call to this function should be treated as if
814 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
815 <dt><tt>nounwind</tt></dt>
816 <dd>This function attribute indicates that the function type does not use
817 the unwind instruction and does not allow stack unwinding to propagate
818 through it.</dd>
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]819 <dt><tt>nest</tt></dt>
820 <dd>This indicates that the parameter can be excised using the
821 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
Duncan Sands13e13f82007-11-22 20:23:04 +0000[diff] [blame]822 <dt><tt>readonly</tt></dt>
Duncan Sandsd69c0e62007-11-14 21:14:02 +0000[diff] [blame]823 <dd>This function attribute indicates that the function has no side-effects
Duncan Sands13e13f82007-11-22 20:23:04 +0000[diff] [blame]824 except for producing a return value or throwing an exception. The value
825 returned must only depend on the function arguments and/or global variables.
826 It may use values obtained by dereferencing pointers.</dd>
827 <dt><tt>readnone</tt></dt>
828 <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]829 function, but in addition it is not allowed to dereference any pointer arguments
830 or global variables.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]831 </dl>
832
833</div>
834
835<!-- ======================================================================= -->
836<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]837 <a name="gc">Garbage Collector Names</a>
838</div>
839
840<div class="doc_text">
841<p>Each function may specify a garbage collector name, which is simply a
842string.</p>
843
844<div class="doc_code"><pre
845>define void @f() gc "name" { ...</pre></div>
846
847<p>The compiler declares the supported values of <i>name</i>. Specifying a
848collector which will cause the compiler to alter its output in order to support
849the named garbage collection algorithm.</p>
850</div>
851
852<!-- ======================================================================= -->
853<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]854 <a name="moduleasm">Module-Level Inline Assembly</a>
855</div>
856
857<div class="doc_text">
858<p>
859Modules may contain "module-level inline asm" blocks, which corresponds to the
860GCC "file scope inline asm" blocks. These blocks are internally concatenated by
861LLVM and treated as a single unit, but may be separated in the .ll file if
862desired. The syntax is very simple:
863</p>
864
865<div class="doc_code">
866<pre>
867module asm "inline asm code goes here"
868module asm "more can go here"
869</pre>
870</div>
871
872<p>The strings can contain any character by escaping non-printable characters.
873 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
874 for the number.
875</p>
876
877<p>
878 The inline asm code is simply printed to the machine code .s file when
879 assembly code is generated.
880</p>
881</div>
882
883<!-- ======================================================================= -->
884<div class="doc_subsection">
885 <a name="datalayout">Data Layout</a>
886</div>
887
888<div class="doc_text">
889<p>A module may specify a target specific data layout string that specifies how
890data is to be laid out in memory. The syntax for the data layout is simply:</p>
891<pre> target datalayout = "<i>layout specification</i>"</pre>
892<p>The <i>layout specification</i> consists of a list of specifications
893separated by the minus sign character ('-'). Each specification starts with a
894letter and may include other information after the letter to define some
895aspect of the data layout. The specifications accepted are as follows: </p>
896<dl>
897 <dt><tt>E</tt></dt>
898 <dd>Specifies that the target lays out data in big-endian form. That is, the
899 bits with the most significance have the lowest address location.</dd>
900 <dt><tt>e</tt></dt>
901 <dd>Specifies that hte target lays out data in little-endian form. That is,
902 the bits with the least significance have the lowest address location.</dd>
903 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
904 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
905 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
906 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
907 too.</dd>
908 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
909 <dd>This specifies the alignment for an integer type of a given bit
910 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
911 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
912 <dd>This specifies the alignment for a vector type of a given bit
913 <i>size</i>.</dd>
914 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
915 <dd>This specifies the alignment for a floating point type of a given bit
916 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
917 (double).</dd>
918 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
919 <dd>This specifies the alignment for an aggregate type of a given bit
920 <i>size</i>.</dd>
921</dl>
922<p>When constructing the data layout for a given target, LLVM starts with a
923default set of specifications which are then (possibly) overriden by the
924specifications in the <tt>datalayout</tt> keyword. The default specifications
925are given in this list:</p>
926<ul>
927 <li><tt>E</tt> - big endian</li>
928 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
929 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
930 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
931 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
932 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
933 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
934 alignment of 64-bits</li>
935 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
936 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
937 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
938 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
939 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
940</ul>
941<p>When llvm is determining the alignment for a given type, it uses the
942following rules:
943<ol>
944 <li>If the type sought is an exact match for one of the specifications, that
945 specification is used.</li>
946 <li>If no match is found, and the type sought is an integer type, then the
947 smallest integer type that is larger than the bitwidth of the sought type is
948 used. If none of the specifications are larger than the bitwidth then the the
949 largest integer type is used. For example, given the default specifications
950 above, the i7 type will use the alignment of i8 (next largest) while both
951 i65 and i256 will use the alignment of i64 (largest specified).</li>
952 <li>If no match is found, and the type sought is a vector type, then the
953 largest vector type that is smaller than the sought vector type will be used
954 as a fall back. This happens because <128 x double> can be implemented in
955 terms of 64 <2 x double>, for example.</li>
956</ol>
957</div>
958
959<!-- *********************************************************************** -->
960<div class="doc_section"> <a name="typesystem">Type System</a> </div>
961<!-- *********************************************************************** -->
962
963<div class="doc_text">
964
965<p>The LLVM type system is one of the most important features of the
966intermediate representation. Being typed enables a number of
967optimizations to be performed on the IR directly, without having to do
968extra analyses on the side before the transformation. A strong type
969system makes it easier to read the generated code and enables novel
970analyses and transformations that are not feasible to perform on normal
971three address code representations.</p>
972
973</div>
974
975<!-- ======================================================================= -->
976<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
977<div class="doc_text">
978<p>The primitive types are the fundamental building blocks of the LLVM
979system. The current set of primitive types is as follows:</p>
980
981<table class="layout">
982 <tr class="layout">
983 <td class="left">
984 <table>
985 <tbody>
986 <tr><th>Type</th><th>Description</th></tr>
987 <tr><td><tt><a name="t_void">void</a></tt></td><td>No value</td></tr>
988 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
989 </tbody>
990 </table>
991 </td>
992 <td class="right">
993 <table>
994 <tbody>
995 <tr><th>Type</th><th>Description</th></tr>
996 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
997 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
998 </tbody>
999 </table>
1000 </td>
1001 </tr>
1002</table>
1003</div>
1004
1005<!-- _______________________________________________________________________ -->
1006<div class="doc_subsubsection"> <a name="t_classifications">Type
1007Classifications</a> </div>
1008<div class="doc_text">
1009<p>These different primitive types fall into a few useful
1010classifications:</p>
1011
1012<table border="1" cellspacing="0" cellpadding="4">
1013 <tbody>
1014 <tr><th>Classification</th><th>Types</th></tr>
1015 <tr>
1016 <td><a name="t_integer">integer</a></td>
1017 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1018 </tr>
1019 <tr>
1020 <td><a name="t_floating">floating point</a></td>
1021 <td><tt>float, double</tt></td>
1022 </tr>
1023 <tr>
1024 <td><a name="t_firstclass">first class</a></td>
1025 <td><tt>i1, ..., float, double, <br/>
1026 <a href="#t_pointer">pointer</a>,<a href="#t_vector">vector</a></tt>
1027 </td>
1028 </tr>
1029 </tbody>
1030</table>
1031
1032<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1033most important. Values of these types are the only ones which can be
1034produced by instructions, passed as arguments, or used as operands to
1035instructions. This means that all structures and arrays must be
1036manipulated either by pointer or by component.</p>
1037</div>
1038
1039<!-- ======================================================================= -->
1040<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1041
1042<div class="doc_text">
1043
1044<p>The real power in LLVM comes from the derived types in the system.
1045This is what allows a programmer to represent arrays, functions,
1046pointers, and other useful types. Note that these derived types may be
1047recursive: For example, it is possible to have a two dimensional array.</p>
1048
1049</div>
1050
1051<!-- _______________________________________________________________________ -->
1052<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1053
1054<div class="doc_text">
1055
1056<h5>Overview:</h5>
1057<p>The integer type is a very simple derived type that simply specifies an
1058arbitrary bit width for the integer type desired. Any bit width from 1 bit to
10592^23-1 (about 8 million) can be specified.</p>
1060
1061<h5>Syntax:</h5>
1062
1063<pre>
1064 iN
1065</pre>
1066
1067<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1068value.</p>
1069
1070<h5>Examples:</h5>
1071<table class="layout">
1072 <tr class="layout">
1073 <td class="left">
1074 <tt>i1</tt><br/>
1075 <tt>i4</tt><br/>
1076 <tt>i8</tt><br/>
1077 <tt>i16</tt><br/>
1078 <tt>i32</tt><br/>
1079 <tt>i42</tt><br/>
1080 <tt>i64</tt><br/>
1081 <tt>i1942652</tt><br/>
1082 </td>
1083 <td class="left">
1084 A boolean integer of 1 bit<br/>
1085 A nibble sized integer of 4 bits.<br/>
1086 A byte sized integer of 8 bits.<br/>
1087 A half word sized integer of 16 bits.<br/>
1088 A word sized integer of 32 bits.<br/>
1089 An integer whose bit width is the answer. <br/>
1090 A double word sized integer of 64 bits.<br/>
1091 A really big integer of over 1 million bits.<br/>
1092 </td>
1093 </tr>
1094</table>
1095</div>
1096
1097<!-- _______________________________________________________________________ -->
1098<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1099
1100<div class="doc_text">
1101
1102<h5>Overview:</h5>
1103
1104<p>The array type is a very simple derived type that arranges elements
1105sequentially in memory. The array type requires a size (number of
1106elements) and an underlying data type.</p>
1107
1108<h5>Syntax:</h5>
1109
1110<pre>
1111 [&lt;# elements&gt; x &lt;elementtype&gt;]
1112</pre>
1113
1114<p>The number of elements is a constant integer value; elementtype may
1115be any type with a size.</p>
1116
1117<h5>Examples:</h5>
1118<table class="layout">
1119 <tr class="layout">
1120 <td class="left">
1121 <tt>[40 x i32 ]</tt><br/>
1122 <tt>[41 x i32 ]</tt><br/>
1123 <tt>[40 x i8]</tt><br/>
1124 </td>
1125 <td class="left">
1126 Array of 40 32-bit integer values.<br/>
1127 Array of 41 32-bit integer values.<br/>
1128 Array of 40 8-bit integer values.<br/>
1129 </td>
1130 </tr>
1131</table>
1132<p>Here are some examples of multidimensional arrays:</p>
1133<table class="layout">
1134 <tr class="layout">
1135 <td class="left">
1136 <tt>[3 x [4 x i32]]</tt><br/>
1137 <tt>[12 x [10 x float]]</tt><br/>
1138 <tt>[2 x [3 x [4 x i16]]]</tt><br/>
1139 </td>
1140 <td class="left">
1141 3x4 array of 32-bit integer values.<br/>
1142 12x10 array of single precision floating point values.<br/>
1143 2x3x4 array of 16-bit integer values.<br/>
1144 </td>
1145 </tr>
1146</table>
1147
1148<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1149length array. Normally, accesses past the end of an array are undefined in
1150LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1151As a special case, however, zero length arrays are recognized to be variable
1152length. This allows implementation of 'pascal style arrays' with the LLVM
1153type "{ i32, [0 x float]}", for example.</p>
1154
1155</div>
1156
1157<!-- _______________________________________________________________________ -->
1158<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1159<div class="doc_text">
1160<h5>Overview:</h5>
1161<p>The function type can be thought of as a function signature. It
1162consists of a return type and a list of formal parameter types.
1163Function types are usually used to build virtual function tables
1164(which are structures of pointers to functions), for indirect function
1165calls, and when defining a function.</p>
1166<p>
1167The return type of a function type cannot be an aggregate type.
1168</p>
1169<h5>Syntax:</h5>
1170<pre> &lt;returntype&gt; (&lt;parameter list&gt;)<br></pre>
1171<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1172specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1173which indicates that the function takes a variable number of arguments.
1174Variable argument functions can access their arguments with the <a
1175 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1176<h5>Examples:</h5>
1177<table class="layout">
1178 <tr class="layout">
1179 <td class="left"><tt>i32 (i32)</tt></td>
1180 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1181 </td>
1182 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1183 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1184 </tt></td>
1185 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1186 an <tt>i16</tt> that should be sign extended and a
1187 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1188 <tt>float</tt>.
1189 </td>
1190 </tr><tr class="layout">
1191 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1192 <td class="left">A vararg function that takes at least one
1193 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1194 which returns an integer. This is the signature for <tt>printf</tt> in
1195 LLVM.
1196 </td>
1197 </tr>
1198</table>
1199
1200</div>
1201<!-- _______________________________________________________________________ -->
1202<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1203<div class="doc_text">
1204<h5>Overview:</h5>
1205<p>The structure type is used to represent a collection of data members
1206together in memory. The packing of the field types is defined to match
1207the ABI of the underlying processor. The elements of a structure may
1208be any type that has a size.</p>
1209<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1210and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1211field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1212instruction.</p>
1213<h5>Syntax:</h5>
1214<pre> { &lt;type list&gt; }<br></pre>
1215<h5>Examples:</h5>
1216<table class="layout">
1217 <tr class="layout">
1218 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1219 <td class="left">A triple of three <tt>i32</tt> values</td>
1220 </tr><tr class="layout">
1221 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1222 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1223 second element is a <a href="#t_pointer">pointer</a> to a
1224 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1225 an <tt>i32</tt>.</td>
1226 </tr>
1227</table>
1228</div>
1229
1230<!-- _______________________________________________________________________ -->
1231<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1232</div>
1233<div class="doc_text">
1234<h5>Overview:</h5>
1235<p>The packed structure type is used to represent a collection of data members
1236together in memory. There is no padding between fields. Further, the alignment
1237of a packed structure is 1 byte. The elements of a packed structure may
1238be any type that has a size.</p>
1239<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1240and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1241field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1242instruction.</p>
1243<h5>Syntax:</h5>
1244<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1245<h5>Examples:</h5>
1246<table class="layout">
1247 <tr class="layout">
1248 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1249 <td class="left">A triple of three <tt>i32</tt> values</td>
1250 </tr><tr class="layout">
1251 <td class="left"><tt>&lt;&nbsp;{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}&nbsp;&gt;</tt></td>
1252 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1253 second element is a <a href="#t_pointer">pointer</a> to a
1254 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1255 an <tt>i32</tt>.</td>
1256 </tr>
1257</table>
1258</div>
1259
1260<!-- _______________________________________________________________________ -->
1261<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1262<div class="doc_text">
1263<h5>Overview:</h5>
1264<p>As in many languages, the pointer type represents a pointer or
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1265reference to another object, which must live in memory. Pointer types may have
1266an optional address space attribute defining the target-specific numbered
1267address space where the pointed-to object resides. The default address space is
1268zero.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1269<h5>Syntax:</h5>
1270<pre> &lt;type&gt; *<br></pre>
1271<h5>Examples:</h5>
1272<table class="layout">
1273 <tr class="layout">
1274 <td class="left">
1275 <tt>[4x i32]*</tt><br/>
1276 <tt>i32 (i32 *) *</tt><br/>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1277 <tt>i32 addrspace(5)*</tt><br/>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1278 </td>
1279 <td class="left">
1280 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
1281 four <tt>i32</tt> values<br/>
1282 A <a href="#t_pointer">pointer</a> to a <a
1283 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1284 <tt>i32</tt>.<br/>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1285 A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value that resides
1286 in address space 5.<br/>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1287 </td>
1288 </tr>
1289</table>
1290</div>
1291
1292<!-- _______________________________________________________________________ -->
1293<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1294<div class="doc_text">
1295
1296<h5>Overview:</h5>
1297
1298<p>A vector type is a simple derived type that represents a vector
1299of elements. Vector types are used when multiple primitive data
1300are operated in parallel using a single instruction (SIMD).
1301A vector type requires a size (number of
1302elements) and an underlying primitive data type. Vectors must have a power
1303of two length (1, 2, 4, 8, 16 ...). Vector types are
1304considered <a href="#t_firstclass">first class</a>.</p>
1305
1306<h5>Syntax:</h5>
1307
1308<pre>
1309 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1310</pre>
1311
1312<p>The number of elements is a constant integer value; elementtype may
1313be any integer or floating point type.</p>
1314
1315<h5>Examples:</h5>
1316
1317<table class="layout">
1318 <tr class="layout">
1319 <td class="left">
1320 <tt>&lt;4 x i32&gt;</tt><br/>
1321 <tt>&lt;8 x float&gt;</tt><br/>
1322 <tt>&lt;2 x i64&gt;</tt><br/>
1323 </td>
1324 <td class="left">
1325 Vector of 4 32-bit integer values.<br/>
1326 Vector of 8 floating-point values.<br/>
1327 Vector of 2 64-bit integer values.<br/>
1328 </td>
1329 </tr>
1330</table>
1331</div>
1332
1333<!-- _______________________________________________________________________ -->
1334<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1335<div class="doc_text">
1336
1337<h5>Overview:</h5>
1338
1339<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1340corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1341In LLVM, opaque types can eventually be resolved to any type (not just a
1342structure type).</p>
1343
1344<h5>Syntax:</h5>
1345
1346<pre>
1347 opaque
1348</pre>
1349
1350<h5>Examples:</h5>
1351
1352<table class="layout">
1353 <tr class="layout">
1354 <td class="left">
1355 <tt>opaque</tt>
1356 </td>
1357 <td class="left">
1358 An opaque type.<br/>
1359 </td>
1360 </tr>
1361</table>
1362</div>
1363
1364
1365<!-- *********************************************************************** -->
1366<div class="doc_section"> <a name="constants">Constants</a> </div>
1367<!-- *********************************************************************** -->
1368
1369<div class="doc_text">
1370
1371<p>LLVM has several different basic types of constants. This section describes
1372them all and their syntax.</p>
1373
1374</div>
1375
1376<!-- ======================================================================= -->
1377<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1378
1379<div class="doc_text">
1380
1381<dl>
1382 <dt><b>Boolean constants</b></dt>
1383
1384 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1385 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1386 </dd>
1387
1388 <dt><b>Integer constants</b></dt>
1389
1390 <dd>Standard integers (such as '4') are constants of the <a
1391 href="#t_integer">integer</a> type. Negative numbers may be used with
1392 integer types.
1393 </dd>
1394
1395 <dt><b>Floating point constants</b></dt>
1396
1397 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1398 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1399 notation (see below). Floating point constants must have a <a
1400 href="#t_floating">floating point</a> type. </dd>
1401
1402 <dt><b>Null pointer constants</b></dt>
1403
1404 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1405 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1406
1407</dl>
1408
1409<p>The one non-intuitive notation for constants is the optional hexadecimal form
1410of floating point constants. For example, the form '<tt>double
14110x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
14124.5e+15</tt>'. The only time hexadecimal floating point constants are required
1413(and the only time that they are generated by the disassembler) is when a
1414floating point constant must be emitted but it cannot be represented as a
1415decimal floating point number. For example, NaN's, infinities, and other
1416special values are represented in their IEEE hexadecimal format so that
1417assembly and disassembly do not cause any bits to change in the constants.</p>
1418
1419</div>
1420
1421<!-- ======================================================================= -->
1422<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1423</div>
1424
1425<div class="doc_text">
1426<p>Aggregate constants arise from aggregation of simple constants
1427and smaller aggregate constants.</p>
1428
1429<dl>
1430 <dt><b>Structure constants</b></dt>
1431
1432 <dd>Structure constants are represented with notation similar to structure
1433 type definitions (a comma separated list of elements, surrounded by braces
1434 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* %G }</tt>",
1435 where "<tt>%G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1436 must have <a href="#t_struct">structure type</a>, and the number and
1437 types of elements must match those specified by the type.
1438 </dd>
1439
1440 <dt><b>Array constants</b></dt>
1441
1442 <dd>Array constants are represented with notation similar to array type
1443 definitions (a comma separated list of elements, surrounded by square brackets
1444 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1445 constants must have <a href="#t_array">array type</a>, and the number and
1446 types of elements must match those specified by the type.
1447 </dd>
1448
1449 <dt><b>Vector constants</b></dt>
1450
1451 <dd>Vector constants are represented with notation similar to vector type
1452 definitions (a comma separated list of elements, surrounded by
1453 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1454 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1455 href="#t_vector">vector type</a>, and the number and types of elements must
1456 match those specified by the type.
1457 </dd>
1458
1459 <dt><b>Zero initialization</b></dt>
1460
1461 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1462 value to zero of <em>any</em> type, including scalar and aggregate types.
1463 This is often used to avoid having to print large zero initializers (e.g. for
1464 large arrays) and is always exactly equivalent to using explicit zero
1465 initializers.
1466 </dd>
1467</dl>
1468
1469</div>
1470
1471<!-- ======================================================================= -->
1472<div class="doc_subsection">
1473 <a name="globalconstants">Global Variable and Function Addresses</a>
1474</div>
1475
1476<div class="doc_text">
1477
1478<p>The addresses of <a href="#globalvars">global variables</a> and <a
1479href="#functionstructure">functions</a> are always implicitly valid (link-time)
1480constants. These constants are explicitly referenced when the <a
1481href="#identifiers">identifier for the global</a> is used and always have <a
1482href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1483file:</p>
1484
1485<div class="doc_code">
1486<pre>
1487@X = global i32 17
1488@Y = global i32 42
1489@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1490</pre>
1491</div>
1492
1493</div>
1494
1495<!-- ======================================================================= -->
1496<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1497<div class="doc_text">
1498 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1499 no specific value. Undefined values may be of any type and be used anywhere
1500 a constant is permitted.</p>
1501
1502 <p>Undefined values indicate to the compiler that the program is well defined
1503 no matter what value is used, giving the compiler more freedom to optimize.
1504 </p>
1505</div>
1506
1507<!-- ======================================================================= -->
1508<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1509</div>
1510
1511<div class="doc_text">
1512
1513<p>Constant expressions are used to allow expressions involving other constants
1514to be used as constants. Constant expressions may be of any <a
1515href="#t_firstclass">first class</a> type and may involve any LLVM operation
1516that does not have side effects (e.g. load and call are not supported). The
1517following is the syntax for constant expressions:</p>
1518
1519<dl>
1520 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1521 <dd>Truncate a constant to another type. The bit size of CST must be larger
1522 than the bit size of TYPE. Both types must be integers.</dd>
1523
1524 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1525 <dd>Zero extend a constant to another type. The bit size of CST must be
1526 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1527
1528 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1529 <dd>Sign extend a constant to another type. The bit size of CST must be
1530 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1531
1532 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1533 <dd>Truncate a floating point constant to another floating point type. The
1534 size of CST must be larger than the size of TYPE. Both types must be
1535 floating point.</dd>
1536
1537 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1538 <dd>Floating point extend a constant to another type. The size of CST must be
1539 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1540
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1541 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1542 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1543 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1544 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1545 of the same number of elements. If the value won't fit in the integer type,
1546 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1547
1548 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1549 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1550 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1551 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1552 of the same number of elements. If the value won't fit in the integer type,
1553 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1554
1555 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1556 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1557 constant. TYPE must be a scalar or vector floating point type. CST must be of
1558 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1559 of the same number of elements. If the value won't fit in the floating point
1560 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1561
1562 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1563 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1564 constant. TYPE must be a scalar or vector floating point type. CST must be of
1565 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1566 of the same number of elements. If the value won't fit in the floating point
1567 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1568
1569 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1570 <dd>Convert a pointer typed constant to the corresponding integer constant
1571 TYPE must be an integer type. CST must be of pointer type. The CST value is
1572 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1573
1574 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1575 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1576 pointer type. CST must be of integer type. The CST value is zero extended,
1577 truncated, or unchanged to make it fit in a pointer size. This one is
1578 <i>really</i> dangerous!</dd>
1579
1580 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1581 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1582 identical (same number of bits). The conversion is done as if the CST value
1583 was stored to memory and read back as TYPE. In other words, no bits change
1584 with this operator, just the type. This can be used for conversion of
1585 vector types to any other type, as long as they have the same bit width. For
1586 pointers it is only valid to cast to another pointer type.
1587 </dd>
1588
1589 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1590
1591 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1592 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1593 instruction, the index list may have zero or more indexes, which are required
1594 to make sense for the type of "CSTPTR".</dd>
1595
1596 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1597
1598 <dd>Perform the <a href="#i_select">select operation</a> on
1599 constants.</dd>
1600
1601 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1602 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1603
1604 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1605 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1606
1607 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1608
1609 <dd>Perform the <a href="#i_extractelement">extractelement
1610 operation</a> on constants.
1611
1612 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1613
1614 <dd>Perform the <a href="#i_insertelement">insertelement
1615 operation</a> on constants.</dd>
1616
1617
1618 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1619
1620 <dd>Perform the <a href="#i_shufflevector">shufflevector
1621 operation</a> on constants.</dd>
1622
1623 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1624
1625 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1626 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1627 binary</a> operations. The constraints on operands are the same as those for
1628 the corresponding instruction (e.g. no bitwise operations on floating point
1629 values are allowed).</dd>
1630</dl>
1631</div>
1632
1633<!-- *********************************************************************** -->
1634<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1635<!-- *********************************************************************** -->
1636
1637<!-- ======================================================================= -->
1638<div class="doc_subsection">
1639<a name="inlineasm">Inline Assembler Expressions</a>
1640</div>
1641
1642<div class="doc_text">
1643
1644<p>
1645LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1646Module-Level Inline Assembly</a>) through the use of a special value. This
1647value represents the inline assembler as a string (containing the instructions
1648to emit), a list of operand constraints (stored as a string), and a flag that
1649indicates whether or not the inline asm expression has side effects. An example
1650inline assembler expression is:
1651</p>
1652
1653<div class="doc_code">
1654<pre>
1655i32 (i32) asm "bswap $0", "=r,r"
1656</pre>
1657</div>
1658
1659<p>
1660Inline assembler expressions may <b>only</b> be used as the callee operand of
1661a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1662</p>
1663
1664<div class="doc_code">
1665<pre>
1666%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1667</pre>
1668</div>
1669
1670<p>
1671Inline asms with side effects not visible in the constraint list must be marked
1672as having side effects. This is done through the use of the
1673'<tt>sideeffect</tt>' keyword, like so:
1674</p>
1675
1676<div class="doc_code">
1677<pre>
1678call void asm sideeffect "eieio", ""()
1679</pre>
1680</div>
1681
1682<p>TODO: The format of the asm and constraints string still need to be
1683documented here. Constraints on what can be done (e.g. duplication, moving, etc
1684need to be documented).
1685</p>
1686
1687</div>
1688
1689<!-- *********************************************************************** -->
1690<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1691<!-- *********************************************************************** -->
1692
1693<div class="doc_text">
1694
1695<p>The LLVM instruction set consists of several different
1696classifications of instructions: <a href="#terminators">terminator
1697instructions</a>, <a href="#binaryops">binary instructions</a>,
1698<a href="#bitwiseops">bitwise binary instructions</a>, <a
1699 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1700instructions</a>.</p>
1701
1702</div>
1703
1704<!-- ======================================================================= -->
1705<div class="doc_subsection"> <a name="terminators">Terminator
1706Instructions</a> </div>
1707
1708<div class="doc_text">
1709
1710<p>As mentioned <a href="#functionstructure">previously</a>, every
1711basic block in a program ends with a "Terminator" instruction, which
1712indicates which block should be executed after the current block is
1713finished. These terminator instructions typically yield a '<tt>void</tt>'
1714value: they produce control flow, not values (the one exception being
1715the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1716<p>There are six different terminator instructions: the '<a
1717 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1718instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1719the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1720 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1721 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1722
1723</div>
1724
1725<!-- _______________________________________________________________________ -->
1726<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1727Instruction</a> </div>
1728<div class="doc_text">
1729<h5>Syntax:</h5>
1730<pre> ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
1731 ret void <i>; Return from void function</i>
1732</pre>
1733<h5>Overview:</h5>
1734<p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1735value) from a function back to the caller.</p>
1736<p>There are two forms of the '<tt>ret</tt>' instruction: one that
1737returns a value and then causes control flow, and one that just causes
1738control flow to occur.</p>
1739<h5>Arguments:</h5>
1740<p>The '<tt>ret</tt>' instruction may return any '<a
1741 href="#t_firstclass">first class</a>' type. Notice that a function is
1742not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1743instruction inside of the function that returns a value that does not
1744match the return type of the function.</p>
1745<h5>Semantics:</h5>
1746<p>When the '<tt>ret</tt>' instruction is executed, control flow
1747returns back to the calling function's context. If the caller is a "<a
1748 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1749the instruction after the call. If the caller was an "<a
1750 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1751at the beginning of the "normal" destination block. If the instruction
1752returns a value, that value shall set the call or invoke instruction's
1753return value.</p>
1754<h5>Example:</h5>
1755<pre> ret i32 5 <i>; Return an integer value of 5</i>
1756 ret void <i>; Return from a void function</i>
1757</pre>
1758</div>
1759<!-- _______________________________________________________________________ -->
1760<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1761<div class="doc_text">
1762<h5>Syntax:</h5>
1763<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
1764</pre>
1765<h5>Overview:</h5>
1766<p>The '<tt>br</tt>' instruction is used to cause control flow to
1767transfer to a different basic block in the current function. There are
1768two forms of this instruction, corresponding to a conditional branch
1769and an unconditional branch.</p>
1770<h5>Arguments:</h5>
1771<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1772single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1773unconditional form of the '<tt>br</tt>' instruction takes a single
1774'<tt>label</tt>' value as a target.</p>
1775<h5>Semantics:</h5>
1776<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1777argument is evaluated. If the value is <tt>true</tt>, control flows
1778to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1779control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1780<h5>Example:</h5>
1781<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
1782 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1783</div>
1784<!-- _______________________________________________________________________ -->
1785<div class="doc_subsubsection">
1786 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1787</div>
1788
1789<div class="doc_text">
1790<h5>Syntax:</h5>
1791
1792<pre>
1793 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
1794</pre>
1795
1796<h5>Overview:</h5>
1797
1798<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1799several different places. It is a generalization of the '<tt>br</tt>'
1800instruction, allowing a branch to occur to one of many possible
1801destinations.</p>
1802
1803
1804<h5>Arguments:</h5>
1805
1806<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1807comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1808an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1809table is not allowed to contain duplicate constant entries.</p>
1810
1811<h5>Semantics:</h5>
1812
1813<p>The <tt>switch</tt> instruction specifies a table of values and
1814destinations. When the '<tt>switch</tt>' instruction is executed, this
1815table is searched for the given value. If the value is found, control flow is
1816transfered to the corresponding destination; otherwise, control flow is
1817transfered to the default destination.</p>
1818
1819<h5>Implementation:</h5>
1820
1821<p>Depending on properties of the target machine and the particular
1822<tt>switch</tt> instruction, this instruction may be code generated in different
1823ways. For example, it could be generated as a series of chained conditional
1824branches or with a lookup table.</p>
1825
1826<h5>Example:</h5>
1827
1828<pre>
1829 <i>; Emulate a conditional br instruction</i>
1830 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1831 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1832
1833 <i>; Emulate an unconditional br instruction</i>
1834 switch i32 0, label %dest [ ]
1835
1836 <i>; Implement a jump table:</i>
1837 switch i32 %val, label %otherwise [ i32 0, label %onzero
1838 i32 1, label %onone
1839 i32 2, label %ontwo ]
1840</pre>
1841</div>
1842
1843<!-- _______________________________________________________________________ -->
1844<div class="doc_subsubsection">
1845 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1846</div>
1847
1848<div class="doc_text">
1849
1850<h5>Syntax:</h5>
1851
1852<pre>
1853 &lt;result&gt; = invoke [<a href="#callingconv">cconv</a>] &lt;ptr to function ty&gt; %&lt;function ptr val&gt;(&lt;function args&gt;)
1854 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
1855</pre>
1856
1857<h5>Overview:</h5>
1858
1859<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1860function, with the possibility of control flow transfer to either the
1861'<tt>normal</tt>' label or the
1862'<tt>exception</tt>' label. If the callee function returns with the
1863"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1864"normal" label. If the callee (or any indirect callees) returns with the "<a
1865href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1866continued at the dynamically nearest "exception" label.</p>
1867
1868<h5>Arguments:</h5>
1869
1870<p>This instruction requires several arguments:</p>
1871
1872<ol>
1873 <li>
1874 The optional "cconv" marker indicates which <a href="#callingconv">calling
1875 convention</a> the call should use. If none is specified, the call defaults
1876 to using C calling conventions.
1877 </li>
1878 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1879 function value being invoked. In most cases, this is a direct function
1880 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1881 an arbitrary pointer to function value.
1882 </li>
1883
1884 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1885 function to be invoked. </li>
1886
1887 <li>'<tt>function args</tt>': argument list whose types match the function
1888 signature argument types. If the function signature indicates the function
1889 accepts a variable number of arguments, the extra arguments can be
1890 specified. </li>
1891
1892 <li>'<tt>normal label</tt>': the label reached when the called function
1893 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1894
1895 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1896 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1897
1898</ol>
1899
1900<h5>Semantics:</h5>
1901
1902<p>This instruction is designed to operate as a standard '<tt><a
1903href="#i_call">call</a></tt>' instruction in most regards. The primary
1904difference is that it establishes an association with a label, which is used by
1905the runtime library to unwind the stack.</p>
1906
1907<p>This instruction is used in languages with destructors to ensure that proper
1908cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1909exception. Additionally, this is important for implementation of
1910'<tt>catch</tt>' clauses in high-level languages that support them.</p>
1911
1912<h5>Example:</h5>
1913<pre>
1914 %retval = invoke i32 %Test(i32 15) to label %Continue
1915 unwind label %TestCleanup <i>; {i32}:retval set</i>
1916 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1917 unwind label %TestCleanup <i>; {i32}:retval set</i>
1918</pre>
1919</div>
1920
1921
1922<!-- _______________________________________________________________________ -->
1923
1924<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1925Instruction</a> </div>
1926
1927<div class="doc_text">
1928
1929<h5>Syntax:</h5>
1930<pre>
1931 unwind
1932</pre>
1933
1934<h5>Overview:</h5>
1935
1936<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1937at the first callee in the dynamic call stack which used an <a
1938href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1939primarily used to implement exception handling.</p>
1940
1941<h5>Semantics:</h5>
1942
1943<p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1944immediately halt. The dynamic call stack is then searched for the first <a
1945href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1946execution continues at the "exceptional" destination block specified by the
1947<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1948dynamic call chain, undefined behavior results.</p>
1949</div>
1950
1951<!-- _______________________________________________________________________ -->
1952
1953<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1954Instruction</a> </div>
1955
1956<div class="doc_text">
1957
1958<h5>Syntax:</h5>
1959<pre>
1960 unreachable
1961</pre>
1962
1963<h5>Overview:</h5>
1964
1965<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1966instruction is used to inform the optimizer that a particular portion of the
1967code is not reachable. This can be used to indicate that the code after a
1968no-return function cannot be reached, and other facts.</p>
1969
1970<h5>Semantics:</h5>
1971
1972<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1973</div>
1974
1975
1976
1977<!-- ======================================================================= -->
1978<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1979<div class="doc_text">
1980<p>Binary operators are used to do most of the computation in a
1981program. They require two operands, execute an operation on them, and
1982produce a single value. The operands might represent
1983multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
1984The result value of a binary operator is not
1985necessarily the same type as its operands.</p>
1986<p>There are several different binary operators:</p>
1987</div>
1988<!-- _______________________________________________________________________ -->
1989<div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1990Instruction</a> </div>
1991<div class="doc_text">
1992<h5>Syntax:</h5>
1993<pre> &lt;result&gt; = add &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
1994</pre>
1995<h5>Overview:</h5>
1996<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1997<h5>Arguments:</h5>
1998<p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1999 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
2000 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2001Both arguments must have identical types.</p>
2002<h5>Semantics:</h5>
2003<p>The value produced is the integer or floating point sum of the two
2004operands.</p>
2005<h5>Example:</h5>
2006<pre> &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2007</pre>
2008</div>
2009<!-- _______________________________________________________________________ -->
2010<div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
2011Instruction</a> </div>
2012<div class="doc_text">
2013<h5>Syntax:</h5>
2014<pre> &lt;result&gt; = sub &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2015</pre>
2016<h5>Overview:</h5>
2017<p>The '<tt>sub</tt>' instruction returns the difference of its two
2018operands.</p>
2019<p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
2020instruction present in most other intermediate representations.</p>
2021<h5>Arguments:</h5>
2022<p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
2023 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2024values.
2025This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2026Both arguments must have identical types.</p>
2027<h5>Semantics:</h5>
2028<p>The value produced is the integer or floating point difference of
2029the two operands.</p>
2030<h5>Example:</h5>
2031<pre>
2032 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2033 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2034</pre>
2035</div>
2036<!-- _______________________________________________________________________ -->
2037<div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
2038Instruction</a> </div>
2039<div class="doc_text">
2040<h5>Syntax:</h5>
2041<pre> &lt;result&gt; = mul &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2042</pre>
2043<h5>Overview:</h5>
2044<p>The '<tt>mul</tt>' instruction returns the product of its two
2045operands.</p>
2046<h5>Arguments:</h5>
2047<p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
2048 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2049values.
2050This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2051Both arguments must have identical types.</p>
2052<h5>Semantics:</h5>
2053<p>The value produced is the integer or floating point product of the
2054two operands.</p>
2055<p>Because the operands are the same width, the result of an integer
2056multiplication is the same whether the operands should be deemed unsigned or
2057signed.</p>
2058<h5>Example:</h5>
2059<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2060</pre>
2061</div>
2062<!-- _______________________________________________________________________ -->
2063<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2064</a></div>
2065<div class="doc_text">
2066<h5>Syntax:</h5>
2067<pre> &lt;result&gt; = udiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2068</pre>
2069<h5>Overview:</h5>
2070<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2071operands.</p>
2072<h5>Arguments:</h5>
2073<p>The two arguments to the '<tt>udiv</tt>' instruction must be
2074<a href="#t_integer">integer</a> values. Both arguments must have identical
2075types. This instruction can also take <a href="#t_vector">vector</a> versions
2076of the values in which case the elements must be integers.</p>
2077<h5>Semantics:</h5>
2078<p>The value produced is the unsigned integer quotient of the two operands. This
2079instruction always performs an unsigned division operation, regardless of
2080whether the arguments are unsigned or not.</p>
2081<h5>Example:</h5>
2082<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2083</pre>
2084</div>
2085<!-- _______________________________________________________________________ -->
2086<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2087</a> </div>
2088<div class="doc_text">
2089<h5>Syntax:</h5>
2090<pre> &lt;result&gt; = sdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2091</pre>
2092<h5>Overview:</h5>
2093<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2094operands.</p>
2095<h5>Arguments:</h5>
2096<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2097<a href="#t_integer">integer</a> values. Both arguments must have identical
2098types. This instruction can also take <a href="#t_vector">vector</a> versions
2099of the values in which case the elements must be integers.</p>
2100<h5>Semantics:</h5>
2101<p>The value produced is the signed integer quotient of the two operands. This
2102instruction always performs a signed division operation, regardless of whether
2103the arguments are signed or not.</p>
2104<h5>Example:</h5>
2105<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2106</pre>
2107</div>
2108<!-- _______________________________________________________________________ -->
2109<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2110Instruction</a> </div>
2111<div class="doc_text">
2112<h5>Syntax:</h5>
2113<pre> &lt;result&gt; = fdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2114</pre>
2115<h5>Overview:</h5>
2116<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2117operands.</p>
2118<h5>Arguments:</h5>
2119<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2120<a href="#t_floating">floating point</a> values. Both arguments must have
2121identical types. This instruction can also take <a href="#t_vector">vector</a>
2122versions of floating point values.</p>
2123<h5>Semantics:</h5>
2124<p>The value produced is the floating point quotient of the two operands.</p>
2125<h5>Example:</h5>
2126<pre> &lt;result&gt; = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2127</pre>
2128</div>
2129<!-- _______________________________________________________________________ -->
2130<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2131</div>
2132<div class="doc_text">
2133<h5>Syntax:</h5>
2134<pre> &lt;result&gt; = urem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2135</pre>
2136<h5>Overview:</h5>
2137<p>The '<tt>urem</tt>' instruction returns the remainder from the
2138unsigned division of its two arguments.</p>
2139<h5>Arguments:</h5>
2140<p>The two arguments to the '<tt>urem</tt>' instruction must be
2141<a href="#t_integer">integer</a> values. Both arguments must have identical
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2142types. This instruction can also take <a href="#t_vector">vector</a> versions
2143of the values in which case the elements must be integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2144<h5>Semantics:</h5>
2145<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2146This instruction always performs an unsigned division to get the remainder,
2147regardless of whether the arguments are unsigned or not.</p>
2148<h5>Example:</h5>
2149<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2150</pre>
2151
2152</div>
2153<!-- _______________________________________________________________________ -->
2154<div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2155Instruction</a> </div>
2156<div class="doc_text">
2157<h5>Syntax:</h5>
2158<pre> &lt;result&gt; = srem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2159</pre>
2160<h5>Overview:</h5>
2161<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2162signed division of its two operands. This instruction can also take
2163<a href="#t_vector">vector</a> versions of the values in which case
2164the elements must be integers.</p>
2165</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2166<h5>Arguments:</h5>
2167<p>The two arguments to the '<tt>srem</tt>' instruction must be
2168<a href="#t_integer">integer</a> values. Both arguments must have identical
2169types.</p>
2170<h5>Semantics:</h5>
2171<p>This instruction returns the <i>remainder</i> of a division (where the result
2172has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2173operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2174a value. For more information about the difference, see <a
2175 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2176Math Forum</a>. For a table of how this is implemented in various languages,
2177please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2178Wikipedia: modulo operation</a>.</p>
2179<h5>Example:</h5>
2180<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2181</pre>
2182
2183</div>
2184<!-- _______________________________________________________________________ -->
2185<div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2186Instruction</a> </div>
2187<div class="doc_text">
2188<h5>Syntax:</h5>
2189<pre> &lt;result&gt; = frem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2190</pre>
2191<h5>Overview:</h5>
2192<p>The '<tt>frem</tt>' instruction returns the remainder from the
2193division of its two operands.</p>
2194<h5>Arguments:</h5>
2195<p>The two arguments to the '<tt>frem</tt>' instruction must be
2196<a href="#t_floating">floating point</a> values. Both arguments must have
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2197identical types. This instruction can also take <a href="#t_vector">vector</a>
2198versions of floating point values.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2199<h5>Semantics:</h5>
2200<p>This instruction returns the <i>remainder</i> of a division.</p>
2201<h5>Example:</h5>
2202<pre> &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2203</pre>
2204</div>
2205
2206<!-- ======================================================================= -->
2207<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2208Operations</a> </div>
2209<div class="doc_text">
2210<p>Bitwise binary operators are used to do various forms of
2211bit-twiddling in a program. They are generally very efficient
2212instructions and can commonly be strength reduced from other
2213instructions. They require two operands, execute an operation on them,
2214and produce a single value. The resulting value of the bitwise binary
2215operators is always the same type as its first operand.</p>
2216</div>
2217
2218<!-- _______________________________________________________________________ -->
2219<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2220Instruction</a> </div>
2221<div class="doc_text">
2222<h5>Syntax:</h5>
2223<pre> &lt;result&gt; = shl &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2224</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2225
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2226<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2227
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2228<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2229the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2230
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2231<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2232
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2233<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2234 href="#t_integer">integer</a> type.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2235
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2236<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2237
2238<p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>. If
2239<tt>var2</tt> is (statically or dynamically) equal to or larger than the number
2240of bits in <tt>var1</tt>, the result is undefined.</p>
2241
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2242<h5>Example:</h5><pre>
2243 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2244 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2245 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2246 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2247</pre>
2248</div>
2249<!-- _______________________________________________________________________ -->
2250<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2251Instruction</a> </div>
2252<div class="doc_text">
2253<h5>Syntax:</h5>
2254<pre> &lt;result&gt; = lshr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2255</pre>
2256
2257<h5>Overview:</h5>
2258<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2259operand shifted to the right a specified number of bits with zero fill.</p>
2260
2261<h5>Arguments:</h5>
2262<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2263<a href="#t_integer">integer</a> type.</p>
2264
2265<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2266
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2267<p>This instruction always performs a logical shift right operation. The most
2268significant bits of the result will be filled with zero bits after the
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2269shift. If <tt>var2</tt> is (statically or dynamically) equal to or larger than
2270the number of bits in <tt>var1</tt>, the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2271
2272<h5>Example:</h5>
2273<pre>
2274 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2275 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2276 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2277 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2278 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2279</pre>
2280</div>
2281
2282<!-- _______________________________________________________________________ -->
2283<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2284Instruction</a> </div>
2285<div class="doc_text">
2286
2287<h5>Syntax:</h5>
2288<pre> &lt;result&gt; = ashr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2289</pre>
2290
2291<h5>Overview:</h5>
2292<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2293operand shifted to the right a specified number of bits with sign extension.</p>
2294
2295<h5>Arguments:</h5>
2296<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2297<a href="#t_integer">integer</a> type.</p>
2298
2299<h5>Semantics:</h5>
2300<p>This instruction always performs an arithmetic shift right operation,
2301The most significant bits of the result will be filled with the sign bit
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2302of <tt>var1</tt>. If <tt>var2</tt> is (statically or dynamically) equal to or
2303larger than the number of bits in <tt>var1</tt>, the result is undefined.
2304</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2305
2306<h5>Example:</h5>
2307<pre>
2308 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2309 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2310 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2311 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2312 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2313</pre>
2314</div>
2315
2316<!-- _______________________________________________________________________ -->
2317<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2318Instruction</a> </div>
2319<div class="doc_text">
2320<h5>Syntax:</h5>
2321<pre> &lt;result&gt; = and &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2322</pre>
2323<h5>Overview:</h5>
2324<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2325its two operands.</p>
2326<h5>Arguments:</h5>
2327<p>The two arguments to the '<tt>and</tt>' instruction must be <a
2328 href="#t_integer">integer</a> values. Both arguments must have
2329identical types.</p>
2330<h5>Semantics:</h5>
2331<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2332<p> </p>
2333<div style="align: center">
2334<table border="1" cellspacing="0" cellpadding="4">
2335 <tbody>
2336 <tr>
2337 <td>In0</td>
2338 <td>In1</td>
2339 <td>Out</td>
2340 </tr>
2341 <tr>
2342 <td>0</td>
2343 <td>0</td>
2344 <td>0</td>
2345 </tr>
2346 <tr>
2347 <td>0</td>
2348 <td>1</td>
2349 <td>0</td>
2350 </tr>
2351 <tr>
2352 <td>1</td>
2353 <td>0</td>
2354 <td>0</td>
2355 </tr>
2356 <tr>
2357 <td>1</td>
2358 <td>1</td>
2359 <td>1</td>
2360 </tr>
2361 </tbody>
2362</table>
2363</div>
2364<h5>Example:</h5>
2365<pre> &lt;result&gt; = and i32 4, %var <i>; yields {i32}:result = 4 &amp; %var</i>
2366 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2367 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2368</pre>
2369</div>
2370<!-- _______________________________________________________________________ -->
2371<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2372<div class="doc_text">
2373<h5>Syntax:</h5>
2374<pre> &lt;result&gt; = or &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2375</pre>
2376<h5>Overview:</h5>
2377<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2378or of its two operands.</p>
2379<h5>Arguments:</h5>
2380<p>The two arguments to the '<tt>or</tt>' instruction must be <a
2381 href="#t_integer">integer</a> values. Both arguments must have
2382identical types.</p>
2383<h5>Semantics:</h5>
2384<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2385<p> </p>
2386<div style="align: center">
2387<table border="1" cellspacing="0" cellpadding="4">
2388 <tbody>
2389 <tr>
2390 <td>In0</td>
2391 <td>In1</td>
2392 <td>Out</td>
2393 </tr>
2394 <tr>
2395 <td>0</td>
2396 <td>0</td>
2397 <td>0</td>
2398 </tr>
2399 <tr>
2400 <td>0</td>
2401 <td>1</td>
2402 <td>1</td>
2403 </tr>
2404 <tr>
2405 <td>1</td>
2406 <td>0</td>
2407 <td>1</td>
2408 </tr>
2409 <tr>
2410 <td>1</td>
2411 <td>1</td>
2412 <td>1</td>
2413 </tr>
2414 </tbody>
2415</table>
2416</div>
2417<h5>Example:</h5>
2418<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2419 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2420 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2421</pre>
2422</div>
2423<!-- _______________________________________________________________________ -->
2424<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2425Instruction</a> </div>
2426<div class="doc_text">
2427<h5>Syntax:</h5>
2428<pre> &lt;result&gt; = xor &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2429</pre>
2430<h5>Overview:</h5>
2431<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2432or of its two operands. The <tt>xor</tt> is used to implement the
2433"one's complement" operation, which is the "~" operator in C.</p>
2434<h5>Arguments:</h5>
2435<p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2436 href="#t_integer">integer</a> values. Both arguments must have
2437identical types.</p>
2438<h5>Semantics:</h5>
2439<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2440<p> </p>
2441<div style="align: center">
2442<table border="1" cellspacing="0" cellpadding="4">
2443 <tbody>
2444 <tr>
2445 <td>In0</td>
2446 <td>In1</td>
2447 <td>Out</td>
2448 </tr>
2449 <tr>
2450 <td>0</td>
2451 <td>0</td>
2452 <td>0</td>
2453 </tr>
2454 <tr>
2455 <td>0</td>
2456 <td>1</td>
2457 <td>1</td>
2458 </tr>
2459 <tr>
2460 <td>1</td>
2461 <td>0</td>
2462 <td>1</td>
2463 </tr>
2464 <tr>
2465 <td>1</td>
2466 <td>1</td>
2467 <td>0</td>
2468 </tr>
2469 </tbody>
2470</table>
2471</div>
2472<p> </p>
2473<h5>Example:</h5>
2474<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2475 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2476 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2477 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2478</pre>
2479</div>
2480
2481<!-- ======================================================================= -->
2482<div class="doc_subsection">
2483 <a name="vectorops">Vector Operations</a>
2484</div>
2485
2486<div class="doc_text">
2487
2488<p>LLVM supports several instructions to represent vector operations in a
2489target-independent manner. These instructions cover the element-access and
2490vector-specific operations needed to process vectors effectively. While LLVM
2491does directly support these vector operations, many sophisticated algorithms
2492will want to use target-specific intrinsics to take full advantage of a specific
2493target.</p>
2494
2495</div>
2496
2497<!-- _______________________________________________________________________ -->
2498<div class="doc_subsubsection">
2499 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2500</div>
2501
2502<div class="doc_text">
2503
2504<h5>Syntax:</h5>
2505
2506<pre>
2507 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
2508</pre>
2509
2510<h5>Overview:</h5>
2511
2512<p>
2513The '<tt>extractelement</tt>' instruction extracts a single scalar
2514element from a vector at a specified index.
2515</p>
2516
2517
2518<h5>Arguments:</h5>
2519
2520<p>
2521The first operand of an '<tt>extractelement</tt>' instruction is a
2522value of <a href="#t_vector">vector</a> type. The second operand is
2523an index indicating the position from which to extract the element.
2524The index may be a variable.</p>
2525
2526<h5>Semantics:</h5>
2527
2528<p>
2529The result is a scalar of the same type as the element type of
2530<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2531<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2532results are undefined.
2533</p>
2534
2535<h5>Example:</h5>
2536
2537<pre>
2538 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
2539</pre>
2540</div>
2541
2542
2543<!-- _______________________________________________________________________ -->
2544<div class="doc_subsubsection">
2545 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2546</div>
2547
2548<div class="doc_text">
2549
2550<h5>Syntax:</h5>
2551
2552<pre>
2553 &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>
2554</pre>
2555
2556<h5>Overview:</h5>
2557
2558<p>
2559The '<tt>insertelement</tt>' instruction inserts a scalar
2560element into a vector at a specified index.
2561</p>
2562
2563
2564<h5>Arguments:</h5>
2565
2566<p>
2567The first operand of an '<tt>insertelement</tt>' instruction is a
2568value of <a href="#t_vector">vector</a> type. The second operand is a
2569scalar value whose type must equal the element type of the first
2570operand. The third operand is an index indicating the position at
2571which to insert the value. The index may be a variable.</p>
2572
2573<h5>Semantics:</h5>
2574
2575<p>
2576The result is a vector of the same type as <tt>val</tt>. Its
2577element values are those of <tt>val</tt> except at position
2578<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2579exceeds the length of <tt>val</tt>, the results are undefined.
2580</p>
2581
2582<h5>Example:</h5>
2583
2584<pre>
2585 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
2586</pre>
2587</div>
2588
2589<!-- _______________________________________________________________________ -->
2590<div class="doc_subsubsection">
2591 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2592</div>
2593
2594<div class="doc_text">
2595
2596<h5>Syntax:</h5>
2597
2598<pre>
2599 &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>
2600</pre>
2601
2602<h5>Overview:</h5>
2603
2604<p>
2605The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2606from two input vectors, returning a vector of the same type.
2607</p>
2608
2609<h5>Arguments:</h5>
2610
2611<p>
2612The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2613with types that match each other and types that match the result of the
2614instruction. The third argument is a shuffle mask, which has the same number
2615of elements as the other vector type, but whose element type is always 'i32'.
2616</p>
2617
2618<p>
2619The shuffle mask operand is required to be a constant vector with either
2620constant integer or undef values.
2621</p>
2622
2623<h5>Semantics:</h5>
2624
2625<p>
2626The elements of the two input vectors are numbered from left to right across
2627both of the vectors. The shuffle mask operand specifies, for each element of
2628the result vector, which element of the two input registers the result element
2629gets. The element selector may be undef (meaning "don't care") and the second
2630operand may be undef if performing a shuffle from only one vector.
2631</p>
2632
2633<h5>Example:</h5>
2634
2635<pre>
2636 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
2637 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
2638 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
2639 &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.
2640</pre>
2641</div>
2642
2643
2644<!-- ======================================================================= -->
2645<div class="doc_subsection">
2646 <a name="memoryops">Memory Access and Addressing Operations</a>
2647</div>
2648
2649<div class="doc_text">
2650
2651<p>A key design point of an SSA-based representation is how it
2652represents memory. In LLVM, no memory locations are in SSA form, which
2653makes things very simple. This section describes how to read, write,
2654allocate, and free memory in LLVM.</p>
2655
2656</div>
2657
2658<!-- _______________________________________________________________________ -->
2659<div class="doc_subsubsection">
2660 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2661</div>
2662
2663<div class="doc_text">
2664
2665<h5>Syntax:</h5>
2666
2667<pre>
2668 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
2669</pre>
2670
2671<h5>Overview:</h5>
2672
2673<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]2674heap and returns a pointer to it. The object is always allocated in the generic
2675address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2676
2677<h5>Arguments:</h5>
2678
2679<p>The '<tt>malloc</tt>' instruction allocates
2680<tt>sizeof(&lt;type&gt;)*NumElements</tt>
2681bytes of memory from the operating system and returns a pointer of the
2682appropriate type to the program. If "NumElements" is specified, it is the
2683number of elements allocated. If an alignment is specified, the value result
2684of the allocation is guaranteed to be aligned to at least that boundary. If
2685not specified, or if zero, the target can choose to align the allocation on any
2686convenient boundary.</p>
2687
2688<p>'<tt>type</tt>' must be a sized type.</p>
2689
2690<h5>Semantics:</h5>
2691
2692<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2693a pointer is returned.</p>
2694
2695<h5>Example:</h5>
2696
2697<pre>
2698 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2699
2700 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2701 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2702 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2703 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2704 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2705</pre>
2706</div>
2707
2708<!-- _______________________________________________________________________ -->
2709<div class="doc_subsubsection">
2710 <a name="i_free">'<tt>free</tt>' Instruction</a>
2711</div>
2712
2713<div class="doc_text">
2714
2715<h5>Syntax:</h5>
2716
2717<pre>
2718 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
2719</pre>
2720
2721<h5>Overview:</h5>
2722
2723<p>The '<tt>free</tt>' instruction returns memory back to the unused
2724memory heap to be reallocated in the future.</p>
2725
2726<h5>Arguments:</h5>
2727
2728<p>'<tt>value</tt>' shall be a pointer value that points to a value
2729that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2730instruction.</p>
2731
2732<h5>Semantics:</h5>
2733
2734<p>Access to the memory pointed to by the pointer is no longer defined
2735after this instruction executes.</p>
2736
2737<h5>Example:</h5>
2738
2739<pre>
2740 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2741 free [4 x i8]* %array
2742</pre>
2743</div>
2744
2745<!-- _______________________________________________________________________ -->
2746<div class="doc_subsubsection">
2747 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2748</div>
2749
2750<div class="doc_text">
2751
2752<h5>Syntax:</h5>
2753
2754<pre>
2755 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
2756</pre>
2757
2758<h5>Overview:</h5>
2759
2760<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
2761currently executing function, to be automatically released when this function
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]2762returns to its caller. The object is always allocated in the generic address
2763space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2764
2765<h5>Arguments:</h5>
2766
2767<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
2768bytes of memory on the runtime stack, returning a pointer of the
2769appropriate type to the program. If "NumElements" is specified, it is the
2770number of elements allocated. If an alignment is specified, the value result
2771of the allocation is guaranteed to be aligned to at least that boundary. If
2772not specified, or if zero, the target can choose to align the allocation on any
2773convenient boundary.</p>
2774
2775<p>'<tt>type</tt>' may be any sized type.</p>
2776
2777<h5>Semantics:</h5>
2778
2779<p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2780memory is automatically released when the function returns. The '<tt>alloca</tt>'
2781instruction is commonly used to represent automatic variables that must
2782have an address available. When the function returns (either with the <tt><a
2783 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2784instructions), the memory is reclaimed.</p>
2785
2786<h5>Example:</h5>
2787
2788<pre>
2789 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2790 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2791 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2792 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2793</pre>
2794</div>
2795
2796<!-- _______________________________________________________________________ -->
2797<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2798Instruction</a> </div>
2799<div class="doc_text">
2800<h5>Syntax:</h5>
2801<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>
2802<h5>Overview:</h5>
2803<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2804<h5>Arguments:</h5>
2805<p>The argument to the '<tt>load</tt>' instruction specifies the memory
2806address from which to load. The pointer must point to a <a
2807 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2808marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2809the number or order of execution of this <tt>load</tt> with other
2810volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2811instructions. </p>
2812<h5>Semantics:</h5>
2813<p>The location of memory pointed to is loaded.</p>
2814<h5>Examples:</h5>
2815<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2816 <a
2817 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2818 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2819</pre>
2820</div>
2821<!-- _______________________________________________________________________ -->
2822<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2823Instruction</a> </div>
2824<div class="doc_text">
2825<h5>Syntax:</h5>
2826<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
2827 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
2828</pre>
2829<h5>Overview:</h5>
2830<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2831<h5>Arguments:</h5>
2832<p>There are two arguments to the '<tt>store</tt>' instruction: a value
2833to store and an address at which to store it. The type of the '<tt>&lt;pointer&gt;</tt>'
2834operand must be a pointer to the type of the '<tt>&lt;value&gt;</tt>'
2835operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2836optimizer is not allowed to modify the number or order of execution of
2837this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2838 href="#i_store">store</a></tt> instructions.</p>
2839<h5>Semantics:</h5>
2840<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
2841at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
2842<h5>Example:</h5>
2843<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]2844 store i32 3, i32* %ptr <i>; yields {void}</i>
2845 %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]2846</pre>
2847</div>
2848
2849<!-- _______________________________________________________________________ -->
2850<div class="doc_subsubsection">
2851 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2852</div>
2853
2854<div class="doc_text">
2855<h5>Syntax:</h5>
2856<pre>
2857 &lt;result&gt; = getelementptr &lt;ty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
2858</pre>
2859
2860<h5>Overview:</h5>
2861
2862<p>
2863The '<tt>getelementptr</tt>' instruction is used to get the address of a
2864subelement of an aggregate data structure.</p>
2865
2866<h5>Arguments:</h5>
2867
2868<p>This instruction takes a list of integer operands that indicate what
2869elements of the aggregate object to index to. The actual types of the arguments
2870provided depend on the type of the first pointer argument. The
2871'<tt>getelementptr</tt>' instruction is used to index down through the type
2872levels of a structure or to a specific index in an array. When indexing into a
2873structure, only <tt>i32</tt> integer constants are allowed. When indexing
2874into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2875be sign extended to 64-bit values.</p>
2876
2877<p>For example, let's consider a C code fragment and how it gets
2878compiled to LLVM:</p>
2879
2880<div class="doc_code">
2881<pre>
2882struct RT {
2883 char A;
2884 int B[10][20];
2885 char C;
2886};
2887struct ST {
2888 int X;
2889 double Y;
2890 struct RT Z;
2891};
2892
2893int *foo(struct ST *s) {
2894 return &amp;s[1].Z.B[5][13];
2895}
2896</pre>
2897</div>
2898
2899<p>The LLVM code generated by the GCC frontend is:</p>
2900
2901<div class="doc_code">
2902<pre>
2903%RT = type { i8 , [10 x [20 x i32]], i8 }
2904%ST = type { i32, double, %RT }
2905
2906define i32* %foo(%ST* %s) {
2907entry:
2908 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
2909 ret i32* %reg
2910}
2911</pre>
2912</div>
2913
2914<h5>Semantics:</h5>
2915
2916<p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2917on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2918and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
2919<a href="#t_integer">integer</a> type but the value will always be sign extended
2920to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
2921<b>constants</b>.</p>
2922
2923<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2924type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
2925}</tt>' type, a structure. The second index indexes into the third element of
2926the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
2927i8 }</tt>' type, another structure. The third index indexes into the second
2928element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
2929array. The two dimensions of the array are subscripted into, yielding an
2930'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2931to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
2932
2933<p>Note that it is perfectly legal to index partially through a
2934structure, returning a pointer to an inner element. Because of this,
2935the LLVM code for the given testcase is equivalent to:</p>
2936
2937<pre>
2938 define i32* %foo(%ST* %s) {
2939 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
2940 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
2941 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
2942 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
2943 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
2944 ret i32* %t5
2945 }
2946</pre>
2947
2948<p>Note that it is undefined to access an array out of bounds: array and
2949pointer indexes must always be within the defined bounds of the array type.
2950The one exception for this rules is zero length arrays. These arrays are
2951defined to be accessible as variable length arrays, which requires access
2952beyond the zero'th element.</p>
2953
2954<p>The getelementptr instruction is often confusing. For some more insight
2955into how it works, see <a href="GetElementPtr.html">the getelementptr
2956FAQ</a>.</p>
2957
2958<h5>Example:</h5>
2959
2960<pre>
2961 <i>; yields [12 x i8]*:aptr</i>
2962 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
2963</pre>
2964</div>
2965
2966<!-- ======================================================================= -->
2967<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2968</div>
2969<div class="doc_text">
2970<p>The instructions in this category are the conversion instructions (casting)
2971which all take a single operand and a type. They perform various bit conversions
2972on the operand.</p>
2973</div>
2974
2975<!-- _______________________________________________________________________ -->
2976<div class="doc_subsubsection">
2977 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2978</div>
2979<div class="doc_text">
2980
2981<h5>Syntax:</h5>
2982<pre>
2983 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
2984</pre>
2985
2986<h5>Overview:</h5>
2987<p>
2988The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2989</p>
2990
2991<h5>Arguments:</h5>
2992<p>
2993The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2994be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2995and type of the result, which must be an <a href="#t_integer">integer</a>
2996type. The bit size of <tt>value</tt> must be larger than the bit size of
2997<tt>ty2</tt>. Equal sized types are not allowed.</p>
2998
2999<h5>Semantics:</h5>
3000<p>
3001The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3002and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3003larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3004It will always truncate bits.</p>
3005
3006<h5>Example:</h5>
3007<pre>
3008 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3009 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3010 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3011</pre>
3012</div>
3013
3014<!-- _______________________________________________________________________ -->
3015<div class="doc_subsubsection">
3016 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3017</div>
3018<div class="doc_text">
3019
3020<h5>Syntax:</h5>
3021<pre>
3022 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3023</pre>
3024
3025<h5>Overview:</h5>
3026<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3027<tt>ty2</tt>.</p>
3028
3029
3030<h5>Arguments:</h5>
3031<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3032<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3033also be of <a href="#t_integer">integer</a> type. The bit size of the
3034<tt>value</tt> must be smaller than the bit size of the destination type,
3035<tt>ty2</tt>.</p>
3036
3037<h5>Semantics:</h5>
3038<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3039bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3040
3041<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3042
3043<h5>Example:</h5>
3044<pre>
3045 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3046 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3047</pre>
3048</div>
3049
3050<!-- _______________________________________________________________________ -->
3051<div class="doc_subsubsection">
3052 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3053</div>
3054<div class="doc_text">
3055
3056<h5>Syntax:</h5>
3057<pre>
3058 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3059</pre>
3060
3061<h5>Overview:</h5>
3062<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3063
3064<h5>Arguments:</h5>
3065<p>
3066The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3067<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3068also be of <a href="#t_integer">integer</a> type. The bit size of the
3069<tt>value</tt> must be smaller than the bit size of the destination type,
3070<tt>ty2</tt>.</p>
3071
3072<h5>Semantics:</h5>
3073<p>
3074The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3075bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3076the type <tt>ty2</tt>.</p>
3077
3078<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3079
3080<h5>Example:</h5>
3081<pre>
3082 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3083 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3084</pre>
3085</div>
3086
3087<!-- _______________________________________________________________________ -->
3088<div class="doc_subsubsection">
3089 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3090</div>
3091
3092<div class="doc_text">
3093
3094<h5>Syntax:</h5>
3095
3096<pre>
3097 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3098</pre>
3099
3100<h5>Overview:</h5>
3101<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3102<tt>ty2</tt>.</p>
3103
3104
3105<h5>Arguments:</h5>
3106<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3107 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3108cast it to. The size of <tt>value</tt> must be larger than the size of
3109<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3110<i>no-op cast</i>.</p>
3111
3112<h5>Semantics:</h5>
3113<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3114<a href="#t_floating">floating point</a> type to a smaller
3115<a href="#t_floating">floating point</a> type. If the value cannot fit within
3116the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3117
3118<h5>Example:</h5>
3119<pre>
3120 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3121 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3122</pre>
3123</div>
3124
3125<!-- _______________________________________________________________________ -->
3126<div class="doc_subsubsection">
3127 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3128</div>
3129<div class="doc_text">
3130
3131<h5>Syntax:</h5>
3132<pre>
3133 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3134</pre>
3135
3136<h5>Overview:</h5>
3137<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3138floating point value.</p>
3139
3140<h5>Arguments:</h5>
3141<p>The '<tt>fpext</tt>' instruction takes a
3142<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3143and a <a href="#t_floating">floating point</a> type to cast it to. The source
3144type must be smaller than the destination type.</p>
3145
3146<h5>Semantics:</h5>
3147<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3148<a href="#t_floating">floating point</a> type to a larger
3149<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3150used to make a <i>no-op cast</i> because it always changes bits. Use
3151<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3152
3153<h5>Example:</h5>
3154<pre>
3155 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3156 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3157</pre>
3158</div>
3159
3160<!-- _______________________________________________________________________ -->
3161<div class="doc_subsubsection">
3162 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3163</div>
3164<div class="doc_text">
3165
3166<h5>Syntax:</h5>
3167<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3168 &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]3169</pre>
3170
3171<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3172<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3173unsigned integer equivalent of type <tt>ty2</tt>.
3174</p>
3175
3176<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3177<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]3178scalar or vector <a href="#t_floating">floating point</a> value, and a type
3179to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3180type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3181vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3182
3183<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3184<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3185<a href="#t_floating">floating point</a> operand into the nearest (rounding
3186towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3187the results are undefined.</p>
3188
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3189<h5>Example:</h5>
3190<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3191 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3192 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3193 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3194</pre>
3195</div>
3196
3197<!-- _______________________________________________________________________ -->
3198<div class="doc_subsubsection">
3199 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3200</div>
3201<div class="doc_text">
3202
3203<h5>Syntax:</h5>
3204<pre>
3205 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3206</pre>
3207
3208<h5>Overview:</h5>
3209<p>The '<tt>fptosi</tt>' instruction converts
3210<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3211</p>
3212
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3213<h5>Arguments:</h5>
3214<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]3215scalar or vector <a href="#t_floating">floating point</a> value, and a type
3216to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3217type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3218vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3219
3220<h5>Semantics:</h5>
3221<p>The '<tt>fptosi</tt>' instruction converts its
3222<a href="#t_floating">floating point</a> operand into the nearest (rounding
3223towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3224the results are undefined.</p>
3225
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3226<h5>Example:</h5>
3227<pre>
3228 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3229 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3230 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3231</pre>
3232</div>
3233
3234<!-- _______________________________________________________________________ -->
3235<div class="doc_subsubsection">
3236 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3237</div>
3238<div class="doc_text">
3239
3240<h5>Syntax:</h5>
3241<pre>
3242 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3243</pre>
3244
3245<h5>Overview:</h5>
3246<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3247integer and converts that value to the <tt>ty2</tt> type.</p>
3248
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3249<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3250<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3251scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3252to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3253type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3254floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3255
3256<h5>Semantics:</h5>
3257<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3258integer quantity and converts it to the corresponding floating point value. If
3259the value cannot fit in the floating point value, the results are undefined.</p>
3260
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3261<h5>Example:</h5>
3262<pre>
3263 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3264 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3265</pre>
3266</div>
3267
3268<!-- _______________________________________________________________________ -->
3269<div class="doc_subsubsection">
3270 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3271</div>
3272<div class="doc_text">
3273
3274<h5>Syntax:</h5>
3275<pre>
3276 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3277</pre>
3278
3279<h5>Overview:</h5>
3280<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3281integer and converts that value to the <tt>ty2</tt> type.</p>
3282
3283<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3284<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3285scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3286to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3287type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3288floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3289
3290<h5>Semantics:</h5>
3291<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3292integer quantity and converts it to the corresponding floating point value. If
3293the value cannot fit in the floating point value, the results are undefined.</p>
3294
3295<h5>Example:</h5>
3296<pre>
3297 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3298 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3299</pre>
3300</div>
3301
3302<!-- _______________________________________________________________________ -->
3303<div class="doc_subsubsection">
3304 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3305</div>
3306<div class="doc_text">
3307
3308<h5>Syntax:</h5>
3309<pre>
3310 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3311</pre>
3312
3313<h5>Overview:</h5>
3314<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3315the integer type <tt>ty2</tt>.</p>
3316
3317<h5>Arguments:</h5>
3318<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3319must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3320<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3321
3322<h5>Semantics:</h5>
3323<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3324<tt>ty2</tt> by interpreting the pointer value as an integer and either
3325truncating or zero extending that value to the size of the integer type. If
3326<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3327<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3328are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3329change.</p>
3330
3331<h5>Example:</h5>
3332<pre>
3333 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3334 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3335</pre>
3336</div>
3337
3338<!-- _______________________________________________________________________ -->
3339<div class="doc_subsubsection">
3340 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3341</div>
3342<div class="doc_text">
3343
3344<h5>Syntax:</h5>
3345<pre>
3346 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3347</pre>
3348
3349<h5>Overview:</h5>
3350<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3351a pointer type, <tt>ty2</tt>.</p>
3352
3353<h5>Arguments:</h5>
3354<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3355value to cast, and a type to cast it to, which must be a
3356<a href="#t_pointer">pointer</a> type.
3357
3358<h5>Semantics:</h5>
3359<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3360<tt>ty2</tt> by applying either a zero extension or a truncation depending on
3361the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3362size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3363the size of a pointer then a zero extension is done. If they are the same size,
3364nothing is done (<i>no-op cast</i>).</p>
3365
3366<h5>Example:</h5>
3367<pre>
3368 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3369 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3370 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3371</pre>
3372</div>
3373
3374<!-- _______________________________________________________________________ -->
3375<div class="doc_subsubsection">
3376 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3377</div>
3378<div class="doc_text">
3379
3380<h5>Syntax:</h5>
3381<pre>
3382 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3383</pre>
3384
3385<h5>Overview:</h5>
3386<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3387<tt>ty2</tt> without changing any bits.</p>
3388
3389<h5>Arguments:</h5>
3390<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3391a first class value, and a type to cast it to, which must also be a <a
3392 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3393and the destination type, <tt>ty2</tt>, must be identical. If the source
3394type is a pointer, the destination type must also be a pointer.</p>
3395
3396<h5>Semantics:</h5>
3397<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3398<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3399this conversion. The conversion is done as if the <tt>value</tt> had been
3400stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3401converted to other pointer types with this instruction. To convert pointers to
3402other types, use the <a href="#i_inttoptr">inttoptr</a> or
3403<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3404
3405<h5>Example:</h5>
3406<pre>
3407 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3408 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3409 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3410</pre>
3411</div>
3412
3413<!-- ======================================================================= -->
3414<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3415<div class="doc_text">
3416<p>The instructions in this category are the "miscellaneous"
3417instructions, which defy better classification.</p>
3418</div>
3419
3420<!-- _______________________________________________________________________ -->
3421<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3422</div>
3423<div class="doc_text">
3424<h5>Syntax:</h5>
3425<pre> &lt;result&gt; = icmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3426</pre>
3427<h5>Overview:</h5>
3428<p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3429of its two integer operands.</p>
3430<h5>Arguments:</h5>
3431<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3432the condition code indicating the kind of comparison to perform. It is not
3433a value, just a keyword. The possible condition code are:
3434<ol>
3435 <li><tt>eq</tt>: equal</li>
3436 <li><tt>ne</tt>: not equal </li>
3437 <li><tt>ugt</tt>: unsigned greater than</li>
3438 <li><tt>uge</tt>: unsigned greater or equal</li>
3439 <li><tt>ult</tt>: unsigned less than</li>
3440 <li><tt>ule</tt>: unsigned less or equal</li>
3441 <li><tt>sgt</tt>: signed greater than</li>
3442 <li><tt>sge</tt>: signed greater or equal</li>
3443 <li><tt>slt</tt>: signed less than</li>
3444 <li><tt>sle</tt>: signed less or equal</li>
3445</ol>
3446<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3447<a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3448<h5>Semantics:</h5>
3449<p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3450the condition code given as <tt>cond</tt>. The comparison performed always
3451yields a <a href="#t_primitive">i1</a> result, as follows:
3452<ol>
3453 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3454 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3455 </li>
3456 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3457 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3458 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3459 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3460 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3461 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3462 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3463 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3464 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3465 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3466 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3467 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3468 <li><tt>sge</tt>: interprets the operands as signed values and yields
3469 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3470 <li><tt>slt</tt>: interprets the operands as signed values and yields
3471 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3472 <li><tt>sle</tt>: interprets the operands as signed values and yields
3473 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3474</ol>
3475<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3476values are compared as if they were integers.</p>
3477
3478<h5>Example:</h5>
3479<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
3480 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
3481 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
3482 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3483 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3484 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
3485</pre>
3486</div>
3487
3488<!-- _______________________________________________________________________ -->
3489<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3490</div>
3491<div class="doc_text">
3492<h5>Syntax:</h5>
3493<pre> &lt;result&gt; = fcmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3494</pre>
3495<h5>Overview:</h5>
3496<p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3497of its floating point operands.</p>
3498<h5>Arguments:</h5>
3499<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3500the condition code indicating the kind of comparison to perform. It is not
3501a value, just a keyword. The possible condition code are:
3502<ol>
3503 <li><tt>false</tt>: no comparison, always returns false</li>
3504 <li><tt>oeq</tt>: ordered and equal</li>
3505 <li><tt>ogt</tt>: ordered and greater than </li>
3506 <li><tt>oge</tt>: ordered and greater than or equal</li>
3507 <li><tt>olt</tt>: ordered and less than </li>
3508 <li><tt>ole</tt>: ordered and less than or equal</li>
3509 <li><tt>one</tt>: ordered and not equal</li>
3510 <li><tt>ord</tt>: ordered (no nans)</li>
3511 <li><tt>ueq</tt>: unordered or equal</li>
3512 <li><tt>ugt</tt>: unordered or greater than </li>
3513 <li><tt>uge</tt>: unordered or greater than or equal</li>
3514 <li><tt>ult</tt>: unordered or less than </li>
3515 <li><tt>ule</tt>: unordered or less than or equal</li>
3516 <li><tt>une</tt>: unordered or not equal</li>
3517 <li><tt>uno</tt>: unordered (either nans)</li>
3518 <li><tt>true</tt>: no comparison, always returns true</li>
3519</ol>
3520<p><i>Ordered</i> means that neither operand is a QNAN while
3521<i>unordered</i> means that either operand may be a QNAN.</p>
3522<p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3523<a href="#t_floating">floating point</a> typed. They must have identical
3524types.</p>
3525<h5>Semantics:</h5>
3526<p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3527the condition code given as <tt>cond</tt>. The comparison performed always
3528yields a <a href="#t_primitive">i1</a> result, as follows:
3529<ol>
3530 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3531 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3532 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3533 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3534 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3535 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3536 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3537 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3538 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3539 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3540 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3541 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3542 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3543 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3544 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3545 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3546 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3547 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3548 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3549 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3550 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3551 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3552 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3553 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3554 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3555 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3556 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3557 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3558</ol>
3559
3560<h5>Example:</h5>
3561<pre> &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3562 &lt;result&gt; = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3563 &lt;result&gt; = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3564 &lt;result&gt; = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3565</pre>
3566</div>
3567
3568<!-- _______________________________________________________________________ -->
3569<div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3570Instruction</a> </div>
3571<div class="doc_text">
3572<h5>Syntax:</h5>
3573<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
3574<h5>Overview:</h5>
3575<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
3576the SSA graph representing the function.</p>
3577<h5>Arguments:</h5>
3578<p>The type of the incoming values is specified with the first type
3579field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3580as arguments, with one pair for each predecessor basic block of the
3581current block. Only values of <a href="#t_firstclass">first class</a>
3582type may be used as the value arguments to the PHI node. Only labels
3583may be used as the label arguments.</p>
3584<p>There must be no non-phi instructions between the start of a basic
3585block and the PHI instructions: i.e. PHI instructions must be first in
3586a basic block.</p>
3587<h5>Semantics:</h5>
3588<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3589specified by the pair corresponding to the predecessor basic block that executed
3590just prior to the current block.</p>
3591<h5>Example:</h5>
3592<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>
3593</div>
3594
3595<!-- _______________________________________________________________________ -->
3596<div class="doc_subsubsection">
3597 <a name="i_select">'<tt>select</tt>' Instruction</a>
3598</div>
3599
3600<div class="doc_text">
3601
3602<h5>Syntax:</h5>
3603
3604<pre>
3605 &lt;result&gt; = select i1 &lt;cond&gt;, &lt;ty&gt; &lt;val1&gt;, &lt;ty&gt; &lt;val2&gt; <i>; yields ty</i>
3606</pre>
3607
3608<h5>Overview:</h5>
3609
3610<p>
3611The '<tt>select</tt>' instruction is used to choose one value based on a
3612condition, without branching.
3613</p>
3614
3615
3616<h5>Arguments:</h5>
3617
3618<p>
3619The '<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.
3620</p>
3621
3622<h5>Semantics:</h5>
3623
3624<p>
3625If the boolean condition evaluates to true, the instruction returns the first
3626value argument; otherwise, it returns the second value argument.
3627</p>
3628
3629<h5>Example:</h5>
3630
3631<pre>
3632 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3633</pre>
3634</div>
3635
3636
3637<!-- _______________________________________________________________________ -->
3638<div class="doc_subsubsection">
3639 <a name="i_call">'<tt>call</tt>' Instruction</a>
3640</div>
3641
3642<div class="doc_text">
3643
3644<h5>Syntax:</h5>
3645<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3646 &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]3647</pre>
3648
3649<h5>Overview:</h5>
3650
3651<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3652
3653<h5>Arguments:</h5>
3654
3655<p>This instruction requires several arguments:</p>
3656
3657<ol>
3658 <li>
3659 <p>The optional "tail" marker indicates whether the callee function accesses
3660 any allocas or varargs in the caller. If the "tail" marker is present, the
3661 function call is eligible for tail call optimization. Note that calls may
3662 be marked "tail" even if they do not occur before a <a
3663 href="#i_ret"><tt>ret</tt></a> instruction.
3664 </li>
3665 <li>
3666 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3667 convention</a> the call should use. If none is specified, the call defaults
3668 to using C calling conventions.
3669 </li>
3670 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3671 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
3672 the type of the return value. Functions that return no value are marked
3673 <tt><a href="#t_void">void</a></tt>.</p>
3674 </li>
3675 <li>
3676 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
3677 value being invoked. The argument types must match the types implied by
3678 this signature. This type can be omitted if the function is not varargs
3679 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3680 </li>
3681 <li>
3682 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3683 be invoked. In most cases, this is a direct function invocation, but
3684 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3685 to function value.</p>
3686 </li>
3687 <li>
3688 <p>'<tt>function args</tt>': argument list whose types match the
3689 function signature argument types. All arguments must be of
3690 <a href="#t_firstclass">first class</a> type. If the function signature
3691 indicates the function accepts a variable number of arguments, the extra
3692 arguments can be specified.</p>
3693 </li>
3694</ol>
3695
3696<h5>Semantics:</h5>
3697
3698<p>The '<tt>call</tt>' instruction is used to cause control flow to
3699transfer to a specified function, with its incoming arguments bound to
3700the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3701instruction in the called function, control flow continues with the
3702instruction after the function call, and the return value of the
3703function is bound to the result argument. This is a simpler case of
3704the <a href="#i_invoke">invoke</a> instruction.</p>
3705
3706<h5>Example:</h5>
3707
3708<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3709 %retval = call i32 @test(i32 %argc)
3710 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42);
3711 %X = tail call i32 @foo()
3712 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo()
3713 %Z = call void %foo(i8 97 signext)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3714</pre>
3715
3716</div>
3717
3718<!-- _______________________________________________________________________ -->
3719<div class="doc_subsubsection">
3720 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3721</div>
3722
3723<div class="doc_text">
3724
3725<h5>Syntax:</h5>
3726
3727<pre>
3728 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
3729</pre>
3730
3731<h5>Overview:</h5>
3732
3733<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3734the "variable argument" area of a function call. It is used to implement the
3735<tt>va_arg</tt> macro in C.</p>
3736
3737<h5>Arguments:</h5>
3738
3739<p>This instruction takes a <tt>va_list*</tt> value and the type of
3740the argument. It returns a value of the specified argument type and
3741increments the <tt>va_list</tt> to point to the next argument. The
3742actual type of <tt>va_list</tt> is target specific.</p>
3743
3744<h5>Semantics:</h5>
3745
3746<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3747type from the specified <tt>va_list</tt> and causes the
3748<tt>va_list</tt> to point to the next argument. For more information,
3749see the variable argument handling <a href="#int_varargs">Intrinsic
3750Functions</a>.</p>
3751
3752<p>It is legal for this instruction to be called in a function which does not
3753take a variable number of arguments, for example, the <tt>vfprintf</tt>
3754function.</p>
3755
3756<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3757href="#intrinsics">intrinsic function</a> because it takes a type as an
3758argument.</p>
3759
3760<h5>Example:</h5>
3761
3762<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3763
3764</div>
3765
3766<!-- *********************************************************************** -->
3767<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3768<!-- *********************************************************************** -->
3769
3770<div class="doc_text">
3771
3772<p>LLVM supports the notion of an "intrinsic function". These functions have
3773well known names and semantics and are required to follow certain restrictions.
3774Overall, these intrinsics represent an extension mechanism for the LLVM
3775language that does not require changing all of the transformations in LLVM when
3776adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
3777
3778<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3779prefix is reserved in LLVM for intrinsic names; thus, function names may not
3780begin with this prefix. Intrinsic functions must always be external functions:
3781you cannot define the body of intrinsic functions. Intrinsic functions may
3782only be used in call or invoke instructions: it is illegal to take the address
3783of an intrinsic function. Additionally, because intrinsic functions are part
3784of the LLVM language, it is required if any are added that they be documented
3785here.</p>
3786
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]3787<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
3788a family of functions that perform the same operation but on different data
3789types. Because LLVM can represent over 8 million different integer types,
3790overloading is used commonly to allow an intrinsic function to operate on any
3791integer type. One or more of the argument types or the result type can be
3792overloaded to accept any integer type. Argument types may also be defined as
3793exactly matching a previous argument's type or the result type. This allows an
3794intrinsic function which accepts multiple arguments, but needs all of them to
3795be of the same type, to only be overloaded with respect to a single argument or
3796the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3797
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]3798<p>Overloaded intrinsics will have the names of its overloaded argument types
3799encoded into its function name, each preceded by a period. Only those types
3800which are overloaded result in a name suffix. Arguments whose type is matched
3801against another type do not. For example, the <tt>llvm.ctpop</tt> function can
3802take an integer of any width and returns an integer of exactly the same integer
3803width. This leads to a family of functions such as
3804<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
3805Only one type, the return type, is overloaded, and only one type suffix is
3806required. Because the argument's type is matched against the return type, it
3807does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3808
3809<p>To learn how to add an intrinsic function, please see the
3810<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3811</p>
3812
3813</div>
3814
3815<!-- ======================================================================= -->
3816<div class="doc_subsection">
3817 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3818</div>
3819
3820<div class="doc_text">
3821
3822<p>Variable argument support is defined in LLVM with the <a
3823 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3824intrinsic functions. These functions are related to the similarly
3825named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
3826
3827<p>All of these functions operate on arguments that use a
3828target-specific value type "<tt>va_list</tt>". The LLVM assembly
3829language reference manual does not define what this type is, so all
3830transformations should be prepared to handle these functions regardless of
3831the type used.</p>
3832
3833<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3834instruction and the variable argument handling intrinsic functions are
3835used.</p>
3836
3837<div class="doc_code">
3838<pre>
3839define i32 @test(i32 %X, ...) {
3840 ; Initialize variable argument processing
3841 %ap = alloca i8*
3842 %ap2 = bitcast i8** %ap to i8*
3843 call void @llvm.va_start(i8* %ap2)
3844
3845 ; Read a single integer argument
3846 %tmp = va_arg i8** %ap, i32
3847
3848 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3849 %aq = alloca i8*
3850 %aq2 = bitcast i8** %aq to i8*
3851 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
3852 call void @llvm.va_end(i8* %aq2)
3853
3854 ; Stop processing of arguments.
3855 call void @llvm.va_end(i8* %ap2)
3856 ret i32 %tmp
3857}
3858
3859declare void @llvm.va_start(i8*)
3860declare void @llvm.va_copy(i8*, i8*)
3861declare void @llvm.va_end(i8*)
3862</pre>
3863</div>
3864
3865</div>
3866
3867<!-- _______________________________________________________________________ -->
3868<div class="doc_subsubsection">
3869 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3870</div>
3871
3872
3873<div class="doc_text">
3874<h5>Syntax:</h5>
3875<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
3876<h5>Overview:</h5>
3877<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3878<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
3879href="#i_va_arg">va_arg</a></tt>.</p>
3880
3881<h5>Arguments:</h5>
3882
3883<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3884
3885<h5>Semantics:</h5>
3886
3887<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3888macro available in C. In a target-dependent way, it initializes the
3889<tt>va_list</tt> element to which the argument points, so that the next call to
3890<tt>va_arg</tt> will produce the first variable argument passed to the function.
3891Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3892last argument of the function as the compiler can figure that out.</p>
3893
3894</div>
3895
3896<!-- _______________________________________________________________________ -->
3897<div class="doc_subsubsection">
3898 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3899</div>
3900
3901<div class="doc_text">
3902<h5>Syntax:</h5>
3903<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
3904<h5>Overview:</h5>
3905
3906<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
3907which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
3908or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3909
3910<h5>Arguments:</h5>
3911
3912<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
3913
3914<h5>Semantics:</h5>
3915
3916<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3917macro available in C. In a target-dependent way, it destroys the
3918<tt>va_list</tt> element to which the argument points. Calls to <a
3919href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
3920<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
3921<tt>llvm.va_end</tt>.</p>
3922
3923</div>
3924
3925<!-- _______________________________________________________________________ -->
3926<div class="doc_subsubsection">
3927 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3928</div>
3929
3930<div class="doc_text">
3931
3932<h5>Syntax:</h5>
3933
3934<pre>
3935 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
3936</pre>
3937
3938<h5>Overview:</h5>
3939
3940<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
3941from the source argument list to the destination argument list.</p>
3942
3943<h5>Arguments:</h5>
3944
3945<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3946The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3947
3948
3949<h5>Semantics:</h5>
3950
3951<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
3952macro available in C. In a target-dependent way, it copies the source
3953<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
3954intrinsic is necessary because the <tt><a href="#int_va_start">
3955llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
3956example, memory allocation.</p>
3957
3958</div>
3959
3960<!-- ======================================================================= -->
3961<div class="doc_subsection">
3962 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3963</div>
3964
3965<div class="doc_text">
3966
3967<p>
3968LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3969Collection</a> requires the implementation and generation of these intrinsics.
3970These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
3971stack</a>, as well as garbage collector implementations that require <a
3972href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
3973Front-ends for type-safe garbage collected languages should generate these
3974intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3975href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3976</p>
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3977
3978<p>The garbage collection intrinsics only operate on objects in the generic
3979 address space (address space zero).</p>
3980
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3981</div>
3982
3983<!-- _______________________________________________________________________ -->
3984<div class="doc_subsubsection">
3985 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3986</div>
3987
3988<div class="doc_text">
3989
3990<h5>Syntax:</h5>
3991
3992<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]3993 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3994</pre>
3995
3996<h5>Overview:</h5>
3997
3998<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3999the code generator, and allows some metadata to be associated with it.</p>
4000
4001<h5>Arguments:</h5>
4002
4003<p>The first argument specifies the address of a stack object that contains the
4004root pointer. The second pointer (which must be either a constant or a global
4005value address) contains the meta-data to be associated with the root.</p>
4006
4007<h5>Semantics:</h5>
4008
4009<p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
4010location. At compile-time, the code generator generates information to allow
4011the runtime to find the pointer at GC safe points.
4012</p>
4013
4014</div>
4015
4016
4017<!-- _______________________________________________________________________ -->
4018<div class="doc_subsubsection">
4019 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4020</div>
4021
4022<div class="doc_text">
4023
4024<h5>Syntax:</h5>
4025
4026<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4027 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4028</pre>
4029
4030<h5>Overview:</h5>
4031
4032<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4033locations, allowing garbage collector implementations that require read
4034barriers.</p>
4035
4036<h5>Arguments:</h5>
4037
4038<p>The second argument is the address to read from, which should be an address
4039allocated from the garbage collector. The first object is a pointer to the
4040start of the referenced object, if needed by the language runtime (otherwise
4041null).</p>
4042
4043<h5>Semantics:</h5>
4044
4045<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4046instruction, but may be replaced with substantially more complex code by the
4047garbage collector runtime, as needed.</p>
4048
4049</div>
4050
4051
4052<!-- _______________________________________________________________________ -->
4053<div class="doc_subsubsection">
4054 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4055</div>
4056
4057<div class="doc_text">
4058
4059<h5>Syntax:</h5>
4060
4061<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4062 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4063</pre>
4064
4065<h5>Overview:</h5>
4066
4067<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4068locations, allowing garbage collector implementations that require write
4069barriers (such as generational or reference counting collectors).</p>
4070
4071<h5>Arguments:</h5>
4072
4073<p>The first argument is the reference to store, the second is the start of the
4074object to store it to, and the third is the address of the field of Obj to
4075store to. If the runtime does not require a pointer to the object, Obj may be
4076null.</p>
4077
4078<h5>Semantics:</h5>
4079
4080<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4081instruction, but may be replaced with substantially more complex code by the
4082garbage collector runtime, as needed.</p>
4083
4084</div>
4085
4086
4087
4088<!-- ======================================================================= -->
4089<div class="doc_subsection">
4090 <a name="int_codegen">Code Generator Intrinsics</a>
4091</div>
4092
4093<div class="doc_text">
4094<p>
4095These intrinsics are provided by LLVM to expose special features that may only
4096be implemented with code generator support.
4097</p>
4098
4099</div>
4100
4101<!-- _______________________________________________________________________ -->
4102<div class="doc_subsubsection">
4103 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4104</div>
4105
4106<div class="doc_text">
4107
4108<h5>Syntax:</h5>
4109<pre>
4110 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
4111</pre>
4112
4113<h5>Overview:</h5>
4114
4115<p>
4116The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4117target-specific value indicating the return address of the current function
4118or one of its callers.
4119</p>
4120
4121<h5>Arguments:</h5>
4122
4123<p>
4124The argument to this intrinsic indicates which function to return the address
4125for. Zero indicates the calling function, one indicates its caller, etc. The
4126argument is <b>required</b> to be a constant integer value.
4127</p>
4128
4129<h5>Semantics:</h5>
4130
4131<p>
4132The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4133the return address of the specified call frame, or zero if it cannot be
4134identified. The value returned by this intrinsic is likely to be incorrect or 0
4135for arguments other than zero, so it should only be used for debugging purposes.
4136</p>
4137
4138<p>
4139Note that calling this intrinsic does not prevent function inlining or other
4140aggressive transformations, so the value returned may not be that of the obvious
4141source-language caller.
4142</p>
4143</div>
4144
4145
4146<!-- _______________________________________________________________________ -->
4147<div class="doc_subsubsection">
4148 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4149</div>
4150
4151<div class="doc_text">
4152
4153<h5>Syntax:</h5>
4154<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4155 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4156</pre>
4157
4158<h5>Overview:</h5>
4159
4160<p>
4161The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4162target-specific frame pointer value for the specified stack frame.
4163</p>
4164
4165<h5>Arguments:</h5>
4166
4167<p>
4168The argument to this intrinsic indicates which function to return the frame
4169pointer for. Zero indicates the calling function, one indicates its caller,
4170etc. The argument is <b>required</b> to be a constant integer value.
4171</p>
4172
4173<h5>Semantics:</h5>
4174
4175<p>
4176The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4177the frame address of the specified call frame, or zero if it cannot be
4178identified. The value returned by this intrinsic is likely to be incorrect or 0
4179for arguments other than zero, so it should only be used for debugging purposes.
4180</p>
4181
4182<p>
4183Note that calling this intrinsic does not prevent function inlining or other
4184aggressive transformations, so the value returned may not be that of the obvious
4185source-language caller.
4186</p>
4187</div>
4188
4189<!-- _______________________________________________________________________ -->
4190<div class="doc_subsubsection">
4191 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4192</div>
4193
4194<div class="doc_text">
4195
4196<h5>Syntax:</h5>
4197<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4198 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4199</pre>
4200
4201<h5>Overview:</h5>
4202
4203<p>
4204The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4205the function stack, for use with <a href="#int_stackrestore">
4206<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4207features like scoped automatic variable sized arrays in C99.
4208</p>
4209
4210<h5>Semantics:</h5>
4211
4212<p>
4213This intrinsic returns a opaque pointer value that can be passed to <a
4214href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4215<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4216<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4217state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4218practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4219that were allocated after the <tt>llvm.stacksave</tt> was executed.
4220</p>
4221
4222</div>
4223
4224<!-- _______________________________________________________________________ -->
4225<div class="doc_subsubsection">
4226 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4227</div>
4228
4229<div class="doc_text">
4230
4231<h5>Syntax:</h5>
4232<pre>
4233 declare void @llvm.stackrestore(i8 * %ptr)
4234</pre>
4235
4236<h5>Overview:</h5>
4237
4238<p>
4239The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4240the function stack to the state it was in when the corresponding <a
4241href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4242useful for implementing language features like scoped automatic variable sized
4243arrays in C99.
4244</p>
4245
4246<h5>Semantics:</h5>
4247
4248<p>
4249See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4250</p>
4251
4252</div>
4253
4254
4255<!-- _______________________________________________________________________ -->
4256<div class="doc_subsubsection">
4257 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4258</div>
4259
4260<div class="doc_text">
4261
4262<h5>Syntax:</h5>
4263<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4264 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]4265</pre>
4266
4267<h5>Overview:</h5>
4268
4269
4270<p>
4271The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4272a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4273no
4274effect on the behavior of the program but can change its performance
4275characteristics.
4276</p>
4277
4278<h5>Arguments:</h5>
4279
4280<p>
4281<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4282determining if the fetch should be for a read (0) or write (1), and
4283<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4284locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4285<tt>locality</tt> arguments must be constant integers.
4286</p>
4287
4288<h5>Semantics:</h5>
4289
4290<p>
4291This intrinsic does not modify the behavior of the program. In particular,
4292prefetches cannot trap and do not produce a value. On targets that support this
4293intrinsic, the prefetch can provide hints to the processor cache for better
4294performance.
4295</p>
4296
4297</div>
4298
4299<!-- _______________________________________________________________________ -->
4300<div class="doc_subsubsection">
4301 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4302</div>
4303
4304<div class="doc_text">
4305
4306<h5>Syntax:</h5>
4307<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4308 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4309</pre>
4310
4311<h5>Overview:</h5>
4312
4313
4314<p>
4315The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4316(PC) in a region of
4317code to simulators and other tools. The method is target specific, but it is
4318expected that the marker will use exported symbols to transmit the PC of the marker.
4319The marker makes no guarantees that it will remain with any specific instruction
4320after optimizations. It is possible that the presence of a marker will inhibit
4321optimizations. The intended use is to be inserted after optimizations to allow
4322correlations of simulation runs.
4323</p>
4324
4325<h5>Arguments:</h5>
4326
4327<p>
4328<tt>id</tt> is a numerical id identifying the marker.
4329</p>
4330
4331<h5>Semantics:</h5>
4332
4333<p>
4334This intrinsic does not modify the behavior of the program. Backends that do not
4335support this intrinisic may ignore it.
4336</p>
4337
4338</div>
4339
4340<!-- _______________________________________________________________________ -->
4341<div class="doc_subsubsection">
4342 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4343</div>
4344
4345<div class="doc_text">
4346
4347<h5>Syntax:</h5>
4348<pre>
4349 declare i64 @llvm.readcyclecounter( )
4350</pre>
4351
4352<h5>Overview:</h5>
4353
4354
4355<p>
4356The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4357counter register (or similar low latency, high accuracy clocks) on those targets
4358that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4359As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4360should only be used for small timings.
4361</p>
4362
4363<h5>Semantics:</h5>
4364
4365<p>
4366When directly supported, reading the cycle counter should not modify any memory.
4367Implementations are allowed to either return a application specific value or a
4368system wide value. On backends without support, this is lowered to a constant 0.
4369</p>
4370
4371</div>
4372
4373<!-- ======================================================================= -->
4374<div class="doc_subsection">
4375 <a name="int_libc">Standard C Library Intrinsics</a>
4376</div>
4377
4378<div class="doc_text">
4379<p>
4380LLVM provides intrinsics for a few important standard C library functions.
4381These intrinsics allow source-language front-ends to pass information about the
4382alignment of the pointer arguments to the code generator, providing opportunity
4383for more efficient code generation.
4384</p>
4385
4386</div>
4387
4388<!-- _______________________________________________________________________ -->
4389<div class="doc_subsubsection">
4390 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4391</div>
4392
4393<div class="doc_text">
4394
4395<h5>Syntax:</h5>
4396<pre>
4397 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4398 i32 &lt;len&gt;, i32 &lt;align&gt;)
4399 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4400 i64 &lt;len&gt;, i32 &lt;align&gt;)
4401</pre>
4402
4403<h5>Overview:</h5>
4404
4405<p>
4406The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4407location to the destination location.
4408</p>
4409
4410<p>
4411Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4412intrinsics do not return a value, and takes an extra alignment argument.
4413</p>
4414
4415<h5>Arguments:</h5>
4416
4417<p>
4418The first argument is a pointer to the destination, the second is a pointer to
4419the source. The third argument is an integer argument
4420specifying the number of bytes to copy, and the fourth argument is the alignment
4421of the source and destination locations.
4422</p>
4423
4424<p>
4425If the call to this intrinisic has an alignment value that is not 0 or 1, then
4426the caller guarantees that both the source and destination pointers are aligned
4427to that boundary.
4428</p>
4429
4430<h5>Semantics:</h5>
4431
4432<p>
4433The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4434location to the destination location, which are not allowed to overlap. It
4435copies "len" bytes of memory over. If the argument is known to be aligned to
4436some boundary, this can be specified as the fourth argument, otherwise it should
4437be set to 0 or 1.
4438</p>
4439</div>
4440
4441
4442<!-- _______________________________________________________________________ -->
4443<div class="doc_subsubsection">
4444 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4445</div>
4446
4447<div class="doc_text">
4448
4449<h5>Syntax:</h5>
4450<pre>
4451 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4452 i32 &lt;len&gt;, i32 &lt;align&gt;)
4453 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4454 i64 &lt;len&gt;, i32 &lt;align&gt;)
4455</pre>
4456
4457<h5>Overview:</h5>
4458
4459<p>
4460The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4461location to the destination location. It is similar to the
4462'<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4463</p>
4464
4465<p>
4466Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4467intrinsics do not return a value, and takes an extra alignment argument.
4468</p>
4469
4470<h5>Arguments:</h5>
4471
4472<p>
4473The first argument is a pointer to the destination, the second is a pointer to
4474the source. The third argument is an integer argument
4475specifying the number of bytes to copy, and the fourth argument is the alignment
4476of the source and destination locations.
4477</p>
4478
4479<p>
4480If the call to this intrinisic has an alignment value that is not 0 or 1, then
4481the caller guarantees that the source and destination pointers are aligned to
4482that boundary.
4483</p>
4484
4485<h5>Semantics:</h5>
4486
4487<p>
4488The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4489location to the destination location, which may overlap. It
4490copies "len" bytes of memory over. If the argument is known to be aligned to
4491some boundary, this can be specified as the fourth argument, otherwise it should
4492be set to 0 or 1.
4493</p>
4494</div>
4495
4496
4497<!-- _______________________________________________________________________ -->
4498<div class="doc_subsubsection">
4499 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4500</div>
4501
4502<div class="doc_text">
4503
4504<h5>Syntax:</h5>
4505<pre>
4506 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
4507 i32 &lt;len&gt;, i32 &lt;align&gt;)
4508 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
4509 i64 &lt;len&gt;, i32 &lt;align&gt;)
4510</pre>
4511
4512<h5>Overview:</h5>
4513
4514<p>
4515The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4516byte value.
4517</p>
4518
4519<p>
4520Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4521does not return a value, and takes an extra alignment argument.
4522</p>
4523
4524<h5>Arguments:</h5>
4525
4526<p>
4527The first argument is a pointer to the destination to fill, the second is the
4528byte value to fill it with, the third argument is an integer
4529argument specifying the number of bytes to fill, and the fourth argument is the
4530known alignment of destination location.
4531</p>
4532
4533<p>
4534If the call to this intrinisic has an alignment value that is not 0 or 1, then
4535the caller guarantees that the destination pointer is aligned to that boundary.
4536</p>
4537
4538<h5>Semantics:</h5>
4539
4540<p>
4541The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4542the
4543destination location. If the argument is known to be aligned to some boundary,
4544this can be specified as the fourth argument, otherwise it should be set to 0 or
45451.
4546</p>
4547</div>
4548
4549
4550<!-- _______________________________________________________________________ -->
4551<div class="doc_subsubsection">
4552 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4553</div>
4554
4555<div class="doc_text">
4556
4557<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4558<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]4559floating point or vector of floating point type. Not all targets support all
4560types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4561<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4562 declare float @llvm.sqrt.f32(float %Val)
4563 declare double @llvm.sqrt.f64(double %Val)
4564 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
4565 declare fp128 @llvm.sqrt.f128(fp128 %Val)
4566 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4567</pre>
4568
4569<h5>Overview:</h5>
4570
4571<p>
4572The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4573returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4574<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4575negative numbers (which allows for better optimization).
4576</p>
4577
4578<h5>Arguments:</h5>
4579
4580<p>
4581The argument and return value are floating point numbers of the same type.
4582</p>
4583
4584<h5>Semantics:</h5>
4585
4586<p>
4587This function returns the sqrt of the specified operand if it is a nonnegative
4588floating point number.
4589</p>
4590</div>
4591
4592<!-- _______________________________________________________________________ -->
4593<div class="doc_subsubsection">
4594 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4595</div>
4596
4597<div class="doc_text">
4598
4599<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4600<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]4601floating point or vector of floating point type. Not all targets support all
4602types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4603<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4604 declare float @llvm.powi.f32(float %Val, i32 %power)
4605 declare double @llvm.powi.f64(double %Val, i32 %power)
4606 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
4607 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
4608 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4609</pre>
4610
4611<h5>Overview:</h5>
4612
4613<p>
4614The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4615specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4616multiplications is not defined. When a vector of floating point type is
4617used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4618</p>
4619
4620<h5>Arguments:</h5>
4621
4622<p>
4623The second argument is an integer power, and the first is a value to raise to
4624that power.
4625</p>
4626
4627<h5>Semantics:</h5>
4628
4629<p>
4630This function returns the first value raised to the second power with an
4631unspecified sequence of rounding operations.</p>
4632</div>
4633
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4634<!-- _______________________________________________________________________ -->
4635<div class="doc_subsubsection">
4636 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
4637</div>
4638
4639<div class="doc_text">
4640
4641<h5>Syntax:</h5>
4642<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
4643floating point or vector of floating point type. Not all targets support all
4644types however.
4645<pre>
4646 declare float @llvm.sin.f32(float %Val)
4647 declare double @llvm.sin.f64(double %Val)
4648 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
4649 declare fp128 @llvm.sin.f128(fp128 %Val)
4650 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
4651</pre>
4652
4653<h5>Overview:</h5>
4654
4655<p>
4656The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
4657</p>
4658
4659<h5>Arguments:</h5>
4660
4661<p>
4662The argument and return value are floating point numbers of the same type.
4663</p>
4664
4665<h5>Semantics:</h5>
4666
4667<p>
4668This function returns the sine of the specified operand, returning the
4669same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4670conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4671</div>
4672
4673<!-- _______________________________________________________________________ -->
4674<div class="doc_subsubsection">
4675 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
4676</div>
4677
4678<div class="doc_text">
4679
4680<h5>Syntax:</h5>
4681<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
4682floating point or vector of floating point type. Not all targets support all
4683types however.
4684<pre>
4685 declare float @llvm.cos.f32(float %Val)
4686 declare double @llvm.cos.f64(double %Val)
4687 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
4688 declare fp128 @llvm.cos.f128(fp128 %Val)
4689 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
4690</pre>
4691
4692<h5>Overview:</h5>
4693
4694<p>
4695The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
4696</p>
4697
4698<h5>Arguments:</h5>
4699
4700<p>
4701The argument and return value are floating point numbers of the same type.
4702</p>
4703
4704<h5>Semantics:</h5>
4705
4706<p>
4707This function returns the cosine of the specified operand, returning the
4708same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4709conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4710</div>
4711
4712<!-- _______________________________________________________________________ -->
4713<div class="doc_subsubsection">
4714 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
4715</div>
4716
4717<div class="doc_text">
4718
4719<h5>Syntax:</h5>
4720<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
4721floating point or vector of floating point type. Not all targets support all
4722types however.
4723<pre>
4724 declare float @llvm.pow.f32(float %Val, float %Power)
4725 declare double @llvm.pow.f64(double %Val, double %Power)
4726 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
4727 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
4728 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
4729</pre>
4730
4731<h5>Overview:</h5>
4732
4733<p>
4734The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
4735specified (positive or negative) power.
4736</p>
4737
4738<h5>Arguments:</h5>
4739
4740<p>
4741The second argument is a floating point power, and the first is a value to
4742raise to that power.
4743</p>
4744
4745<h5>Semantics:</h5>
4746
4747<p>
4748This function returns the first value raised to the second power,
4749returning the
4750same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4751conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4752</div>
4753
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4754
4755<!-- ======================================================================= -->
4756<div class="doc_subsection">
4757 <a name="int_manip">Bit Manipulation Intrinsics</a>
4758</div>
4759
4760<div class="doc_text">
4761<p>
4762LLVM provides intrinsics for a few important bit manipulation operations.
4763These allow efficient code generation for some algorithms.
4764</p>
4765
4766</div>
4767
4768<!-- _______________________________________________________________________ -->
4769<div class="doc_subsubsection">
4770 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4771</div>
4772
4773<div class="doc_text">
4774
4775<h5>Syntax:</h5>
4776<p>This is an overloaded intrinsic function. You can use bswap on any integer
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4777type that is an even number of bytes (i.e. BitWidth % 16 == 0).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4778<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4779 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
4780 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
4781 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4782</pre>
4783
4784<h5>Overview:</h5>
4785
4786<p>
4787The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
4788values with an even number of bytes (positive multiple of 16 bits). These are
4789useful for performing operations on data that is not in the target's native
4790byte order.
4791</p>
4792
4793<h5>Semantics:</h5>
4794
4795<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4796The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4797and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4798intrinsic returns an i32 value that has the four bytes of the input i32
4799swapped, 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]4800i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
4801<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4802additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4803</p>
4804
4805</div>
4806
4807<!-- _______________________________________________________________________ -->
4808<div class="doc_subsubsection">
4809 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4810</div>
4811
4812<div class="doc_text">
4813
4814<h5>Syntax:</h5>
4815<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4816width. Not all targets support all bit widths however.
4817<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4818 declare i8 @llvm.ctpop.i8 (i8 &lt;src&gt;)
4819 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4820 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4821 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
4822 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4823</pre>
4824
4825<h5>Overview:</h5>
4826
4827<p>
4828The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4829value.
4830</p>
4831
4832<h5>Arguments:</h5>
4833
4834<p>
4835The only argument is the value to be counted. The argument may be of any
4836integer type. The return type must match the argument type.
4837</p>
4838
4839<h5>Semantics:</h5>
4840
4841<p>
4842The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4843</p>
4844</div>
4845
4846<!-- _______________________________________________________________________ -->
4847<div class="doc_subsubsection">
4848 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4849</div>
4850
4851<div class="doc_text">
4852
4853<h5>Syntax:</h5>
4854<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4855integer bit width. Not all targets support all bit widths however.
4856<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4857 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
4858 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4859 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4860 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
4861 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4862</pre>
4863
4864<h5>Overview:</h5>
4865
4866<p>
4867The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4868leading zeros in a variable.
4869</p>
4870
4871<h5>Arguments:</h5>
4872
4873<p>
4874The only argument is the value to be counted. The argument may be of any
4875integer type. The return type must match the argument type.
4876</p>
4877
4878<h5>Semantics:</h5>
4879
4880<p>
4881The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4882in a variable. If the src == 0 then the result is the size in bits of the type
4883of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4884</p>
4885</div>
4886
4887
4888
4889<!-- _______________________________________________________________________ -->
4890<div class="doc_subsubsection">
4891 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4892</div>
4893
4894<div class="doc_text">
4895
4896<h5>Syntax:</h5>
4897<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
4898integer bit width. Not all targets support all bit widths however.
4899<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4900 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
4901 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4902 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4903 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
4904 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4905</pre>
4906
4907<h5>Overview:</h5>
4908
4909<p>
4910The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4911trailing zeros.
4912</p>
4913
4914<h5>Arguments:</h5>
4915
4916<p>
4917The only argument is the value to be counted. The argument may be of any
4918integer type. The return type must match the argument type.
4919</p>
4920
4921<h5>Semantics:</h5>
4922
4923<p>
4924The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4925in a variable. If the src == 0 then the result is the size in bits of the type
4926of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4927</p>
4928</div>
4929
4930<!-- _______________________________________________________________________ -->
4931<div class="doc_subsubsection">
4932 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
4933</div>
4934
4935<div class="doc_text">
4936
4937<h5>Syntax:</h5>
4938<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
4939on any integer bit width.
4940<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4941 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
4942 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4943</pre>
4944
4945<h5>Overview:</h5>
4946<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
4947range of bits from an integer value and returns them in the same bit width as
4948the original value.</p>
4949
4950<h5>Arguments:</h5>
4951<p>The first argument, <tt>%val</tt> and the result may be integer types of
4952any bit width but they must have the same bit width. The second and third
4953arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
4954
4955<h5>Semantics:</h5>
4956<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
4957of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
4958<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
4959operates in forward mode.</p>
4960<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
4961right by <tt>%loBit</tt> bits and then ANDing it with a mask with
4962only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
4963<ol>
4964 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
4965 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
4966 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
4967 to determine the number of bits to retain.</li>
4968 <li>A mask of the retained bits is created by shifting a -1 value.</li>
4969 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
4970</ol>
4971<p>In reverse mode, a similar computation is made except that the bits are
4972returned in the reverse order. So, for example, if <tt>X</tt> has the value
4973<tt>i16 0x0ACF (101011001111)</tt> and we apply
4974<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
4975<tt>i16 0x0026 (000000100110)</tt>.</p>
4976</div>
4977
4978<div class="doc_subsubsection">
4979 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
4980</div>
4981
4982<div class="doc_text">
4983
4984<h5>Syntax:</h5>
4985<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
4986on any integer bit width.
4987<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4988 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
4989 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]4990</pre>
4991
4992<h5>Overview:</h5>
4993<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
4994of bits in an integer value with another integer value. It returns the integer
4995with the replaced bits.</p>
4996
4997<h5>Arguments:</h5>
4998<p>The first argument, <tt>%val</tt> and the result may be integer types of
4999any bit width but they must have the same bit width. <tt>%val</tt> is the value
5000whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5001integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5002type since they specify only a bit index.</p>
5003
5004<h5>Semantics:</h5>
5005<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5006of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5007<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5008operates in forward mode.</p>
5009<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5010truncating it down to the size of the replacement area or zero extending it
5011up to that size.</p>
5012<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5013are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5014in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5015to the <tt>%hi</tt>th bit.
5016<p>In reverse mode, a similar computation is made except that the bits are
5017reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5018<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
5019<h5>Examples:</h5>
5020<pre>
5021 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
5022 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
5023 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
5024 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
5025 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
5026</pre>
5027</div>
5028
5029<!-- ======================================================================= -->
5030<div class="doc_subsection">
5031 <a name="int_debugger">Debugger Intrinsics</a>
5032</div>
5033
5034<div class="doc_text">
5035<p>
5036The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5037are described in the <a
5038href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5039Debugging</a> document.
5040</p>
5041</div>
5042
5043
5044<!-- ======================================================================= -->
5045<div class="doc_subsection">
5046 <a name="int_eh">Exception Handling Intrinsics</a>
5047</div>
5048
5049<div class="doc_text">
5050<p> The LLVM exception handling intrinsics (which all start with
5051<tt>llvm.eh.</tt> prefix), are described in the <a
5052href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5053Handling</a> document. </p>
5054</div>
5055
5056<!-- ======================================================================= -->
5057<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5058 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5059</div>
5060
5061<div class="doc_text">
5062<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5063 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5064 the <tt>nest</tt> attribute, from a function. The result is a callable
5065 function pointer lacking the nest parameter - the caller does not need
5066 to provide a value for it. Instead, the value to use is stored in
5067 advance in a "trampoline", a block of memory usually allocated
5068 on the stack, which also contains code to splice the nest value into the
5069 argument list. This is used to implement the GCC nested function address
5070 extension.
5071</p>
5072<p>
5073 For example, if the function is
5074 <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]5075 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]5076<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5077 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5078 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5079 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5080 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5081</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]5082 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5083 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5084</div>
5085
5086<!-- _______________________________________________________________________ -->
5087<div class="doc_subsubsection">
5088 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5089</div>
5090<div class="doc_text">
5091<h5>Syntax:</h5>
5092<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5093declare 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]5094</pre>
5095<h5>Overview:</h5>
5096<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5097 This fills the memory pointed to by <tt>tramp</tt> with code
5098 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5099</p>
5100<h5>Arguments:</h5>
5101<p>
5102 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5103 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5104 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]5105 intrinsic. Note that the size and the alignment are target-specific - LLVM
5106 currently provides no portable way of determining them, so a front-end that
5107 generates this intrinsic needs to have some target-specific knowledge.
5108 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]5109</p>
5110<h5>Semantics:</h5>
5111<p>
5112 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5113 dependent code, turning it into a function. A pointer to this function is
5114 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5115 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5116 before being called. The new function's signature is the same as that of
5117 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5118 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5119 of pointer type. Calling the new function is equivalent to calling
5120 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5121 missing <tt>nest</tt> argument. If, after calling
5122 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5123 modified, then the effect of any later call to the returned function pointer is
5124 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5125</p>
5126</div>
5127
5128<!-- ======================================================================= -->
5129<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5130 <a name="int_general">General Intrinsics</a>
5131</div>
5132
5133<div class="doc_text">
5134<p> This class of intrinsics is designed to be generic and has
5135no specific purpose. </p>
5136</div>
5137
5138<!-- _______________________________________________________________________ -->
5139<div class="doc_subsubsection">
5140 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
5141</div>
5142
5143<div class="doc_text">
5144
5145<h5>Syntax:</h5>
5146<pre>
5147 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5148</pre>
5149
5150<h5>Overview:</h5>
5151
5152<p>
5153The '<tt>llvm.var.annotation</tt>' intrinsic
5154</p>
5155
5156<h5>Arguments:</h5>
5157
5158<p>
5159The first argument is a pointer to a value, the second is a pointer to a
5160global string, the third is a pointer to a global string which is the source
5161file name, and the last argument is the line number.
5162</p>
5163
5164<h5>Semantics:</h5>
5165
5166<p>
5167This intrinsic allows annotation of local variables with arbitrary strings.
5168This can be useful for special purpose optimizations that want to look for these
5169 annotations. These have no other defined use, they are ignored by code
5170 generation and optimization.
5171</div>
5172
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5173<!-- _______________________________________________________________________ -->
5174<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]5175 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5176</div>
5177
5178<div class="doc_text">
5179
5180<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5181<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
5182any integer bit width.
5183</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5184<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]5185 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5186 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5187 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5188 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5189 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]5190</pre>
5191
5192<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5193
5194<p>
5195The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5196</p>
5197
5198<h5>Arguments:</h5>
5199
5200<p>
5201The first argument is an integer value (result of some expression),
5202the second is a pointer to a global string, the third is a pointer to a global
5203string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5204It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5205</p>
5206
5207<h5>Semantics:</h5>
5208
5209<p>
5210This intrinsic allows annotations to be put on arbitrary expressions
5211with arbitrary strings. This can be useful for special purpose optimizations
5212that want to look for these annotations. These have no other defined use, they
5213are ignored by code generation and optimization.
5214</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5215
5216<!-- *********************************************************************** -->
5217<hr>
5218<address>
5219 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
5220 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
5221 <a href="http://validator.w3.org/check/referer"><img
5222 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
5223
5224 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
5225 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
5226 Last modified: $Date$
5227</address>
5228</body>
5229</html>