Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 78246cae8f37740879b3b099b388f0595106db6b / . / docs / LangRef.html
blob: 9c4bbf094d3fd17f0227dbfd0c0eddb62e90e65c [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>
29 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
30 <li><a href="#datalayout">Data Layout</a></li>
31 </ol>
32 </li>
33 <li><a href="#typesystem">Type System</a>
34 <ol>
35 <li><a href="#t_primitive">Primitive Types</a>
36 <ol>
37 <li><a href="#t_classifications">Type Classifications</a></li>
38 </ol>
39 </li>
40 <li><a href="#t_derived">Derived Types</a>
41 <ol>
42 <li><a href="#t_array">Array Type</a></li>
43 <li><a href="#t_function">Function Type</a></li>
44 <li><a href="#t_pointer">Pointer Type</a></li>
45 <li><a href="#t_struct">Structure Type</a></li>
46 <li><a href="#t_pstruct">Packed Structure Type</a></li>
47 <li><a href="#t_vector">Vector Type</a></li>
48 <li><a href="#t_opaque">Opaque Type</a></li>
49 </ol>
50 </li>
51 </ol>
52 </li>
53 <li><a href="#constants">Constants</a>
54 <ol>
55 <li><a href="#simpleconstants">Simple Constants</a>
56 <li><a href="#aggregateconstants">Aggregate Constants</a>
57 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
58 <li><a href="#undefvalues">Undefined Values</a>
59 <li><a href="#constantexprs">Constant Expressions</a>
60 </ol>
61 </li>
62 <li><a href="#othervalues">Other Values</a>
63 <ol>
64 <li><a href="#inlineasm">Inline Assembler Expressions</a>
65 </ol>
66 </li>
67 <li><a href="#instref">Instruction Reference</a>
68 <ol>
69 <li><a href="#terminators">Terminator Instructions</a>
70 <ol>
71 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
72 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
73 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
74 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
75 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
76 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
77 </ol>
78 </li>
79 <li><a href="#binaryops">Binary Operations</a>
80 <ol>
81 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
82 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
83 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
84 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
85 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
86 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
87 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
88 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
89 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
90 </ol>
91 </li>
92 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
93 <ol>
94 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
95 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
96 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
97 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
98 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
99 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
100 </ol>
101 </li>
102 <li><a href="#vectorops">Vector Operations</a>
103 <ol>
104 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
105 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
106 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
107 </ol>
108 </li>
109 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
110 <ol>
111 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
112 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
113 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
114 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
115 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
116 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
117 </ol>
118 </li>
119 <li><a href="#convertops">Conversion Operations</a>
120 <ol>
121 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
122 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
123 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
124 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
125 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
126 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
127 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
128 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
129 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
130 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
131 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
132 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
133 </ol>
134 <li><a href="#otherops">Other Operations</a>
135 <ol>
136 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
137 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
138 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
139 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
140 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
141 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
142 </ol>
143 </li>
144 </ol>
145 </li>
146 <li><a href="#intrinsics">Intrinsic Functions</a>
147 <ol>
148 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
149 <ol>
150 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
151 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
152 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
153 </ol>
154 </li>
155 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
156 <ol>
157 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
158 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
159 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
160 </ol>
161 </li>
162 <li><a href="#int_codegen">Code Generator Intrinsics</a>
163 <ol>
164 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
165 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
166 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
167 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
168 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
169 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
170 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
171 </ol>
172 </li>
173 <li><a href="#int_libc">Standard C Library Intrinsics</a>
174 <ol>
175 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
176 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
177 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
178 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
179 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]180 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
181 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
182 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]183 </ol>
184 </li>
185 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
186 <ol>
187 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
188 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
189 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
190 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
191 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
192 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
193 </ol>
194 </li>
195 <li><a href="#int_debugger">Debugger intrinsics</a></li>
196 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]197 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]198 <ol>
199 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]200 </ol>
201 </li>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]202 <li><a href="#int_general">General intrinsics</a>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]203 <ol>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]204 <li><a href="#int_var_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]205 <tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Reid Spencerb043f672007-07-20 19:59:11 +0000[diff] [blame]206 </ol>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]207 <ol>
208 <li><a href="#int_annotation">
Tanya Lattner51369f32007-09-22 00:01:26 +0000[diff] [blame]209 <tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]210 </ol>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]211 </li>
212 </ol>
213 </li>
214</ol>
215
216<div class="doc_author">
217 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
218 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
219</div>
220
221<!-- *********************************************************************** -->
222<div class="doc_section"> <a name="abstract">Abstract </a></div>
223<!-- *********************************************************************** -->
224
225<div class="doc_text">
226<p>This document is a reference manual for the LLVM assembly language.
227LLVM is an SSA based representation that provides type safety,
228low-level operations, flexibility, and the capability of representing
229'all' high-level languages cleanly. It is the common code
230representation used throughout all phases of the LLVM compilation
231strategy.</p>
232</div>
233
234<!-- *********************************************************************** -->
235<div class="doc_section"> <a name="introduction">Introduction</a> </div>
236<!-- *********************************************************************** -->
237
238<div class="doc_text">
239
240<p>The LLVM code representation is designed to be used in three
241different forms: as an in-memory compiler IR, as an on-disk bitcode
242representation (suitable for fast loading by a Just-In-Time compiler),
243and as a human readable assembly language representation. This allows
244LLVM to provide a powerful intermediate representation for efficient
245compiler transformations and analysis, while providing a natural means
246to debug and visualize the transformations. The three different forms
247of LLVM are all equivalent. This document describes the human readable
248representation and notation.</p>
249
250<p>The LLVM representation aims to be light-weight and low-level
251while being expressive, typed, and extensible at the same time. It
252aims to be a "universal IR" of sorts, by being at a low enough level
253that high-level ideas may be cleanly mapped to it (similar to how
254microprocessors are "universal IR's", allowing many source languages to
255be mapped to them). By providing type information, LLVM can be used as
256the target of optimizations: for example, through pointer analysis, it
257can be proven that a C automatic variable is never accessed outside of
258the current function... allowing it to be promoted to a simple SSA
259value instead of a memory location.</p>
260
261</div>
262
263<!-- _______________________________________________________________________ -->
264<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
265
266<div class="doc_text">
267
268<p>It is important to note that this document describes 'well formed'
269LLVM assembly language. There is a difference between what the parser
270accepts and what is considered 'well formed'. For example, the
271following instruction is syntactically okay, but not well formed:</p>
272
273<div class="doc_code">
274<pre>
275%x = <a href="#i_add">add</a> i32 1, %x
276</pre>
277</div>
278
279<p>...because the definition of <tt>%x</tt> does not dominate all of
280its uses. The LLVM infrastructure provides a verification pass that may
281be used to verify that an LLVM module is well formed. This pass is
282automatically run by the parser after parsing input assembly and by
283the optimizer before it outputs bitcode. The violations pointed out
284by the verifier pass indicate bugs in transformation passes or input to
285the parser.</p>
286</div>
287
Chris Lattnera83fdc02007-10-03 17:34:29 +0000[diff] [blame]288<!-- Describe the typesetting conventions here. -->
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]289
290<!-- *********************************************************************** -->
291<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
292<!-- *********************************************************************** -->
293
294<div class="doc_text">
295
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]296 <p>LLVM identifiers come in two basic types: global and local. Global
297 identifiers (functions, global variables) begin with the @ character. Local
298 identifiers (register names, types) begin with the % character. Additionally,
299 there are three different formats for identifiers, for different purposes:
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]300
301<ol>
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]302 <li>Named values are represented as a string of characters with their prefix.
303 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
304 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]305 Identifiers which require other characters in their names can be surrounded
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]306 with quotes. In this way, anything except a <tt>&quot;</tt> character can
307 be used in a named value.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]308
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]309 <li>Unnamed values are represented as an unsigned numeric value with their
310 prefix. For example, %12, @2, %44.</li>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]311
312 <li>Constants, which are described in a <a href="#constants">section about
313 constants</a>, below.</li>
314</ol>
315
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]316<p>LLVM requires that values start with a prefix for two reasons: Compilers
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]317don't need to worry about name clashes with reserved words, and the set of
318reserved words may be expanded in the future without penalty. Additionally,
319unnamed identifiers allow a compiler to quickly come up with a temporary
320variable without having to avoid symbol table conflicts.</p>
321
322<p>Reserved words in LLVM are very similar to reserved words in other
323languages. There are keywords for different opcodes
324('<tt><a href="#i_add">add</a></tt>',
325 '<tt><a href="#i_bitcast">bitcast</a></tt>',
326 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
327href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
328and others. These reserved words cannot conflict with variable names, because
Reid Spencerc8245b02007-08-07 14:34:28 +0000[diff] [blame]329none of them start with a prefix character ('%' or '@').</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]330
331<p>Here is an example of LLVM code to multiply the integer variable
332'<tt>%X</tt>' by 8:</p>
333
334<p>The easy way:</p>
335
336<div class="doc_code">
337<pre>
338%result = <a href="#i_mul">mul</a> i32 %X, 8
339</pre>
340</div>
341
342<p>After strength reduction:</p>
343
344<div class="doc_code">
345<pre>
346%result = <a href="#i_shl">shl</a> i32 %X, i8 3
347</pre>
348</div>
349
350<p>And the hard way:</p>
351
352<div class="doc_code">
353<pre>
354<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
355<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
356%result = <a href="#i_add">add</a> i32 %1, %1
357</pre>
358</div>
359
360<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
361important lexical features of LLVM:</p>
362
363<ol>
364
365 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
366 line.</li>
367
368 <li>Unnamed temporaries are created when the result of a computation is not
369 assigned to a named value.</li>
370
371 <li>Unnamed temporaries are numbered sequentially</li>
372
373</ol>
374
375<p>...and it also shows a convention that we follow in this document. When
376demonstrating instructions, we will follow an instruction with a comment that
377defines the type and name of value produced. Comments are shown in italic
378text.</p>
379
380</div>
381
382<!-- *********************************************************************** -->
383<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
384<!-- *********************************************************************** -->
385
386<!-- ======================================================================= -->
387<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
388</div>
389
390<div class="doc_text">
391
392<p>LLVM programs are composed of "Module"s, each of which is a
393translation unit of the input programs. Each module consists of
394functions, global variables, and symbol table entries. Modules may be
395combined together with the LLVM linker, which merges function (and
396global variable) definitions, resolves forward declarations, and merges
397symbol table entries. Here is an example of the "hello world" module:</p>
398
399<div class="doc_code">
400<pre><i>; Declare the string constant as a global constant...</i>
401<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
402 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
403
404<i>; External declaration of the puts function</i>
405<a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
406
407<i>; Definition of main function</i>
408define i32 @main() { <i>; i32()* </i>
409 <i>; Convert [13x i8 ]* to i8 *...</i>
410 %cast210 = <a
411 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
412
413 <i>; Call puts function to write out the string to stdout...</i>
414 <a
415 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
416 <a
417 href="#i_ret">ret</a> i32 0<br>}<br>
418</pre>
419</div>
420
421<p>This example is made up of a <a href="#globalvars">global variable</a>
422named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
423function, and a <a href="#functionstructure">function definition</a>
424for "<tt>main</tt>".</p>
425
426<p>In general, a module is made up of a list of global values,
427where both functions and global variables are global values. Global values are
428represented by a pointer to a memory location (in this case, a pointer to an
429array of char, and a pointer to a function), and have one of the following <a
430href="#linkage">linkage types</a>.</p>
431
432</div>
433
434<!-- ======================================================================= -->
435<div class="doc_subsection">
436 <a name="linkage">Linkage Types</a>
437</div>
438
439<div class="doc_text">
440
441<p>
442All Global Variables and Functions have one of the following types of linkage:
443</p>
444
445<dl>
446
447 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
448
449 <dd>Global values with internal linkage are only directly accessible by
450 objects in the current module. In particular, linking code into a module with
451 an internal global value may cause the internal to be renamed as necessary to
452 avoid collisions. Because the symbol is internal to the module, all
453 references can be updated. This corresponds to the notion of the
454 '<tt>static</tt>' keyword in C.
455 </dd>
456
457 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
458
459 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
460 the same name when linkage occurs. This is typically used to implement
461 inline functions, templates, or other code which must be generated in each
462 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
463 allowed to be discarded.
464 </dd>
465
466 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
467
468 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
469 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
470 used for globals that may be emitted in multiple translation units, but that
471 are not guaranteed to be emitted into every translation unit that uses them.
472 One example of this are common globals in C, such as "<tt>int X;</tt>" at
473 global scope.
474 </dd>
475
476 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
477
478 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
479 pointer to array type. When two global variables with appending linkage are
480 linked together, the two global arrays are appended together. This is the
481 LLVM, typesafe, equivalent of having the system linker append together
482 "sections" with identical names when .o files are linked.
483 </dd>
484
485 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
486 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
487 until linked, if not linked, the symbol becomes null instead of being an
488 undefined reference.
489 </dd>
490
491 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
492
493 <dd>If none of the above identifiers are used, the global is externally
494 visible, meaning that it participates in linkage and can be used to resolve
495 external symbol references.
496 </dd>
497</dl>
498
499 <p>
500 The next two types of linkage are targeted for Microsoft Windows platform
501 only. They are designed to support importing (exporting) symbols from (to)
502 DLLs.
503 </p>
504
505 <dl>
506 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
507
508 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
509 or variable via a global pointer to a pointer that is set up by the DLL
510 exporting the symbol. On Microsoft Windows targets, the pointer name is
511 formed by combining <code>_imp__</code> and the function or variable name.
512 </dd>
513
514 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
515
516 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
517 pointer to a pointer in a DLL, so that it can be referenced with the
518 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
519 name is formed by combining <code>_imp__</code> and the function or variable
520 name.
521 </dd>
522
523</dl>
524
525<p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
526variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
527variable and was linked with this one, one of the two would be renamed,
528preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
529external (i.e., lacking any linkage declarations), they are accessible
530outside of the current module.</p>
531<p>It is illegal for a function <i>declaration</i>
532to have any linkage type other than "externally visible", <tt>dllimport</tt>,
533or <tt>extern_weak</tt>.</p>
534<p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
535linkages.
536</div>
537
538<!-- ======================================================================= -->
539<div class="doc_subsection">
540 <a name="callingconv">Calling Conventions</a>
541</div>
542
543<div class="doc_text">
544
545<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
546and <a href="#i_invoke">invokes</a> can all have an optional calling convention
547specified for the call. The calling convention of any pair of dynamic
548caller/callee must match, or the behavior of the program is undefined. The
549following calling conventions are supported by LLVM, and more may be added in
550the future:</p>
551
552<dl>
553 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
554
555 <dd>This calling convention (the default if no other calling convention is
556 specified) matches the target C calling conventions. This calling convention
557 supports varargs function calls and tolerates some mismatch in the declared
558 prototype and implemented declaration of the function (as does normal C).
559 </dd>
560
561 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
562
563 <dd>This calling convention attempts to make calls as fast as possible
564 (e.g. by passing things in registers). This calling convention allows the
565 target to use whatever tricks it wants to produce fast code for the target,
566 without having to conform to an externally specified ABI. Implementations of
567 this convention should allow arbitrary tail call optimization to be supported.
568 This calling convention does not support varargs and requires the prototype of
569 all callees to exactly match the prototype of the function definition.
570 </dd>
571
572 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
573
574 <dd>This calling convention attempts to make code in the caller as efficient
575 as possible under the assumption that the call is not commonly executed. As
576 such, these calls often preserve all registers so that the call does not break
577 any live ranges in the caller side. This calling convention does not support
578 varargs and requires the prototype of all callees to exactly match the
579 prototype of the function definition.
580 </dd>
581
582 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
583
584 <dd>Any calling convention may be specified by number, allowing
585 target-specific calling conventions to be used. Target specific calling
586 conventions start at 64.
587 </dd>
588</dl>
589
590<p>More calling conventions can be added/defined on an as-needed basis, to
591support pascal conventions or any other well-known target-independent
592convention.</p>
593
594</div>
595
596<!-- ======================================================================= -->
597<div class="doc_subsection">
598 <a name="visibility">Visibility Styles</a>
599</div>
600
601<div class="doc_text">
602
603<p>
604All Global Variables and Functions have one of the following visibility styles:
605</p>
606
607<dl>
608 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
609
610 <dd>On ELF, default visibility means that the declaration is visible to other
611 modules and, in shared libraries, means that the declared entity may be
612 overridden. On Darwin, default visibility means that the declaration is
613 visible to other modules. Default visibility corresponds to "external
614 linkage" in the language.
615 </dd>
616
617 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
618
619 <dd>Two declarations of an object with hidden visibility refer to the same
620 object if they are in the same shared object. Usually, hidden visibility
621 indicates that the symbol will not be placed into the dynamic symbol table,
622 so no other module (executable or shared library) can reference it
623 directly.
624 </dd>
625
626 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
627
628 <dd>On ELF, protected visibility indicates that the symbol will be placed in
629 the dynamic symbol table, but that references within the defining module will
630 bind to the local symbol. That is, the symbol cannot be overridden by another
631 module.
632 </dd>
633</dl>
634
635</div>
636
637<!-- ======================================================================= -->
638<div class="doc_subsection">
639 <a name="globalvars">Global Variables</a>
640</div>
641
642<div class="doc_text">
643
644<p>Global variables define regions of memory allocated at compilation time
645instead of run-time. Global variables may optionally be initialized, may have
646an explicit section to be placed in, and may have an optional explicit alignment
647specified. A variable may be defined as "thread_local", which means that it
648will not be shared by threads (each thread will have a separated copy of the
649variable). A variable may be defined as a global "constant," which indicates
650that the contents of the variable will <b>never</b> be modified (enabling better
651optimization, allowing the global data to be placed in the read-only section of
652an executable, etc). Note that variables that need runtime initialization
653cannot be marked "constant" as there is a store to the variable.</p>
654
655<p>
656LLVM explicitly allows <em>declarations</em> of global variables to be marked
657constant, even if the final definition of the global is not. This capability
658can be used to enable slightly better optimization of the program, but requires
659the language definition to guarantee that optimizations based on the
660'constantness' are valid for the translation units that do not include the
661definition.
662</p>
663
664<p>As SSA values, global variables define pointer values that are in
665scope (i.e. they dominate) all basic blocks in the program. Global
666variables always define a pointer to their "content" type because they
667describe a region of memory, and all memory objects in LLVM are
668accessed through pointers.</p>
669
670<p>LLVM allows an explicit section to be specified for globals. If the target
671supports it, it will emit globals to the section specified.</p>
672
673<p>An explicit alignment may be specified for a global. If not present, or if
674the alignment is set to zero, the alignment of the global is set by the target
675to whatever it feels convenient. If an explicit alignment is specified, the
676global is forced to have at least that much alignment. All alignments must be
677a power of 2.</p>
678
679<p>For example, the following defines a global with an initializer, section,
680 and alignment:</p>
681
682<div class="doc_code">
683<pre>
684@G = constant float 1.0, section "foo", align 4
685</pre>
686</div>
687
688</div>
689
690
691<!-- ======================================================================= -->
692<div class="doc_subsection">
693 <a name="functionstructure">Functions</a>
694</div>
695
696<div class="doc_text">
697
698<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
699an optional <a href="#linkage">linkage type</a>, an optional
700<a href="#visibility">visibility style</a>, an optional
701<a href="#callingconv">calling convention</a>, a return type, an optional
702<a href="#paramattrs">parameter attribute</a> for the return type, a function
703name, a (possibly empty) argument list (each with optional
704<a href="#paramattrs">parameter attributes</a>), an optional section, an
705optional alignment, an opening curly brace, a list of basic blocks, and a
706closing curly brace.
707
708LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
709optional <a href="#linkage">linkage type</a>, an optional
710<a href="#visibility">visibility style</a>, an optional
711<a href="#callingconv">calling convention</a>, a return type, an optional
712<a href="#paramattrs">parameter attribute</a> for the return type, a function
713name, a possibly empty list of arguments, and an optional alignment.</p>
714
715<p>A function definition contains a list of basic blocks, forming the CFG for
716the function. Each basic block may optionally start with a label (giving the
717basic block a symbol table entry), contains a list of instructions, and ends
718with a <a href="#terminators">terminator</a> instruction (such as a branch or
719function return).</p>
720
721<p>The first basic block in a function is special in two ways: it is immediately
722executed on entrance to the function, and it is not allowed to have predecessor
723basic blocks (i.e. there can not be any branches to the entry block of a
724function). Because the block can have no predecessors, it also cannot have any
725<a href="#i_phi">PHI nodes</a>.</p>
726
727<p>LLVM allows an explicit section to be specified for functions. If the target
728supports it, it will emit functions to the section specified.</p>
729
730<p>An explicit alignment may be specified for a function. If not present, or if
731the alignment is set to zero, the alignment of the function is set by the target
732to whatever it feels convenient. If an explicit alignment is specified, the
733function is forced to have at least that much alignment. All alignments must be
734a power of 2.</p>
735
736</div>
737
738
739<!-- ======================================================================= -->
740<div class="doc_subsection">
741 <a name="aliasstructure">Aliases</a>
742</div>
743<div class="doc_text">
744 <p>Aliases act as "second name" for the aliasee value (which can be either
745 function or global variable or bitcast of global value). Aliases may have an
746 optional <a href="#linkage">linkage type</a>, and an
747 optional <a href="#visibility">visibility style</a>.</p>
748
749 <h5>Syntax:</h5>
750
751<div class="doc_code">
752<pre>
753@&lt;Name&gt; = [Linkage] [Visibility] alias &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
754</pre>
755</div>
756
757</div>
758
759
760
761<!-- ======================================================================= -->
762<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
763<div class="doc_text">
764 <p>The return type and each parameter of a function type may have a set of
765 <i>parameter attributes</i> associated with them. Parameter attributes are
766 used to communicate additional information about the result or parameters of
767 a function. Parameter attributes are considered to be part of the function
768 type so two functions types that differ only by the parameter attributes
769 are different function types.</p>
770
771 <p>Parameter attributes are simple keywords that follow the type specified. If
772 multiple parameter attributes are needed, they are space separated. For
773 example:</p>
774
775<div class="doc_code">
776<pre>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]777%someFunc = i16 (i8 signext %someParam) zeroext
778%someFunc = i16 (i8 zeroext %someParam) zeroext
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]779</pre>
780</div>
781
782 <p>Note that the two function types above are unique because the parameter has
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]783 a different attribute (<tt>signext</tt> in the first one, <tt>zeroext</tt> in
784 the second). Also note that the attribute for the function result
785 (<tt>zeroext</tt>) comes immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]786
787 <p>Currently, only the following parameter attributes are defined:</p>
788 <dl>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]789 <dt><tt>zeroext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]790 <dd>This indicates that the parameter should be zero extended just before
791 a call to this function.</dd>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]792 <dt><tt>signext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]793 <dd>This indicates that the parameter should be sign extended just before
794 a call to this function.</dd>
795 <dt><tt>inreg</tt></dt>
796 <dd>This indicates that the parameter should be placed in register (if
797 possible) during assembling function call. Support for this attribute is
798 target-specific</dd>
799 <dt><tt>sret</tt></dt>
800 <dd>This indicates that the parameter specifies the address of a structure
801 that is the return value of the function in the source program.</dd>
802 <dt><tt>noalias</tt></dt>
803 <dd>This indicates that the parameter not alias any other object or any
804 other "noalias" objects during the function call.
805 <dt><tt>noreturn</tt></dt>
806 <dd>This function attribute indicates that the function never returns. This
807 indicates to LLVM that every call to this function should be treated as if
808 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
809 <dt><tt>nounwind</tt></dt>
810 <dd>This function attribute indicates that the function type does not use
811 the unwind instruction and does not allow stack unwinding to propagate
812 through it.</dd>
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]813 <dt><tt>nest</tt></dt>
814 <dd>This indicates that the parameter can be excised using the
815 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
Anton Korobeynikov012eebf2007-11-14 10:30:13 +0000[diff] [blame]816 <dt><tt>pure</tt></dt>
Duncan Sandsd69c0e62007-11-14 21:14:02 +0000[diff] [blame]817 <dd>This function attribute indicates that the function has no side-effects
818 except for producing a return value. The value returned must only depend on
819 the function arguments and/or global variables. It may use values obtained
820 by dereferencing pointers.</dd>
Anton Korobeynikov012eebf2007-11-14 10:30:13 +0000[diff] [blame]821 <dt><tt>const</tt></dt>
Duncan Sandsd69c0e62007-11-14 21:14:02 +0000[diff] [blame]822 <dd>A <tt>const</tt> function has the same restrictions as a <tt>pure</tt>
823 function, but in addition it is not allowed to dereference any pointer arguments
824 or global variables.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]825 </dl>
826
827</div>
828
829<!-- ======================================================================= -->
830<div class="doc_subsection">
831 <a name="moduleasm">Module-Level Inline Assembly</a>
832</div>
833
834<div class="doc_text">
835<p>
836Modules may contain "module-level inline asm" blocks, which corresponds to the
837GCC "file scope inline asm" blocks. These blocks are internally concatenated by
838LLVM and treated as a single unit, but may be separated in the .ll file if
839desired. The syntax is very simple:
840</p>
841
842<div class="doc_code">
843<pre>
844module asm "inline asm code goes here"
845module asm "more can go here"
846</pre>
847</div>
848
849<p>The strings can contain any character by escaping non-printable characters.
850 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
851 for the number.
852</p>
853
854<p>
855 The inline asm code is simply printed to the machine code .s file when
856 assembly code is generated.
857</p>
858</div>
859
860<!-- ======================================================================= -->
861<div class="doc_subsection">
862 <a name="datalayout">Data Layout</a>
863</div>
864
865<div class="doc_text">
866<p>A module may specify a target specific data layout string that specifies how
867data is to be laid out in memory. The syntax for the data layout is simply:</p>
868<pre> target datalayout = "<i>layout specification</i>"</pre>
869<p>The <i>layout specification</i> consists of a list of specifications
870separated by the minus sign character ('-'). Each specification starts with a
871letter and may include other information after the letter to define some
872aspect of the data layout. The specifications accepted are as follows: </p>
873<dl>
874 <dt><tt>E</tt></dt>
875 <dd>Specifies that the target lays out data in big-endian form. That is, the
876 bits with the most significance have the lowest address location.</dd>
877 <dt><tt>e</tt></dt>
878 <dd>Specifies that hte target lays out data in little-endian form. That is,
879 the bits with the least significance have the lowest address location.</dd>
880 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
881 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
882 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
883 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
884 too.</dd>
885 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
886 <dd>This specifies the alignment for an integer type of a given bit
887 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
888 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
889 <dd>This specifies the alignment for a vector type of a given bit
890 <i>size</i>.</dd>
891 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
892 <dd>This specifies the alignment for a floating point type of a given bit
893 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
894 (double).</dd>
895 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
896 <dd>This specifies the alignment for an aggregate type of a given bit
897 <i>size</i>.</dd>
898</dl>
899<p>When constructing the data layout for a given target, LLVM starts with a
900default set of specifications which are then (possibly) overriden by the
901specifications in the <tt>datalayout</tt> keyword. The default specifications
902are given in this list:</p>
903<ul>
904 <li><tt>E</tt> - big endian</li>
905 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
906 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
907 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
908 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
909 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
910 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
911 alignment of 64-bits</li>
912 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
913 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
914 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
915 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
916 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
917</ul>
918<p>When llvm is determining the alignment for a given type, it uses the
919following rules:
920<ol>
921 <li>If the type sought is an exact match for one of the specifications, that
922 specification is used.</li>
923 <li>If no match is found, and the type sought is an integer type, then the
924 smallest integer type that is larger than the bitwidth of the sought type is
925 used. If none of the specifications are larger than the bitwidth then the the
926 largest integer type is used. For example, given the default specifications
927 above, the i7 type will use the alignment of i8 (next largest) while both
928 i65 and i256 will use the alignment of i64 (largest specified).</li>
929 <li>If no match is found, and the type sought is a vector type, then the
930 largest vector type that is smaller than the sought vector type will be used
931 as a fall back. This happens because <128 x double> can be implemented in
932 terms of 64 <2 x double>, for example.</li>
933</ol>
934</div>
935
936<!-- *********************************************************************** -->
937<div class="doc_section"> <a name="typesystem">Type System</a> </div>
938<!-- *********************************************************************** -->
939
940<div class="doc_text">
941
942<p>The LLVM type system is one of the most important features of the
943intermediate representation. Being typed enables a number of
944optimizations to be performed on the IR directly, without having to do
945extra analyses on the side before the transformation. A strong type
946system makes it easier to read the generated code and enables novel
947analyses and transformations that are not feasible to perform on normal
948three address code representations.</p>
949
950</div>
951
952<!-- ======================================================================= -->
953<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
954<div class="doc_text">
955<p>The primitive types are the fundamental building blocks of the LLVM
956system. The current set of primitive types is as follows:</p>
957
958<table class="layout">
959 <tr class="layout">
960 <td class="left">
961 <table>
962 <tbody>
963 <tr><th>Type</th><th>Description</th></tr>
964 <tr><td><tt><a name="t_void">void</a></tt></td><td>No value</td></tr>
965 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
966 </tbody>
967 </table>
968 </td>
969 <td class="right">
970 <table>
971 <tbody>
972 <tr><th>Type</th><th>Description</th></tr>
973 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
974 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
975 </tbody>
976 </table>
977 </td>
978 </tr>
979</table>
980</div>
981
982<!-- _______________________________________________________________________ -->
983<div class="doc_subsubsection"> <a name="t_classifications">Type
984Classifications</a> </div>
985<div class="doc_text">
986<p>These different primitive types fall into a few useful
987classifications:</p>
988
989<table border="1" cellspacing="0" cellpadding="4">
990 <tbody>
991 <tr><th>Classification</th><th>Types</th></tr>
992 <tr>
993 <td><a name="t_integer">integer</a></td>
994 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
995 </tr>
996 <tr>
997 <td><a name="t_floating">floating point</a></td>
998 <td><tt>float, double</tt></td>
999 </tr>
1000 <tr>
1001 <td><a name="t_firstclass">first class</a></td>
1002 <td><tt>i1, ..., float, double, <br/>
1003 <a href="#t_pointer">pointer</a>,<a href="#t_vector">vector</a></tt>
1004 </td>
1005 </tr>
1006 </tbody>
1007</table>
1008
1009<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1010most important. Values of these types are the only ones which can be
1011produced by instructions, passed as arguments, or used as operands to
1012instructions. This means that all structures and arrays must be
1013manipulated either by pointer or by component.</p>
1014</div>
1015
1016<!-- ======================================================================= -->
1017<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1018
1019<div class="doc_text">
1020
1021<p>The real power in LLVM comes from the derived types in the system.
1022This is what allows a programmer to represent arrays, functions,
1023pointers, and other useful types. Note that these derived types may be
1024recursive: For example, it is possible to have a two dimensional array.</p>
1025
1026</div>
1027
1028<!-- _______________________________________________________________________ -->
1029<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1030
1031<div class="doc_text">
1032
1033<h5>Overview:</h5>
1034<p>The integer type is a very simple derived type that simply specifies an
1035arbitrary bit width for the integer type desired. Any bit width from 1 bit to
10362^23-1 (about 8 million) can be specified.</p>
1037
1038<h5>Syntax:</h5>
1039
1040<pre>
1041 iN
1042</pre>
1043
1044<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1045value.</p>
1046
1047<h5>Examples:</h5>
1048<table class="layout">
1049 <tr class="layout">
1050 <td class="left">
1051 <tt>i1</tt><br/>
1052 <tt>i4</tt><br/>
1053 <tt>i8</tt><br/>
1054 <tt>i16</tt><br/>
1055 <tt>i32</tt><br/>
1056 <tt>i42</tt><br/>
1057 <tt>i64</tt><br/>
1058 <tt>i1942652</tt><br/>
1059 </td>
1060 <td class="left">
1061 A boolean integer of 1 bit<br/>
1062 A nibble sized integer of 4 bits.<br/>
1063 A byte sized integer of 8 bits.<br/>
1064 A half word sized integer of 16 bits.<br/>
1065 A word sized integer of 32 bits.<br/>
1066 An integer whose bit width is the answer. <br/>
1067 A double word sized integer of 64 bits.<br/>
1068 A really big integer of over 1 million bits.<br/>
1069 </td>
1070 </tr>
1071</table>
1072</div>
1073
1074<!-- _______________________________________________________________________ -->
1075<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1076
1077<div class="doc_text">
1078
1079<h5>Overview:</h5>
1080
1081<p>The array type is a very simple derived type that arranges elements
1082sequentially in memory. The array type requires a size (number of
1083elements) and an underlying data type.</p>
1084
1085<h5>Syntax:</h5>
1086
1087<pre>
1088 [&lt;# elements&gt; x &lt;elementtype&gt;]
1089</pre>
1090
1091<p>The number of elements is a constant integer value; elementtype may
1092be any type with a size.</p>
1093
1094<h5>Examples:</h5>
1095<table class="layout">
1096 <tr class="layout">
1097 <td class="left">
1098 <tt>[40 x i32 ]</tt><br/>
1099 <tt>[41 x i32 ]</tt><br/>
1100 <tt>[40 x i8]</tt><br/>
1101 </td>
1102 <td class="left">
1103 Array of 40 32-bit integer values.<br/>
1104 Array of 41 32-bit integer values.<br/>
1105 Array of 40 8-bit integer values.<br/>
1106 </td>
1107 </tr>
1108</table>
1109<p>Here are some examples of multidimensional arrays:</p>
1110<table class="layout">
1111 <tr class="layout">
1112 <td class="left">
1113 <tt>[3 x [4 x i32]]</tt><br/>
1114 <tt>[12 x [10 x float]]</tt><br/>
1115 <tt>[2 x [3 x [4 x i16]]]</tt><br/>
1116 </td>
1117 <td class="left">
1118 3x4 array of 32-bit integer values.<br/>
1119 12x10 array of single precision floating point values.<br/>
1120 2x3x4 array of 16-bit integer values.<br/>
1121 </td>
1122 </tr>
1123</table>
1124
1125<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1126length array. Normally, accesses past the end of an array are undefined in
1127LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1128As a special case, however, zero length arrays are recognized to be variable
1129length. This allows implementation of 'pascal style arrays' with the LLVM
1130type "{ i32, [0 x float]}", for example.</p>
1131
1132</div>
1133
1134<!-- _______________________________________________________________________ -->
1135<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1136<div class="doc_text">
1137<h5>Overview:</h5>
1138<p>The function type can be thought of as a function signature. It
1139consists of a return type and a list of formal parameter types.
1140Function types are usually used to build virtual function tables
1141(which are structures of pointers to functions), for indirect function
1142calls, and when defining a function.</p>
1143<p>
1144The return type of a function type cannot be an aggregate type.
1145</p>
1146<h5>Syntax:</h5>
1147<pre> &lt;returntype&gt; (&lt;parameter list&gt;)<br></pre>
1148<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1149specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1150which indicates that the function takes a variable number of arguments.
1151Variable argument functions can access their arguments with the <a
1152 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1153<h5>Examples:</h5>
1154<table class="layout">
1155 <tr class="layout">
1156 <td class="left"><tt>i32 (i32)</tt></td>
1157 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1158 </td>
1159 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1160 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1161 </tt></td>
1162 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1163 an <tt>i16</tt> that should be sign extended and a
1164 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1165 <tt>float</tt>.
1166 </td>
1167 </tr><tr class="layout">
1168 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1169 <td class="left">A vararg function that takes at least one
1170 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1171 which returns an integer. This is the signature for <tt>printf</tt> in
1172 LLVM.
1173 </td>
1174 </tr>
1175</table>
1176
1177</div>
1178<!-- _______________________________________________________________________ -->
1179<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1180<div class="doc_text">
1181<h5>Overview:</h5>
1182<p>The structure type is used to represent a collection of data members
1183together in memory. The packing of the field types is defined to match
1184the ABI of the underlying processor. The elements of a structure may
1185be any type that has a size.</p>
1186<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1187and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1188field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1189instruction.</p>
1190<h5>Syntax:</h5>
1191<pre> { &lt;type list&gt; }<br></pre>
1192<h5>Examples:</h5>
1193<table class="layout">
1194 <tr class="layout">
1195 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1196 <td class="left">A triple of three <tt>i32</tt> values</td>
1197 </tr><tr class="layout">
1198 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1199 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1200 second element is a <a href="#t_pointer">pointer</a> to a
1201 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1202 an <tt>i32</tt>.</td>
1203 </tr>
1204</table>
1205</div>
1206
1207<!-- _______________________________________________________________________ -->
1208<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1209</div>
1210<div class="doc_text">
1211<h5>Overview:</h5>
1212<p>The packed structure type is used to represent a collection of data members
1213together in memory. There is no padding between fields. Further, the alignment
1214of a packed structure is 1 byte. The elements of a packed structure may
1215be any type that has a size.</p>
1216<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1217and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1218field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1219instruction.</p>
1220<h5>Syntax:</h5>
1221<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1222<h5>Examples:</h5>
1223<table class="layout">
1224 <tr class="layout">
1225 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1226 <td class="left">A triple of three <tt>i32</tt> values</td>
1227 </tr><tr class="layout">
1228 <td class="left"><tt>&lt;&nbsp;{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}&nbsp;&gt;</tt></td>
1229 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1230 second element is a <a href="#t_pointer">pointer</a> to a
1231 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1232 an <tt>i32</tt>.</td>
1233 </tr>
1234</table>
1235</div>
1236
1237<!-- _______________________________________________________________________ -->
1238<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1239<div class="doc_text">
1240<h5>Overview:</h5>
1241<p>As in many languages, the pointer type represents a pointer or
1242reference to another object, which must live in memory.</p>
1243<h5>Syntax:</h5>
1244<pre> &lt;type&gt; *<br></pre>
1245<h5>Examples:</h5>
1246<table class="layout">
1247 <tr class="layout">
1248 <td class="left">
1249 <tt>[4x i32]*</tt><br/>
1250 <tt>i32 (i32 *) *</tt><br/>
1251 </td>
1252 <td class="left">
1253 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
1254 four <tt>i32</tt> values<br/>
1255 A <a href="#t_pointer">pointer</a> to a <a
1256 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1257 <tt>i32</tt>.<br/>
1258 </td>
1259 </tr>
1260</table>
1261</div>
1262
1263<!-- _______________________________________________________________________ -->
1264<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1265<div class="doc_text">
1266
1267<h5>Overview:</h5>
1268
1269<p>A vector type is a simple derived type that represents a vector
1270of elements. Vector types are used when multiple primitive data
1271are operated in parallel using a single instruction (SIMD).
1272A vector type requires a size (number of
1273elements) and an underlying primitive data type. Vectors must have a power
1274of two length (1, 2, 4, 8, 16 ...). Vector types are
1275considered <a href="#t_firstclass">first class</a>.</p>
1276
1277<h5>Syntax:</h5>
1278
1279<pre>
1280 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1281</pre>
1282
1283<p>The number of elements is a constant integer value; elementtype may
1284be any integer or floating point type.</p>
1285
1286<h5>Examples:</h5>
1287
1288<table class="layout">
1289 <tr class="layout">
1290 <td class="left">
1291 <tt>&lt;4 x i32&gt;</tt><br/>
1292 <tt>&lt;8 x float&gt;</tt><br/>
1293 <tt>&lt;2 x i64&gt;</tt><br/>
1294 </td>
1295 <td class="left">
1296 Vector of 4 32-bit integer values.<br/>
1297 Vector of 8 floating-point values.<br/>
1298 Vector of 2 64-bit integer values.<br/>
1299 </td>
1300 </tr>
1301</table>
1302</div>
1303
1304<!-- _______________________________________________________________________ -->
1305<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1306<div class="doc_text">
1307
1308<h5>Overview:</h5>
1309
1310<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1311corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1312In LLVM, opaque types can eventually be resolved to any type (not just a
1313structure type).</p>
1314
1315<h5>Syntax:</h5>
1316
1317<pre>
1318 opaque
1319</pre>
1320
1321<h5>Examples:</h5>
1322
1323<table class="layout">
1324 <tr class="layout">
1325 <td class="left">
1326 <tt>opaque</tt>
1327 </td>
1328 <td class="left">
1329 An opaque type.<br/>
1330 </td>
1331 </tr>
1332</table>
1333</div>
1334
1335
1336<!-- *********************************************************************** -->
1337<div class="doc_section"> <a name="constants">Constants</a> </div>
1338<!-- *********************************************************************** -->
1339
1340<div class="doc_text">
1341
1342<p>LLVM has several different basic types of constants. This section describes
1343them all and their syntax.</p>
1344
1345</div>
1346
1347<!-- ======================================================================= -->
1348<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1349
1350<div class="doc_text">
1351
1352<dl>
1353 <dt><b>Boolean constants</b></dt>
1354
1355 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1356 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1357 </dd>
1358
1359 <dt><b>Integer constants</b></dt>
1360
1361 <dd>Standard integers (such as '4') are constants of the <a
1362 href="#t_integer">integer</a> type. Negative numbers may be used with
1363 integer types.
1364 </dd>
1365
1366 <dt><b>Floating point constants</b></dt>
1367
1368 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1369 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1370 notation (see below). Floating point constants must have a <a
1371 href="#t_floating">floating point</a> type. </dd>
1372
1373 <dt><b>Null pointer constants</b></dt>
1374
1375 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1376 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1377
1378</dl>
1379
1380<p>The one non-intuitive notation for constants is the optional hexadecimal form
1381of floating point constants. For example, the form '<tt>double
13820x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
13834.5e+15</tt>'. The only time hexadecimal floating point constants are required
1384(and the only time that they are generated by the disassembler) is when a
1385floating point constant must be emitted but it cannot be represented as a
1386decimal floating point number. For example, NaN's, infinities, and other
1387special values are represented in their IEEE hexadecimal format so that
1388assembly and disassembly do not cause any bits to change in the constants.</p>
1389
1390</div>
1391
1392<!-- ======================================================================= -->
1393<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1394</div>
1395
1396<div class="doc_text">
1397<p>Aggregate constants arise from aggregation of simple constants
1398and smaller aggregate constants.</p>
1399
1400<dl>
1401 <dt><b>Structure constants</b></dt>
1402
1403 <dd>Structure constants are represented with notation similar to structure
1404 type definitions (a comma separated list of elements, surrounded by braces
1405 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* %G }</tt>",
1406 where "<tt>%G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1407 must have <a href="#t_struct">structure type</a>, and the number and
1408 types of elements must match those specified by the type.
1409 </dd>
1410
1411 <dt><b>Array constants</b></dt>
1412
1413 <dd>Array constants are represented with notation similar to array type
1414 definitions (a comma separated list of elements, surrounded by square brackets
1415 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1416 constants must have <a href="#t_array">array type</a>, and the number and
1417 types of elements must match those specified by the type.
1418 </dd>
1419
1420 <dt><b>Vector constants</b></dt>
1421
1422 <dd>Vector constants are represented with notation similar to vector type
1423 definitions (a comma separated list of elements, surrounded by
1424 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1425 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1426 href="#t_vector">vector type</a>, and the number and types of elements must
1427 match those specified by the type.
1428 </dd>
1429
1430 <dt><b>Zero initialization</b></dt>
1431
1432 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1433 value to zero of <em>any</em> type, including scalar and aggregate types.
1434 This is often used to avoid having to print large zero initializers (e.g. for
1435 large arrays) and is always exactly equivalent to using explicit zero
1436 initializers.
1437 </dd>
1438</dl>
1439
1440</div>
1441
1442<!-- ======================================================================= -->
1443<div class="doc_subsection">
1444 <a name="globalconstants">Global Variable and Function Addresses</a>
1445</div>
1446
1447<div class="doc_text">
1448
1449<p>The addresses of <a href="#globalvars">global variables</a> and <a
1450href="#functionstructure">functions</a> are always implicitly valid (link-time)
1451constants. These constants are explicitly referenced when the <a
1452href="#identifiers">identifier for the global</a> is used and always have <a
1453href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1454file:</p>
1455
1456<div class="doc_code">
1457<pre>
1458@X = global i32 17
1459@Y = global i32 42
1460@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1461</pre>
1462</div>
1463
1464</div>
1465
1466<!-- ======================================================================= -->
1467<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1468<div class="doc_text">
1469 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1470 no specific value. Undefined values may be of any type and be used anywhere
1471 a constant is permitted.</p>
1472
1473 <p>Undefined values indicate to the compiler that the program is well defined
1474 no matter what value is used, giving the compiler more freedom to optimize.
1475 </p>
1476</div>
1477
1478<!-- ======================================================================= -->
1479<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1480</div>
1481
1482<div class="doc_text">
1483
1484<p>Constant expressions are used to allow expressions involving other constants
1485to be used as constants. Constant expressions may be of any <a
1486href="#t_firstclass">first class</a> type and may involve any LLVM operation
1487that does not have side effects (e.g. load and call are not supported). The
1488following is the syntax for constant expressions:</p>
1489
1490<dl>
1491 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1492 <dd>Truncate a constant to another type. The bit size of CST must be larger
1493 than the bit size of TYPE. Both types must be integers.</dd>
1494
1495 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1496 <dd>Zero extend a constant to another type. The bit size of CST must be
1497 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1498
1499 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1500 <dd>Sign extend a constant to another type. The bit size of CST must be
1501 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1502
1503 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1504 <dd>Truncate a floating point constant to another floating point type. The
1505 size of CST must be larger than the size of TYPE. Both types must be
1506 floating point.</dd>
1507
1508 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1509 <dd>Floating point extend a constant to another type. The size of CST must be
1510 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1511
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1512 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1513 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame^]1514 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1515 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1516 of the same number of elements. If the value won't fit in the integer type,
1517 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1518
1519 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1520 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame^]1521 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1522 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1523 of the same number of elements. If the value won't fit in the integer type,
1524 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1525
1526 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1527 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame^]1528 constant. TYPE must be a scalar or vector floating point type. CST must be of
1529 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1530 of the same number of elements. If the value won't fit in the floating point
1531 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1532
1533 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1534 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame^]1535 constant. TYPE must be a scalar or vector floating point type. CST must be of
1536 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1537 of the same number of elements. If the value won't fit in the floating point
1538 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1539
1540 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1541 <dd>Convert a pointer typed constant to the corresponding integer constant
1542 TYPE must be an integer type. CST must be of pointer type. The CST value is
1543 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1544
1545 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1546 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1547 pointer type. CST must be of integer type. The CST value is zero extended,
1548 truncated, or unchanged to make it fit in a pointer size. This one is
1549 <i>really</i> dangerous!</dd>
1550
1551 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1552 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1553 identical (same number of bits). The conversion is done as if the CST value
1554 was stored to memory and read back as TYPE. In other words, no bits change
1555 with this operator, just the type. This can be used for conversion of
1556 vector types to any other type, as long as they have the same bit width. For
1557 pointers it is only valid to cast to another pointer type.
1558 </dd>
1559
1560 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1561
1562 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1563 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1564 instruction, the index list may have zero or more indexes, which are required
1565 to make sense for the type of "CSTPTR".</dd>
1566
1567 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1568
1569 <dd>Perform the <a href="#i_select">select operation</a> on
1570 constants.</dd>
1571
1572 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1573 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1574
1575 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1576 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1577
1578 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1579
1580 <dd>Perform the <a href="#i_extractelement">extractelement
1581 operation</a> on constants.
1582
1583 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1584
1585 <dd>Perform the <a href="#i_insertelement">insertelement
1586 operation</a> on constants.</dd>
1587
1588
1589 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1590
1591 <dd>Perform the <a href="#i_shufflevector">shufflevector
1592 operation</a> on constants.</dd>
1593
1594 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1595
1596 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1597 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1598 binary</a> operations. The constraints on operands are the same as those for
1599 the corresponding instruction (e.g. no bitwise operations on floating point
1600 values are allowed).</dd>
1601</dl>
1602</div>
1603
1604<!-- *********************************************************************** -->
1605<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1606<!-- *********************************************************************** -->
1607
1608<!-- ======================================================================= -->
1609<div class="doc_subsection">
1610<a name="inlineasm">Inline Assembler Expressions</a>
1611</div>
1612
1613<div class="doc_text">
1614
1615<p>
1616LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1617Module-Level Inline Assembly</a>) through the use of a special value. This
1618value represents the inline assembler as a string (containing the instructions
1619to emit), a list of operand constraints (stored as a string), and a flag that
1620indicates whether or not the inline asm expression has side effects. An example
1621inline assembler expression is:
1622</p>
1623
1624<div class="doc_code">
1625<pre>
1626i32 (i32) asm "bswap $0", "=r,r"
1627</pre>
1628</div>
1629
1630<p>
1631Inline assembler expressions may <b>only</b> be used as the callee operand of
1632a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1633</p>
1634
1635<div class="doc_code">
1636<pre>
1637%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1638</pre>
1639</div>
1640
1641<p>
1642Inline asms with side effects not visible in the constraint list must be marked
1643as having side effects. This is done through the use of the
1644'<tt>sideeffect</tt>' keyword, like so:
1645</p>
1646
1647<div class="doc_code">
1648<pre>
1649call void asm sideeffect "eieio", ""()
1650</pre>
1651</div>
1652
1653<p>TODO: The format of the asm and constraints string still need to be
1654documented here. Constraints on what can be done (e.g. duplication, moving, etc
1655need to be documented).
1656</p>
1657
1658</div>
1659
1660<!-- *********************************************************************** -->
1661<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1662<!-- *********************************************************************** -->
1663
1664<div class="doc_text">
1665
1666<p>The LLVM instruction set consists of several different
1667classifications of instructions: <a href="#terminators">terminator
1668instructions</a>, <a href="#binaryops">binary instructions</a>,
1669<a href="#bitwiseops">bitwise binary instructions</a>, <a
1670 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1671instructions</a>.</p>
1672
1673</div>
1674
1675<!-- ======================================================================= -->
1676<div class="doc_subsection"> <a name="terminators">Terminator
1677Instructions</a> </div>
1678
1679<div class="doc_text">
1680
1681<p>As mentioned <a href="#functionstructure">previously</a>, every
1682basic block in a program ends with a "Terminator" instruction, which
1683indicates which block should be executed after the current block is
1684finished. These terminator instructions typically yield a '<tt>void</tt>'
1685value: they produce control flow, not values (the one exception being
1686the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1687<p>There are six different terminator instructions: the '<a
1688 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1689instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1690the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1691 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1692 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1693
1694</div>
1695
1696<!-- _______________________________________________________________________ -->
1697<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1698Instruction</a> </div>
1699<div class="doc_text">
1700<h5>Syntax:</h5>
1701<pre> ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
1702 ret void <i>; Return from void function</i>
1703</pre>
1704<h5>Overview:</h5>
1705<p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1706value) from a function back to the caller.</p>
1707<p>There are two forms of the '<tt>ret</tt>' instruction: one that
1708returns a value and then causes control flow, and one that just causes
1709control flow to occur.</p>
1710<h5>Arguments:</h5>
1711<p>The '<tt>ret</tt>' instruction may return any '<a
1712 href="#t_firstclass">first class</a>' type. Notice that a function is
1713not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1714instruction inside of the function that returns a value that does not
1715match the return type of the function.</p>
1716<h5>Semantics:</h5>
1717<p>When the '<tt>ret</tt>' instruction is executed, control flow
1718returns back to the calling function's context. If the caller is a "<a
1719 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1720the instruction after the call. If the caller was an "<a
1721 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1722at the beginning of the "normal" destination block. If the instruction
1723returns a value, that value shall set the call or invoke instruction's
1724return value.</p>
1725<h5>Example:</h5>
1726<pre> ret i32 5 <i>; Return an integer value of 5</i>
1727 ret void <i>; Return from a void function</i>
1728</pre>
1729</div>
1730<!-- _______________________________________________________________________ -->
1731<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1732<div class="doc_text">
1733<h5>Syntax:</h5>
1734<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
1735</pre>
1736<h5>Overview:</h5>
1737<p>The '<tt>br</tt>' instruction is used to cause control flow to
1738transfer to a different basic block in the current function. There are
1739two forms of this instruction, corresponding to a conditional branch
1740and an unconditional branch.</p>
1741<h5>Arguments:</h5>
1742<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1743single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1744unconditional form of the '<tt>br</tt>' instruction takes a single
1745'<tt>label</tt>' value as a target.</p>
1746<h5>Semantics:</h5>
1747<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1748argument is evaluated. If the value is <tt>true</tt>, control flows
1749to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1750control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1751<h5>Example:</h5>
1752<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
1753 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1754</div>
1755<!-- _______________________________________________________________________ -->
1756<div class="doc_subsubsection">
1757 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1758</div>
1759
1760<div class="doc_text">
1761<h5>Syntax:</h5>
1762
1763<pre>
1764 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
1765</pre>
1766
1767<h5>Overview:</h5>
1768
1769<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1770several different places. It is a generalization of the '<tt>br</tt>'
1771instruction, allowing a branch to occur to one of many possible
1772destinations.</p>
1773
1774
1775<h5>Arguments:</h5>
1776
1777<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1778comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1779an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1780table is not allowed to contain duplicate constant entries.</p>
1781
1782<h5>Semantics:</h5>
1783
1784<p>The <tt>switch</tt> instruction specifies a table of values and
1785destinations. When the '<tt>switch</tt>' instruction is executed, this
1786table is searched for the given value. If the value is found, control flow is
1787transfered to the corresponding destination; otherwise, control flow is
1788transfered to the default destination.</p>
1789
1790<h5>Implementation:</h5>
1791
1792<p>Depending on properties of the target machine and the particular
1793<tt>switch</tt> instruction, this instruction may be code generated in different
1794ways. For example, it could be generated as a series of chained conditional
1795branches or with a lookup table.</p>
1796
1797<h5>Example:</h5>
1798
1799<pre>
1800 <i>; Emulate a conditional br instruction</i>
1801 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1802 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1803
1804 <i>; Emulate an unconditional br instruction</i>
1805 switch i32 0, label %dest [ ]
1806
1807 <i>; Implement a jump table:</i>
1808 switch i32 %val, label %otherwise [ i32 0, label %onzero
1809 i32 1, label %onone
1810 i32 2, label %ontwo ]
1811</pre>
1812</div>
1813
1814<!-- _______________________________________________________________________ -->
1815<div class="doc_subsubsection">
1816 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1817</div>
1818
1819<div class="doc_text">
1820
1821<h5>Syntax:</h5>
1822
1823<pre>
1824 &lt;result&gt; = invoke [<a href="#callingconv">cconv</a>] &lt;ptr to function ty&gt; %&lt;function ptr val&gt;(&lt;function args&gt;)
1825 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
1826</pre>
1827
1828<h5>Overview:</h5>
1829
1830<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1831function, with the possibility of control flow transfer to either the
1832'<tt>normal</tt>' label or the
1833'<tt>exception</tt>' label. If the callee function returns with the
1834"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1835"normal" label. If the callee (or any indirect callees) returns with the "<a
1836href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1837continued at the dynamically nearest "exception" label.</p>
1838
1839<h5>Arguments:</h5>
1840
1841<p>This instruction requires several arguments:</p>
1842
1843<ol>
1844 <li>
1845 The optional "cconv" marker indicates which <a href="#callingconv">calling
1846 convention</a> the call should use. If none is specified, the call defaults
1847 to using C calling conventions.
1848 </li>
1849 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1850 function value being invoked. In most cases, this is a direct function
1851 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1852 an arbitrary pointer to function value.
1853 </li>
1854
1855 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1856 function to be invoked. </li>
1857
1858 <li>'<tt>function args</tt>': argument list whose types match the function
1859 signature argument types. If the function signature indicates the function
1860 accepts a variable number of arguments, the extra arguments can be
1861 specified. </li>
1862
1863 <li>'<tt>normal label</tt>': the label reached when the called function
1864 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1865
1866 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1867 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1868
1869</ol>
1870
1871<h5>Semantics:</h5>
1872
1873<p>This instruction is designed to operate as a standard '<tt><a
1874href="#i_call">call</a></tt>' instruction in most regards. The primary
1875difference is that it establishes an association with a label, which is used by
1876the runtime library to unwind the stack.</p>
1877
1878<p>This instruction is used in languages with destructors to ensure that proper
1879cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1880exception. Additionally, this is important for implementation of
1881'<tt>catch</tt>' clauses in high-level languages that support them.</p>
1882
1883<h5>Example:</h5>
1884<pre>
1885 %retval = invoke i32 %Test(i32 15) to label %Continue
1886 unwind label %TestCleanup <i>; {i32}:retval set</i>
1887 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1888 unwind label %TestCleanup <i>; {i32}:retval set</i>
1889</pre>
1890</div>
1891
1892
1893<!-- _______________________________________________________________________ -->
1894
1895<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1896Instruction</a> </div>
1897
1898<div class="doc_text">
1899
1900<h5>Syntax:</h5>
1901<pre>
1902 unwind
1903</pre>
1904
1905<h5>Overview:</h5>
1906
1907<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1908at the first callee in the dynamic call stack which used an <a
1909href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1910primarily used to implement exception handling.</p>
1911
1912<h5>Semantics:</h5>
1913
1914<p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1915immediately halt. The dynamic call stack is then searched for the first <a
1916href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1917execution continues at the "exceptional" destination block specified by the
1918<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1919dynamic call chain, undefined behavior results.</p>
1920</div>
1921
1922<!-- _______________________________________________________________________ -->
1923
1924<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1925Instruction</a> </div>
1926
1927<div class="doc_text">
1928
1929<h5>Syntax:</h5>
1930<pre>
1931 unreachable
1932</pre>
1933
1934<h5>Overview:</h5>
1935
1936<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1937instruction is used to inform the optimizer that a particular portion of the
1938code is not reachable. This can be used to indicate that the code after a
1939no-return function cannot be reached, and other facts.</p>
1940
1941<h5>Semantics:</h5>
1942
1943<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1944</div>
1945
1946
1947
1948<!-- ======================================================================= -->
1949<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1950<div class="doc_text">
1951<p>Binary operators are used to do most of the computation in a
1952program. They require two operands, execute an operation on them, and
1953produce a single value. The operands might represent
1954multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
1955The result value of a binary operator is not
1956necessarily the same type as its operands.</p>
1957<p>There are several different binary operators:</p>
1958</div>
1959<!-- _______________________________________________________________________ -->
1960<div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1961Instruction</a> </div>
1962<div class="doc_text">
1963<h5>Syntax:</h5>
1964<pre> &lt;result&gt; = add &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
1965</pre>
1966<h5>Overview:</h5>
1967<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1968<h5>Arguments:</h5>
1969<p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1970 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1971 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1972Both arguments must have identical types.</p>
1973<h5>Semantics:</h5>
1974<p>The value produced is the integer or floating point sum of the two
1975operands.</p>
1976<h5>Example:</h5>
1977<pre> &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
1978</pre>
1979</div>
1980<!-- _______________________________________________________________________ -->
1981<div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1982Instruction</a> </div>
1983<div class="doc_text">
1984<h5>Syntax:</h5>
1985<pre> &lt;result&gt; = sub &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
1986</pre>
1987<h5>Overview:</h5>
1988<p>The '<tt>sub</tt>' instruction returns the difference of its two
1989operands.</p>
1990<p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1991instruction present in most other intermediate representations.</p>
1992<h5>Arguments:</h5>
1993<p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1994 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1995values.
1996This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1997Both arguments must have identical types.</p>
1998<h5>Semantics:</h5>
1999<p>The value produced is the integer or floating point difference of
2000the two operands.</p>
2001<h5>Example:</h5>
2002<pre>
2003 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2004 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2005</pre>
2006</div>
2007<!-- _______________________________________________________________________ -->
2008<div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
2009Instruction</a> </div>
2010<div class="doc_text">
2011<h5>Syntax:</h5>
2012<pre> &lt;result&gt; = mul &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2013</pre>
2014<h5>Overview:</h5>
2015<p>The '<tt>mul</tt>' instruction returns the product of its two
2016operands.</p>
2017<h5>Arguments:</h5>
2018<p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
2019 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2020values.
2021This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2022Both arguments must have identical types.</p>
2023<h5>Semantics:</h5>
2024<p>The value produced is the integer or floating point product of the
2025two operands.</p>
2026<p>Because the operands are the same width, the result of an integer
2027multiplication is the same whether the operands should be deemed unsigned or
2028signed.</p>
2029<h5>Example:</h5>
2030<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2031</pre>
2032</div>
2033<!-- _______________________________________________________________________ -->
2034<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2035</a></div>
2036<div class="doc_text">
2037<h5>Syntax:</h5>
2038<pre> &lt;result&gt; = udiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2039</pre>
2040<h5>Overview:</h5>
2041<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2042operands.</p>
2043<h5>Arguments:</h5>
2044<p>The two arguments to the '<tt>udiv</tt>' instruction must be
2045<a href="#t_integer">integer</a> values. Both arguments must have identical
2046types. This instruction can also take <a href="#t_vector">vector</a> versions
2047of the values in which case the elements must be integers.</p>
2048<h5>Semantics:</h5>
2049<p>The value produced is the unsigned integer quotient of the two operands. This
2050instruction always performs an unsigned division operation, regardless of
2051whether the arguments are unsigned or not.</p>
2052<h5>Example:</h5>
2053<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2054</pre>
2055</div>
2056<!-- _______________________________________________________________________ -->
2057<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2058</a> </div>
2059<div class="doc_text">
2060<h5>Syntax:</h5>
2061<pre> &lt;result&gt; = sdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2062</pre>
2063<h5>Overview:</h5>
2064<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2065operands.</p>
2066<h5>Arguments:</h5>
2067<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2068<a href="#t_integer">integer</a> values. Both arguments must have identical
2069types. This instruction can also take <a href="#t_vector">vector</a> versions
2070of the values in which case the elements must be integers.</p>
2071<h5>Semantics:</h5>
2072<p>The value produced is the signed integer quotient of the two operands. This
2073instruction always performs a signed division operation, regardless of whether
2074the arguments are signed or not.</p>
2075<h5>Example:</h5>
2076<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2077</pre>
2078</div>
2079<!-- _______________________________________________________________________ -->
2080<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2081Instruction</a> </div>
2082<div class="doc_text">
2083<h5>Syntax:</h5>
2084<pre> &lt;result&gt; = fdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2085</pre>
2086<h5>Overview:</h5>
2087<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2088operands.</p>
2089<h5>Arguments:</h5>
2090<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2091<a href="#t_floating">floating point</a> values. Both arguments must have
2092identical types. This instruction can also take <a href="#t_vector">vector</a>
2093versions of floating point values.</p>
2094<h5>Semantics:</h5>
2095<p>The value produced is the floating point quotient of the two operands.</p>
2096<h5>Example:</h5>
2097<pre> &lt;result&gt; = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2098</pre>
2099</div>
2100<!-- _______________________________________________________________________ -->
2101<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2102</div>
2103<div class="doc_text">
2104<h5>Syntax:</h5>
2105<pre> &lt;result&gt; = urem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2106</pre>
2107<h5>Overview:</h5>
2108<p>The '<tt>urem</tt>' instruction returns the remainder from the
2109unsigned division of its two arguments.</p>
2110<h5>Arguments:</h5>
2111<p>The two arguments to the '<tt>urem</tt>' instruction must be
2112<a href="#t_integer">integer</a> values. Both arguments must have identical
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2113types. This instruction can also take <a href="#t_vector">vector</a> versions
2114of the values in which case the elements must be integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2115<h5>Semantics:</h5>
2116<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2117This instruction always performs an unsigned division to get the remainder,
2118regardless of whether the arguments are unsigned or not.</p>
2119<h5>Example:</h5>
2120<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2121</pre>
2122
2123</div>
2124<!-- _______________________________________________________________________ -->
2125<div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2126Instruction</a> </div>
2127<div class="doc_text">
2128<h5>Syntax:</h5>
2129<pre> &lt;result&gt; = srem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2130</pre>
2131<h5>Overview:</h5>
2132<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2133signed division of its two operands. This instruction can also take
2134<a href="#t_vector">vector</a> versions of the values in which case
2135the elements must be integers.</p>
2136</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2137<h5>Arguments:</h5>
2138<p>The two arguments to the '<tt>srem</tt>' instruction must be
2139<a href="#t_integer">integer</a> values. Both arguments must have identical
2140types.</p>
2141<h5>Semantics:</h5>
2142<p>This instruction returns the <i>remainder</i> of a division (where the result
2143has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2144operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2145a value. For more information about the difference, see <a
2146 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2147Math Forum</a>. For a table of how this is implemented in various languages,
2148please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2149Wikipedia: modulo operation</a>.</p>
2150<h5>Example:</h5>
2151<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2152</pre>
2153
2154</div>
2155<!-- _______________________________________________________________________ -->
2156<div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2157Instruction</a> </div>
2158<div class="doc_text">
2159<h5>Syntax:</h5>
2160<pre> &lt;result&gt; = frem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2161</pre>
2162<h5>Overview:</h5>
2163<p>The '<tt>frem</tt>' instruction returns the remainder from the
2164division of its two operands.</p>
2165<h5>Arguments:</h5>
2166<p>The two arguments to the '<tt>frem</tt>' instruction must be
2167<a href="#t_floating">floating point</a> values. Both arguments must have
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2168identical types. This instruction can also take <a href="#t_vector">vector</a>
2169versions of floating point values.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2170<h5>Semantics:</h5>
2171<p>This instruction returns the <i>remainder</i> of a division.</p>
2172<h5>Example:</h5>
2173<pre> &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2174</pre>
2175</div>
2176
2177<!-- ======================================================================= -->
2178<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2179Operations</a> </div>
2180<div class="doc_text">
2181<p>Bitwise binary operators are used to do various forms of
2182bit-twiddling in a program. They are generally very efficient
2183instructions and can commonly be strength reduced from other
2184instructions. They require two operands, execute an operation on them,
2185and produce a single value. The resulting value of the bitwise binary
2186operators is always the same type as its first operand.</p>
2187</div>
2188
2189<!-- _______________________________________________________________________ -->
2190<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2191Instruction</a> </div>
2192<div class="doc_text">
2193<h5>Syntax:</h5>
2194<pre> &lt;result&gt; = shl &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2195</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2196
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2197<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2198
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2199<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2200the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2201
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2202<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2203
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2204<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2205 href="#t_integer">integer</a> type.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2206
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2207<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2208
2209<p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>. If
2210<tt>var2</tt> is (statically or dynamically) equal to or larger than the number
2211of bits in <tt>var1</tt>, the result is undefined.</p>
2212
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2213<h5>Example:</h5><pre>
2214 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2215 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2216 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2217 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2218</pre>
2219</div>
2220<!-- _______________________________________________________________________ -->
2221<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2222Instruction</a> </div>
2223<div class="doc_text">
2224<h5>Syntax:</h5>
2225<pre> &lt;result&gt; = lshr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2226</pre>
2227
2228<h5>Overview:</h5>
2229<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2230operand shifted to the right a specified number of bits with zero fill.</p>
2231
2232<h5>Arguments:</h5>
2233<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2234<a href="#t_integer">integer</a> type.</p>
2235
2236<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2237
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2238<p>This instruction always performs a logical shift right operation. The most
2239significant bits of the result will be filled with zero bits after the
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2240shift. If <tt>var2</tt> is (statically or dynamically) equal to or larger than
2241the number of bits in <tt>var1</tt>, the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2242
2243<h5>Example:</h5>
2244<pre>
2245 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2246 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2247 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2248 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2249 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2250</pre>
2251</div>
2252
2253<!-- _______________________________________________________________________ -->
2254<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2255Instruction</a> </div>
2256<div class="doc_text">
2257
2258<h5>Syntax:</h5>
2259<pre> &lt;result&gt; = ashr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2260</pre>
2261
2262<h5>Overview:</h5>
2263<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2264operand shifted to the right a specified number of bits with sign extension.</p>
2265
2266<h5>Arguments:</h5>
2267<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2268<a href="#t_integer">integer</a> type.</p>
2269
2270<h5>Semantics:</h5>
2271<p>This instruction always performs an arithmetic shift right operation,
2272The most significant bits of the result will be filled with the sign bit
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2273of <tt>var1</tt>. If <tt>var2</tt> is (statically or dynamically) equal to or
2274larger than the number of bits in <tt>var1</tt>, the result is undefined.
2275</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2276
2277<h5>Example:</h5>
2278<pre>
2279 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2280 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2281 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2282 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2283 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2284</pre>
2285</div>
2286
2287<!-- _______________________________________________________________________ -->
2288<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2289Instruction</a> </div>
2290<div class="doc_text">
2291<h5>Syntax:</h5>
2292<pre> &lt;result&gt; = and &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2293</pre>
2294<h5>Overview:</h5>
2295<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2296its two operands.</p>
2297<h5>Arguments:</h5>
2298<p>The two arguments to the '<tt>and</tt>' instruction must be <a
2299 href="#t_integer">integer</a> values. Both arguments must have
2300identical types.</p>
2301<h5>Semantics:</h5>
2302<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2303<p> </p>
2304<div style="align: center">
2305<table border="1" cellspacing="0" cellpadding="4">
2306 <tbody>
2307 <tr>
2308 <td>In0</td>
2309 <td>In1</td>
2310 <td>Out</td>
2311 </tr>
2312 <tr>
2313 <td>0</td>
2314 <td>0</td>
2315 <td>0</td>
2316 </tr>
2317 <tr>
2318 <td>0</td>
2319 <td>1</td>
2320 <td>0</td>
2321 </tr>
2322 <tr>
2323 <td>1</td>
2324 <td>0</td>
2325 <td>0</td>
2326 </tr>
2327 <tr>
2328 <td>1</td>
2329 <td>1</td>
2330 <td>1</td>
2331 </tr>
2332 </tbody>
2333</table>
2334</div>
2335<h5>Example:</h5>
2336<pre> &lt;result&gt; = and i32 4, %var <i>; yields {i32}:result = 4 &amp; %var</i>
2337 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2338 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2339</pre>
2340</div>
2341<!-- _______________________________________________________________________ -->
2342<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2343<div class="doc_text">
2344<h5>Syntax:</h5>
2345<pre> &lt;result&gt; = or &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2346</pre>
2347<h5>Overview:</h5>
2348<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2349or of its two operands.</p>
2350<h5>Arguments:</h5>
2351<p>The two arguments to the '<tt>or</tt>' instruction must be <a
2352 href="#t_integer">integer</a> values. Both arguments must have
2353identical types.</p>
2354<h5>Semantics:</h5>
2355<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2356<p> </p>
2357<div style="align: center">
2358<table border="1" cellspacing="0" cellpadding="4">
2359 <tbody>
2360 <tr>
2361 <td>In0</td>
2362 <td>In1</td>
2363 <td>Out</td>
2364 </tr>
2365 <tr>
2366 <td>0</td>
2367 <td>0</td>
2368 <td>0</td>
2369 </tr>
2370 <tr>
2371 <td>0</td>
2372 <td>1</td>
2373 <td>1</td>
2374 </tr>
2375 <tr>
2376 <td>1</td>
2377 <td>0</td>
2378 <td>1</td>
2379 </tr>
2380 <tr>
2381 <td>1</td>
2382 <td>1</td>
2383 <td>1</td>
2384 </tr>
2385 </tbody>
2386</table>
2387</div>
2388<h5>Example:</h5>
2389<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2390 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2391 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2392</pre>
2393</div>
2394<!-- _______________________________________________________________________ -->
2395<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2396Instruction</a> </div>
2397<div class="doc_text">
2398<h5>Syntax:</h5>
2399<pre> &lt;result&gt; = xor &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2400</pre>
2401<h5>Overview:</h5>
2402<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2403or of its two operands. The <tt>xor</tt> is used to implement the
2404"one's complement" operation, which is the "~" operator in C.</p>
2405<h5>Arguments:</h5>
2406<p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2407 href="#t_integer">integer</a> values. Both arguments must have
2408identical types.</p>
2409<h5>Semantics:</h5>
2410<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2411<p> </p>
2412<div style="align: center">
2413<table border="1" cellspacing="0" cellpadding="4">
2414 <tbody>
2415 <tr>
2416 <td>In0</td>
2417 <td>In1</td>
2418 <td>Out</td>
2419 </tr>
2420 <tr>
2421 <td>0</td>
2422 <td>0</td>
2423 <td>0</td>
2424 </tr>
2425 <tr>
2426 <td>0</td>
2427 <td>1</td>
2428 <td>1</td>
2429 </tr>
2430 <tr>
2431 <td>1</td>
2432 <td>0</td>
2433 <td>1</td>
2434 </tr>
2435 <tr>
2436 <td>1</td>
2437 <td>1</td>
2438 <td>0</td>
2439 </tr>
2440 </tbody>
2441</table>
2442</div>
2443<p> </p>
2444<h5>Example:</h5>
2445<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2446 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2447 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2448 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2449</pre>
2450</div>
2451
2452<!-- ======================================================================= -->
2453<div class="doc_subsection">
2454 <a name="vectorops">Vector Operations</a>
2455</div>
2456
2457<div class="doc_text">
2458
2459<p>LLVM supports several instructions to represent vector operations in a
2460target-independent manner. These instructions cover the element-access and
2461vector-specific operations needed to process vectors effectively. While LLVM
2462does directly support these vector operations, many sophisticated algorithms
2463will want to use target-specific intrinsics to take full advantage of a specific
2464target.</p>
2465
2466</div>
2467
2468<!-- _______________________________________________________________________ -->
2469<div class="doc_subsubsection">
2470 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2471</div>
2472
2473<div class="doc_text">
2474
2475<h5>Syntax:</h5>
2476
2477<pre>
2478 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
2479</pre>
2480
2481<h5>Overview:</h5>
2482
2483<p>
2484The '<tt>extractelement</tt>' instruction extracts a single scalar
2485element from a vector at a specified index.
2486</p>
2487
2488
2489<h5>Arguments:</h5>
2490
2491<p>
2492The first operand of an '<tt>extractelement</tt>' instruction is a
2493value of <a href="#t_vector">vector</a> type. The second operand is
2494an index indicating the position from which to extract the element.
2495The index may be a variable.</p>
2496
2497<h5>Semantics:</h5>
2498
2499<p>
2500The result is a scalar of the same type as the element type of
2501<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2502<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2503results are undefined.
2504</p>
2505
2506<h5>Example:</h5>
2507
2508<pre>
2509 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
2510</pre>
2511</div>
2512
2513
2514<!-- _______________________________________________________________________ -->
2515<div class="doc_subsubsection">
2516 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2517</div>
2518
2519<div class="doc_text">
2520
2521<h5>Syntax:</h5>
2522
2523<pre>
2524 &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>
2525</pre>
2526
2527<h5>Overview:</h5>
2528
2529<p>
2530The '<tt>insertelement</tt>' instruction inserts a scalar
2531element into a vector at a specified index.
2532</p>
2533
2534
2535<h5>Arguments:</h5>
2536
2537<p>
2538The first operand of an '<tt>insertelement</tt>' instruction is a
2539value of <a href="#t_vector">vector</a> type. The second operand is a
2540scalar value whose type must equal the element type of the first
2541operand. The third operand is an index indicating the position at
2542which to insert the value. The index may be a variable.</p>
2543
2544<h5>Semantics:</h5>
2545
2546<p>
2547The result is a vector of the same type as <tt>val</tt>. Its
2548element values are those of <tt>val</tt> except at position
2549<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2550exceeds the length of <tt>val</tt>, the results are undefined.
2551</p>
2552
2553<h5>Example:</h5>
2554
2555<pre>
2556 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
2557</pre>
2558</div>
2559
2560<!-- _______________________________________________________________________ -->
2561<div class="doc_subsubsection">
2562 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2563</div>
2564
2565<div class="doc_text">
2566
2567<h5>Syntax:</h5>
2568
2569<pre>
2570 &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>
2571</pre>
2572
2573<h5>Overview:</h5>
2574
2575<p>
2576The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2577from two input vectors, returning a vector of the same type.
2578</p>
2579
2580<h5>Arguments:</h5>
2581
2582<p>
2583The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2584with types that match each other and types that match the result of the
2585instruction. The third argument is a shuffle mask, which has the same number
2586of elements as the other vector type, but whose element type is always 'i32'.
2587</p>
2588
2589<p>
2590The shuffle mask operand is required to be a constant vector with either
2591constant integer or undef values.
2592</p>
2593
2594<h5>Semantics:</h5>
2595
2596<p>
2597The elements of the two input vectors are numbered from left to right across
2598both of the vectors. The shuffle mask operand specifies, for each element of
2599the result vector, which element of the two input registers the result element
2600gets. The element selector may be undef (meaning "don't care") and the second
2601operand may be undef if performing a shuffle from only one vector.
2602</p>
2603
2604<h5>Example:</h5>
2605
2606<pre>
2607 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
2608 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
2609 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
2610 &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.
2611</pre>
2612</div>
2613
2614
2615<!-- ======================================================================= -->
2616<div class="doc_subsection">
2617 <a name="memoryops">Memory Access and Addressing Operations</a>
2618</div>
2619
2620<div class="doc_text">
2621
2622<p>A key design point of an SSA-based representation is how it
2623represents memory. In LLVM, no memory locations are in SSA form, which
2624makes things very simple. This section describes how to read, write,
2625allocate, and free memory in LLVM.</p>
2626
2627</div>
2628
2629<!-- _______________________________________________________________________ -->
2630<div class="doc_subsubsection">
2631 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2632</div>
2633
2634<div class="doc_text">
2635
2636<h5>Syntax:</h5>
2637
2638<pre>
2639 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
2640</pre>
2641
2642<h5>Overview:</h5>
2643
2644<p>The '<tt>malloc</tt>' instruction allocates memory from the system
2645heap and returns a pointer to it.</p>
2646
2647<h5>Arguments:</h5>
2648
2649<p>The '<tt>malloc</tt>' instruction allocates
2650<tt>sizeof(&lt;type&gt;)*NumElements</tt>
2651bytes of memory from the operating system and returns a pointer of the
2652appropriate type to the program. If "NumElements" is specified, it is the
2653number of elements allocated. If an alignment is specified, the value result
2654of the allocation is guaranteed to be aligned to at least that boundary. If
2655not specified, or if zero, the target can choose to align the allocation on any
2656convenient boundary.</p>
2657
2658<p>'<tt>type</tt>' must be a sized type.</p>
2659
2660<h5>Semantics:</h5>
2661
2662<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2663a pointer is returned.</p>
2664
2665<h5>Example:</h5>
2666
2667<pre>
2668 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2669
2670 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2671 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2672 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2673 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2674 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2675</pre>
2676</div>
2677
2678<!-- _______________________________________________________________________ -->
2679<div class="doc_subsubsection">
2680 <a name="i_free">'<tt>free</tt>' Instruction</a>
2681</div>
2682
2683<div class="doc_text">
2684
2685<h5>Syntax:</h5>
2686
2687<pre>
2688 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
2689</pre>
2690
2691<h5>Overview:</h5>
2692
2693<p>The '<tt>free</tt>' instruction returns memory back to the unused
2694memory heap to be reallocated in the future.</p>
2695
2696<h5>Arguments:</h5>
2697
2698<p>'<tt>value</tt>' shall be a pointer value that points to a value
2699that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2700instruction.</p>
2701
2702<h5>Semantics:</h5>
2703
2704<p>Access to the memory pointed to by the pointer is no longer defined
2705after this instruction executes.</p>
2706
2707<h5>Example:</h5>
2708
2709<pre>
2710 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2711 free [4 x i8]* %array
2712</pre>
2713</div>
2714
2715<!-- _______________________________________________________________________ -->
2716<div class="doc_subsubsection">
2717 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2718</div>
2719
2720<div class="doc_text">
2721
2722<h5>Syntax:</h5>
2723
2724<pre>
2725 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
2726</pre>
2727
2728<h5>Overview:</h5>
2729
2730<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
2731currently executing function, to be automatically released when this function
2732returns to its caller.</p>
2733
2734<h5>Arguments:</h5>
2735
2736<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
2737bytes of memory on the runtime stack, returning a pointer of the
2738appropriate type to the program. If "NumElements" is specified, it is the
2739number of elements allocated. If an alignment is specified, the value result
2740of the allocation is guaranteed to be aligned to at least that boundary. If
2741not specified, or if zero, the target can choose to align the allocation on any
2742convenient boundary.</p>
2743
2744<p>'<tt>type</tt>' may be any sized type.</p>
2745
2746<h5>Semantics:</h5>
2747
2748<p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2749memory is automatically released when the function returns. The '<tt>alloca</tt>'
2750instruction is commonly used to represent automatic variables that must
2751have an address available. When the function returns (either with the <tt><a
2752 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2753instructions), the memory is reclaimed.</p>
2754
2755<h5>Example:</h5>
2756
2757<pre>
2758 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2759 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2760 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2761 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2762</pre>
2763</div>
2764
2765<!-- _______________________________________________________________________ -->
2766<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2767Instruction</a> </div>
2768<div class="doc_text">
2769<h5>Syntax:</h5>
2770<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>
2771<h5>Overview:</h5>
2772<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2773<h5>Arguments:</h5>
2774<p>The argument to the '<tt>load</tt>' instruction specifies the memory
2775address from which to load. The pointer must point to a <a
2776 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2777marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2778the number or order of execution of this <tt>load</tt> with other
2779volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2780instructions. </p>
2781<h5>Semantics:</h5>
2782<p>The location of memory pointed to is loaded.</p>
2783<h5>Examples:</h5>
2784<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2785 <a
2786 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2787 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2788</pre>
2789</div>
2790<!-- _______________________________________________________________________ -->
2791<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2792Instruction</a> </div>
2793<div class="doc_text">
2794<h5>Syntax:</h5>
2795<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
2796 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
2797</pre>
2798<h5>Overview:</h5>
2799<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2800<h5>Arguments:</h5>
2801<p>There are two arguments to the '<tt>store</tt>' instruction: a value
2802to store and an address at which to store it. The type of the '<tt>&lt;pointer&gt;</tt>'
2803operand must be a pointer to the type of the '<tt>&lt;value&gt;</tt>'
2804operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2805optimizer is not allowed to modify the number or order of execution of
2806this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2807 href="#i_store">store</a></tt> instructions.</p>
2808<h5>Semantics:</h5>
2809<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
2810at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
2811<h5>Example:</h5>
2812<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]2813 store i32 3, i32* %ptr <i>; yields {void}</i>
2814 %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]2815</pre>
2816</div>
2817
2818<!-- _______________________________________________________________________ -->
2819<div class="doc_subsubsection">
2820 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2821</div>
2822
2823<div class="doc_text">
2824<h5>Syntax:</h5>
2825<pre>
2826 &lt;result&gt; = getelementptr &lt;ty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
2827</pre>
2828
2829<h5>Overview:</h5>
2830
2831<p>
2832The '<tt>getelementptr</tt>' instruction is used to get the address of a
2833subelement of an aggregate data structure.</p>
2834
2835<h5>Arguments:</h5>
2836
2837<p>This instruction takes a list of integer operands that indicate what
2838elements of the aggregate object to index to. The actual types of the arguments
2839provided depend on the type of the first pointer argument. The
2840'<tt>getelementptr</tt>' instruction is used to index down through the type
2841levels of a structure or to a specific index in an array. When indexing into a
2842structure, only <tt>i32</tt> integer constants are allowed. When indexing
2843into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2844be sign extended to 64-bit values.</p>
2845
2846<p>For example, let's consider a C code fragment and how it gets
2847compiled to LLVM:</p>
2848
2849<div class="doc_code">
2850<pre>
2851struct RT {
2852 char A;
2853 int B[10][20];
2854 char C;
2855};
2856struct ST {
2857 int X;
2858 double Y;
2859 struct RT Z;
2860};
2861
2862int *foo(struct ST *s) {
2863 return &amp;s[1].Z.B[5][13];
2864}
2865</pre>
2866</div>
2867
2868<p>The LLVM code generated by the GCC frontend is:</p>
2869
2870<div class="doc_code">
2871<pre>
2872%RT = type { i8 , [10 x [20 x i32]], i8 }
2873%ST = type { i32, double, %RT }
2874
2875define i32* %foo(%ST* %s) {
2876entry:
2877 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
2878 ret i32* %reg
2879}
2880</pre>
2881</div>
2882
2883<h5>Semantics:</h5>
2884
2885<p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2886on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2887and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
2888<a href="#t_integer">integer</a> type but the value will always be sign extended
2889to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
2890<b>constants</b>.</p>
2891
2892<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2893type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
2894}</tt>' type, a structure. The second index indexes into the third element of
2895the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
2896i8 }</tt>' type, another structure. The third index indexes into the second
2897element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
2898array. The two dimensions of the array are subscripted into, yielding an
2899'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2900to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
2901
2902<p>Note that it is perfectly legal to index partially through a
2903structure, returning a pointer to an inner element. Because of this,
2904the LLVM code for the given testcase is equivalent to:</p>
2905
2906<pre>
2907 define i32* %foo(%ST* %s) {
2908 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
2909 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
2910 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
2911 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
2912 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
2913 ret i32* %t5
2914 }
2915</pre>
2916
2917<p>Note that it is undefined to access an array out of bounds: array and
2918pointer indexes must always be within the defined bounds of the array type.
2919The one exception for this rules is zero length arrays. These arrays are
2920defined to be accessible as variable length arrays, which requires access
2921beyond the zero'th element.</p>
2922
2923<p>The getelementptr instruction is often confusing. For some more insight
2924into how it works, see <a href="GetElementPtr.html">the getelementptr
2925FAQ</a>.</p>
2926
2927<h5>Example:</h5>
2928
2929<pre>
2930 <i>; yields [12 x i8]*:aptr</i>
2931 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
2932</pre>
2933</div>
2934
2935<!-- ======================================================================= -->
2936<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2937</div>
2938<div class="doc_text">
2939<p>The instructions in this category are the conversion instructions (casting)
2940which all take a single operand and a type. They perform various bit conversions
2941on the operand.</p>
2942</div>
2943
2944<!-- _______________________________________________________________________ -->
2945<div class="doc_subsubsection">
2946 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2947</div>
2948<div class="doc_text">
2949
2950<h5>Syntax:</h5>
2951<pre>
2952 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
2953</pre>
2954
2955<h5>Overview:</h5>
2956<p>
2957The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2958</p>
2959
2960<h5>Arguments:</h5>
2961<p>
2962The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2963be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2964and type of the result, which must be an <a href="#t_integer">integer</a>
2965type. The bit size of <tt>value</tt> must be larger than the bit size of
2966<tt>ty2</tt>. Equal sized types are not allowed.</p>
2967
2968<h5>Semantics:</h5>
2969<p>
2970The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
2971and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
2972larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
2973It will always truncate bits.</p>
2974
2975<h5>Example:</h5>
2976<pre>
2977 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
2978 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
2979 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
2980</pre>
2981</div>
2982
2983<!-- _______________________________________________________________________ -->
2984<div class="doc_subsubsection">
2985 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
2986</div>
2987<div class="doc_text">
2988
2989<h5>Syntax:</h5>
2990<pre>
2991 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
2992</pre>
2993
2994<h5>Overview:</h5>
2995<p>The '<tt>zext</tt>' instruction zero extends its operand to type
2996<tt>ty2</tt>.</p>
2997
2998
2999<h5>Arguments:</h5>
3000<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3001<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3002also be of <a href="#t_integer">integer</a> type. The bit size of the
3003<tt>value</tt> must be smaller than the bit size of the destination type,
3004<tt>ty2</tt>.</p>
3005
3006<h5>Semantics:</h5>
3007<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3008bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3009
3010<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3011
3012<h5>Example:</h5>
3013<pre>
3014 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3015 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3016</pre>
3017</div>
3018
3019<!-- _______________________________________________________________________ -->
3020<div class="doc_subsubsection">
3021 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3022</div>
3023<div class="doc_text">
3024
3025<h5>Syntax:</h5>
3026<pre>
3027 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3028</pre>
3029
3030<h5>Overview:</h5>
3031<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3032
3033<h5>Arguments:</h5>
3034<p>
3035The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3036<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3037also be of <a href="#t_integer">integer</a> type. The bit size of the
3038<tt>value</tt> must be smaller than the bit size of the destination type,
3039<tt>ty2</tt>.</p>
3040
3041<h5>Semantics:</h5>
3042<p>
3043The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3044bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3045the type <tt>ty2</tt>.</p>
3046
3047<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3048
3049<h5>Example:</h5>
3050<pre>
3051 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3052 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3053</pre>
3054</div>
3055
3056<!-- _______________________________________________________________________ -->
3057<div class="doc_subsubsection">
3058 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3059</div>
3060
3061<div class="doc_text">
3062
3063<h5>Syntax:</h5>
3064
3065<pre>
3066 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3067</pre>
3068
3069<h5>Overview:</h5>
3070<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3071<tt>ty2</tt>.</p>
3072
3073
3074<h5>Arguments:</h5>
3075<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3076 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3077cast it to. The size of <tt>value</tt> must be larger than the size of
3078<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3079<i>no-op cast</i>.</p>
3080
3081<h5>Semantics:</h5>
3082<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3083<a href="#t_floating">floating point</a> type to a smaller
3084<a href="#t_floating">floating point</a> type. If the value cannot fit within
3085the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3086
3087<h5>Example:</h5>
3088<pre>
3089 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3090 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3091</pre>
3092</div>
3093
3094<!-- _______________________________________________________________________ -->
3095<div class="doc_subsubsection">
3096 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3097</div>
3098<div class="doc_text">
3099
3100<h5>Syntax:</h5>
3101<pre>
3102 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3103</pre>
3104
3105<h5>Overview:</h5>
3106<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3107floating point value.</p>
3108
3109<h5>Arguments:</h5>
3110<p>The '<tt>fpext</tt>' instruction takes a
3111<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3112and a <a href="#t_floating">floating point</a> type to cast it to. The source
3113type must be smaller than the destination type.</p>
3114
3115<h5>Semantics:</h5>
3116<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3117<a href="#t_floating">floating point</a> type to a larger
3118<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3119used to make a <i>no-op cast</i> because it always changes bits. Use
3120<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3121
3122<h5>Example:</h5>
3123<pre>
3124 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3125 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3126</pre>
3127</div>
3128
3129<!-- _______________________________________________________________________ -->
3130<div class="doc_subsubsection">
3131 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3132</div>
3133<div class="doc_text">
3134
3135<h5>Syntax:</h5>
3136<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3137 &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]3138</pre>
3139
3140<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3141<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3142unsigned integer equivalent of type <tt>ty2</tt>.
3143</p>
3144
3145<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3146<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^]3147scalar or vector <a href="#t_floating">floating point</a> value, and a type
3148to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3149type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3150vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3151
3152<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3153<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3154<a href="#t_floating">floating point</a> operand into the nearest (rounding
3155towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3156the results are undefined.</p>
3157
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3158<h5>Example:</h5>
3159<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3160 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3161 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3162 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3163</pre>
3164</div>
3165
3166<!-- _______________________________________________________________________ -->
3167<div class="doc_subsubsection">
3168 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3169</div>
3170<div class="doc_text">
3171
3172<h5>Syntax:</h5>
3173<pre>
3174 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3175</pre>
3176
3177<h5>Overview:</h5>
3178<p>The '<tt>fptosi</tt>' instruction converts
3179<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3180</p>
3181
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3182<h5>Arguments:</h5>
3183<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^]3184scalar or vector <a href="#t_floating">floating point</a> value, and a type
3185to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3186type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3187vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3188
3189<h5>Semantics:</h5>
3190<p>The '<tt>fptosi</tt>' instruction converts its
3191<a href="#t_floating">floating point</a> operand into the nearest (rounding
3192towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3193the results are undefined.</p>
3194
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3195<h5>Example:</h5>
3196<pre>
3197 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3198 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3199 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3200</pre>
3201</div>
3202
3203<!-- _______________________________________________________________________ -->
3204<div class="doc_subsubsection">
3205 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3206</div>
3207<div class="doc_text">
3208
3209<h5>Syntax:</h5>
3210<pre>
3211 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3212</pre>
3213
3214<h5>Overview:</h5>
3215<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3216integer and converts that value to the <tt>ty2</tt> type.</p>
3217
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3218<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame^]3219<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3220scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3221to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3222type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3223floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3224
3225<h5>Semantics:</h5>
3226<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3227integer quantity and converts it to the corresponding floating point value. If
3228the value cannot fit in the floating point value, the results are undefined.</p>
3229
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3230<h5>Example:</h5>
3231<pre>
3232 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3233 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3234</pre>
3235</div>
3236
3237<!-- _______________________________________________________________________ -->
3238<div class="doc_subsubsection">
3239 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3240</div>
3241<div class="doc_text">
3242
3243<h5>Syntax:</h5>
3244<pre>
3245 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3246</pre>
3247
3248<h5>Overview:</h5>
3249<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3250integer and converts that value to the <tt>ty2</tt> type.</p>
3251
3252<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame^]3253<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3254scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3255to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3256type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3257floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3258
3259<h5>Semantics:</h5>
3260<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3261integer quantity and converts it to the corresponding floating point value. If
3262the value cannot fit in the floating point value, the results are undefined.</p>
3263
3264<h5>Example:</h5>
3265<pre>
3266 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3267 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3268</pre>
3269</div>
3270
3271<!-- _______________________________________________________________________ -->
3272<div class="doc_subsubsection">
3273 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3274</div>
3275<div class="doc_text">
3276
3277<h5>Syntax:</h5>
3278<pre>
3279 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3280</pre>
3281
3282<h5>Overview:</h5>
3283<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3284the integer type <tt>ty2</tt>.</p>
3285
3286<h5>Arguments:</h5>
3287<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3288must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3289<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3290
3291<h5>Semantics:</h5>
3292<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3293<tt>ty2</tt> by interpreting the pointer value as an integer and either
3294truncating or zero extending that value to the size of the integer type. If
3295<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3296<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3297are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3298change.</p>
3299
3300<h5>Example:</h5>
3301<pre>
3302 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3303 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3304</pre>
3305</div>
3306
3307<!-- _______________________________________________________________________ -->
3308<div class="doc_subsubsection">
3309 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3310</div>
3311<div class="doc_text">
3312
3313<h5>Syntax:</h5>
3314<pre>
3315 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3316</pre>
3317
3318<h5>Overview:</h5>
3319<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3320a pointer type, <tt>ty2</tt>.</p>
3321
3322<h5>Arguments:</h5>
3323<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3324value to cast, and a type to cast it to, which must be a
3325<a href="#t_pointer">pointer</a> type.
3326
3327<h5>Semantics:</h5>
3328<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3329<tt>ty2</tt> by applying either a zero extension or a truncation depending on
3330the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3331size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3332the size of a pointer then a zero extension is done. If they are the same size,
3333nothing is done (<i>no-op cast</i>).</p>
3334
3335<h5>Example:</h5>
3336<pre>
3337 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3338 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3339 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3340</pre>
3341</div>
3342
3343<!-- _______________________________________________________________________ -->
3344<div class="doc_subsubsection">
3345 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3346</div>
3347<div class="doc_text">
3348
3349<h5>Syntax:</h5>
3350<pre>
3351 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3352</pre>
3353
3354<h5>Overview:</h5>
3355<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3356<tt>ty2</tt> without changing any bits.</p>
3357
3358<h5>Arguments:</h5>
3359<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3360a first class value, and a type to cast it to, which must also be a <a
3361 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3362and the destination type, <tt>ty2</tt>, must be identical. If the source
3363type is a pointer, the destination type must also be a pointer.</p>
3364
3365<h5>Semantics:</h5>
3366<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3367<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3368this conversion. The conversion is done as if the <tt>value</tt> had been
3369stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3370converted to other pointer types with this instruction. To convert pointers to
3371other types, use the <a href="#i_inttoptr">inttoptr</a> or
3372<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3373
3374<h5>Example:</h5>
3375<pre>
3376 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3377 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3378 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3379</pre>
3380</div>
3381
3382<!-- ======================================================================= -->
3383<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3384<div class="doc_text">
3385<p>The instructions in this category are the "miscellaneous"
3386instructions, which defy better classification.</p>
3387</div>
3388
3389<!-- _______________________________________________________________________ -->
3390<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3391</div>
3392<div class="doc_text">
3393<h5>Syntax:</h5>
3394<pre> &lt;result&gt; = icmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3395</pre>
3396<h5>Overview:</h5>
3397<p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3398of its two integer operands.</p>
3399<h5>Arguments:</h5>
3400<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3401the condition code indicating the kind of comparison to perform. It is not
3402a value, just a keyword. The possible condition code are:
3403<ol>
3404 <li><tt>eq</tt>: equal</li>
3405 <li><tt>ne</tt>: not equal </li>
3406 <li><tt>ugt</tt>: unsigned greater than</li>
3407 <li><tt>uge</tt>: unsigned greater or equal</li>
3408 <li><tt>ult</tt>: unsigned less than</li>
3409 <li><tt>ule</tt>: unsigned less or equal</li>
3410 <li><tt>sgt</tt>: signed greater than</li>
3411 <li><tt>sge</tt>: signed greater or equal</li>
3412 <li><tt>slt</tt>: signed less than</li>
3413 <li><tt>sle</tt>: signed less or equal</li>
3414</ol>
3415<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3416<a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3417<h5>Semantics:</h5>
3418<p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3419the condition code given as <tt>cond</tt>. The comparison performed always
3420yields a <a href="#t_primitive">i1</a> result, as follows:
3421<ol>
3422 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3423 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3424 </li>
3425 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3426 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3427 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3428 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3429 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3430 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3431 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3432 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3433 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3434 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3435 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3436 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3437 <li><tt>sge</tt>: interprets the operands as signed values and yields
3438 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3439 <li><tt>slt</tt>: interprets the operands as signed values and yields
3440 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3441 <li><tt>sle</tt>: interprets the operands as signed values and yields
3442 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3443</ol>
3444<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3445values are compared as if they were integers.</p>
3446
3447<h5>Example:</h5>
3448<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
3449 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
3450 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
3451 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3452 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3453 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
3454</pre>
3455</div>
3456
3457<!-- _______________________________________________________________________ -->
3458<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3459</div>
3460<div class="doc_text">
3461<h5>Syntax:</h5>
3462<pre> &lt;result&gt; = fcmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3463</pre>
3464<h5>Overview:</h5>
3465<p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3466of its floating point operands.</p>
3467<h5>Arguments:</h5>
3468<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3469the condition code indicating the kind of comparison to perform. It is not
3470a value, just a keyword. The possible condition code are:
3471<ol>
3472 <li><tt>false</tt>: no comparison, always returns false</li>
3473 <li><tt>oeq</tt>: ordered and equal</li>
3474 <li><tt>ogt</tt>: ordered and greater than </li>
3475 <li><tt>oge</tt>: ordered and greater than or equal</li>
3476 <li><tt>olt</tt>: ordered and less than </li>
3477 <li><tt>ole</tt>: ordered and less than or equal</li>
3478 <li><tt>one</tt>: ordered and not equal</li>
3479 <li><tt>ord</tt>: ordered (no nans)</li>
3480 <li><tt>ueq</tt>: unordered or equal</li>
3481 <li><tt>ugt</tt>: unordered or greater than </li>
3482 <li><tt>uge</tt>: unordered or greater than or equal</li>
3483 <li><tt>ult</tt>: unordered or less than </li>
3484 <li><tt>ule</tt>: unordered or less than or equal</li>
3485 <li><tt>une</tt>: unordered or not equal</li>
3486 <li><tt>uno</tt>: unordered (either nans)</li>
3487 <li><tt>true</tt>: no comparison, always returns true</li>
3488</ol>
3489<p><i>Ordered</i> means that neither operand is a QNAN while
3490<i>unordered</i> means that either operand may be a QNAN.</p>
3491<p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3492<a href="#t_floating">floating point</a> typed. They must have identical
3493types.</p>
3494<h5>Semantics:</h5>
3495<p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3496the condition code given as <tt>cond</tt>. The comparison performed always
3497yields a <a href="#t_primitive">i1</a> result, as follows:
3498<ol>
3499 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3500 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3501 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3502 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3503 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3504 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3505 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3506 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3507 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3508 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3509 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3510 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3511 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3512 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3513 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3514 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3515 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3516 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3517 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3518 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3519 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3520 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3521 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3522 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3523 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3524 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3525 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3526 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3527</ol>
3528
3529<h5>Example:</h5>
3530<pre> &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3531 &lt;result&gt; = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3532 &lt;result&gt; = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3533 &lt;result&gt; = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3534</pre>
3535</div>
3536
3537<!-- _______________________________________________________________________ -->
3538<div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3539Instruction</a> </div>
3540<div class="doc_text">
3541<h5>Syntax:</h5>
3542<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
3543<h5>Overview:</h5>
3544<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
3545the SSA graph representing the function.</p>
3546<h5>Arguments:</h5>
3547<p>The type of the incoming values is specified with the first type
3548field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3549as arguments, with one pair for each predecessor basic block of the
3550current block. Only values of <a href="#t_firstclass">first class</a>
3551type may be used as the value arguments to the PHI node. Only labels
3552may be used as the label arguments.</p>
3553<p>There must be no non-phi instructions between the start of a basic
3554block and the PHI instructions: i.e. PHI instructions must be first in
3555a basic block.</p>
3556<h5>Semantics:</h5>
3557<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3558specified by the pair corresponding to the predecessor basic block that executed
3559just prior to the current block.</p>
3560<h5>Example:</h5>
3561<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>
3562</div>
3563
3564<!-- _______________________________________________________________________ -->
3565<div class="doc_subsubsection">
3566 <a name="i_select">'<tt>select</tt>' Instruction</a>
3567</div>
3568
3569<div class="doc_text">
3570
3571<h5>Syntax:</h5>
3572
3573<pre>
3574 &lt;result&gt; = select i1 &lt;cond&gt;, &lt;ty&gt; &lt;val1&gt;, &lt;ty&gt; &lt;val2&gt; <i>; yields ty</i>
3575</pre>
3576
3577<h5>Overview:</h5>
3578
3579<p>
3580The '<tt>select</tt>' instruction is used to choose one value based on a
3581condition, without branching.
3582</p>
3583
3584
3585<h5>Arguments:</h5>
3586
3587<p>
3588The '<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.
3589</p>
3590
3591<h5>Semantics:</h5>
3592
3593<p>
3594If the boolean condition evaluates to true, the instruction returns the first
3595value argument; otherwise, it returns the second value argument.
3596</p>
3597
3598<h5>Example:</h5>
3599
3600<pre>
3601 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3602</pre>
3603</div>
3604
3605
3606<!-- _______________________________________________________________________ -->
3607<div class="doc_subsubsection">
3608 <a name="i_call">'<tt>call</tt>' Instruction</a>
3609</div>
3610
3611<div class="doc_text">
3612
3613<h5>Syntax:</h5>
3614<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3615 &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]3616</pre>
3617
3618<h5>Overview:</h5>
3619
3620<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3621
3622<h5>Arguments:</h5>
3623
3624<p>This instruction requires several arguments:</p>
3625
3626<ol>
3627 <li>
3628 <p>The optional "tail" marker indicates whether the callee function accesses
3629 any allocas or varargs in the caller. If the "tail" marker is present, the
3630 function call is eligible for tail call optimization. Note that calls may
3631 be marked "tail" even if they do not occur before a <a
3632 href="#i_ret"><tt>ret</tt></a> instruction.
3633 </li>
3634 <li>
3635 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3636 convention</a> the call should use. If none is specified, the call defaults
3637 to using C calling conventions.
3638 </li>
3639 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3640 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
3641 the type of the return value. Functions that return no value are marked
3642 <tt><a href="#t_void">void</a></tt>.</p>
3643 </li>
3644 <li>
3645 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
3646 value being invoked. The argument types must match the types implied by
3647 this signature. This type can be omitted if the function is not varargs
3648 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3649 </li>
3650 <li>
3651 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3652 be invoked. In most cases, this is a direct function invocation, but
3653 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3654 to function value.</p>
3655 </li>
3656 <li>
3657 <p>'<tt>function args</tt>': argument list whose types match the
3658 function signature argument types. All arguments must be of
3659 <a href="#t_firstclass">first class</a> type. If the function signature
3660 indicates the function accepts a variable number of arguments, the extra
3661 arguments can be specified.</p>
3662 </li>
3663</ol>
3664
3665<h5>Semantics:</h5>
3666
3667<p>The '<tt>call</tt>' instruction is used to cause control flow to
3668transfer to a specified function, with its incoming arguments bound to
3669the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3670instruction in the called function, control flow continues with the
3671instruction after the function call, and the return value of the
3672function is bound to the result argument. This is a simpler case of
3673the <a href="#i_invoke">invoke</a> instruction.</p>
3674
3675<h5>Example:</h5>
3676
3677<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3678 %retval = call i32 @test(i32 %argc)
3679 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42);
3680 %X = tail call i32 @foo()
3681 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo()
3682 %Z = call void %foo(i8 97 signext)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3683</pre>
3684
3685</div>
3686
3687<!-- _______________________________________________________________________ -->
3688<div class="doc_subsubsection">
3689 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3690</div>
3691
3692<div class="doc_text">
3693
3694<h5>Syntax:</h5>
3695
3696<pre>
3697 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
3698</pre>
3699
3700<h5>Overview:</h5>
3701
3702<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3703the "variable argument" area of a function call. It is used to implement the
3704<tt>va_arg</tt> macro in C.</p>
3705
3706<h5>Arguments:</h5>
3707
3708<p>This instruction takes a <tt>va_list*</tt> value and the type of
3709the argument. It returns a value of the specified argument type and
3710increments the <tt>va_list</tt> to point to the next argument. The
3711actual type of <tt>va_list</tt> is target specific.</p>
3712
3713<h5>Semantics:</h5>
3714
3715<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3716type from the specified <tt>va_list</tt> and causes the
3717<tt>va_list</tt> to point to the next argument. For more information,
3718see the variable argument handling <a href="#int_varargs">Intrinsic
3719Functions</a>.</p>
3720
3721<p>It is legal for this instruction to be called in a function which does not
3722take a variable number of arguments, for example, the <tt>vfprintf</tt>
3723function.</p>
3724
3725<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3726href="#intrinsics">intrinsic function</a> because it takes a type as an
3727argument.</p>
3728
3729<h5>Example:</h5>
3730
3731<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3732
3733</div>
3734
3735<!-- *********************************************************************** -->
3736<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3737<!-- *********************************************************************** -->
3738
3739<div class="doc_text">
3740
3741<p>LLVM supports the notion of an "intrinsic function". These functions have
3742well known names and semantics and are required to follow certain restrictions.
3743Overall, these intrinsics represent an extension mechanism for the LLVM
3744language that does not require changing all of the transformations in LLVM when
3745adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
3746
3747<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3748prefix is reserved in LLVM for intrinsic names; thus, function names may not
3749begin with this prefix. Intrinsic functions must always be external functions:
3750you cannot define the body of intrinsic functions. Intrinsic functions may
3751only be used in call or invoke instructions: it is illegal to take the address
3752of an intrinsic function. Additionally, because intrinsic functions are part
3753of the LLVM language, it is required if any are added that they be documented
3754here.</p>
3755
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]3756<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
3757a family of functions that perform the same operation but on different data
3758types. Because LLVM can represent over 8 million different integer types,
3759overloading is used commonly to allow an intrinsic function to operate on any
3760integer type. One or more of the argument types or the result type can be
3761overloaded to accept any integer type. Argument types may also be defined as
3762exactly matching a previous argument's type or the result type. This allows an
3763intrinsic function which accepts multiple arguments, but needs all of them to
3764be of the same type, to only be overloaded with respect to a single argument or
3765the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3766
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]3767<p>Overloaded intrinsics will have the names of its overloaded argument types
3768encoded into its function name, each preceded by a period. Only those types
3769which are overloaded result in a name suffix. Arguments whose type is matched
3770against another type do not. For example, the <tt>llvm.ctpop</tt> function can
3771take an integer of any width and returns an integer of exactly the same integer
3772width. This leads to a family of functions such as
3773<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
3774Only one type, the return type, is overloaded, and only one type suffix is
3775required. Because the argument's type is matched against the return type, it
3776does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3777
3778<p>To learn how to add an intrinsic function, please see the
3779<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3780</p>
3781
3782</div>
3783
3784<!-- ======================================================================= -->
3785<div class="doc_subsection">
3786 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3787</div>
3788
3789<div class="doc_text">
3790
3791<p>Variable argument support is defined in LLVM with the <a
3792 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3793intrinsic functions. These functions are related to the similarly
3794named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
3795
3796<p>All of these functions operate on arguments that use a
3797target-specific value type "<tt>va_list</tt>". The LLVM assembly
3798language reference manual does not define what this type is, so all
3799transformations should be prepared to handle these functions regardless of
3800the type used.</p>
3801
3802<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3803instruction and the variable argument handling intrinsic functions are
3804used.</p>
3805
3806<div class="doc_code">
3807<pre>
3808define i32 @test(i32 %X, ...) {
3809 ; Initialize variable argument processing
3810 %ap = alloca i8*
3811 %ap2 = bitcast i8** %ap to i8*
3812 call void @llvm.va_start(i8* %ap2)
3813
3814 ; Read a single integer argument
3815 %tmp = va_arg i8** %ap, i32
3816
3817 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3818 %aq = alloca i8*
3819 %aq2 = bitcast i8** %aq to i8*
3820 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
3821 call void @llvm.va_end(i8* %aq2)
3822
3823 ; Stop processing of arguments.
3824 call void @llvm.va_end(i8* %ap2)
3825 ret i32 %tmp
3826}
3827
3828declare void @llvm.va_start(i8*)
3829declare void @llvm.va_copy(i8*, i8*)
3830declare void @llvm.va_end(i8*)
3831</pre>
3832</div>
3833
3834</div>
3835
3836<!-- _______________________________________________________________________ -->
3837<div class="doc_subsubsection">
3838 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3839</div>
3840
3841
3842<div class="doc_text">
3843<h5>Syntax:</h5>
3844<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
3845<h5>Overview:</h5>
3846<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3847<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
3848href="#i_va_arg">va_arg</a></tt>.</p>
3849
3850<h5>Arguments:</h5>
3851
3852<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3853
3854<h5>Semantics:</h5>
3855
3856<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3857macro available in C. In a target-dependent way, it initializes the
3858<tt>va_list</tt> element to which the argument points, so that the next call to
3859<tt>va_arg</tt> will produce the first variable argument passed to the function.
3860Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3861last argument of the function as the compiler can figure that out.</p>
3862
3863</div>
3864
3865<!-- _______________________________________________________________________ -->
3866<div class="doc_subsubsection">
3867 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3868</div>
3869
3870<div class="doc_text">
3871<h5>Syntax:</h5>
3872<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
3873<h5>Overview:</h5>
3874
3875<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
3876which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
3877or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3878
3879<h5>Arguments:</h5>
3880
3881<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
3882
3883<h5>Semantics:</h5>
3884
3885<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3886macro available in C. In a target-dependent way, it destroys the
3887<tt>va_list</tt> element to which the argument points. Calls to <a
3888href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
3889<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
3890<tt>llvm.va_end</tt>.</p>
3891
3892</div>
3893
3894<!-- _______________________________________________________________________ -->
3895<div class="doc_subsubsection">
3896 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3897</div>
3898
3899<div class="doc_text">
3900
3901<h5>Syntax:</h5>
3902
3903<pre>
3904 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
3905</pre>
3906
3907<h5>Overview:</h5>
3908
3909<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
3910from the source argument list to the destination argument list.</p>
3911
3912<h5>Arguments:</h5>
3913
3914<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3915The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3916
3917
3918<h5>Semantics:</h5>
3919
3920<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
3921macro available in C. In a target-dependent way, it copies the source
3922<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
3923intrinsic is necessary because the <tt><a href="#int_va_start">
3924llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
3925example, memory allocation.</p>
3926
3927</div>
3928
3929<!-- ======================================================================= -->
3930<div class="doc_subsection">
3931 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3932</div>
3933
3934<div class="doc_text">
3935
3936<p>
3937LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3938Collection</a> requires the implementation and generation of these intrinsics.
3939These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
3940stack</a>, as well as garbage collector implementations that require <a
3941href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
3942Front-ends for type-safe garbage collected languages should generate these
3943intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3944href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3945</p>
3946</div>
3947
3948<!-- _______________________________________________________________________ -->
3949<div class="doc_subsubsection">
3950 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3951</div>
3952
3953<div class="doc_text">
3954
3955<h5>Syntax:</h5>
3956
3957<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]3958 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3959</pre>
3960
3961<h5>Overview:</h5>
3962
3963<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3964the code generator, and allows some metadata to be associated with it.</p>
3965
3966<h5>Arguments:</h5>
3967
3968<p>The first argument specifies the address of a stack object that contains the
3969root pointer. The second pointer (which must be either a constant or a global
3970value address) contains the meta-data to be associated with the root.</p>
3971
3972<h5>Semantics:</h5>
3973
3974<p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3975location. At compile-time, the code generator generates information to allow
3976the runtime to find the pointer at GC safe points.
3977</p>
3978
3979</div>
3980
3981
3982<!-- _______________________________________________________________________ -->
3983<div class="doc_subsubsection">
3984 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
3985</div>
3986
3987<div class="doc_text">
3988
3989<h5>Syntax:</h5>
3990
3991<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]3992 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3993</pre>
3994
3995<h5>Overview:</h5>
3996
3997<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
3998locations, allowing garbage collector implementations that require read
3999barriers.</p>
4000
4001<h5>Arguments:</h5>
4002
4003<p>The second argument is the address to read from, which should be an address
4004allocated from the garbage collector. The first object is a pointer to the
4005start of the referenced object, if needed by the language runtime (otherwise
4006null).</p>
4007
4008<h5>Semantics:</h5>
4009
4010<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4011instruction, but may be replaced with substantially more complex code by the
4012garbage collector runtime, as needed.</p>
4013
4014</div>
4015
4016
4017<!-- _______________________________________________________________________ -->
4018<div class="doc_subsubsection">
4019 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4020</div>
4021
4022<div class="doc_text">
4023
4024<h5>Syntax:</h5>
4025
4026<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4027 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4028</pre>
4029
4030<h5>Overview:</h5>
4031
4032<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4033locations, allowing garbage collector implementations that require write
4034barriers (such as generational or reference counting collectors).</p>
4035
4036<h5>Arguments:</h5>
4037
4038<p>The first argument is the reference to store, the second is the start of the
4039object to store it to, and the third is the address of the field of Obj to
4040store to. If the runtime does not require a pointer to the object, Obj may be
4041null.</p>
4042
4043<h5>Semantics:</h5>
4044
4045<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4046instruction, but may be replaced with substantially more complex code by the
4047garbage collector runtime, as needed.</p>
4048
4049</div>
4050
4051
4052
4053<!-- ======================================================================= -->
4054<div class="doc_subsection">
4055 <a name="int_codegen">Code Generator Intrinsics</a>
4056</div>
4057
4058<div class="doc_text">
4059<p>
4060These intrinsics are provided by LLVM to expose special features that may only
4061be implemented with code generator support.
4062</p>
4063
4064</div>
4065
4066<!-- _______________________________________________________________________ -->
4067<div class="doc_subsubsection">
4068 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4069</div>
4070
4071<div class="doc_text">
4072
4073<h5>Syntax:</h5>
4074<pre>
4075 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
4076</pre>
4077
4078<h5>Overview:</h5>
4079
4080<p>
4081The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4082target-specific value indicating the return address of the current function
4083or one of its callers.
4084</p>
4085
4086<h5>Arguments:</h5>
4087
4088<p>
4089The argument to this intrinsic indicates which function to return the address
4090for. Zero indicates the calling function, one indicates its caller, etc. The
4091argument is <b>required</b> to be a constant integer value.
4092</p>
4093
4094<h5>Semantics:</h5>
4095
4096<p>
4097The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4098the return address of the specified call frame, or zero if it cannot be
4099identified. The value returned by this intrinsic is likely to be incorrect or 0
4100for arguments other than zero, so it should only be used for debugging purposes.
4101</p>
4102
4103<p>
4104Note that calling this intrinsic does not prevent function inlining or other
4105aggressive transformations, so the value returned may not be that of the obvious
4106source-language caller.
4107</p>
4108</div>
4109
4110
4111<!-- _______________________________________________________________________ -->
4112<div class="doc_subsubsection">
4113 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4114</div>
4115
4116<div class="doc_text">
4117
4118<h5>Syntax:</h5>
4119<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4120 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4121</pre>
4122
4123<h5>Overview:</h5>
4124
4125<p>
4126The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4127target-specific frame pointer value for the specified stack frame.
4128</p>
4129
4130<h5>Arguments:</h5>
4131
4132<p>
4133The argument to this intrinsic indicates which function to return the frame
4134pointer for. Zero indicates the calling function, one indicates its caller,
4135etc. The argument is <b>required</b> to be a constant integer value.
4136</p>
4137
4138<h5>Semantics:</h5>
4139
4140<p>
4141The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4142the frame address of the specified call frame, or zero if it cannot be
4143identified. The value returned by this intrinsic is likely to be incorrect or 0
4144for arguments other than zero, so it should only be used for debugging purposes.
4145</p>
4146
4147<p>
4148Note that calling this intrinsic does not prevent function inlining or other
4149aggressive transformations, so the value returned may not be that of the obvious
4150source-language caller.
4151</p>
4152</div>
4153
4154<!-- _______________________________________________________________________ -->
4155<div class="doc_subsubsection">
4156 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4157</div>
4158
4159<div class="doc_text">
4160
4161<h5>Syntax:</h5>
4162<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4163 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4164</pre>
4165
4166<h5>Overview:</h5>
4167
4168<p>
4169The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4170the function stack, for use with <a href="#int_stackrestore">
4171<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4172features like scoped automatic variable sized arrays in C99.
4173</p>
4174
4175<h5>Semantics:</h5>
4176
4177<p>
4178This intrinsic returns a opaque pointer value that can be passed to <a
4179href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4180<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4181<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4182state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4183practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4184that were allocated after the <tt>llvm.stacksave</tt> was executed.
4185</p>
4186
4187</div>
4188
4189<!-- _______________________________________________________________________ -->
4190<div class="doc_subsubsection">
4191 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4192</div>
4193
4194<div class="doc_text">
4195
4196<h5>Syntax:</h5>
4197<pre>
4198 declare void @llvm.stackrestore(i8 * %ptr)
4199</pre>
4200
4201<h5>Overview:</h5>
4202
4203<p>
4204The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4205the function stack to the state it was in when the corresponding <a
4206href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4207useful for implementing language features like scoped automatic variable sized
4208arrays in C99.
4209</p>
4210
4211<h5>Semantics:</h5>
4212
4213<p>
4214See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4215</p>
4216
4217</div>
4218
4219
4220<!-- _______________________________________________________________________ -->
4221<div class="doc_subsubsection">
4222 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4223</div>
4224
4225<div class="doc_text">
4226
4227<h5>Syntax:</h5>
4228<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4229 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]4230</pre>
4231
4232<h5>Overview:</h5>
4233
4234
4235<p>
4236The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4237a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4238no
4239effect on the behavior of the program but can change its performance
4240characteristics.
4241</p>
4242
4243<h5>Arguments:</h5>
4244
4245<p>
4246<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4247determining if the fetch should be for a read (0) or write (1), and
4248<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4249locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4250<tt>locality</tt> arguments must be constant integers.
4251</p>
4252
4253<h5>Semantics:</h5>
4254
4255<p>
4256This intrinsic does not modify the behavior of the program. In particular,
4257prefetches cannot trap and do not produce a value. On targets that support this
4258intrinsic, the prefetch can provide hints to the processor cache for better
4259performance.
4260</p>
4261
4262</div>
4263
4264<!-- _______________________________________________________________________ -->
4265<div class="doc_subsubsection">
4266 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4267</div>
4268
4269<div class="doc_text">
4270
4271<h5>Syntax:</h5>
4272<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4273 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4274</pre>
4275
4276<h5>Overview:</h5>
4277
4278
4279<p>
4280The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4281(PC) in a region of
4282code to simulators and other tools. The method is target specific, but it is
4283expected that the marker will use exported symbols to transmit the PC of the marker.
4284The marker makes no guarantees that it will remain with any specific instruction
4285after optimizations. It is possible that the presence of a marker will inhibit
4286optimizations. The intended use is to be inserted after optimizations to allow
4287correlations of simulation runs.
4288</p>
4289
4290<h5>Arguments:</h5>
4291
4292<p>
4293<tt>id</tt> is a numerical id identifying the marker.
4294</p>
4295
4296<h5>Semantics:</h5>
4297
4298<p>
4299This intrinsic does not modify the behavior of the program. Backends that do not
4300support this intrinisic may ignore it.
4301</p>
4302
4303</div>
4304
4305<!-- _______________________________________________________________________ -->
4306<div class="doc_subsubsection">
4307 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4308</div>
4309
4310<div class="doc_text">
4311
4312<h5>Syntax:</h5>
4313<pre>
4314 declare i64 @llvm.readcyclecounter( )
4315</pre>
4316
4317<h5>Overview:</h5>
4318
4319
4320<p>
4321The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4322counter register (or similar low latency, high accuracy clocks) on those targets
4323that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4324As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4325should only be used for small timings.
4326</p>
4327
4328<h5>Semantics:</h5>
4329
4330<p>
4331When directly supported, reading the cycle counter should not modify any memory.
4332Implementations are allowed to either return a application specific value or a
4333system wide value. On backends without support, this is lowered to a constant 0.
4334</p>
4335
4336</div>
4337
4338<!-- ======================================================================= -->
4339<div class="doc_subsection">
4340 <a name="int_libc">Standard C Library Intrinsics</a>
4341</div>
4342
4343<div class="doc_text">
4344<p>
4345LLVM provides intrinsics for a few important standard C library functions.
4346These intrinsics allow source-language front-ends to pass information about the
4347alignment of the pointer arguments to the code generator, providing opportunity
4348for more efficient code generation.
4349</p>
4350
4351</div>
4352
4353<!-- _______________________________________________________________________ -->
4354<div class="doc_subsubsection">
4355 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4356</div>
4357
4358<div class="doc_text">
4359
4360<h5>Syntax:</h5>
4361<pre>
4362 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4363 i32 &lt;len&gt;, i32 &lt;align&gt;)
4364 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4365 i64 &lt;len&gt;, i32 &lt;align&gt;)
4366</pre>
4367
4368<h5>Overview:</h5>
4369
4370<p>
4371The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4372location to the destination location.
4373</p>
4374
4375<p>
4376Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4377intrinsics do not return a value, and takes an extra alignment argument.
4378</p>
4379
4380<h5>Arguments:</h5>
4381
4382<p>
4383The first argument is a pointer to the destination, the second is a pointer to
4384the source. The third argument is an integer argument
4385specifying the number of bytes to copy, and the fourth argument is the alignment
4386of the source and destination locations.
4387</p>
4388
4389<p>
4390If the call to this intrinisic has an alignment value that is not 0 or 1, then
4391the caller guarantees that both the source and destination pointers are aligned
4392to that boundary.
4393</p>
4394
4395<h5>Semantics:</h5>
4396
4397<p>
4398The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4399location to the destination location, which are not allowed to overlap. It
4400copies "len" bytes of memory over. If the argument is known to be aligned to
4401some boundary, this can be specified as the fourth argument, otherwise it should
4402be set to 0 or 1.
4403</p>
4404</div>
4405
4406
4407<!-- _______________________________________________________________________ -->
4408<div class="doc_subsubsection">
4409 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4410</div>
4411
4412<div class="doc_text">
4413
4414<h5>Syntax:</h5>
4415<pre>
4416 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4417 i32 &lt;len&gt;, i32 &lt;align&gt;)
4418 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4419 i64 &lt;len&gt;, i32 &lt;align&gt;)
4420</pre>
4421
4422<h5>Overview:</h5>
4423
4424<p>
4425The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4426location to the destination location. It is similar to the
4427'<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4428</p>
4429
4430<p>
4431Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4432intrinsics do not return a value, and takes an extra alignment argument.
4433</p>
4434
4435<h5>Arguments:</h5>
4436
4437<p>
4438The first argument is a pointer to the destination, the second is a pointer to
4439the source. The third argument is an integer argument
4440specifying the number of bytes to copy, and the fourth argument is the alignment
4441of the source and destination locations.
4442</p>
4443
4444<p>
4445If the call to this intrinisic has an alignment value that is not 0 or 1, then
4446the caller guarantees that the source and destination pointers are aligned to
4447that boundary.
4448</p>
4449
4450<h5>Semantics:</h5>
4451
4452<p>
4453The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4454location to the destination location, which may overlap. It
4455copies "len" bytes of memory over. If the argument is known to be aligned to
4456some boundary, this can be specified as the fourth argument, otherwise it should
4457be set to 0 or 1.
4458</p>
4459</div>
4460
4461
4462<!-- _______________________________________________________________________ -->
4463<div class="doc_subsubsection">
4464 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4465</div>
4466
4467<div class="doc_text">
4468
4469<h5>Syntax:</h5>
4470<pre>
4471 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
4472 i32 &lt;len&gt;, i32 &lt;align&gt;)
4473 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
4474 i64 &lt;len&gt;, i32 &lt;align&gt;)
4475</pre>
4476
4477<h5>Overview:</h5>
4478
4479<p>
4480The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4481byte value.
4482</p>
4483
4484<p>
4485Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4486does not return a value, and takes an extra alignment argument.
4487</p>
4488
4489<h5>Arguments:</h5>
4490
4491<p>
4492The first argument is a pointer to the destination to fill, the second is the
4493byte value to fill it with, the third argument is an integer
4494argument specifying the number of bytes to fill, and the fourth argument is the
4495known alignment of destination location.
4496</p>
4497
4498<p>
4499If the call to this intrinisic has an alignment value that is not 0 or 1, then
4500the caller guarantees that the destination pointer is aligned to that boundary.
4501</p>
4502
4503<h5>Semantics:</h5>
4504
4505<p>
4506The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4507the
4508destination location. If the argument is known to be aligned to some boundary,
4509this can be specified as the fourth argument, otherwise it should be set to 0 or
45101.
4511</p>
4512</div>
4513
4514
4515<!-- _______________________________________________________________________ -->
4516<div class="doc_subsubsection">
4517 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4518</div>
4519
4520<div class="doc_text">
4521
4522<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4523<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]4524floating point or vector of floating point type. Not all targets support all
4525types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4526<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4527 declare float @llvm.sqrt.f32(float %Val)
4528 declare double @llvm.sqrt.f64(double %Val)
4529 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
4530 declare fp128 @llvm.sqrt.f128(fp128 %Val)
4531 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4532</pre>
4533
4534<h5>Overview:</h5>
4535
4536<p>
4537The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4538returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4539<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4540negative numbers (which allows for better optimization).
4541</p>
4542
4543<h5>Arguments:</h5>
4544
4545<p>
4546The argument and return value are floating point numbers of the same type.
4547</p>
4548
4549<h5>Semantics:</h5>
4550
4551<p>
4552This function returns the sqrt of the specified operand if it is a nonnegative
4553floating point number.
4554</p>
4555</div>
4556
4557<!-- _______________________________________________________________________ -->
4558<div class="doc_subsubsection">
4559 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4560</div>
4561
4562<div class="doc_text">
4563
4564<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4565<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]4566floating point or vector of floating point type. Not all targets support all
4567types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4568<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4569 declare float @llvm.powi.f32(float %Val, i32 %power)
4570 declare double @llvm.powi.f64(double %Val, i32 %power)
4571 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
4572 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
4573 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4574</pre>
4575
4576<h5>Overview:</h5>
4577
4578<p>
4579The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4580specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4581multiplications is not defined. When a vector of floating point type is
4582used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4583</p>
4584
4585<h5>Arguments:</h5>
4586
4587<p>
4588The second argument is an integer power, and the first is a value to raise to
4589that power.
4590</p>
4591
4592<h5>Semantics:</h5>
4593
4594<p>
4595This function returns the first value raised to the second power with an
4596unspecified sequence of rounding operations.</p>
4597</div>
4598
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4599<!-- _______________________________________________________________________ -->
4600<div class="doc_subsubsection">
4601 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
4602</div>
4603
4604<div class="doc_text">
4605
4606<h5>Syntax:</h5>
4607<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
4608floating point or vector of floating point type. Not all targets support all
4609types however.
4610<pre>
4611 declare float @llvm.sin.f32(float %Val)
4612 declare double @llvm.sin.f64(double %Val)
4613 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
4614 declare fp128 @llvm.sin.f128(fp128 %Val)
4615 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
4616</pre>
4617
4618<h5>Overview:</h5>
4619
4620<p>
4621The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
4622</p>
4623
4624<h5>Arguments:</h5>
4625
4626<p>
4627The argument and return value are floating point numbers of the same type.
4628</p>
4629
4630<h5>Semantics:</h5>
4631
4632<p>
4633This function returns the sine of the specified operand, returning the
4634same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4635conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4636</div>
4637
4638<!-- _______________________________________________________________________ -->
4639<div class="doc_subsubsection">
4640 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
4641</div>
4642
4643<div class="doc_text">
4644
4645<h5>Syntax:</h5>
4646<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
4647floating point or vector of floating point type. Not all targets support all
4648types however.
4649<pre>
4650 declare float @llvm.cos.f32(float %Val)
4651 declare double @llvm.cos.f64(double %Val)
4652 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
4653 declare fp128 @llvm.cos.f128(fp128 %Val)
4654 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
4655</pre>
4656
4657<h5>Overview:</h5>
4658
4659<p>
4660The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
4661</p>
4662
4663<h5>Arguments:</h5>
4664
4665<p>
4666The argument and return value are floating point numbers of the same type.
4667</p>
4668
4669<h5>Semantics:</h5>
4670
4671<p>
4672This function returns the cosine of the specified operand, returning the
4673same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4674conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4675</div>
4676
4677<!-- _______________________________________________________________________ -->
4678<div class="doc_subsubsection">
4679 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
4680</div>
4681
4682<div class="doc_text">
4683
4684<h5>Syntax:</h5>
4685<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
4686floating point or vector of floating point type. Not all targets support all
4687types however.
4688<pre>
4689 declare float @llvm.pow.f32(float %Val, float %Power)
4690 declare double @llvm.pow.f64(double %Val, double %Power)
4691 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
4692 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
4693 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
4694</pre>
4695
4696<h5>Overview:</h5>
4697
4698<p>
4699The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
4700specified (positive or negative) power.
4701</p>
4702
4703<h5>Arguments:</h5>
4704
4705<p>
4706The second argument is a floating point power, and the first is a value to
4707raise to that power.
4708</p>
4709
4710<h5>Semantics:</h5>
4711
4712<p>
4713This function returns the first value raised to the second power,
4714returning the
4715same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4716conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4717</div>
4718
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4719
4720<!-- ======================================================================= -->
4721<div class="doc_subsection">
4722 <a name="int_manip">Bit Manipulation Intrinsics</a>
4723</div>
4724
4725<div class="doc_text">
4726<p>
4727LLVM provides intrinsics for a few important bit manipulation operations.
4728These allow efficient code generation for some algorithms.
4729</p>
4730
4731</div>
4732
4733<!-- _______________________________________________________________________ -->
4734<div class="doc_subsubsection">
4735 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4736</div>
4737
4738<div class="doc_text">
4739
4740<h5>Syntax:</h5>
4741<p>This is an overloaded intrinsic function. You can use bswap on any integer
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4742type that is an even number of bytes (i.e. BitWidth % 16 == 0).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4743<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4744 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
4745 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
4746 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4747</pre>
4748
4749<h5>Overview:</h5>
4750
4751<p>
4752The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
4753values with an even number of bytes (positive multiple of 16 bits). These are
4754useful for performing operations on data that is not in the target's native
4755byte order.
4756</p>
4757
4758<h5>Semantics:</h5>
4759
4760<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4761The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4762and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4763intrinsic returns an i32 value that has the four bytes of the input i32
4764swapped, 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]4765i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
4766<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4767additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4768</p>
4769
4770</div>
4771
4772<!-- _______________________________________________________________________ -->
4773<div class="doc_subsubsection">
4774 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4775</div>
4776
4777<div class="doc_text">
4778
4779<h5>Syntax:</h5>
4780<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4781width. Not all targets support all bit widths however.
4782<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4783 declare i8 @llvm.ctpop.i8 (i8 &lt;src&gt;)
4784 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4785 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4786 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
4787 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4788</pre>
4789
4790<h5>Overview:</h5>
4791
4792<p>
4793The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4794value.
4795</p>
4796
4797<h5>Arguments:</h5>
4798
4799<p>
4800The only argument is the value to be counted. The argument may be of any
4801integer type. The return type must match the argument type.
4802</p>
4803
4804<h5>Semantics:</h5>
4805
4806<p>
4807The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4808</p>
4809</div>
4810
4811<!-- _______________________________________________________________________ -->
4812<div class="doc_subsubsection">
4813 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4814</div>
4815
4816<div class="doc_text">
4817
4818<h5>Syntax:</h5>
4819<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4820integer bit width. Not all targets support all bit widths however.
4821<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4822 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
4823 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4824 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4825 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
4826 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4827</pre>
4828
4829<h5>Overview:</h5>
4830
4831<p>
4832The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4833leading zeros in a variable.
4834</p>
4835
4836<h5>Arguments:</h5>
4837
4838<p>
4839The only argument is the value to be counted. The argument may be of any
4840integer type. The return type must match the argument type.
4841</p>
4842
4843<h5>Semantics:</h5>
4844
4845<p>
4846The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4847in a variable. If the src == 0 then the result is the size in bits of the type
4848of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4849</p>
4850</div>
4851
4852
4853
4854<!-- _______________________________________________________________________ -->
4855<div class="doc_subsubsection">
4856 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4857</div>
4858
4859<div class="doc_text">
4860
4861<h5>Syntax:</h5>
4862<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
4863integer bit width. Not all targets support all bit widths however.
4864<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4865 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
4866 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4867 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4868 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
4869 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4870</pre>
4871
4872<h5>Overview:</h5>
4873
4874<p>
4875The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4876trailing zeros.
4877</p>
4878
4879<h5>Arguments:</h5>
4880
4881<p>
4882The only argument is the value to be counted. The argument may be of any
4883integer type. The return type must match the argument type.
4884</p>
4885
4886<h5>Semantics:</h5>
4887
4888<p>
4889The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4890in a variable. If the src == 0 then the result is the size in bits of the type
4891of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4892</p>
4893</div>
4894
4895<!-- _______________________________________________________________________ -->
4896<div class="doc_subsubsection">
4897 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
4898</div>
4899
4900<div class="doc_text">
4901
4902<h5>Syntax:</h5>
4903<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
4904on any integer bit width.
4905<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4906 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
4907 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4908</pre>
4909
4910<h5>Overview:</h5>
4911<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
4912range of bits from an integer value and returns them in the same bit width as
4913the original value.</p>
4914
4915<h5>Arguments:</h5>
4916<p>The first argument, <tt>%val</tt> and the result may be integer types of
4917any bit width but they must have the same bit width. The second and third
4918arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
4919
4920<h5>Semantics:</h5>
4921<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
4922of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
4923<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
4924operates in forward mode.</p>
4925<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
4926right by <tt>%loBit</tt> bits and then ANDing it with a mask with
4927only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
4928<ol>
4929 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
4930 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
4931 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
4932 to determine the number of bits to retain.</li>
4933 <li>A mask of the retained bits is created by shifting a -1 value.</li>
4934 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
4935</ol>
4936<p>In reverse mode, a similar computation is made except that the bits are
4937returned in the reverse order. So, for example, if <tt>X</tt> has the value
4938<tt>i16 0x0ACF (101011001111)</tt> and we apply
4939<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
4940<tt>i16 0x0026 (000000100110)</tt>.</p>
4941</div>
4942
4943<div class="doc_subsubsection">
4944 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
4945</div>
4946
4947<div class="doc_text">
4948
4949<h5>Syntax:</h5>
4950<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
4951on any integer bit width.
4952<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4953 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
4954 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]4955</pre>
4956
4957<h5>Overview:</h5>
4958<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
4959of bits in an integer value with another integer value. It returns the integer
4960with the replaced bits.</p>
4961
4962<h5>Arguments:</h5>
4963<p>The first argument, <tt>%val</tt> and the result may be integer types of
4964any bit width but they must have the same bit width. <tt>%val</tt> is the value
4965whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
4966integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
4967type since they specify only a bit index.</p>
4968
4969<h5>Semantics:</h5>
4970<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
4971of operation: forwards and reverse. If <tt>%lo</tt> is greater than
4972<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
4973operates in forward mode.</p>
4974<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
4975truncating it down to the size of the replacement area or zero extending it
4976up to that size.</p>
4977<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
4978are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
4979in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
4980to the <tt>%hi</tt>th bit.
4981<p>In reverse mode, a similar computation is made except that the bits are
4982reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
4983<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
4984<h5>Examples:</h5>
4985<pre>
4986 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
4987 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
4988 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
4989 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
4990 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
4991</pre>
4992</div>
4993
4994<!-- ======================================================================= -->
4995<div class="doc_subsection">
4996 <a name="int_debugger">Debugger Intrinsics</a>
4997</div>
4998
4999<div class="doc_text">
5000<p>
5001The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5002are described in the <a
5003href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5004Debugging</a> document.
5005</p>
5006</div>
5007
5008
5009<!-- ======================================================================= -->
5010<div class="doc_subsection">
5011 <a name="int_eh">Exception Handling Intrinsics</a>
5012</div>
5013
5014<div class="doc_text">
5015<p> The LLVM exception handling intrinsics (which all start with
5016<tt>llvm.eh.</tt> prefix), are described in the <a
5017href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5018Handling</a> document. </p>
5019</div>
5020
5021<!-- ======================================================================= -->
5022<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5023 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5024</div>
5025
5026<div class="doc_text">
5027<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5028 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5029 the <tt>nest</tt> attribute, from a function. The result is a callable
5030 function pointer lacking the nest parameter - the caller does not need
5031 to provide a value for it. Instead, the value to use is stored in
5032 advance in a "trampoline", a block of memory usually allocated
5033 on the stack, which also contains code to splice the nest value into the
5034 argument list. This is used to implement the GCC nested function address
5035 extension.
5036</p>
5037<p>
5038 For example, if the function is
5039 <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]5040 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]5041<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5042 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5043 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5044 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5045 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5046</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]5047 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5048 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5049</div>
5050
5051<!-- _______________________________________________________________________ -->
5052<div class="doc_subsubsection">
5053 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5054</div>
5055<div class="doc_text">
5056<h5>Syntax:</h5>
5057<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5058declare 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]5059</pre>
5060<h5>Overview:</h5>
5061<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5062 This fills the memory pointed to by <tt>tramp</tt> with code
5063 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5064</p>
5065<h5>Arguments:</h5>
5066<p>
5067 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5068 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5069 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]5070 intrinsic. Note that the size and the alignment are target-specific - LLVM
5071 currently provides no portable way of determining them, so a front-end that
5072 generates this intrinsic needs to have some target-specific knowledge.
5073 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]5074</p>
5075<h5>Semantics:</h5>
5076<p>
5077 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5078 dependent code, turning it into a function. A pointer to this function is
5079 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5080 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5081 before being called. The new function's signature is the same as that of
5082 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5083 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5084 of pointer type. Calling the new function is equivalent to calling
5085 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5086 missing <tt>nest</tt> argument. If, after calling
5087 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5088 modified, then the effect of any later call to the returned function pointer is
5089 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5090</p>
5091</div>
5092
5093<!-- ======================================================================= -->
5094<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5095 <a name="int_general">General Intrinsics</a>
5096</div>
5097
5098<div class="doc_text">
5099<p> This class of intrinsics is designed to be generic and has
5100no specific purpose. </p>
5101</div>
5102
5103<!-- _______________________________________________________________________ -->
5104<div class="doc_subsubsection">
5105 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
5106</div>
5107
5108<div class="doc_text">
5109
5110<h5>Syntax:</h5>
5111<pre>
5112 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5113</pre>
5114
5115<h5>Overview:</h5>
5116
5117<p>
5118The '<tt>llvm.var.annotation</tt>' intrinsic
5119</p>
5120
5121<h5>Arguments:</h5>
5122
5123<p>
5124The first argument is a pointer to a value, the second is a pointer to a
5125global string, the third is a pointer to a global string which is the source
5126file name, and the last argument is the line number.
5127</p>
5128
5129<h5>Semantics:</h5>
5130
5131<p>
5132This intrinsic allows annotation of local variables with arbitrary strings.
5133This can be useful for special purpose optimizations that want to look for these
5134 annotations. These have no other defined use, they are ignored by code
5135 generation and optimization.
5136</div>
5137
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5138<!-- _______________________________________________________________________ -->
5139<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]5140 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5141</div>
5142
5143<div class="doc_text">
5144
5145<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5146<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
5147any integer bit width.
5148</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5149<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]5150 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5151 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5152 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5153 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5154 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]5155</pre>
5156
5157<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5158
5159<p>
5160The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5161</p>
5162
5163<h5>Arguments:</h5>
5164
5165<p>
5166The first argument is an integer value (result of some expression),
5167the second is a pointer to a global string, the third is a pointer to a global
5168string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5169It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5170</p>
5171
5172<h5>Semantics:</h5>
5173
5174<p>
5175This intrinsic allows annotations to be put on arbitrary expressions
5176with arbitrary strings. This can be useful for special purpose optimizations
5177that want to look for these annotations. These have no other defined use, they
5178are ignored by code generation and optimization.
5179</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5180
5181<!-- *********************************************************************** -->
5182<hr>
5183<address>
5184 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
5185 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
5186 <a href="http://validator.w3.org/check/referer"><img
5187 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
5188
5189 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
5190 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
5191 Last modified: $Date$
5192</address>
5193</body>
5194</html>