Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 500d39425a2376eb8b5c5b94526ef33e89308d57 / . / docs / LangRef.html
blob: 8adc4d7a0f60157385817809bb16f279342cbc3d [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>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]114 <li><a href="#aggregateops">Aggregate Operations</a>
115 <ol>
116 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
117 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
118 </ol>
119 </li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]120 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
121 <ol>
122 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
123 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
124 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
125 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
126 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
127 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
128 </ol>
129 </li>
130 <li><a href="#convertops">Conversion Operations</a>
131 <ol>
132 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
133 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
134 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
135 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
136 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
137 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
138 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
139 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
140 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
141 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
142 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
143 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
144 </ol>
145 <li><a href="#otherops">Other Operations</a>
146 <ol>
147 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
148 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]149 <li><a href="#i_vicmp">'<tt>vicmp</tt>' Instruction</a></li>
150 <li><a href="#i_vfcmp">'<tt>vfcmp</tt>' Instruction</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]151 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
152 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
153 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
154 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]155 <li><a href="#i_getresult">'<tt>getresult</tt>' Instruction</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]156 </ol>
157 </li>
158 </ol>
159 </li>
160 <li><a href="#intrinsics">Intrinsic Functions</a>
161 <ol>
162 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
163 <ol>
164 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
165 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
166 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
167 </ol>
168 </li>
169 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
170 <ol>
171 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
172 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
173 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
174 </ol>
175 </li>
176 <li><a href="#int_codegen">Code Generator Intrinsics</a>
177 <ol>
178 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
179 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
180 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
181 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
182 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
183 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
184 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
185 </ol>
186 </li>
187 <li><a href="#int_libc">Standard C Library Intrinsics</a>
188 <ol>
189 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
190 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
191 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
192 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
193 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]194 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
195 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
196 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]197 </ol>
198 </li>
199 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
200 <ol>
201 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
202 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
203 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
204 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
205 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
206 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
207 </ol>
208 </li>
209 <li><a href="#int_debugger">Debugger intrinsics</a></li>
210 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]211 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]212 <ol>
213 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]214 </ol>
215 </li>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]216 <li><a href="#int_atomics">Atomic intrinsics</a>
217 <ol>
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]218 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
219 <li><a href="#int_atomic_lcs"><tt>llvm.atomic.lcs</tt></a></li>
220 <li><a href="#int_atomic_las"><tt>llvm.atomic.las</tt></a></li>
221 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]222 </ol>
223 </li>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]224 <li><a href="#int_general">General intrinsics</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]225 <ol>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]226 <li><a href="#int_var_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]227 <tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]228 <li><a href="#int_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]229 <tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]230 <li><a href="#int_trap">
231 <tt>llvm.trap</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]232 </ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]233 </li>
234 </ol>
235 </li>
236</ol>
237
238<div class="doc_author">
239 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
240 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
241</div>
242
243<!-- *********************************************************************** -->
244<div class="doc_section"> <a name="abstract">Abstract </a></div>
245<!-- *********************************************************************** -->
246
247<div class="doc_text">
248<p>This document is a reference manual for the LLVM assembly language.
249LLVM is an SSA based representation that provides type safety,
250low-level operations, flexibility, and the capability of representing
251'all' high-level languages cleanly. It is the common code
252representation used throughout all phases of the LLVM compilation
253strategy.</p>
254</div>
255
256<!-- *********************************************************************** -->
257<div class="doc_section"> <a name="introduction">Introduction</a> </div>
258<!-- *********************************************************************** -->
259
260<div class="doc_text">
261
262<p>The LLVM code representation is designed to be used in three
263different forms: as an in-memory compiler IR, as an on-disk bitcode
264representation (suitable for fast loading by a Just-In-Time compiler),
265and as a human readable assembly language representation. This allows
266LLVM to provide a powerful intermediate representation for efficient
267compiler transformations and analysis, while providing a natural means
268to debug and visualize the transformations. The three different forms
269of LLVM are all equivalent. This document describes the human readable
270representation and notation.</p>
271
272<p>The LLVM representation aims to be light-weight and low-level
273while being expressive, typed, and extensible at the same time. It
274aims to be a "universal IR" of sorts, by being at a low enough level
275that high-level ideas may be cleanly mapped to it (similar to how
276microprocessors are "universal IR's", allowing many source languages to
277be mapped to them). By providing type information, LLVM can be used as
278the target of optimizations: for example, through pointer analysis, it
279can be proven that a C automatic variable is never accessed outside of
280the current function... allowing it to be promoted to a simple SSA
281value instead of a memory location.</p>
282
283</div>
284
285<!-- _______________________________________________________________________ -->
286<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
287
288<div class="doc_text">
289
290<p>It is important to note that this document describes 'well formed'
291LLVM assembly language. There is a difference between what the parser
292accepts and what is considered 'well formed'. For example, the
293following instruction is syntactically okay, but not well formed:</p>
294
295<div class="doc_code">
296<pre>
297%x = <a href="#i_add">add</a> i32 1, %x
298</pre>
299</div>
300
301<p>...because the definition of <tt>%x</tt> does not dominate all of
302its uses. The LLVM infrastructure provides a verification pass that may
303be used to verify that an LLVM module is well formed. This pass is
304automatically run by the parser after parsing input assembly and by
305the optimizer before it outputs bitcode. The violations pointed out
306by the verifier pass indicate bugs in transformation passes or input to
307the parser.</p>
308</div>
309
Chris Lattnera83fdc02007-10-03 17:34:29 +0000[diff] [blame]310<!-- Describe the typesetting conventions here. -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]311
312<!-- *********************************************************************** -->
313<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
314<!-- *********************************************************************** -->
315
316<div class="doc_text">
317
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]318 <p>LLVM identifiers come in two basic types: global and local. Global
319 identifiers (functions, global variables) begin with the @ character. Local
320 identifiers (register names, types) begin with the % character. Additionally,
321 there are three different formats for identifiers, for different purposes:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]322
323<ol>
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]324 <li>Named values are represented as a string of characters with their prefix.
325 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
326 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]327 Identifiers which require other characters in their names can be surrounded
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]328 with quotes. In this way, anything except a <tt>&quot;</tt> character can
329 be used in a named value.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]330
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]331 <li>Unnamed values are represented as an unsigned numeric value with their
332 prefix. For example, %12, @2, %44.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]333
334 <li>Constants, which are described in a <a href="#constants">section about
335 constants</a>, below.</li>
336</ol>
337
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]338<p>LLVM requires that values start with a prefix for two reasons: Compilers
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]339don't need to worry about name clashes with reserved words, and the set of
340reserved words may be expanded in the future without penalty. Additionally,
341unnamed identifiers allow a compiler to quickly come up with a temporary
342variable without having to avoid symbol table conflicts.</p>
343
344<p>Reserved words in LLVM are very similar to reserved words in other
345languages. There are keywords for different opcodes
346('<tt><a href="#i_add">add</a></tt>',
347 '<tt><a href="#i_bitcast">bitcast</a></tt>',
348 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
349href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
350and others. These reserved words cannot conflict with variable names, because
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]351none of them start with a prefix character ('%' or '@').</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]352
353<p>Here is an example of LLVM code to multiply the integer variable
354'<tt>%X</tt>' by 8:</p>
355
356<p>The easy way:</p>
357
358<div class="doc_code">
359<pre>
360%result = <a href="#i_mul">mul</a> i32 %X, 8
361</pre>
362</div>
363
364<p>After strength reduction:</p>
365
366<div class="doc_code">
367<pre>
368%result = <a href="#i_shl">shl</a> i32 %X, i8 3
369</pre>
370</div>
371
372<p>And the hard way:</p>
373
374<div class="doc_code">
375<pre>
376<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
377<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
378%result = <a href="#i_add">add</a> i32 %1, %1
379</pre>
380</div>
381
382<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
383important lexical features of LLVM:</p>
384
385<ol>
386
387 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
388 line.</li>
389
390 <li>Unnamed temporaries are created when the result of a computation is not
391 assigned to a named value.</li>
392
393 <li>Unnamed temporaries are numbered sequentially</li>
394
395</ol>
396
397<p>...and it also shows a convention that we follow in this document. When
398demonstrating instructions, we will follow an instruction with a comment that
399defines the type and name of value produced. Comments are shown in italic
400text.</p>
401
402</div>
403
404<!-- *********************************************************************** -->
405<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
406<!-- *********************************************************************** -->
407
408<!-- ======================================================================= -->
409<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
410</div>
411
412<div class="doc_text">
413
414<p>LLVM programs are composed of "Module"s, each of which is a
415translation unit of the input programs. Each module consists of
416functions, global variables, and symbol table entries. Modules may be
417combined together with the LLVM linker, which merges function (and
418global variable) definitions, resolves forward declarations, and merges
419symbol table entries. Here is an example of the "hello world" module:</p>
420
421<div class="doc_code">
422<pre><i>; Declare the string constant as a global constant...</i>
423<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
424 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
425
426<i>; External declaration of the puts function</i>
427<a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
428
429<i>; Definition of main function</i>
430define i32 @main() { <i>; i32()* </i>
431 <i>; Convert [13x i8 ]* to i8 *...</i>
432 %cast210 = <a
433 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
434
435 <i>; Call puts function to write out the string to stdout...</i>
436 <a
437 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
438 <a
439 href="#i_ret">ret</a> i32 0<br>}<br>
440</pre>
441</div>
442
443<p>This example is made up of a <a href="#globalvars">global variable</a>
444named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
445function, and a <a href="#functionstructure">function definition</a>
446for "<tt>main</tt>".</p>
447
448<p>In general, a module is made up of a list of global values,
449where both functions and global variables are global values. Global values are
450represented by a pointer to a memory location (in this case, a pointer to an
451array of char, and a pointer to a function), and have one of the following <a
452href="#linkage">linkage types</a>.</p>
453
454</div>
455
456<!-- ======================================================================= -->
457<div class="doc_subsection">
458 <a name="linkage">Linkage Types</a>
459</div>
460
461<div class="doc_text">
462
463<p>
464All Global Variables and Functions have one of the following types of linkage:
465</p>
466
467<dl>
468
469 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
470
471 <dd>Global values with internal linkage are only directly accessible by
472 objects in the current module. In particular, linking code into a module with
473 an internal global value may cause the internal to be renamed as necessary to
474 avoid collisions. Because the symbol is internal to the module, all
475 references can be updated. This corresponds to the notion of the
476 '<tt>static</tt>' keyword in C.
477 </dd>
478
479 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
480
481 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
482 the same name when linkage occurs. This is typically used to implement
483 inline functions, templates, or other code which must be generated in each
484 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
485 allowed to be discarded.
486 </dd>
487
488 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
489
490 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
491 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
492 used for globals that may be emitted in multiple translation units, but that
493 are not guaranteed to be emitted into every translation unit that uses them.
494 One example of this are common globals in C, such as "<tt>int X;</tt>" at
495 global scope.
496 </dd>
497
498 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
499
500 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
501 pointer to array type. When two global variables with appending linkage are
502 linked together, the two global arrays are appended together. This is the
503 LLVM, typesafe, equivalent of having the system linker append together
504 "sections" with identical names when .o files are linked.
505 </dd>
506
507 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
508 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
509 until linked, if not linked, the symbol becomes null instead of being an
510 undefined reference.
511 </dd>
512
513 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
514
515 <dd>If none of the above identifiers are used, the global is externally
516 visible, meaning that it participates in linkage and can be used to resolve
517 external symbol references.
518 </dd>
519</dl>
520
521 <p>
522 The next two types of linkage are targeted for Microsoft Windows platform
523 only. They are designed to support importing (exporting) symbols from (to)
524 DLLs.
525 </p>
526
527 <dl>
528 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
529
530 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
531 or variable via a global pointer to a pointer that is set up by the DLL
532 exporting the symbol. On Microsoft Windows targets, the pointer name is
533 formed by combining <code>_imp__</code> and the function or variable name.
534 </dd>
535
536 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
537
538 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
539 pointer to a pointer in a DLL, so that it can be referenced with the
540 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
541 name is formed by combining <code>_imp__</code> and the function or variable
542 name.
543 </dd>
544
545</dl>
546
547<p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
548variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
549variable and was linked with this one, one of the two would be renamed,
550preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
551external (i.e., lacking any linkage declarations), they are accessible
552outside of the current module.</p>
553<p>It is illegal for a function <i>declaration</i>
554to have any linkage type other than "externally visible", <tt>dllimport</tt>,
555or <tt>extern_weak</tt>.</p>
556<p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
557linkages.
558</div>
559
560<!-- ======================================================================= -->
561<div class="doc_subsection">
562 <a name="callingconv">Calling Conventions</a>
563</div>
564
565<div class="doc_text">
566
567<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
568and <a href="#i_invoke">invokes</a> can all have an optional calling convention
569specified for the call. The calling convention of any pair of dynamic
570caller/callee must match, or the behavior of the program is undefined. The
571following calling conventions are supported by LLVM, and more may be added in
572the future:</p>
573
574<dl>
575 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
576
577 <dd>This calling convention (the default if no other calling convention is
578 specified) matches the target C calling conventions. This calling convention
579 supports varargs function calls and tolerates some mismatch in the declared
580 prototype and implemented declaration of the function (as does normal C).
581 </dd>
582
583 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
584
585 <dd>This calling convention attempts to make calls as fast as possible
586 (e.g. by passing things in registers). This calling convention allows the
587 target to use whatever tricks it wants to produce fast code for the target,
588 without having to conform to an externally specified ABI. Implementations of
Arnold Schwaighofer07444922008-05-14 09:17:12 +0000[diff] [blame]589 this convention should allow arbitrary
590 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> to be
591 supported. This calling convention does not support varargs and requires the
592 prototype of all callees to exactly match the prototype of the function
593 definition.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]594 </dd>
595
596 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
597
598 <dd>This calling convention attempts to make code in the caller as efficient
599 as possible under the assumption that the call is not commonly executed. As
600 such, these calls often preserve all registers so that the call does not break
601 any live ranges in the caller side. This calling convention does not support
602 varargs and requires the prototype of all callees to exactly match the
603 prototype of the function definition.
604 </dd>
605
606 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
607
608 <dd>Any calling convention may be specified by number, allowing
609 target-specific calling conventions to be used. Target specific calling
610 conventions start at 64.
611 </dd>
612</dl>
613
614<p>More calling conventions can be added/defined on an as-needed basis, to
615support pascal conventions or any other well-known target-independent
616convention.</p>
617
618</div>
619
620<!-- ======================================================================= -->
621<div class="doc_subsection">
622 <a name="visibility">Visibility Styles</a>
623</div>
624
625<div class="doc_text">
626
627<p>
628All Global Variables and Functions have one of the following visibility styles:
629</p>
630
631<dl>
632 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
633
634 <dd>On ELF, default visibility means that the declaration is visible to other
635 modules and, in shared libraries, means that the declared entity may be
636 overridden. On Darwin, default visibility means that the declaration is
637 visible to other modules. Default visibility corresponds to "external
638 linkage" in the language.
639 </dd>
640
641 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
642
643 <dd>Two declarations of an object with hidden visibility refer to the same
644 object if they are in the same shared object. Usually, hidden visibility
645 indicates that the symbol will not be placed into the dynamic symbol table,
646 so no other module (executable or shared library) can reference it
647 directly.
648 </dd>
649
650 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
651
652 <dd>On ELF, protected visibility indicates that the symbol will be placed in
653 the dynamic symbol table, but that references within the defining module will
654 bind to the local symbol. That is, the symbol cannot be overridden by another
655 module.
656 </dd>
657</dl>
658
659</div>
660
661<!-- ======================================================================= -->
662<div class="doc_subsection">
663 <a name="globalvars">Global Variables</a>
664</div>
665
666<div class="doc_text">
667
668<p>Global variables define regions of memory allocated at compilation time
669instead of run-time. Global variables may optionally be initialized, may have
670an explicit section to be placed in, and may have an optional explicit alignment
671specified. A variable may be defined as "thread_local", which means that it
672will not be shared by threads (each thread will have a separated copy of the
673variable). A variable may be defined as a global "constant," which indicates
674that the contents of the variable will <b>never</b> be modified (enabling better
675optimization, allowing the global data to be placed in the read-only section of
676an executable, etc). Note that variables that need runtime initialization
677cannot be marked "constant" as there is a store to the variable.</p>
678
679<p>
680LLVM explicitly allows <em>declarations</em> of global variables to be marked
681constant, even if the final definition of the global is not. This capability
682can be used to enable slightly better optimization of the program, but requires
683the language definition to guarantee that optimizations based on the
684'constantness' are valid for the translation units that do not include the
685definition.
686</p>
687
688<p>As SSA values, global variables define pointer values that are in
689scope (i.e. they dominate) all basic blocks in the program. Global
690variables always define a pointer to their "content" type because they
691describe a region of memory, and all memory objects in LLVM are
692accessed through pointers.</p>
693
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]694<p>A global variable may be declared to reside in a target-specifc numbered
695address space. For targets that support them, address spaces may affect how
696optimizations are performed and/or what target instructions are used to access
Christopher Lamb20a39e92007-12-12 08:44:39 +0000[diff] [blame]697the variable. The default address space is zero. The address space qualifier
698must precede any other attributes.</p>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]699
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]700<p>LLVM allows an explicit section to be specified for globals. If the target
701supports it, it will emit globals to the section specified.</p>
702
703<p>An explicit alignment may be specified for a global. If not present, or if
704the alignment is set to zero, the alignment of the global is set by the target
705to whatever it feels convenient. If an explicit alignment is specified, the
706global is forced to have at least that much alignment. All alignments must be
707a power of 2.</p>
708
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]709<p>For example, the following defines a global in a numbered address space with
710an initializer, section, and alignment:</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]711
712<div class="doc_code">
713<pre>
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]714@G = constant float 1.0 addrspace(5), section "foo", align 4
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]715</pre>
716</div>
717
718</div>
719
720
721<!-- ======================================================================= -->
722<div class="doc_subsection">
723 <a name="functionstructure">Functions</a>
724</div>
725
726<div class="doc_text">
727
728<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
729an optional <a href="#linkage">linkage type</a>, an optional
730<a href="#visibility">visibility style</a>, an optional
731<a href="#callingconv">calling convention</a>, a return type, an optional
732<a href="#paramattrs">parameter attribute</a> for the return type, a function
733name, a (possibly empty) argument list (each with optional
734<a href="#paramattrs">parameter attributes</a>), an optional section, an
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]735optional alignment, an optional <a href="#gc">garbage collector name</a>, an
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]736opening curly brace, a list of basic blocks, and a closing curly brace.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]737
738LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
739optional <a href="#linkage">linkage type</a>, an optional
740<a href="#visibility">visibility style</a>, an optional
741<a href="#callingconv">calling convention</a>, a return type, an optional
742<a href="#paramattrs">parameter attribute</a> for the return type, a function
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]743name, a possibly empty list of arguments, an optional alignment, and an optional
Gordon Henriksend606f5b2007-12-10 03:30:21 +0000[diff] [blame]744<a href="#gc">garbage collector name</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]745
746<p>A function definition contains a list of basic blocks, forming the CFG for
747the function. Each basic block may optionally start with a label (giving the
748basic block a symbol table entry), contains a list of instructions, and ends
749with a <a href="#terminators">terminator</a> instruction (such as a branch or
750function return).</p>
751
752<p>The first basic block in a function is special in two ways: it is immediately
753executed on entrance to the function, and it is not allowed to have predecessor
754basic blocks (i.e. there can not be any branches to the entry block of a
755function). Because the block can have no predecessors, it also cannot have any
756<a href="#i_phi">PHI nodes</a>.</p>
757
758<p>LLVM allows an explicit section to be specified for functions. If the target
759supports it, it will emit functions to the section specified.</p>
760
761<p>An explicit alignment may be specified for a function. If not present, or if
762the alignment is set to zero, the alignment of the function is set by the target
763to whatever it feels convenient. If an explicit alignment is specified, the
764function is forced to have at least that much alignment. All alignments must be
765a power of 2.</p>
766
767</div>
768
769
770<!-- ======================================================================= -->
771<div class="doc_subsection">
772 <a name="aliasstructure">Aliases</a>
773</div>
774<div class="doc_text">
775 <p>Aliases act as "second name" for the aliasee value (which can be either
Anton Korobeynikov96822812008-03-22 08:36:14 +0000[diff] [blame]776 function, global variable, another alias or bitcast of global value). Aliases
777 may have an optional <a href="#linkage">linkage type</a>, and an
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]778 optional <a href="#visibility">visibility style</a>.</p>
779
780 <h5>Syntax:</h5>
781
782<div class="doc_code">
783<pre>
784@&lt;Name&gt; = [Linkage] [Visibility] alias &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
785</pre>
786</div>
787
788</div>
789
790
791
792<!-- ======================================================================= -->
793<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
794<div class="doc_text">
795 <p>The return type and each parameter of a function type may have a set of
796 <i>parameter attributes</i> associated with them. Parameter attributes are
797 used to communicate additional information about the result or parameters of
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]798 a function. Parameter attributes are considered to be part of the function,
799 not of the function type, so functions with different parameter attributes
800 can have the same function type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]801
802 <p>Parameter attributes are simple keywords that follow the type specified. If
803 multiple parameter attributes are needed, they are space separated. For
804 example:</p>
805
806<div class="doc_code">
807<pre>
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]808declare i32 @printf(i8* noalias , ...) nounwind
809declare i32 @atoi(i8*) nounwind readonly
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]810</pre>
811</div>
812
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]813 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
814 <tt>readonly</tt>) come immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]815
816 <p>Currently, only the following parameter attributes are defined:</p>
817 <dl>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]818 <dt><tt>zeroext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]819 <dd>This indicates that the parameter should be zero extended just before
820 a call to this function.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]821
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]822 <dt><tt>signext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]823 <dd>This indicates that the parameter should be sign extended just before
824 a call to this function.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]825
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]826 <dt><tt>inreg</tt></dt>
827 <dd>This indicates that the parameter should be placed in register (if
828 possible) during assembling function call. Support for this attribute is
829 target-specific</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]830
831 <dt><tt>byval</tt></dt>
Chris Lattner04c86182008-01-15 04:34:22 +0000[diff] [blame]832 <dd>This indicates that the pointer parameter should really be passed by
833 value to the function. The attribute implies that a hidden copy of the
834 pointee is made between the caller and the callee, so the callee is unable
835 to modify the value in the callee. This attribute is only valid on llvm
836 pointer arguments. It is generally used to pass structs and arrays by
837 value, but is also valid on scalars (even though this is silly).</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]838
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]839 <dt><tt>sret</tt></dt>
Duncan Sands616cc032008-02-18 04:19:38 +0000[diff] [blame]840 <dd>This indicates that the pointer parameter specifies the address of a
841 structure that is the return value of the function in the source program.
Duncan Sandsef0e9e42008-03-17 12:17:41 +0000[diff] [blame]842 Loads and stores to the structure are assumed not to trap.
Duncan Sands616cc032008-02-18 04:19:38 +0000[diff] [blame]843 May only be applied to the first parameter.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]844
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]845 <dt><tt>noalias</tt></dt>
Owen Andersonc4fc4cd2008-02-18 04:09:01 +0000[diff] [blame]846 <dd>This indicates that the parameter does not alias any global or any other
847 parameter. The caller is responsible for ensuring that this is the case,
848 usually by placing the value in a stack allocation.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]849
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]850 <dt><tt>noreturn</tt></dt>
851 <dd>This function attribute indicates that the function never returns. This
852 indicates to LLVM that every call to this function should be treated as if
853 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
Chris Lattner275e6be2008-01-11 06:20:47 +0000[diff] [blame]854
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]855 <dt><tt>nounwind</tt></dt>
Duncan Sandsef0e9e42008-03-17 12:17:41 +0000[diff] [blame]856 <dd>This function attribute indicates that no exceptions unwind out of the
857 function. Usually this is because the function makes no use of exceptions,
858 but it may also be that the function catches any exceptions thrown when
859 executing it.</dd>
860
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]861 <dt><tt>nest</tt></dt>
862 <dd>This indicates that the parameter can be excised using the
863 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
Duncan Sands13e13f82007-11-22 20:23:04 +0000[diff] [blame]864 <dt><tt>readonly</tt></dt>
Duncan Sandsd69c0e62007-11-14 21:14:02 +0000[diff] [blame]865 <dd>This function attribute indicates that the function has no side-effects
Duncan Sands13e13f82007-11-22 20:23:04 +0000[diff] [blame]866 except for producing a return value or throwing an exception. The value
867 returned must only depend on the function arguments and/or global variables.
868 It may use values obtained by dereferencing pointers.</dd>
869 <dt><tt>readnone</tt></dt>
870 <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]871 function, but in addition it is not allowed to dereference any pointer arguments
872 or global variables.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]873 </dl>
874
875</div>
876
877<!-- ======================================================================= -->
878<div class="doc_subsection">
Gordon Henriksen13fe5e32007-12-10 03:18:06 +0000[diff] [blame]879 <a name="gc">Garbage Collector Names</a>
880</div>
881
882<div class="doc_text">
883<p>Each function may specify a garbage collector name, which is simply a
884string.</p>
885
886<div class="doc_code"><pre
887>define void @f() gc "name" { ...</pre></div>
888
889<p>The compiler declares the supported values of <i>name</i>. Specifying a
890collector which will cause the compiler to alter its output in order to support
891the named garbage collection algorithm.</p>
892</div>
893
894<!-- ======================================================================= -->
895<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]896 <a name="moduleasm">Module-Level Inline Assembly</a>
897</div>
898
899<div class="doc_text">
900<p>
901Modules may contain "module-level inline asm" blocks, which corresponds to the
902GCC "file scope inline asm" blocks. These blocks are internally concatenated by
903LLVM and treated as a single unit, but may be separated in the .ll file if
904desired. The syntax is very simple:
905</p>
906
907<div class="doc_code">
908<pre>
909module asm "inline asm code goes here"
910module asm "more can go here"
911</pre>
912</div>
913
914<p>The strings can contain any character by escaping non-printable characters.
915 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
916 for the number.
917</p>
918
919<p>
920 The inline asm code is simply printed to the machine code .s file when
921 assembly code is generated.
922</p>
923</div>
924
925<!-- ======================================================================= -->
926<div class="doc_subsection">
927 <a name="datalayout">Data Layout</a>
928</div>
929
930<div class="doc_text">
931<p>A module may specify a target specific data layout string that specifies how
932data is to be laid out in memory. The syntax for the data layout is simply:</p>
933<pre> target datalayout = "<i>layout specification</i>"</pre>
934<p>The <i>layout specification</i> consists of a list of specifications
935separated by the minus sign character ('-'). Each specification starts with a
936letter and may include other information after the letter to define some
937aspect of the data layout. The specifications accepted are as follows: </p>
938<dl>
939 <dt><tt>E</tt></dt>
940 <dd>Specifies that the target lays out data in big-endian form. That is, the
941 bits with the most significance have the lowest address location.</dd>
942 <dt><tt>e</tt></dt>
943 <dd>Specifies that hte target lays out data in little-endian form. That is,
944 the bits with the least significance have the lowest address location.</dd>
945 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
946 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
947 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
948 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
949 too.</dd>
950 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
951 <dd>This specifies the alignment for an integer type of a given bit
952 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
953 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
954 <dd>This specifies the alignment for a vector type of a given bit
955 <i>size</i>.</dd>
956 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
957 <dd>This specifies the alignment for a floating point type of a given bit
958 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
959 (double).</dd>
960 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
961 <dd>This specifies the alignment for an aggregate type of a given bit
962 <i>size</i>.</dd>
963</dl>
964<p>When constructing the data layout for a given target, LLVM starts with a
965default set of specifications which are then (possibly) overriden by the
966specifications in the <tt>datalayout</tt> keyword. The default specifications
967are given in this list:</p>
968<ul>
969 <li><tt>E</tt> - big endian</li>
970 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
971 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
972 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
973 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
974 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
975 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
976 alignment of 64-bits</li>
977 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
978 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
979 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
980 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
981 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
982</ul>
983<p>When llvm is determining the alignment for a given type, it uses the
984following rules:
985<ol>
986 <li>If the type sought is an exact match for one of the specifications, that
987 specification is used.</li>
988 <li>If no match is found, and the type sought is an integer type, then the
989 smallest integer type that is larger than the bitwidth of the sought type is
990 used. If none of the specifications are larger than the bitwidth then the the
991 largest integer type is used. For example, given the default specifications
992 above, the i7 type will use the alignment of i8 (next largest) while both
993 i65 and i256 will use the alignment of i64 (largest specified).</li>
994 <li>If no match is found, and the type sought is a vector type, then the
995 largest vector type that is smaller than the sought vector type will be used
996 as a fall back. This happens because <128 x double> can be implemented in
997 terms of 64 <2 x double>, for example.</li>
998</ol>
999</div>
1000
1001<!-- *********************************************************************** -->
1002<div class="doc_section"> <a name="typesystem">Type System</a> </div>
1003<!-- *********************************************************************** -->
1004
1005<div class="doc_text">
1006
1007<p>The LLVM type system is one of the most important features of the
1008intermediate representation. Being typed enables a number of
1009optimizations to be performed on the IR directly, without having to do
1010extra analyses on the side before the transformation. A strong type
1011system makes it easier to read the generated code and enables novel
1012analyses and transformations that are not feasible to perform on normal
1013three address code representations.</p>
1014
1015</div>
1016
1017<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1018<div class="doc_subsection"> <a name="t_classifications">Type
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1019Classifications</a> </div>
1020<div class="doc_text">
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1021<p>The types fall into a few useful
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1022classifications:</p>
1023
1024<table border="1" cellspacing="0" cellpadding="4">
1025 <tbody>
1026 <tr><th>Classification</th><th>Types</th></tr>
1027 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1028 <td><a href="#t_integer">integer</a></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1029 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1030 </tr>
1031 <tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1032 <td><a href="#t_floating">floating point</a></td>
1033 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1034 </tr>
1035 <tr>
1036 <td><a name="t_firstclass">first class</a></td>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1037 <td><a href="#t_integer">integer</a>,
1038 <a href="#t_floating">floating point</a>,
1039 <a href="#t_pointer">pointer</a>,
1040 <a href="#t_vector">vector</a>
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]1041 <a href="#t_struct">structure</a>,
1042 <a href="#t_array">array</a>,
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1043 </td>
1044 </tr>
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1045 <tr>
1046 <td><a href="#t_primitive">primitive</a></td>
1047 <td><a href="#t_label">label</a>,
1048 <a href="#t_void">void</a>,
1049 <a href="#t_integer">integer</a>,
1050 <a href="#t_floating">floating point</a>.</td>
1051 </tr>
1052 <tr>
1053 <td><a href="#t_derived">derived</a></td>
1054 <td><a href="#t_integer">integer</a>,
1055 <a href="#t_array">array</a>,
1056 <a href="#t_function">function</a>,
1057 <a href="#t_pointer">pointer</a>,
1058 <a href="#t_struct">structure</a>,
1059 <a href="#t_pstruct">packed structure</a>,
1060 <a href="#t_vector">vector</a>,
1061 <a href="#t_opaque">opaque</a>.
1062 </tr>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1063 </tbody>
1064</table>
1065
1066<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1067most important. Values of these types are the only ones which can be
1068produced by instructions, passed as arguments, or used as operands to
1069instructions. This means that all structures and arrays must be
1070manipulated either by pointer or by component.</p>
1071</div>
1072
1073<!-- ======================================================================= -->
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1074<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1075
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1076<div class="doc_text">
1077<p>The primitive types are the fundamental building blocks of the LLVM
1078system.</p>
1079
Chris Lattner86437612008-01-04 04:34:14 +0000[diff] [blame]1080</div>
1081
Chris Lattner488772f2008-01-04 04:32:38 +0000[diff] [blame]1082<!-- _______________________________________________________________________ -->
1083<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1084
1085<div class="doc_text">
1086 <table>
1087 <tbody>
1088 <tr><th>Type</th><th>Description</th></tr>
1089 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1090 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1091 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1092 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1093 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1094 </tbody>
1095 </table>
1096</div>
1097
1098<!-- _______________________________________________________________________ -->
1099<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1100
1101<div class="doc_text">
1102<h5>Overview:</h5>
1103<p>The void type does not represent any value and has no size.</p>
1104
1105<h5>Syntax:</h5>
1106
1107<pre>
1108 void
1109</pre>
1110</div>
1111
1112<!-- _______________________________________________________________________ -->
1113<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1114
1115<div class="doc_text">
1116<h5>Overview:</h5>
1117<p>The label type represents code labels.</p>
1118
1119<h5>Syntax:</h5>
1120
1121<pre>
1122 label
1123</pre>
1124</div>
1125
1126
1127<!-- ======================================================================= -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1128<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1129
1130<div class="doc_text">
1131
1132<p>The real power in LLVM comes from the derived types in the system.
1133This is what allows a programmer to represent arrays, functions,
1134pointers, and other useful types. Note that these derived types may be
1135recursive: For example, it is possible to have a two dimensional array.</p>
1136
1137</div>
1138
1139<!-- _______________________________________________________________________ -->
1140<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1141
1142<div class="doc_text">
1143
1144<h5>Overview:</h5>
1145<p>The integer type is a very simple derived type that simply specifies an
1146arbitrary bit width for the integer type desired. Any bit width from 1 bit to
11472^23-1 (about 8 million) can be specified.</p>
1148
1149<h5>Syntax:</h5>
1150
1151<pre>
1152 iN
1153</pre>
1154
1155<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1156value.</p>
1157
1158<h5>Examples:</h5>
1159<table class="layout">
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1160 <tbody>
1161 <tr>
1162 <td><tt>i1</tt></td>
1163 <td>a single-bit integer.</td>
1164 </tr><tr>
1165 <td><tt>i32</tt></td>
1166 <td>a 32-bit integer.</td>
1167 </tr><tr>
1168 <td><tt>i1942652</tt></td>
1169 <td>a really big integer of over 1 million bits.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1170 </tr>
Chris Lattner251ab812007-12-18 06:18:21 +0000[diff] [blame]1171 </tbody>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1172</table>
1173</div>
1174
1175<!-- _______________________________________________________________________ -->
1176<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1177
1178<div class="doc_text">
1179
1180<h5>Overview:</h5>
1181
1182<p>The array type is a very simple derived type that arranges elements
1183sequentially in memory. The array type requires a size (number of
1184elements) and an underlying data type.</p>
1185
1186<h5>Syntax:</h5>
1187
1188<pre>
1189 [&lt;# elements&gt; x &lt;elementtype&gt;]
1190</pre>
1191
1192<p>The number of elements is a constant integer value; elementtype may
1193be any type with a size.</p>
1194
1195<h5>Examples:</h5>
1196<table class="layout">
1197 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1198 <td class="left"><tt>[40 x i32]</tt></td>
1199 <td class="left">Array of 40 32-bit integer values.</td>
1200 </tr>
1201 <tr class="layout">
1202 <td class="left"><tt>[41 x i32]</tt></td>
1203 <td class="left">Array of 41 32-bit integer values.</td>
1204 </tr>
1205 <tr class="layout">
1206 <td class="left"><tt>[4 x i8]</tt></td>
1207 <td class="left">Array of 4 8-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1208 </tr>
1209</table>
1210<p>Here are some examples of multidimensional arrays:</p>
1211<table class="layout">
1212 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1213 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1214 <td class="left">3x4 array of 32-bit integer values.</td>
1215 </tr>
1216 <tr class="layout">
1217 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1218 <td class="left">12x10 array of single precision floating point values.</td>
1219 </tr>
1220 <tr class="layout">
1221 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1222 <td class="left">2x3x4 array of 16-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1223 </tr>
1224</table>
1225
1226<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1227length array. Normally, accesses past the end of an array are undefined in
1228LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1229As a special case, however, zero length arrays are recognized to be variable
1230length. This allows implementation of 'pascal style arrays' with the LLVM
1231type "{ i32, [0 x float]}", for example.</p>
1232
1233</div>
1234
1235<!-- _______________________________________________________________________ -->
1236<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1237<div class="doc_text">
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1238
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1239<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1240
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1241<p>The function type can be thought of as a function signature. It
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1242consists of a return type and a list of formal parameter types. The
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1243return type of a function type is a scalar type, a void type, or a struct type.
Devang Pateld5404c02008-03-24 20:52:42 +0000[diff] [blame]1244If the return type is a struct type then all struct elements must be of first
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1245class types, and the struct must have at least one element.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1246
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1247<h5>Syntax:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1248
1249<pre>
1250 &lt;returntype list&gt; (&lt;parameter list&gt;)
1251</pre>
1252
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1253<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1254specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1255which indicates that the function takes a variable number of arguments.
1256Variable argument functions can access their arguments with the <a
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1257 href="#int_varargs">variable argument handling intrinsic</a> functions.
1258'<tt>&lt;returntype list&gt;</tt>' is a comma-separated list of
1259<a href="#t_firstclass">first class</a> type specifiers.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1260
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1261<h5>Examples:</h5>
1262<table class="layout">
1263 <tr class="layout">
1264 <td class="left"><tt>i32 (i32)</tt></td>
1265 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1266 </td>
1267 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1268 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1269 </tt></td>
1270 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1271 an <tt>i16</tt> that should be sign extended and a
1272 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1273 <tt>float</tt>.
1274 </td>
1275 </tr><tr class="layout">
1276 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1277 <td class="left">A vararg function that takes at least one
1278 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1279 which returns an integer. This is the signature for <tt>printf</tt> in
1280 LLVM.
1281 </td>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1282 </tr><tr class="layout">
1283 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Devang Patel6dddba22008-03-24 18:10:52 +0000[diff] [blame]1284 <td class="left">A function taking an <tt>i32></tt>, returning two
1285 <tt> i32 </tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
Devang Pateld4ba41d2008-03-24 05:35:41 +0000[diff] [blame]1286 </td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1287 </tr>
1288</table>
1289
1290</div>
1291<!-- _______________________________________________________________________ -->
1292<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1293<div class="doc_text">
1294<h5>Overview:</h5>
1295<p>The structure type is used to represent a collection of data members
1296together in memory. The packing of the field types is defined to match
1297the ABI of the underlying processor. The elements of a structure may
1298be any type that has a size.</p>
1299<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1300and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1301field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1302instruction.</p>
1303<h5>Syntax:</h5>
1304<pre> { &lt;type list&gt; }<br></pre>
1305<h5>Examples:</h5>
1306<table class="layout">
1307 <tr class="layout">
1308 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1309 <td class="left">A triple of three <tt>i32</tt> values</td>
1310 </tr><tr class="layout">
1311 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1312 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1313 second element is a <a href="#t_pointer">pointer</a> to a
1314 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1315 an <tt>i32</tt>.</td>
1316 </tr>
1317</table>
1318</div>
1319
1320<!-- _______________________________________________________________________ -->
1321<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1322</div>
1323<div class="doc_text">
1324<h5>Overview:</h5>
1325<p>The packed structure type is used to represent a collection of data members
1326together in memory. There is no padding between fields. Further, the alignment
1327of a packed structure is 1 byte. The elements of a packed structure may
1328be any type that has a size.</p>
1329<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1330and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1331field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1332instruction.</p>
1333<h5>Syntax:</h5>
1334<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1335<h5>Examples:</h5>
1336<table class="layout">
1337 <tr class="layout">
1338 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1339 <td class="left">A triple of three <tt>i32</tt> values</td>
1340 </tr><tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1341 <td class="left"><tt>&lt; { float, i32 (i32)* } &gt;</tt></td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1342 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1343 second element is a <a href="#t_pointer">pointer</a> to a
1344 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1345 an <tt>i32</tt>.</td>
1346 </tr>
1347</table>
1348</div>
1349
1350<!-- _______________________________________________________________________ -->
1351<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1352<div class="doc_text">
1353<h5>Overview:</h5>
1354<p>As in many languages, the pointer type represents a pointer or
Christopher Lambdd0049d2007-12-11 09:31:00 +0000[diff] [blame]1355reference to another object, which must live in memory. Pointer types may have
1356an optional address space attribute defining the target-specific numbered
1357address space where the pointed-to object resides. The default address space is
1358zero.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1359<h5>Syntax:</h5>
1360<pre> &lt;type&gt; *<br></pre>
1361<h5>Examples:</h5>
1362<table class="layout">
1363 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1364 <td class="left"><tt>[4x i32]*</tt></td>
1365 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1366 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1367 </tr>
1368 <tr class="layout">
1369 <td class="left"><tt>i32 (i32 *) *</tt></td>
1370 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1371 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1372 <tt>i32</tt>.</td>
1373 </tr>
1374 <tr class="layout">
1375 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1376 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1377 that resides in address space #5.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1378 </tr>
1379</table>
1380</div>
1381
1382<!-- _______________________________________________________________________ -->
1383<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1384<div class="doc_text">
1385
1386<h5>Overview:</h5>
1387
1388<p>A vector type is a simple derived type that represents a vector
1389of elements. Vector types are used when multiple primitive data
1390are operated in parallel using a single instruction (SIMD).
1391A vector type requires a size (number of
1392elements) and an underlying primitive data type. Vectors must have a power
1393of two length (1, 2, 4, 8, 16 ...). Vector types are
1394considered <a href="#t_firstclass">first class</a>.</p>
1395
1396<h5>Syntax:</h5>
1397
1398<pre>
1399 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1400</pre>
1401
1402<p>The number of elements is a constant integer value; elementtype may
1403be any integer or floating point type.</p>
1404
1405<h5>Examples:</h5>
1406
1407<table class="layout">
1408 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1409 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1410 <td class="left">Vector of 4 32-bit integer values.</td>
1411 </tr>
1412 <tr class="layout">
1413 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1414 <td class="left">Vector of 8 32-bit floating-point values.</td>
1415 </tr>
1416 <tr class="layout">
1417 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1418 <td class="left">Vector of 2 64-bit integer values.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1419 </tr>
1420</table>
1421</div>
1422
1423<!-- _______________________________________________________________________ -->
1424<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1425<div class="doc_text">
1426
1427<h5>Overview:</h5>
1428
1429<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1430corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1431In LLVM, opaque types can eventually be resolved to any type (not just a
1432structure type).</p>
1433
1434<h5>Syntax:</h5>
1435
1436<pre>
1437 opaque
1438</pre>
1439
1440<h5>Examples:</h5>
1441
1442<table class="layout">
1443 <tr class="layout">
Chris Lattner7311d222007-12-19 05:04:11 +0000[diff] [blame]1444 <td class="left"><tt>opaque</tt></td>
1445 <td class="left">An opaque type.</td>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1446 </tr>
1447</table>
1448</div>
1449
1450
1451<!-- *********************************************************************** -->
1452<div class="doc_section"> <a name="constants">Constants</a> </div>
1453<!-- *********************************************************************** -->
1454
1455<div class="doc_text">
1456
1457<p>LLVM has several different basic types of constants. This section describes
1458them all and their syntax.</p>
1459
1460</div>
1461
1462<!-- ======================================================================= -->
1463<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1464
1465<div class="doc_text">
1466
1467<dl>
1468 <dt><b>Boolean constants</b></dt>
1469
1470 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1471 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1472 </dd>
1473
1474 <dt><b>Integer constants</b></dt>
1475
1476 <dd>Standard integers (such as '4') are constants of the <a
1477 href="#t_integer">integer</a> type. Negative numbers may be used with
1478 integer types.
1479 </dd>
1480
1481 <dt><b>Floating point constants</b></dt>
1482
1483 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1484 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]1485 notation (see below). The assembler requires the exact decimal value of
1486 a floating-point constant. For example, the assembler accepts 1.25 but
1487 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1488 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1489
1490 <dt><b>Null pointer constants</b></dt>
1491
1492 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1493 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1494
1495</dl>
1496
1497<p>The one non-intuitive notation for constants is the optional hexadecimal form
1498of floating point constants. For example, the form '<tt>double
14990x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
15004.5e+15</tt>'. The only time hexadecimal floating point constants are required
1501(and the only time that they are generated by the disassembler) is when a
1502floating point constant must be emitted but it cannot be represented as a
1503decimal floating point number. For example, NaN's, infinities, and other
1504special values are represented in their IEEE hexadecimal format so that
1505assembly and disassembly do not cause any bits to change in the constants.</p>
1506
1507</div>
1508
1509<!-- ======================================================================= -->
1510<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1511</div>
1512
1513<div class="doc_text">
1514<p>Aggregate constants arise from aggregation of simple constants
1515and smaller aggregate constants.</p>
1516
1517<dl>
1518 <dt><b>Structure constants</b></dt>
1519
1520 <dd>Structure constants are represented with notation similar to structure
1521 type definitions (a comma separated list of elements, surrounded by braces
Chris Lattner6c8de962007-12-25 20:34:52 +0000[diff] [blame]1522 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1523 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]1524 must have <a href="#t_struct">structure type</a>, and the number and
1525 types of elements must match those specified by the type.
1526 </dd>
1527
1528 <dt><b>Array constants</b></dt>
1529
1530 <dd>Array constants are represented with notation similar to array type
1531 definitions (a comma separated list of elements, surrounded by square brackets
1532 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1533 constants must have <a href="#t_array">array type</a>, and the number and
1534 types of elements must match those specified by the type.
1535 </dd>
1536
1537 <dt><b>Vector constants</b></dt>
1538
1539 <dd>Vector constants are represented with notation similar to vector type
1540 definitions (a comma separated list of elements, surrounded by
1541 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1542 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1543 href="#t_vector">vector type</a>, and the number and types of elements must
1544 match those specified by the type.
1545 </dd>
1546
1547 <dt><b>Zero initialization</b></dt>
1548
1549 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1550 value to zero of <em>any</em> type, including scalar and aggregate types.
1551 This is often used to avoid having to print large zero initializers (e.g. for
1552 large arrays) and is always exactly equivalent to using explicit zero
1553 initializers.
1554 </dd>
1555</dl>
1556
1557</div>
1558
1559<!-- ======================================================================= -->
1560<div class="doc_subsection">
1561 <a name="globalconstants">Global Variable and Function Addresses</a>
1562</div>
1563
1564<div class="doc_text">
1565
1566<p>The addresses of <a href="#globalvars">global variables</a> and <a
1567href="#functionstructure">functions</a> are always implicitly valid (link-time)
1568constants. These constants are explicitly referenced when the <a
1569href="#identifiers">identifier for the global</a> is used and always have <a
1570href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1571file:</p>
1572
1573<div class="doc_code">
1574<pre>
1575@X = global i32 17
1576@Y = global i32 42
1577@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1578</pre>
1579</div>
1580
1581</div>
1582
1583<!-- ======================================================================= -->
1584<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1585<div class="doc_text">
1586 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1587 no specific value. Undefined values may be of any type and be used anywhere
1588 a constant is permitted.</p>
1589
1590 <p>Undefined values indicate to the compiler that the program is well defined
1591 no matter what value is used, giving the compiler more freedom to optimize.
1592 </p>
1593</div>
1594
1595<!-- ======================================================================= -->
1596<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1597</div>
1598
1599<div class="doc_text">
1600
1601<p>Constant expressions are used to allow expressions involving other constants
1602to be used as constants. Constant expressions may be of any <a
1603href="#t_firstclass">first class</a> type and may involve any LLVM operation
1604that does not have side effects (e.g. load and call are not supported). The
1605following is the syntax for constant expressions:</p>
1606
1607<dl>
1608 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1609 <dd>Truncate a constant to another type. The bit size of CST must be larger
1610 than the bit size of TYPE. Both types must be integers.</dd>
1611
1612 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1613 <dd>Zero extend a constant to another type. The bit size of CST must be
1614 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1615
1616 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1617 <dd>Sign extend a constant to another type. The bit size of CST must be
1618 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1619
1620 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1621 <dd>Truncate a floating point constant to another floating point type. The
1622 size of CST must be larger than the size of TYPE. Both types must be
1623 floating point.</dd>
1624
1625 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1626 <dd>Floating point extend a constant to another type. The size of CST must be
1627 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1628
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1629 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1630 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1631 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1632 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1633 of the same number of elements. If the value won't fit in the integer type,
1634 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1635
1636 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1637 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1638 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1639 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1640 of the same number of elements. If the value won't fit in the integer type,
1641 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1642
1643 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1644 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1645 constant. TYPE must be a scalar or vector floating point type. CST must be of
1646 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1647 of the same number of elements. If the value won't fit in the floating point
1648 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1649
1650 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1651 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1652 constant. TYPE must be a scalar or vector floating point type. CST must be of
1653 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1654 of the same number of elements. If the value won't fit in the floating point
1655 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1656
1657 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1658 <dd>Convert a pointer typed constant to the corresponding integer constant
1659 TYPE must be an integer type. CST must be of pointer type. The CST value is
1660 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1661
1662 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1663 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1664 pointer type. CST must be of integer type. The CST value is zero extended,
1665 truncated, or unchanged to make it fit in a pointer size. This one is
1666 <i>really</i> dangerous!</dd>
1667
1668 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1669 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1670 identical (same number of bits). The conversion is done as if the CST value
1671 was stored to memory and read back as TYPE. In other words, no bits change
1672 with this operator, just the type. This can be used for conversion of
1673 vector types to any other type, as long as they have the same bit width. For
1674 pointers it is only valid to cast to another pointer type.
1675 </dd>
1676
1677 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1678
1679 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1680 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1681 instruction, the index list may have zero or more indexes, which are required
1682 to make sense for the type of "CSTPTR".</dd>
1683
1684 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1685
1686 <dd>Perform the <a href="#i_select">select operation</a> on
1687 constants.</dd>
1688
1689 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1690 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1691
1692 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1693 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1694
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]1695 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
1696 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
1697
1698 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
1699 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
1700
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1701 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1702
1703 <dd>Perform the <a href="#i_extractelement">extractelement
1704 operation</a> on constants.
1705
1706 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1707
1708 <dd>Perform the <a href="#i_insertelement">insertelement
1709 operation</a> on constants.</dd>
1710
1711
1712 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1713
1714 <dd>Perform the <a href="#i_shufflevector">shufflevector
1715 operation</a> on constants.</dd>
1716
1717 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1718
1719 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1720 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1721 binary</a> operations. The constraints on operands are the same as those for
1722 the corresponding instruction (e.g. no bitwise operations on floating point
1723 values are allowed).</dd>
1724</dl>
1725</div>
1726
1727<!-- *********************************************************************** -->
1728<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1729<!-- *********************************************************************** -->
1730
1731<!-- ======================================================================= -->
1732<div class="doc_subsection">
1733<a name="inlineasm">Inline Assembler Expressions</a>
1734</div>
1735
1736<div class="doc_text">
1737
1738<p>
1739LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1740Module-Level Inline Assembly</a>) through the use of a special value. This
1741value represents the inline assembler as a string (containing the instructions
1742to emit), a list of operand constraints (stored as a string), and a flag that
1743indicates whether or not the inline asm expression has side effects. An example
1744inline assembler expression is:
1745</p>
1746
1747<div class="doc_code">
1748<pre>
1749i32 (i32) asm "bswap $0", "=r,r"
1750</pre>
1751</div>
1752
1753<p>
1754Inline assembler expressions may <b>only</b> be used as the callee operand of
1755a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1756</p>
1757
1758<div class="doc_code">
1759<pre>
1760%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1761</pre>
1762</div>
1763
1764<p>
1765Inline asms with side effects not visible in the constraint list must be marked
1766as having side effects. This is done through the use of the
1767'<tt>sideeffect</tt>' keyword, like so:
1768</p>
1769
1770<div class="doc_code">
1771<pre>
1772call void asm sideeffect "eieio", ""()
1773</pre>
1774</div>
1775
1776<p>TODO: The format of the asm and constraints string still need to be
1777documented here. Constraints on what can be done (e.g. duplication, moving, etc
1778need to be documented).
1779</p>
1780
1781</div>
1782
1783<!-- *********************************************************************** -->
1784<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1785<!-- *********************************************************************** -->
1786
1787<div class="doc_text">
1788
1789<p>The LLVM instruction set consists of several different
1790classifications of instructions: <a href="#terminators">terminator
1791instructions</a>, <a href="#binaryops">binary instructions</a>,
1792<a href="#bitwiseops">bitwise binary instructions</a>, <a
1793 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1794instructions</a>.</p>
1795
1796</div>
1797
1798<!-- ======================================================================= -->
1799<div class="doc_subsection"> <a name="terminators">Terminator
1800Instructions</a> </div>
1801
1802<div class="doc_text">
1803
1804<p>As mentioned <a href="#functionstructure">previously</a>, every
1805basic block in a program ends with a "Terminator" instruction, which
1806indicates which block should be executed after the current block is
1807finished. These terminator instructions typically yield a '<tt>void</tt>'
1808value: they produce control flow, not values (the one exception being
1809the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1810<p>There are six different terminator instructions: the '<a
1811 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1812instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1813the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1814 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1815 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1816
1817</div>
1818
1819<!-- _______________________________________________________________________ -->
1820<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1821Instruction</a> </div>
1822<div class="doc_text">
1823<h5>Syntax:</h5>
1824<pre> ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
1825 ret void <i>; Return from void function</i>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1826 ret &lt;type&gt; &lt;value&gt;, &lt;type&gt; &lt;value&gt; <i>; Return two values from a non-void function </i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1827</pre>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1828
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1829<h5>Overview:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1830
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1831<p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1832value) from a function back to the caller.</p>
1833<p>There are two forms of the '<tt>ret</tt>' instruction: one that
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1834returns value(s) and then causes control flow, and one that just causes
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1835control flow to occur.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1836
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1837<h5>Arguments:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1838
1839<p>The '<tt>ret</tt>' instruction may return zero, one or multiple values.
1840The type of each return value must be a '<a href="#t_firstclass">first
1841class</a>' type. Note that a function is not <a href="#wellformed">well
1842formed</a> if there exists a '<tt>ret</tt>' instruction inside of the
1843function that returns values that do not match the return type of the
1844function.</p>
1845
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1846<h5>Semantics:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1847
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1848<p>When the '<tt>ret</tt>' instruction is executed, control flow
1849returns back to the calling function's context. If the caller is a "<a
1850 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1851the instruction after the call. If the caller was an "<a
1852 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1853at the beginning of the "normal" destination block. If the instruction
1854returns a value, that value shall set the call or invoke instruction's
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1855return value. If the instruction returns multiple values then these
Devang Patelec8a5b02008-03-11 05:51:59 +0000[diff] [blame]1856values can only be accessed through a '<a href="#i_getresult"><tt>getresult</tt>
1857</a>' instruction.</p>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1858
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1859<h5>Example:</h5>
Chris Lattner43030e72008-04-23 04:59:35 +0000[diff] [blame]1860
1861<pre>
1862 ret i32 5 <i>; Return an integer value of 5</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1863 ret void <i>; Return from a void function</i>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1864 ret i32 4, i8 2 <i>; Return two values 4 and 2 </i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1865</pre>
1866</div>
1867<!-- _______________________________________________________________________ -->
1868<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1869<div class="doc_text">
1870<h5>Syntax:</h5>
1871<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
1872</pre>
1873<h5>Overview:</h5>
1874<p>The '<tt>br</tt>' instruction is used to cause control flow to
1875transfer to a different basic block in the current function. There are
1876two forms of this instruction, corresponding to a conditional branch
1877and an unconditional branch.</p>
1878<h5>Arguments:</h5>
1879<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1880single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1881unconditional form of the '<tt>br</tt>' instruction takes a single
1882'<tt>label</tt>' value as a target.</p>
1883<h5>Semantics:</h5>
1884<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1885argument is evaluated. If the value is <tt>true</tt>, control flows
1886to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1887control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1888<h5>Example:</h5>
1889<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
1890 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1891</div>
1892<!-- _______________________________________________________________________ -->
1893<div class="doc_subsubsection">
1894 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1895</div>
1896
1897<div class="doc_text">
1898<h5>Syntax:</h5>
1899
1900<pre>
1901 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
1902</pre>
1903
1904<h5>Overview:</h5>
1905
1906<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1907several different places. It is a generalization of the '<tt>br</tt>'
1908instruction, allowing a branch to occur to one of many possible
1909destinations.</p>
1910
1911
1912<h5>Arguments:</h5>
1913
1914<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1915comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1916an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1917table is not allowed to contain duplicate constant entries.</p>
1918
1919<h5>Semantics:</h5>
1920
1921<p>The <tt>switch</tt> instruction specifies a table of values and
1922destinations. When the '<tt>switch</tt>' instruction is executed, this
1923table is searched for the given value. If the value is found, control flow is
1924transfered to the corresponding destination; otherwise, control flow is
1925transfered to the default destination.</p>
1926
1927<h5>Implementation:</h5>
1928
1929<p>Depending on properties of the target machine and the particular
1930<tt>switch</tt> instruction, this instruction may be code generated in different
1931ways. For example, it could be generated as a series of chained conditional
1932branches or with a lookup table.</p>
1933
1934<h5>Example:</h5>
1935
1936<pre>
1937 <i>; Emulate a conditional br instruction</i>
1938 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1939 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1940
1941 <i>; Emulate an unconditional br instruction</i>
1942 switch i32 0, label %dest [ ]
1943
1944 <i>; Implement a jump table:</i>
1945 switch i32 %val, label %otherwise [ i32 0, label %onzero
1946 i32 1, label %onone
1947 i32 2, label %ontwo ]
1948</pre>
1949</div>
1950
1951<!-- _______________________________________________________________________ -->
1952<div class="doc_subsubsection">
1953 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1954</div>
1955
1956<div class="doc_text">
1957
1958<h5>Syntax:</h5>
1959
1960<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]1961 &lt;result&gt; = invoke [<a href="#callingconv">cconv</a>] &lt;ptr to function ty&gt; &lt;function ptr val&gt;(&lt;function args&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1962 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
1963</pre>
1964
1965<h5>Overview:</h5>
1966
1967<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1968function, with the possibility of control flow transfer to either the
1969'<tt>normal</tt>' label or the
1970'<tt>exception</tt>' label. If the callee function returns with the
1971"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1972"normal" label. If the callee (or any indirect callees) returns with the "<a
1973href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]1974continued at the dynamically nearest "exception" label. If the callee function
Devang Patelec8a5b02008-03-11 05:51:59 +0000[diff] [blame]1975returns multiple values then individual return values are only accessible through
1976a '<tt><a href="#i_getresult">getresult</a></tt>' instruction.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1977
1978<h5>Arguments:</h5>
1979
1980<p>This instruction requires several arguments:</p>
1981
1982<ol>
1983 <li>
1984 The optional "cconv" marker indicates which <a href="#callingconv">calling
1985 convention</a> the call should use. If none is specified, the call defaults
1986 to using C calling conventions.
1987 </li>
1988 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1989 function value being invoked. In most cases, this is a direct function
1990 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1991 an arbitrary pointer to function value.
1992 </li>
1993
1994 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1995 function to be invoked. </li>
1996
1997 <li>'<tt>function args</tt>': argument list whose types match the function
1998 signature argument types. If the function signature indicates the function
1999 accepts a variable number of arguments, the extra arguments can be
2000 specified. </li>
2001
2002 <li>'<tt>normal label</tt>': the label reached when the called function
2003 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2004
2005 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2006 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2007
2008</ol>
2009
2010<h5>Semantics:</h5>
2011
2012<p>This instruction is designed to operate as a standard '<tt><a
2013href="#i_call">call</a></tt>' instruction in most regards. The primary
2014difference is that it establishes an association with a label, which is used by
2015the runtime library to unwind the stack.</p>
2016
2017<p>This instruction is used in languages with destructors to ensure that proper
2018cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2019exception. Additionally, this is important for implementation of
2020'<tt>catch</tt>' clauses in high-level languages that support them.</p>
2021
2022<h5>Example:</h5>
2023<pre>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2024 %retval = invoke i32 @Test(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2025 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewyckya1c11a12008-03-16 07:18:12 +0000[diff] [blame]2026 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2027 unwind label %TestCleanup <i>; {i32}:retval set</i>
2028</pre>
2029</div>
2030
2031
2032<!-- _______________________________________________________________________ -->
2033
2034<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2035Instruction</a> </div>
2036
2037<div class="doc_text">
2038
2039<h5>Syntax:</h5>
2040<pre>
2041 unwind
2042</pre>
2043
2044<h5>Overview:</h5>
2045
2046<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2047at the first callee in the dynamic call stack which used an <a
2048href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2049primarily used to implement exception handling.</p>
2050
2051<h5>Semantics:</h5>
2052
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2053<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2054immediately halt. The dynamic call stack is then searched for the first <a
2055href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2056execution continues at the "exceptional" destination block specified by the
2057<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2058dynamic call chain, undefined behavior results.</p>
2059</div>
2060
2061<!-- _______________________________________________________________________ -->
2062
2063<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2064Instruction</a> </div>
2065
2066<div class="doc_text">
2067
2068<h5>Syntax:</h5>
2069<pre>
2070 unreachable
2071</pre>
2072
2073<h5>Overview:</h5>
2074
2075<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2076instruction is used to inform the optimizer that a particular portion of the
2077code is not reachable. This can be used to indicate that the code after a
2078no-return function cannot be reached, and other facts.</p>
2079
2080<h5>Semantics:</h5>
2081
2082<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2083</div>
2084
2085
2086
2087<!-- ======================================================================= -->
2088<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2089<div class="doc_text">
2090<p>Binary operators are used to do most of the computation in a
Chris Lattnerab596d92008-04-01 18:47:32 +0000[diff] [blame]2091program. They require two operands of the same type, execute an operation on them, and
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2092produce a single value. The operands might represent
2093multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
Chris Lattnerab596d92008-04-01 18:47:32 +0000[diff] [blame]2094The result value has the same type as its operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2095<p>There are several different binary operators:</p>
2096</div>
2097<!-- _______________________________________________________________________ -->
2098<div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
2099Instruction</a> </div>
2100<div class="doc_text">
2101<h5>Syntax:</h5>
2102<pre> &lt;result&gt; = add &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2103</pre>
2104<h5>Overview:</h5>
2105<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2106<h5>Arguments:</h5>
2107<p>The two arguments to the '<tt>add</tt>' instruction must be either <a
2108 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
2109 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2110Both arguments must have identical types.</p>
2111<h5>Semantics:</h5>
2112<p>The value produced is the integer or floating point sum of the two
2113operands.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2114<p>If an integer sum has unsigned overflow, the result returned is the
2115mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2116the result.</p>
2117<p>Because LLVM integers use a two's complement representation, this
2118instruction is appropriate for both signed and unsigned integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2119<h5>Example:</h5>
2120<pre> &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2121</pre>
2122</div>
2123<!-- _______________________________________________________________________ -->
2124<div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
2125Instruction</a> </div>
2126<div class="doc_text">
2127<h5>Syntax:</h5>
2128<pre> &lt;result&gt; = sub &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2129</pre>
2130<h5>Overview:</h5>
2131<p>The '<tt>sub</tt>' instruction returns the difference of its two
2132operands.</p>
2133<p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
2134instruction present in most other intermediate representations.</p>
2135<h5>Arguments:</h5>
2136<p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
2137 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2138values.
2139This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2140Both arguments must have identical types.</p>
2141<h5>Semantics:</h5>
2142<p>The value produced is the integer or floating point difference of
2143the two operands.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2144<p>If an integer difference has unsigned overflow, the result returned is the
2145mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2146the result.</p>
2147<p>Because LLVM integers use a two's complement representation, this
2148instruction is appropriate for both signed and unsigned integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2149<h5>Example:</h5>
2150<pre>
2151 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2152 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2153</pre>
2154</div>
2155<!-- _______________________________________________________________________ -->
2156<div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
2157Instruction</a> </div>
2158<div class="doc_text">
2159<h5>Syntax:</h5>
2160<pre> &lt;result&gt; = mul &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2161</pre>
2162<h5>Overview:</h5>
2163<p>The '<tt>mul</tt>' instruction returns the product of its two
2164operands.</p>
2165<h5>Arguments:</h5>
2166<p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
2167 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2168values.
2169This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2170Both arguments must have identical types.</p>
2171<h5>Semantics:</h5>
2172<p>The value produced is the integer or floating point product of the
2173two operands.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2174<p>If the result of an integer multiplication has unsigned overflow,
2175the result returned is the mathematical result modulo
21762<sup>n</sup>, where n is the bit width of the result.</p>
2177<p>Because LLVM integers use a two's complement representation, and the
2178result is the same width as the operands, this instruction returns the
2179correct result for both signed and unsigned integers. If a full product
2180(e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2181should be sign-extended or zero-extended as appropriate to the
2182width of the full product.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2183<h5>Example:</h5>
2184<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2185</pre>
2186</div>
2187<!-- _______________________________________________________________________ -->
2188<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2189</a></div>
2190<div class="doc_text">
2191<h5>Syntax:</h5>
2192<pre> &lt;result&gt; = udiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2193</pre>
2194<h5>Overview:</h5>
2195<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2196operands.</p>
2197<h5>Arguments:</h5>
2198<p>The two arguments to the '<tt>udiv</tt>' instruction must be
2199<a href="#t_integer">integer</a> values. Both arguments must have identical
2200types. This instruction can also take <a href="#t_vector">vector</a> versions
2201of the values in which case the elements must be integers.</p>
2202<h5>Semantics:</h5>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2203<p>The value produced is the unsigned integer quotient of the two operands.</p>
2204<p>Note that unsigned integer division and signed integer division are distinct
2205operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2206<p>Division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2207<h5>Example:</h5>
2208<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2209</pre>
2210</div>
2211<!-- _______________________________________________________________________ -->
2212<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2213</a> </div>
2214<div class="doc_text">
2215<h5>Syntax:</h5>
2216<pre> &lt;result&gt; = sdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2217</pre>
2218<h5>Overview:</h5>
2219<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2220operands.</p>
2221<h5>Arguments:</h5>
2222<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2223<a href="#t_integer">integer</a> values. Both arguments must have identical
2224types. This instruction can also take <a href="#t_vector">vector</a> versions
2225of the values in which case the elements must be integers.</p>
2226<h5>Semantics:</h5>
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2227<p>The value produced is the signed integer quotient of the two operands rounded towards zero.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2228<p>Note that signed integer division and unsigned integer division are distinct
2229operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2230<p>Division by zero leads to undefined behavior. Overflow also leads to
2231undefined behavior; this is a rare case, but can occur, for example,
2232by doing a 32-bit division of -2147483648 by -1.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2233<h5>Example:</h5>
2234<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2235</pre>
2236</div>
2237<!-- _______________________________________________________________________ -->
2238<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2239Instruction</a> </div>
2240<div class="doc_text">
2241<h5>Syntax:</h5>
2242<pre> &lt;result&gt; = fdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2243</pre>
2244<h5>Overview:</h5>
2245<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2246operands.</p>
2247<h5>Arguments:</h5>
2248<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2249<a href="#t_floating">floating point</a> values. Both arguments must have
2250identical types. This instruction can also take <a href="#t_vector">vector</a>
2251versions of floating point values.</p>
2252<h5>Semantics:</h5>
2253<p>The value produced is the floating point quotient of the two operands.</p>
2254<h5>Example:</h5>
2255<pre> &lt;result&gt; = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2256</pre>
2257</div>
2258<!-- _______________________________________________________________________ -->
2259<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2260</div>
2261<div class="doc_text">
2262<h5>Syntax:</h5>
2263<pre> &lt;result&gt; = urem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2264</pre>
2265<h5>Overview:</h5>
2266<p>The '<tt>urem</tt>' instruction returns the remainder from the
2267unsigned division of its two arguments.</p>
2268<h5>Arguments:</h5>
2269<p>The two arguments to the '<tt>urem</tt>' instruction must be
2270<a href="#t_integer">integer</a> values. Both arguments must have identical
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2271types. This instruction can also take <a href="#t_vector">vector</a> versions
2272of the values in which case the elements must be integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2273<h5>Semantics:</h5>
2274<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2275This instruction always performs an unsigned division to get the remainder.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2276<p>Note that unsigned integer remainder and signed integer remainder are
2277distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2278<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2279<h5>Example:</h5>
2280<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2281</pre>
2282
2283</div>
2284<!-- _______________________________________________________________________ -->
2285<div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2286Instruction</a> </div>
2287<div class="doc_text">
2288<h5>Syntax:</h5>
2289<pre> &lt;result&gt; = srem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2290</pre>
2291<h5>Overview:</h5>
2292<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2293signed division of its two operands. This instruction can also take
2294<a href="#t_vector">vector</a> versions of the values in which case
2295the elements must be integers.</p>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]2296
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2297<h5>Arguments:</h5>
2298<p>The two arguments to the '<tt>srem</tt>' instruction must be
2299<a href="#t_integer">integer</a> values. Both arguments must have identical
2300types.</p>
2301<h5>Semantics:</h5>
2302<p>This instruction returns the <i>remainder</i> of a division (where the result
2303has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2304operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2305a value. For more information about the difference, see <a
2306 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2307Math Forum</a>. For a table of how this is implemented in various languages,
2308please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2309Wikipedia: modulo operation</a>.</p>
Chris Lattner9aba1e22008-01-28 00:36:27 +0000[diff] [blame]2310<p>Note that signed integer remainder and unsigned integer remainder are
2311distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2312<p>Taking the remainder of a division by zero leads to undefined behavior.
2313Overflow also leads to undefined behavior; this is a rare case, but can occur,
2314for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2315(The remainder doesn't actually overflow, but this rule lets srem be
2316implemented using instructions that return both the result of the division
2317and the remainder.)</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2318<h5>Example:</h5>
2319<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2320</pre>
2321
2322</div>
2323<!-- _______________________________________________________________________ -->
2324<div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2325Instruction</a> </div>
2326<div class="doc_text">
2327<h5>Syntax:</h5>
2328<pre> &lt;result&gt; = frem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2329</pre>
2330<h5>Overview:</h5>
2331<p>The '<tt>frem</tt>' instruction returns the remainder from the
2332division of its two operands.</p>
2333<h5>Arguments:</h5>
2334<p>The two arguments to the '<tt>frem</tt>' instruction must be
2335<a href="#t_floating">floating point</a> values. Both arguments must have
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2336identical types. This instruction can also take <a href="#t_vector">vector</a>
2337versions of floating point values.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2338<h5>Semantics:</h5>
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2339<p>This instruction returns the <i>remainder</i> of a division.
2340The remainder has the same sign as the dividend.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2341<h5>Example:</h5>
2342<pre> &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2343</pre>
2344</div>
2345
2346<!-- ======================================================================= -->
2347<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2348Operations</a> </div>
2349<div class="doc_text">
2350<p>Bitwise binary operators are used to do various forms of
2351bit-twiddling in a program. They are generally very efficient
2352instructions and can commonly be strength reduced from other
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2353instructions. They require two operands of the same type, execute an operation on them,
2354and produce a single value. The resulting value is the same type as its operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2355</div>
2356
2357<!-- _______________________________________________________________________ -->
2358<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2359Instruction</a> </div>
2360<div class="doc_text">
2361<h5>Syntax:</h5>
2362<pre> &lt;result&gt; = shl &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2363</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2364
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2365<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2366
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2367<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2368the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2369
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2370<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2371
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2372<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2373 href="#t_integer">integer</a> type. '<tt>var2</tt>' is treated as an
2374unsigned value.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2375
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2376<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2377
Chris Lattnerdc15b1d2008-04-01 18:45:27 +0000[diff] [blame]2378<p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup> mod 2<sup>n</sup>,
2379where n is the width of the result. If <tt>var2</tt> is (statically or dynamically) negative or
2380equal to or larger than the number of bits in <tt>var1</tt>, the result is undefined.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2381
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2382<h5>Example:</h5><pre>
2383 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2384 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2385 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2386 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2387</pre>
2388</div>
2389<!-- _______________________________________________________________________ -->
2390<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2391Instruction</a> </div>
2392<div class="doc_text">
2393<h5>Syntax:</h5>
2394<pre> &lt;result&gt; = lshr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2395</pre>
2396
2397<h5>Overview:</h5>
2398<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2399operand shifted to the right a specified number of bits with zero fill.</p>
2400
2401<h5>Arguments:</h5>
2402<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2403<a href="#t_integer">integer</a> type. '<tt>var2</tt>' is treated as an
2404unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2405
2406<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2407
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2408<p>This instruction always performs a logical shift right operation. The most
2409significant bits of the result will be filled with zero bits after the
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2410shift. If <tt>var2</tt> is (statically or dynamically) equal to or larger than
2411the number of bits in <tt>var1</tt>, the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2412
2413<h5>Example:</h5>
2414<pre>
2415 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2416 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2417 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2418 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2419 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2420</pre>
2421</div>
2422
2423<!-- _______________________________________________________________________ -->
2424<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2425Instruction</a> </div>
2426<div class="doc_text">
2427
2428<h5>Syntax:</h5>
2429<pre> &lt;result&gt; = ashr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2430</pre>
2431
2432<h5>Overview:</h5>
2433<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2434operand shifted to the right a specified number of bits with sign extension.</p>
2435
2436<h5>Arguments:</h5>
2437<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2438<a href="#t_integer">integer</a> type. '<tt>var2</tt>' is treated as an
2439unsigned value.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2440
2441<h5>Semantics:</h5>
2442<p>This instruction always performs an arithmetic shift right operation,
2443The most significant bits of the result will be filled with the sign bit
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2444of <tt>var1</tt>. If <tt>var2</tt> is (statically or dynamically) equal to or
2445larger than the number of bits in <tt>var1</tt>, the result is undefined.
2446</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2447
2448<h5>Example:</h5>
2449<pre>
2450 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2451 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2452 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2453 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2454 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2455</pre>
2456</div>
2457
2458<!-- _______________________________________________________________________ -->
2459<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2460Instruction</a> </div>
2461<div class="doc_text">
2462<h5>Syntax:</h5>
2463<pre> &lt;result&gt; = and &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2464</pre>
2465<h5>Overview:</h5>
2466<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2467its two operands.</p>
2468<h5>Arguments:</h5>
2469<p>The two arguments to the '<tt>and</tt>' instruction must be <a
2470 href="#t_integer">integer</a> values. Both arguments must have
2471identical types.</p>
2472<h5>Semantics:</h5>
2473<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2474<p> </p>
2475<div style="align: center">
2476<table border="1" cellspacing="0" cellpadding="4">
2477 <tbody>
2478 <tr>
2479 <td>In0</td>
2480 <td>In1</td>
2481 <td>Out</td>
2482 </tr>
2483 <tr>
2484 <td>0</td>
2485 <td>0</td>
2486 <td>0</td>
2487 </tr>
2488 <tr>
2489 <td>0</td>
2490 <td>1</td>
2491 <td>0</td>
2492 </tr>
2493 <tr>
2494 <td>1</td>
2495 <td>0</td>
2496 <td>0</td>
2497 </tr>
2498 <tr>
2499 <td>1</td>
2500 <td>1</td>
2501 <td>1</td>
2502 </tr>
2503 </tbody>
2504</table>
2505</div>
2506<h5>Example:</h5>
2507<pre> &lt;result&gt; = and i32 4, %var <i>; yields {i32}:result = 4 &amp; %var</i>
2508 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2509 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2510</pre>
2511</div>
2512<!-- _______________________________________________________________________ -->
2513<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2514<div class="doc_text">
2515<h5>Syntax:</h5>
2516<pre> &lt;result&gt; = or &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2517</pre>
2518<h5>Overview:</h5>
2519<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2520or of its two operands.</p>
2521<h5>Arguments:</h5>
2522<p>The two arguments to the '<tt>or</tt>' instruction must be <a
2523 href="#t_integer">integer</a> values. Both arguments must have
2524identical types.</p>
2525<h5>Semantics:</h5>
2526<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2527<p> </p>
2528<div style="align: center">
2529<table border="1" cellspacing="0" cellpadding="4">
2530 <tbody>
2531 <tr>
2532 <td>In0</td>
2533 <td>In1</td>
2534 <td>Out</td>
2535 </tr>
2536 <tr>
2537 <td>0</td>
2538 <td>0</td>
2539 <td>0</td>
2540 </tr>
2541 <tr>
2542 <td>0</td>
2543 <td>1</td>
2544 <td>1</td>
2545 </tr>
2546 <tr>
2547 <td>1</td>
2548 <td>0</td>
2549 <td>1</td>
2550 </tr>
2551 <tr>
2552 <td>1</td>
2553 <td>1</td>
2554 <td>1</td>
2555 </tr>
2556 </tbody>
2557</table>
2558</div>
2559<h5>Example:</h5>
2560<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2561 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2562 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2563</pre>
2564</div>
2565<!-- _______________________________________________________________________ -->
2566<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2567Instruction</a> </div>
2568<div class="doc_text">
2569<h5>Syntax:</h5>
2570<pre> &lt;result&gt; = xor &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2571</pre>
2572<h5>Overview:</h5>
2573<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2574or of its two operands. The <tt>xor</tt> is used to implement the
2575"one's complement" operation, which is the "~" operator in C.</p>
2576<h5>Arguments:</h5>
2577<p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2578 href="#t_integer">integer</a> values. Both arguments must have
2579identical types.</p>
2580<h5>Semantics:</h5>
2581<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2582<p> </p>
2583<div style="align: center">
2584<table border="1" cellspacing="0" cellpadding="4">
2585 <tbody>
2586 <tr>
2587 <td>In0</td>
2588 <td>In1</td>
2589 <td>Out</td>
2590 </tr>
2591 <tr>
2592 <td>0</td>
2593 <td>0</td>
2594 <td>0</td>
2595 </tr>
2596 <tr>
2597 <td>0</td>
2598 <td>1</td>
2599 <td>1</td>
2600 </tr>
2601 <tr>
2602 <td>1</td>
2603 <td>0</td>
2604 <td>1</td>
2605 </tr>
2606 <tr>
2607 <td>1</td>
2608 <td>1</td>
2609 <td>0</td>
2610 </tr>
2611 </tbody>
2612</table>
2613</div>
2614<p> </p>
2615<h5>Example:</h5>
2616<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2617 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2618 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2619 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2620</pre>
2621</div>
2622
2623<!-- ======================================================================= -->
2624<div class="doc_subsection">
2625 <a name="vectorops">Vector Operations</a>
2626</div>
2627
2628<div class="doc_text">
2629
2630<p>LLVM supports several instructions to represent vector operations in a
2631target-independent manner. These instructions cover the element-access and
2632vector-specific operations needed to process vectors effectively. While LLVM
2633does directly support these vector operations, many sophisticated algorithms
2634will want to use target-specific intrinsics to take full advantage of a specific
2635target.</p>
2636
2637</div>
2638
2639<!-- _______________________________________________________________________ -->
2640<div class="doc_subsubsection">
2641 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2642</div>
2643
2644<div class="doc_text">
2645
2646<h5>Syntax:</h5>
2647
2648<pre>
2649 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
2650</pre>
2651
2652<h5>Overview:</h5>
2653
2654<p>
2655The '<tt>extractelement</tt>' instruction extracts a single scalar
2656element from a vector at a specified index.
2657</p>
2658
2659
2660<h5>Arguments:</h5>
2661
2662<p>
2663The first operand of an '<tt>extractelement</tt>' instruction is a
2664value of <a href="#t_vector">vector</a> type. The second operand is
2665an index indicating the position from which to extract the element.
2666The index may be a variable.</p>
2667
2668<h5>Semantics:</h5>
2669
2670<p>
2671The result is a scalar of the same type as the element type of
2672<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2673<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2674results are undefined.
2675</p>
2676
2677<h5>Example:</h5>
2678
2679<pre>
2680 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
2681</pre>
2682</div>
2683
2684
2685<!-- _______________________________________________________________________ -->
2686<div class="doc_subsubsection">
2687 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2688</div>
2689
2690<div class="doc_text">
2691
2692<h5>Syntax:</h5>
2693
2694<pre>
Dan Gohmanbcc3c502008-05-12 23:38:42 +0000[diff] [blame]2695 &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>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2696</pre>
2697
2698<h5>Overview:</h5>
2699
2700<p>
2701The '<tt>insertelement</tt>' instruction inserts a scalar
2702element into a vector at a specified index.
2703</p>
2704
2705
2706<h5>Arguments:</h5>
2707
2708<p>
2709The first operand of an '<tt>insertelement</tt>' instruction is a
2710value of <a href="#t_vector">vector</a> type. The second operand is a
2711scalar value whose type must equal the element type of the first
2712operand. The third operand is an index indicating the position at
2713which to insert the value. The index may be a variable.</p>
2714
2715<h5>Semantics:</h5>
2716
2717<p>
2718The result is a vector of the same type as <tt>val</tt>. Its
2719element values are those of <tt>val</tt> except at position
2720<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2721exceeds the length of <tt>val</tt>, the results are undefined.
2722</p>
2723
2724<h5>Example:</h5>
2725
2726<pre>
2727 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
2728</pre>
2729</div>
2730
2731<!-- _______________________________________________________________________ -->
2732<div class="doc_subsubsection">
2733 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2734</div>
2735
2736<div class="doc_text">
2737
2738<h5>Syntax:</h5>
2739
2740<pre>
2741 &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>
2742</pre>
2743
2744<h5>Overview:</h5>
2745
2746<p>
2747The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2748from two input vectors, returning a vector of the same type.
2749</p>
2750
2751<h5>Arguments:</h5>
2752
2753<p>
2754The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2755with types that match each other and types that match the result of the
2756instruction. The third argument is a shuffle mask, which has the same number
2757of elements as the other vector type, but whose element type is always 'i32'.
2758</p>
2759
2760<p>
2761The shuffle mask operand is required to be a constant vector with either
2762constant integer or undef values.
2763</p>
2764
2765<h5>Semantics:</h5>
2766
2767<p>
2768The elements of the two input vectors are numbered from left to right across
2769both of the vectors. The shuffle mask operand specifies, for each element of
2770the result vector, which element of the two input registers the result element
2771gets. The element selector may be undef (meaning "don't care") and the second
2772operand may be undef if performing a shuffle from only one vector.
2773</p>
2774
2775<h5>Example:</h5>
2776
2777<pre>
2778 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
2779 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
2780 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
2781 &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.
2782</pre>
2783</div>
2784
2785
2786<!-- ======================================================================= -->
2787<div class="doc_subsection">
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2788 <a name="aggregateops">Aggregate Operations</a>
2789</div>
2790
2791<div class="doc_text">
2792
2793<p>LLVM supports several instructions for working with aggregate values.
2794</p>
2795
2796</div>
2797
2798<!-- _______________________________________________________________________ -->
2799<div class="doc_subsubsection">
2800 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
2801</div>
2802
2803<div class="doc_text">
2804
2805<h5>Syntax:</h5>
2806
2807<pre>
2808 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
2809</pre>
2810
2811<h5>Overview:</h5>
2812
2813<p>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2814The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
2815or array element from an aggregate value.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2816</p>
2817
2818
2819<h5>Arguments:</h5>
2820
2821<p>
2822The first operand of an '<tt>extractvalue</tt>' instruction is a
2823value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2824type. The operands are constant indices to specify which value to extract
2825in the same manner as indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2826'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
2827</p>
2828
2829<h5>Semantics:</h5>
2830
2831<p>
2832The result is the value at the position in the aggregate specified by
2833the index operands.
2834</p>
2835
2836<h5>Example:</h5>
2837
2838<pre>
2839 %result = extractvalue {i32, float} %agg, i32 0 <i>; yields i32</i>
2840</pre>
2841</div>
2842
2843
2844<!-- _______________________________________________________________________ -->
2845<div class="doc_subsubsection">
2846 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
2847</div>
2848
2849<div class="doc_text">
2850
2851<h5>Syntax:</h5>
2852
2853<pre>
2854 &lt;result&gt; = insertvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;ty&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;n x &lt;ty&gt;&gt;</i>
2855</pre>
2856
2857<h5>Overview:</h5>
2858
2859<p>
2860The '<tt>insertvalue</tt>' instruction inserts a value
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2861into a struct field or array element in an aggregate.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2862</p>
2863
2864
2865<h5>Arguments:</h5>
2866
2867<p>
2868The first operand of an '<tt>insertvalue</tt>' instruction is a
2869value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
2870The second operand is a first-class value to insert.
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2871type of the first operand. The following operands are constant indices
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2872indicating the position at which to insert the value in the same manner as
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2873indices in a
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2874'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
2875The value to insert must have the same type as the value identified
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2876by the indices.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2877
2878<h5>Semantics:</h5>
2879
2880<p>
2881The result is an aggregate of the same type as <tt>val</tt>. Its
2882value is that of <tt>val</tt> except that the value at the position
Dan Gohman6c6dea02008-05-13 18:16:06 +0000[diff] [blame]2883specified by the indices is that of <tt>elt</tt>.
Dan Gohman74d6faf2008-05-12 23:51:09 +0000[diff] [blame]2884</p>
2885
2886<h5>Example:</h5>
2887
2888<pre>
2889 %result = insertvalue {i32, float} %agg, i32 1, i32 0 <i>; yields {i32, float}</i>
2890</pre>
2891</div>
2892
2893
2894<!-- ======================================================================= -->
2895<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2896 <a name="memoryops">Memory Access and Addressing Operations</a>
2897</div>
2898
2899<div class="doc_text">
2900
2901<p>A key design point of an SSA-based representation is how it
2902represents memory. In LLVM, no memory locations are in SSA form, which
2903makes things very simple. This section describes how to read, write,
2904allocate, and free memory in LLVM.</p>
2905
2906</div>
2907
2908<!-- _______________________________________________________________________ -->
2909<div class="doc_subsubsection">
2910 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2911</div>
2912
2913<div class="doc_text">
2914
2915<h5>Syntax:</h5>
2916
2917<pre>
2918 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
2919</pre>
2920
2921<h5>Overview:</h5>
2922
2923<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]2924heap and returns a pointer to it. The object is always allocated in the generic
2925address space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2926
2927<h5>Arguments:</h5>
2928
2929<p>The '<tt>malloc</tt>' instruction allocates
2930<tt>sizeof(&lt;type&gt;)*NumElements</tt>
2931bytes of memory from the operating system and returns a pointer of the
2932appropriate type to the program. If "NumElements" is specified, it is the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]2933number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]2934If a constant alignment is specified, the value result of the allocation is guaranteed to
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]2935be aligned to at least that boundary. If not specified, or if zero, the target can
2936choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2937
2938<p>'<tt>type</tt>' must be a sized type.</p>
2939
2940<h5>Semantics:</h5>
2941
2942<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]2943a pointer is returned. The result of a zero byte allocattion is undefined. The
2944result is null if there is insufficient memory available.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2945
2946<h5>Example:</h5>
2947
2948<pre>
2949 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2950
2951 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2952 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2953 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2954 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2955 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2956</pre>
2957</div>
2958
2959<!-- _______________________________________________________________________ -->
2960<div class="doc_subsubsection">
2961 <a name="i_free">'<tt>free</tt>' Instruction</a>
2962</div>
2963
2964<div class="doc_text">
2965
2966<h5>Syntax:</h5>
2967
2968<pre>
2969 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
2970</pre>
2971
2972<h5>Overview:</h5>
2973
2974<p>The '<tt>free</tt>' instruction returns memory back to the unused
2975memory heap to be reallocated in the future.</p>
2976
2977<h5>Arguments:</h5>
2978
2979<p>'<tt>value</tt>' shall be a pointer value that points to a value
2980that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2981instruction.</p>
2982
2983<h5>Semantics:</h5>
2984
2985<p>Access to the memory pointed to by the pointer is no longer defined
Chris Lattner329d6532008-04-19 22:41:32 +0000[diff] [blame]2986after this instruction executes. If the pointer is null, the operation
2987is a noop.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2988
2989<h5>Example:</h5>
2990
2991<pre>
2992 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2993 free [4 x i8]* %array
2994</pre>
2995</div>
2996
2997<!-- _______________________________________________________________________ -->
2998<div class="doc_subsubsection">
2999 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3000</div>
3001
3002<div class="doc_text">
3003
3004<h5>Syntax:</h5>
3005
3006<pre>
3007 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
3008</pre>
3009
3010<h5>Overview:</h5>
3011
3012<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3013currently executing function, to be automatically released when this function
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]3014returns to its caller. The object is always allocated in the generic address
3015space (address space zero).</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3016
3017<h5>Arguments:</h5>
3018
3019<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
3020bytes of memory on the runtime stack, returning a pointer of the
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3021appropriate type to the program. If "NumElements" is specified, it is the
3022number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3023If a constant alignment is specified, the value result of the allocation is guaranteed
Gabor Greif5082cf42008-02-09 22:24:34 +0000[diff] [blame]3024to be aligned to at least that boundary. If not specified, or if zero, the target
3025can choose to align the allocation on any convenient boundary.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3026
3027<p>'<tt>type</tt>' may be any sized type.</p>
3028
3029<h5>Semantics:</h5>
3030
Chris Lattner8b094fc2008-04-19 21:01:16 +0000[diff] [blame]3031<p>Memory is allocated; a pointer is returned. The operation is undefiend if
3032there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3033memory is automatically released when the function returns. The '<tt>alloca</tt>'
3034instruction is commonly used to represent automatic variables that must
3035have an address available. When the function returns (either with the <tt><a
3036 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3037instructions), the memory is reclaimed. Allocating zero bytes
3038is legal, but the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3039
3040<h5>Example:</h5>
3041
3042<pre>
3043 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3044 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3045 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3046 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3047</pre>
3048</div>
3049
3050<!-- _______________________________________________________________________ -->
3051<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3052Instruction</a> </div>
3053<div class="doc_text">
3054<h5>Syntax:</h5>
3055<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>
3056<h5>Overview:</h5>
3057<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3058<h5>Arguments:</h5>
3059<p>The argument to the '<tt>load</tt>' instruction specifies the memory
3060address from which to load. The pointer must point to a <a
3061 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3062marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3063the number or order of execution of this <tt>load</tt> with other
3064volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3065instructions. </p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3066<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3067The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3068(that is, the alignment of the memory address). A value of 0 or an
3069omitted "align" argument means that the operation has the preferential
3070alignment for the target. It is the responsibility of the code emitter
3071to ensure that the alignment information is correct. Overestimating
3072the alignment results in an undefined behavior. Underestimating the
3073alignment may produce less efficient code. An alignment of 1 is always
3074safe.
3075</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3076<h5>Semantics:</h5>
3077<p>The location of memory pointed to is loaded.</p>
3078<h5>Examples:</h5>
3079<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3080 <a
3081 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3082 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3083</pre>
3084</div>
3085<!-- _______________________________________________________________________ -->
3086<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3087Instruction</a> </div>
3088<div class="doc_text">
3089<h5>Syntax:</h5>
3090<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3091 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3092</pre>
3093<h5>Overview:</h5>
3094<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3095<h5>Arguments:</h5>
3096<p>There are two arguments to the '<tt>store</tt>' instruction: a value
3097to store and an address at which to store it. The type of the '<tt>&lt;pointer&gt;</tt>'
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3098operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3099of the '<tt>&lt;value&gt;</tt>'
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3100operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3101optimizer is not allowed to modify the number or order of execution of
3102this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3103 href="#i_store">store</a></tt> instructions.</p>
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3104<p>
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3105The optional constant "align" argument specifies the alignment of the operation
Chris Lattner21003c82008-01-06 21:04:43 +0000[diff] [blame]3106(that is, the alignment of the memory address). A value of 0 or an
3107omitted "align" argument means that the operation has the preferential
3108alignment for the target. It is the responsibility of the code emitter
3109to ensure that the alignment information is correct. Overestimating
3110the alignment results in an undefined behavior. Underestimating the
3111alignment may produce less efficient code. An alignment of 1 is always
3112safe.
3113</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3114<h5>Semantics:</h5>
3115<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
3116at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
3117<h5>Example:</h5>
3118<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]3119 store i32 3, i32* %ptr <i>; yields {void}</i>
3120 %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]3121</pre>
3122</div>
3123
3124<!-- _______________________________________________________________________ -->
3125<div class="doc_subsubsection">
3126 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3127</div>
3128
3129<div class="doc_text">
3130<h5>Syntax:</h5>
3131<pre>
3132 &lt;result&gt; = getelementptr &lt;ty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
3133</pre>
3134
3135<h5>Overview:</h5>
3136
3137<p>
3138The '<tt>getelementptr</tt>' instruction is used to get the address of a
3139subelement of an aggregate data structure.</p>
3140
3141<h5>Arguments:</h5>
3142
3143<p>This instruction takes a list of integer operands that indicate what
3144elements of the aggregate object to index to. The actual types of the arguments
3145provided depend on the type of the first pointer argument. The
3146'<tt>getelementptr</tt>' instruction is used to index down through the type
3147levels of a structure or to a specific index in an array. When indexing into a
3148structure, only <tt>i32</tt> integer constants are allowed. When indexing
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]3149into an array or pointer, only integers of 32 or 64 bits are allowed; 32-bit
3150values will be sign extended to 64-bits if required.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3151
3152<p>For example, let's consider a C code fragment and how it gets
3153compiled to LLVM:</p>
3154
3155<div class="doc_code">
3156<pre>
3157struct RT {
3158 char A;
3159 int B[10][20];
3160 char C;
3161};
3162struct ST {
3163 int X;
3164 double Y;
3165 struct RT Z;
3166};
3167
3168int *foo(struct ST *s) {
3169 return &amp;s[1].Z.B[5][13];
3170}
3171</pre>
3172</div>
3173
3174<p>The LLVM code generated by the GCC frontend is:</p>
3175
3176<div class="doc_code">
3177<pre>
3178%RT = type { i8 , [10 x [20 x i32]], i8 }
3179%ST = type { i32, double, %RT }
3180
3181define i32* %foo(%ST* %s) {
3182entry:
3183 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3184 ret i32* %reg
3185}
3186</pre>
3187</div>
3188
3189<h5>Semantics:</h5>
3190
3191<p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
3192on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
3193and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
3194<a href="#t_integer">integer</a> type but the value will always be sign extended
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3195to 64-bits. <a href="#t_struct">Structure</a> and <a href="#t_pstruct">packed
3196structure</a> types require <tt>i32</tt> <b>constants</b>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3197
3198<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3199type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3200}</tt>' type, a structure. The second index indexes into the third element of
3201the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3202i8 }</tt>' type, another structure. The third index indexes into the second
3203element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3204array. The two dimensions of the array are subscripted into, yielding an
3205'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3206to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3207
3208<p>Note that it is perfectly legal to index partially through a
3209structure, returning a pointer to an inner element. Because of this,
3210the LLVM code for the given testcase is equivalent to:</p>
3211
3212<pre>
3213 define i32* %foo(%ST* %s) {
3214 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3215 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3216 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3217 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3218 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3219 ret i32* %t5
3220 }
3221</pre>
3222
3223<p>Note that it is undefined to access an array out of bounds: array and
3224pointer indexes must always be within the defined bounds of the array type.
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]3225The one exception for this rule is zero length arrays. These arrays are
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3226defined to be accessible as variable length arrays, which requires access
3227beyond the zero'th element.</p>
3228
3229<p>The getelementptr instruction is often confusing. For some more insight
3230into how it works, see <a href="GetElementPtr.html">the getelementptr
3231FAQ</a>.</p>
3232
3233<h5>Example:</h5>
3234
3235<pre>
3236 <i>; yields [12 x i8]*:aptr</i>
3237 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
3238</pre>
3239</div>
3240
3241<!-- ======================================================================= -->
3242<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3243</div>
3244<div class="doc_text">
3245<p>The instructions in this category are the conversion instructions (casting)
3246which all take a single operand and a type. They perform various bit conversions
3247on the operand.</p>
3248</div>
3249
3250<!-- _______________________________________________________________________ -->
3251<div class="doc_subsubsection">
3252 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3253</div>
3254<div class="doc_text">
3255
3256<h5>Syntax:</h5>
3257<pre>
3258 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3259</pre>
3260
3261<h5>Overview:</h5>
3262<p>
3263The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3264</p>
3265
3266<h5>Arguments:</h5>
3267<p>
3268The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3269be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3270and type of the result, which must be an <a href="#t_integer">integer</a>
3271type. The bit size of <tt>value</tt> must be larger than the bit size of
3272<tt>ty2</tt>. Equal sized types are not allowed.</p>
3273
3274<h5>Semantics:</h5>
3275<p>
3276The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3277and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3278larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3279It will always truncate bits.</p>
3280
3281<h5>Example:</h5>
3282<pre>
3283 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3284 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3285 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3286</pre>
3287</div>
3288
3289<!-- _______________________________________________________________________ -->
3290<div class="doc_subsubsection">
3291 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3292</div>
3293<div class="doc_text">
3294
3295<h5>Syntax:</h5>
3296<pre>
3297 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3298</pre>
3299
3300<h5>Overview:</h5>
3301<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3302<tt>ty2</tt>.</p>
3303
3304
3305<h5>Arguments:</h5>
3306<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3307<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3308also be of <a href="#t_integer">integer</a> type. The bit size of the
3309<tt>value</tt> must be smaller than the bit size of the destination type,
3310<tt>ty2</tt>.</p>
3311
3312<h5>Semantics:</h5>
3313<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3314bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3315
3316<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3317
3318<h5>Example:</h5>
3319<pre>
3320 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3321 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3322</pre>
3323</div>
3324
3325<!-- _______________________________________________________________________ -->
3326<div class="doc_subsubsection">
3327 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3328</div>
3329<div class="doc_text">
3330
3331<h5>Syntax:</h5>
3332<pre>
3333 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3334</pre>
3335
3336<h5>Overview:</h5>
3337<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3338
3339<h5>Arguments:</h5>
3340<p>
3341The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3342<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3343also be of <a href="#t_integer">integer</a> type. The bit size of the
3344<tt>value</tt> must be smaller than the bit size of the destination type,
3345<tt>ty2</tt>.</p>
3346
3347<h5>Semantics:</h5>
3348<p>
3349The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3350bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3351the type <tt>ty2</tt>.</p>
3352
3353<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3354
3355<h5>Example:</h5>
3356<pre>
3357 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3358 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3359</pre>
3360</div>
3361
3362<!-- _______________________________________________________________________ -->
3363<div class="doc_subsubsection">
3364 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3365</div>
3366
3367<div class="doc_text">
3368
3369<h5>Syntax:</h5>
3370
3371<pre>
3372 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3373</pre>
3374
3375<h5>Overview:</h5>
3376<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3377<tt>ty2</tt>.</p>
3378
3379
3380<h5>Arguments:</h5>
3381<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3382 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3383cast it to. The size of <tt>value</tt> must be larger than the size of
3384<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3385<i>no-op cast</i>.</p>
3386
3387<h5>Semantics:</h5>
3388<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3389<a href="#t_floating">floating point</a> type to a smaller
3390<a href="#t_floating">floating point</a> type. If the value cannot fit within
3391the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3392
3393<h5>Example:</h5>
3394<pre>
3395 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3396 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3397</pre>
3398</div>
3399
3400<!-- _______________________________________________________________________ -->
3401<div class="doc_subsubsection">
3402 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3403</div>
3404<div class="doc_text">
3405
3406<h5>Syntax:</h5>
3407<pre>
3408 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3409</pre>
3410
3411<h5>Overview:</h5>
3412<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3413floating point value.</p>
3414
3415<h5>Arguments:</h5>
3416<p>The '<tt>fpext</tt>' instruction takes a
3417<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3418and a <a href="#t_floating">floating point</a> type to cast it to. The source
3419type must be smaller than the destination type.</p>
3420
3421<h5>Semantics:</h5>
3422<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3423<a href="#t_floating">floating point</a> type to a larger
3424<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3425used to make a <i>no-op cast</i> because it always changes bits. Use
3426<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3427
3428<h5>Example:</h5>
3429<pre>
3430 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3431 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3432</pre>
3433</div>
3434
3435<!-- _______________________________________________________________________ -->
3436<div class="doc_subsubsection">
3437 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3438</div>
3439<div class="doc_text">
3440
3441<h5>Syntax:</h5>
3442<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3443 &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]3444</pre>
3445
3446<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3447<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3448unsigned integer equivalent of type <tt>ty2</tt>.
3449</p>
3450
3451<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3452<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]3453scalar or vector <a href="#t_floating">floating point</a> value, and a type
3454to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3455type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3456vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3457
3458<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3459<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3460<a href="#t_floating">floating point</a> operand into the nearest (rounding
3461towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3462the results are undefined.</p>
3463
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3464<h5>Example:</h5>
3465<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3466 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3467 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3468 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3469</pre>
3470</div>
3471
3472<!-- _______________________________________________________________________ -->
3473<div class="doc_subsubsection">
3474 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3475</div>
3476<div class="doc_text">
3477
3478<h5>Syntax:</h5>
3479<pre>
3480 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3481</pre>
3482
3483<h5>Overview:</h5>
3484<p>The '<tt>fptosi</tt>' instruction converts
3485<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3486</p>
3487
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3488<h5>Arguments:</h5>
3489<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]3490scalar or vector <a href="#t_floating">floating point</a> value, and a type
3491to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3492type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3493vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3494
3495<h5>Semantics:</h5>
3496<p>The '<tt>fptosi</tt>' instruction converts its
3497<a href="#t_floating">floating point</a> operand into the nearest (rounding
3498towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3499the results are undefined.</p>
3500
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3501<h5>Example:</h5>
3502<pre>
3503 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3504 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3505 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3506</pre>
3507</div>
3508
3509<!-- _______________________________________________________________________ -->
3510<div class="doc_subsubsection">
3511 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3512</div>
3513<div class="doc_text">
3514
3515<h5>Syntax:</h5>
3516<pre>
3517 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3518</pre>
3519
3520<h5>Overview:</h5>
3521<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3522integer and converts that value to the <tt>ty2</tt> type.</p>
3523
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3524<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3525<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3526scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3527to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3528type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3529floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3530
3531<h5>Semantics:</h5>
3532<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3533integer quantity and converts it to the corresponding floating point value. If
3534the value cannot fit in the floating point value, the results are undefined.</p>
3535
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3536<h5>Example:</h5>
3537<pre>
3538 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3539 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3540</pre>
3541</div>
3542
3543<!-- _______________________________________________________________________ -->
3544<div class="doc_subsubsection">
3545 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3546</div>
3547<div class="doc_text">
3548
3549<h5>Syntax:</h5>
3550<pre>
3551 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3552</pre>
3553
3554<h5>Overview:</h5>
3555<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3556integer and converts that value to the <tt>ty2</tt> type.</p>
3557
3558<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3559<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3560scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3561to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3562type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3563floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3564
3565<h5>Semantics:</h5>
3566<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3567integer quantity and converts it to the corresponding floating point value. If
3568the value cannot fit in the floating point value, the results are undefined.</p>
3569
3570<h5>Example:</h5>
3571<pre>
3572 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3573 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3574</pre>
3575</div>
3576
3577<!-- _______________________________________________________________________ -->
3578<div class="doc_subsubsection">
3579 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3580</div>
3581<div class="doc_text">
3582
3583<h5>Syntax:</h5>
3584<pre>
3585 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3586</pre>
3587
3588<h5>Overview:</h5>
3589<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3590the integer type <tt>ty2</tt>.</p>
3591
3592<h5>Arguments:</h5>
3593<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3594must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3595<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3596
3597<h5>Semantics:</h5>
3598<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3599<tt>ty2</tt> by interpreting the pointer value as an integer and either
3600truncating or zero extending that value to the size of the integer type. If
3601<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3602<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3603are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3604change.</p>
3605
3606<h5>Example:</h5>
3607<pre>
3608 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3609 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3610</pre>
3611</div>
3612
3613<!-- _______________________________________________________________________ -->
3614<div class="doc_subsubsection">
3615 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3616</div>
3617<div class="doc_text">
3618
3619<h5>Syntax:</h5>
3620<pre>
3621 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3622</pre>
3623
3624<h5>Overview:</h5>
3625<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3626a pointer type, <tt>ty2</tt>.</p>
3627
3628<h5>Arguments:</h5>
3629<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3630value to cast, and a type to cast it to, which must be a
3631<a href="#t_pointer">pointer</a> type.
3632
3633<h5>Semantics:</h5>
3634<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3635<tt>ty2</tt> by applying either a zero extension or a truncation depending on
3636the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3637size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3638the size of a pointer then a zero extension is done. If they are the same size,
3639nothing is done (<i>no-op cast</i>).</p>
3640
3641<h5>Example:</h5>
3642<pre>
3643 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3644 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3645 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3646</pre>
3647</div>
3648
3649<!-- _______________________________________________________________________ -->
3650<div class="doc_subsubsection">
3651 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3652</div>
3653<div class="doc_text">
3654
3655<h5>Syntax:</h5>
3656<pre>
3657 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3658</pre>
3659
3660<h5>Overview:</h5>
3661<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3662<tt>ty2</tt> without changing any bits.</p>
3663
3664<h5>Arguments:</h5>
3665<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3666a first class value, and a type to cast it to, which must also be a <a
3667 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3668and the destination type, <tt>ty2</tt>, must be identical. If the source
3669type is a pointer, the destination type must also be a pointer.</p>
3670
3671<h5>Semantics:</h5>
3672<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3673<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3674this conversion. The conversion is done as if the <tt>value</tt> had been
3675stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3676converted to other pointer types with this instruction. To convert pointers to
3677other types, use the <a href="#i_inttoptr">inttoptr</a> or
3678<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3679
3680<h5>Example:</h5>
3681<pre>
3682 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3683 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3684 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3685</pre>
3686</div>
3687
3688<!-- ======================================================================= -->
3689<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3690<div class="doc_text">
3691<p>The instructions in this category are the "miscellaneous"
3692instructions, which defy better classification.</p>
3693</div>
3694
3695<!-- _______________________________________________________________________ -->
3696<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3697</div>
3698<div class="doc_text">
3699<h5>Syntax:</h5>
3700<pre> &lt;result&gt; = icmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3701</pre>
3702<h5>Overview:</h5>
3703<p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
Chris Lattner10368b62008-04-02 00:38:26 +0000[diff] [blame]3704of its two integer or pointer operands.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3705<h5>Arguments:</h5>
3706<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3707the condition code indicating the kind of comparison to perform. It is not
3708a value, just a keyword. The possible condition code are:
3709<ol>
3710 <li><tt>eq</tt>: equal</li>
3711 <li><tt>ne</tt>: not equal </li>
3712 <li><tt>ugt</tt>: unsigned greater than</li>
3713 <li><tt>uge</tt>: unsigned greater or equal</li>
3714 <li><tt>ult</tt>: unsigned less than</li>
3715 <li><tt>ule</tt>: unsigned less or equal</li>
3716 <li><tt>sgt</tt>: signed greater than</li>
3717 <li><tt>sge</tt>: signed greater or equal</li>
3718 <li><tt>slt</tt>: signed less than</li>
3719 <li><tt>sle</tt>: signed less or equal</li>
3720</ol>
3721<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3722<a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3723<h5>Semantics:</h5>
3724<p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3725the condition code given as <tt>cond</tt>. The comparison performed always
3726yields a <a href="#t_primitive">i1</a> result, as follows:
3727<ol>
3728 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3729 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3730 </li>
3731 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3732 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3733 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3734 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3735 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3736 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3737 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3738 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3739 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3740 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3741 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3742 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3743 <li><tt>sge</tt>: interprets the operands as signed values and yields
3744 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3745 <li><tt>slt</tt>: interprets the operands as signed values and yields
3746 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3747 <li><tt>sle</tt>: interprets the operands as signed values and yields
3748 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3749</ol>
3750<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3751values are compared as if they were integers.</p>
3752
3753<h5>Example:</h5>
3754<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
3755 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
3756 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
3757 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3758 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3759 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
3760</pre>
3761</div>
3762
3763<!-- _______________________________________________________________________ -->
3764<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3765</div>
3766<div class="doc_text">
3767<h5>Syntax:</h5>
3768<pre> &lt;result&gt; = fcmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3769</pre>
3770<h5>Overview:</h5>
3771<p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3772of its floating point operands.</p>
3773<h5>Arguments:</h5>
3774<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3775the condition code indicating the kind of comparison to perform. It is not
3776a value, just a keyword. The possible condition code are:
3777<ol>
3778 <li><tt>false</tt>: no comparison, always returns false</li>
3779 <li><tt>oeq</tt>: ordered and equal</li>
3780 <li><tt>ogt</tt>: ordered and greater than </li>
3781 <li><tt>oge</tt>: ordered and greater than or equal</li>
3782 <li><tt>olt</tt>: ordered and less than </li>
3783 <li><tt>ole</tt>: ordered and less than or equal</li>
3784 <li><tt>one</tt>: ordered and not equal</li>
3785 <li><tt>ord</tt>: ordered (no nans)</li>
3786 <li><tt>ueq</tt>: unordered or equal</li>
3787 <li><tt>ugt</tt>: unordered or greater than </li>
3788 <li><tt>uge</tt>: unordered or greater than or equal</li>
3789 <li><tt>ult</tt>: unordered or less than </li>
3790 <li><tt>ule</tt>: unordered or less than or equal</li>
3791 <li><tt>une</tt>: unordered or not equal</li>
3792 <li><tt>uno</tt>: unordered (either nans)</li>
3793 <li><tt>true</tt>: no comparison, always returns true</li>
3794</ol>
3795<p><i>Ordered</i> means that neither operand is a QNAN while
3796<i>unordered</i> means that either operand may be a QNAN.</p>
3797<p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3798<a href="#t_floating">floating point</a> typed. They must have identical
3799types.</p>
3800<h5>Semantics:</h5>
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]3801<p>The '<tt>fcmp</tt>' instruction compares <tt>var1</tt> and <tt>var2</tt>
3802according to the condition code given as <tt>cond</tt>. The comparison performed
3803always yields a <a href="#t_primitive">i1</a> result, as follows:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3804<ol>
3805 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3806 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3807 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3808 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3809 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3810 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3811 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3812 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3813 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3814 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3815 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3816 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3817 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3818 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3819 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3820 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3821 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3822 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3823 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3824 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3825 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3826 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3827 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3828 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3829 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3830 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3831 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3832 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3833</ol>
3834
3835<h5>Example:</h5>
3836<pre> &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3837 &lt;result&gt; = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3838 &lt;result&gt; = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3839 &lt;result&gt; = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3840</pre>
3841</div>
3842
3843<!-- _______________________________________________________________________ -->
Nate Begeman646fa482008-05-12 19:01:56 +0000[diff] [blame]3844<div class="doc_subsubsection">
3845 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
3846</div>
3847<div class="doc_text">
3848<h5>Syntax:</h5>
3849<pre> &lt;result&gt; = vicmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
3850</pre>
3851<h5>Overview:</h5>
3852<p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
3853element-wise comparison of its two integer vector operands.</p>
3854<h5>Arguments:</h5>
3855<p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
3856the condition code indicating the kind of comparison to perform. It is not
3857a value, just a keyword. The possible condition code are:
3858<ol>
3859 <li><tt>eq</tt>: equal</li>
3860 <li><tt>ne</tt>: not equal </li>
3861 <li><tt>ugt</tt>: unsigned greater than</li>
3862 <li><tt>uge</tt>: unsigned greater or equal</li>
3863 <li><tt>ult</tt>: unsigned less than</li>
3864 <li><tt>ule</tt>: unsigned less or equal</li>
3865 <li><tt>sgt</tt>: signed greater than</li>
3866 <li><tt>sge</tt>: signed greater or equal</li>
3867 <li><tt>slt</tt>: signed less than</li>
3868 <li><tt>sle</tt>: signed less or equal</li>
3869</ol>
3870<p>The remaining two arguments must be <a href="#t_vector">vector</a> of
3871<a href="#t_integer">integer</a> typed. They must also be identical types.</p>
3872<h5>Semantics:</h5>
3873<p>The '<tt>vicmp</tt>' instruction compares <tt>var1</tt> and <tt>var2</tt>
3874according to the condition code given as <tt>cond</tt>. The comparison yields a
3875<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
3876identical type as the values being compared. The most significant bit in each
3877element is 1 if the element-wise comparison evaluates to true, and is 0
3878otherwise. All other bits of the result are undefined. The condition codes
3879are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
3880instruction</a>.
3881
3882<h5>Example:</h5>
3883<pre>
3884 &lt;result&gt; = vicmp eq <2 x i32> < i32 4, i32 0 >, < i32 5, i32 0 > <i>; yields: result=<2 x i32> < i32 0, i32 -1 ></i>
3885 &lt;result&gt; = vicmp ult <2 x i8> < i8 1, i8 2 >, < i8 2, i8 2> <i>; yields: result=<2 x i8> < i8 -1, i8 0 ></i>
3886</pre>
3887</div>
3888
3889<!-- _______________________________________________________________________ -->
3890<div class="doc_subsubsection">
3891 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
3892</div>
3893<div class="doc_text">
3894<h5>Syntax:</h5>
3895<pre> &lt;result&gt; = vfcmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt;</pre>
3896<h5>Overview:</h5>
3897<p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
3898element-wise comparison of its two floating point vector operands. The output
3899elements have the same width as the input elements.</p>
3900<h5>Arguments:</h5>
3901<p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
3902the condition code indicating the kind of comparison to perform. It is not
3903a value, just a keyword. The possible condition code are:
3904<ol>
3905 <li><tt>false</tt>: no comparison, always returns false</li>
3906 <li><tt>oeq</tt>: ordered and equal</li>
3907 <li><tt>ogt</tt>: ordered and greater than </li>
3908 <li><tt>oge</tt>: ordered and greater than or equal</li>
3909 <li><tt>olt</tt>: ordered and less than </li>
3910 <li><tt>ole</tt>: ordered and less than or equal</li>
3911 <li><tt>one</tt>: ordered and not equal</li>
3912 <li><tt>ord</tt>: ordered (no nans)</li>
3913 <li><tt>ueq</tt>: unordered or equal</li>
3914 <li><tt>ugt</tt>: unordered or greater than </li>
3915 <li><tt>uge</tt>: unordered or greater than or equal</li>
3916 <li><tt>ult</tt>: unordered or less than </li>
3917 <li><tt>ule</tt>: unordered or less than or equal</li>
3918 <li><tt>une</tt>: unordered or not equal</li>
3919 <li><tt>uno</tt>: unordered (either nans)</li>
3920 <li><tt>true</tt>: no comparison, always returns true</li>
3921</ol>
3922<p>The remaining two arguments must be <a href="#t_vector">vector</a> of
3923<a href="#t_floating">floating point</a> typed. They must also be identical
3924types.</p>
3925<h5>Semantics:</h5>
3926<p>The '<tt>vfcmp</tt>' instruction compares <tt>var1</tt> and <tt>var2</tt>
3927according to the condition code given as <tt>cond</tt>. The comparison yields a
3928<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
3929an identical number of elements as the values being compared, and each element
3930having identical with to the width of the floating point elements. The most
3931significant bit in each element is 1 if the element-wise comparison evaluates to
3932true, and is 0 otherwise. All other bits of the result are undefined. The
3933condition codes are evaluated identically to the
3934<a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.
3935
3936<h5>Example:</h5>
3937<pre>
3938 &lt;result&gt; = vfcmp oeq <2 x float> < float 4, float 0 >, < float 5, float 0 > <i>; yields: result=<2 x i32> < i32 0, i32 -1 ></i>
3939 &lt;result&gt; = vfcmp ult <2 x double> < double 1, double 2 >, < double 2, double 2> <i>; yields: result=<2 x i64> < i64 -1, i64 0 ></i>
3940</pre>
3941</div>
3942
3943<!-- _______________________________________________________________________ -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3944<div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3945Instruction</a> </div>
3946<div class="doc_text">
3947<h5>Syntax:</h5>
3948<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
3949<h5>Overview:</h5>
3950<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
3951the SSA graph representing the function.</p>
3952<h5>Arguments:</h5>
3953<p>The type of the incoming values is specified with the first type
3954field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3955as arguments, with one pair for each predecessor basic block of the
3956current block. Only values of <a href="#t_firstclass">first class</a>
3957type may be used as the value arguments to the PHI node. Only labels
3958may be used as the label arguments.</p>
3959<p>There must be no non-phi instructions between the start of a basic
3960block and the PHI instructions: i.e. PHI instructions must be first in
3961a basic block.</p>
3962<h5>Semantics:</h5>
3963<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3964specified by the pair corresponding to the predecessor basic block that executed
3965just prior to the current block.</p>
3966<h5>Example:</h5>
3967<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>
3968</div>
3969
3970<!-- _______________________________________________________________________ -->
3971<div class="doc_subsubsection">
3972 <a name="i_select">'<tt>select</tt>' Instruction</a>
3973</div>
3974
3975<div class="doc_text">
3976
3977<h5>Syntax:</h5>
3978
3979<pre>
3980 &lt;result&gt; = select i1 &lt;cond&gt;, &lt;ty&gt; &lt;val1&gt;, &lt;ty&gt; &lt;val2&gt; <i>; yields ty</i>
3981</pre>
3982
3983<h5>Overview:</h5>
3984
3985<p>
3986The '<tt>select</tt>' instruction is used to choose one value based on a
3987condition, without branching.
3988</p>
3989
3990
3991<h5>Arguments:</h5>
3992
3993<p>
3994The '<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.
3995</p>
3996
3997<h5>Semantics:</h5>
3998
3999<p>
4000If the boolean condition evaluates to true, the instruction returns the first
4001value argument; otherwise, it returns the second value argument.
4002</p>
4003
4004<h5>Example:</h5>
4005
4006<pre>
4007 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4008</pre>
4009</div>
4010
4011
4012<!-- _______________________________________________________________________ -->
4013<div class="doc_subsubsection">
4014 <a name="i_call">'<tt>call</tt>' Instruction</a>
4015</div>
4016
4017<div class="doc_text">
4018
4019<h5>Syntax:</h5>
4020<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4021 &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]4022</pre>
4023
4024<h5>Overview:</h5>
4025
4026<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4027
4028<h5>Arguments:</h5>
4029
4030<p>This instruction requires several arguments:</p>
4031
4032<ol>
4033 <li>
4034 <p>The optional "tail" marker indicates whether the callee function accesses
4035 any allocas or varargs in the caller. If the "tail" marker is present, the
4036 function call is eligible for tail call optimization. Note that calls may
4037 be marked "tail" even if they do not occur before a <a
4038 href="#i_ret"><tt>ret</tt></a> instruction.
4039 </li>
4040 <li>
4041 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4042 convention</a> the call should use. If none is specified, the call defaults
4043 to using C calling conventions.
4044 </li>
4045 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4046 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4047 the type of the return value. Functions that return no value are marked
4048 <tt><a href="#t_void">void</a></tt>.</p>
4049 </li>
4050 <li>
4051 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4052 value being invoked. The argument types must match the types implied by
4053 this signature. This type can be omitted if the function is not varargs
4054 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4055 </li>
4056 <li>
4057 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4058 be invoked. In most cases, this is a direct function invocation, but
4059 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4060 to function value.</p>
4061 </li>
4062 <li>
4063 <p>'<tt>function args</tt>': argument list whose types match the
4064 function signature argument types. All arguments must be of
4065 <a href="#t_firstclass">first class</a> type. If the function signature
4066 indicates the function accepts a variable number of arguments, the extra
4067 arguments can be specified.</p>
4068 </li>
4069</ol>
4070
4071<h5>Semantics:</h5>
4072
4073<p>The '<tt>call</tt>' instruction is used to cause control flow to
4074transfer to a specified function, with its incoming arguments bound to
4075the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4076instruction in the called function, control flow continues with the
4077instruction after the function call, and the return value of the
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4078function is bound to the result argument. If the callee returns multiple
4079values then the return values of the function are only accessible through
4080the '<tt><a href="#i_getresult">getresult</a></tt>' instruction.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4081
4082<h5>Example:</h5>
4083
4084<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]4085 %retval = call i32 @test(i32 %argc)
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4086 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4087 %X = tail call i32 @foo() <i>; yields i32</i>
4088 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4089 call void %foo(i8 97 signext)
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4090
4091 %struct.A = type { i32, i8 }
Chris Lattner5e893ef2008-03-21 17:24:17 +0000[diff] [blame]4092 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
4093 %gr = getresult %struct.A %r, 0 <i>; yields i32</i>
4094 %gr1 = getresult %struct.A %r, 1 <i>; yields i8</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4095</pre>
4096
4097</div>
4098
4099<!-- _______________________________________________________________________ -->
4100<div class="doc_subsubsection">
4101 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4102</div>
4103
4104<div class="doc_text">
4105
4106<h5>Syntax:</h5>
4107
4108<pre>
4109 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
4110</pre>
4111
4112<h5>Overview:</h5>
4113
4114<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4115the "variable argument" area of a function call. It is used to implement the
4116<tt>va_arg</tt> macro in C.</p>
4117
4118<h5>Arguments:</h5>
4119
4120<p>This instruction takes a <tt>va_list*</tt> value and the type of
4121the argument. It returns a value of the specified argument type and
4122increments the <tt>va_list</tt> to point to the next argument. The
4123actual type of <tt>va_list</tt> is target specific.</p>
4124
4125<h5>Semantics:</h5>
4126
4127<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4128type from the specified <tt>va_list</tt> and causes the
4129<tt>va_list</tt> to point to the next argument. For more information,
4130see the variable argument handling <a href="#int_varargs">Intrinsic
4131Functions</a>.</p>
4132
4133<p>It is legal for this instruction to be called in a function which does not
4134take a variable number of arguments, for example, the <tt>vfprintf</tt>
4135function.</p>
4136
4137<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4138href="#intrinsics">intrinsic function</a> because it takes a type as an
4139argument.</p>
4140
4141<h5>Example:</h5>
4142
4143<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4144
4145</div>
4146
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4147<!-- _______________________________________________________________________ -->
4148<div class="doc_subsubsection">
4149 <a name="i_getresult">'<tt>getresult</tt>' Instruction</a>
4150</div>
4151
4152<div class="doc_text">
4153
4154<h5>Syntax:</h5>
4155<pre>
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4156 &lt;resultval&gt; = getresult &lt;type&gt; &lt;retval&gt;, &lt;index&gt;
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4157</pre>
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4158
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4159<h5>Overview:</h5>
4160
4161<p> The '<tt>getresult</tt>' instruction is used to extract individual values
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4162from a '<tt><a href="#i_call">call</a></tt>'
4163or '<tt><a href="#i_invoke">invoke</a></tt>' instruction that returns multiple
4164results.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4165
4166<h5>Arguments:</h5>
4167
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4168<p>The '<tt>getresult</tt>' instruction takes a call or invoke value as its
Chris Lattnerd8dd3522008-04-23 04:06:52 +0000[diff] [blame]4169first argument, or an undef value. The value must have <a
4170href="#t_struct">structure type</a>. The second argument is a constant
4171unsigned index value which must be in range for the number of values returned
4172by the call.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4173
4174<h5>Semantics:</h5>
4175
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4176<p>The '<tt>getresult</tt>' instruction extracts the element identified by
4177'<tt>index</tt>' from the aggregate value.</p>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4178
4179<h5>Example:</h5>
4180
4181<pre>
4182 %struct.A = type { i32, i8 }
4183
4184 %r = call %struct.A @foo()
Chris Lattneree9da3f2008-03-21 17:20:51 +0000[diff] [blame]4185 %gr = getresult %struct.A %r, 0 <i>; yields i32:%gr</i>
4186 %gr1 = getresult %struct.A %r, 1 <i>; yields i8:%gr1</i>
Devang Patela3cc5372008-03-10 20:49:15 +0000[diff] [blame]4187 add i32 %gr, 42
4188 add i8 %gr1, 41
4189</pre>
4190
4191</div>
4192
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4193<!-- *********************************************************************** -->
4194<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4195<!-- *********************************************************************** -->
4196
4197<div class="doc_text">
4198
4199<p>LLVM supports the notion of an "intrinsic function". These functions have
4200well known names and semantics and are required to follow certain restrictions.
4201Overall, these intrinsics represent an extension mechanism for the LLVM
4202language that does not require changing all of the transformations in LLVM when
4203adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4204
4205<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4206prefix is reserved in LLVM for intrinsic names; thus, function names may not
4207begin with this prefix. Intrinsic functions must always be external functions:
4208you cannot define the body of intrinsic functions. Intrinsic functions may
4209only be used in call or invoke instructions: it is illegal to take the address
4210of an intrinsic function. Additionally, because intrinsic functions are part
4211of the LLVM language, it is required if any are added that they be documented
4212here.</p>
4213
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4214<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4215a family of functions that perform the same operation but on different data
4216types. Because LLVM can represent over 8 million different integer types,
4217overloading is used commonly to allow an intrinsic function to operate on any
4218integer type. One or more of the argument types or the result type can be
4219overloaded to accept any integer type. Argument types may also be defined as
4220exactly matching a previous argument's type or the result type. This allows an
4221intrinsic function which accepts multiple arguments, but needs all of them to
4222be of the same type, to only be overloaded with respect to a single argument or
4223the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4224
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4225<p>Overloaded intrinsics will have the names of its overloaded argument types
4226encoded into its function name, each preceded by a period. Only those types
4227which are overloaded result in a name suffix. Arguments whose type is matched
4228against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4229take an integer of any width and returns an integer of exactly the same integer
4230width. This leads to a family of functions such as
4231<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4232Only one type, the return type, is overloaded, and only one type suffix is
4233required. Because the argument's type is matched against the return type, it
4234does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4235
4236<p>To learn how to add an intrinsic function, please see the
4237<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4238</p>
4239
4240</div>
4241
4242<!-- ======================================================================= -->
4243<div class="doc_subsection">
4244 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4245</div>
4246
4247<div class="doc_text">
4248
4249<p>Variable argument support is defined in LLVM with the <a
4250 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4251intrinsic functions. These functions are related to the similarly
4252named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
4253
4254<p>All of these functions operate on arguments that use a
4255target-specific value type "<tt>va_list</tt>". The LLVM assembly
4256language reference manual does not define what this type is, so all
4257transformations should be prepared to handle these functions regardless of
4258the type used.</p>
4259
4260<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4261instruction and the variable argument handling intrinsic functions are
4262used.</p>
4263
4264<div class="doc_code">
4265<pre>
4266define i32 @test(i32 %X, ...) {
4267 ; Initialize variable argument processing
4268 %ap = alloca i8*
4269 %ap2 = bitcast i8** %ap to i8*
4270 call void @llvm.va_start(i8* %ap2)
4271
4272 ; Read a single integer argument
4273 %tmp = va_arg i8** %ap, i32
4274
4275 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4276 %aq = alloca i8*
4277 %aq2 = bitcast i8** %aq to i8*
4278 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4279 call void @llvm.va_end(i8* %aq2)
4280
4281 ; Stop processing of arguments.
4282 call void @llvm.va_end(i8* %ap2)
4283 ret i32 %tmp
4284}
4285
4286declare void @llvm.va_start(i8*)
4287declare void @llvm.va_copy(i8*, i8*)
4288declare void @llvm.va_end(i8*)
4289</pre>
4290</div>
4291
4292</div>
4293
4294<!-- _______________________________________________________________________ -->
4295<div class="doc_subsubsection">
4296 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4297</div>
4298
4299
4300<div class="doc_text">
4301<h5>Syntax:</h5>
4302<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
4303<h5>Overview:</h5>
4304<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
4305<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
4306href="#i_va_arg">va_arg</a></tt>.</p>
4307
4308<h5>Arguments:</h5>
4309
4310<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
4311
4312<h5>Semantics:</h5>
4313
4314<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
4315macro available in C. In a target-dependent way, it initializes the
4316<tt>va_list</tt> element to which the argument points, so that the next call to
4317<tt>va_arg</tt> will produce the first variable argument passed to the function.
4318Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4319last argument of the function as the compiler can figure that out.</p>
4320
4321</div>
4322
4323<!-- _______________________________________________________________________ -->
4324<div class="doc_subsubsection">
4325 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4326</div>
4327
4328<div class="doc_text">
4329<h5>Syntax:</h5>
4330<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
4331<h5>Overview:</h5>
4332
4333<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
4334which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4335or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4336
4337<h5>Arguments:</h5>
4338
4339<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4340
4341<h5>Semantics:</h5>
4342
4343<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4344macro available in C. In a target-dependent way, it destroys the
4345<tt>va_list</tt> element to which the argument points. Calls to <a
4346href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4347<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4348<tt>llvm.va_end</tt>.</p>
4349
4350</div>
4351
4352<!-- _______________________________________________________________________ -->
4353<div class="doc_subsubsection">
4354 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4355</div>
4356
4357<div class="doc_text">
4358
4359<h5>Syntax:</h5>
4360
4361<pre>
4362 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
4363</pre>
4364
4365<h5>Overview:</h5>
4366
4367<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4368from the source argument list to the destination argument list.</p>
4369
4370<h5>Arguments:</h5>
4371
4372<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4373The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4374
4375
4376<h5>Semantics:</h5>
4377
4378<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4379macro available in C. In a target-dependent way, it copies the source
4380<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4381intrinsic is necessary because the <tt><a href="#int_va_start">
4382llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4383example, memory allocation.</p>
4384
4385</div>
4386
4387<!-- ======================================================================= -->
4388<div class="doc_subsection">
4389 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4390</div>
4391
4392<div class="doc_text">
4393
4394<p>
4395LLVM support for <a href="GarbageCollection.html">Accurate Garbage
4396Collection</a> requires the implementation and generation of these intrinsics.
4397These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4398stack</a>, as well as garbage collector implementations that require <a
4399href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4400Front-ends for type-safe garbage collected languages should generate these
4401intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4402href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4403</p>
Christopher Lambcfe00962007-12-17 01:00:21 +0000[diff] [blame]4404
4405<p>The garbage collection intrinsics only operate on objects in the generic
4406 address space (address space zero).</p>
4407
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4408</div>
4409
4410<!-- _______________________________________________________________________ -->
4411<div class="doc_subsubsection">
4412 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4413</div>
4414
4415<div class="doc_text">
4416
4417<h5>Syntax:</h5>
4418
4419<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4420 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4421</pre>
4422
4423<h5>Overview:</h5>
4424
4425<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4426the code generator, and allows some metadata to be associated with it.</p>
4427
4428<h5>Arguments:</h5>
4429
4430<p>The first argument specifies the address of a stack object that contains the
4431root pointer. The second pointer (which must be either a constant or a global
4432value address) contains the meta-data to be associated with the root.</p>
4433
4434<h5>Semantics:</h5>
4435
Chris Lattnera7d94ba2008-04-24 05:59:56 +0000[diff] [blame]4436<p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4437location. At compile-time, the code generator generates information to allow
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4438the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4439intrinsic may only be used in a function which <a href="#gc">specifies a GC
4440algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4441
4442</div>
4443
4444
4445<!-- _______________________________________________________________________ -->
4446<div class="doc_subsubsection">
4447 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4448</div>
4449
4450<div class="doc_text">
4451
4452<h5>Syntax:</h5>
4453
4454<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4455 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4456</pre>
4457
4458<h5>Overview:</h5>
4459
4460<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4461locations, allowing garbage collector implementations that require read
4462barriers.</p>
4463
4464<h5>Arguments:</h5>
4465
4466<p>The second argument is the address to read from, which should be an address
4467allocated from the garbage collector. The first object is a pointer to the
4468start of the referenced object, if needed by the language runtime (otherwise
4469null).</p>
4470
4471<h5>Semantics:</h5>
4472
4473<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4474instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4475garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4476may only be used in a function which <a href="#gc">specifies a GC
4477algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4478
4479</div>
4480
4481
4482<!-- _______________________________________________________________________ -->
4483<div class="doc_subsubsection">
4484 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4485</div>
4486
4487<div class="doc_text">
4488
4489<h5>Syntax:</h5>
4490
4491<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4492 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4493</pre>
4494
4495<h5>Overview:</h5>
4496
4497<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4498locations, allowing garbage collector implementations that require write
4499barriers (such as generational or reference counting collectors).</p>
4500
4501<h5>Arguments:</h5>
4502
4503<p>The first argument is the reference to store, the second is the start of the
4504object to store it to, and the third is the address of the field of Obj to
4505store to. If the runtime does not require a pointer to the object, Obj may be
4506null.</p>
4507
4508<h5>Semantics:</h5>
4509
4510<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4511instruction, but may be replaced with substantially more complex code by the
Gordon Henriksen40393542007-12-25 02:31:26 +0000[diff] [blame]4512garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4513may only be used in a function which <a href="#gc">specifies a GC
4514algorithm</a>.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4515
4516</div>
4517
4518
4519
4520<!-- ======================================================================= -->
4521<div class="doc_subsection">
4522 <a name="int_codegen">Code Generator Intrinsics</a>
4523</div>
4524
4525<div class="doc_text">
4526<p>
4527These intrinsics are provided by LLVM to expose special features that may only
4528be implemented with code generator support.
4529</p>
4530
4531</div>
4532
4533<!-- _______________________________________________________________________ -->
4534<div class="doc_subsubsection">
4535 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4536</div>
4537
4538<div class="doc_text">
4539
4540<h5>Syntax:</h5>
4541<pre>
4542 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
4543</pre>
4544
4545<h5>Overview:</h5>
4546
4547<p>
4548The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4549target-specific value indicating the return address of the current function
4550or one of its callers.
4551</p>
4552
4553<h5>Arguments:</h5>
4554
4555<p>
4556The argument to this intrinsic indicates which function to return the address
4557for. Zero indicates the calling function, one indicates its caller, etc. The
4558argument is <b>required</b> to be a constant integer value.
4559</p>
4560
4561<h5>Semantics:</h5>
4562
4563<p>
4564The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4565the return address of the specified call frame, or zero if it cannot be
4566identified. The value returned by this intrinsic is likely to be incorrect or 0
4567for arguments other than zero, so it should only be used for debugging purposes.
4568</p>
4569
4570<p>
4571Note that calling this intrinsic does not prevent function inlining or other
4572aggressive transformations, so the value returned may not be that of the obvious
4573source-language caller.
4574</p>
4575</div>
4576
4577
4578<!-- _______________________________________________________________________ -->
4579<div class="doc_subsubsection">
4580 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4581</div>
4582
4583<div class="doc_text">
4584
4585<h5>Syntax:</h5>
4586<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4587 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4588</pre>
4589
4590<h5>Overview:</h5>
4591
4592<p>
4593The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4594target-specific frame pointer value for the specified stack frame.
4595</p>
4596
4597<h5>Arguments:</h5>
4598
4599<p>
4600The argument to this intrinsic indicates which function to return the frame
4601pointer for. Zero indicates the calling function, one indicates its caller,
4602etc. The argument is <b>required</b> to be a constant integer value.
4603</p>
4604
4605<h5>Semantics:</h5>
4606
4607<p>
4608The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4609the frame address of the specified call frame, or zero if it cannot be
4610identified. The value returned by this intrinsic is likely to be incorrect or 0
4611for arguments other than zero, so it should only be used for debugging purposes.
4612</p>
4613
4614<p>
4615Note that calling this intrinsic does not prevent function inlining or other
4616aggressive transformations, so the value returned may not be that of the obvious
4617source-language caller.
4618</p>
4619</div>
4620
4621<!-- _______________________________________________________________________ -->
4622<div class="doc_subsubsection">
4623 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4624</div>
4625
4626<div class="doc_text">
4627
4628<h5>Syntax:</h5>
4629<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4630 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4631</pre>
4632
4633<h5>Overview:</h5>
4634
4635<p>
4636The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4637the function stack, for use with <a href="#int_stackrestore">
4638<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4639features like scoped automatic variable sized arrays in C99.
4640</p>
4641
4642<h5>Semantics:</h5>
4643
4644<p>
4645This intrinsic returns a opaque pointer value that can be passed to <a
4646href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4647<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4648<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4649state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4650practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4651that were allocated after the <tt>llvm.stacksave</tt> was executed.
4652</p>
4653
4654</div>
4655
4656<!-- _______________________________________________________________________ -->
4657<div class="doc_subsubsection">
4658 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4659</div>
4660
4661<div class="doc_text">
4662
4663<h5>Syntax:</h5>
4664<pre>
4665 declare void @llvm.stackrestore(i8 * %ptr)
4666</pre>
4667
4668<h5>Overview:</h5>
4669
4670<p>
4671The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4672the function stack to the state it was in when the corresponding <a
4673href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4674useful for implementing language features like scoped automatic variable sized
4675arrays in C99.
4676</p>
4677
4678<h5>Semantics:</h5>
4679
4680<p>
4681See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4682</p>
4683
4684</div>
4685
4686
4687<!-- _______________________________________________________________________ -->
4688<div class="doc_subsubsection">
4689 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4690</div>
4691
4692<div class="doc_text">
4693
4694<h5>Syntax:</h5>
4695<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4696 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]4697</pre>
4698
4699<h5>Overview:</h5>
4700
4701
4702<p>
4703The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4704a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4705no
4706effect on the behavior of the program but can change its performance
4707characteristics.
4708</p>
4709
4710<h5>Arguments:</h5>
4711
4712<p>
4713<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4714determining if the fetch should be for a read (0) or write (1), and
4715<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4716locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4717<tt>locality</tt> arguments must be constant integers.
4718</p>
4719
4720<h5>Semantics:</h5>
4721
4722<p>
4723This intrinsic does not modify the behavior of the program. In particular,
4724prefetches cannot trap and do not produce a value. On targets that support this
4725intrinsic, the prefetch can provide hints to the processor cache for better
4726performance.
4727</p>
4728
4729</div>
4730
4731<!-- _______________________________________________________________________ -->
4732<div class="doc_subsubsection">
4733 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4734</div>
4735
4736<div class="doc_text">
4737
4738<h5>Syntax:</h5>
4739<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4740 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4741</pre>
4742
4743<h5>Overview:</h5>
4744
4745
4746<p>
4747The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4748(PC) in a region of
4749code to simulators and other tools. The method is target specific, but it is
4750expected that the marker will use exported symbols to transmit the PC of the marker.
4751The marker makes no guarantees that it will remain with any specific instruction
4752after optimizations. It is possible that the presence of a marker will inhibit
4753optimizations. The intended use is to be inserted after optimizations to allow
4754correlations of simulation runs.
4755</p>
4756
4757<h5>Arguments:</h5>
4758
4759<p>
4760<tt>id</tt> is a numerical id identifying the marker.
4761</p>
4762
4763<h5>Semantics:</h5>
4764
4765<p>
4766This intrinsic does not modify the behavior of the program. Backends that do not
4767support this intrinisic may ignore it.
4768</p>
4769
4770</div>
4771
4772<!-- _______________________________________________________________________ -->
4773<div class="doc_subsubsection">
4774 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4775</div>
4776
4777<div class="doc_text">
4778
4779<h5>Syntax:</h5>
4780<pre>
4781 declare i64 @llvm.readcyclecounter( )
4782</pre>
4783
4784<h5>Overview:</h5>
4785
4786
4787<p>
4788The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4789counter register (or similar low latency, high accuracy clocks) on those targets
4790that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4791As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4792should only be used for small timings.
4793</p>
4794
4795<h5>Semantics:</h5>
4796
4797<p>
4798When directly supported, reading the cycle counter should not modify any memory.
4799Implementations are allowed to either return a application specific value or a
4800system wide value. On backends without support, this is lowered to a constant 0.
4801</p>
4802
4803</div>
4804
4805<!-- ======================================================================= -->
4806<div class="doc_subsection">
4807 <a name="int_libc">Standard C Library Intrinsics</a>
4808</div>
4809
4810<div class="doc_text">
4811<p>
4812LLVM provides intrinsics for a few important standard C library functions.
4813These intrinsics allow source-language front-ends to pass information about the
4814alignment of the pointer arguments to the code generator, providing opportunity
4815for more efficient code generation.
4816</p>
4817
4818</div>
4819
4820<!-- _______________________________________________________________________ -->
4821<div class="doc_subsubsection">
4822 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4823</div>
4824
4825<div class="doc_text">
4826
4827<h5>Syntax:</h5>
4828<pre>
4829 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4830 i32 &lt;len&gt;, i32 &lt;align&gt;)
4831 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4832 i64 &lt;len&gt;, i32 &lt;align&gt;)
4833</pre>
4834
4835<h5>Overview:</h5>
4836
4837<p>
4838The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4839location to the destination location.
4840</p>
4841
4842<p>
4843Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4844intrinsics do not return a value, and takes an extra alignment argument.
4845</p>
4846
4847<h5>Arguments:</h5>
4848
4849<p>
4850The first argument is a pointer to the destination, the second is a pointer to
4851the source. The third argument is an integer argument
4852specifying the number of bytes to copy, and the fourth argument is the alignment
4853of the source and destination locations.
4854</p>
4855
4856<p>
4857If the call to this intrinisic has an alignment value that is not 0 or 1, then
4858the caller guarantees that both the source and destination pointers are aligned
4859to that boundary.
4860</p>
4861
4862<h5>Semantics:</h5>
4863
4864<p>
4865The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4866location to the destination location, which are not allowed to overlap. It
4867copies "len" bytes of memory over. If the argument is known to be aligned to
4868some boundary, this can be specified as the fourth argument, otherwise it should
4869be set to 0 or 1.
4870</p>
4871</div>
4872
4873
4874<!-- _______________________________________________________________________ -->
4875<div class="doc_subsubsection">
4876 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4877</div>
4878
4879<div class="doc_text">
4880
4881<h5>Syntax:</h5>
4882<pre>
4883 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4884 i32 &lt;len&gt;, i32 &lt;align&gt;)
4885 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4886 i64 &lt;len&gt;, i32 &lt;align&gt;)
4887</pre>
4888
4889<h5>Overview:</h5>
4890
4891<p>
4892The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4893location to the destination location. It is similar to the
Chris Lattnerdba16ea2008-01-06 19:51:52 +0000[diff] [blame]4894'<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4895</p>
4896
4897<p>
4898Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4899intrinsics do not return a value, and takes an extra alignment argument.
4900</p>
4901
4902<h5>Arguments:</h5>
4903
4904<p>
4905The first argument is a pointer to the destination, the second is a pointer to
4906the source. The third argument is an integer argument
4907specifying the number of bytes to copy, and the fourth argument is the alignment
4908of the source and destination locations.
4909</p>
4910
4911<p>
4912If the call to this intrinisic has an alignment value that is not 0 or 1, then
4913the caller guarantees that the source and destination pointers are aligned to
4914that boundary.
4915</p>
4916
4917<h5>Semantics:</h5>
4918
4919<p>
4920The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4921location to the destination location, which may overlap. It
4922copies "len" bytes of memory over. If the argument is known to be aligned to
4923some boundary, this can be specified as the fourth argument, otherwise it should
4924be set to 0 or 1.
4925</p>
4926</div>
4927
4928
4929<!-- _______________________________________________________________________ -->
4930<div class="doc_subsubsection">
4931 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4932</div>
4933
4934<div class="doc_text">
4935
4936<h5>Syntax:</h5>
4937<pre>
4938 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
4939 i32 &lt;len&gt;, i32 &lt;align&gt;)
4940 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
4941 i64 &lt;len&gt;, i32 &lt;align&gt;)
4942</pre>
4943
4944<h5>Overview:</h5>
4945
4946<p>
4947The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4948byte value.
4949</p>
4950
4951<p>
4952Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4953does not return a value, and takes an extra alignment argument.
4954</p>
4955
4956<h5>Arguments:</h5>
4957
4958<p>
4959The first argument is a pointer to the destination to fill, the second is the
4960byte value to fill it with, the third argument is an integer
4961argument specifying the number of bytes to fill, and the fourth argument is the
4962known alignment of destination location.
4963</p>
4964
4965<p>
4966If the call to this intrinisic has an alignment value that is not 0 or 1, then
4967the caller guarantees that the destination pointer is aligned to that boundary.
4968</p>
4969
4970<h5>Semantics:</h5>
4971
4972<p>
4973The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4974the
4975destination location. If the argument is known to be aligned to some boundary,
4976this can be specified as the fourth argument, otherwise it should be set to 0 or
49771.
4978</p>
4979</div>
4980
4981
4982<!-- _______________________________________________________________________ -->
4983<div class="doc_subsubsection">
4984 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4985</div>
4986
4987<div class="doc_text">
4988
4989<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4990<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]4991floating point or vector of floating point type. Not all targets support all
4992types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4993<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4994 declare float @llvm.sqrt.f32(float %Val)
4995 declare double @llvm.sqrt.f64(double %Val)
4996 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
4997 declare fp128 @llvm.sqrt.f128(fp128 %Val)
4998 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4999</pre>
5000
5001<h5>Overview:</h5>
5002
5003<p>
5004The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5005returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5006<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
Chris Lattnerf914a002008-01-29 07:00:44 +0000[diff] [blame]5007negative numbers other than -0.0 (which allows for better optimization, because
5008there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5009defined to return -0.0 like IEEE sqrt.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5010</p>
5011
5012<h5>Arguments:</h5>
5013
5014<p>
5015The argument and return value are floating point numbers of the same type.
5016</p>
5017
5018<h5>Semantics:</h5>
5019
5020<p>
5021This function returns the sqrt of the specified operand if it is a nonnegative
5022floating point number.
5023</p>
5024</div>
5025
5026<!-- _______________________________________________________________________ -->
5027<div class="doc_subsubsection">
5028 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5029</div>
5030
5031<div class="doc_text">
5032
5033<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5034<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]5035floating point or vector of floating point type. Not all targets support all
5036types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5037<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]5038 declare float @llvm.powi.f32(float %Val, i32 %power)
5039 declare double @llvm.powi.f64(double %Val, i32 %power)
5040 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5041 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5042 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5043</pre>
5044
5045<h5>Overview:</h5>
5046
5047<p>
5048The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5049specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5050multiplications is not defined. When a vector of floating point type is
5051used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5052</p>
5053
5054<h5>Arguments:</h5>
5055
5056<p>
5057The second argument is an integer power, and the first is a value to raise to
5058that power.
5059</p>
5060
5061<h5>Semantics:</h5>
5062
5063<p>
5064This function returns the first value raised to the second power with an
5065unspecified sequence of rounding operations.</p>
5066</div>
5067
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5068<!-- _______________________________________________________________________ -->
5069<div class="doc_subsubsection">
5070 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5071</div>
5072
5073<div class="doc_text">
5074
5075<h5>Syntax:</h5>
5076<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5077floating point or vector of floating point type. Not all targets support all
5078types however.
5079<pre>
5080 declare float @llvm.sin.f32(float %Val)
5081 declare double @llvm.sin.f64(double %Val)
5082 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5083 declare fp128 @llvm.sin.f128(fp128 %Val)
5084 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5085</pre>
5086
5087<h5>Overview:</h5>
5088
5089<p>
5090The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5091</p>
5092
5093<h5>Arguments:</h5>
5094
5095<p>
5096The argument and return value are floating point numbers of the same type.
5097</p>
5098
5099<h5>Semantics:</h5>
5100
5101<p>
5102This function returns the sine of the specified operand, returning the
5103same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5104conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5105</div>
5106
5107<!-- _______________________________________________________________________ -->
5108<div class="doc_subsubsection">
5109 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5110</div>
5111
5112<div class="doc_text">
5113
5114<h5>Syntax:</h5>
5115<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5116floating point or vector of floating point type. Not all targets support all
5117types however.
5118<pre>
5119 declare float @llvm.cos.f32(float %Val)
5120 declare double @llvm.cos.f64(double %Val)
5121 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5122 declare fp128 @llvm.cos.f128(fp128 %Val)
5123 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5124</pre>
5125
5126<h5>Overview:</h5>
5127
5128<p>
5129The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5130</p>
5131
5132<h5>Arguments:</h5>
5133
5134<p>
5135The argument and return value are floating point numbers of the same type.
5136</p>
5137
5138<h5>Semantics:</h5>
5139
5140<p>
5141This function returns the cosine of the specified operand, returning the
5142same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5143conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5144</div>
5145
5146<!-- _______________________________________________________________________ -->
5147<div class="doc_subsubsection">
5148 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5149</div>
5150
5151<div class="doc_text">
5152
5153<h5>Syntax:</h5>
5154<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5155floating point or vector of floating point type. Not all targets support all
5156types however.
5157<pre>
5158 declare float @llvm.pow.f32(float %Val, float %Power)
5159 declare double @llvm.pow.f64(double %Val, double %Power)
5160 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5161 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5162 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5163</pre>
5164
5165<h5>Overview:</h5>
5166
5167<p>
5168The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5169specified (positive or negative) power.
5170</p>
5171
5172<h5>Arguments:</h5>
5173
5174<p>
5175The second argument is a floating point power, and the first is a value to
5176raise to that power.
5177</p>
5178
5179<h5>Semantics:</h5>
5180
5181<p>
5182This function returns the first value raised to the second power,
5183returning the
5184same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]5185conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]5186</div>
5187
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5188
5189<!-- ======================================================================= -->
5190<div class="doc_subsection">
5191 <a name="int_manip">Bit Manipulation Intrinsics</a>
5192</div>
5193
5194<div class="doc_text">
5195<p>
5196LLVM provides intrinsics for a few important bit manipulation operations.
5197These allow efficient code generation for some algorithms.
5198</p>
5199
5200</div>
5201
5202<!-- _______________________________________________________________________ -->
5203<div class="doc_subsubsection">
5204 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5205</div>
5206
5207<div class="doc_text">
5208
5209<h5>Syntax:</h5>
5210<p>This is an overloaded intrinsic function. You can use bswap on any integer
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5211type that is an even number of bytes (i.e. BitWidth % 16 == 0).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5212<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5213 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
5214 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
5215 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5216</pre>
5217
5218<h5>Overview:</h5>
5219
5220<p>
5221The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5222values with an even number of bytes (positive multiple of 16 bits). These are
5223useful for performing operations on data that is not in the target's native
5224byte order.
5225</p>
5226
5227<h5>Semantics:</h5>
5228
5229<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5230The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5231and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5232intrinsic returns an i32 value that has the four bytes of the input i32
5233swapped, 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]5234i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5235<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5236additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5237</p>
5238
5239</div>
5240
5241<!-- _______________________________________________________________________ -->
5242<div class="doc_subsubsection">
5243 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5244</div>
5245
5246<div class="doc_text">
5247
5248<h5>Syntax:</h5>
5249<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
5250width. Not all targets support all bit widths however.
5251<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5252 declare i8 @llvm.ctpop.i8 (i8 &lt;src&gt;)
5253 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5254 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5255 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
5256 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5257</pre>
5258
5259<h5>Overview:</h5>
5260
5261<p>
5262The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5263value.
5264</p>
5265
5266<h5>Arguments:</h5>
5267
5268<p>
5269The only argument is the value to be counted. The argument may be of any
5270integer type. The return type must match the argument type.
5271</p>
5272
5273<h5>Semantics:</h5>
5274
5275<p>
5276The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5277</p>
5278</div>
5279
5280<!-- _______________________________________________________________________ -->
5281<div class="doc_subsubsection">
5282 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5283</div>
5284
5285<div class="doc_text">
5286
5287<h5>Syntax:</h5>
5288<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5289integer bit width. Not all targets support all bit widths however.
5290<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5291 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
5292 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5293 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5294 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
5295 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5296</pre>
5297
5298<h5>Overview:</h5>
5299
5300<p>
5301The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5302leading zeros in a variable.
5303</p>
5304
5305<h5>Arguments:</h5>
5306
5307<p>
5308The only argument is the value to be counted. The argument may be of any
5309integer type. The return type must match the argument type.
5310</p>
5311
5312<h5>Semantics:</h5>
5313
5314<p>
5315The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5316in a variable. If the src == 0 then the result is the size in bits of the type
5317of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5318</p>
5319</div>
5320
5321
5322
5323<!-- _______________________________________________________________________ -->
5324<div class="doc_subsubsection">
5325 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5326</div>
5327
5328<div class="doc_text">
5329
5330<h5>Syntax:</h5>
5331<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5332integer bit width. Not all targets support all bit widths however.
5333<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5334 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
5335 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5336 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5337 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
5338 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5339</pre>
5340
5341<h5>Overview:</h5>
5342
5343<p>
5344The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5345trailing zeros.
5346</p>
5347
5348<h5>Arguments:</h5>
5349
5350<p>
5351The only argument is the value to be counted. The argument may be of any
5352integer type. The return type must match the argument type.
5353</p>
5354
5355<h5>Semantics:</h5>
5356
5357<p>
5358The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5359in a variable. If the src == 0 then the result is the size in bits of the type
5360of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5361</p>
5362</div>
5363
5364<!-- _______________________________________________________________________ -->
5365<div class="doc_subsubsection">
5366 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5367</div>
5368
5369<div class="doc_text">
5370
5371<h5>Syntax:</h5>
5372<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
5373on any integer bit width.
5374<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5375 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5376 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5377</pre>
5378
5379<h5>Overview:</h5>
5380<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5381range of bits from an integer value and returns them in the same bit width as
5382the original value.</p>
5383
5384<h5>Arguments:</h5>
5385<p>The first argument, <tt>%val</tt> and the result may be integer types of
5386any bit width but they must have the same bit width. The second and third
5387arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5388
5389<h5>Semantics:</h5>
5390<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5391of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5392<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5393operates in forward mode.</p>
5394<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5395right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5396only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5397<ol>
5398 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5399 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5400 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5401 to determine the number of bits to retain.</li>
5402 <li>A mask of the retained bits is created by shifting a -1 value.</li>
5403 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
5404</ol>
5405<p>In reverse mode, a similar computation is made except that the bits are
5406returned in the reverse order. So, for example, if <tt>X</tt> has the value
5407<tt>i16 0x0ACF (101011001111)</tt> and we apply
5408<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5409<tt>i16 0x0026 (000000100110)</tt>.</p>
5410</div>
5411
5412<div class="doc_subsubsection">
5413 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5414</div>
5415
5416<div class="doc_text">
5417
5418<h5>Syntax:</h5>
5419<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
5420on any integer bit width.
5421<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]5422 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5423 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]5424</pre>
5425
5426<h5>Overview:</h5>
5427<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5428of bits in an integer value with another integer value. It returns the integer
5429with the replaced bits.</p>
5430
5431<h5>Arguments:</h5>
5432<p>The first argument, <tt>%val</tt> and the result may be integer types of
5433any bit width but they must have the same bit width. <tt>%val</tt> is the value
5434whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5435integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5436type since they specify only a bit index.</p>
5437
5438<h5>Semantics:</h5>
5439<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5440of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5441<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5442operates in forward mode.</p>
5443<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5444truncating it down to the size of the replacement area or zero extending it
5445up to that size.</p>
5446<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5447are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5448in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5449to the <tt>%hi</tt>th bit.
5450<p>In reverse mode, a similar computation is made except that the bits are
5451reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5452<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
5453<h5>Examples:</h5>
5454<pre>
5455 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
5456 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
5457 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
5458 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
5459 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
5460</pre>
5461</div>
5462
5463<!-- ======================================================================= -->
5464<div class="doc_subsection">
5465 <a name="int_debugger">Debugger Intrinsics</a>
5466</div>
5467
5468<div class="doc_text">
5469<p>
5470The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5471are described in the <a
5472href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5473Debugging</a> document.
5474</p>
5475</div>
5476
5477
5478<!-- ======================================================================= -->
5479<div class="doc_subsection">
5480 <a name="int_eh">Exception Handling Intrinsics</a>
5481</div>
5482
5483<div class="doc_text">
5484<p> The LLVM exception handling intrinsics (which all start with
5485<tt>llvm.eh.</tt> prefix), are described in the <a
5486href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5487Handling</a> document. </p>
5488</div>
5489
5490<!-- ======================================================================= -->
5491<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5492 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5493</div>
5494
5495<div class="doc_text">
5496<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5497 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5498 the <tt>nest</tt> attribute, from a function. The result is a callable
5499 function pointer lacking the nest parameter - the caller does not need
5500 to provide a value for it. Instead, the value to use is stored in
5501 advance in a "trampoline", a block of memory usually allocated
5502 on the stack, which also contains code to splice the nest value into the
5503 argument list. This is used to implement the GCC nested function address
5504 extension.
5505</p>
5506<p>
5507 For example, if the function is
5508 <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]5509 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]5510<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5511 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5512 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5513 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5514 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5515</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]5516 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5517 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5518</div>
5519
5520<!-- _______________________________________________________________________ -->
5521<div class="doc_subsubsection">
5522 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5523</div>
5524<div class="doc_text">
5525<h5>Syntax:</h5>
5526<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5527declare 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]5528</pre>
5529<h5>Overview:</h5>
5530<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5531 This fills the memory pointed to by <tt>tramp</tt> with code
5532 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5533</p>
5534<h5>Arguments:</h5>
5535<p>
5536 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5537 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5538 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]5539 intrinsic. Note that the size and the alignment are target-specific - LLVM
5540 currently provides no portable way of determining them, so a front-end that
5541 generates this intrinsic needs to have some target-specific knowledge.
5542 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]5543</p>
5544<h5>Semantics:</h5>
5545<p>
5546 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5547 dependent code, turning it into a function. A pointer to this function is
5548 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5549 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5550 before being called. The new function's signature is the same as that of
5551 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5552 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5553 of pointer type. Calling the new function is equivalent to calling
5554 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5555 missing <tt>nest</tt> argument. If, after calling
5556 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5557 modified, then the effect of any later call to the returned function pointer is
5558 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5559</p>
5560</div>
5561
5562<!-- ======================================================================= -->
5563<div class="doc_subsection">
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]5564 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
5565</div>
5566
5567<div class="doc_text">
5568<p>
5569 These intrinsic functions expand the "universal IR" of LLVM to represent
5570 hardware constructs for atomic operations and memory synchronization. This
5571 provides an interface to the hardware, not an interface to the programmer. It
5572 is aimed at a low enough level to allow any programming models or APIs which
5573 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
5574 hardware behavior. Just as hardware provides a "universal IR" for source
5575 languages, it also provides a starting point for developing a "universal"
5576 atomic operation and synchronization IR.
5577</p>
5578<p>
5579 These do <em>not</em> form an API such as high-level threading libraries,
5580 software transaction memory systems, atomic primitives, and intrinsic
5581 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
5582 application libraries. The hardware interface provided by LLVM should allow
5583 a clean implementation of all of these APIs and parallel programming models.
5584 No one model or paradigm should be selected above others unless the hardware
5585 itself ubiquitously does so.
5586
5587</p>
5588</div>
5589
5590<!-- _______________________________________________________________________ -->
5591<div class="doc_subsubsection">
5592 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
5593</div>
5594<div class="doc_text">
5595<h5>Syntax:</h5>
5596<pre>
5597declare void @llvm.memory.barrier( i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt;,
5598i1 &lt;device&gt; )
5599
5600</pre>
5601<h5>Overview:</h5>
5602<p>
5603 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
5604 specific pairs of memory access types.
5605</p>
5606<h5>Arguments:</h5>
5607<p>
5608 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
5609 The first four arguments enables a specific barrier as listed below. The fith
5610 argument specifies that the barrier applies to io or device or uncached memory.
5611
5612</p>
5613 <ul>
5614 <li><tt>ll</tt>: load-load barrier</li>
5615 <li><tt>ls</tt>: load-store barrier</li>
5616 <li><tt>sl</tt>: store-load barrier</li>
5617 <li><tt>ss</tt>: store-store barrier</li>
5618 <li><tt>device</tt>: barrier applies to device and uncached memory also.
5619 </ul>
5620<h5>Semantics:</h5>
5621<p>
5622 This intrinsic causes the system to enforce some ordering constraints upon
5623 the loads and stores of the program. This barrier does not indicate
5624 <em>when</em> any events will occur, it only enforces an <em>order</em> in
5625 which they occur. For any of the specified pairs of load and store operations
5626 (f.ex. load-load, or store-load), all of the first operations preceding the
5627 barrier will complete before any of the second operations succeeding the
5628 barrier begin. Specifically the semantics for each pairing is as follows:
5629</p>
5630 <ul>
5631 <li><tt>ll</tt>: All loads before the barrier must complete before any load
5632 after the barrier begins.</li>
5633
5634 <li><tt>ls</tt>: All loads before the barrier must complete before any
5635 store after the barrier begins.</li>
5636 <li><tt>ss</tt>: All stores before the barrier must complete before any
5637 store after the barrier begins.</li>
5638 <li><tt>sl</tt>: All stores before the barrier must complete before any
5639 load after the barrier begins.</li>
5640 </ul>
5641<p>
5642 These semantics are applied with a logical "and" behavior when more than one
5643 is enabled in a single memory barrier intrinsic.
5644</p>
5645<p>
5646 Backends may implement stronger barriers than those requested when they do not
5647 support as fine grained a barrier as requested. Some architectures do not
5648 need all types of barriers and on such architectures, these become noops.
5649</p>
5650<h5>Example:</h5>
5651<pre>
5652%ptr = malloc i32
5653 store i32 4, %ptr
5654
5655%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
5656 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
5657 <i>; guarantee the above finishes</i>
5658 store i32 8, %ptr <i>; before this begins</i>
5659</pre>
5660</div>
5661
Andrew Lenharthe44f3902008-02-21 06:45:13 +0000[diff] [blame]5662<!-- _______________________________________________________________________ -->
5663<div class="doc_subsubsection">
5664 <a name="int_atomic_lcs">'<tt>llvm.atomic.lcs.*</tt>' Intrinsic</a>
5665</div>
5666<div class="doc_text">
5667<h5>Syntax:</h5>
5668<p>
5669 This is an overloaded intrinsic. You can use <tt>llvm.atomic.lcs</tt> on any
5670 integer bit width. Not all targets support all bit widths however.</p>
5671
5672<pre>
5673declare i8 @llvm.atomic.lcs.i8( i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt; )
5674declare i16 @llvm.atomic.lcs.i16( i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt; )
5675declare i32 @llvm.atomic.lcs.i32( i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt; )
5676declare i64 @llvm.atomic.lcs.i64( i64* &lt;ptr&gt;, i64 &lt;cmp&gt;, i64 &lt;val&gt; )
5677
5678</pre>
5679<h5>Overview:</h5>
5680<p>
5681 This loads a value in memory and compares it to a given value. If they are
5682 equal, it stores a new value into the memory.
5683</p>
5684<h5>Arguments:</h5>
5685<p>
5686 The <tt>llvm.atomic.lcs</tt> intrinsic takes three arguments. The result as
5687 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
5688 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
5689 this integer type. While any bit width integer may be used, targets may only
5690 lower representations they support in hardware.
5691
5692</p>
5693<h5>Semantics:</h5>
5694<p>
5695 This entire intrinsic must be executed atomically. It first loads the value
5696 in memory pointed to by <tt>ptr</tt> and compares it with the value
5697 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
5698 loaded value is yielded in all cases. This provides the equivalent of an
5699 atomic compare-and-swap operation within the SSA framework.
5700</p>
5701<h5>Examples:</h5>
5702
5703<pre>
5704%ptr = malloc i32
5705 store i32 4, %ptr
5706
5707%val1 = add i32 4, 4
5708%result1 = call i32 @llvm.atomic.lcs.i32( i32* %ptr, i32 4, %val1 )
5709 <i>; yields {i32}:result1 = 4</i>
5710%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5711%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5712
5713%val2 = add i32 1, 1
5714%result2 = call i32 @llvm.atomic.lcs.i32( i32* %ptr, i32 5, %val2 )
5715 <i>; yields {i32}:result2 = 8</i>
5716%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
5717
5718%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
5719</pre>
5720</div>
5721
5722<!-- _______________________________________________________________________ -->
5723<div class="doc_subsubsection">
5724 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
5725</div>
5726<div class="doc_text">
5727<h5>Syntax:</h5>
5728
5729<p>
5730 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
5731 integer bit width. Not all targets support all bit widths however.</p>
5732<pre>
5733declare i8 @llvm.atomic.swap.i8( i8* &lt;ptr&gt;, i8 &lt;val&gt; )
5734declare i16 @llvm.atomic.swap.i16( i16* &lt;ptr&gt;, i16 &lt;val&gt; )
5735declare i32 @llvm.atomic.swap.i32( i32* &lt;ptr&gt;, i32 &lt;val&gt; )
5736declare i64 @llvm.atomic.swap.i64( i64* &lt;ptr&gt;, i64 &lt;val&gt; )
5737
5738</pre>
5739<h5>Overview:</h5>
5740<p>
5741 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
5742 the value from memory. It then stores the value in <tt>val</tt> in the memory
5743 at <tt>ptr</tt>.
5744</p>
5745<h5>Arguments:</h5>
5746
5747<p>
5748 The <tt>llvm.atomic.ls</tt> intrinsic takes two arguments. Both the
5749 <tt>val</tt> argument and the result must be integers of the same bit width.
5750 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
5751 integer type. The targets may only lower integer representations they
5752 support.
5753</p>
5754<h5>Semantics:</h5>
5755<p>
5756 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
5757 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
5758 equivalent of an atomic swap operation within the SSA framework.
5759
5760</p>
5761<h5>Examples:</h5>
5762<pre>
5763%ptr = malloc i32
5764 store i32 4, %ptr
5765
5766%val1 = add i32 4, 4
5767%result1 = call i32 @llvm.atomic.swap.i32( i32* %ptr, i32 %val1 )
5768 <i>; yields {i32}:result1 = 4</i>
5769%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
5770%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
5771
5772%val2 = add i32 1, 1
5773%result2 = call i32 @llvm.atomic.swap.i32( i32* %ptr, i32 %val2 )
5774 <i>; yields {i32}:result2 = 8</i>
5775
5776%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
5777%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
5778</pre>
5779</div>
5780
5781<!-- _______________________________________________________________________ -->
5782<div class="doc_subsubsection">
5783 <a name="int_atomic_las">'<tt>llvm.atomic.las.*</tt>' Intrinsic</a>
5784
5785</div>
5786<div class="doc_text">
5787<h5>Syntax:</h5>
5788<p>
5789 This is an overloaded intrinsic. You can use <tt>llvm.atomic.las</tt> on any
5790 integer bit width. Not all targets support all bit widths however.</p>
5791<pre>
5792declare i8 @llvm.atomic.las.i8.( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
5793declare i16 @llvm.atomic.las.i16.( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
5794declare i32 @llvm.atomic.las.i32.( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
5795declare i64 @llvm.atomic.las.i64.( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
5796
5797</pre>
5798<h5>Overview:</h5>
5799<p>
5800 This intrinsic adds <tt>delta</tt> to the value stored in memory at
5801 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
5802</p>
5803<h5>Arguments:</h5>
5804<p>
5805
5806 The intrinsic takes two arguments, the first a pointer to an integer value
5807 and the second an integer value. The result is also an integer value. These
5808 integer types can have any bit width, but they must all have the same bit
5809 width. The targets may only lower integer representations they support.
5810</p>
5811<h5>Semantics:</h5>
5812<p>
5813 This intrinsic does a series of operations atomically. It first loads the
5814 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
5815 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
5816</p>
5817
5818<h5>Examples:</h5>
5819<pre>
5820%ptr = malloc i32
5821 store i32 4, %ptr
5822%result1 = call i32 @llvm.atomic.las.i32( i32* %ptr, i32 4 )
5823 <i>; yields {i32}:result1 = 4</i>
5824%result2 = call i32 @llvm.atomic.las.i32( i32* %ptr, i32 2 )
5825 <i>; yields {i32}:result2 = 8</i>
5826%result3 = call i32 @llvm.atomic.las.i32( i32* %ptr, i32 5 )
5827 <i>; yields {i32}:result3 = 10</i>
5828%memval = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
5829</pre>
5830</div>
5831
Andrew Lenharth785610d2008-02-16 01:24:58 +0000[diff] [blame]5832
5833<!-- ======================================================================= -->
5834<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5835 <a name="int_general">General Intrinsics</a>
5836</div>
5837
5838<div class="doc_text">
5839<p> This class of intrinsics is designed to be generic and has
5840no specific purpose. </p>
5841</div>
5842
5843<!-- _______________________________________________________________________ -->
5844<div class="doc_subsubsection">
5845 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
5846</div>
5847
5848<div class="doc_text">
5849
5850<h5>Syntax:</h5>
5851<pre>
5852 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5853</pre>
5854
5855<h5>Overview:</h5>
5856
5857<p>
5858The '<tt>llvm.var.annotation</tt>' intrinsic
5859</p>
5860
5861<h5>Arguments:</h5>
5862
5863<p>
5864The first argument is a pointer to a value, the second is a pointer to a
5865global string, the third is a pointer to a global string which is the source
5866file name, and the last argument is the line number.
5867</p>
5868
5869<h5>Semantics:</h5>
5870
5871<p>
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]5872This intrinsic allows annotation of local variables with arbitrary strings.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5873This can be useful for special purpose optimizations that want to look for these
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]5874annotations. These have no other defined use, they are ignored by code
5875generation and optimization.
5876</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5877</div>
5878
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5879<!-- _______________________________________________________________________ -->
5880<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]5881 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5882</div>
5883
5884<div class="doc_text">
5885
5886<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5887<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
5888any integer bit width.
5889</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5890<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]5891 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5892 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5893 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5894 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5895 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]5896</pre>
5897
5898<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5899
5900<p>
5901The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5902</p>
5903
5904<h5>Arguments:</h5>
5905
5906<p>
5907The first argument is an integer value (result of some expression),
5908the second is a pointer to a global string, the third is a pointer to a global
5909string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5910It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5911</p>
5912
5913<h5>Semantics:</h5>
5914
5915<p>
5916This intrinsic allows annotations to be put on arbitrary expressions
5917with arbitrary strings. This can be useful for special purpose optimizations
5918that want to look for these annotations. These have no other defined use, they
5919are ignored by code generation and optimization.
5920</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5921
Anton Korobeynikove6e764f2008-01-15 22:31:34 +0000[diff] [blame]5922<!-- _______________________________________________________________________ -->
5923<div class="doc_subsubsection">
5924 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
5925</div>
5926
5927<div class="doc_text">
5928
5929<h5>Syntax:</h5>
5930<pre>
5931 declare void @llvm.trap()
5932</pre>
5933
5934<h5>Overview:</h5>
5935
5936<p>
5937The '<tt>llvm.trap</tt>' intrinsic
5938</p>
5939
5940<h5>Arguments:</h5>
5941
5942<p>
5943None
5944</p>
5945
5946<h5>Semantics:</h5>
5947
5948<p>
5949This intrinsics is lowered to the target dependent trap instruction. If the
5950target does not have a trap instruction, this intrinsic will be lowered to the
5951call of the abort() function.
5952</p>
5953</div>
5954
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5955<!-- *********************************************************************** -->
5956<hr>
5957<address>
5958 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
5959 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
5960 <a href="http://validator.w3.org/check/referer"><img
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]5961 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5962
5963 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
5964 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
5965 Last modified: $Date$
5966</address>
Chris Lattner08497ce2008-01-04 04:33:49 +0000[diff] [blame]5967
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5968</body>
5969</html>