Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 2594d94fcc0f5c4ffc32059dec5209efa97d55a6 / . / docs / LangRef.html
blob: d09d34860e25522c7515e65a84f4cdd65ca121ae [file] [log] [blame]
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
3<html>
4<head>
5 <title>LLVM Assembly Language Reference Manual</title>
6 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
8 <meta name="description"
9 content="LLVM Assembly Language Reference Manual.">
10 <link rel="stylesheet" href="llvm.css" type="text/css">
11</head>
12
13<body>
14
15<div class="doc_title"> LLVM Language Reference Manual </div>
16<ol>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
20 <li><a href="#highlevel">High Level Structure</a>
21 <ol>
22 <li><a href="#modulestructure">Module Structure</a></li>
23 <li><a href="#linkage">Linkage Types</a></li>
24 <li><a href="#callingconv">Calling Conventions</a></li>
25 <li><a href="#globalvars">Global Variables</a></li>
26 <li><a href="#functionstructure">Functions</a></li>
27 <li><a href="#aliasstructure">Aliases</a>
28 <li><a href="#paramattrs">Parameter Attributes</a></li>
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]29 <li><a href="#gc">Garbage Collector Names</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]30 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
31 <li><a href="#datalayout">Data Layout</a></li>
32 </ol>
33 </li>
34 <li><a href="#typesystem">Type System</a>
35 <ol>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]36 <li><a href="#t_classifications">Type Classifications</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]37 <li><a href="#t_primitive">Primitive Types</a>
38 <ol>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]39 <li><a href="#t_floating">Floating Point Types</a></li>
40 <li><a href="#t_void">Void Type</a></li>
41 <li><a href="#t_label">Label Type</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]42 </ol>
43 </li>
44 <li><a href="#t_derived">Derived Types</a>
45 <ol>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]46 <li><a href="#t_integer">Integer Type</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]47 <li><a href="#t_array">Array Type</a></li>
48 <li><a href="#t_function">Function Type</a></li>
49 <li><a href="#t_pointer">Pointer Type</a></li>
50 <li><a href="#t_struct">Structure Type</a></li>
51 <li><a href="#t_pstruct">Packed Structure Type</a></li>
52 <li><a href="#t_vector">Vector Type</a></li>
53 <li><a href="#t_opaque">Opaque Type</a></li>
54 </ol>
55 </li>
56 </ol>
57 </li>
58 <li><a href="#constants">Constants</a>
59 <ol>
60 <li><a href="#simpleconstants">Simple Constants</a>
61 <li><a href="#aggregateconstants">Aggregate Constants</a>
62 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
63 <li><a href="#undefvalues">Undefined Values</a>
64 <li><a href="#constantexprs">Constant Expressions</a>
65 </ol>
66 </li>
67 <li><a href="#othervalues">Other Values</a>
68 <ol>
69 <li><a href="#inlineasm">Inline Assembler Expressions</a>
70 </ol>
71 </li>
72 <li><a href="#instref">Instruction Reference</a>
73 <ol>
74 <li><a href="#terminators">Terminator Instructions</a>
75 <ol>
76 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
77 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
78 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
79 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
80 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
81 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
82 </ol>
83 </li>
84 <li><a href="#binaryops">Binary Operations</a>
85 <ol>
86 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
87 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
88 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
89 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
90 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
91 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
92 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
93 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
94 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
95 </ol>
96 </li>
97 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
98 <ol>
99 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
100 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
101 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
102 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
103 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
104 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
105 </ol>
106 </li>
107 <li><a href="#vectorops">Vector Operations</a>
108 <ol>
109 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
110 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
111 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
112 </ol>
113 </li>
114 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
115 <ol>
116 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
117 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
118 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
119 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
120 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
121 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
122 </ol>
123 </li>
124 <li><a href="#convertops">Conversion Operations</a>
125 <ol>
126 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
127 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
128 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
129 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
130 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
131 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
132 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
133 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
134 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
135 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
136 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
137 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
138 </ol>
139 <li><a href="#otherops">Other Operations</a>
140 <ol>
141 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
142 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
143 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
144 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
145 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
146 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
147 </ol>
148 </li>
149 </ol>
150 </li>
151 <li><a href="#intrinsics">Intrinsic Functions</a>
152 <ol>
153 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
154 <ol>
155 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
156 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
157 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
158 </ol>
159 </li>
160 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
161 <ol>
162 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
163 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
164 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
165 </ol>
166 </li>
167 <li><a href="#int_codegen">Code Generator Intrinsics</a>
168 <ol>
169 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
170 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
171 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
172 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
173 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
174 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
175 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
176 </ol>
177 </li>
178 <li><a href="#int_libc">Standard C Library Intrinsics</a>
179 <ol>
180 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
181 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
182 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
183 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
184 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]185 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
186 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
187 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]188 </ol>
189 </li>
190 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
191 <ol>
192 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
193 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
194 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
195 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
196 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
197 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
198 </ol>
199 </li>
200 <li><a href="#int_debugger">Debugger intrinsics</a></li>
201 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]202 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]203 <ol>
204 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]205 </ol>
206 </li>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]207 <li><a href="#int_atomics">Atomic intrinsics</a>
208 <ol>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]209 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
210 <li><a href="#int_atomic_lcs"><tt>llvm.atomic.lcs</tt></a></li>
211 <li><a href="#int_atomic_las"><tt>llvm.atomic.las</tt></a></li>
212 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]213 </ol>
214 </li>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]215 <li><a href="#int_general">General intrinsics</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]216 <ol>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]217 <li><a href="#int_var_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]218 <tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]219 <li><a href="#int_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]220 <tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]221 <li><a href="#int_trap">
222 <tt>llvm.trap</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]223 </ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]224 </li>
225 </ol>
226 </li>
227</ol>
228
229<div class="doc_author">
230 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
231 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
232</div>
233
234<!-- *********************************************************************** -->
235<div class="doc_section"> <a name="abstract">Abstract </a></div>
236<!-- *********************************************************************** -->
237
238<div class="doc_text">
239<p>This document is a reference manual for the LLVM assembly language.
240LLVM is an SSA based representation that provides type safety,
241low-level operations, flexibility, and the capability of representing
242'all' high-level languages cleanly. It is the common code
243representation used throughout all phases of the LLVM compilation
244strategy.</p>
245</div>
246
247<!-- *********************************************************************** -->
248<div class="doc_section"> <a name="introduction">Introduction</a> </div>
249<!-- *********************************************************************** -->
250
251<div class="doc_text">
252
253<p>The LLVM code representation is designed to be used in three
254different forms: as an in-memory compiler IR, as an on-disk bitcode
255representation (suitable for fast loading by a Just-In-Time compiler),
256and as a human readable assembly language representation. This allows
257LLVM to provide a powerful intermediate representation for efficient
258compiler transformations and analysis, while providing a natural means
259to debug and visualize the transformations. The three different forms
260of LLVM are all equivalent. This document describes the human readable
261representation and notation.</p>
262
263<p>The LLVM representation aims to be light-weight and low-level
264while being expressive, typed, and extensible at the same time. It
265aims to be a "universal IR" of sorts, by being at a low enough level
266that high-level ideas may be cleanly mapped to it (similar to how
267microprocessors are "universal IR's", allowing many source languages to
268be mapped to them). By providing type information, LLVM can be used as
269the target of optimizations: for example, through pointer analysis, it
270can be proven that a C automatic variable is never accessed outside of
271the current function... allowing it to be promoted to a simple SSA
272value instead of a memory location.</p>
273
274</div>
275
276<!-- _______________________________________________________________________ -->
277<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
278
279<div class="doc_text">
280
281<p>It is important to note that this document describes 'well formed'
282LLVM assembly language. There is a difference between what the parser
283accepts and what is considered 'well formed'. For example, the
284following instruction is syntactically okay, but not well formed:</p>
285
286<div class="doc_code">
287<pre>
288%x = <a href="#i_add">add</a> i32 1, %x
289</pre>
290</div>
291
292<p>...because the definition of <tt>%x</tt> does not dominate all of
293its uses. The LLVM infrastructure provides a verification pass that may
294be used to verify that an LLVM module is well formed. This pass is
295automatically run by the parser after parsing input assembly and by
296the optimizer before it outputs bitcode. The violations pointed out
297by the verifier pass indicate bugs in transformation passes or input to
298the parser.</p>
299</div>
300
Chris Lattnera83fdc02007-10-03 17:34:29 +0000[diff] [blame]301<!-- Describe the typesetting conventions here. -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]302
303<!-- *********************************************************************** -->
304<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
305<!-- *********************************************************************** -->
306
307<div class="doc_text">
308
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]309 <p>LLVM identifiers come in two basic types: global and local. Global
310 identifiers (functions, global variables) begin with the @ character. Local
311 identifiers (register names, types) begin with the % character. Additionally,
312 there are three different formats for identifiers, for different purposes:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]313
314<ol>
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]315 <li>Named values are represented as a string of characters with their prefix.
316 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
317 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]318 Identifiers which require other characters in their names can be surrounded
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]319 with quotes. In this way, anything except a <tt>&quot;</tt> character can
320 be used in a named value.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]321
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]322 <li>Unnamed values are represented as an unsigned numeric value with their
323 prefix. For example, %12, @2, %44.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]324
325 <li>Constants, which are described in a <a href="#constants">section about
326 constants</a>, below.</li>
327</ol>
328
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]329<p>LLVM requires that values start with a prefix for two reasons: Compilers
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]330don't need to worry about name clashes with reserved words, and the set of
331reserved words may be expanded in the future without penalty. Additionally,
332unnamed identifiers allow a compiler to quickly come up with a temporary
333variable without having to avoid symbol table conflicts.</p>
334
335<p>Reserved words in LLVM are very similar to reserved words in other
336languages. There are keywords for different opcodes
337('<tt><a href="#i_add">add</a></tt>',
338 '<tt><a href="#i_bitcast">bitcast</a></tt>',
339 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
340href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
341and others. These reserved words cannot conflict with variable names, because
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]342none of them start with a prefix character ('%' or '@').</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]343
344<p>Here is an example of LLVM code to multiply the integer variable
345'<tt>%X</tt>' by 8:</p>
346
347<p>The easy way:</p>
348
349<div class="doc_code">
350<pre>
351%result = <a href="#i_mul">mul</a> i32 %X, 8
352</pre>
353</div>
354
355<p>After strength reduction:</p>
356
357<div class="doc_code">
358<pre>
359%result = <a href="#i_shl">shl</a> i32 %X, i8 3
360</pre>
361</div>
362
363<p>And the hard way:</p>
364
365<div class="doc_code">
366<pre>
367<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
368<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
369%result = <a href="#i_add">add</a> i32 %1, %1
370</pre>
371</div>
372
373<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
374important lexical features of LLVM:</p>
375
376<ol>
377
378 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
379 line.</li>
380
381 <li>Unnamed temporaries are created when the result of a computation is not
382 assigned to a named value.</li>
383
384 <li>Unnamed temporaries are numbered sequentially</li>
385
386</ol>
387
388<p>...and it also shows a convention that we follow in this document. When
389demonstrating instructions, we will follow an instruction with a comment that
390defines the type and name of value produced. Comments are shown in italic
391text.</p>
392
393</div>
394
395<!-- *********************************************************************** -->
396<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
397<!-- *********************************************************************** -->
398
399<!-- ======================================================================= -->
400<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
401</div>
402
403<div class="doc_text">
404
405<p>LLVM programs are composed of "Module"s, each of which is a
406translation unit of the input programs. Each module consists of
407functions, global variables, and symbol table entries. Modules may be
408combined together with the LLVM linker, which merges function (and
409global variable) definitions, resolves forward declarations, and merges
410symbol table entries. Here is an example of the "hello world" module:</p>
411
412<div class="doc_code">
413<pre><i>; Declare the string constant as a global constant...</i>
414<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
415 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
416
417<i>; External declaration of the puts function</i>
418<a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
419
420<i>; Definition of main function</i>
421define i32 @main() { <i>; i32()* </i>
422 <i>; Convert [13x i8 ]* to i8 *...</i>
423 %cast210 = <a
424 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
425
426 <i>; Call puts function to write out the string to stdout...</i>
427 <a
428 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
429 <a
430 href="#i_ret">ret</a> i32 0<br>}<br>
431</pre>
432</div>
433
434<p>This example is made up of a <a href="#globalvars">global variable</a>
435named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
436function, and a <a href="#functionstructure">function definition</a>
437for "<tt>main</tt>".</p>
438
439<p>In general, a module is made up of a list of global values,
440where both functions and global variables are global values. Global values are
441represented by a pointer to a memory location (in this case, a pointer to an
442array of char, and a pointer to a function), and have one of the following <a
443href="#linkage">linkage types</a>.</p>
444
445</div>
446
447<!-- ======================================================================= -->
448<div class="doc_subsection">
449 <a name="linkage">Linkage Types</a>
450</div>
451
452<div class="doc_text">
453
454<p>
455All Global Variables and Functions have one of the following types of linkage:
456</p>
457
458<dl>
459
460 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
461
462 <dd>Global values with internal linkage are only directly accessible by
463 objects in the current module. In particular, linking code into a module with
464 an internal global value may cause the internal to be renamed as necessary to
465 avoid collisions. Because the symbol is internal to the module, all
466 references can be updated. This corresponds to the notion of the
467 '<tt>static</tt>' keyword in C.
468 </dd>
469
470 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
471
472 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
473 the same name when linkage occurs. This is typically used to implement
474 inline functions, templates, or other code which must be generated in each
475 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
476 allowed to be discarded.
477 </dd>
478
479 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
480
481 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
482 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
483 used for globals that may be emitted in multiple translation units, but that
484 are not guaranteed to be emitted into every translation unit that uses them.
485 One example of this are common globals in C, such as "<tt>int X;</tt>" at
486 global scope.
487 </dd>
488
489 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
490
491 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
492 pointer to array type. When two global variables with appending linkage are
493 linked together, the two global arrays are appended together. This is the
494 LLVM, typesafe, equivalent of having the system linker append together
495 "sections" with identical names when .o files are linked.
496 </dd>
497
498 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
499 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
500 until linked, if not linked, the symbol becomes null instead of being an
501 undefined reference.
502 </dd>
503
504 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
505
506 <dd>If none of the above identifiers are used, the global is externally
507 visible, meaning that it participates in linkage and can be used to resolve
508 external symbol references.
509 </dd>
510</dl>
511
512 <p>
513 The next two types of linkage are targeted for Microsoft Windows platform
514 only. They are designed to support importing (exporting) symbols from (to)
515 DLLs.
516 </p>
517
518 <dl>
519 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
520
521 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
522 or variable via a global pointer to a pointer that is set up by the DLL
523 exporting the symbol. On Microsoft Windows targets, the pointer name is
524 formed by combining <code>_imp__</code> and the function or variable name.
525 </dd>
526
527 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
528
529 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
530 pointer to a pointer in a DLL, so that it can be referenced with the
531 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
532 name is formed by combining <code>_imp__</code> and the function or variable
533 name.
534 </dd>
535
536</dl>
537
538<p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
539variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
540variable and was linked with this one, one of the two would be renamed,
541preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
542external (i.e., lacking any linkage declarations), they are accessible
543outside of the current module.</p>
544<p>It is illegal for a function <i>declaration</i>
545to have any linkage type other than "externally visible", <tt>dllimport</tt>,
546or <tt>extern_weak</tt>.</p>
547<p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
548linkages.
549</div>
550
551<!-- ======================================================================= -->
552<div class="doc_subsection">
553 <a name="callingconv">Calling Conventions</a>
554</div>
555
556<div class="doc_text">
557
558<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
559and <a href="#i_invoke">invokes</a> can all have an optional calling convention
560specified for the call. The calling convention of any pair of dynamic
561caller/callee must match, or the behavior of the program is undefined. The
562following calling conventions are supported by LLVM, and more may be added in
563the future:</p>
564
565<dl>
566 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
567
568 <dd>This calling convention (the default if no other calling convention is
569 specified) matches the target C calling conventions. This calling convention
570 supports varargs function calls and tolerates some mismatch in the declared
571 prototype and implemented declaration of the function (as does normal C).
572 </dd>
573
574 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
575
576 <dd>This calling convention attempts to make calls as fast as possible
577 (e.g. by passing things in registers). This calling convention allows the
578 target to use whatever tricks it wants to produce fast code for the target,
579 without having to conform to an externally specified ABI. Implementations of
580 this convention should allow arbitrary tail call optimization to be supported.
581 This calling convention does not support varargs and requires the prototype of
582 all callees to exactly match the prototype of the function definition.
583 </dd>
584
585 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
586
587 <dd>This calling convention attempts to make code in the caller as efficient
588 as possible under the assumption that the call is not commonly executed. As
589 such, these calls often preserve all registers so that the call does not break
590 any live ranges in the caller side. This calling convention does not support
591 varargs and requires the prototype of all callees to exactly match the
592 prototype of the function definition.
593 </dd>
594
595 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
596
597 <dd>Any calling convention may be specified by number, allowing
598 target-specific calling conventions to be used. Target specific calling
599 conventions start at 64.
600 </dd>
601</dl>
602
603<p>More calling conventions can be added/defined on an as-needed basis, to
604support pascal conventions or any other well-known target-independent
605convention.</p>
606
607</div>
608
609<!-- ======================================================================= -->
610<div class="doc_subsection">
611 <a name="visibility">Visibility Styles</a>
612</div>
613
614<div class="doc_text">
615
616<p>
617All Global Variables and Functions have one of the following visibility styles:
618</p>
619
620<dl>
621 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
622
623 <dd>On ELF, default visibility means that the declaration is visible to other
624 modules and, in shared libraries, means that the declared entity may be
625 overridden. On Darwin, default visibility means that the declaration is
626 visible to other modules. Default visibility corresponds to "external
627 linkage" in the language.
628 </dd>
629
630 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
631
632 <dd>Two declarations of an object with hidden visibility refer to the same
633 object if they are in the same shared object. Usually, hidden visibility
634 indicates that the symbol will not be placed into the dynamic symbol table,
635 so no other module (executable or shared library) can reference it
636 directly.
637 </dd>
638
639 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
640
641 <dd>On ELF, protected visibility indicates that the symbol will be placed in
642 the dynamic symbol table, but that references within the defining module will
643 bind to the local symbol. That is, the symbol cannot be overridden by another
644 module.
645 </dd>
646</dl>
647
648</div>
649
650<!-- ======================================================================= -->
651<div class="doc_subsection">
652 <a name="globalvars">Global Variables</a>
653</div>
654
655<div class="doc_text">
656
657<p>Global variables define regions of memory allocated at compilation time
658instead of run-time. Global variables may optionally be initialized, may have
659an explicit section to be placed in, and may have an optional explicit alignment
660specified. A variable may be defined as "thread_local", which means that it
661will not be shared by threads (each thread will have a separated copy of the
662variable). A variable may be defined as a global "constant," which indicates
663that the contents of the variable will <b>never</b> be modified (enabling better
664optimization, allowing the global data to be placed in the read-only section of
665an executable, etc). Note that variables that need runtime initialization
666cannot be marked "constant" as there is a store to the variable.</p>
667
668<p>
669LLVM explicitly allows <em>declarations</em> of global variables to be marked
670constant, even if the final definition of the global is not. This capability
671can be used to enable slightly better optimization of the program, but requires
672the language definition to guarantee that optimizations based on the
673'constantness' are valid for the translation units that do not include the
674definition.
675</p>
676
677<p>As SSA values, global variables define pointer values that are in
678scope (i.e. they dominate) all basic blocks in the program. Global
679variables always define a pointer to their "content" type because they
680describe a region of memory, and all memory objects in LLVM are
681accessed through pointers.</p>
682
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]683<p>A global variable may be declared to reside in a target-specifc numbered
684address space. For targets that support them, address spaces may affect how
685optimizations are performed and/or what target instructions are used to access
Christopher Lamb20a39e92007-12-12 08:44:39 +0000[diff] [blame]686the variable. The default address space is zero. The address space qualifier
687must precede any other attributes.</p>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]688
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]689<p>LLVM allows an explicit section to be specified for globals. If the target
690supports it, it will emit globals to the section specified.</p>
691
692<p>An explicit alignment may be specified for a global. If not present, or if
693the alignment is set to zero, the alignment of the global is set by the target
694to whatever it feels convenient. If an explicit alignment is specified, the
695global is forced to have at least that much alignment. All alignments must be
696a power of 2.</p>
697
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]698<p>For example, the following defines a global in a numbered address space with
699an initializer, section, and alignment:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]700
701<div class="doc_code">
702<pre>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]703@G = constant float 1.0 addrspace(5), section "foo", align 4
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]704</pre>
705</div>
706
707</div>
708
709
710<!-- ======================================================================= -->
711<div class="doc_subsection">
712 <a name="functionstructure">Functions</a>
713</div>
714
715<div class="doc_text">
716
717<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
718an optional <a href="#linkage">linkage type</a>, an optional
719<a href="#visibility">visibility style</a>, an optional
720<a href="#callingconv">calling convention</a>, a return type, an optional
721<a href="#paramattrs">parameter attribute</a> for the return type, a function
722name, a (possibly empty) argument list (each with optional
723<a href="#paramattrs">parameter attributes</a>), an optional section, an
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]724optional alignment, an optional <a href="#gc">garbage collector name</a>, an
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]725opening curly brace, a list of basic blocks, and a closing curly brace.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]726
727LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
728optional <a href="#linkage">linkage type</a>, an optional
729<a href="#visibility">visibility style</a>, an optional
730<a href="#callingconv">calling convention</a>, a return type, an optional
731<a href="#paramattrs">parameter attribute</a> for the return type, a function
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]732name, a possibly empty list of arguments, an optional alignment, and an optional
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]733<a href="#gc">garbage collector name</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]734
735<p>A function definition contains a list of basic blocks, forming the CFG for
736the function. Each basic block may optionally start with a label (giving the
737basic block a symbol table entry), contains a list of instructions, and ends
738with a <a href="#terminators">terminator</a> instruction (such as a branch or
739function return).</p>
740
741<p>The first basic block in a function is special in two ways: it is immediately
742executed on entrance to the function, and it is not allowed to have predecessor
743basic blocks (i.e. there can not be any branches to the entry block of a
744function). Because the block can have no predecessors, it also cannot have any
745<a href="#i_phi">PHI nodes</a>.</p>
746
747<p>LLVM allows an explicit section to be specified for functions. If the target
748supports it, it will emit functions to the section specified.</p>
749
750<p>An explicit alignment may be specified for a function. If not present, or if
751the alignment is set to zero, the alignment of the function is set by the target
752to whatever it feels convenient. If an explicit alignment is specified, the
753function is forced to have at least that much alignment. All alignments must be
754a power of 2.</p>
755
756</div>
757
758
759<!-- ======================================================================= -->
760<div class="doc_subsection">
761 <a name="aliasstructure">Aliases</a>
762</div>
763<div class="doc_text">
764 <p>Aliases act as "second name" for the aliasee value (which can be either
765 function or global variable or bitcast of global value). Aliases may have an
766 optional <a href="#linkage">linkage type</a>, and an
767 optional <a href="#visibility">visibility style</a>.</p>
768
769 <h5>Syntax:</h5>
770
771<div class="doc_code">
772<pre>
773@&lt;Name&gt; = [Linkage] [Visibility] alias &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
774</pre>
775</div>
776
777</div>
778
779
780
781<!-- ======================================================================= -->
782<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
783<div class="doc_text">
784 <p>The return type and each parameter of a function type may have a set of
785 <i>parameter attributes</i> associated with them. Parameter attributes are
786 used to communicate additional information about the result or parameters of
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]787 a function. Parameter attributes are considered to be part of the function,
788 not of the function type, so functions with different parameter attributes
789 can have the same function type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]790
791 <p>Parameter attributes are simple keywords that follow the type specified. If
792 multiple parameter attributes are needed, they are space separated. For
793 example:</p>
794
795<div class="doc_code">
796<pre>
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]797declare i32 @printf(i8* noalias , ...) nounwind
798declare i32 @atoi(i8*) nounwind readonly
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]799</pre>
800</div>
801
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]802 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
803 <tt>readonly</tt>) come immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]804
805 <p>Currently, only the following parameter attributes are defined:</p>
806 <dl>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]807 <dt><tt>zeroext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]808 <dd>This indicates that the parameter should be zero extended just before
809 a call to this function.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]810
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]811 <dt><tt>signext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]812 <dd>This indicates that the parameter should be sign extended just before
813 a call to this function.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]814
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]815 <dt><tt>inreg</tt></dt>
816 <dd>This indicates that the parameter should be placed in register (if
817 possible) during assembling function call. Support for this attribute is
818 target-specific</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]819
820 <dt><tt>byval</tt></dt>
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]821 <dd>This indicates that the pointer parameter should really be passed by
822 value to the function. The attribute implies that a hidden copy of the
823 pointee is made between the caller and the callee, so the callee is unable
824 to modify the value in the callee. This attribute is only valid on llvm
825 pointer arguments. It is generally used to pass structs and arrays by
826 value, but is also valid on scalars (even though this is silly).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]827
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]828 <dt><tt>sret</tt></dt>
Duncan Sands616cc032008-02-18 04:19:38 +0000[diff] [blame]829 <dd>This indicates that the pointer parameter specifies the address of a
830 structure that is the return value of the function in the source program.
831 May only be applied to the first parameter.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]832
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]833 <dt><tt>noalias</tt></dt>
Owen Andersonc4fc4cd2008-02-18 04:09:01 +0000[diff] [blame]834 <dd>This indicates that the parameter does not alias any global or any other
835 parameter. The caller is responsible for ensuring that this is the case,
836 usually by placing the value in a stack allocation.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]837
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]838 <dt><tt>noreturn</tt></dt>
839 <dd>This function attribute indicates that the function never returns. This
840 indicates to LLVM that every call to this function should be treated as if
841 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]842
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]843 <dt><tt>nounwind</tt></dt>
844 <dd>This function attribute indicates that the function type does not use
845 the unwind instruction and does not allow stack unwinding to propagate
846 through it.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]847
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]848 <dt><tt>nest</tt></dt>
849 <dd>This indicates that the parameter can be excised using the
850 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
Duncan Sands13e13f82007-11-22 20:23:04 +0000[diff] [blame]851 <dt><tt>readonly</tt></dt>
Duncan Sandsd69c0e62007-11-14 21:14:02 +0000[diff] [blame]852 <dd>This function attribute indicates that the function has no side-effects
Duncan Sands13e13f82007-11-22 20:23:04 +0000[diff] [blame]853 except for producing a return value or throwing an exception. The value
854 returned must only depend on the function arguments and/or global variables.
855 It may use values obtained by dereferencing pointers.</dd>
856 <dt><tt>readnone</tt></dt>
857 <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]858 function, but in addition it is not allowed to dereference any pointer arguments
859 or global variables.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]860 </dl>
861
862</div>
863
864<!-- ======================================================================= -->
865<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]866 <a name="gc">Garbage Collector Names</a>
867</div>
868
869<div class="doc_text">
870<p>Each function may specify a garbage collector name, which is simply a
871string.</p>
872
873<div class="doc_code"><pre
874>define void @f() gc "name" { ...</pre></div>
875
876<p>The compiler declares the supported values of <i>name</i>. Specifying a
877collector which will cause the compiler to alter its output in order to support
878the named garbage collection algorithm.</p>
879</div>
880
881<!-- ======================================================================= -->
882<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]883 <a name="moduleasm">Module-Level Inline Assembly</a>
884</div>
885
886<div class="doc_text">
887<p>
888Modules may contain "module-level inline asm" blocks, which corresponds to the
889GCC "file scope inline asm" blocks. These blocks are internally concatenated by
890LLVM and treated as a single unit, but may be separated in the .ll file if
891desired. The syntax is very simple:
892</p>
893
894<div class="doc_code">
895<pre>
896module asm "inline asm code goes here"
897module asm "more can go here"
898</pre>
899</div>
900
901<p>The strings can contain any character by escaping non-printable characters.
902 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
903 for the number.
904</p>
905
906<p>
907 The inline asm code is simply printed to the machine code .s file when
908 assembly code is generated.
909</p>
910</div>
911
912<!-- ======================================================================= -->
913<div class="doc_subsection">
914 <a name="datalayout">Data Layout</a>
915</div>
916
917<div class="doc_text">
918<p>A module may specify a target specific data layout string that specifies how
919data is to be laid out in memory. The syntax for the data layout is simply:</p>
920<pre> target datalayout = "<i>layout specification</i>"</pre>
921<p>The <i>layout specification</i> consists of a list of specifications
922separated by the minus sign character ('-'). Each specification starts with a
923letter and may include other information after the letter to define some
924aspect of the data layout. The specifications accepted are as follows: </p>
925<dl>
926 <dt><tt>E</tt></dt>
927 <dd>Specifies that the target lays out data in big-endian form. That is, the
928 bits with the most significance have the lowest address location.</dd>
929 <dt><tt>e</tt></dt>
930 <dd>Specifies that hte target lays out data in little-endian form. That is,
931 the bits with the least significance have the lowest address location.</dd>
932 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
933 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
934 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
935 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
936 too.</dd>
937 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
938 <dd>This specifies the alignment for an integer type of a given bit
939 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
940 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
941 <dd>This specifies the alignment for a vector type of a given bit
942 <i>size</i>.</dd>
943 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
944 <dd>This specifies the alignment for a floating point type of a given bit
945 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
946 (double).</dd>
947 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
948 <dd>This specifies the alignment for an aggregate type of a given bit
949 <i>size</i>.</dd>
950</dl>
951<p>When constructing the data layout for a given target, LLVM starts with a
952default set of specifications which are then (possibly) overriden by the
953specifications in the <tt>datalayout</tt> keyword. The default specifications
954are given in this list:</p>
955<ul>
956 <li><tt>E</tt> - big endian</li>
957 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
958 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
959 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
960 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
961 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
962 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
963 alignment of 64-bits</li>
964 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
965 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
966 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
967 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
968 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
969</ul>
970<p>When llvm is determining the alignment for a given type, it uses the
971following rules:
972<ol>
973 <li>If the type sought is an exact match for one of the specifications, that
974 specification is used.</li>
975 <li>If no match is found, and the type sought is an integer type, then the
976 smallest integer type that is larger than the bitwidth of the sought type is
977 used. If none of the specifications are larger than the bitwidth then the the
978 largest integer type is used. For example, given the default specifications
979 above, the i7 type will use the alignment of i8 (next largest) while both
980 i65 and i256 will use the alignment of i64 (largest specified).</li>
981 <li>If no match is found, and the type sought is a vector type, then the
982 largest vector type that is smaller than the sought vector type will be used
983 as a fall back. This happens because <128 x double> can be implemented in
984 terms of 64 <2 x double>, for example.</li>
985</ol>
986</div>
987
988<!-- *********************************************************************** -->
989<div class="doc_section"> <a name="typesystem">Type System</a> </div>
990<!-- *********************************************************************** -->
991
992<div class="doc_text">
993
994<p>The LLVM type system is one of the most important features of the
995intermediate representation. Being typed enables a number of
996optimizations to be performed on the IR directly, without having to do
997extra analyses on the side before the transformation. A strong type
998system makes it easier to read the generated code and enables novel
999analyses and transformations that are not feasible to perform on normal
1000three address code representations.</p>
1001
1002</div>
1003
1004<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1005<div class="doc_subsection"> <a name="t_classifications">Type
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1006Classifications</a> </div>
1007<div class="doc_text">
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1008<p>The types fall into a few useful
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1009classifications:</p>
1010
1011<table border="1" cellspacing="0" cellpadding="4">
1012 <tbody>
1013 <tr><th>Classification</th><th>Types</th></tr>
1014 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1015 <td><a href="#t_integer">integer</a></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1016 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1017 </tr>
1018 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1019 <td><a href="#t_floating">floating point</a></td>
1020 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1021 </tr>
1022 <tr>
1023 <td><a name="t_firstclass">first class</a></td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1024 <td><a href="#t_integer">integer</a>,
1025 <a href="#t_floating">floating point</a>,
1026 <a href="#t_pointer">pointer</a>,
1027 <a href="#t_vector">vector</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1028 </td>
1029 </tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1030 <tr>
1031 <td><a href="#t_primitive">primitive</a></td>
1032 <td><a href="#t_label">label</a>,
1033 <a href="#t_void">void</a>,
1034 <a href="#t_integer">integer</a>,
1035 <a href="#t_floating">floating point</a>.</td>
1036 </tr>
1037 <tr>
1038 <td><a href="#t_derived">derived</a></td>
1039 <td><a href="#t_integer">integer</a>,
1040 <a href="#t_array">array</a>,
1041 <a href="#t_function">function</a>,
1042 <a href="#t_pointer">pointer</a>,
1043 <a href="#t_struct">structure</a>,
1044 <a href="#t_pstruct">packed structure</a>,
1045 <a href="#t_vector">vector</a>,
1046 <a href="#t_opaque">opaque</a>.
1047 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1048 </tbody>
1049</table>
1050
1051<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1052most important. Values of these types are the only ones which can be
1053produced by instructions, passed as arguments, or used as operands to
1054instructions. This means that all structures and arrays must be
1055manipulated either by pointer or by component.</p>
1056</div>
1057
1058<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1059<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1060
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1061<div class="doc_text">
1062<p>The primitive types are the fundamental building blocks of the LLVM
1063system.</p>
1064
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1065</div>
1066
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1067<!-- _______________________________________________________________________ -->
1068<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1069
1070<div class="doc_text">
1071 <table>
1072 <tbody>
1073 <tr><th>Type</th><th>Description</th></tr>
1074 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1075 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1076 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1077 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1078 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1079 </tbody>
1080 </table>
1081</div>
1082
1083<!-- _______________________________________________________________________ -->
1084<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1085
1086<div class="doc_text">
1087<h5>Overview:</h5>
1088<p>The void type does not represent any value and has no size.</p>
1089
1090<h5>Syntax:</h5>
1091
1092<pre>
1093 void
1094</pre>
1095</div>
1096
1097<!-- _______________________________________________________________________ -->
1098<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1099
1100<div class="doc_text">
1101<h5>Overview:</h5>
1102<p>The label type represents code labels.</p>
1103
1104<h5>Syntax:</h5>
1105
1106<pre>
1107 label
1108</pre>
1109</div>
1110
1111
1112<!-- ======================================================================= -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1113<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1114
1115<div class="doc_text">
1116
1117<p>The real power in LLVM comes from the derived types in the system.
1118This is what allows a programmer to represent arrays, functions,
1119pointers, and other useful types. Note that these derived types may be
1120recursive: For example, it is possible to have a two dimensional array.</p>
1121
1122</div>
1123
1124<!-- _______________________________________________________________________ -->
1125<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1126
1127<div class="doc_text">
1128
1129<h5>Overview:</h5>
1130<p>The integer type is a very simple derived type that simply specifies an
1131arbitrary bit width for the integer type desired. Any bit width from 1 bit to
11322^23-1 (about 8 million) can be specified.</p>
1133
1134<h5>Syntax:</h5>
1135
1136<pre>
1137 iN
1138</pre>
1139
1140<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1141value.</p>
1142
1143<h5>Examples:</h5>
1144<table class="layout">
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1145 <tbody>
1146 <tr>
1147 <td><tt>i1</tt></td>
1148 <td>a single-bit integer.</td>
1149 </tr><tr>
1150 <td><tt>i32</tt></td>
1151 <td>a 32-bit integer.</td>
1152 </tr><tr>
1153 <td><tt>i1942652</tt></td>
1154 <td>a really big integer of over 1 million bits.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1155 </tr>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1156 </tbody>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1157</table>
1158</div>
1159
1160<!-- _______________________________________________________________________ -->
1161<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1162
1163<div class="doc_text">
1164
1165<h5>Overview:</h5>
1166
1167<p>The array type is a very simple derived type that arranges elements
1168sequentially in memory. The array type requires a size (number of
1169elements) and an underlying data type.</p>
1170
1171<h5>Syntax:</h5>
1172
1173<pre>
1174 [&lt;# elements&gt; x &lt;elementtype&gt;]
1175</pre>
1176
1177<p>The number of elements is a constant integer value; elementtype may
1178be any type with a size.</p>
1179
1180<h5>Examples:</h5>
1181<table class="layout">
1182 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1183 <td class="left"><tt>[40 x i32]</tt></td>
1184 <td class="left">Array of 40 32-bit integer values.</td>
1185 </tr>
1186 <tr class="layout">
1187 <td class="left"><tt>[41 x i32]</tt></td>
1188 <td class="left">Array of 41 32-bit integer values.</td>
1189 </tr>
1190 <tr class="layout">
1191 <td class="left"><tt>[4 x i8]</tt></td>
1192 <td class="left">Array of 4 8-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1193 </tr>
1194</table>
1195<p>Here are some examples of multidimensional arrays:</p>
1196<table class="layout">
1197 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1198 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1199 <td class="left">3x4 array of 32-bit integer values.</td>
1200 </tr>
1201 <tr class="layout">
1202 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1203 <td class="left">12x10 array of single precision floating point values.</td>
1204 </tr>
1205 <tr class="layout">
1206 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1207 <td class="left">2x3x4 array of 16-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1208 </tr>
1209</table>
1210
1211<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1212length array. Normally, accesses past the end of an array are undefined in
1213LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1214As a special case, however, zero length arrays are recognized to be variable
1215length. This allows implementation of 'pascal style arrays' with the LLVM
1216type "{ i32, [0 x float]}", for example.</p>
1217
1218</div>
1219
1220<!-- _______________________________________________________________________ -->
1221<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1222<div class="doc_text">
1223<h5>Overview:</h5>
1224<p>The function type can be thought of as a function signature. It
1225consists of a return type and a list of formal parameter types.
1226Function types are usually used to build virtual function tables
1227(which are structures of pointers to functions), for indirect function
1228calls, and when defining a function.</p>
1229<p>
1230The return type of a function type cannot be an aggregate type.
1231</p>
1232<h5>Syntax:</h5>
1233<pre> &lt;returntype&gt; (&lt;parameter list&gt;)<br></pre>
1234<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1235specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1236which indicates that the function takes a variable number of arguments.
1237Variable argument functions can access their arguments with the <a
1238 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1239<h5>Examples:</h5>
1240<table class="layout">
1241 <tr class="layout">
1242 <td class="left"><tt>i32 (i32)</tt></td>
1243 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1244 </td>
1245 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1246 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1247 </tt></td>
1248 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1249 an <tt>i16</tt> that should be sign extended and a
1250 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1251 <tt>float</tt>.
1252 </td>
1253 </tr><tr class="layout">
1254 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1255 <td class="left">A vararg function that takes at least one
1256 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1257 which returns an integer. This is the signature for <tt>printf</tt> in
1258 LLVM.
1259 </td>
1260 </tr>
1261</table>
1262
1263</div>
1264<!-- _______________________________________________________________________ -->
1265<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1266<div class="doc_text">
1267<h5>Overview:</h5>
1268<p>The structure type is used to represent a collection of data members
1269together in memory. The packing of the field types is defined to match
1270the ABI of the underlying processor. The elements of a structure may
1271be any type that has a size.</p>
1272<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1273and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1274field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1275instruction.</p>
1276<h5>Syntax:</h5>
1277<pre> { &lt;type list&gt; }<br></pre>
1278<h5>Examples:</h5>
1279<table class="layout">
1280 <tr class="layout">
1281 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1282 <td class="left">A triple of three <tt>i32</tt> values</td>
1283 </tr><tr class="layout">
1284 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1285 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1286 second element is a <a href="#t_pointer">pointer</a> to a
1287 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1288 an <tt>i32</tt>.</td>
1289 </tr>
1290</table>
1291</div>
1292
1293<!-- _______________________________________________________________________ -->
1294<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1295</div>
1296<div class="doc_text">
1297<h5>Overview:</h5>
1298<p>The packed structure type is used to represent a collection of data members
1299together in memory. There is no padding between fields. Further, the alignment
1300of a packed structure is 1 byte. The elements of a packed structure may
1301be any type that has a size.</p>
1302<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1303and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1304field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1305instruction.</p>
1306<h5>Syntax:</h5>
1307<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1308<h5>Examples:</h5>
1309<table class="layout">
1310 <tr class="layout">
1311 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1312 <td class="left">A triple of three <tt>i32</tt> values</td>
1313 </tr><tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1314 <td class="left"><tt>&lt; { float, i32 (i32)* } &gt;</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1315 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1316 second element is a <a href="#t_pointer">pointer</a> to a
1317 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1318 an <tt>i32</tt>.</td>
1319 </tr>
1320</table>
1321</div>
1322
1323<!-- _______________________________________________________________________ -->
1324<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1325<div class="doc_text">
1326<h5>Overview:</h5>
1327<p>As in many languages, the pointer type represents a pointer or
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1328reference to another object, which must live in memory. Pointer types may have
1329an optional address space attribute defining the target-specific numbered
1330address space where the pointed-to object resides. The default address space is
1331zero.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1332<h5>Syntax:</h5>
1333<pre> &lt;type&gt; *<br></pre>
1334<h5>Examples:</h5>
1335<table class="layout">
1336 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1337 <td class="left"><tt>[4x i32]*</tt></td>
1338 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1339 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1340 </tr>
1341 <tr class="layout">
1342 <td class="left"><tt>i32 (i32 *) *</tt></td>
1343 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1344 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1345 <tt>i32</tt>.</td>
1346 </tr>
1347 <tr class="layout">
1348 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1349 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1350 that resides in address space #5.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1351 </tr>
1352</table>
1353</div>
1354
1355<!-- _______________________________________________________________________ -->
1356<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1357<div class="doc_text">
1358
1359<h5>Overview:</h5>
1360
1361<p>A vector type is a simple derived type that represents a vector
1362of elements. Vector types are used when multiple primitive data
1363are operated in parallel using a single instruction (SIMD).
1364A vector type requires a size (number of
1365elements) and an underlying primitive data type. Vectors must have a power
1366of two length (1, 2, 4, 8, 16 ...). Vector types are
1367considered <a href="#t_firstclass">first class</a>.</p>
1368
1369<h5>Syntax:</h5>
1370
1371<pre>
1372 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1373</pre>
1374
1375<p>The number of elements is a constant integer value; elementtype may
1376be any integer or floating point type.</p>
1377
1378<h5>Examples:</h5>
1379
1380<table class="layout">
1381 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1382 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1383 <td class="left">Vector of 4 32-bit integer values.</td>
1384 </tr>
1385 <tr class="layout">
1386 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1387 <td class="left">Vector of 8 32-bit floating-point values.</td>
1388 </tr>
1389 <tr class="layout">
1390 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1391 <td class="left">Vector of 2 64-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1392 </tr>
1393</table>
1394</div>
1395
1396<!-- _______________________________________________________________________ -->
1397<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1398<div class="doc_text">
1399
1400<h5>Overview:</h5>
1401
1402<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1403corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1404In LLVM, opaque types can eventually be resolved to any type (not just a
1405structure type).</p>
1406
1407<h5>Syntax:</h5>
1408
1409<pre>
1410 opaque
1411</pre>
1412
1413<h5>Examples:</h5>
1414
1415<table class="layout">
1416 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1417 <td class="left"><tt>opaque</tt></td>
1418 <td class="left">An opaque type.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1419 </tr>
1420</table>
1421</div>
1422
1423
1424<!-- *********************************************************************** -->
1425<div class="doc_section"> <a name="constants">Constants</a> </div>
1426<!-- *********************************************************************** -->
1427
1428<div class="doc_text">
1429
1430<p>LLVM has several different basic types of constants. This section describes
1431them all and their syntax.</p>
1432
1433</div>
1434
1435<!-- ======================================================================= -->
1436<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1437
1438<div class="doc_text">
1439
1440<dl>
1441 <dt><b>Boolean constants</b></dt>
1442
1443 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1444 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1445 </dd>
1446
1447 <dt><b>Integer constants</b></dt>
1448
1449 <dd>Standard integers (such as '4') are constants of the <a
1450 href="#t_integer">integer</a> type. Negative numbers may be used with
1451 integer types.
1452 </dd>
1453
1454 <dt><b>Floating point constants</b></dt>
1455
1456 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1457 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1458 notation (see below). Floating point constants must have a <a
1459 href="#t_floating">floating point</a> type. </dd>
1460
1461 <dt><b>Null pointer constants</b></dt>
1462
1463 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1464 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1465
1466</dl>
1467
1468<p>The one non-intuitive notation for constants is the optional hexadecimal form
1469of floating point constants. For example, the form '<tt>double
14700x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
14714.5e+15</tt>'. The only time hexadecimal floating point constants are required
1472(and the only time that they are generated by the disassembler) is when a
1473floating point constant must be emitted but it cannot be represented as a
1474decimal floating point number. For example, NaN's, infinities, and other
1475special values are represented in their IEEE hexadecimal format so that
1476assembly and disassembly do not cause any bits to change in the constants.</p>
1477
1478</div>
1479
1480<!-- ======================================================================= -->
1481<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1482</div>
1483
1484<div class="doc_text">
1485<p>Aggregate constants arise from aggregation of simple constants
1486and smaller aggregate constants.</p>
1487
1488<dl>
1489 <dt><b>Structure constants</b></dt>
1490
1491 <dd>Structure constants are represented with notation similar to structure
1492 type definitions (a comma separated list of elements, surrounded by braces
Chris Lattner6c8de962007-12-25 20:34:52 +0000[diff] [blame]1493 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1494 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1495 must have <a href="#t_struct">structure type</a>, and the number and
1496 types of elements must match those specified by the type.
1497 </dd>
1498
1499 <dt><b>Array constants</b></dt>
1500
1501 <dd>Array constants are represented with notation similar to array type
1502 definitions (a comma separated list of elements, surrounded by square brackets
1503 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1504 constants must have <a href="#t_array">array type</a>, and the number and
1505 types of elements must match those specified by the type.
1506 </dd>
1507
1508 <dt><b>Vector constants</b></dt>
1509
1510 <dd>Vector constants are represented with notation similar to vector type
1511 definitions (a comma separated list of elements, surrounded by
1512 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1513 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1514 href="#t_vector">vector type</a>, and the number and types of elements must
1515 match those specified by the type.
1516 </dd>
1517
1518 <dt><b>Zero initialization</b></dt>
1519
1520 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1521 value to zero of <em>any</em> type, including scalar and aggregate types.
1522 This is often used to avoid having to print large zero initializers (e.g. for
1523 large arrays) and is always exactly equivalent to using explicit zero
1524 initializers.
1525 </dd>
1526</dl>
1527
1528</div>
1529
1530<!-- ======================================================================= -->
1531<div class="doc_subsection">
1532 <a name="globalconstants">Global Variable and Function Addresses</a>
1533</div>
1534
1535<div class="doc_text">
1536
1537<p>The addresses of <a href="#globalvars">global variables</a> and <a
1538href="#functionstructure">functions</a> are always implicitly valid (link-time)
1539constants. These constants are explicitly referenced when the <a
1540href="#identifiers">identifier for the global</a> is used and always have <a
1541href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1542file:</p>
1543
1544<div class="doc_code">
1545<pre>
1546@X = global i32 17
1547@Y = global i32 42
1548@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1549</pre>
1550</div>
1551
1552</div>
1553
1554<!-- ======================================================================= -->
1555<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1556<div class="doc_text">
1557 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1558 no specific value. Undefined values may be of any type and be used anywhere
1559 a constant is permitted.</p>
1560
1561 <p>Undefined values indicate to the compiler that the program is well defined
1562 no matter what value is used, giving the compiler more freedom to optimize.
1563 </p>
1564</div>
1565
1566<!-- ======================================================================= -->
1567<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1568</div>
1569
1570<div class="doc_text">
1571
1572<p>Constant expressions are used to allow expressions involving other constants
1573to be used as constants. Constant expressions may be of any <a
1574href="#t_firstclass">first class</a> type and may involve any LLVM operation
1575that does not have side effects (e.g. load and call are not supported). The
1576following is the syntax for constant expressions:</p>
1577
1578<dl>
1579 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1580 <dd>Truncate a constant to another type. The bit size of CST must be larger
1581 than the bit size of TYPE. Both types must be integers.</dd>
1582
1583 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1584 <dd>Zero extend a constant to another type. The bit size of CST must be
1585 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1586
1587 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1588 <dd>Sign extend a constant to another type. The bit size of CST must be
1589 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1590
1591 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1592 <dd>Truncate a floating point constant to another floating point type. The
1593 size of CST must be larger than the size of TYPE. Both types must be
1594 floating point.</dd>
1595
1596 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1597 <dd>Floating point extend a constant to another type. The size of CST must be
1598 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1599
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1600 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1601 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1602 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1603 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1604 of the same number of elements. If the value won't fit in the integer type,
1605 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1606
1607 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1608 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1609 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1610 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1611 of the same number of elements. If the value won't fit in the integer type,
1612 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1613
1614 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1615 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1616 constant. TYPE must be a scalar or vector floating point type. CST must be of
1617 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1618 of the same number of elements. If the value won't fit in the floating point
1619 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1620
1621 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1622 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1623 constant. TYPE must be a scalar or vector floating point type. CST must be of
1624 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1625 of the same number of elements. If the value won't fit in the floating point
1626 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1627
1628 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1629 <dd>Convert a pointer typed constant to the corresponding integer constant
1630 TYPE must be an integer type. CST must be of pointer type. The CST value is
1631 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1632
1633 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1634 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1635 pointer type. CST must be of integer type. The CST value is zero extended,
1636 truncated, or unchanged to make it fit in a pointer size. This one is
1637 <i>really</i> dangerous!</dd>
1638
1639 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1640 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1641 identical (same number of bits). The conversion is done as if the CST value
1642 was stored to memory and read back as TYPE. In other words, no bits change
1643 with this operator, just the type. This can be used for conversion of
1644 vector types to any other type, as long as they have the same bit width. For
1645 pointers it is only valid to cast to another pointer type.
1646 </dd>
1647
1648 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1649
1650 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1651 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1652 instruction, the index list may have zero or more indexes, which are required
1653 to make sense for the type of "CSTPTR".</dd>
1654
1655 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1656
1657 <dd>Perform the <a href="#i_select">select operation</a> on
1658 constants.</dd>
1659
1660 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1661 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1662
1663 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1664 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1665
1666 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1667
1668 <dd>Perform the <a href="#i_extractelement">extractelement
1669 operation</a> on constants.
1670
1671 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1672
1673 <dd>Perform the <a href="#i_insertelement">insertelement
1674 operation</a> on constants.</dd>
1675
1676
1677 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1678
1679 <dd>Perform the <a href="#i_shufflevector">shufflevector
1680 operation</a> on constants.</dd>
1681
1682 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1683
1684 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1685 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1686 binary</a> operations. The constraints on operands are the same as those for
1687 the corresponding instruction (e.g. no bitwise operations on floating point
1688 values are allowed).</dd>
1689</dl>
1690</div>
1691
1692<!-- *********************************************************************** -->
1693<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1694<!-- *********************************************************************** -->
1695
1696<!-- ======================================================================= -->
1697<div class="doc_subsection">
1698<a name="inlineasm">Inline Assembler Expressions</a>
1699</div>
1700
1701<div class="doc_text">
1702
1703<p>
1704LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1705Module-Level Inline Assembly</a>) through the use of a special value. This
1706value represents the inline assembler as a string (containing the instructions
1707to emit), a list of operand constraints (stored as a string), and a flag that
1708indicates whether or not the inline asm expression has side effects. An example
1709inline assembler expression is:
1710</p>
1711
1712<div class="doc_code">
1713<pre>
1714i32 (i32) asm "bswap $0", "=r,r"
1715</pre>
1716</div>
1717
1718<p>
1719Inline assembler expressions may <b>only</b> be used as the callee operand of
1720a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1721</p>
1722
1723<div class="doc_code">
1724<pre>
1725%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1726</pre>
1727</div>
1728
1729<p>
1730Inline asms with side effects not visible in the constraint list must be marked
1731as having side effects. This is done through the use of the
1732'<tt>sideeffect</tt>' keyword, like so:
1733</p>
1734
1735<div class="doc_code">
1736<pre>
1737call void asm sideeffect "eieio", ""()
1738</pre>
1739</div>
1740
1741<p>TODO: The format of the asm and constraints string still need to be
1742documented here. Constraints on what can be done (e.g. duplication, moving, etc
1743need to be documented).
1744</p>
1745
1746</div>
1747
1748<!-- *********************************************************************** -->
1749<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1750<!-- *********************************************************************** -->
1751
1752<div class="doc_text">
1753
1754<p>The LLVM instruction set consists of several different
1755classifications of instructions: <a href="#terminators">terminator
1756instructions</a>, <a href="#binaryops">binary instructions</a>,
1757<a href="#bitwiseops">bitwise binary instructions</a>, <a
1758 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1759instructions</a>.</p>
1760
1761</div>
1762
1763<!-- ======================================================================= -->
1764<div class="doc_subsection"> <a name="terminators">Terminator
1765Instructions</a> </div>
1766
1767<div class="doc_text">
1768
1769<p>As mentioned <a href="#functionstructure">previously</a>, every
1770basic block in a program ends with a "Terminator" instruction, which
1771indicates which block should be executed after the current block is
1772finished. These terminator instructions typically yield a '<tt>void</tt>'
1773value: they produce control flow, not values (the one exception being
1774the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1775<p>There are six different terminator instructions: the '<a
1776 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1777instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1778the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1779 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1780 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1781
1782</div>
1783
1784<!-- _______________________________________________________________________ -->
1785<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1786Instruction</a> </div>
1787<div class="doc_text">
1788<h5>Syntax:</h5>
1789<pre> ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
1790 ret void <i>; Return from void function</i>
1791</pre>
1792<h5>Overview:</h5>
1793<p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1794value) from a function back to the caller.</p>
1795<p>There are two forms of the '<tt>ret</tt>' instruction: one that
1796returns a value and then causes control flow, and one that just causes
1797control flow to occur.</p>
1798<h5>Arguments:</h5>
1799<p>The '<tt>ret</tt>' instruction may return any '<a
1800 href="#t_firstclass">first class</a>' type. Notice that a function is
1801not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1802instruction inside of the function that returns a value that does not
1803match the return type of the function.</p>
1804<h5>Semantics:</h5>
1805<p>When the '<tt>ret</tt>' instruction is executed, control flow
1806returns back to the calling function's context. If the caller is a "<a
1807 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1808the instruction after the call. If the caller was an "<a
1809 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1810at the beginning of the "normal" destination block. If the instruction
1811returns a value, that value shall set the call or invoke instruction's
1812return value.</p>
1813<h5>Example:</h5>
1814<pre> ret i32 5 <i>; Return an integer value of 5</i>
1815 ret void <i>; Return from a void function</i>
1816</pre>
1817</div>
1818<!-- _______________________________________________________________________ -->
1819<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1820<div class="doc_text">
1821<h5>Syntax:</h5>
1822<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
1823</pre>
1824<h5>Overview:</h5>
1825<p>The '<tt>br</tt>' instruction is used to cause control flow to
1826transfer to a different basic block in the current function. There are
1827two forms of this instruction, corresponding to a conditional branch
1828and an unconditional branch.</p>
1829<h5>Arguments:</h5>
1830<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1831single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1832unconditional form of the '<tt>br</tt>' instruction takes a single
1833'<tt>label</tt>' value as a target.</p>
1834<h5>Semantics:</h5>
1835<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1836argument is evaluated. If the value is <tt>true</tt>, control flows
1837to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1838control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1839<h5>Example:</h5>
1840<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
1841 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1842</div>
1843<!-- _______________________________________________________________________ -->
1844<div class="doc_subsubsection">
1845 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1846</div>
1847
1848<div class="doc_text">
1849<h5>Syntax:</h5>
1850
1851<pre>
1852 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
1853</pre>
1854
1855<h5>Overview:</h5>
1856
1857<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1858several different places. It is a generalization of the '<tt>br</tt>'
1859instruction, allowing a branch to occur to one of many possible
1860destinations.</p>
1861
1862
1863<h5>Arguments:</h5>
1864
1865<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1866comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1867an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1868table is not allowed to contain duplicate constant entries.</p>
1869
1870<h5>Semantics:</h5>
1871
1872<p>The <tt>switch</tt> instruction specifies a table of values and
1873destinations. When the '<tt>switch</tt>' instruction is executed, this
1874table is searched for the given value. If the value is found, control flow is
1875transfered to the corresponding destination; otherwise, control flow is
1876transfered to the default destination.</p>
1877
1878<h5>Implementation:</h5>
1879
1880<p>Depending on properties of the target machine and the particular
1881<tt>switch</tt> instruction, this instruction may be code generated in different
1882ways. For example, it could be generated as a series of chained conditional
1883branches or with a lookup table.</p>
1884
1885<h5>Example:</h5>
1886
1887<pre>
1888 <i>; Emulate a conditional br instruction</i>
1889 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1890 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1891
1892 <i>; Emulate an unconditional br instruction</i>
1893 switch i32 0, label %dest [ ]
1894
1895 <i>; Implement a jump table:</i>
1896 switch i32 %val, label %otherwise [ i32 0, label %onzero
1897 i32 1, label %onone
1898 i32 2, label %ontwo ]
1899</pre>
1900</div>
1901
1902<!-- _______________________________________________________________________ -->
1903<div class="doc_subsubsection">
1904 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1905</div>
1906
1907<div class="doc_text">
1908
1909<h5>Syntax:</h5>
1910
1911<pre>
1912 &lt;result&gt; = invoke [<a href="#callingconv">cconv</a>] &lt;ptr to function ty&gt; %&lt;function ptr val&gt;(&lt;function args&gt;)
1913 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
1914</pre>
1915
1916<h5>Overview:</h5>
1917
1918<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1919function, with the possibility of control flow transfer to either the
1920'<tt>normal</tt>' label or the
1921'<tt>exception</tt>' label. If the callee function returns with the
1922"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1923"normal" label. If the callee (or any indirect callees) returns with the "<a
1924href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1925continued at the dynamically nearest "exception" label.</p>
1926
1927<h5>Arguments:</h5>
1928
1929<p>This instruction requires several arguments:</p>
1930
1931<ol>
1932 <li>
1933 The optional "cconv" marker indicates which <a href="#callingconv">calling
1934 convention</a> the call should use. If none is specified, the call defaults
1935 to using C calling conventions.
1936 </li>
1937 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1938 function value being invoked. In most cases, this is a direct function
1939 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1940 an arbitrary pointer to function value.
1941 </li>
1942
1943 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1944 function to be invoked. </li>
1945
1946 <li>'<tt>function args</tt>': argument list whose types match the function
1947 signature argument types. If the function signature indicates the function
1948 accepts a variable number of arguments, the extra arguments can be
1949 specified. </li>
1950
1951 <li>'<tt>normal label</tt>': the label reached when the called function
1952 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1953
1954 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1955 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1956
1957</ol>
1958
1959<h5>Semantics:</h5>
1960
1961<p>This instruction is designed to operate as a standard '<tt><a
1962href="#i_call">call</a></tt>' instruction in most regards. The primary
1963difference is that it establishes an association with a label, which is used by
1964the runtime library to unwind the stack.</p>
1965
1966<p>This instruction is used in languages with destructors to ensure that proper
1967cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1968exception. Additionally, this is important for implementation of
1969'<tt>catch</tt>' clauses in high-level languages that support them.</p>
1970
1971<h5>Example:</h5>
1972<pre>
1973 %retval = invoke i32 %Test(i32 15) to label %Continue
1974 unwind label %TestCleanup <i>; {i32}:retval set</i>
1975 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1976 unwind label %TestCleanup <i>; {i32}:retval set</i>
1977</pre>
1978</div>
1979
1980
1981<!-- _______________________________________________________________________ -->
1982
1983<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1984Instruction</a> </div>
1985
1986<div class="doc_text">
1987
1988<h5>Syntax:</h5>
1989<pre>
1990 unwind
1991</pre>
1992
1993<h5>Overview:</h5>
1994
1995<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1996at the first callee in the dynamic call stack which used an <a
1997href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1998primarily used to implement exception handling.</p>
1999
2000<h5>Semantics:</h5>
2001
2002<p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
2003immediately halt. The dynamic call stack is then searched for the first <a
2004href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2005execution continues at the "exceptional" destination block specified by the
2006<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2007dynamic call chain, undefined behavior results.</p>
2008</div>
2009
2010<!-- _______________________________________________________________________ -->
2011
2012<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2013Instruction</a> </div>
2014
2015<div class="doc_text">
2016
2017<h5>Syntax:</h5>
2018<pre>
2019 unreachable
2020</pre>
2021
2022<h5>Overview:</h5>
2023
2024<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2025instruction is used to inform the optimizer that a particular portion of the
2026code is not reachable. This can be used to indicate that the code after a
2027no-return function cannot be reached, and other facts.</p>
2028
2029<h5>Semantics:</h5>
2030
2031<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2032</div>
2033
2034
2035
2036<!-- ======================================================================= -->
2037<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2038<div class="doc_text">
2039<p>Binary operators are used to do most of the computation in a
2040program. They require two operands, execute an operation on them, and
2041produce a single value. The operands might represent
2042multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
2043The result value of a binary operator is not
2044necessarily the same type as its operands.</p>
2045<p>There are several different binary operators:</p>
2046</div>
2047<!-- _______________________________________________________________________ -->
2048<div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
2049Instruction</a> </div>
2050<div class="doc_text">
2051<h5>Syntax:</h5>
2052<pre> &lt;result&gt; = add &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2053</pre>
2054<h5>Overview:</h5>
2055<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2056<h5>Arguments:</h5>
2057<p>The two arguments to the '<tt>add</tt>' instruction must be either <a
2058 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
2059 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2060Both arguments must have identical types.</p>
2061<h5>Semantics:</h5>
2062<p>The value produced is the integer or floating point sum of the two
2063operands.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2064<p>If an integer sum has unsigned overflow, the result returned is the
2065mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2066the result.</p>
2067<p>Because LLVM integers use a two's complement representation, this
2068instruction is appropriate for both signed and unsigned integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2069<h5>Example:</h5>
2070<pre> &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2071</pre>
2072</div>
2073<!-- _______________________________________________________________________ -->
2074<div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
2075Instruction</a> </div>
2076<div class="doc_text">
2077<h5>Syntax:</h5>
2078<pre> &lt;result&gt; = sub &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2079</pre>
2080<h5>Overview:</h5>
2081<p>The '<tt>sub</tt>' instruction returns the difference of its two
2082operands.</p>
2083<p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
2084instruction present in most other intermediate representations.</p>
2085<h5>Arguments:</h5>
2086<p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
2087 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2088values.
2089This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2090Both arguments must have identical types.</p>
2091<h5>Semantics:</h5>
2092<p>The value produced is the integer or floating point difference of
2093the two operands.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2094<p>If an integer difference has unsigned overflow, the result returned is the
2095mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2096the result.</p>
2097<p>Because LLVM integers use a two's complement representation, this
2098instruction is appropriate for both signed and unsigned integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2099<h5>Example:</h5>
2100<pre>
2101 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2102 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2103</pre>
2104</div>
2105<!-- _______________________________________________________________________ -->
2106<div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
2107Instruction</a> </div>
2108<div class="doc_text">
2109<h5>Syntax:</h5>
2110<pre> &lt;result&gt; = mul &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2111</pre>
2112<h5>Overview:</h5>
2113<p>The '<tt>mul</tt>' instruction returns the product of its two
2114operands.</p>
2115<h5>Arguments:</h5>
2116<p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
2117 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2118values.
2119This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2120Both arguments must have identical types.</p>
2121<h5>Semantics:</h5>
2122<p>The value produced is the integer or floating point product of the
2123two operands.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2124<p>If the result of an integer multiplication has unsigned overflow,
2125the result returned is the mathematical result modulo
21262<sup>n</sup>, where n is the bit width of the result.</p>
2127<p>Because LLVM integers use a two's complement representation, and the
2128result is the same width as the operands, this instruction returns the
2129correct result for both signed and unsigned integers. If a full product
2130(e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2131should be sign-extended or zero-extended as appropriate to the
2132width of the full product.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2133<h5>Example:</h5>
2134<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2135</pre>
2136</div>
2137<!-- _______________________________________________________________________ -->
2138<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2139</a></div>
2140<div class="doc_text">
2141<h5>Syntax:</h5>
2142<pre> &lt;result&gt; = udiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2143</pre>
2144<h5>Overview:</h5>
2145<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2146operands.</p>
2147<h5>Arguments:</h5>
2148<p>The two arguments to the '<tt>udiv</tt>' instruction must be
2149<a href="#t_integer">integer</a> values. Both arguments must have identical
2150types. This instruction can also take <a href="#t_vector">vector</a> versions
2151of the values in which case the elements must be integers.</p>
2152<h5>Semantics:</h5>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2153<p>The value produced is the unsigned integer quotient of the two operands.</p>
2154<p>Note that unsigned integer division and signed integer division are distinct
2155operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2156<p>Division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2157<h5>Example:</h5>
2158<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2159</pre>
2160</div>
2161<!-- _______________________________________________________________________ -->
2162<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2163</a> </div>
2164<div class="doc_text">
2165<h5>Syntax:</h5>
2166<pre> &lt;result&gt; = sdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2167</pre>
2168<h5>Overview:</h5>
2169<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2170operands.</p>
2171<h5>Arguments:</h5>
2172<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2173<a href="#t_integer">integer</a> values. Both arguments must have identical
2174types. This instruction can also take <a href="#t_vector">vector</a> versions
2175of the values in which case the elements must be integers.</p>
2176<h5>Semantics:</h5>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2177<p>The value produced is the signed integer quotient of the two operands.</p>
2178<p>Note that signed integer division and unsigned integer division are distinct
2179operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2180<p>Division by zero leads to undefined behavior. Overflow also leads to
2181undefined behavior; this is a rare case, but can occur, for example,
2182by doing a 32-bit division of -2147483648 by -1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2183<h5>Example:</h5>
2184<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2185</pre>
2186</div>
2187<!-- _______________________________________________________________________ -->
2188<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2189Instruction</a> </div>
2190<div class="doc_text">
2191<h5>Syntax:</h5>
2192<pre> &lt;result&gt; = fdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2193</pre>
2194<h5>Overview:</h5>
2195<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2196operands.</p>
2197<h5>Arguments:</h5>
2198<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2199<a href="#t_floating">floating point</a> values. Both arguments must have
2200identical types. This instruction can also take <a href="#t_vector">vector</a>
2201versions of floating point values.</p>
2202<h5>Semantics:</h5>
2203<p>The value produced is the floating point quotient of the two operands.</p>
2204<h5>Example:</h5>
2205<pre> &lt;result&gt; = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2206</pre>
2207</div>
2208<!-- _______________________________________________________________________ -->
2209<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2210</div>
2211<div class="doc_text">
2212<h5>Syntax:</h5>
2213<pre> &lt;result&gt; = urem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2214</pre>
2215<h5>Overview:</h5>
2216<p>The '<tt>urem</tt>' instruction returns the remainder from the
2217unsigned division of its two arguments.</p>
2218<h5>Arguments:</h5>
2219<p>The two arguments to the '<tt>urem</tt>' instruction must be
2220<a href="#t_integer">integer</a> values. Both arguments must have identical
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2221types. This instruction can also take <a href="#t_vector">vector</a> versions
2222of the values in which case the elements must be integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2223<h5>Semantics:</h5>
2224<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2225This instruction always performs an unsigned division to get the remainder,
2226regardless of whether the arguments are unsigned or not.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2227<p>Note that unsigned integer remainder and signed integer remainder are
2228distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2229<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2230<h5>Example:</h5>
2231<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2232</pre>
2233
2234</div>
2235<!-- _______________________________________________________________________ -->
2236<div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2237Instruction</a> </div>
2238<div class="doc_text">
2239<h5>Syntax:</h5>
2240<pre> &lt;result&gt; = srem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2241</pre>
2242<h5>Overview:</h5>
2243<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2244signed division of its two operands. This instruction can also take
2245<a href="#t_vector">vector</a> versions of the values in which case
2246the elements must be integers.</p>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]2247
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2248<h5>Arguments:</h5>
2249<p>The two arguments to the '<tt>srem</tt>' instruction must be
2250<a href="#t_integer">integer</a> values. Both arguments must have identical
2251types.</p>
2252<h5>Semantics:</h5>
2253<p>This instruction returns the <i>remainder</i> of a division (where the result
2254has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2255operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2256a value. For more information about the difference, see <a
2257 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2258Math Forum</a>. For a table of how this is implemented in various languages,
2259please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2260Wikipedia: modulo operation</a>.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2261<p>Note that signed integer remainder and unsigned integer remainder are
2262distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2263<p>Taking the remainder of a division by zero leads to undefined behavior.
2264Overflow also leads to undefined behavior; this is a rare case, but can occur,
2265for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2266(The remainder doesn't actually overflow, but this rule lets srem be
2267implemented using instructions that return both the result of the division
2268and the remainder.)</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2269<h5>Example:</h5>
2270<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2271</pre>
2272
2273</div>
2274<!-- _______________________________________________________________________ -->
2275<div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2276Instruction</a> </div>
2277<div class="doc_text">
2278<h5>Syntax:</h5>
2279<pre> &lt;result&gt; = frem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2280</pre>
2281<h5>Overview:</h5>
2282<p>The '<tt>frem</tt>' instruction returns the remainder from the
2283division of its two operands.</p>
2284<h5>Arguments:</h5>
2285<p>The two arguments to the '<tt>frem</tt>' instruction must be
2286<a href="#t_floating">floating point</a> values. Both arguments must have
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2287identical types. This instruction can also take <a href="#t_vector">vector</a>
2288versions of floating point values.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2289<h5>Semantics:</h5>
2290<p>This instruction returns the <i>remainder</i> of a division.</p>
2291<h5>Example:</h5>
2292<pre> &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2293</pre>
2294</div>
2295
2296<!-- ======================================================================= -->
2297<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2298Operations</a> </div>
2299<div class="doc_text">
2300<p>Bitwise binary operators are used to do various forms of
2301bit-twiddling in a program. They are generally very efficient
2302instructions and can commonly be strength reduced from other
2303instructions. They require two operands, execute an operation on them,
2304and produce a single value. The resulting value of the bitwise binary
2305operators is always the same type as its first operand.</p>
2306</div>
2307
2308<!-- _______________________________________________________________________ -->
2309<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2310Instruction</a> </div>
2311<div class="doc_text">
2312<h5>Syntax:</h5>
2313<pre> &lt;result&gt; = shl &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2314</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2315
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2316<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2317
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2318<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2319the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2320
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2321<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2322
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2323<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2324 href="#t_integer">integer</a> type.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2325
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2326<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2327
2328<p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>. If
2329<tt>var2</tt> is (statically or dynamically) equal to or larger than the number
2330of bits in <tt>var1</tt>, the result is undefined.</p>
2331
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2332<h5>Example:</h5><pre>
2333 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2334 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2335 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2336 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2337</pre>
2338</div>
2339<!-- _______________________________________________________________________ -->
2340<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2341Instruction</a> </div>
2342<div class="doc_text">
2343<h5>Syntax:</h5>
2344<pre> &lt;result&gt; = lshr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2345</pre>
2346
2347<h5>Overview:</h5>
2348<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2349operand shifted to the right a specified number of bits with zero fill.</p>
2350
2351<h5>Arguments:</h5>
2352<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2353<a href="#t_integer">integer</a> type.</p>
2354
2355<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2356
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2357<p>This instruction always performs a logical shift right operation. The most
2358significant bits of the result will be filled with zero bits after the
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2359shift. If <tt>var2</tt> is (statically or dynamically) equal to or larger than
2360the number of bits in <tt>var1</tt>, the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2361
2362<h5>Example:</h5>
2363<pre>
2364 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2365 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2366 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2367 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2368 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2369</pre>
2370</div>
2371
2372<!-- _______________________________________________________________________ -->
2373<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2374Instruction</a> </div>
2375<div class="doc_text">
2376
2377<h5>Syntax:</h5>
2378<pre> &lt;result&gt; = ashr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2379</pre>
2380
2381<h5>Overview:</h5>
2382<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2383operand shifted to the right a specified number of bits with sign extension.</p>
2384
2385<h5>Arguments:</h5>
2386<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2387<a href="#t_integer">integer</a> type.</p>
2388
2389<h5>Semantics:</h5>
2390<p>This instruction always performs an arithmetic shift right operation,
2391The most significant bits of the result will be filled with the sign bit
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2392of <tt>var1</tt>. If <tt>var2</tt> is (statically or dynamically) equal to or
2393larger than the number of bits in <tt>var1</tt>, the result is undefined.
2394</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2395
2396<h5>Example:</h5>
2397<pre>
2398 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2399 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2400 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2401 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2402 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2403</pre>
2404</div>
2405
2406<!-- _______________________________________________________________________ -->
2407<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2408Instruction</a> </div>
2409<div class="doc_text">
2410<h5>Syntax:</h5>
2411<pre> &lt;result&gt; = and &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2412</pre>
2413<h5>Overview:</h5>
2414<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2415its two operands.</p>
2416<h5>Arguments:</h5>
2417<p>The two arguments to the '<tt>and</tt>' instruction must be <a
2418 href="#t_integer">integer</a> values. Both arguments must have
2419identical types.</p>
2420<h5>Semantics:</h5>
2421<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2422<p> </p>
2423<div style="align: center">
2424<table border="1" cellspacing="0" cellpadding="4">
2425 <tbody>
2426 <tr>
2427 <td>In0</td>
2428 <td>In1</td>
2429 <td>Out</td>
2430 </tr>
2431 <tr>
2432 <td>0</td>
2433 <td>0</td>
2434 <td>0</td>
2435 </tr>
2436 <tr>
2437 <td>0</td>
2438 <td>1</td>
2439 <td>0</td>
2440 </tr>
2441 <tr>
2442 <td>1</td>
2443 <td>0</td>
2444 <td>0</td>
2445 </tr>
2446 <tr>
2447 <td>1</td>
2448 <td>1</td>
2449 <td>1</td>
2450 </tr>
2451 </tbody>
2452</table>
2453</div>
2454<h5>Example:</h5>
2455<pre> &lt;result&gt; = and i32 4, %var <i>; yields {i32}:result = 4 &amp; %var</i>
2456 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2457 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2458</pre>
2459</div>
2460<!-- _______________________________________________________________________ -->
2461<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2462<div class="doc_text">
2463<h5>Syntax:</h5>
2464<pre> &lt;result&gt; = or &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2465</pre>
2466<h5>Overview:</h5>
2467<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2468or of its two operands.</p>
2469<h5>Arguments:</h5>
2470<p>The two arguments to the '<tt>or</tt>' instruction must be <a
2471 href="#t_integer">integer</a> values. Both arguments must have
2472identical types.</p>
2473<h5>Semantics:</h5>
2474<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2475<p> </p>
2476<div style="align: center">
2477<table border="1" cellspacing="0" cellpadding="4">
2478 <tbody>
2479 <tr>
2480 <td>In0</td>
2481 <td>In1</td>
2482 <td>Out</td>
2483 </tr>
2484 <tr>
2485 <td>0</td>
2486 <td>0</td>
2487 <td>0</td>
2488 </tr>
2489 <tr>
2490 <td>0</td>
2491 <td>1</td>
2492 <td>1</td>
2493 </tr>
2494 <tr>
2495 <td>1</td>
2496 <td>0</td>
2497 <td>1</td>
2498 </tr>
2499 <tr>
2500 <td>1</td>
2501 <td>1</td>
2502 <td>1</td>
2503 </tr>
2504 </tbody>
2505</table>
2506</div>
2507<h5>Example:</h5>
2508<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2509 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2510 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2511</pre>
2512</div>
2513<!-- _______________________________________________________________________ -->
2514<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2515Instruction</a> </div>
2516<div class="doc_text">
2517<h5>Syntax:</h5>
2518<pre> &lt;result&gt; = xor &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2519</pre>
2520<h5>Overview:</h5>
2521<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2522or of its two operands. The <tt>xor</tt> is used to implement the
2523"one's complement" operation, which is the "~" operator in C.</p>
2524<h5>Arguments:</h5>
2525<p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2526 href="#t_integer">integer</a> values. Both arguments must have
2527identical types.</p>
2528<h5>Semantics:</h5>
2529<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2530<p> </p>
2531<div style="align: center">
2532<table border="1" cellspacing="0" cellpadding="4">
2533 <tbody>
2534 <tr>
2535 <td>In0</td>
2536 <td>In1</td>
2537 <td>Out</td>
2538 </tr>
2539 <tr>
2540 <td>0</td>
2541 <td>0</td>
2542 <td>0</td>
2543 </tr>
2544 <tr>
2545 <td>0</td>
2546 <td>1</td>
2547 <td>1</td>
2548 </tr>
2549 <tr>
2550 <td>1</td>
2551 <td>0</td>
2552 <td>1</td>
2553 </tr>
2554 <tr>
2555 <td>1</td>
2556 <td>1</td>
2557 <td>0</td>
2558 </tr>
2559 </tbody>
2560</table>
2561</div>
2562<p> </p>
2563<h5>Example:</h5>
2564<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2565 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2566 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2567 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2568</pre>
2569</div>
2570
2571<!-- ======================================================================= -->
2572<div class="doc_subsection">
2573 <a name="vectorops">Vector Operations</a>
2574</div>
2575
2576<div class="doc_text">
2577
2578<p>LLVM supports several instructions to represent vector operations in a
2579target-independent manner. These instructions cover the element-access and
2580vector-specific operations needed to process vectors effectively. While LLVM
2581does directly support these vector operations, many sophisticated algorithms
2582will want to use target-specific intrinsics to take full advantage of a specific
2583target.</p>
2584
2585</div>
2586
2587<!-- _______________________________________________________________________ -->
2588<div class="doc_subsubsection">
2589 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2590</div>
2591
2592<div class="doc_text">
2593
2594<h5>Syntax:</h5>
2595
2596<pre>
2597 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
2598</pre>
2599
2600<h5>Overview:</h5>
2601
2602<p>
2603The '<tt>extractelement</tt>' instruction extracts a single scalar
2604element from a vector at a specified index.
2605</p>
2606
2607
2608<h5>Arguments:</h5>
2609
2610<p>
2611The first operand of an '<tt>extractelement</tt>' instruction is a
2612value of <a href="#t_vector">vector</a> type. The second operand is
2613an index indicating the position from which to extract the element.
2614The index may be a variable.</p>
2615
2616<h5>Semantics:</h5>
2617
2618<p>
2619The result is a scalar of the same type as the element type of
2620<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2621<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2622results are undefined.
2623</p>
2624
2625<h5>Example:</h5>
2626
2627<pre>
2628 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
2629</pre>
2630</div>
2631
2632
2633<!-- _______________________________________________________________________ -->
2634<div class="doc_subsubsection">
2635 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2636</div>
2637
2638<div class="doc_text">
2639
2640<h5>Syntax:</h5>
2641
2642<pre>
2643 &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>
2644</pre>
2645
2646<h5>Overview:</h5>
2647
2648<p>
2649The '<tt>insertelement</tt>' instruction inserts a scalar
2650element into a vector at a specified index.
2651</p>
2652
2653
2654<h5>Arguments:</h5>
2655
2656<p>
2657The first operand of an '<tt>insertelement</tt>' instruction is a
2658value of <a href="#t_vector">vector</a> type. The second operand is a
2659scalar value whose type must equal the element type of the first
2660operand. The third operand is an index indicating the position at
2661which to insert the value. The index may be a variable.</p>
2662
2663<h5>Semantics:</h5>
2664
2665<p>
2666The result is a vector of the same type as <tt>val</tt>. Its
2667element values are those of <tt>val</tt> except at position
2668<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2669exceeds the length of <tt>val</tt>, the results are undefined.
2670</p>
2671
2672<h5>Example:</h5>
2673
2674<pre>
2675 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
2676</pre>
2677</div>
2678
2679<!-- _______________________________________________________________________ -->
2680<div class="doc_subsubsection">
2681 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2682</div>
2683
2684<div class="doc_text">
2685
2686<h5>Syntax:</h5>
2687
2688<pre>
2689 &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>
2690</pre>
2691
2692<h5>Overview:</h5>
2693
2694<p>
2695The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2696from two input vectors, returning a vector of the same type.
2697</p>
2698
2699<h5>Arguments:</h5>
2700
2701<p>
2702The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2703with types that match each other and types that match the result of the
2704instruction. The third argument is a shuffle mask, which has the same number
2705of elements as the other vector type, but whose element type is always 'i32'.
2706</p>
2707
2708<p>
2709The shuffle mask operand is required to be a constant vector with either
2710constant integer or undef values.
2711</p>
2712
2713<h5>Semantics:</h5>
2714
2715<p>
2716The elements of the two input vectors are numbered from left to right across
2717both of the vectors. The shuffle mask operand specifies, for each element of
2718the result vector, which element of the two input registers the result element
2719gets. The element selector may be undef (meaning "don't care") and the second
2720operand may be undef if performing a shuffle from only one vector.
2721</p>
2722
2723<h5>Example:</h5>
2724
2725<pre>
2726 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
2727 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
2728 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
2729 &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.
2730</pre>
2731</div>
2732
2733
2734<!-- ======================================================================= -->
2735<div class="doc_subsection">
2736 <a name="memoryops">Memory Access and Addressing Operations</a>
2737</div>
2738
2739<div class="doc_text">
2740
2741<p>A key design point of an SSA-based representation is how it
2742represents memory. In LLVM, no memory locations are in SSA form, which
2743makes things very simple. This section describes how to read, write,
2744allocate, and free memory in LLVM.</p>
2745
2746</div>
2747
2748<!-- _______________________________________________________________________ -->
2749<div class="doc_subsubsection">
2750 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2751</div>
2752
2753<div class="doc_text">
2754
2755<h5>Syntax:</h5>
2756
2757<pre>
2758 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
2759</pre>
2760
2761<h5>Overview:</h5>
2762
2763<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]2764heap and returns a pointer to it. The object is always allocated in the generic
2765address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2766
2767<h5>Arguments:</h5>
2768
2769<p>The '<tt>malloc</tt>' instruction allocates
2770<tt>sizeof(&lt;type&gt;)*NumElements</tt>
2771bytes of memory from the operating system and returns a pointer of the
2772appropriate type to the program. If "NumElements" is specified, it is the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]2773number of elements allocated, otherwise "NumElements" is defaulted to be one.
2774If an alignment is specified, the value result of the allocation is guaranteed to
2775be aligned to at least that boundary. If not specified, or if zero, the target can
2776choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2777
2778<p>'<tt>type</tt>' must be a sized type.</p>
2779
2780<h5>Semantics:</h5>
2781
2782<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2783a pointer is returned.</p>
2784
2785<h5>Example:</h5>
2786
2787<pre>
2788 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2789
2790 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2791 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2792 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2793 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2794 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2795</pre>
2796</div>
2797
2798<!-- _______________________________________________________________________ -->
2799<div class="doc_subsubsection">
2800 <a name="i_free">'<tt>free</tt>' Instruction</a>
2801</div>
2802
2803<div class="doc_text">
2804
2805<h5>Syntax:</h5>
2806
2807<pre>
2808 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
2809</pre>
2810
2811<h5>Overview:</h5>
2812
2813<p>The '<tt>free</tt>' instruction returns memory back to the unused
2814memory heap to be reallocated in the future.</p>
2815
2816<h5>Arguments:</h5>
2817
2818<p>'<tt>value</tt>' shall be a pointer value that points to a value
2819that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2820instruction.</p>
2821
2822<h5>Semantics:</h5>
2823
2824<p>Access to the memory pointed to by the pointer is no longer defined
2825after this instruction executes.</p>
2826
2827<h5>Example:</h5>
2828
2829<pre>
2830 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2831 free [4 x i8]* %array
2832</pre>
2833</div>
2834
2835<!-- _______________________________________________________________________ -->
2836<div class="doc_subsubsection">
2837 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2838</div>
2839
2840<div class="doc_text">
2841
2842<h5>Syntax:</h5>
2843
2844<pre>
2845 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
2846</pre>
2847
2848<h5>Overview:</h5>
2849
2850<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
2851currently executing function, to be automatically released when this function
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]2852returns to its caller. The object is always allocated in the generic address
2853space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2854
2855<h5>Arguments:</h5>
2856
2857<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
2858bytes of memory on the runtime stack, returning a pointer of the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]2859appropriate type to the program. If "NumElements" is specified, it is the
2860number of elements allocated, otherwise "NumElements" is defaulted to be one.
2861If an alignment is specified, the value result of the allocation is guaranteed
2862to be aligned to at least that boundary. If not specified, or if zero, the target
2863can choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2864
2865<p>'<tt>type</tt>' may be any sized type.</p>
2866
2867<h5>Semantics:</h5>
2868
2869<p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2870memory is automatically released when the function returns. The '<tt>alloca</tt>'
2871instruction is commonly used to represent automatic variables that must
2872have an address available. When the function returns (either with the <tt><a
2873 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2874instructions), the memory is reclaimed.</p>
2875
2876<h5>Example:</h5>
2877
2878<pre>
2879 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2880 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2881 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2882 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2883</pre>
2884</div>
2885
2886<!-- _______________________________________________________________________ -->
2887<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2888Instruction</a> </div>
2889<div class="doc_text">
2890<h5>Syntax:</h5>
2891<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>
2892<h5>Overview:</h5>
2893<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2894<h5>Arguments:</h5>
2895<p>The argument to the '<tt>load</tt>' instruction specifies the memory
2896address from which to load. The pointer must point to a <a
2897 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2898marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2899the number or order of execution of this <tt>load</tt> with other
2900volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2901instructions. </p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]2902<p>
2903The optional "align" argument specifies the alignment of the operation
2904(that is, the alignment of the memory address). A value of 0 or an
2905omitted "align" argument means that the operation has the preferential
2906alignment for the target. It is the responsibility of the code emitter
2907to ensure that the alignment information is correct. Overestimating
2908the alignment results in an undefined behavior. Underestimating the
2909alignment may produce less efficient code. An alignment of 1 is always
2910safe.
2911</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2912<h5>Semantics:</h5>
2913<p>The location of memory pointed to is loaded.</p>
2914<h5>Examples:</h5>
2915<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2916 <a
2917 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2918 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2919</pre>
2920</div>
2921<!-- _______________________________________________________________________ -->
2922<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2923Instruction</a> </div>
2924<div class="doc_text">
2925<h5>Syntax:</h5>
2926<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
2927 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
2928</pre>
2929<h5>Overview:</h5>
2930<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2931<h5>Arguments:</h5>
2932<p>There are two arguments to the '<tt>store</tt>' instruction: a value
2933to store and an address at which to store it. The type of the '<tt>&lt;pointer&gt;</tt>'
2934operand must be a pointer to the type of the '<tt>&lt;value&gt;</tt>'
2935operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2936optimizer is not allowed to modify the number or order of execution of
2937this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2938 href="#i_store">store</a></tt> instructions.</p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]2939<p>
2940The optional "align" argument specifies the alignment of the operation
2941(that is, the alignment of the memory address). A value of 0 or an
2942omitted "align" argument means that the operation has the preferential
2943alignment for the target. It is the responsibility of the code emitter
2944to ensure that the alignment information is correct. Overestimating
2945the alignment results in an undefined behavior. Underestimating the
2946alignment may produce less efficient code. An alignment of 1 is always
2947safe.
2948</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2949<h5>Semantics:</h5>
2950<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
2951at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
2952<h5>Example:</h5>
2953<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]2954 store i32 3, i32* %ptr <i>; yields {void}</i>
2955 %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]2956</pre>
2957</div>
2958
2959<!-- _______________________________________________________________________ -->
2960<div class="doc_subsubsection">
2961 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2962</div>
2963
2964<div class="doc_text">
2965<h5>Syntax:</h5>
2966<pre>
2967 &lt;result&gt; = getelementptr &lt;ty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
2968</pre>
2969
2970<h5>Overview:</h5>
2971
2972<p>
2973The '<tt>getelementptr</tt>' instruction is used to get the address of a
2974subelement of an aggregate data structure.</p>
2975
2976<h5>Arguments:</h5>
2977
2978<p>This instruction takes a list of integer operands that indicate what
2979elements of the aggregate object to index to. The actual types of the arguments
2980provided depend on the type of the first pointer argument. The
2981'<tt>getelementptr</tt>' instruction is used to index down through the type
2982levels of a structure or to a specific index in an array. When indexing into a
2983structure, only <tt>i32</tt> integer constants are allowed. When indexing
2984into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2985be sign extended to 64-bit values.</p>
2986
2987<p>For example, let's consider a C code fragment and how it gets
2988compiled to LLVM:</p>
2989
2990<div class="doc_code">
2991<pre>
2992struct RT {
2993 char A;
2994 int B[10][20];
2995 char C;
2996};
2997struct ST {
2998 int X;
2999 double Y;
3000 struct RT Z;
3001};
3002
3003int *foo(struct ST *s) {
3004 return &amp;s[1].Z.B[5][13];
3005}
3006</pre>
3007</div>
3008
3009<p>The LLVM code generated by the GCC frontend is:</p>
3010
3011<div class="doc_code">
3012<pre>
3013%RT = type { i8 , [10 x [20 x i32]], i8 }
3014%ST = type { i32, double, %RT }
3015
3016define i32* %foo(%ST* %s) {
3017entry:
3018 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3019 ret i32* %reg
3020}
3021</pre>
3022</div>
3023
3024<h5>Semantics:</h5>
3025
3026<p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
3027on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
3028and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
3029<a href="#t_integer">integer</a> type but the value will always be sign extended
3030to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
3031<b>constants</b>.</p>
3032
3033<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3034type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3035}</tt>' type, a structure. The second index indexes into the third element of
3036the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3037i8 }</tt>' type, another structure. The third index indexes into the second
3038element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3039array. The two dimensions of the array are subscripted into, yielding an
3040'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3041to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3042
3043<p>Note that it is perfectly legal to index partially through a
3044structure, returning a pointer to an inner element. Because of this,
3045the LLVM code for the given testcase is equivalent to:</p>
3046
3047<pre>
3048 define i32* %foo(%ST* %s) {
3049 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3050 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3051 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3052 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3053 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3054 ret i32* %t5
3055 }
3056</pre>
3057
3058<p>Note that it is undefined to access an array out of bounds: array and
3059pointer indexes must always be within the defined bounds of the array type.
3060The one exception for this rules is zero length arrays. These arrays are
3061defined to be accessible as variable length arrays, which requires access
3062beyond the zero'th element.</p>
3063
3064<p>The getelementptr instruction is often confusing. For some more insight
3065into how it works, see <a href="GetElementPtr.html">the getelementptr
3066FAQ</a>.</p>
3067
3068<h5>Example:</h5>
3069
3070<pre>
3071 <i>; yields [12 x i8]*:aptr</i>
3072 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
3073</pre>
3074</div>
3075
3076<!-- ======================================================================= -->
3077<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3078</div>
3079<div class="doc_text">
3080<p>The instructions in this category are the conversion instructions (casting)
3081which all take a single operand and a type. They perform various bit conversions
3082on the operand.</p>
3083</div>
3084
3085<!-- _______________________________________________________________________ -->
3086<div class="doc_subsubsection">
3087 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3088</div>
3089<div class="doc_text">
3090
3091<h5>Syntax:</h5>
3092<pre>
3093 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3094</pre>
3095
3096<h5>Overview:</h5>
3097<p>
3098The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3099</p>
3100
3101<h5>Arguments:</h5>
3102<p>
3103The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3104be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3105and type of the result, which must be an <a href="#t_integer">integer</a>
3106type. The bit size of <tt>value</tt> must be larger than the bit size of
3107<tt>ty2</tt>. Equal sized types are not allowed.</p>
3108
3109<h5>Semantics:</h5>
3110<p>
3111The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3112and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3113larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3114It will always truncate bits.</p>
3115
3116<h5>Example:</h5>
3117<pre>
3118 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3119 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3120 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3121</pre>
3122</div>
3123
3124<!-- _______________________________________________________________________ -->
3125<div class="doc_subsubsection">
3126 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3127</div>
3128<div class="doc_text">
3129
3130<h5>Syntax:</h5>
3131<pre>
3132 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3133</pre>
3134
3135<h5>Overview:</h5>
3136<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3137<tt>ty2</tt>.</p>
3138
3139
3140<h5>Arguments:</h5>
3141<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3142<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3143also be of <a href="#t_integer">integer</a> type. The bit size of the
3144<tt>value</tt> must be smaller than the bit size of the destination type,
3145<tt>ty2</tt>.</p>
3146
3147<h5>Semantics:</h5>
3148<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3149bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3150
3151<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3152
3153<h5>Example:</h5>
3154<pre>
3155 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3156 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3157</pre>
3158</div>
3159
3160<!-- _______________________________________________________________________ -->
3161<div class="doc_subsubsection">
3162 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3163</div>
3164<div class="doc_text">
3165
3166<h5>Syntax:</h5>
3167<pre>
3168 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3169</pre>
3170
3171<h5>Overview:</h5>
3172<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3173
3174<h5>Arguments:</h5>
3175<p>
3176The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3177<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3178also be of <a href="#t_integer">integer</a> type. The bit size of the
3179<tt>value</tt> must be smaller than the bit size of the destination type,
3180<tt>ty2</tt>.</p>
3181
3182<h5>Semantics:</h5>
3183<p>
3184The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3185bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3186the type <tt>ty2</tt>.</p>
3187
3188<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3189
3190<h5>Example:</h5>
3191<pre>
3192 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3193 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3194</pre>
3195</div>
3196
3197<!-- _______________________________________________________________________ -->
3198<div class="doc_subsubsection">
3199 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3200</div>
3201
3202<div class="doc_text">
3203
3204<h5>Syntax:</h5>
3205
3206<pre>
3207 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3208</pre>
3209
3210<h5>Overview:</h5>
3211<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3212<tt>ty2</tt>.</p>
3213
3214
3215<h5>Arguments:</h5>
3216<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3217 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3218cast it to. The size of <tt>value</tt> must be larger than the size of
3219<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3220<i>no-op cast</i>.</p>
3221
3222<h5>Semantics:</h5>
3223<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3224<a href="#t_floating">floating point</a> type to a smaller
3225<a href="#t_floating">floating point</a> type. If the value cannot fit within
3226the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3227
3228<h5>Example:</h5>
3229<pre>
3230 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3231 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3232</pre>
3233</div>
3234
3235<!-- _______________________________________________________________________ -->
3236<div class="doc_subsubsection">
3237 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3238</div>
3239<div class="doc_text">
3240
3241<h5>Syntax:</h5>
3242<pre>
3243 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3244</pre>
3245
3246<h5>Overview:</h5>
3247<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3248floating point value.</p>
3249
3250<h5>Arguments:</h5>
3251<p>The '<tt>fpext</tt>' instruction takes a
3252<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3253and a <a href="#t_floating">floating point</a> type to cast it to. The source
3254type must be smaller than the destination type.</p>
3255
3256<h5>Semantics:</h5>
3257<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3258<a href="#t_floating">floating point</a> type to a larger
3259<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3260used to make a <i>no-op cast</i> because it always changes bits. Use
3261<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3262
3263<h5>Example:</h5>
3264<pre>
3265 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3266 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3267</pre>
3268</div>
3269
3270<!-- _______________________________________________________________________ -->
3271<div class="doc_subsubsection">
3272 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3273</div>
3274<div class="doc_text">
3275
3276<h5>Syntax:</h5>
3277<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3278 &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]3279</pre>
3280
3281<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3282<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3283unsigned integer equivalent of type <tt>ty2</tt>.
3284</p>
3285
3286<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3287<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]3288scalar or vector <a href="#t_floating">floating point</a> value, and a type
3289to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3290type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3291vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3292
3293<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3294<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3295<a href="#t_floating">floating point</a> operand into the nearest (rounding
3296towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3297the results are undefined.</p>
3298
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3299<h5>Example:</h5>
3300<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3301 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3302 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3303 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3304</pre>
3305</div>
3306
3307<!-- _______________________________________________________________________ -->
3308<div class="doc_subsubsection">
3309 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3310</div>
3311<div class="doc_text">
3312
3313<h5>Syntax:</h5>
3314<pre>
3315 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3316</pre>
3317
3318<h5>Overview:</h5>
3319<p>The '<tt>fptosi</tt>' instruction converts
3320<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3321</p>
3322
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3323<h5>Arguments:</h5>
3324<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]3325scalar or vector <a href="#t_floating">floating point</a> value, and a type
3326to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3327type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3328vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3329
3330<h5>Semantics:</h5>
3331<p>The '<tt>fptosi</tt>' instruction converts its
3332<a href="#t_floating">floating point</a> operand into the nearest (rounding
3333towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3334the results are undefined.</p>
3335
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3336<h5>Example:</h5>
3337<pre>
3338 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3339 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3340 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3341</pre>
3342</div>
3343
3344<!-- _______________________________________________________________________ -->
3345<div class="doc_subsubsection">
3346 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3347</div>
3348<div class="doc_text">
3349
3350<h5>Syntax:</h5>
3351<pre>
3352 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3353</pre>
3354
3355<h5>Overview:</h5>
3356<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3357integer and converts that value to the <tt>ty2</tt> type.</p>
3358
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3359<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3360<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3361scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3362to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3363type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3364floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3365
3366<h5>Semantics:</h5>
3367<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3368integer quantity and converts it to the corresponding floating point value. If
3369the value cannot fit in the floating point value, the results are undefined.</p>
3370
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3371<h5>Example:</h5>
3372<pre>
3373 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3374 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3375</pre>
3376</div>
3377
3378<!-- _______________________________________________________________________ -->
3379<div class="doc_subsubsection">
3380 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3381</div>
3382<div class="doc_text">
3383
3384<h5>Syntax:</h5>
3385<pre>
3386 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3387</pre>
3388
3389<h5>Overview:</h5>
3390<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3391integer and converts that value to the <tt>ty2</tt> type.</p>
3392
3393<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3394<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3395scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3396to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3397type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3398floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3399
3400<h5>Semantics:</h5>
3401<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3402integer quantity and converts it to the corresponding floating point value. If
3403the value cannot fit in the floating point value, the results are undefined.</p>
3404
3405<h5>Example:</h5>
3406<pre>
3407 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3408 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3409</pre>
3410</div>
3411
3412<!-- _______________________________________________________________________ -->
3413<div class="doc_subsubsection">
3414 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3415</div>
3416<div class="doc_text">
3417
3418<h5>Syntax:</h5>
3419<pre>
3420 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3421</pre>
3422
3423<h5>Overview:</h5>
3424<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3425the integer type <tt>ty2</tt>.</p>
3426
3427<h5>Arguments:</h5>
3428<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3429must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3430<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3431
3432<h5>Semantics:</h5>
3433<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3434<tt>ty2</tt> by interpreting the pointer value as an integer and either
3435truncating or zero extending that value to the size of the integer type. If
3436<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3437<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3438are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3439change.</p>
3440
3441<h5>Example:</h5>
3442<pre>
3443 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3444 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3445</pre>
3446</div>
3447
3448<!-- _______________________________________________________________________ -->
3449<div class="doc_subsubsection">
3450 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3451</div>
3452<div class="doc_text">
3453
3454<h5>Syntax:</h5>
3455<pre>
3456 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3457</pre>
3458
3459<h5>Overview:</h5>
3460<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3461a pointer type, <tt>ty2</tt>.</p>
3462
3463<h5>Arguments:</h5>
3464<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3465value to cast, and a type to cast it to, which must be a
3466<a href="#t_pointer">pointer</a> type.
3467
3468<h5>Semantics:</h5>
3469<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3470<tt>ty2</tt> by applying either a zero extension or a truncation depending on
3471the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3472size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3473the size of a pointer then a zero extension is done. If they are the same size,
3474nothing is done (<i>no-op cast</i>).</p>
3475
3476<h5>Example:</h5>
3477<pre>
3478 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3479 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3480 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3481</pre>
3482</div>
3483
3484<!-- _______________________________________________________________________ -->
3485<div class="doc_subsubsection">
3486 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3487</div>
3488<div class="doc_text">
3489
3490<h5>Syntax:</h5>
3491<pre>
3492 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3493</pre>
3494
3495<h5>Overview:</h5>
3496<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3497<tt>ty2</tt> without changing any bits.</p>
3498
3499<h5>Arguments:</h5>
3500<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3501a first class value, and a type to cast it to, which must also be a <a
3502 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3503and the destination type, <tt>ty2</tt>, must be identical. If the source
3504type is a pointer, the destination type must also be a pointer.</p>
3505
3506<h5>Semantics:</h5>
3507<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3508<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3509this conversion. The conversion is done as if the <tt>value</tt> had been
3510stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3511converted to other pointer types with this instruction. To convert pointers to
3512other types, use the <a href="#i_inttoptr">inttoptr</a> or
3513<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3514
3515<h5>Example:</h5>
3516<pre>
3517 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3518 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3519 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3520</pre>
3521</div>
3522
3523<!-- ======================================================================= -->
3524<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3525<div class="doc_text">
3526<p>The instructions in this category are the "miscellaneous"
3527instructions, which defy better classification.</p>
3528</div>
3529
3530<!-- _______________________________________________________________________ -->
3531<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3532</div>
3533<div class="doc_text">
3534<h5>Syntax:</h5>
3535<pre> &lt;result&gt; = icmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3536</pre>
3537<h5>Overview:</h5>
3538<p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3539of its two integer operands.</p>
3540<h5>Arguments:</h5>
3541<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3542the condition code indicating the kind of comparison to perform. It is not
3543a value, just a keyword. The possible condition code are:
3544<ol>
3545 <li><tt>eq</tt>: equal</li>
3546 <li><tt>ne</tt>: not equal </li>
3547 <li><tt>ugt</tt>: unsigned greater than</li>
3548 <li><tt>uge</tt>: unsigned greater or equal</li>
3549 <li><tt>ult</tt>: unsigned less than</li>
3550 <li><tt>ule</tt>: unsigned less or equal</li>
3551 <li><tt>sgt</tt>: signed greater than</li>
3552 <li><tt>sge</tt>: signed greater or equal</li>
3553 <li><tt>slt</tt>: signed less than</li>
3554 <li><tt>sle</tt>: signed less or equal</li>
3555</ol>
3556<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3557<a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3558<h5>Semantics:</h5>
3559<p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3560the condition code given as <tt>cond</tt>. The comparison performed always
3561yields a <a href="#t_primitive">i1</a> result, as follows:
3562<ol>
3563 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3564 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3565 </li>
3566 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3567 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3568 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3569 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3570 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3571 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3572 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3573 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3574 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3575 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3576 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3577 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3578 <li><tt>sge</tt>: interprets the operands as signed values and yields
3579 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3580 <li><tt>slt</tt>: interprets the operands as signed values and yields
3581 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3582 <li><tt>sle</tt>: interprets the operands as signed values and yields
3583 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3584</ol>
3585<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3586values are compared as if they were integers.</p>
3587
3588<h5>Example:</h5>
3589<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
3590 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
3591 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
3592 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3593 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3594 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
3595</pre>
3596</div>
3597
3598<!-- _______________________________________________________________________ -->
3599<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3600</div>
3601<div class="doc_text">
3602<h5>Syntax:</h5>
3603<pre> &lt;result&gt; = fcmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3604</pre>
3605<h5>Overview:</h5>
3606<p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3607of its floating point operands.</p>
3608<h5>Arguments:</h5>
3609<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3610the condition code indicating the kind of comparison to perform. It is not
3611a value, just a keyword. The possible condition code are:
3612<ol>
3613 <li><tt>false</tt>: no comparison, always returns false</li>
3614 <li><tt>oeq</tt>: ordered and equal</li>
3615 <li><tt>ogt</tt>: ordered and greater than </li>
3616 <li><tt>oge</tt>: ordered and greater than or equal</li>
3617 <li><tt>olt</tt>: ordered and less than </li>
3618 <li><tt>ole</tt>: ordered and less than or equal</li>
3619 <li><tt>one</tt>: ordered and not equal</li>
3620 <li><tt>ord</tt>: ordered (no nans)</li>
3621 <li><tt>ueq</tt>: unordered or equal</li>
3622 <li><tt>ugt</tt>: unordered or greater than </li>
3623 <li><tt>uge</tt>: unordered or greater than or equal</li>
3624 <li><tt>ult</tt>: unordered or less than </li>
3625 <li><tt>ule</tt>: unordered or less than or equal</li>
3626 <li><tt>une</tt>: unordered or not equal</li>
3627 <li><tt>uno</tt>: unordered (either nans)</li>
3628 <li><tt>true</tt>: no comparison, always returns true</li>
3629</ol>
3630<p><i>Ordered</i> means that neither operand is a QNAN while
3631<i>unordered</i> means that either operand may be a QNAN.</p>
3632<p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3633<a href="#t_floating">floating point</a> typed. They must have identical
3634types.</p>
3635<h5>Semantics:</h5>
3636<p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3637the condition code given as <tt>cond</tt>. The comparison performed always
3638yields a <a href="#t_primitive">i1</a> result, as follows:
3639<ol>
3640 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3641 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3642 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3643 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3644 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3645 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3646 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3647 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3648 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3649 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3650 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3651 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3652 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3653 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3654 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3655 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3656 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3657 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3658 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3659 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3660 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3661 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3662 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3663 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3664 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3665 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3666 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3667 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3668</ol>
3669
3670<h5>Example:</h5>
3671<pre> &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3672 &lt;result&gt; = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3673 &lt;result&gt; = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3674 &lt;result&gt; = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3675</pre>
3676</div>
3677
3678<!-- _______________________________________________________________________ -->
3679<div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3680Instruction</a> </div>
3681<div class="doc_text">
3682<h5>Syntax:</h5>
3683<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
3684<h5>Overview:</h5>
3685<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
3686the SSA graph representing the function.</p>
3687<h5>Arguments:</h5>
3688<p>The type of the incoming values is specified with the first type
3689field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3690as arguments, with one pair for each predecessor basic block of the
3691current block. Only values of <a href="#t_firstclass">first class</a>
3692type may be used as the value arguments to the PHI node. Only labels
3693may be used as the label arguments.</p>
3694<p>There must be no non-phi instructions between the start of a basic
3695block and the PHI instructions: i.e. PHI instructions must be first in
3696a basic block.</p>
3697<h5>Semantics:</h5>
3698<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3699specified by the pair corresponding to the predecessor basic block that executed
3700just prior to the current block.</p>
3701<h5>Example:</h5>
3702<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>
3703</div>
3704
3705<!-- _______________________________________________________________________ -->
3706<div class="doc_subsubsection">
3707 <a name="i_select">'<tt>select</tt>' Instruction</a>
3708</div>
3709
3710<div class="doc_text">
3711
3712<h5>Syntax:</h5>
3713
3714<pre>
3715 &lt;result&gt; = select i1 &lt;cond&gt;, &lt;ty&gt; &lt;val1&gt;, &lt;ty&gt; &lt;val2&gt; <i>; yields ty</i>
3716</pre>
3717
3718<h5>Overview:</h5>
3719
3720<p>
3721The '<tt>select</tt>' instruction is used to choose one value based on a
3722condition, without branching.
3723</p>
3724
3725
3726<h5>Arguments:</h5>
3727
3728<p>
3729The '<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.
3730</p>
3731
3732<h5>Semantics:</h5>
3733
3734<p>
3735If the boolean condition evaluates to true, the instruction returns the first
3736value argument; otherwise, it returns the second value argument.
3737</p>
3738
3739<h5>Example:</h5>
3740
3741<pre>
3742 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3743</pre>
3744</div>
3745
3746
3747<!-- _______________________________________________________________________ -->
3748<div class="doc_subsubsection">
3749 <a name="i_call">'<tt>call</tt>' Instruction</a>
3750</div>
3751
3752<div class="doc_text">
3753
3754<h5>Syntax:</h5>
3755<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3756 &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]3757</pre>
3758
3759<h5>Overview:</h5>
3760
3761<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3762
3763<h5>Arguments:</h5>
3764
3765<p>This instruction requires several arguments:</p>
3766
3767<ol>
3768 <li>
3769 <p>The optional "tail" marker indicates whether the callee function accesses
3770 any allocas or varargs in the caller. If the "tail" marker is present, the
3771 function call is eligible for tail call optimization. Note that calls may
3772 be marked "tail" even if they do not occur before a <a
3773 href="#i_ret"><tt>ret</tt></a> instruction.
3774 </li>
3775 <li>
3776 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3777 convention</a> the call should use. If none is specified, the call defaults
3778 to using C calling conventions.
3779 </li>
3780 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3781 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
3782 the type of the return value. Functions that return no value are marked
3783 <tt><a href="#t_void">void</a></tt>.</p>
3784 </li>
3785 <li>
3786 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
3787 value being invoked. The argument types must match the types implied by
3788 this signature. This type can be omitted if the function is not varargs
3789 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3790 </li>
3791 <li>
3792 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3793 be invoked. In most cases, this is a direct function invocation, but
3794 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3795 to function value.</p>
3796 </li>
3797 <li>
3798 <p>'<tt>function args</tt>': argument list whose types match the
3799 function signature argument types. All arguments must be of
3800 <a href="#t_firstclass">first class</a> type. If the function signature
3801 indicates the function accepts a variable number of arguments, the extra
3802 arguments can be specified.</p>
3803 </li>
3804</ol>
3805
3806<h5>Semantics:</h5>
3807
3808<p>The '<tt>call</tt>' instruction is used to cause control flow to
3809transfer to a specified function, with its incoming arguments bound to
3810the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3811instruction in the called function, control flow continues with the
3812instruction after the function call, and the return value of the
3813function is bound to the result argument. This is a simpler case of
3814the <a href="#i_invoke">invoke</a> instruction.</p>
3815
3816<h5>Example:</h5>
3817
3818<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3819 %retval = call i32 @test(i32 %argc)
3820 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42);
3821 %X = tail call i32 @foo()
3822 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo()
3823 %Z = call void %foo(i8 97 signext)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3824</pre>
3825
3826</div>
3827
3828<!-- _______________________________________________________________________ -->
3829<div class="doc_subsubsection">
3830 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3831</div>
3832
3833<div class="doc_text">
3834
3835<h5>Syntax:</h5>
3836
3837<pre>
3838 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
3839</pre>
3840
3841<h5>Overview:</h5>
3842
3843<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3844the "variable argument" area of a function call. It is used to implement the
3845<tt>va_arg</tt> macro in C.</p>
3846
3847<h5>Arguments:</h5>
3848
3849<p>This instruction takes a <tt>va_list*</tt> value and the type of
3850the argument. It returns a value of the specified argument type and
3851increments the <tt>va_list</tt> to point to the next argument. The
3852actual type of <tt>va_list</tt> is target specific.</p>
3853
3854<h5>Semantics:</h5>
3855
3856<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3857type from the specified <tt>va_list</tt> and causes the
3858<tt>va_list</tt> to point to the next argument. For more information,
3859see the variable argument handling <a href="#int_varargs">Intrinsic
3860Functions</a>.</p>
3861
3862<p>It is legal for this instruction to be called in a function which does not
3863take a variable number of arguments, for example, the <tt>vfprintf</tt>
3864function.</p>
3865
3866<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3867href="#intrinsics">intrinsic function</a> because it takes a type as an
3868argument.</p>
3869
3870<h5>Example:</h5>
3871
3872<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3873
3874</div>
3875
3876<!-- *********************************************************************** -->
3877<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3878<!-- *********************************************************************** -->
3879
3880<div class="doc_text">
3881
3882<p>LLVM supports the notion of an "intrinsic function". These functions have
3883well known names and semantics and are required to follow certain restrictions.
3884Overall, these intrinsics represent an extension mechanism for the LLVM
3885language that does not require changing all of the transformations in LLVM when
3886adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
3887
3888<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3889prefix is reserved in LLVM for intrinsic names; thus, function names may not
3890begin with this prefix. Intrinsic functions must always be external functions:
3891you cannot define the body of intrinsic functions. Intrinsic functions may
3892only be used in call or invoke instructions: it is illegal to take the address
3893of an intrinsic function. Additionally, because intrinsic functions are part
3894of the LLVM language, it is required if any are added that they be documented
3895here.</p>
3896
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]3897<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
3898a family of functions that perform the same operation but on different data
3899types. Because LLVM can represent over 8 million different integer types,
3900overloading is used commonly to allow an intrinsic function to operate on any
3901integer type. One or more of the argument types or the result type can be
3902overloaded to accept any integer type. Argument types may also be defined as
3903exactly matching a previous argument's type or the result type. This allows an
3904intrinsic function which accepts multiple arguments, but needs all of them to
3905be of the same type, to only be overloaded with respect to a single argument or
3906the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3907
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]3908<p>Overloaded intrinsics will have the names of its overloaded argument types
3909encoded into its function name, each preceded by a period. Only those types
3910which are overloaded result in a name suffix. Arguments whose type is matched
3911against another type do not. For example, the <tt>llvm.ctpop</tt> function can
3912take an integer of any width and returns an integer of exactly the same integer
3913width. This leads to a family of functions such as
3914<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
3915Only one type, the return type, is overloaded, and only one type suffix is
3916required. Because the argument's type is matched against the return type, it
3917does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3918
3919<p>To learn how to add an intrinsic function, please see the
3920<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3921</p>
3922
3923</div>
3924
3925<!-- ======================================================================= -->
3926<div class="doc_subsection">
3927 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3928</div>
3929
3930<div class="doc_text">
3931
3932<p>Variable argument support is defined in LLVM with the <a
3933 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3934intrinsic functions. These functions are related to the similarly
3935named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
3936
3937<p>All of these functions operate on arguments that use a
3938target-specific value type "<tt>va_list</tt>". The LLVM assembly
3939language reference manual does not define what this type is, so all
3940transformations should be prepared to handle these functions regardless of
3941the type used.</p>
3942
3943<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3944instruction and the variable argument handling intrinsic functions are
3945used.</p>
3946
3947<div class="doc_code">
3948<pre>
3949define i32 @test(i32 %X, ...) {
3950 ; Initialize variable argument processing
3951 %ap = alloca i8*
3952 %ap2 = bitcast i8** %ap to i8*
3953 call void @llvm.va_start(i8* %ap2)
3954
3955 ; Read a single integer argument
3956 %tmp = va_arg i8** %ap, i32
3957
3958 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3959 %aq = alloca i8*
3960 %aq2 = bitcast i8** %aq to i8*
3961 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
3962 call void @llvm.va_end(i8* %aq2)
3963
3964 ; Stop processing of arguments.
3965 call void @llvm.va_end(i8* %ap2)
3966 ret i32 %tmp
3967}
3968
3969declare void @llvm.va_start(i8*)
3970declare void @llvm.va_copy(i8*, i8*)
3971declare void @llvm.va_end(i8*)
3972</pre>
3973</div>
3974
3975</div>
3976
3977<!-- _______________________________________________________________________ -->
3978<div class="doc_subsubsection">
3979 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3980</div>
3981
3982
3983<div class="doc_text">
3984<h5>Syntax:</h5>
3985<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
3986<h5>Overview:</h5>
3987<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3988<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
3989href="#i_va_arg">va_arg</a></tt>.</p>
3990
3991<h5>Arguments:</h5>
3992
3993<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3994
3995<h5>Semantics:</h5>
3996
3997<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3998macro available in C. In a target-dependent way, it initializes the
3999<tt>va_list</tt> element to which the argument points, so that the next call to
4000<tt>va_arg</tt> will produce the first variable argument passed to the function.
4001Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4002last argument of the function as the compiler can figure that out.</p>
4003
4004</div>
4005
4006<!-- _______________________________________________________________________ -->
4007<div class="doc_subsubsection">
4008 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4009</div>
4010
4011<div class="doc_text">
4012<h5>Syntax:</h5>
4013<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
4014<h5>Overview:</h5>
4015
4016<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
4017which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4018or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4019
4020<h5>Arguments:</h5>
4021
4022<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4023
4024<h5>Semantics:</h5>
4025
4026<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4027macro available in C. In a target-dependent way, it destroys the
4028<tt>va_list</tt> element to which the argument points. Calls to <a
4029href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4030<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4031<tt>llvm.va_end</tt>.</p>
4032
4033</div>
4034
4035<!-- _______________________________________________________________________ -->
4036<div class="doc_subsubsection">
4037 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4038</div>
4039
4040<div class="doc_text">
4041
4042<h5>Syntax:</h5>
4043
4044<pre>
4045 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
4046</pre>
4047
4048<h5>Overview:</h5>
4049
4050<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4051from the source argument list to the destination argument list.</p>
4052
4053<h5>Arguments:</h5>
4054
4055<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4056The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4057
4058
4059<h5>Semantics:</h5>
4060
4061<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4062macro available in C. In a target-dependent way, it copies the source
4063<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4064intrinsic is necessary because the <tt><a href="#int_va_start">
4065llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4066example, memory allocation.</p>
4067
4068</div>
4069
4070<!-- ======================================================================= -->
4071<div class="doc_subsection">
4072 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4073</div>
4074
4075<div class="doc_text">
4076
4077<p>
4078LLVM support for <a href="GarbageCollection.html">Accurate Garbage
4079Collection</a> requires the implementation and generation of these intrinsics.
4080These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4081stack</a>, as well as garbage collector implementations that require <a
4082href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4083Front-ends for type-safe garbage collected languages should generate these
4084intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4085href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4086</p>
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]4087
4088<p>The garbage collection intrinsics only operate on objects in the generic
4089 address space (address space zero).</p>
4090
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4091</div>
4092
4093<!-- _______________________________________________________________________ -->
4094<div class="doc_subsubsection">
4095 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4096</div>
4097
4098<div class="doc_text">
4099
4100<h5>Syntax:</h5>
4101
4102<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4103 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4104</pre>
4105
4106<h5>Overview:</h5>
4107
4108<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4109the code generator, and allows some metadata to be associated with it.</p>
4110
4111<h5>Arguments:</h5>
4112
4113<p>The first argument specifies the address of a stack object that contains the
4114root pointer. The second pointer (which must be either a constant or a global
4115value address) contains the meta-data to be associated with the root.</p>
4116
4117<h5>Semantics:</h5>
4118
4119<p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
4120location. At compile-time, the code generator generates information to allow
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4121the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4122intrinsic may only be used in a function which <a href="#gc">specifies a GC
4123algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4124
4125</div>
4126
4127
4128<!-- _______________________________________________________________________ -->
4129<div class="doc_subsubsection">
4130 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4131</div>
4132
4133<div class="doc_text">
4134
4135<h5>Syntax:</h5>
4136
4137<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4138 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4139</pre>
4140
4141<h5>Overview:</h5>
4142
4143<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4144locations, allowing garbage collector implementations that require read
4145barriers.</p>
4146
4147<h5>Arguments:</h5>
4148
4149<p>The second argument is the address to read from, which should be an address
4150allocated from the garbage collector. The first object is a pointer to the
4151start of the referenced object, if needed by the language runtime (otherwise
4152null).</p>
4153
4154<h5>Semantics:</h5>
4155
4156<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4157instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4158garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4159may only be used in a function which <a href="#gc">specifies a GC
4160algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4161
4162</div>
4163
4164
4165<!-- _______________________________________________________________________ -->
4166<div class="doc_subsubsection">
4167 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4168</div>
4169
4170<div class="doc_text">
4171
4172<h5>Syntax:</h5>
4173
4174<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4175 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4176</pre>
4177
4178<h5>Overview:</h5>
4179
4180<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4181locations, allowing garbage collector implementations that require write
4182barriers (such as generational or reference counting collectors).</p>
4183
4184<h5>Arguments:</h5>
4185
4186<p>The first argument is the reference to store, the second is the start of the
4187object to store it to, and the third is the address of the field of Obj to
4188store to. If the runtime does not require a pointer to the object, Obj may be
4189null.</p>
4190
4191<h5>Semantics:</h5>
4192
4193<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4194instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4195garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4196may only be used in a function which <a href="#gc">specifies a GC
4197algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4198
4199</div>
4200
4201
4202
4203<!-- ======================================================================= -->
4204<div class="doc_subsection">
4205 <a name="int_codegen">Code Generator Intrinsics</a>
4206</div>
4207
4208<div class="doc_text">
4209<p>
4210These intrinsics are provided by LLVM to expose special features that may only
4211be implemented with code generator support.
4212</p>
4213
4214</div>
4215
4216<!-- _______________________________________________________________________ -->
4217<div class="doc_subsubsection">
4218 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4219</div>
4220
4221<div class="doc_text">
4222
4223<h5>Syntax:</h5>
4224<pre>
4225 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
4226</pre>
4227
4228<h5>Overview:</h5>
4229
4230<p>
4231The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4232target-specific value indicating the return address of the current function
4233or one of its callers.
4234</p>
4235
4236<h5>Arguments:</h5>
4237
4238<p>
4239The argument to this intrinsic indicates which function to return the address
4240for. Zero indicates the calling function, one indicates its caller, etc. The
4241argument is <b>required</b> to be a constant integer value.
4242</p>
4243
4244<h5>Semantics:</h5>
4245
4246<p>
4247The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4248the return address of the specified call frame, or zero if it cannot be
4249identified. The value returned by this intrinsic is likely to be incorrect or 0
4250for arguments other than zero, so it should only be used for debugging purposes.
4251</p>
4252
4253<p>
4254Note that calling this intrinsic does not prevent function inlining or other
4255aggressive transformations, so the value returned may not be that of the obvious
4256source-language caller.
4257</p>
4258</div>
4259
4260
4261<!-- _______________________________________________________________________ -->
4262<div class="doc_subsubsection">
4263 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4264</div>
4265
4266<div class="doc_text">
4267
4268<h5>Syntax:</h5>
4269<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4270 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4271</pre>
4272
4273<h5>Overview:</h5>
4274
4275<p>
4276The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4277target-specific frame pointer value for the specified stack frame.
4278</p>
4279
4280<h5>Arguments:</h5>
4281
4282<p>
4283The argument to this intrinsic indicates which function to return the frame
4284pointer for. Zero indicates the calling function, one indicates its caller,
4285etc. The argument is <b>required</b> to be a constant integer value.
4286</p>
4287
4288<h5>Semantics:</h5>
4289
4290<p>
4291The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4292the frame address of the specified call frame, or zero if it cannot be
4293identified. The value returned by this intrinsic is likely to be incorrect or 0
4294for arguments other than zero, so it should only be used for debugging purposes.
4295</p>
4296
4297<p>
4298Note that calling this intrinsic does not prevent function inlining or other
4299aggressive transformations, so the value returned may not be that of the obvious
4300source-language caller.
4301</p>
4302</div>
4303
4304<!-- _______________________________________________________________________ -->
4305<div class="doc_subsubsection">
4306 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4307</div>
4308
4309<div class="doc_text">
4310
4311<h5>Syntax:</h5>
4312<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4313 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4314</pre>
4315
4316<h5>Overview:</h5>
4317
4318<p>
4319The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4320the function stack, for use with <a href="#int_stackrestore">
4321<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4322features like scoped automatic variable sized arrays in C99.
4323</p>
4324
4325<h5>Semantics:</h5>
4326
4327<p>
4328This intrinsic returns a opaque pointer value that can be passed to <a
4329href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4330<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4331<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4332state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4333practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4334that were allocated after the <tt>llvm.stacksave</tt> was executed.
4335</p>
4336
4337</div>
4338
4339<!-- _______________________________________________________________________ -->
4340<div class="doc_subsubsection">
4341 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4342</div>
4343
4344<div class="doc_text">
4345
4346<h5>Syntax:</h5>
4347<pre>
4348 declare void @llvm.stackrestore(i8 * %ptr)
4349</pre>
4350
4351<h5>Overview:</h5>
4352
4353<p>
4354The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4355the function stack to the state it was in when the corresponding <a
4356href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4357useful for implementing language features like scoped automatic variable sized
4358arrays in C99.
4359</p>
4360
4361<h5>Semantics:</h5>
4362
4363<p>
4364See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4365</p>
4366
4367</div>
4368
4369
4370<!-- _______________________________________________________________________ -->
4371<div class="doc_subsubsection">
4372 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4373</div>
4374
4375<div class="doc_text">
4376
4377<h5>Syntax:</h5>
4378<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4379 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]4380</pre>
4381
4382<h5>Overview:</h5>
4383
4384
4385<p>
4386The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4387a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4388no
4389effect on the behavior of the program but can change its performance
4390characteristics.
4391</p>
4392
4393<h5>Arguments:</h5>
4394
4395<p>
4396<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4397determining if the fetch should be for a read (0) or write (1), and
4398<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4399locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4400<tt>locality</tt> arguments must be constant integers.
4401</p>
4402
4403<h5>Semantics:</h5>
4404
4405<p>
4406This intrinsic does not modify the behavior of the program. In particular,
4407prefetches cannot trap and do not produce a value. On targets that support this
4408intrinsic, the prefetch can provide hints to the processor cache for better
4409performance.
4410</p>
4411
4412</div>
4413
4414<!-- _______________________________________________________________________ -->
4415<div class="doc_subsubsection">
4416 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4417</div>
4418
4419<div class="doc_text">
4420
4421<h5>Syntax:</h5>
4422<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4423 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4424</pre>
4425
4426<h5>Overview:</h5>
4427
4428
4429<p>
4430The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4431(PC) in a region of
4432code to simulators and other tools. The method is target specific, but it is
4433expected that the marker will use exported symbols to transmit the PC of the marker.
4434The marker makes no guarantees that it will remain with any specific instruction
4435after optimizations. It is possible that the presence of a marker will inhibit
4436optimizations. The intended use is to be inserted after optimizations to allow
4437correlations of simulation runs.
4438</p>
4439
4440<h5>Arguments:</h5>
4441
4442<p>
4443<tt>id</tt> is a numerical id identifying the marker.
4444</p>
4445
4446<h5>Semantics:</h5>
4447
4448<p>
4449This intrinsic does not modify the behavior of the program. Backends that do not
4450support this intrinisic may ignore it.
4451</p>
4452
4453</div>
4454
4455<!-- _______________________________________________________________________ -->
4456<div class="doc_subsubsection">
4457 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4458</div>
4459
4460<div class="doc_text">
4461
4462<h5>Syntax:</h5>
4463<pre>
4464 declare i64 @llvm.readcyclecounter( )
4465</pre>
4466
4467<h5>Overview:</h5>
4468
4469
4470<p>
4471The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4472counter register (or similar low latency, high accuracy clocks) on those targets
4473that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4474As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4475should only be used for small timings.
4476</p>
4477
4478<h5>Semantics:</h5>
4479
4480<p>
4481When directly supported, reading the cycle counter should not modify any memory.
4482Implementations are allowed to either return a application specific value or a
4483system wide value. On backends without support, this is lowered to a constant 0.
4484</p>
4485
4486</div>
4487
4488<!-- ======================================================================= -->
4489<div class="doc_subsection">
4490 <a name="int_libc">Standard C Library Intrinsics</a>
4491</div>
4492
4493<div class="doc_text">
4494<p>
4495LLVM provides intrinsics for a few important standard C library functions.
4496These intrinsics allow source-language front-ends to pass information about the
4497alignment of the pointer arguments to the code generator, providing opportunity
4498for more efficient code generation.
4499</p>
4500
4501</div>
4502
4503<!-- _______________________________________________________________________ -->
4504<div class="doc_subsubsection">
4505 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4506</div>
4507
4508<div class="doc_text">
4509
4510<h5>Syntax:</h5>
4511<pre>
4512 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4513 i32 &lt;len&gt;, i32 &lt;align&gt;)
4514 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4515 i64 &lt;len&gt;, i32 &lt;align&gt;)
4516</pre>
4517
4518<h5>Overview:</h5>
4519
4520<p>
4521The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4522location to the destination location.
4523</p>
4524
4525<p>
4526Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4527intrinsics do not return a value, and takes an extra alignment argument.
4528</p>
4529
4530<h5>Arguments:</h5>
4531
4532<p>
4533The first argument is a pointer to the destination, the second is a pointer to
4534the source. The third argument is an integer argument
4535specifying the number of bytes to copy, and the fourth argument is the alignment
4536of the source and destination locations.
4537</p>
4538
4539<p>
4540If the call to this intrinisic has an alignment value that is not 0 or 1, then
4541the caller guarantees that both the source and destination pointers are aligned
4542to that boundary.
4543</p>
4544
4545<h5>Semantics:</h5>
4546
4547<p>
4548The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4549location to the destination location, which are not allowed to overlap. It
4550copies "len" bytes of memory over. If the argument is known to be aligned to
4551some boundary, this can be specified as the fourth argument, otherwise it should
4552be set to 0 or 1.
4553</p>
4554</div>
4555
4556
4557<!-- _______________________________________________________________________ -->
4558<div class="doc_subsubsection">
4559 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4560</div>
4561
4562<div class="doc_text">
4563
4564<h5>Syntax:</h5>
4565<pre>
4566 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4567 i32 &lt;len&gt;, i32 &lt;align&gt;)
4568 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4569 i64 &lt;len&gt;, i32 &lt;align&gt;)
4570</pre>
4571
4572<h5>Overview:</h5>
4573
4574<p>
4575The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4576location to the destination location. It is similar to the
Chris Lattnerdba16ea2008-01-06 19:51:52 +0000[diff] [blame]4577'<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4578</p>
4579
4580<p>
4581Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4582intrinsics do not return a value, and takes an extra alignment argument.
4583</p>
4584
4585<h5>Arguments:</h5>
4586
4587<p>
4588The first argument is a pointer to the destination, the second is a pointer to
4589the source. The third argument is an integer argument
4590specifying the number of bytes to copy, and the fourth argument is the alignment
4591of the source and destination locations.
4592</p>
4593
4594<p>
4595If the call to this intrinisic has an alignment value that is not 0 or 1, then
4596the caller guarantees that the source and destination pointers are aligned to
4597that boundary.
4598</p>
4599
4600<h5>Semantics:</h5>
4601
4602<p>
4603The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4604location to the destination location, which may overlap. It
4605copies "len" bytes of memory over. If the argument is known to be aligned to
4606some boundary, this can be specified as the fourth argument, otherwise it should
4607be set to 0 or 1.
4608</p>
4609</div>
4610
4611
4612<!-- _______________________________________________________________________ -->
4613<div class="doc_subsubsection">
4614 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4615</div>
4616
4617<div class="doc_text">
4618
4619<h5>Syntax:</h5>
4620<pre>
4621 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
4622 i32 &lt;len&gt;, i32 &lt;align&gt;)
4623 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
4624 i64 &lt;len&gt;, i32 &lt;align&gt;)
4625</pre>
4626
4627<h5>Overview:</h5>
4628
4629<p>
4630The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4631byte value.
4632</p>
4633
4634<p>
4635Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4636does not return a value, and takes an extra alignment argument.
4637</p>
4638
4639<h5>Arguments:</h5>
4640
4641<p>
4642The first argument is a pointer to the destination to fill, the second is the
4643byte value to fill it with, the third argument is an integer
4644argument specifying the number of bytes to fill, and the fourth argument is the
4645known alignment of destination location.
4646</p>
4647
4648<p>
4649If the call to this intrinisic has an alignment value that is not 0 or 1, then
4650the caller guarantees that the destination pointer is aligned to that boundary.
4651</p>
4652
4653<h5>Semantics:</h5>
4654
4655<p>
4656The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4657the
4658destination location. If the argument is known to be aligned to some boundary,
4659this can be specified as the fourth argument, otherwise it should be set to 0 or
46601.
4661</p>
4662</div>
4663
4664
4665<!-- _______________________________________________________________________ -->
4666<div class="doc_subsubsection">
4667 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4668</div>
4669
4670<div class="doc_text">
4671
4672<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4673<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]4674floating point or vector of floating point type. Not all targets support all
4675types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4676<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4677 declare float @llvm.sqrt.f32(float %Val)
4678 declare double @llvm.sqrt.f64(double %Val)
4679 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
4680 declare fp128 @llvm.sqrt.f128(fp128 %Val)
4681 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4682</pre>
4683
4684<h5>Overview:</h5>
4685
4686<p>
4687The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4688returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4689<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
Chris Lattnerf914a002008-01-29 07:00:44 +0000[diff] [blame]4690negative numbers other than -0.0 (which allows for better optimization, because
4691there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
4692defined to return -0.0 like IEEE sqrt.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4693</p>
4694
4695<h5>Arguments:</h5>
4696
4697<p>
4698The argument and return value are floating point numbers of the same type.
4699</p>
4700
4701<h5>Semantics:</h5>
4702
4703<p>
4704This function returns the sqrt of the specified operand if it is a nonnegative
4705floating point number.
4706</p>
4707</div>
4708
4709<!-- _______________________________________________________________________ -->
4710<div class="doc_subsubsection">
4711 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4712</div>
4713
4714<div class="doc_text">
4715
4716<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4717<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]4718floating point or vector of floating point type. Not all targets support all
4719types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4720<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4721 declare float @llvm.powi.f32(float %Val, i32 %power)
4722 declare double @llvm.powi.f64(double %Val, i32 %power)
4723 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
4724 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
4725 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4726</pre>
4727
4728<h5>Overview:</h5>
4729
4730<p>
4731The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4732specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4733multiplications is not defined. When a vector of floating point type is
4734used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4735</p>
4736
4737<h5>Arguments:</h5>
4738
4739<p>
4740The second argument is an integer power, and the first is a value to raise to
4741that power.
4742</p>
4743
4744<h5>Semantics:</h5>
4745
4746<p>
4747This function returns the first value raised to the second power with an
4748unspecified sequence of rounding operations.</p>
4749</div>
4750
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4751<!-- _______________________________________________________________________ -->
4752<div class="doc_subsubsection">
4753 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
4754</div>
4755
4756<div class="doc_text">
4757
4758<h5>Syntax:</h5>
4759<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
4760floating point or vector of floating point type. Not all targets support all
4761types however.
4762<pre>
4763 declare float @llvm.sin.f32(float %Val)
4764 declare double @llvm.sin.f64(double %Val)
4765 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
4766 declare fp128 @llvm.sin.f128(fp128 %Val)
4767 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
4768</pre>
4769
4770<h5>Overview:</h5>
4771
4772<p>
4773The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
4774</p>
4775
4776<h5>Arguments:</h5>
4777
4778<p>
4779The argument and return value are floating point numbers of the same type.
4780</p>
4781
4782<h5>Semantics:</h5>
4783
4784<p>
4785This function returns the sine of the specified operand, returning the
4786same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4787conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4788</div>
4789
4790<!-- _______________________________________________________________________ -->
4791<div class="doc_subsubsection">
4792 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
4793</div>
4794
4795<div class="doc_text">
4796
4797<h5>Syntax:</h5>
4798<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
4799floating point or vector of floating point type. Not all targets support all
4800types however.
4801<pre>
4802 declare float @llvm.cos.f32(float %Val)
4803 declare double @llvm.cos.f64(double %Val)
4804 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
4805 declare fp128 @llvm.cos.f128(fp128 %Val)
4806 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
4807</pre>
4808
4809<h5>Overview:</h5>
4810
4811<p>
4812The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
4813</p>
4814
4815<h5>Arguments:</h5>
4816
4817<p>
4818The argument and return value are floating point numbers of the same type.
4819</p>
4820
4821<h5>Semantics:</h5>
4822
4823<p>
4824This function returns the cosine of the specified operand, returning the
4825same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4826conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4827</div>
4828
4829<!-- _______________________________________________________________________ -->
4830<div class="doc_subsubsection">
4831 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
4832</div>
4833
4834<div class="doc_text">
4835
4836<h5>Syntax:</h5>
4837<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
4838floating point or vector of floating point type. Not all targets support all
4839types however.
4840<pre>
4841 declare float @llvm.pow.f32(float %Val, float %Power)
4842 declare double @llvm.pow.f64(double %Val, double %Power)
4843 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
4844 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
4845 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
4846</pre>
4847
4848<h5>Overview:</h5>
4849
4850<p>
4851The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
4852specified (positive or negative) power.
4853</p>
4854
4855<h5>Arguments:</h5>
4856
4857<p>
4858The second argument is a floating point power, and the first is a value to
4859raise to that power.
4860</p>
4861
4862<h5>Semantics:</h5>
4863
4864<p>
4865This function returns the first value raised to the second power,
4866returning the
4867same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4868conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4869</div>
4870
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4871
4872<!-- ======================================================================= -->
4873<div class="doc_subsection">
4874 <a name="int_manip">Bit Manipulation Intrinsics</a>
4875</div>
4876
4877<div class="doc_text">
4878<p>
4879LLVM provides intrinsics for a few important bit manipulation operations.
4880These allow efficient code generation for some algorithms.
4881</p>
4882
4883</div>
4884
4885<!-- _______________________________________________________________________ -->
4886<div class="doc_subsubsection">
4887 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4888</div>
4889
4890<div class="doc_text">
4891
4892<h5>Syntax:</h5>
4893<p>This is an overloaded intrinsic function. You can use bswap on any integer
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4894type that is an even number of bytes (i.e. BitWidth % 16 == 0).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4895<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4896 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
4897 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
4898 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4899</pre>
4900
4901<h5>Overview:</h5>
4902
4903<p>
4904The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
4905values with an even number of bytes (positive multiple of 16 bits). These are
4906useful for performing operations on data that is not in the target's native
4907byte order.
4908</p>
4909
4910<h5>Semantics:</h5>
4911
4912<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4913The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4914and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4915intrinsic returns an i32 value that has the four bytes of the input i32
4916swapped, 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]4917i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
4918<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4919additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4920</p>
4921
4922</div>
4923
4924<!-- _______________________________________________________________________ -->
4925<div class="doc_subsubsection">
4926 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4927</div>
4928
4929<div class="doc_text">
4930
4931<h5>Syntax:</h5>
4932<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4933width. Not all targets support all bit widths however.
4934<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4935 declare i8 @llvm.ctpop.i8 (i8 &lt;src&gt;)
4936 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4937 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4938 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
4939 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4940</pre>
4941
4942<h5>Overview:</h5>
4943
4944<p>
4945The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4946value.
4947</p>
4948
4949<h5>Arguments:</h5>
4950
4951<p>
4952The only argument is the value to be counted. The argument may be of any
4953integer type. The return type must match the argument type.
4954</p>
4955
4956<h5>Semantics:</h5>
4957
4958<p>
4959The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4960</p>
4961</div>
4962
4963<!-- _______________________________________________________________________ -->
4964<div class="doc_subsubsection">
4965 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4966</div>
4967
4968<div class="doc_text">
4969
4970<h5>Syntax:</h5>
4971<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4972integer bit width. Not all targets support all bit widths however.
4973<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4974 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
4975 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4976 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4977 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
4978 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4979</pre>
4980
4981<h5>Overview:</h5>
4982
4983<p>
4984The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4985leading zeros in a variable.
4986</p>
4987
4988<h5>Arguments:</h5>
4989
4990<p>
4991The only argument is the value to be counted. The argument may be of any
4992integer type. The return type must match the argument type.
4993</p>
4994
4995<h5>Semantics:</h5>
4996
4997<p>
4998The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4999in a variable. If the src == 0 then the result is the size in bits of the type
5000of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5001</p>
5002</div>
5003
5004
5005
5006<!-- _______________________________________________________________________ -->
5007<div class="doc_subsubsection">
5008 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5009</div>
5010
5011<div class="doc_text">
5012
5013<h5>Syntax:</h5>
5014<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5015integer bit width. Not all targets support all bit widths however.
5016<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5017 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
5018 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5019 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5020 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
5021 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5022</pre>
5023
5024<h5>Overview:</h5>
5025
5026<p>
5027The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5028trailing zeros.
5029</p>
5030
5031<h5>Arguments:</h5>
5032
5033<p>
5034The only argument is the value to be counted. The argument may be of any
5035integer type. The return type must match the argument type.
5036</p>
5037
5038<h5>Semantics:</h5>
5039
5040<p>
5041The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5042in a variable. If the src == 0 then the result is the size in bits of the type
5043of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5044</p>
5045</div>
5046
5047<!-- _______________________________________________________________________ -->
5048<div class="doc_subsubsection">
5049 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5050</div>
5051
5052<div class="doc_text">
5053
5054<h5>Syntax:</h5>
5055<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
5056on any integer bit width.
5057<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5058 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5059 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5060</pre>
5061
5062<h5>Overview:</h5>
5063<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5064range of bits from an integer value and returns them in the same bit width as
5065the original value.</p>
5066
5067<h5>Arguments:</h5>
5068<p>The first argument, <tt>%val</tt> and the result may be integer types of
5069any bit width but they must have the same bit width. The second and third
5070arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5071
5072<h5>Semantics:</h5>
5073<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5074of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5075<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5076operates in forward mode.</p>
5077<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5078right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5079only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5080<ol>
5081 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5082 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5083 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5084 to determine the number of bits to retain.</li>
5085 <li>A mask of the retained bits is created by shifting a -1 value.</li>
5086 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
5087</ol>
5088<p>In reverse mode, a similar computation is made except that the bits are
5089returned in the reverse order. So, for example, if <tt>X</tt> has the value
5090<tt>i16 0x0ACF (101011001111)</tt> and we apply
5091<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5092<tt>i16 0x0026 (000000100110)</tt>.</p>
5093</div>
5094
5095<div class="doc_subsubsection">
5096 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5097</div>
5098
5099<div class="doc_text">
5100
5101<h5>Syntax:</h5>
5102<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
5103on any integer bit width.
5104<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5105 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5106 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]5107</pre>
5108
5109<h5>Overview:</h5>
5110<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5111of bits in an integer value with another integer value. It returns the integer
5112with the replaced bits.</p>
5113
5114<h5>Arguments:</h5>
5115<p>The first argument, <tt>%val</tt> and the result may be integer types of
5116any bit width but they must have the same bit width. <tt>%val</tt> is the value
5117whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5118integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5119type since they specify only a bit index.</p>
5120
5121<h5>Semantics:</h5>
5122<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5123of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5124<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5125operates in forward mode.</p>
5126<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5127truncating it down to the size of the replacement area or zero extending it
5128up to that size.</p>
5129<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5130are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5131in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5132to the <tt>%hi</tt>th bit.
5133<p>In reverse mode, a similar computation is made except that the bits are
5134reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5135<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
5136<h5>Examples:</h5>
5137<pre>
5138 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
5139 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
5140 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
5141 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
5142 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
5143</pre>
5144</div>
5145
5146<!-- ======================================================================= -->
5147<div class="doc_subsection">
5148 <a name="int_debugger">Debugger Intrinsics</a>
5149</div>
5150
5151<div class="doc_text">
5152<p>
5153The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5154are described in the <a
5155href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5156Debugging</a> document.
5157</p>
5158</div>
5159
5160
5161<!-- ======================================================================= -->
5162<div class="doc_subsection">
5163 <a name="int_eh">Exception Handling Intrinsics</a>
5164</div>
5165
5166<div class="doc_text">
5167<p> The LLVM exception handling intrinsics (which all start with
5168<tt>llvm.eh.</tt> prefix), are described in the <a
5169href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5170Handling</a> document. </p>
5171</div>
5172
5173<!-- ======================================================================= -->
5174<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5175 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5176</div>
5177
5178<div class="doc_text">
5179<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5180 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5181 the <tt>nest</tt> attribute, from a function. The result is a callable
5182 function pointer lacking the nest parameter - the caller does not need
5183 to provide a value for it. Instead, the value to use is stored in
5184 advance in a "trampoline", a block of memory usually allocated
5185 on the stack, which also contains code to splice the nest value into the
5186 argument list. This is used to implement the GCC nested function address
5187 extension.
5188</p>
5189<p>
5190 For example, if the function is
5191 <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]5192 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]5193<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5194 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5195 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5196 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5197 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5198</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]5199 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5200 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5201</div>
5202
5203<!-- _______________________________________________________________________ -->
5204<div class="doc_subsubsection">
5205 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5206</div>
5207<div class="doc_text">
5208<h5>Syntax:</h5>
5209<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5210declare 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]5211</pre>
5212<h5>Overview:</h5>
5213<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5214 This fills the memory pointed to by <tt>tramp</tt> with code
5215 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5216</p>
5217<h5>Arguments:</h5>
5218<p>
5219 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5220 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5221 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]5222 intrinsic. Note that the size and the alignment are target-specific - LLVM
5223 currently provides no portable way of determining them, so a front-end that
5224 generates this intrinsic needs to have some target-specific knowledge.
5225 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]5226</p>
5227<h5>Semantics:</h5>
5228<p>
5229 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5230 dependent code, turning it into a function. A pointer to this function is
5231 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5232 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5233 before being called. The new function's signature is the same as that of
5234 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5235 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5236 of pointer type. Calling the new function is equivalent to calling
5237 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5238 missing <tt>nest</tt> argument. If, after calling
5239 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5240 modified, then the effect of any later call to the returned function pointer is
5241 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5242</p>
5243</div>
5244
5245<!-- ======================================================================= -->
5246<div class="doc_subsection">
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]5247 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
5248</div>
5249
5250<div class="doc_text">
5251<p>
5252 These intrinsic functions expand the "universal IR" of LLVM to represent
5253 hardware constructs for atomic operations and memory synchronization. This
5254 provides an interface to the hardware, not an interface to the programmer. It
5255 is aimed at a low enough level to allow any programming models or APIs which
5256 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
5257 hardware behavior. Just as hardware provides a "universal IR" for source
5258 languages, it also provides a starting point for developing a "universal"
5259 atomic operation and synchronization IR.
5260</p>
5261<p>
5262 These do <em>not</em> form an API such as high-level threading libraries,
5263 software transaction memory systems, atomic primitives, and intrinsic
5264 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
5265 application libraries. The hardware interface provided by LLVM should allow
5266 a clean implementation of all of these APIs and parallel programming models.
5267 No one model or paradigm should be selected above others unless the hardware
5268 itself ubiquitously does so.
5269
5270</p>
5271</div>
5272
5273<!-- _______________________________________________________________________ -->
5274<div class="doc_subsubsection">
5275 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
5276</div>
5277<div class="doc_text">
5278<h5>Syntax:</h5>
5279<pre>
5280declare void @llvm.memory.barrier( i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt;,
5281i1 &lt;device&gt; )
5282
5283</pre>
5284<h5>Overview:</h5>
5285<p>
5286 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
5287 specific pairs of memory access types.
5288</p>
5289<h5>Arguments:</h5>
5290<p>
5291 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
5292 The first four arguments enables a specific barrier as listed below. The fith
5293 argument specifies that the barrier applies to io or device or uncached memory.
5294
5295</p>
5296 <ul>
5297 <li><tt>ll</tt>: load-load barrier</li>
5298 <li><tt>ls</tt>: load-store barrier</li>
5299 <li><tt>sl</tt>: store-load barrier</li>
5300 <li><tt>ss</tt>: store-store barrier</li>
5301 <li><tt>device</tt>: barrier applies to device and uncached memory also.
5302 </ul>
5303<h5>Semantics:</h5>
5304<p>
5305 This intrinsic causes the system to enforce some ordering constraints upon
5306 the loads and stores of the program. This barrier does not indicate
5307 <em>when</em> any events will occur, it only enforces an <em>order</em> in
5308 which they occur. For any of the specified pairs of load and store operations
5309 (f.ex. load-load, or store-load), all of the first operations preceding the
5310 barrier will complete before any of the second operations succeeding the
5311 barrier begin. Specifically the semantics for each pairing is as follows:
5312</p>
5313 <ul>
5314 <li><tt>ll</tt>: All loads before the barrier must complete before any load
5315 after the barrier begins.</li>
5316
5317 <li><tt>ls</tt>: All loads before the barrier must complete before any
5318 store after the barrier begins.</li>
5319 <li><tt>ss</tt>: All stores before the barrier must complete before any
5320 store after the barrier begins.</li>
5321 <li><tt>sl</tt>: All stores before the barrier must complete before any
5322 load after the barrier begins.</li>
5323 </ul>
5324<p>
5325 These semantics are applied with a logical "and" behavior when more than one
5326 is enabled in a single memory barrier intrinsic.
5327</p>
5328<p>
5329 Backends may implement stronger barriers than those requested when they do not
5330 support as fine grained a barrier as requested. Some architectures do not
5331 need all types of barriers and on such architectures, these become noops.
5332</p>
5333<h5>Example:</h5>
5334<pre>
5335%ptr = malloc i32
5336 store i32 4, %ptr
5337
5338%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
5339 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
5340 <i>; guarantee the above finishes</i>
5341 store i32 8, %ptr <i>; before this begins</i>
5342</pre>
5343</div>
5344
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5345<!-- _______________________________________________________________________ -->
5346<div class="doc_subsubsection">
5347 <a name="int_atomic_lcs">'<tt>llvm.atomic.lcs.*</tt>' Intrinsic</a>
5348</div>
5349<div class="doc_text">
5350<h5>Syntax:</h5>
5351<p>
5352 This is an overloaded intrinsic. You can use <tt>llvm.atomic.lcs</tt> on any
5353 integer bit width. Not all targets support all bit widths however.</p>
5354
5355<pre>
5356declare i8 @llvm.atomic.lcs.i8( i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt; )
5357declare i16 @llvm.atomic.lcs.i16( i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt; )
5358declare i32 @llvm.atomic.lcs.i32( i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt; )
5359declare i64 @llvm.atomic.lcs.i64( i64* &lt;ptr&gt;, i64 &lt;cmp&gt;, i64 &lt;val&gt; )
5360
5361</pre>
5362<h5>Overview:</h5>
5363<p>
5364 This loads a value in memory and compares it to a given value. If they are
5365 equal, it stores a new value into the memory.
5366</p>
5367<h5>Arguments:</h5>
5368<p>
5369 The <tt>llvm.atomic.lcs</tt> intrinsic takes three arguments. The result as
5370 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
5371 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
5372 this integer type. While any bit width integer may be used, targets may only
5373 lower representations they support in hardware.
5374
5375</p>
5376<h5>Semantics:</h5>
5377<p>
5378 This entire intrinsic must be executed atomically. It first loads the value
5379 in memory pointed to by <tt>ptr</tt> and compares it with the value
5380 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
5381 loaded value is yielded in all cases. This provides the equivalent of an
5382 atomic compare-and-swap operation within the SSA framework.
5383</p>
5384<h5>Examples:</h5>
5385
5386<pre>
5387%ptr = malloc i32
5388 store i32 4, %ptr
5389
5390%val1 = add i32 4, 4
5391%result1 = call i32 @llvm.atomic.lcs.i32( i32* %ptr, i32 4, %val1 )
5392 <i>; yields {i32}:result1 = 4</i>
5393%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5394%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5395
5396%val2 = add i32 1, 1
5397%result2 = call i32 @llvm.atomic.lcs.i32( i32* %ptr, i32 5, %val2 )
5398 <i>; yields {i32}:result2 = 8</i>
5399%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
5400
5401%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
5402</pre>
5403</div>
5404
5405<!-- _______________________________________________________________________ -->
5406<div class="doc_subsubsection">
5407 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
5408</div>
5409<div class="doc_text">
5410<h5>Syntax:</h5>
5411
5412<p>
5413 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
5414 integer bit width. Not all targets support all bit widths however.</p>
5415<pre>
5416declare i8 @llvm.atomic.swap.i8( i8* &lt;ptr&gt;, i8 &lt;val&gt; )
5417declare i16 @llvm.atomic.swap.i16( i16* &lt;ptr&gt;, i16 &lt;val&gt; )
5418declare i32 @llvm.atomic.swap.i32( i32* &lt;ptr&gt;, i32 &lt;val&gt; )
5419declare i64 @llvm.atomic.swap.i64( i64* &lt;ptr&gt;, i64 &lt;val&gt; )
5420
5421</pre>
5422<h5>Overview:</h5>
5423<p>
5424 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
5425 the value from memory. It then stores the value in <tt>val</tt> in the memory
5426 at <tt>ptr</tt>.
5427</p>
5428<h5>Arguments:</h5>
5429
5430<p>
5431 The <tt>llvm.atomic.ls</tt> intrinsic takes two arguments. Both the
5432 <tt>val</tt> argument and the result must be integers of the same bit width.
5433 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
5434 integer type. The targets may only lower integer representations they
5435 support.
5436</p>
5437<h5>Semantics:</h5>
5438<p>
5439 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
5440 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
5441 equivalent of an atomic swap operation within the SSA framework.
5442
5443</p>
5444<h5>Examples:</h5>
5445<pre>
5446%ptr = malloc i32
5447 store i32 4, %ptr
5448
5449%val1 = add i32 4, 4
5450%result1 = call i32 @llvm.atomic.swap.i32( i32* %ptr, i32 %val1 )
5451 <i>; yields {i32}:result1 = 4</i>
5452%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5453%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5454
5455%val2 = add i32 1, 1
5456%result2 = call i32 @llvm.atomic.swap.i32( i32* %ptr, i32 %val2 )
5457 <i>; yields {i32}:result2 = 8</i>
5458
5459%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
5460%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
5461</pre>
5462</div>
5463
5464<!-- _______________________________________________________________________ -->
5465<div class="doc_subsubsection">
5466 <a name="int_atomic_las">'<tt>llvm.atomic.las.*</tt>' Intrinsic</a>
5467
5468</div>
5469<div class="doc_text">
5470<h5>Syntax:</h5>
5471<p>
5472 This is an overloaded intrinsic. You can use <tt>llvm.atomic.las</tt> on any
5473 integer bit width. Not all targets support all bit widths however.</p>
5474<pre>
5475declare i8 @llvm.atomic.las.i8.( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
5476declare i16 @llvm.atomic.las.i16.( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
5477declare i32 @llvm.atomic.las.i32.( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
5478declare i64 @llvm.atomic.las.i64.( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
5479
5480</pre>
5481<h5>Overview:</h5>
5482<p>
5483 This intrinsic adds <tt>delta</tt> to the value stored in memory at
5484 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
5485</p>
5486<h5>Arguments:</h5>
5487<p>
5488
5489 The intrinsic takes two arguments, the first a pointer to an integer value
5490 and the second an integer value. The result is also an integer value. These
5491 integer types can have any bit width, but they must all have the same bit
5492 width. The targets may only lower integer representations they support.
5493</p>
5494<h5>Semantics:</h5>
5495<p>
5496 This intrinsic does a series of operations atomically. It first loads the
5497 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
5498 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
5499</p>
5500
5501<h5>Examples:</h5>
5502<pre>
5503%ptr = malloc i32
5504 store i32 4, %ptr
5505%result1 = call i32 @llvm.atomic.las.i32( i32* %ptr, i32 4 )
5506 <i>; yields {i32}:result1 = 4</i>
5507%result2 = call i32 @llvm.atomic.las.i32( i32* %ptr, i32 2 )
5508 <i>; yields {i32}:result2 = 8</i>
5509%result3 = call i32 @llvm.atomic.las.i32( i32* %ptr, i32 5 )
5510 <i>; yields {i32}:result3 = 10</i>
5511%memval = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
5512</pre>
5513</div>
5514
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]5515
5516<!-- ======================================================================= -->
5517<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5518 <a name="int_general">General Intrinsics</a>
5519</div>
5520
5521<div class="doc_text">
5522<p> This class of intrinsics is designed to be generic and has
5523no specific purpose. </p>
5524</div>
5525
5526<!-- _______________________________________________________________________ -->
5527<div class="doc_subsubsection">
5528 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
5529</div>
5530
5531<div class="doc_text">
5532
5533<h5>Syntax:</h5>
5534<pre>
5535 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5536</pre>
5537
5538<h5>Overview:</h5>
5539
5540<p>
5541The '<tt>llvm.var.annotation</tt>' intrinsic
5542</p>
5543
5544<h5>Arguments:</h5>
5545
5546<p>
5547The first argument is a pointer to a value, the second is a pointer to a
5548global string, the third is a pointer to a global string which is the source
5549file name, and the last argument is the line number.
5550</p>
5551
5552<h5>Semantics:</h5>
5553
5554<p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]5555This intrinsic allows annotation of local variables with arbitrary strings.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5556This can be useful for special purpose optimizations that want to look for these
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]5557annotations. These have no other defined use, they are ignored by code
5558generation and optimization.
5559</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5560</div>
5561
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5562<!-- _______________________________________________________________________ -->
5563<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]5564 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5565</div>
5566
5567<div class="doc_text">
5568
5569<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5570<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
5571any integer bit width.
5572</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5573<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]5574 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5575 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5576 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5577 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5578 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]5579</pre>
5580
5581<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5582
5583<p>
5584The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5585</p>
5586
5587<h5>Arguments:</h5>
5588
5589<p>
5590The first argument is an integer value (result of some expression),
5591the second is a pointer to a global string, the third is a pointer to a global
5592string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5593It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5594</p>
5595
5596<h5>Semantics:</h5>
5597
5598<p>
5599This intrinsic allows annotations to be put on arbitrary expressions
5600with arbitrary strings. This can be useful for special purpose optimizations
5601that want to look for these annotations. These have no other defined use, they
5602are ignored by code generation and optimization.
5603</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5604
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]5605<!-- _______________________________________________________________________ -->
5606<div class="doc_subsubsection">
5607 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
5608</div>
5609
5610<div class="doc_text">
5611
5612<h5>Syntax:</h5>
5613<pre>
5614 declare void @llvm.trap()
5615</pre>
5616
5617<h5>Overview:</h5>
5618
5619<p>
5620The '<tt>llvm.trap</tt>' intrinsic
5621</p>
5622
5623<h5>Arguments:</h5>
5624
5625<p>
5626None
5627</p>
5628
5629<h5>Semantics:</h5>
5630
5631<p>
5632This intrinsics is lowered to the target dependent trap instruction. If the
5633target does not have a trap instruction, this intrinsic will be lowered to the
5634call of the abort() function.
5635</p>
5636</div>
5637
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5638<!-- *********************************************************************** -->
5639<hr>
5640<address>
5641 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
5642 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
5643 <a href="http://validator.w3.org/check/referer"><img
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]5644 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5645
5646 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
5647 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
5648 Last modified: $Date$
5649</address>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]5650
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5651</body>
5652</html>