Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / c8ddfa60b8711b95e3c46a73257cab15b3f7bbed / . / docs / LangRef.html
blob: f85a441e5317f7f1917bb0cf999001f42115407f [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
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]767 a function. Parameter attributes are considered to be part of the function,
768 not of the function type, so functions with different parameter attributes
769 can have the same function type.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]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>
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]777declare i32 @printf(i8* noalias , ...) nounwind
778declare i32 @atoi(i8*) nounwind readonly
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]779</pre>
780</div>
781
Duncan Sandsf5588dc2007-11-27 13:23:08 +0000[diff] [blame]782 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
783 <tt>readonly</tt>) come immediately after the argument list.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]784
785 <p>Currently, only the following parameter attributes are defined:</p>
786 <dl>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]787 <dt><tt>zeroext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]788 <dd>This indicates that the parameter should be zero extended just before
789 a call to this function.</dd>
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]790 <dt><tt>signext</tt></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]791 <dd>This indicates that the parameter should be sign extended just before
792 a call to this function.</dd>
793 <dt><tt>inreg</tt></dt>
794 <dd>This indicates that the parameter should be placed in register (if
795 possible) during assembling function call. Support for this attribute is
796 target-specific</dd>
797 <dt><tt>sret</tt></dt>
798 <dd>This indicates that the parameter specifies the address of a structure
799 that is the return value of the function in the source program.</dd>
800 <dt><tt>noalias</tt></dt>
801 <dd>This indicates that the parameter not alias any other object or any
802 other "noalias" objects during the function call.
803 <dt><tt>noreturn</tt></dt>
804 <dd>This function attribute indicates that the function never returns. This
805 indicates to LLVM that every call to this function should be treated as if
806 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
807 <dt><tt>nounwind</tt></dt>
808 <dd>This function attribute indicates that the function type does not use
809 the unwind instruction and does not allow stack unwinding to propagate
810 through it.</dd>
Duncan Sands4ee46812007-07-27 19:57:41 +0000[diff] [blame]811 <dt><tt>nest</tt></dt>
812 <dd>This indicates that the parameter can be excised using the
813 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
Duncan Sands13e13f82007-11-22 20:23:04 +0000[diff] [blame]814 <dt><tt>readonly</tt></dt>
Duncan Sandsd69c0e62007-11-14 21:14:02 +0000[diff] [blame]815 <dd>This function attribute indicates that the function has no side-effects
Duncan Sands13e13f82007-11-22 20:23:04 +0000[diff] [blame]816 except for producing a return value or throwing an exception. The value
817 returned must only depend on the function arguments and/or global variables.
818 It may use values obtained by dereferencing pointers.</dd>
819 <dt><tt>readnone</tt></dt>
820 <dd>A <tt>readnone</tt> function has the same restrictions as a <tt>readonly</tt>
Duncan Sandsd69c0e62007-11-14 21:14:02 +0000[diff] [blame]821 function, but in addition it is not allowed to dereference any pointer arguments
822 or global variables.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]823 </dl>
824
825</div>
826
827<!-- ======================================================================= -->
828<div class="doc_subsection">
829 <a name="moduleasm">Module-Level Inline Assembly</a>
830</div>
831
832<div class="doc_text">
833<p>
834Modules may contain "module-level inline asm" blocks, which corresponds to the
835GCC "file scope inline asm" blocks. These blocks are internally concatenated by
836LLVM and treated as a single unit, but may be separated in the .ll file if
837desired. The syntax is very simple:
838</p>
839
840<div class="doc_code">
841<pre>
842module asm "inline asm code goes here"
843module asm "more can go here"
844</pre>
845</div>
846
847<p>The strings can contain any character by escaping non-printable characters.
848 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
849 for the number.
850</p>
851
852<p>
853 The inline asm code is simply printed to the machine code .s file when
854 assembly code is generated.
855</p>
856</div>
857
858<!-- ======================================================================= -->
859<div class="doc_subsection">
860 <a name="datalayout">Data Layout</a>
861</div>
862
863<div class="doc_text">
864<p>A module may specify a target specific data layout string that specifies how
865data is to be laid out in memory. The syntax for the data layout is simply:</p>
866<pre> target datalayout = "<i>layout specification</i>"</pre>
867<p>The <i>layout specification</i> consists of a list of specifications
868separated by the minus sign character ('-'). Each specification starts with a
869letter and may include other information after the letter to define some
870aspect of the data layout. The specifications accepted are as follows: </p>
871<dl>
872 <dt><tt>E</tt></dt>
873 <dd>Specifies that the target lays out data in big-endian form. That is, the
874 bits with the most significance have the lowest address location.</dd>
875 <dt><tt>e</tt></dt>
876 <dd>Specifies that hte target lays out data in little-endian form. That is,
877 the bits with the least significance have the lowest address location.</dd>
878 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
879 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
880 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
881 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
882 too.</dd>
883 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
884 <dd>This specifies the alignment for an integer type of a given bit
885 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
886 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
887 <dd>This specifies the alignment for a vector type of a given bit
888 <i>size</i>.</dd>
889 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
890 <dd>This specifies the alignment for a floating point type of a given bit
891 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
892 (double).</dd>
893 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
894 <dd>This specifies the alignment for an aggregate type of a given bit
895 <i>size</i>.</dd>
896</dl>
897<p>When constructing the data layout for a given target, LLVM starts with a
898default set of specifications which are then (possibly) overriden by the
899specifications in the <tt>datalayout</tt> keyword. The default specifications
900are given in this list:</p>
901<ul>
902 <li><tt>E</tt> - big endian</li>
903 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
904 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
905 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
906 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
907 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
908 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
909 alignment of 64-bits</li>
910 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
911 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
912 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
913 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
914 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
915</ul>
916<p>When llvm is determining the alignment for a given type, it uses the
917following rules:
918<ol>
919 <li>If the type sought is an exact match for one of the specifications, that
920 specification is used.</li>
921 <li>If no match is found, and the type sought is an integer type, then the
922 smallest integer type that is larger than the bitwidth of the sought type is
923 used. If none of the specifications are larger than the bitwidth then the the
924 largest integer type is used. For example, given the default specifications
925 above, the i7 type will use the alignment of i8 (next largest) while both
926 i65 and i256 will use the alignment of i64 (largest specified).</li>
927 <li>If no match is found, and the type sought is a vector type, then the
928 largest vector type that is smaller than the sought vector type will be used
929 as a fall back. This happens because <128 x double> can be implemented in
930 terms of 64 <2 x double>, for example.</li>
931</ol>
932</div>
933
934<!-- *********************************************************************** -->
935<div class="doc_section"> <a name="typesystem">Type System</a> </div>
936<!-- *********************************************************************** -->
937
938<div class="doc_text">
939
940<p>The LLVM type system is one of the most important features of the
941intermediate representation. Being typed enables a number of
942optimizations to be performed on the IR directly, without having to do
943extra analyses on the side before the transformation. A strong type
944system makes it easier to read the generated code and enables novel
945analyses and transformations that are not feasible to perform on normal
946three address code representations.</p>
947
948</div>
949
950<!-- ======================================================================= -->
951<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
952<div class="doc_text">
953<p>The primitive types are the fundamental building blocks of the LLVM
954system. The current set of primitive types is as follows:</p>
955
956<table class="layout">
957 <tr class="layout">
958 <td class="left">
959 <table>
960 <tbody>
961 <tr><th>Type</th><th>Description</th></tr>
962 <tr><td><tt><a name="t_void">void</a></tt></td><td>No value</td></tr>
963 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
964 </tbody>
965 </table>
966 </td>
967 <td class="right">
968 <table>
969 <tbody>
970 <tr><th>Type</th><th>Description</th></tr>
971 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
972 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
973 </tbody>
974 </table>
975 </td>
976 </tr>
977</table>
978</div>
979
980<!-- _______________________________________________________________________ -->
981<div class="doc_subsubsection"> <a name="t_classifications">Type
982Classifications</a> </div>
983<div class="doc_text">
984<p>These different primitive types fall into a few useful
985classifications:</p>
986
987<table border="1" cellspacing="0" cellpadding="4">
988 <tbody>
989 <tr><th>Classification</th><th>Types</th></tr>
990 <tr>
991 <td><a name="t_integer">integer</a></td>
992 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
993 </tr>
994 <tr>
995 <td><a name="t_floating">floating point</a></td>
996 <td><tt>float, double</tt></td>
997 </tr>
998 <tr>
999 <td><a name="t_firstclass">first class</a></td>
1000 <td><tt>i1, ..., float, double, <br/>
1001 <a href="#t_pointer">pointer</a>,<a href="#t_vector">vector</a></tt>
1002 </td>
1003 </tr>
1004 </tbody>
1005</table>
1006
1007<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1008most important. Values of these types are the only ones which can be
1009produced by instructions, passed as arguments, or used as operands to
1010instructions. This means that all structures and arrays must be
1011manipulated either by pointer or by component.</p>
1012</div>
1013
1014<!-- ======================================================================= -->
1015<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1016
1017<div class="doc_text">
1018
1019<p>The real power in LLVM comes from the derived types in the system.
1020This is what allows a programmer to represent arrays, functions,
1021pointers, and other useful types. Note that these derived types may be
1022recursive: For example, it is possible to have a two dimensional array.</p>
1023
1024</div>
1025
1026<!-- _______________________________________________________________________ -->
1027<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1028
1029<div class="doc_text">
1030
1031<h5>Overview:</h5>
1032<p>The integer type is a very simple derived type that simply specifies an
1033arbitrary bit width for the integer type desired. Any bit width from 1 bit to
10342^23-1 (about 8 million) can be specified.</p>
1035
1036<h5>Syntax:</h5>
1037
1038<pre>
1039 iN
1040</pre>
1041
1042<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1043value.</p>
1044
1045<h5>Examples:</h5>
1046<table class="layout">
1047 <tr class="layout">
1048 <td class="left">
1049 <tt>i1</tt><br/>
1050 <tt>i4</tt><br/>
1051 <tt>i8</tt><br/>
1052 <tt>i16</tt><br/>
1053 <tt>i32</tt><br/>
1054 <tt>i42</tt><br/>
1055 <tt>i64</tt><br/>
1056 <tt>i1942652</tt><br/>
1057 </td>
1058 <td class="left">
1059 A boolean integer of 1 bit<br/>
1060 A nibble sized integer of 4 bits.<br/>
1061 A byte sized integer of 8 bits.<br/>
1062 A half word sized integer of 16 bits.<br/>
1063 A word sized integer of 32 bits.<br/>
1064 An integer whose bit width is the answer. <br/>
1065 A double word sized integer of 64 bits.<br/>
1066 A really big integer of over 1 million bits.<br/>
1067 </td>
1068 </tr>
1069</table>
1070</div>
1071
1072<!-- _______________________________________________________________________ -->
1073<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1074
1075<div class="doc_text">
1076
1077<h5>Overview:</h5>
1078
1079<p>The array type is a very simple derived type that arranges elements
1080sequentially in memory. The array type requires a size (number of
1081elements) and an underlying data type.</p>
1082
1083<h5>Syntax:</h5>
1084
1085<pre>
1086 [&lt;# elements&gt; x &lt;elementtype&gt;]
1087</pre>
1088
1089<p>The number of elements is a constant integer value; elementtype may
1090be any type with a size.</p>
1091
1092<h5>Examples:</h5>
1093<table class="layout">
1094 <tr class="layout">
1095 <td class="left">
1096 <tt>[40 x i32 ]</tt><br/>
1097 <tt>[41 x i32 ]</tt><br/>
1098 <tt>[40 x i8]</tt><br/>
1099 </td>
1100 <td class="left">
1101 Array of 40 32-bit integer values.<br/>
1102 Array of 41 32-bit integer values.<br/>
1103 Array of 40 8-bit integer values.<br/>
1104 </td>
1105 </tr>
1106</table>
1107<p>Here are some examples of multidimensional arrays:</p>
1108<table class="layout">
1109 <tr class="layout">
1110 <td class="left">
1111 <tt>[3 x [4 x i32]]</tt><br/>
1112 <tt>[12 x [10 x float]]</tt><br/>
1113 <tt>[2 x [3 x [4 x i16]]]</tt><br/>
1114 </td>
1115 <td class="left">
1116 3x4 array of 32-bit integer values.<br/>
1117 12x10 array of single precision floating point values.<br/>
1118 2x3x4 array of 16-bit integer values.<br/>
1119 </td>
1120 </tr>
1121</table>
1122
1123<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1124length array. Normally, accesses past the end of an array are undefined in
1125LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1126As a special case, however, zero length arrays are recognized to be variable
1127length. This allows implementation of 'pascal style arrays' with the LLVM
1128type "{ i32, [0 x float]}", for example.</p>
1129
1130</div>
1131
1132<!-- _______________________________________________________________________ -->
1133<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1134<div class="doc_text">
1135<h5>Overview:</h5>
1136<p>The function type can be thought of as a function signature. It
1137consists of a return type and a list of formal parameter types.
1138Function types are usually used to build virtual function tables
1139(which are structures of pointers to functions), for indirect function
1140calls, and when defining a function.</p>
1141<p>
1142The return type of a function type cannot be an aggregate type.
1143</p>
1144<h5>Syntax:</h5>
1145<pre> &lt;returntype&gt; (&lt;parameter list&gt;)<br></pre>
1146<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
1147specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1148which indicates that the function takes a variable number of arguments.
1149Variable argument functions can access their arguments with the <a
1150 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1151<h5>Examples:</h5>
1152<table class="layout">
1153 <tr class="layout">
1154 <td class="left"><tt>i32 (i32)</tt></td>
1155 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1156 </td>
1157 </tr><tr class="layout">
Reid Spencerf234bed2007-07-19 23:13:04 +0000[diff] [blame]1158 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1159 </tt></td>
1160 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1161 an <tt>i16</tt> that should be sign extended and a
1162 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1163 <tt>float</tt>.
1164 </td>
1165 </tr><tr class="layout">
1166 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1167 <td class="left">A vararg function that takes at least one
1168 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1169 which returns an integer. This is the signature for <tt>printf</tt> in
1170 LLVM.
1171 </td>
1172 </tr>
1173</table>
1174
1175</div>
1176<!-- _______________________________________________________________________ -->
1177<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1178<div class="doc_text">
1179<h5>Overview:</h5>
1180<p>The structure type is used to represent a collection of data members
1181together in memory. The packing of the field types is defined to match
1182the ABI of the underlying processor. The elements of a structure may
1183be any type that has a size.</p>
1184<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1185and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1186field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1187instruction.</p>
1188<h5>Syntax:</h5>
1189<pre> { &lt;type list&gt; }<br></pre>
1190<h5>Examples:</h5>
1191<table class="layout">
1192 <tr class="layout">
1193 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1194 <td class="left">A triple of three <tt>i32</tt> values</td>
1195 </tr><tr class="layout">
1196 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1197 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1198 second element is a <a href="#t_pointer">pointer</a> to a
1199 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1200 an <tt>i32</tt>.</td>
1201 </tr>
1202</table>
1203</div>
1204
1205<!-- _______________________________________________________________________ -->
1206<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1207</div>
1208<div class="doc_text">
1209<h5>Overview:</h5>
1210<p>The packed structure type is used to represent a collection of data members
1211together in memory. There is no padding between fields. Further, the alignment
1212of a packed structure is 1 byte. The elements of a packed structure may
1213be any type that has a size.</p>
1214<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1215and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1216field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1217instruction.</p>
1218<h5>Syntax:</h5>
1219<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1220<h5>Examples:</h5>
1221<table class="layout">
1222 <tr class="layout">
1223 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1224 <td class="left">A triple of three <tt>i32</tt> values</td>
1225 </tr><tr class="layout">
1226 <td class="left"><tt>&lt;&nbsp;{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}&nbsp;&gt;</tt></td>
1227 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1228 second element is a <a href="#t_pointer">pointer</a> to a
1229 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1230 an <tt>i32</tt>.</td>
1231 </tr>
1232</table>
1233</div>
1234
1235<!-- _______________________________________________________________________ -->
1236<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1237<div class="doc_text">
1238<h5>Overview:</h5>
1239<p>As in many languages, the pointer type represents a pointer or
1240reference to another object, which must live in memory.</p>
1241<h5>Syntax:</h5>
1242<pre> &lt;type&gt; *<br></pre>
1243<h5>Examples:</h5>
1244<table class="layout">
1245 <tr class="layout">
1246 <td class="left">
1247 <tt>[4x i32]*</tt><br/>
1248 <tt>i32 (i32 *) *</tt><br/>
1249 </td>
1250 <td class="left">
1251 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
1252 four <tt>i32</tt> values<br/>
1253 A <a href="#t_pointer">pointer</a> to a <a
1254 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1255 <tt>i32</tt>.<br/>
1256 </td>
1257 </tr>
1258</table>
1259</div>
1260
1261<!-- _______________________________________________________________________ -->
1262<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1263<div class="doc_text">
1264
1265<h5>Overview:</h5>
1266
1267<p>A vector type is a simple derived type that represents a vector
1268of elements. Vector types are used when multiple primitive data
1269are operated in parallel using a single instruction (SIMD).
1270A vector type requires a size (number of
1271elements) and an underlying primitive data type. Vectors must have a power
1272of two length (1, 2, 4, 8, 16 ...). Vector types are
1273considered <a href="#t_firstclass">first class</a>.</p>
1274
1275<h5>Syntax:</h5>
1276
1277<pre>
1278 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1279</pre>
1280
1281<p>The number of elements is a constant integer value; elementtype may
1282be any integer or floating point type.</p>
1283
1284<h5>Examples:</h5>
1285
1286<table class="layout">
1287 <tr class="layout">
1288 <td class="left">
1289 <tt>&lt;4 x i32&gt;</tt><br/>
1290 <tt>&lt;8 x float&gt;</tt><br/>
1291 <tt>&lt;2 x i64&gt;</tt><br/>
1292 </td>
1293 <td class="left">
1294 Vector of 4 32-bit integer values.<br/>
1295 Vector of 8 floating-point values.<br/>
1296 Vector of 2 64-bit integer values.<br/>
1297 </td>
1298 </tr>
1299</table>
1300</div>
1301
1302<!-- _______________________________________________________________________ -->
1303<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1304<div class="doc_text">
1305
1306<h5>Overview:</h5>
1307
1308<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksenda0706e2007-10-14 00:34:53 +0000[diff] [blame]1309corresponds (for example) to the C notion of a forward declared structure type.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1310In LLVM, opaque types can eventually be resolved to any type (not just a
1311structure type).</p>
1312
1313<h5>Syntax:</h5>
1314
1315<pre>
1316 opaque
1317</pre>
1318
1319<h5>Examples:</h5>
1320
1321<table class="layout">
1322 <tr class="layout">
1323 <td class="left">
1324 <tt>opaque</tt>
1325 </td>
1326 <td class="left">
1327 An opaque type.<br/>
1328 </td>
1329 </tr>
1330</table>
1331</div>
1332
1333
1334<!-- *********************************************************************** -->
1335<div class="doc_section"> <a name="constants">Constants</a> </div>
1336<!-- *********************************************************************** -->
1337
1338<div class="doc_text">
1339
1340<p>LLVM has several different basic types of constants. This section describes
1341them all and their syntax.</p>
1342
1343</div>
1344
1345<!-- ======================================================================= -->
1346<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1347
1348<div class="doc_text">
1349
1350<dl>
1351 <dt><b>Boolean constants</b></dt>
1352
1353 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1354 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1355 </dd>
1356
1357 <dt><b>Integer constants</b></dt>
1358
1359 <dd>Standard integers (such as '4') are constants of the <a
1360 href="#t_integer">integer</a> type. Negative numbers may be used with
1361 integer types.
1362 </dd>
1363
1364 <dt><b>Floating point constants</b></dt>
1365
1366 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1367 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1368 notation (see below). Floating point constants must have a <a
1369 href="#t_floating">floating point</a> type. </dd>
1370
1371 <dt><b>Null pointer constants</b></dt>
1372
1373 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1374 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1375
1376</dl>
1377
1378<p>The one non-intuitive notation for constants is the optional hexadecimal form
1379of floating point constants. For example, the form '<tt>double
13800x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
13814.5e+15</tt>'. The only time hexadecimal floating point constants are required
1382(and the only time that they are generated by the disassembler) is when a
1383floating point constant must be emitted but it cannot be represented as a
1384decimal floating point number. For example, NaN's, infinities, and other
1385special values are represented in their IEEE hexadecimal format so that
1386assembly and disassembly do not cause any bits to change in the constants.</p>
1387
1388</div>
1389
1390<!-- ======================================================================= -->
1391<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1392</div>
1393
1394<div class="doc_text">
1395<p>Aggregate constants arise from aggregation of simple constants
1396and smaller aggregate constants.</p>
1397
1398<dl>
1399 <dt><b>Structure constants</b></dt>
1400
1401 <dd>Structure constants are represented with notation similar to structure
1402 type definitions (a comma separated list of elements, surrounded by braces
1403 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* %G }</tt>",
1404 where "<tt>%G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1405 must have <a href="#t_struct">structure type</a>, and the number and
1406 types of elements must match those specified by the type.
1407 </dd>
1408
1409 <dt><b>Array constants</b></dt>
1410
1411 <dd>Array constants are represented with notation similar to array type
1412 definitions (a comma separated list of elements, surrounded by square brackets
1413 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1414 constants must have <a href="#t_array">array type</a>, and the number and
1415 types of elements must match those specified by the type.
1416 </dd>
1417
1418 <dt><b>Vector constants</b></dt>
1419
1420 <dd>Vector constants are represented with notation similar to vector type
1421 definitions (a comma separated list of elements, surrounded by
1422 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
1423 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
1424 href="#t_vector">vector type</a>, and the number and types of elements must
1425 match those specified by the type.
1426 </dd>
1427
1428 <dt><b>Zero initialization</b></dt>
1429
1430 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1431 value to zero of <em>any</em> type, including scalar and aggregate types.
1432 This is often used to avoid having to print large zero initializers (e.g. for
1433 large arrays) and is always exactly equivalent to using explicit zero
1434 initializers.
1435 </dd>
1436</dl>
1437
1438</div>
1439
1440<!-- ======================================================================= -->
1441<div class="doc_subsection">
1442 <a name="globalconstants">Global Variable and Function Addresses</a>
1443</div>
1444
1445<div class="doc_text">
1446
1447<p>The addresses of <a href="#globalvars">global variables</a> and <a
1448href="#functionstructure">functions</a> are always implicitly valid (link-time)
1449constants. These constants are explicitly referenced when the <a
1450href="#identifiers">identifier for the global</a> is used and always have <a
1451href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1452file:</p>
1453
1454<div class="doc_code">
1455<pre>
1456@X = global i32 17
1457@Y = global i32 42
1458@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1459</pre>
1460</div>
1461
1462</div>
1463
1464<!-- ======================================================================= -->
1465<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1466<div class="doc_text">
1467 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1468 no specific value. Undefined values may be of any type and be used anywhere
1469 a constant is permitted.</p>
1470
1471 <p>Undefined values indicate to the compiler that the program is well defined
1472 no matter what value is used, giving the compiler more freedom to optimize.
1473 </p>
1474</div>
1475
1476<!-- ======================================================================= -->
1477<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1478</div>
1479
1480<div class="doc_text">
1481
1482<p>Constant expressions are used to allow expressions involving other constants
1483to be used as constants. Constant expressions may be of any <a
1484href="#t_firstclass">first class</a> type and may involve any LLVM operation
1485that does not have side effects (e.g. load and call are not supported). The
1486following is the syntax for constant expressions:</p>
1487
1488<dl>
1489 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1490 <dd>Truncate a constant to another type. The bit size of CST must be larger
1491 than the bit size of TYPE. Both types must be integers.</dd>
1492
1493 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1494 <dd>Zero extend a constant to another type. The bit size of CST must be
1495 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1496
1497 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1498 <dd>Sign extend a constant to another type. The bit size of CST must be
1499 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1500
1501 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1502 <dd>Truncate a floating point constant to another floating point type. The
1503 size of CST must be larger than the size of TYPE. Both types must be
1504 floating point.</dd>
1505
1506 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1507 <dd>Floating point extend a constant to another type. The size of CST must be
1508 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1509
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]1510 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1511 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1512 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1513 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1514 of the same number of elements. If the value won't fit in the integer type,
1515 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1516
1517 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1518 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1519 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1520 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1521 of the same number of elements. If the value won't fit in the integer type,
1522 the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1523
1524 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1525 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1526 constant. TYPE must be a scalar or vector floating point type. CST must be of
1527 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1528 of the same number of elements. If the value won't fit in the floating point
1529 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1530
1531 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1532 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]1533 constant. TYPE must be a scalar or vector floating point type. CST must be of
1534 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1535 of the same number of elements. If the value won't fit in the floating point
1536 type, the results are undefined.</dd>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]1537
1538 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1539 <dd>Convert a pointer typed constant to the corresponding integer constant
1540 TYPE must be an integer type. CST must be of pointer type. The CST value is
1541 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1542
1543 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1544 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1545 pointer type. CST must be of integer type. The CST value is zero extended,
1546 truncated, or unchanged to make it fit in a pointer size. This one is
1547 <i>really</i> dangerous!</dd>
1548
1549 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1550 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1551 identical (same number of bits). The conversion is done as if the CST value
1552 was stored to memory and read back as TYPE. In other words, no bits change
1553 with this operator, just the type. This can be used for conversion of
1554 vector types to any other type, as long as they have the same bit width. For
1555 pointers it is only valid to cast to another pointer type.
1556 </dd>
1557
1558 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1559
1560 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1561 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1562 instruction, the index list may have zero or more indexes, which are required
1563 to make sense for the type of "CSTPTR".</dd>
1564
1565 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1566
1567 <dd>Perform the <a href="#i_select">select operation</a> on
1568 constants.</dd>
1569
1570 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1571 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1572
1573 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1574 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1575
1576 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1577
1578 <dd>Perform the <a href="#i_extractelement">extractelement
1579 operation</a> on constants.
1580
1581 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1582
1583 <dd>Perform the <a href="#i_insertelement">insertelement
1584 operation</a> on constants.</dd>
1585
1586
1587 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1588
1589 <dd>Perform the <a href="#i_shufflevector">shufflevector
1590 operation</a> on constants.</dd>
1591
1592 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1593
1594 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1595 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1596 binary</a> operations. The constraints on operands are the same as those for
1597 the corresponding instruction (e.g. no bitwise operations on floating point
1598 values are allowed).</dd>
1599</dl>
1600</div>
1601
1602<!-- *********************************************************************** -->
1603<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1604<!-- *********************************************************************** -->
1605
1606<!-- ======================================================================= -->
1607<div class="doc_subsection">
1608<a name="inlineasm">Inline Assembler Expressions</a>
1609</div>
1610
1611<div class="doc_text">
1612
1613<p>
1614LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1615Module-Level Inline Assembly</a>) through the use of a special value. This
1616value represents the inline assembler as a string (containing the instructions
1617to emit), a list of operand constraints (stored as a string), and a flag that
1618indicates whether or not the inline asm expression has side effects. An example
1619inline assembler expression is:
1620</p>
1621
1622<div class="doc_code">
1623<pre>
1624i32 (i32) asm "bswap $0", "=r,r"
1625</pre>
1626</div>
1627
1628<p>
1629Inline assembler expressions may <b>only</b> be used as the callee operand of
1630a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1631</p>
1632
1633<div class="doc_code">
1634<pre>
1635%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1636</pre>
1637</div>
1638
1639<p>
1640Inline asms with side effects not visible in the constraint list must be marked
1641as having side effects. This is done through the use of the
1642'<tt>sideeffect</tt>' keyword, like so:
1643</p>
1644
1645<div class="doc_code">
1646<pre>
1647call void asm sideeffect "eieio", ""()
1648</pre>
1649</div>
1650
1651<p>TODO: The format of the asm and constraints string still need to be
1652documented here. Constraints on what can be done (e.g. duplication, moving, etc
1653need to be documented).
1654</p>
1655
1656</div>
1657
1658<!-- *********************************************************************** -->
1659<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1660<!-- *********************************************************************** -->
1661
1662<div class="doc_text">
1663
1664<p>The LLVM instruction set consists of several different
1665classifications of instructions: <a href="#terminators">terminator
1666instructions</a>, <a href="#binaryops">binary instructions</a>,
1667<a href="#bitwiseops">bitwise binary instructions</a>, <a
1668 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1669instructions</a>.</p>
1670
1671</div>
1672
1673<!-- ======================================================================= -->
1674<div class="doc_subsection"> <a name="terminators">Terminator
1675Instructions</a> </div>
1676
1677<div class="doc_text">
1678
1679<p>As mentioned <a href="#functionstructure">previously</a>, every
1680basic block in a program ends with a "Terminator" instruction, which
1681indicates which block should be executed after the current block is
1682finished. These terminator instructions typically yield a '<tt>void</tt>'
1683value: they produce control flow, not values (the one exception being
1684the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1685<p>There are six different terminator instructions: the '<a
1686 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1687instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1688the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1689 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1690 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1691
1692</div>
1693
1694<!-- _______________________________________________________________________ -->
1695<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1696Instruction</a> </div>
1697<div class="doc_text">
1698<h5>Syntax:</h5>
1699<pre> ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
1700 ret void <i>; Return from void function</i>
1701</pre>
1702<h5>Overview:</h5>
1703<p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1704value) from a function back to the caller.</p>
1705<p>There are two forms of the '<tt>ret</tt>' instruction: one that
1706returns a value and then causes control flow, and one that just causes
1707control flow to occur.</p>
1708<h5>Arguments:</h5>
1709<p>The '<tt>ret</tt>' instruction may return any '<a
1710 href="#t_firstclass">first class</a>' type. Notice that a function is
1711not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1712instruction inside of the function that returns a value that does not
1713match the return type of the function.</p>
1714<h5>Semantics:</h5>
1715<p>When the '<tt>ret</tt>' instruction is executed, control flow
1716returns back to the calling function's context. If the caller is a "<a
1717 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1718the instruction after the call. If the caller was an "<a
1719 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1720at the beginning of the "normal" destination block. If the instruction
1721returns a value, that value shall set the call or invoke instruction's
1722return value.</p>
1723<h5>Example:</h5>
1724<pre> ret i32 5 <i>; Return an integer value of 5</i>
1725 ret void <i>; Return from a void function</i>
1726</pre>
1727</div>
1728<!-- _______________________________________________________________________ -->
1729<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1730<div class="doc_text">
1731<h5>Syntax:</h5>
1732<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
1733</pre>
1734<h5>Overview:</h5>
1735<p>The '<tt>br</tt>' instruction is used to cause control flow to
1736transfer to a different basic block in the current function. There are
1737two forms of this instruction, corresponding to a conditional branch
1738and an unconditional branch.</p>
1739<h5>Arguments:</h5>
1740<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1741single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1742unconditional form of the '<tt>br</tt>' instruction takes a single
1743'<tt>label</tt>' value as a target.</p>
1744<h5>Semantics:</h5>
1745<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1746argument is evaluated. If the value is <tt>true</tt>, control flows
1747to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1748control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1749<h5>Example:</h5>
1750<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
1751 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1752</div>
1753<!-- _______________________________________________________________________ -->
1754<div class="doc_subsubsection">
1755 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1756</div>
1757
1758<div class="doc_text">
1759<h5>Syntax:</h5>
1760
1761<pre>
1762 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
1763</pre>
1764
1765<h5>Overview:</h5>
1766
1767<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1768several different places. It is a generalization of the '<tt>br</tt>'
1769instruction, allowing a branch to occur to one of many possible
1770destinations.</p>
1771
1772
1773<h5>Arguments:</h5>
1774
1775<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1776comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1777an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1778table is not allowed to contain duplicate constant entries.</p>
1779
1780<h5>Semantics:</h5>
1781
1782<p>The <tt>switch</tt> instruction specifies a table of values and
1783destinations. When the '<tt>switch</tt>' instruction is executed, this
1784table is searched for the given value. If the value is found, control flow is
1785transfered to the corresponding destination; otherwise, control flow is
1786transfered to the default destination.</p>
1787
1788<h5>Implementation:</h5>
1789
1790<p>Depending on properties of the target machine and the particular
1791<tt>switch</tt> instruction, this instruction may be code generated in different
1792ways. For example, it could be generated as a series of chained conditional
1793branches or with a lookup table.</p>
1794
1795<h5>Example:</h5>
1796
1797<pre>
1798 <i>; Emulate a conditional br instruction</i>
1799 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1800 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1801
1802 <i>; Emulate an unconditional br instruction</i>
1803 switch i32 0, label %dest [ ]
1804
1805 <i>; Implement a jump table:</i>
1806 switch i32 %val, label %otherwise [ i32 0, label %onzero
1807 i32 1, label %onone
1808 i32 2, label %ontwo ]
1809</pre>
1810</div>
1811
1812<!-- _______________________________________________________________________ -->
1813<div class="doc_subsubsection">
1814 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1815</div>
1816
1817<div class="doc_text">
1818
1819<h5>Syntax:</h5>
1820
1821<pre>
1822 &lt;result&gt; = invoke [<a href="#callingconv">cconv</a>] &lt;ptr to function ty&gt; %&lt;function ptr val&gt;(&lt;function args&gt;)
1823 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
1824</pre>
1825
1826<h5>Overview:</h5>
1827
1828<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1829function, with the possibility of control flow transfer to either the
1830'<tt>normal</tt>' label or the
1831'<tt>exception</tt>' label. If the callee function returns with the
1832"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1833"normal" label. If the callee (or any indirect callees) returns with the "<a
1834href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1835continued at the dynamically nearest "exception" label.</p>
1836
1837<h5>Arguments:</h5>
1838
1839<p>This instruction requires several arguments:</p>
1840
1841<ol>
1842 <li>
1843 The optional "cconv" marker indicates which <a href="#callingconv">calling
1844 convention</a> the call should use. If none is specified, the call defaults
1845 to using C calling conventions.
1846 </li>
1847 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1848 function value being invoked. In most cases, this is a direct function
1849 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1850 an arbitrary pointer to function value.
1851 </li>
1852
1853 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1854 function to be invoked. </li>
1855
1856 <li>'<tt>function args</tt>': argument list whose types match the function
1857 signature argument types. If the function signature indicates the function
1858 accepts a variable number of arguments, the extra arguments can be
1859 specified. </li>
1860
1861 <li>'<tt>normal label</tt>': the label reached when the called function
1862 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1863
1864 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1865 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1866
1867</ol>
1868
1869<h5>Semantics:</h5>
1870
1871<p>This instruction is designed to operate as a standard '<tt><a
1872href="#i_call">call</a></tt>' instruction in most regards. The primary
1873difference is that it establishes an association with a label, which is used by
1874the runtime library to unwind the stack.</p>
1875
1876<p>This instruction is used in languages with destructors to ensure that proper
1877cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1878exception. Additionally, this is important for implementation of
1879'<tt>catch</tt>' clauses in high-level languages that support them.</p>
1880
1881<h5>Example:</h5>
1882<pre>
1883 %retval = invoke i32 %Test(i32 15) to label %Continue
1884 unwind label %TestCleanup <i>; {i32}:retval set</i>
1885 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1886 unwind label %TestCleanup <i>; {i32}:retval set</i>
1887</pre>
1888</div>
1889
1890
1891<!-- _______________________________________________________________________ -->
1892
1893<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1894Instruction</a> </div>
1895
1896<div class="doc_text">
1897
1898<h5>Syntax:</h5>
1899<pre>
1900 unwind
1901</pre>
1902
1903<h5>Overview:</h5>
1904
1905<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1906at the first callee in the dynamic call stack which used an <a
1907href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1908primarily used to implement exception handling.</p>
1909
1910<h5>Semantics:</h5>
1911
1912<p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1913immediately halt. The dynamic call stack is then searched for the first <a
1914href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1915execution continues at the "exceptional" destination block specified by the
1916<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1917dynamic call chain, undefined behavior results.</p>
1918</div>
1919
1920<!-- _______________________________________________________________________ -->
1921
1922<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1923Instruction</a> </div>
1924
1925<div class="doc_text">
1926
1927<h5>Syntax:</h5>
1928<pre>
1929 unreachable
1930</pre>
1931
1932<h5>Overview:</h5>
1933
1934<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1935instruction is used to inform the optimizer that a particular portion of the
1936code is not reachable. This can be used to indicate that the code after a
1937no-return function cannot be reached, and other facts.</p>
1938
1939<h5>Semantics:</h5>
1940
1941<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1942</div>
1943
1944
1945
1946<!-- ======================================================================= -->
1947<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1948<div class="doc_text">
1949<p>Binary operators are used to do most of the computation in a
1950program. They require two operands, execute an operation on them, and
1951produce a single value. The operands might represent
1952multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
1953The result value of a binary operator is not
1954necessarily the same type as its operands.</p>
1955<p>There are several different binary operators:</p>
1956</div>
1957<!-- _______________________________________________________________________ -->
1958<div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1959Instruction</a> </div>
1960<div class="doc_text">
1961<h5>Syntax:</h5>
1962<pre> &lt;result&gt; = add &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
1963</pre>
1964<h5>Overview:</h5>
1965<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1966<h5>Arguments:</h5>
1967<p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1968 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1969 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1970Both arguments must have identical types.</p>
1971<h5>Semantics:</h5>
1972<p>The value produced is the integer or floating point sum of the two
1973operands.</p>
1974<h5>Example:</h5>
1975<pre> &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
1976</pre>
1977</div>
1978<!-- _______________________________________________________________________ -->
1979<div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1980Instruction</a> </div>
1981<div class="doc_text">
1982<h5>Syntax:</h5>
1983<pre> &lt;result&gt; = sub &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
1984</pre>
1985<h5>Overview:</h5>
1986<p>The '<tt>sub</tt>' instruction returns the difference of its two
1987operands.</p>
1988<p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1989instruction present in most other intermediate representations.</p>
1990<h5>Arguments:</h5>
1991<p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1992 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1993values.
1994This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1995Both arguments must have identical types.</p>
1996<h5>Semantics:</h5>
1997<p>The value produced is the integer or floating point difference of
1998the two operands.</p>
1999<h5>Example:</h5>
2000<pre>
2001 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2002 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2003</pre>
2004</div>
2005<!-- _______________________________________________________________________ -->
2006<div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
2007Instruction</a> </div>
2008<div class="doc_text">
2009<h5>Syntax:</h5>
2010<pre> &lt;result&gt; = mul &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2011</pre>
2012<h5>Overview:</h5>
2013<p>The '<tt>mul</tt>' instruction returns the product of its two
2014operands.</p>
2015<h5>Arguments:</h5>
2016<p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
2017 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2018values.
2019This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2020Both arguments must have identical types.</p>
2021<h5>Semantics:</h5>
2022<p>The value produced is the integer or floating point product of the
2023two operands.</p>
2024<p>Because the operands are the same width, the result of an integer
2025multiplication is the same whether the operands should be deemed unsigned or
2026signed.</p>
2027<h5>Example:</h5>
2028<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2029</pre>
2030</div>
2031<!-- _______________________________________________________________________ -->
2032<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2033</a></div>
2034<div class="doc_text">
2035<h5>Syntax:</h5>
2036<pre> &lt;result&gt; = udiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2037</pre>
2038<h5>Overview:</h5>
2039<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2040operands.</p>
2041<h5>Arguments:</h5>
2042<p>The two arguments to the '<tt>udiv</tt>' instruction must be
2043<a href="#t_integer">integer</a> values. Both arguments must have identical
2044types. This instruction can also take <a href="#t_vector">vector</a> versions
2045of the values in which case the elements must be integers.</p>
2046<h5>Semantics:</h5>
2047<p>The value produced is the unsigned integer quotient of the two operands. This
2048instruction always performs an unsigned division operation, regardless of
2049whether the arguments are unsigned or not.</p>
2050<h5>Example:</h5>
2051<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2052</pre>
2053</div>
2054<!-- _______________________________________________________________________ -->
2055<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2056</a> </div>
2057<div class="doc_text">
2058<h5>Syntax:</h5>
2059<pre> &lt;result&gt; = sdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2060</pre>
2061<h5>Overview:</h5>
2062<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2063operands.</p>
2064<h5>Arguments:</h5>
2065<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2066<a href="#t_integer">integer</a> values. Both arguments must have identical
2067types. This instruction can also take <a href="#t_vector">vector</a> versions
2068of the values in which case the elements must be integers.</p>
2069<h5>Semantics:</h5>
2070<p>The value produced is the signed integer quotient of the two operands. This
2071instruction always performs a signed division operation, regardless of whether
2072the arguments are signed or not.</p>
2073<h5>Example:</h5>
2074<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2075</pre>
2076</div>
2077<!-- _______________________________________________________________________ -->
2078<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2079Instruction</a> </div>
2080<div class="doc_text">
2081<h5>Syntax:</h5>
2082<pre> &lt;result&gt; = fdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2083</pre>
2084<h5>Overview:</h5>
2085<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2086operands.</p>
2087<h5>Arguments:</h5>
2088<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2089<a href="#t_floating">floating point</a> values. Both arguments must have
2090identical types. This instruction can also take <a href="#t_vector">vector</a>
2091versions of floating point values.</p>
2092<h5>Semantics:</h5>
2093<p>The value produced is the floating point quotient of the two operands.</p>
2094<h5>Example:</h5>
2095<pre> &lt;result&gt; = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2096</pre>
2097</div>
2098<!-- _______________________________________________________________________ -->
2099<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2100</div>
2101<div class="doc_text">
2102<h5>Syntax:</h5>
2103<pre> &lt;result&gt; = urem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2104</pre>
2105<h5>Overview:</h5>
2106<p>The '<tt>urem</tt>' instruction returns the remainder from the
2107unsigned division of its two arguments.</p>
2108<h5>Arguments:</h5>
2109<p>The two arguments to the '<tt>urem</tt>' instruction must be
2110<a href="#t_integer">integer</a> values. Both arguments must have identical
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2111types. This instruction can also take <a href="#t_vector">vector</a> versions
2112of the values in which case the elements must be integers.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2113<h5>Semantics:</h5>
2114<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2115This instruction always performs an unsigned division to get the remainder,
2116regardless of whether the arguments are unsigned or not.</p>
2117<h5>Example:</h5>
2118<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2119</pre>
2120
2121</div>
2122<!-- _______________________________________________________________________ -->
2123<div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2124Instruction</a> </div>
2125<div class="doc_text">
2126<h5>Syntax:</h5>
2127<pre> &lt;result&gt; = srem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2128</pre>
2129<h5>Overview:</h5>
2130<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2131signed division of its two operands. This instruction can also take
2132<a href="#t_vector">vector</a> versions of the values in which case
2133the elements must be integers.</p>
2134</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2135<h5>Arguments:</h5>
2136<p>The two arguments to the '<tt>srem</tt>' instruction must be
2137<a href="#t_integer">integer</a> values. Both arguments must have identical
2138types.</p>
2139<h5>Semantics:</h5>
2140<p>This instruction returns the <i>remainder</i> of a division (where the result
2141has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2142operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2143a value. For more information about the difference, see <a
2144 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2145Math Forum</a>. For a table of how this is implemented in various languages,
2146please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2147Wikipedia: modulo operation</a>.</p>
2148<h5>Example:</h5>
2149<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2150</pre>
2151
2152</div>
2153<!-- _______________________________________________________________________ -->
2154<div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2155Instruction</a> </div>
2156<div class="doc_text">
2157<h5>Syntax:</h5>
2158<pre> &lt;result&gt; = frem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2159</pre>
2160<h5>Overview:</h5>
2161<p>The '<tt>frem</tt>' instruction returns the remainder from the
2162division of its two operands.</p>
2163<h5>Arguments:</h5>
2164<p>The two arguments to the '<tt>frem</tt>' instruction must be
2165<a href="#t_floating">floating point</a> values. Both arguments must have
Dan Gohman3e3fd8c2007-11-05 23:35:22 +0000[diff] [blame]2166identical types. This instruction can also take <a href="#t_vector">vector</a>
2167versions of floating point values.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2168<h5>Semantics:</h5>
2169<p>This instruction returns the <i>remainder</i> of a division.</p>
2170<h5>Example:</h5>
2171<pre> &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2172</pre>
2173</div>
2174
2175<!-- ======================================================================= -->
2176<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2177Operations</a> </div>
2178<div class="doc_text">
2179<p>Bitwise binary operators are used to do various forms of
2180bit-twiddling in a program. They are generally very efficient
2181instructions and can commonly be strength reduced from other
2182instructions. They require two operands, execute an operation on them,
2183and produce a single value. The resulting value of the bitwise binary
2184operators is always the same type as its first operand.</p>
2185</div>
2186
2187<!-- _______________________________________________________________________ -->
2188<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2189Instruction</a> </div>
2190<div class="doc_text">
2191<h5>Syntax:</h5>
2192<pre> &lt;result&gt; = shl &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2193</pre>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2194
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2195<h5>Overview:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2196
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2197<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2198the left a specified number of bits.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2199
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2200<h5>Arguments:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2201
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2202<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2203 href="#t_integer">integer</a> type.</p>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2204
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2205<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2206
2207<p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>. If
2208<tt>var2</tt> is (statically or dynamically) equal to or larger than the number
2209of bits in <tt>var1</tt>, the result is undefined.</p>
2210
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2211<h5>Example:</h5><pre>
2212 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2213 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2214 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2215 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2216</pre>
2217</div>
2218<!-- _______________________________________________________________________ -->
2219<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2220Instruction</a> </div>
2221<div class="doc_text">
2222<h5>Syntax:</h5>
2223<pre> &lt;result&gt; = lshr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2224</pre>
2225
2226<h5>Overview:</h5>
2227<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2228operand shifted to the right a specified number of bits with zero fill.</p>
2229
2230<h5>Arguments:</h5>
2231<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2232<a href="#t_integer">integer</a> type.</p>
2233
2234<h5>Semantics:</h5>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2235
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2236<p>This instruction always performs a logical shift right operation. The most
2237significant bits of the result will be filled with zero bits after the
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2238shift. If <tt>var2</tt> is (statically or dynamically) equal to or larger than
2239the number of bits in <tt>var1</tt>, the result is undefined.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2240
2241<h5>Example:</h5>
2242<pre>
2243 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2244 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2245 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2246 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2247 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2248</pre>
2249</div>
2250
2251<!-- _______________________________________________________________________ -->
2252<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2253Instruction</a> </div>
2254<div class="doc_text">
2255
2256<h5>Syntax:</h5>
2257<pre> &lt;result&gt; = ashr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2258</pre>
2259
2260<h5>Overview:</h5>
2261<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2262operand shifted to the right a specified number of bits with sign extension.</p>
2263
2264<h5>Arguments:</h5>
2265<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2266<a href="#t_integer">integer</a> type.</p>
2267
2268<h5>Semantics:</h5>
2269<p>This instruction always performs an arithmetic shift right operation,
2270The most significant bits of the result will be filled with the sign bit
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2271of <tt>var1</tt>. If <tt>var2</tt> is (statically or dynamically) equal to or
2272larger than the number of bits in <tt>var1</tt>, the result is undefined.
2273</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2274
2275<h5>Example:</h5>
2276<pre>
2277 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2278 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2279 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2280 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerd939d9f2007-10-03 21:01:14 +0000[diff] [blame]2281 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]2282</pre>
2283</div>
2284
2285<!-- _______________________________________________________________________ -->
2286<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2287Instruction</a> </div>
2288<div class="doc_text">
2289<h5>Syntax:</h5>
2290<pre> &lt;result&gt; = and &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2291</pre>
2292<h5>Overview:</h5>
2293<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2294its two operands.</p>
2295<h5>Arguments:</h5>
2296<p>The two arguments to the '<tt>and</tt>' instruction must be <a
2297 href="#t_integer">integer</a> values. Both arguments must have
2298identical types.</p>
2299<h5>Semantics:</h5>
2300<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2301<p> </p>
2302<div style="align: center">
2303<table border="1" cellspacing="0" cellpadding="4">
2304 <tbody>
2305 <tr>
2306 <td>In0</td>
2307 <td>In1</td>
2308 <td>Out</td>
2309 </tr>
2310 <tr>
2311 <td>0</td>
2312 <td>0</td>
2313 <td>0</td>
2314 </tr>
2315 <tr>
2316 <td>0</td>
2317 <td>1</td>
2318 <td>0</td>
2319 </tr>
2320 <tr>
2321 <td>1</td>
2322 <td>0</td>
2323 <td>0</td>
2324 </tr>
2325 <tr>
2326 <td>1</td>
2327 <td>1</td>
2328 <td>1</td>
2329 </tr>
2330 </tbody>
2331</table>
2332</div>
2333<h5>Example:</h5>
2334<pre> &lt;result&gt; = and i32 4, %var <i>; yields {i32}:result = 4 &amp; %var</i>
2335 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2336 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2337</pre>
2338</div>
2339<!-- _______________________________________________________________________ -->
2340<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2341<div class="doc_text">
2342<h5>Syntax:</h5>
2343<pre> &lt;result&gt; = or &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2344</pre>
2345<h5>Overview:</h5>
2346<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2347or of its two operands.</p>
2348<h5>Arguments:</h5>
2349<p>The two arguments to the '<tt>or</tt>' instruction must be <a
2350 href="#t_integer">integer</a> values. Both arguments must have
2351identical types.</p>
2352<h5>Semantics:</h5>
2353<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2354<p> </p>
2355<div style="align: center">
2356<table border="1" cellspacing="0" cellpadding="4">
2357 <tbody>
2358 <tr>
2359 <td>In0</td>
2360 <td>In1</td>
2361 <td>Out</td>
2362 </tr>
2363 <tr>
2364 <td>0</td>
2365 <td>0</td>
2366 <td>0</td>
2367 </tr>
2368 <tr>
2369 <td>0</td>
2370 <td>1</td>
2371 <td>1</td>
2372 </tr>
2373 <tr>
2374 <td>1</td>
2375 <td>0</td>
2376 <td>1</td>
2377 </tr>
2378 <tr>
2379 <td>1</td>
2380 <td>1</td>
2381 <td>1</td>
2382 </tr>
2383 </tbody>
2384</table>
2385</div>
2386<h5>Example:</h5>
2387<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2388 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2389 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2390</pre>
2391</div>
2392<!-- _______________________________________________________________________ -->
2393<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2394Instruction</a> </div>
2395<div class="doc_text">
2396<h5>Syntax:</h5>
2397<pre> &lt;result&gt; = xor &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2398</pre>
2399<h5>Overview:</h5>
2400<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2401or of its two operands. The <tt>xor</tt> is used to implement the
2402"one's complement" operation, which is the "~" operator in C.</p>
2403<h5>Arguments:</h5>
2404<p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2405 href="#t_integer">integer</a> values. Both arguments must have
2406identical types.</p>
2407<h5>Semantics:</h5>
2408<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2409<p> </p>
2410<div style="align: center">
2411<table border="1" cellspacing="0" cellpadding="4">
2412 <tbody>
2413 <tr>
2414 <td>In0</td>
2415 <td>In1</td>
2416 <td>Out</td>
2417 </tr>
2418 <tr>
2419 <td>0</td>
2420 <td>0</td>
2421 <td>0</td>
2422 </tr>
2423 <tr>
2424 <td>0</td>
2425 <td>1</td>
2426 <td>1</td>
2427 </tr>
2428 <tr>
2429 <td>1</td>
2430 <td>0</td>
2431 <td>1</td>
2432 </tr>
2433 <tr>
2434 <td>1</td>
2435 <td>1</td>
2436 <td>0</td>
2437 </tr>
2438 </tbody>
2439</table>
2440</div>
2441<p> </p>
2442<h5>Example:</h5>
2443<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2444 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2445 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2446 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2447</pre>
2448</div>
2449
2450<!-- ======================================================================= -->
2451<div class="doc_subsection">
2452 <a name="vectorops">Vector Operations</a>
2453</div>
2454
2455<div class="doc_text">
2456
2457<p>LLVM supports several instructions to represent vector operations in a
2458target-independent manner. These instructions cover the element-access and
2459vector-specific operations needed to process vectors effectively. While LLVM
2460does directly support these vector operations, many sophisticated algorithms
2461will want to use target-specific intrinsics to take full advantage of a specific
2462target.</p>
2463
2464</div>
2465
2466<!-- _______________________________________________________________________ -->
2467<div class="doc_subsubsection">
2468 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2469</div>
2470
2471<div class="doc_text">
2472
2473<h5>Syntax:</h5>
2474
2475<pre>
2476 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
2477</pre>
2478
2479<h5>Overview:</h5>
2480
2481<p>
2482The '<tt>extractelement</tt>' instruction extracts a single scalar
2483element from a vector at a specified index.
2484</p>
2485
2486
2487<h5>Arguments:</h5>
2488
2489<p>
2490The first operand of an '<tt>extractelement</tt>' instruction is a
2491value of <a href="#t_vector">vector</a> type. The second operand is
2492an index indicating the position from which to extract the element.
2493The index may be a variable.</p>
2494
2495<h5>Semantics:</h5>
2496
2497<p>
2498The result is a scalar of the same type as the element type of
2499<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2500<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2501results are undefined.
2502</p>
2503
2504<h5>Example:</h5>
2505
2506<pre>
2507 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
2508</pre>
2509</div>
2510
2511
2512<!-- _______________________________________________________________________ -->
2513<div class="doc_subsubsection">
2514 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2515</div>
2516
2517<div class="doc_text">
2518
2519<h5>Syntax:</h5>
2520
2521<pre>
2522 &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>
2523</pre>
2524
2525<h5>Overview:</h5>
2526
2527<p>
2528The '<tt>insertelement</tt>' instruction inserts a scalar
2529element into a vector at a specified index.
2530</p>
2531
2532
2533<h5>Arguments:</h5>
2534
2535<p>
2536The first operand of an '<tt>insertelement</tt>' instruction is a
2537value of <a href="#t_vector">vector</a> type. The second operand is a
2538scalar value whose type must equal the element type of the first
2539operand. The third operand is an index indicating the position at
2540which to insert the value. The index may be a variable.</p>
2541
2542<h5>Semantics:</h5>
2543
2544<p>
2545The result is a vector of the same type as <tt>val</tt>. Its
2546element values are those of <tt>val</tt> except at position
2547<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2548exceeds the length of <tt>val</tt>, the results are undefined.
2549</p>
2550
2551<h5>Example:</h5>
2552
2553<pre>
2554 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
2555</pre>
2556</div>
2557
2558<!-- _______________________________________________________________________ -->
2559<div class="doc_subsubsection">
2560 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2561</div>
2562
2563<div class="doc_text">
2564
2565<h5>Syntax:</h5>
2566
2567<pre>
2568 &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>
2569</pre>
2570
2571<h5>Overview:</h5>
2572
2573<p>
2574The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2575from two input vectors, returning a vector of the same type.
2576</p>
2577
2578<h5>Arguments:</h5>
2579
2580<p>
2581The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2582with types that match each other and types that match the result of the
2583instruction. The third argument is a shuffle mask, which has the same number
2584of elements as the other vector type, but whose element type is always 'i32'.
2585</p>
2586
2587<p>
2588The shuffle mask operand is required to be a constant vector with either
2589constant integer or undef values.
2590</p>
2591
2592<h5>Semantics:</h5>
2593
2594<p>
2595The elements of the two input vectors are numbered from left to right across
2596both of the vectors. The shuffle mask operand specifies, for each element of
2597the result vector, which element of the two input registers the result element
2598gets. The element selector may be undef (meaning "don't care") and the second
2599operand may be undef if performing a shuffle from only one vector.
2600</p>
2601
2602<h5>Example:</h5>
2603
2604<pre>
2605 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
2606 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
2607 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
2608 &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.
2609</pre>
2610</div>
2611
2612
2613<!-- ======================================================================= -->
2614<div class="doc_subsection">
2615 <a name="memoryops">Memory Access and Addressing Operations</a>
2616</div>
2617
2618<div class="doc_text">
2619
2620<p>A key design point of an SSA-based representation is how it
2621represents memory. In LLVM, no memory locations are in SSA form, which
2622makes things very simple. This section describes how to read, write,
2623allocate, and free memory in LLVM.</p>
2624
2625</div>
2626
2627<!-- _______________________________________________________________________ -->
2628<div class="doc_subsubsection">
2629 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2630</div>
2631
2632<div class="doc_text">
2633
2634<h5>Syntax:</h5>
2635
2636<pre>
2637 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
2638</pre>
2639
2640<h5>Overview:</h5>
2641
2642<p>The '<tt>malloc</tt>' instruction allocates memory from the system
2643heap and returns a pointer to it.</p>
2644
2645<h5>Arguments:</h5>
2646
2647<p>The '<tt>malloc</tt>' instruction allocates
2648<tt>sizeof(&lt;type&gt;)*NumElements</tt>
2649bytes of memory from the operating system and returns a pointer of the
2650appropriate type to the program. If "NumElements" is specified, it is the
2651number of elements allocated. If an alignment is specified, the value result
2652of the allocation is guaranteed to be aligned to at least that boundary. If
2653not specified, or if zero, the target can choose to align the allocation on any
2654convenient boundary.</p>
2655
2656<p>'<tt>type</tt>' must be a sized type.</p>
2657
2658<h5>Semantics:</h5>
2659
2660<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2661a pointer is returned.</p>
2662
2663<h5>Example:</h5>
2664
2665<pre>
2666 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2667
2668 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2669 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2670 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2671 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2672 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2673</pre>
2674</div>
2675
2676<!-- _______________________________________________________________________ -->
2677<div class="doc_subsubsection">
2678 <a name="i_free">'<tt>free</tt>' Instruction</a>
2679</div>
2680
2681<div class="doc_text">
2682
2683<h5>Syntax:</h5>
2684
2685<pre>
2686 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
2687</pre>
2688
2689<h5>Overview:</h5>
2690
2691<p>The '<tt>free</tt>' instruction returns memory back to the unused
2692memory heap to be reallocated in the future.</p>
2693
2694<h5>Arguments:</h5>
2695
2696<p>'<tt>value</tt>' shall be a pointer value that points to a value
2697that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2698instruction.</p>
2699
2700<h5>Semantics:</h5>
2701
2702<p>Access to the memory pointed to by the pointer is no longer defined
2703after this instruction executes.</p>
2704
2705<h5>Example:</h5>
2706
2707<pre>
2708 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2709 free [4 x i8]* %array
2710</pre>
2711</div>
2712
2713<!-- _______________________________________________________________________ -->
2714<div class="doc_subsubsection">
2715 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2716</div>
2717
2718<div class="doc_text">
2719
2720<h5>Syntax:</h5>
2721
2722<pre>
2723 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
2724</pre>
2725
2726<h5>Overview:</h5>
2727
2728<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
2729currently executing function, to be automatically released when this function
2730returns to its caller.</p>
2731
2732<h5>Arguments:</h5>
2733
2734<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
2735bytes of memory on the runtime stack, returning a pointer of the
2736appropriate type to the program. If "NumElements" is specified, it is the
2737number of elements allocated. If an alignment is specified, the value result
2738of the allocation is guaranteed to be aligned to at least that boundary. If
2739not specified, or if zero, the target can choose to align the allocation on any
2740convenient boundary.</p>
2741
2742<p>'<tt>type</tt>' may be any sized type.</p>
2743
2744<h5>Semantics:</h5>
2745
2746<p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2747memory is automatically released when the function returns. The '<tt>alloca</tt>'
2748instruction is commonly used to represent automatic variables that must
2749have an address available. When the function returns (either with the <tt><a
2750 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2751instructions), the memory is reclaimed.</p>
2752
2753<h5>Example:</h5>
2754
2755<pre>
2756 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2757 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2758 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2759 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2760</pre>
2761</div>
2762
2763<!-- _______________________________________________________________________ -->
2764<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2765Instruction</a> </div>
2766<div class="doc_text">
2767<h5>Syntax:</h5>
2768<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>
2769<h5>Overview:</h5>
2770<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2771<h5>Arguments:</h5>
2772<p>The argument to the '<tt>load</tt>' instruction specifies the memory
2773address from which to load. The pointer must point to a <a
2774 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2775marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2776the number or order of execution of this <tt>load</tt> with other
2777volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2778instructions. </p>
2779<h5>Semantics:</h5>
2780<p>The location of memory pointed to is loaded.</p>
2781<h5>Examples:</h5>
2782<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2783 <a
2784 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2785 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2786</pre>
2787</div>
2788<!-- _______________________________________________________________________ -->
2789<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2790Instruction</a> </div>
2791<div class="doc_text">
2792<h5>Syntax:</h5>
2793<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
2794 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
2795</pre>
2796<h5>Overview:</h5>
2797<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2798<h5>Arguments:</h5>
2799<p>There are two arguments to the '<tt>store</tt>' instruction: a value
2800to store and an address at which to store it. The type of the '<tt>&lt;pointer&gt;</tt>'
2801operand must be a pointer to the type of the '<tt>&lt;value&gt;</tt>'
2802operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2803optimizer is not allowed to modify the number or order of execution of
2804this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2805 href="#i_store">store</a></tt> instructions.</p>
2806<h5>Semantics:</h5>
2807<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
2808at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
2809<h5>Example:</h5>
2810<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling63ffa142007-10-22 05:10:05 +0000[diff] [blame]2811 store i32 3, i32* %ptr <i>; yields {void}</i>
2812 %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]2813</pre>
2814</div>
2815
2816<!-- _______________________________________________________________________ -->
2817<div class="doc_subsubsection">
2818 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2819</div>
2820
2821<div class="doc_text">
2822<h5>Syntax:</h5>
2823<pre>
2824 &lt;result&gt; = getelementptr &lt;ty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
2825</pre>
2826
2827<h5>Overview:</h5>
2828
2829<p>
2830The '<tt>getelementptr</tt>' instruction is used to get the address of a
2831subelement of an aggregate data structure.</p>
2832
2833<h5>Arguments:</h5>
2834
2835<p>This instruction takes a list of integer operands that indicate what
2836elements of the aggregate object to index to. The actual types of the arguments
2837provided depend on the type of the first pointer argument. The
2838'<tt>getelementptr</tt>' instruction is used to index down through the type
2839levels of a structure or to a specific index in an array. When indexing into a
2840structure, only <tt>i32</tt> integer constants are allowed. When indexing
2841into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2842be sign extended to 64-bit values.</p>
2843
2844<p>For example, let's consider a C code fragment and how it gets
2845compiled to LLVM:</p>
2846
2847<div class="doc_code">
2848<pre>
2849struct RT {
2850 char A;
2851 int B[10][20];
2852 char C;
2853};
2854struct ST {
2855 int X;
2856 double Y;
2857 struct RT Z;
2858};
2859
2860int *foo(struct ST *s) {
2861 return &amp;s[1].Z.B[5][13];
2862}
2863</pre>
2864</div>
2865
2866<p>The LLVM code generated by the GCC frontend is:</p>
2867
2868<div class="doc_code">
2869<pre>
2870%RT = type { i8 , [10 x [20 x i32]], i8 }
2871%ST = type { i32, double, %RT }
2872
2873define i32* %foo(%ST* %s) {
2874entry:
2875 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
2876 ret i32* %reg
2877}
2878</pre>
2879</div>
2880
2881<h5>Semantics:</h5>
2882
2883<p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2884on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2885and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
2886<a href="#t_integer">integer</a> type but the value will always be sign extended
2887to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
2888<b>constants</b>.</p>
2889
2890<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2891type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
2892}</tt>' type, a structure. The second index indexes into the third element of
2893the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
2894i8 }</tt>' type, another structure. The third index indexes into the second
2895element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
2896array. The two dimensions of the array are subscripted into, yielding an
2897'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2898to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
2899
2900<p>Note that it is perfectly legal to index partially through a
2901structure, returning a pointer to an inner element. Because of this,
2902the LLVM code for the given testcase is equivalent to:</p>
2903
2904<pre>
2905 define i32* %foo(%ST* %s) {
2906 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
2907 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
2908 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
2909 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
2910 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
2911 ret i32* %t5
2912 }
2913</pre>
2914
2915<p>Note that it is undefined to access an array out of bounds: array and
2916pointer indexes must always be within the defined bounds of the array type.
2917The one exception for this rules is zero length arrays. These arrays are
2918defined to be accessible as variable length arrays, which requires access
2919beyond the zero'th element.</p>
2920
2921<p>The getelementptr instruction is often confusing. For some more insight
2922into how it works, see <a href="GetElementPtr.html">the getelementptr
2923FAQ</a>.</p>
2924
2925<h5>Example:</h5>
2926
2927<pre>
2928 <i>; yields [12 x i8]*:aptr</i>
2929 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
2930</pre>
2931</div>
2932
2933<!-- ======================================================================= -->
2934<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2935</div>
2936<div class="doc_text">
2937<p>The instructions in this category are the conversion instructions (casting)
2938which all take a single operand and a type. They perform various bit conversions
2939on the operand.</p>
2940</div>
2941
2942<!-- _______________________________________________________________________ -->
2943<div class="doc_subsubsection">
2944 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2945</div>
2946<div class="doc_text">
2947
2948<h5>Syntax:</h5>
2949<pre>
2950 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
2951</pre>
2952
2953<h5>Overview:</h5>
2954<p>
2955The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2956</p>
2957
2958<h5>Arguments:</h5>
2959<p>
2960The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2961be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2962and type of the result, which must be an <a href="#t_integer">integer</a>
2963type. The bit size of <tt>value</tt> must be larger than the bit size of
2964<tt>ty2</tt>. Equal sized types are not allowed.</p>
2965
2966<h5>Semantics:</h5>
2967<p>
2968The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
2969and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
2970larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
2971It will always truncate bits.</p>
2972
2973<h5>Example:</h5>
2974<pre>
2975 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
2976 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
2977 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
2978</pre>
2979</div>
2980
2981<!-- _______________________________________________________________________ -->
2982<div class="doc_subsubsection">
2983 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
2984</div>
2985<div class="doc_text">
2986
2987<h5>Syntax:</h5>
2988<pre>
2989 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
2990</pre>
2991
2992<h5>Overview:</h5>
2993<p>The '<tt>zext</tt>' instruction zero extends its operand to type
2994<tt>ty2</tt>.</p>
2995
2996
2997<h5>Arguments:</h5>
2998<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
2999<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3000also be of <a href="#t_integer">integer</a> type. The bit size of the
3001<tt>value</tt> must be smaller than the bit size of the destination type,
3002<tt>ty2</tt>.</p>
3003
3004<h5>Semantics:</h5>
3005<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3006bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3007
3008<p>When zero extending from i1, the result will always be either 0 or 1.</p>
3009
3010<h5>Example:</h5>
3011<pre>
3012 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3013 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3014</pre>
3015</div>
3016
3017<!-- _______________________________________________________________________ -->
3018<div class="doc_subsubsection">
3019 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3020</div>
3021<div class="doc_text">
3022
3023<h5>Syntax:</h5>
3024<pre>
3025 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3026</pre>
3027
3028<h5>Overview:</h5>
3029<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3030
3031<h5>Arguments:</h5>
3032<p>
3033The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3034<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3035also be of <a href="#t_integer">integer</a> type. The bit size of the
3036<tt>value</tt> must be smaller than the bit size of the destination type,
3037<tt>ty2</tt>.</p>
3038
3039<h5>Semantics:</h5>
3040<p>
3041The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3042bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3043the type <tt>ty2</tt>.</p>
3044
3045<p>When sign extending from i1, the extension always results in -1 or 0.</p>
3046
3047<h5>Example:</h5>
3048<pre>
3049 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3050 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3051</pre>
3052</div>
3053
3054<!-- _______________________________________________________________________ -->
3055<div class="doc_subsubsection">
3056 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3057</div>
3058
3059<div class="doc_text">
3060
3061<h5>Syntax:</h5>
3062
3063<pre>
3064 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3065</pre>
3066
3067<h5>Overview:</h5>
3068<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3069<tt>ty2</tt>.</p>
3070
3071
3072<h5>Arguments:</h5>
3073<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3074 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3075cast it to. The size of <tt>value</tt> must be larger than the size of
3076<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3077<i>no-op cast</i>.</p>
3078
3079<h5>Semantics:</h5>
3080<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3081<a href="#t_floating">floating point</a> type to a smaller
3082<a href="#t_floating">floating point</a> type. If the value cannot fit within
3083the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3084
3085<h5>Example:</h5>
3086<pre>
3087 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3088 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3089</pre>
3090</div>
3091
3092<!-- _______________________________________________________________________ -->
3093<div class="doc_subsubsection">
3094 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3095</div>
3096<div class="doc_text">
3097
3098<h5>Syntax:</h5>
3099<pre>
3100 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3101</pre>
3102
3103<h5>Overview:</h5>
3104<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3105floating point value.</p>
3106
3107<h5>Arguments:</h5>
3108<p>The '<tt>fpext</tt>' instruction takes a
3109<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3110and a <a href="#t_floating">floating point</a> type to cast it to. The source
3111type must be smaller than the destination type.</p>
3112
3113<h5>Semantics:</h5>
3114<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3115<a href="#t_floating">floating point</a> type to a larger
3116<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3117used to make a <i>no-op cast</i> because it always changes bits. Use
3118<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3119
3120<h5>Example:</h5>
3121<pre>
3122 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3123 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3124</pre>
3125</div>
3126
3127<!-- _______________________________________________________________________ -->
3128<div class="doc_subsubsection">
3129 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3130</div>
3131<div class="doc_text">
3132
3133<h5>Syntax:</h5>
3134<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3135 &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]3136</pre>
3137
3138<h5>Overview:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3139<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3140unsigned integer equivalent of type <tt>ty2</tt>.
3141</p>
3142
3143<h5>Arguments:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3144<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]3145scalar or vector <a href="#t_floating">floating point</a> value, and a type
3146to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3147type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3148vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3149
3150<h5>Semantics:</h5>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3151<p> The '<tt>fptoui</tt>' instruction converts its
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3152<a href="#t_floating">floating point</a> operand into the nearest (rounding
3153towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3154the results are undefined.</p>
3155
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3156<h5>Example:</h5>
3157<pre>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3158 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3159 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencere6adee82007-07-31 14:40:14 +0000[diff] [blame]3160 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3161</pre>
3162</div>
3163
3164<!-- _______________________________________________________________________ -->
3165<div class="doc_subsubsection">
3166 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3167</div>
3168<div class="doc_text">
3169
3170<h5>Syntax:</h5>
3171<pre>
3172 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3173</pre>
3174
3175<h5>Overview:</h5>
3176<p>The '<tt>fptosi</tt>' instruction converts
3177<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3178</p>
3179
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3180<h5>Arguments:</h5>
3181<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]3182scalar or vector <a href="#t_floating">floating point</a> value, and a type
3183to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3184type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3185vector integer type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3186
3187<h5>Semantics:</h5>
3188<p>The '<tt>fptosi</tt>' instruction converts its
3189<a href="#t_floating">floating point</a> operand into the nearest (rounding
3190towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3191the results are undefined.</p>
3192
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3193<h5>Example:</h5>
3194<pre>
3195 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner681f1e82007-09-22 03:17:52 +0000[diff] [blame]3196 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3197 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3198</pre>
3199</div>
3200
3201<!-- _______________________________________________________________________ -->
3202<div class="doc_subsubsection">
3203 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3204</div>
3205<div class="doc_text">
3206
3207<h5>Syntax:</h5>
3208<pre>
3209 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3210</pre>
3211
3212<h5>Overview:</h5>
3213<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3214integer and converts that value to the <tt>ty2</tt> type.</p>
3215
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3216<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3217<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3218scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3219to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3220type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3221floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3222
3223<h5>Semantics:</h5>
3224<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3225integer quantity and converts it to the corresponding floating point value. If
3226the value cannot fit in the floating point value, the results are undefined.</p>
3227
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3228<h5>Example:</h5>
3229<pre>
3230 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3231 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3232</pre>
3233</div>
3234
3235<!-- _______________________________________________________________________ -->
3236<div class="doc_subsubsection">
3237 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3238</div>
3239<div class="doc_text">
3240
3241<h5>Syntax:</h5>
3242<pre>
3243 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3244</pre>
3245
3246<h5>Overview:</h5>
3247<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3248integer and converts that value to the <tt>ty2</tt> type.</p>
3249
3250<h5>Arguments:</h5>
Nate Begeman78246ca2007-11-17 03:58:34 +0000[diff] [blame]3251<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3252scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3253to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3254type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3255floating point type with the same number of elements as <tt>ty</tt></p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3256
3257<h5>Semantics:</h5>
3258<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3259integer quantity and converts it to the corresponding floating point value. If
3260the value cannot fit in the floating point value, the results are undefined.</p>
3261
3262<h5>Example:</h5>
3263<pre>
3264 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3265 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3266</pre>
3267</div>
3268
3269<!-- _______________________________________________________________________ -->
3270<div class="doc_subsubsection">
3271 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3272</div>
3273<div class="doc_text">
3274
3275<h5>Syntax:</h5>
3276<pre>
3277 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3278</pre>
3279
3280<h5>Overview:</h5>
3281<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3282the integer type <tt>ty2</tt>.</p>
3283
3284<h5>Arguments:</h5>
3285<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3286must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3287<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3288
3289<h5>Semantics:</h5>
3290<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3291<tt>ty2</tt> by interpreting the pointer value as an integer and either
3292truncating or zero extending that value to the size of the integer type. If
3293<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3294<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3295are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3296change.</p>
3297
3298<h5>Example:</h5>
3299<pre>
3300 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3301 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3302</pre>
3303</div>
3304
3305<!-- _______________________________________________________________________ -->
3306<div class="doc_subsubsection">
3307 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3308</div>
3309<div class="doc_text">
3310
3311<h5>Syntax:</h5>
3312<pre>
3313 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3314</pre>
3315
3316<h5>Overview:</h5>
3317<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3318a pointer type, <tt>ty2</tt>.</p>
3319
3320<h5>Arguments:</h5>
3321<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3322value to cast, and a type to cast it to, which must be a
3323<a href="#t_pointer">pointer</a> type.
3324
3325<h5>Semantics:</h5>
3326<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3327<tt>ty2</tt> by applying either a zero extension or a truncation depending on
3328the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3329size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3330the size of a pointer then a zero extension is done. If they are the same size,
3331nothing is done (<i>no-op cast</i>).</p>
3332
3333<h5>Example:</h5>
3334<pre>
3335 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3336 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3337 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3338</pre>
3339</div>
3340
3341<!-- _______________________________________________________________________ -->
3342<div class="doc_subsubsection">
3343 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3344</div>
3345<div class="doc_text">
3346
3347<h5>Syntax:</h5>
3348<pre>
3349 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3350</pre>
3351
3352<h5>Overview:</h5>
3353<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3354<tt>ty2</tt> without changing any bits.</p>
3355
3356<h5>Arguments:</h5>
3357<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3358a first class value, and a type to cast it to, which must also be a <a
3359 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3360and the destination type, <tt>ty2</tt>, must be identical. If the source
3361type is a pointer, the destination type must also be a pointer.</p>
3362
3363<h5>Semantics:</h5>
3364<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3365<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3366this conversion. The conversion is done as if the <tt>value</tt> had been
3367stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3368converted to other pointer types with this instruction. To convert pointers to
3369other types, use the <a href="#i_inttoptr">inttoptr</a> or
3370<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3371
3372<h5>Example:</h5>
3373<pre>
3374 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3375 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3376 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3377</pre>
3378</div>
3379
3380<!-- ======================================================================= -->
3381<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3382<div class="doc_text">
3383<p>The instructions in this category are the "miscellaneous"
3384instructions, which defy better classification.</p>
3385</div>
3386
3387<!-- _______________________________________________________________________ -->
3388<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3389</div>
3390<div class="doc_text">
3391<h5>Syntax:</h5>
3392<pre> &lt;result&gt; = icmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3393</pre>
3394<h5>Overview:</h5>
3395<p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3396of its two integer operands.</p>
3397<h5>Arguments:</h5>
3398<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3399the condition code indicating the kind of comparison to perform. It is not
3400a value, just a keyword. The possible condition code are:
3401<ol>
3402 <li><tt>eq</tt>: equal</li>
3403 <li><tt>ne</tt>: not equal </li>
3404 <li><tt>ugt</tt>: unsigned greater than</li>
3405 <li><tt>uge</tt>: unsigned greater or equal</li>
3406 <li><tt>ult</tt>: unsigned less than</li>
3407 <li><tt>ule</tt>: unsigned less or equal</li>
3408 <li><tt>sgt</tt>: signed greater than</li>
3409 <li><tt>sge</tt>: signed greater or equal</li>
3410 <li><tt>slt</tt>: signed less than</li>
3411 <li><tt>sle</tt>: signed less or equal</li>
3412</ol>
3413<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3414<a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3415<h5>Semantics:</h5>
3416<p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3417the condition code given as <tt>cond</tt>. The comparison performed always
3418yields a <a href="#t_primitive">i1</a> result, as follows:
3419<ol>
3420 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3421 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3422 </li>
3423 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3424 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3425 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3426 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3427 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3428 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3429 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3430 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3431 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3432 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3433 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3434 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3435 <li><tt>sge</tt>: interprets the operands as signed values and yields
3436 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3437 <li><tt>slt</tt>: interprets the operands as signed values and yields
3438 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3439 <li><tt>sle</tt>: interprets the operands as signed values and yields
3440 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3441</ol>
3442<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3443values are compared as if they were integers.</p>
3444
3445<h5>Example:</h5>
3446<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
3447 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
3448 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
3449 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3450 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3451 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
3452</pre>
3453</div>
3454
3455<!-- _______________________________________________________________________ -->
3456<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3457</div>
3458<div class="doc_text">
3459<h5>Syntax:</h5>
3460<pre> &lt;result&gt; = fcmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
3461</pre>
3462<h5>Overview:</h5>
3463<p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3464of its floating point operands.</p>
3465<h5>Arguments:</h5>
3466<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3467the condition code indicating the kind of comparison to perform. It is not
3468a value, just a keyword. The possible condition code are:
3469<ol>
3470 <li><tt>false</tt>: no comparison, always returns false</li>
3471 <li><tt>oeq</tt>: ordered and equal</li>
3472 <li><tt>ogt</tt>: ordered and greater than </li>
3473 <li><tt>oge</tt>: ordered and greater than or equal</li>
3474 <li><tt>olt</tt>: ordered and less than </li>
3475 <li><tt>ole</tt>: ordered and less than or equal</li>
3476 <li><tt>one</tt>: ordered and not equal</li>
3477 <li><tt>ord</tt>: ordered (no nans)</li>
3478 <li><tt>ueq</tt>: unordered or equal</li>
3479 <li><tt>ugt</tt>: unordered or greater than </li>
3480 <li><tt>uge</tt>: unordered or greater than or equal</li>
3481 <li><tt>ult</tt>: unordered or less than </li>
3482 <li><tt>ule</tt>: unordered or less than or equal</li>
3483 <li><tt>une</tt>: unordered or not equal</li>
3484 <li><tt>uno</tt>: unordered (either nans)</li>
3485 <li><tt>true</tt>: no comparison, always returns true</li>
3486</ol>
3487<p><i>Ordered</i> means that neither operand is a QNAN while
3488<i>unordered</i> means that either operand may be a QNAN.</p>
3489<p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3490<a href="#t_floating">floating point</a> typed. They must have identical
3491types.</p>
3492<h5>Semantics:</h5>
3493<p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3494the condition code given as <tt>cond</tt>. The comparison performed always
3495yields a <a href="#t_primitive">i1</a> result, as follows:
3496<ol>
3497 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3498 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3499 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3500 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3501 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3502 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3503 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3504 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3505 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3506 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3507 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3508 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3509 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3510 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3511 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3512 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3513 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3514 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3515 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3516 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3517 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3518 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3519 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3520 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3521 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3522 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3523 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3524 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3525</ol>
3526
3527<h5>Example:</h5>
3528<pre> &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3529 &lt;result&gt; = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3530 &lt;result&gt; = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3531 &lt;result&gt; = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3532</pre>
3533</div>
3534
3535<!-- _______________________________________________________________________ -->
3536<div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3537Instruction</a> </div>
3538<div class="doc_text">
3539<h5>Syntax:</h5>
3540<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
3541<h5>Overview:</h5>
3542<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
3543the SSA graph representing the function.</p>
3544<h5>Arguments:</h5>
3545<p>The type of the incoming values is specified with the first type
3546field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3547as arguments, with one pair for each predecessor basic block of the
3548current block. Only values of <a href="#t_firstclass">first class</a>
3549type may be used as the value arguments to the PHI node. Only labels
3550may be used as the label arguments.</p>
3551<p>There must be no non-phi instructions between the start of a basic
3552block and the PHI instructions: i.e. PHI instructions must be first in
3553a basic block.</p>
3554<h5>Semantics:</h5>
3555<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3556specified by the pair corresponding to the predecessor basic block that executed
3557just prior to the current block.</p>
3558<h5>Example:</h5>
3559<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>
3560</div>
3561
3562<!-- _______________________________________________________________________ -->
3563<div class="doc_subsubsection">
3564 <a name="i_select">'<tt>select</tt>' Instruction</a>
3565</div>
3566
3567<div class="doc_text">
3568
3569<h5>Syntax:</h5>
3570
3571<pre>
3572 &lt;result&gt; = select i1 &lt;cond&gt;, &lt;ty&gt; &lt;val1&gt;, &lt;ty&gt; &lt;val2&gt; <i>; yields ty</i>
3573</pre>
3574
3575<h5>Overview:</h5>
3576
3577<p>
3578The '<tt>select</tt>' instruction is used to choose one value based on a
3579condition, without branching.
3580</p>
3581
3582
3583<h5>Arguments:</h5>
3584
3585<p>
3586The '<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.
3587</p>
3588
3589<h5>Semantics:</h5>
3590
3591<p>
3592If the boolean condition evaluates to true, the instruction returns the first
3593value argument; otherwise, it returns the second value argument.
3594</p>
3595
3596<h5>Example:</h5>
3597
3598<pre>
3599 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3600</pre>
3601</div>
3602
3603
3604<!-- _______________________________________________________________________ -->
3605<div class="doc_subsubsection">
3606 <a name="i_call">'<tt>call</tt>' Instruction</a>
3607</div>
3608
3609<div class="doc_text">
3610
3611<h5>Syntax:</h5>
3612<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3613 &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]3614</pre>
3615
3616<h5>Overview:</h5>
3617
3618<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3619
3620<h5>Arguments:</h5>
3621
3622<p>This instruction requires several arguments:</p>
3623
3624<ol>
3625 <li>
3626 <p>The optional "tail" marker indicates whether the callee function accesses
3627 any allocas or varargs in the caller. If the "tail" marker is present, the
3628 function call is eligible for tail call optimization. Note that calls may
3629 be marked "tail" even if they do not occur before a <a
3630 href="#i_ret"><tt>ret</tt></a> instruction.
3631 </li>
3632 <li>
3633 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3634 convention</a> the call should use. If none is specified, the call defaults
3635 to using C calling conventions.
3636 </li>
3637 <li>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3638 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
3639 the type of the return value. Functions that return no value are marked
3640 <tt><a href="#t_void">void</a></tt>.</p>
3641 </li>
3642 <li>
3643 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
3644 value being invoked. The argument types must match the types implied by
3645 this signature. This type can be omitted if the function is not varargs
3646 and if the function type does not return a pointer to a function.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3647 </li>
3648 <li>
3649 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3650 be invoked. In most cases, this is a direct function invocation, but
3651 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3652 to function value.</p>
3653 </li>
3654 <li>
3655 <p>'<tt>function args</tt>': argument list whose types match the
3656 function signature argument types. All arguments must be of
3657 <a href="#t_firstclass">first class</a> type. If the function signature
3658 indicates the function accepts a variable number of arguments, the extra
3659 arguments can be specified.</p>
3660 </li>
3661</ol>
3662
3663<h5>Semantics:</h5>
3664
3665<p>The '<tt>call</tt>' instruction is used to cause control flow to
3666transfer to a specified function, with its incoming arguments bound to
3667the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3668instruction in the called function, control flow continues with the
3669instruction after the function call, and the return value of the
3670function is bound to the result argument. This is a simpler case of
3671the <a href="#i_invoke">invoke</a> instruction.</p>
3672
3673<h5>Example:</h5>
3674
3675<pre>
Nick Lewycky93082fc2007-09-08 13:57:50 +0000[diff] [blame]3676 %retval = call i32 @test(i32 %argc)
3677 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42);
3678 %X = tail call i32 @foo()
3679 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo()
3680 %Z = call void %foo(i8 97 signext)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3681</pre>
3682
3683</div>
3684
3685<!-- _______________________________________________________________________ -->
3686<div class="doc_subsubsection">
3687 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3688</div>
3689
3690<div class="doc_text">
3691
3692<h5>Syntax:</h5>
3693
3694<pre>
3695 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
3696</pre>
3697
3698<h5>Overview:</h5>
3699
3700<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3701the "variable argument" area of a function call. It is used to implement the
3702<tt>va_arg</tt> macro in C.</p>
3703
3704<h5>Arguments:</h5>
3705
3706<p>This instruction takes a <tt>va_list*</tt> value and the type of
3707the argument. It returns a value of the specified argument type and
3708increments the <tt>va_list</tt> to point to the next argument. The
3709actual type of <tt>va_list</tt> is target specific.</p>
3710
3711<h5>Semantics:</h5>
3712
3713<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3714type from the specified <tt>va_list</tt> and causes the
3715<tt>va_list</tt> to point to the next argument. For more information,
3716see the variable argument handling <a href="#int_varargs">Intrinsic
3717Functions</a>.</p>
3718
3719<p>It is legal for this instruction to be called in a function which does not
3720take a variable number of arguments, for example, the <tt>vfprintf</tt>
3721function.</p>
3722
3723<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3724href="#intrinsics">intrinsic function</a> because it takes a type as an
3725argument.</p>
3726
3727<h5>Example:</h5>
3728
3729<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3730
3731</div>
3732
3733<!-- *********************************************************************** -->
3734<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3735<!-- *********************************************************************** -->
3736
3737<div class="doc_text">
3738
3739<p>LLVM supports the notion of an "intrinsic function". These functions have
3740well known names and semantics and are required to follow certain restrictions.
3741Overall, these intrinsics represent an extension mechanism for the LLVM
3742language that does not require changing all of the transformations in LLVM when
3743adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
3744
3745<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3746prefix is reserved in LLVM for intrinsic names; thus, function names may not
3747begin with this prefix. Intrinsic functions must always be external functions:
3748you cannot define the body of intrinsic functions. Intrinsic functions may
3749only be used in call or invoke instructions: it is illegal to take the address
3750of an intrinsic function. Additionally, because intrinsic functions are part
3751of the LLVM language, it is required if any are added that they be documented
3752here.</p>
3753
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]3754<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
3755a family of functions that perform the same operation but on different data
3756types. Because LLVM can represent over 8 million different integer types,
3757overloading is used commonly to allow an intrinsic function to operate on any
3758integer type. One or more of the argument types or the result type can be
3759overloaded to accept any integer type. Argument types may also be defined as
3760exactly matching a previous argument's type or the result type. This allows an
3761intrinsic function which accepts multiple arguments, but needs all of them to
3762be of the same type, to only be overloaded with respect to a single argument or
3763the result.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3764
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]3765<p>Overloaded intrinsics will have the names of its overloaded argument types
3766encoded into its function name, each preceded by a period. Only those types
3767which are overloaded result in a name suffix. Arguments whose type is matched
3768against another type do not. For example, the <tt>llvm.ctpop</tt> function can
3769take an integer of any width and returns an integer of exactly the same integer
3770width. This leads to a family of functions such as
3771<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
3772Only one type, the return type, is overloaded, and only one type suffix is
3773required. Because the argument's type is matched against the return type, it
3774does not require its own name suffix.</p>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3775
3776<p>To learn how to add an intrinsic function, please see the
3777<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3778</p>
3779
3780</div>
3781
3782<!-- ======================================================================= -->
3783<div class="doc_subsection">
3784 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3785</div>
3786
3787<div class="doc_text">
3788
3789<p>Variable argument support is defined in LLVM with the <a
3790 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3791intrinsic functions. These functions are related to the similarly
3792named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
3793
3794<p>All of these functions operate on arguments that use a
3795target-specific value type "<tt>va_list</tt>". The LLVM assembly
3796language reference manual does not define what this type is, so all
3797transformations should be prepared to handle these functions regardless of
3798the type used.</p>
3799
3800<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3801instruction and the variable argument handling intrinsic functions are
3802used.</p>
3803
3804<div class="doc_code">
3805<pre>
3806define i32 @test(i32 %X, ...) {
3807 ; Initialize variable argument processing
3808 %ap = alloca i8*
3809 %ap2 = bitcast i8** %ap to i8*
3810 call void @llvm.va_start(i8* %ap2)
3811
3812 ; Read a single integer argument
3813 %tmp = va_arg i8** %ap, i32
3814
3815 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3816 %aq = alloca i8*
3817 %aq2 = bitcast i8** %aq to i8*
3818 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
3819 call void @llvm.va_end(i8* %aq2)
3820
3821 ; Stop processing of arguments.
3822 call void @llvm.va_end(i8* %ap2)
3823 ret i32 %tmp
3824}
3825
3826declare void @llvm.va_start(i8*)
3827declare void @llvm.va_copy(i8*, i8*)
3828declare void @llvm.va_end(i8*)
3829</pre>
3830</div>
3831
3832</div>
3833
3834<!-- _______________________________________________________________________ -->
3835<div class="doc_subsubsection">
3836 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3837</div>
3838
3839
3840<div class="doc_text">
3841<h5>Syntax:</h5>
3842<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
3843<h5>Overview:</h5>
3844<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3845<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
3846href="#i_va_arg">va_arg</a></tt>.</p>
3847
3848<h5>Arguments:</h5>
3849
3850<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3851
3852<h5>Semantics:</h5>
3853
3854<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3855macro available in C. In a target-dependent way, it initializes the
3856<tt>va_list</tt> element to which the argument points, so that the next call to
3857<tt>va_arg</tt> will produce the first variable argument passed to the function.
3858Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3859last argument of the function as the compiler can figure that out.</p>
3860
3861</div>
3862
3863<!-- _______________________________________________________________________ -->
3864<div class="doc_subsubsection">
3865 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3866</div>
3867
3868<div class="doc_text">
3869<h5>Syntax:</h5>
3870<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
3871<h5>Overview:</h5>
3872
3873<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
3874which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
3875or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3876
3877<h5>Arguments:</h5>
3878
3879<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
3880
3881<h5>Semantics:</h5>
3882
3883<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3884macro available in C. In a target-dependent way, it destroys the
3885<tt>va_list</tt> element to which the argument points. Calls to <a
3886href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
3887<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
3888<tt>llvm.va_end</tt>.</p>
3889
3890</div>
3891
3892<!-- _______________________________________________________________________ -->
3893<div class="doc_subsubsection">
3894 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3895</div>
3896
3897<div class="doc_text">
3898
3899<h5>Syntax:</h5>
3900
3901<pre>
3902 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
3903</pre>
3904
3905<h5>Overview:</h5>
3906
3907<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
3908from the source argument list to the destination argument list.</p>
3909
3910<h5>Arguments:</h5>
3911
3912<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3913The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3914
3915
3916<h5>Semantics:</h5>
3917
3918<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
3919macro available in C. In a target-dependent way, it copies the source
3920<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
3921intrinsic is necessary because the <tt><a href="#int_va_start">
3922llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
3923example, memory allocation.</p>
3924
3925</div>
3926
3927<!-- ======================================================================= -->
3928<div class="doc_subsection">
3929 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3930</div>
3931
3932<div class="doc_text">
3933
3934<p>
3935LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3936Collection</a> requires the implementation and generation of these intrinsics.
3937These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
3938stack</a>, as well as garbage collector implementations that require <a
3939href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
3940Front-ends for type-safe garbage collected languages should generate these
3941intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3942href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3943</p>
3944</div>
3945
3946<!-- _______________________________________________________________________ -->
3947<div class="doc_subsubsection">
3948 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3949</div>
3950
3951<div class="doc_text">
3952
3953<h5>Syntax:</h5>
3954
3955<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]3956 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3957</pre>
3958
3959<h5>Overview:</h5>
3960
3961<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3962the code generator, and allows some metadata to be associated with it.</p>
3963
3964<h5>Arguments:</h5>
3965
3966<p>The first argument specifies the address of a stack object that contains the
3967root pointer. The second pointer (which must be either a constant or a global
3968value address) contains the meta-data to be associated with the root.</p>
3969
3970<h5>Semantics:</h5>
3971
3972<p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3973location. At compile-time, the code generator generates information to allow
3974the runtime to find the pointer at GC safe points.
3975</p>
3976
3977</div>
3978
3979
3980<!-- _______________________________________________________________________ -->
3981<div class="doc_subsubsection">
3982 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
3983</div>
3984
3985<div class="doc_text">
3986
3987<h5>Syntax:</h5>
3988
3989<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]3990 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]3991</pre>
3992
3993<h5>Overview:</h5>
3994
3995<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
3996locations, allowing garbage collector implementations that require read
3997barriers.</p>
3998
3999<h5>Arguments:</h5>
4000
4001<p>The second argument is the address to read from, which should be an address
4002allocated from the garbage collector. The first object is a pointer to the
4003start of the referenced object, if needed by the language runtime (otherwise
4004null).</p>
4005
4006<h5>Semantics:</h5>
4007
4008<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4009instruction, but may be replaced with substantially more complex code by the
4010garbage collector runtime, as needed.</p>
4011
4012</div>
4013
4014
4015<!-- _______________________________________________________________________ -->
4016<div class="doc_subsubsection">
4017 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4018</div>
4019
4020<div class="doc_text">
4021
4022<h5>Syntax:</h5>
4023
4024<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4025 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4026</pre>
4027
4028<h5>Overview:</h5>
4029
4030<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4031locations, allowing garbage collector implementations that require write
4032barriers (such as generational or reference counting collectors).</p>
4033
4034<h5>Arguments:</h5>
4035
4036<p>The first argument is the reference to store, the second is the start of the
4037object to store it to, and the third is the address of the field of Obj to
4038store to. If the runtime does not require a pointer to the object, Obj may be
4039null.</p>
4040
4041<h5>Semantics:</h5>
4042
4043<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4044instruction, but may be replaced with substantially more complex code by the
4045garbage collector runtime, as needed.</p>
4046
4047</div>
4048
4049
4050
4051<!-- ======================================================================= -->
4052<div class="doc_subsection">
4053 <a name="int_codegen">Code Generator Intrinsics</a>
4054</div>
4055
4056<div class="doc_text">
4057<p>
4058These intrinsics are provided by LLVM to expose special features that may only
4059be implemented with code generator support.
4060</p>
4061
4062</div>
4063
4064<!-- _______________________________________________________________________ -->
4065<div class="doc_subsubsection">
4066 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4067</div>
4068
4069<div class="doc_text">
4070
4071<h5>Syntax:</h5>
4072<pre>
4073 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
4074</pre>
4075
4076<h5>Overview:</h5>
4077
4078<p>
4079The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4080target-specific value indicating the return address of the current function
4081or one of its callers.
4082</p>
4083
4084<h5>Arguments:</h5>
4085
4086<p>
4087The argument to this intrinsic indicates which function to return the address
4088for. Zero indicates the calling function, one indicates its caller, etc. The
4089argument is <b>required</b> to be a constant integer value.
4090</p>
4091
4092<h5>Semantics:</h5>
4093
4094<p>
4095The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4096the return address of the specified call frame, or zero if it cannot be
4097identified. The value returned by this intrinsic is likely to be incorrect or 0
4098for arguments other than zero, so it should only be used for debugging purposes.
4099</p>
4100
4101<p>
4102Note that calling this intrinsic does not prevent function inlining or other
4103aggressive transformations, so the value returned may not be that of the obvious
4104source-language caller.
4105</p>
4106</div>
4107
4108
4109<!-- _______________________________________________________________________ -->
4110<div class="doc_subsubsection">
4111 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4112</div>
4113
4114<div class="doc_text">
4115
4116<h5>Syntax:</h5>
4117<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4118 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4119</pre>
4120
4121<h5>Overview:</h5>
4122
4123<p>
4124The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4125target-specific frame pointer value for the specified stack frame.
4126</p>
4127
4128<h5>Arguments:</h5>
4129
4130<p>
4131The argument to this intrinsic indicates which function to return the frame
4132pointer for. Zero indicates the calling function, one indicates its caller,
4133etc. The argument is <b>required</b> to be a constant integer value.
4134</p>
4135
4136<h5>Semantics:</h5>
4137
4138<p>
4139The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4140the frame address of the specified call frame, or zero if it cannot be
4141identified. The value returned by this intrinsic is likely to be incorrect or 0
4142for arguments other than zero, so it should only be used for debugging purposes.
4143</p>
4144
4145<p>
4146Note that calling this intrinsic does not prevent function inlining or other
4147aggressive transformations, so the value returned may not be that of the obvious
4148source-language caller.
4149</p>
4150</div>
4151
4152<!-- _______________________________________________________________________ -->
4153<div class="doc_subsubsection">
4154 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4155</div>
4156
4157<div class="doc_text">
4158
4159<h5>Syntax:</h5>
4160<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4161 declare i8 *@llvm.stacksave()
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4162</pre>
4163
4164<h5>Overview:</h5>
4165
4166<p>
4167The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4168the function stack, for use with <a href="#int_stackrestore">
4169<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4170features like scoped automatic variable sized arrays in C99.
4171</p>
4172
4173<h5>Semantics:</h5>
4174
4175<p>
4176This intrinsic returns a opaque pointer value that can be passed to <a
4177href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4178<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4179<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4180state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4181practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4182that were allocated after the <tt>llvm.stacksave</tt> was executed.
4183</p>
4184
4185</div>
4186
4187<!-- _______________________________________________________________________ -->
4188<div class="doc_subsubsection">
4189 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4190</div>
4191
4192<div class="doc_text">
4193
4194<h5>Syntax:</h5>
4195<pre>
4196 declare void @llvm.stackrestore(i8 * %ptr)
4197</pre>
4198
4199<h5>Overview:</h5>
4200
4201<p>
4202The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4203the function stack to the state it was in when the corresponding <a
4204href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4205useful for implementing language features like scoped automatic variable sized
4206arrays in C99.
4207</p>
4208
4209<h5>Semantics:</h5>
4210
4211<p>
4212See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4213</p>
4214
4215</div>
4216
4217
4218<!-- _______________________________________________________________________ -->
4219<div class="doc_subsubsection">
4220 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4221</div>
4222
4223<div class="doc_text">
4224
4225<h5>Syntax:</h5>
4226<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4227 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]4228</pre>
4229
4230<h5>Overview:</h5>
4231
4232
4233<p>
4234The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4235a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4236no
4237effect on the behavior of the program but can change its performance
4238characteristics.
4239</p>
4240
4241<h5>Arguments:</h5>
4242
4243<p>
4244<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4245determining if the fetch should be for a read (0) or write (1), and
4246<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4247locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4248<tt>locality</tt> arguments must be constant integers.
4249</p>
4250
4251<h5>Semantics:</h5>
4252
4253<p>
4254This intrinsic does not modify the behavior of the program. In particular,
4255prefetches cannot trap and do not produce a value. On targets that support this
4256intrinsic, the prefetch can provide hints to the processor cache for better
4257performance.
4258</p>
4259
4260</div>
4261
4262<!-- _______________________________________________________________________ -->
4263<div class="doc_subsubsection">
4264 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4265</div>
4266
4267<div class="doc_text">
4268
4269<h5>Syntax:</h5>
4270<pre>
Chris Lattner38bd5dd2007-09-21 17:30:40 +0000[diff] [blame]4271 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4272</pre>
4273
4274<h5>Overview:</h5>
4275
4276
4277<p>
4278The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4279(PC) in a region of
4280code to simulators and other tools. The method is target specific, but it is
4281expected that the marker will use exported symbols to transmit the PC of the marker.
4282The marker makes no guarantees that it will remain with any specific instruction
4283after optimizations. It is possible that the presence of a marker will inhibit
4284optimizations. The intended use is to be inserted after optimizations to allow
4285correlations of simulation runs.
4286</p>
4287
4288<h5>Arguments:</h5>
4289
4290<p>
4291<tt>id</tt> is a numerical id identifying the marker.
4292</p>
4293
4294<h5>Semantics:</h5>
4295
4296<p>
4297This intrinsic does not modify the behavior of the program. Backends that do not
4298support this intrinisic may ignore it.
4299</p>
4300
4301</div>
4302
4303<!-- _______________________________________________________________________ -->
4304<div class="doc_subsubsection">
4305 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4306</div>
4307
4308<div class="doc_text">
4309
4310<h5>Syntax:</h5>
4311<pre>
4312 declare i64 @llvm.readcyclecounter( )
4313</pre>
4314
4315<h5>Overview:</h5>
4316
4317
4318<p>
4319The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4320counter register (or similar low latency, high accuracy clocks) on those targets
4321that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4322As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4323should only be used for small timings.
4324</p>
4325
4326<h5>Semantics:</h5>
4327
4328<p>
4329When directly supported, reading the cycle counter should not modify any memory.
4330Implementations are allowed to either return a application specific value or a
4331system wide value. On backends without support, this is lowered to a constant 0.
4332</p>
4333
4334</div>
4335
4336<!-- ======================================================================= -->
4337<div class="doc_subsection">
4338 <a name="int_libc">Standard C Library Intrinsics</a>
4339</div>
4340
4341<div class="doc_text">
4342<p>
4343LLVM provides intrinsics for a few important standard C library functions.
4344These intrinsics allow source-language front-ends to pass information about the
4345alignment of the pointer arguments to the code generator, providing opportunity
4346for more efficient code generation.
4347</p>
4348
4349</div>
4350
4351<!-- _______________________________________________________________________ -->
4352<div class="doc_subsubsection">
4353 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4354</div>
4355
4356<div class="doc_text">
4357
4358<h5>Syntax:</h5>
4359<pre>
4360 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4361 i32 &lt;len&gt;, i32 &lt;align&gt;)
4362 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4363 i64 &lt;len&gt;, i32 &lt;align&gt;)
4364</pre>
4365
4366<h5>Overview:</h5>
4367
4368<p>
4369The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4370location to the destination location.
4371</p>
4372
4373<p>
4374Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4375intrinsics do not return a value, and takes an extra alignment argument.
4376</p>
4377
4378<h5>Arguments:</h5>
4379
4380<p>
4381The first argument is a pointer to the destination, the second is a pointer to
4382the source. The third argument is an integer argument
4383specifying the number of bytes to copy, and the fourth argument is the alignment
4384of the source and destination locations.
4385</p>
4386
4387<p>
4388If the call to this intrinisic has an alignment value that is not 0 or 1, then
4389the caller guarantees that both the source and destination pointers are aligned
4390to that boundary.
4391</p>
4392
4393<h5>Semantics:</h5>
4394
4395<p>
4396The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4397location to the destination location, which are not allowed to overlap. It
4398copies "len" bytes of memory over. If the argument is known to be aligned to
4399some boundary, this can be specified as the fourth argument, otherwise it should
4400be set to 0 or 1.
4401</p>
4402</div>
4403
4404
4405<!-- _______________________________________________________________________ -->
4406<div class="doc_subsubsection">
4407 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4408</div>
4409
4410<div class="doc_text">
4411
4412<h5>Syntax:</h5>
4413<pre>
4414 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4415 i32 &lt;len&gt;, i32 &lt;align&gt;)
4416 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
4417 i64 &lt;len&gt;, i32 &lt;align&gt;)
4418</pre>
4419
4420<h5>Overview:</h5>
4421
4422<p>
4423The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4424location to the destination location. It is similar to the
4425'<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4426</p>
4427
4428<p>
4429Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4430intrinsics do not return a value, and takes an extra alignment argument.
4431</p>
4432
4433<h5>Arguments:</h5>
4434
4435<p>
4436The first argument is a pointer to the destination, the second is a pointer to
4437the source. The third argument is an integer argument
4438specifying the number of bytes to copy, and the fourth argument is the alignment
4439of the source and destination locations.
4440</p>
4441
4442<p>
4443If the call to this intrinisic has an alignment value that is not 0 or 1, then
4444the caller guarantees that the source and destination pointers are aligned to
4445that boundary.
4446</p>
4447
4448<h5>Semantics:</h5>
4449
4450<p>
4451The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4452location to the destination location, which may overlap. It
4453copies "len" bytes of memory over. If the argument is known to be aligned to
4454some boundary, this can be specified as the fourth argument, otherwise it should
4455be set to 0 or 1.
4456</p>
4457</div>
4458
4459
4460<!-- _______________________________________________________________________ -->
4461<div class="doc_subsubsection">
4462 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4463</div>
4464
4465<div class="doc_text">
4466
4467<h5>Syntax:</h5>
4468<pre>
4469 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
4470 i32 &lt;len&gt;, i32 &lt;align&gt;)
4471 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
4472 i64 &lt;len&gt;, i32 &lt;align&gt;)
4473</pre>
4474
4475<h5>Overview:</h5>
4476
4477<p>
4478The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4479byte value.
4480</p>
4481
4482<p>
4483Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4484does not return a value, and takes an extra alignment argument.
4485</p>
4486
4487<h5>Arguments:</h5>
4488
4489<p>
4490The first argument is a pointer to the destination to fill, the second is the
4491byte value to fill it with, the third argument is an integer
4492argument specifying the number of bytes to fill, and the fourth argument is the
4493known alignment of destination location.
4494</p>
4495
4496<p>
4497If the call to this intrinisic has an alignment value that is not 0 or 1, then
4498the caller guarantees that the destination pointer is aligned to that boundary.
4499</p>
4500
4501<h5>Semantics:</h5>
4502
4503<p>
4504The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4505the
4506destination location. If the argument is known to be aligned to some boundary,
4507this can be specified as the fourth argument, otherwise it should be set to 0 or
45081.
4509</p>
4510</div>
4511
4512
4513<!-- _______________________________________________________________________ -->
4514<div class="doc_subsubsection">
4515 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4516</div>
4517
4518<div class="doc_text">
4519
4520<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4521<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]4522floating point or vector of floating point type. Not all targets support all
4523types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4524<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4525 declare float @llvm.sqrt.f32(float %Val)
4526 declare double @llvm.sqrt.f64(double %Val)
4527 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
4528 declare fp128 @llvm.sqrt.f128(fp128 %Val)
4529 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4530</pre>
4531
4532<h5>Overview:</h5>
4533
4534<p>
4535The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4536returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4537<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4538negative numbers (which allows for better optimization).
4539</p>
4540
4541<h5>Arguments:</h5>
4542
4543<p>
4544The argument and return value are floating point numbers of the same type.
4545</p>
4546
4547<h5>Semantics:</h5>
4548
4549<p>
4550This function returns the sqrt of the specified operand if it is a nonnegative
4551floating point number.
4552</p>
4553</div>
4554
4555<!-- _______________________________________________________________________ -->
4556<div class="doc_subsubsection">
4557 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4558</div>
4559
4560<div class="doc_text">
4561
4562<h5>Syntax:</h5>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4563<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]4564floating point or vector of floating point type. Not all targets support all
4565types however.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4566<pre>
Dale Johannesenf9adbb62007-10-02 17:47:38 +0000[diff] [blame]4567 declare float @llvm.powi.f32(float %Val, i32 %power)
4568 declare double @llvm.powi.f64(double %Val, i32 %power)
4569 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
4570 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
4571 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4572</pre>
4573
4574<h5>Overview:</h5>
4575
4576<p>
4577The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4578specified (positive or negative) power. The order of evaluation of
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4579multiplications is not defined. When a vector of floating point type is
4580used, the second argument remains a scalar integer value.
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4581</p>
4582
4583<h5>Arguments:</h5>
4584
4585<p>
4586The second argument is an integer power, and the first is a value to raise to
4587that power.
4588</p>
4589
4590<h5>Semantics:</h5>
4591
4592<p>
4593This function returns the first value raised to the second power with an
4594unspecified sequence of rounding operations.</p>
4595</div>
4596
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4597<!-- _______________________________________________________________________ -->
4598<div class="doc_subsubsection">
4599 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
4600</div>
4601
4602<div class="doc_text">
4603
4604<h5>Syntax:</h5>
4605<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
4606floating point or vector of floating point type. Not all targets support all
4607types however.
4608<pre>
4609 declare float @llvm.sin.f32(float %Val)
4610 declare double @llvm.sin.f64(double %Val)
4611 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
4612 declare fp128 @llvm.sin.f128(fp128 %Val)
4613 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
4614</pre>
4615
4616<h5>Overview:</h5>
4617
4618<p>
4619The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
4620</p>
4621
4622<h5>Arguments:</h5>
4623
4624<p>
4625The argument and return value are floating point numbers of the same type.
4626</p>
4627
4628<h5>Semantics:</h5>
4629
4630<p>
4631This function returns the sine of the specified operand, returning the
4632same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4633conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4634</div>
4635
4636<!-- _______________________________________________________________________ -->
4637<div class="doc_subsubsection">
4638 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
4639</div>
4640
4641<div class="doc_text">
4642
4643<h5>Syntax:</h5>
4644<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
4645floating point or vector of floating point type. Not all targets support all
4646types however.
4647<pre>
4648 declare float @llvm.cos.f32(float %Val)
4649 declare double @llvm.cos.f64(double %Val)
4650 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
4651 declare fp128 @llvm.cos.f128(fp128 %Val)
4652 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
4653</pre>
4654
4655<h5>Overview:</h5>
4656
4657<p>
4658The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
4659</p>
4660
4661<h5>Arguments:</h5>
4662
4663<p>
4664The argument and return value are floating point numbers of the same type.
4665</p>
4666
4667<h5>Semantics:</h5>
4668
4669<p>
4670This function returns the cosine of the specified operand, returning the
4671same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4672conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4673</div>
4674
4675<!-- _______________________________________________________________________ -->
4676<div class="doc_subsubsection">
4677 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
4678</div>
4679
4680<div class="doc_text">
4681
4682<h5>Syntax:</h5>
4683<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
4684floating point or vector of floating point type. Not all targets support all
4685types however.
4686<pre>
4687 declare float @llvm.pow.f32(float %Val, float %Power)
4688 declare double @llvm.pow.f64(double %Val, double %Power)
4689 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
4690 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
4691 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
4692</pre>
4693
4694<h5>Overview:</h5>
4695
4696<p>
4697The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
4698specified (positive or negative) power.
4699</p>
4700
4701<h5>Arguments:</h5>
4702
4703<p>
4704The second argument is a floating point power, and the first is a value to
4705raise to that power.
4706</p>
4707
4708<h5>Semantics:</h5>
4709
4710<p>
4711This function returns the first value raised to the second power,
4712returning the
4713same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmaneaba92e2007-10-17 18:05:13 +0000[diff] [blame]4714conditions in the same way.</p>
Dan Gohman361079c2007-10-15 20:30:11 +0000[diff] [blame]4715</div>
4716
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4717
4718<!-- ======================================================================= -->
4719<div class="doc_subsection">
4720 <a name="int_manip">Bit Manipulation Intrinsics</a>
4721</div>
4722
4723<div class="doc_text">
4724<p>
4725LLVM provides intrinsics for a few important bit manipulation operations.
4726These allow efficient code generation for some algorithms.
4727</p>
4728
4729</div>
4730
4731<!-- _______________________________________________________________________ -->
4732<div class="doc_subsubsection">
4733 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4734</div>
4735
4736<div class="doc_text">
4737
4738<h5>Syntax:</h5>
4739<p>This is an overloaded intrinsic function. You can use bswap on any integer
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4740type that is an even number of bytes (i.e. BitWidth % 16 == 0).
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4741<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4742 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
4743 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
4744 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4745</pre>
4746
4747<h5>Overview:</h5>
4748
4749<p>
4750The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
4751values with an even number of bytes (positive multiple of 16 bits). These are
4752useful for performing operations on data that is not in the target's native
4753byte order.
4754</p>
4755
4756<h5>Semantics:</h5>
4757
4758<p>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4759The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4760and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4761intrinsic returns an i32 value that has the four bytes of the input i32
4762swapped, 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]4763i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
4764<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4765additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4766</p>
4767
4768</div>
4769
4770<!-- _______________________________________________________________________ -->
4771<div class="doc_subsubsection">
4772 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4773</div>
4774
4775<div class="doc_text">
4776
4777<h5>Syntax:</h5>
4778<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4779width. Not all targets support all bit widths however.
4780<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4781 declare i8 @llvm.ctpop.i8 (i8 &lt;src&gt;)
4782 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4783 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4784 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
4785 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4786</pre>
4787
4788<h5>Overview:</h5>
4789
4790<p>
4791The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4792value.
4793</p>
4794
4795<h5>Arguments:</h5>
4796
4797<p>
4798The only argument is the value to be counted. The argument may be of any
4799integer type. The return type must match the argument type.
4800</p>
4801
4802<h5>Semantics:</h5>
4803
4804<p>
4805The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4806</p>
4807</div>
4808
4809<!-- _______________________________________________________________________ -->
4810<div class="doc_subsubsection">
4811 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4812</div>
4813
4814<div class="doc_text">
4815
4816<h5>Syntax:</h5>
4817<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4818integer bit width. Not all targets support all bit widths however.
4819<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4820 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
4821 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4822 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4823 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
4824 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4825</pre>
4826
4827<h5>Overview:</h5>
4828
4829<p>
4830The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4831leading zeros in a variable.
4832</p>
4833
4834<h5>Arguments:</h5>
4835
4836<p>
4837The only argument is the value to be counted. The argument may be of any
4838integer type. The return type must match the argument type.
4839</p>
4840
4841<h5>Semantics:</h5>
4842
4843<p>
4844The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4845in a variable. If the src == 0 then the result is the size in bits of the type
4846of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4847</p>
4848</div>
4849
4850
4851
4852<!-- _______________________________________________________________________ -->
4853<div class="doc_subsubsection">
4854 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4855</div>
4856
4857<div class="doc_text">
4858
4859<h5>Syntax:</h5>
4860<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
4861integer bit width. Not all targets support all bit widths however.
4862<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4863 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
4864 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4865 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4866 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
4867 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4868</pre>
4869
4870<h5>Overview:</h5>
4871
4872<p>
4873The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4874trailing zeros.
4875</p>
4876
4877<h5>Arguments:</h5>
4878
4879<p>
4880The only argument is the value to be counted. The argument may be of any
4881integer type. The return type must match the argument type.
4882</p>
4883
4884<h5>Semantics:</h5>
4885
4886<p>
4887The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4888in a variable. If the src == 0 then the result is the size in bits of the type
4889of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4890</p>
4891</div>
4892
4893<!-- _______________________________________________________________________ -->
4894<div class="doc_subsubsection">
4895 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
4896</div>
4897
4898<div class="doc_text">
4899
4900<h5>Syntax:</h5>
4901<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
4902on any integer bit width.
4903<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4904 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
4905 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]4906</pre>
4907
4908<h5>Overview:</h5>
4909<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
4910range of bits from an integer value and returns them in the same bit width as
4911the original value.</p>
4912
4913<h5>Arguments:</h5>
4914<p>The first argument, <tt>%val</tt> and the result may be integer types of
4915any bit width but they must have the same bit width. The second and third
4916arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
4917
4918<h5>Semantics:</h5>
4919<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
4920of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
4921<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
4922operates in forward mode.</p>
4923<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
4924right by <tt>%loBit</tt> bits and then ANDing it with a mask with
4925only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
4926<ol>
4927 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
4928 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
4929 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
4930 to determine the number of bits to retain.</li>
4931 <li>A mask of the retained bits is created by shifting a -1 value.</li>
4932 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
4933</ol>
4934<p>In reverse mode, a similar computation is made except that the bits are
4935returned in the reverse order. So, for example, if <tt>X</tt> has the value
4936<tt>i16 0x0ACF (101011001111)</tt> and we apply
4937<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
4938<tt>i16 0x0026 (000000100110)</tt>.</p>
4939</div>
4940
4941<div class="doc_subsubsection">
4942 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
4943</div>
4944
4945<div class="doc_text">
4946
4947<h5>Syntax:</h5>
4948<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
4949on any integer bit width.
4950<pre>
Chandler Carrutha228e392007-08-04 01:51:18 +0000[diff] [blame]4951 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
4952 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]4953</pre>
4954
4955<h5>Overview:</h5>
4956<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
4957of bits in an integer value with another integer value. It returns the integer
4958with the replaced bits.</p>
4959
4960<h5>Arguments:</h5>
4961<p>The first argument, <tt>%val</tt> and the result may be integer types of
4962any bit width but they must have the same bit width. <tt>%val</tt> is the value
4963whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
4964integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
4965type since they specify only a bit index.</p>
4966
4967<h5>Semantics:</h5>
4968<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
4969of operation: forwards and reverse. If <tt>%lo</tt> is greater than
4970<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
4971operates in forward mode.</p>
4972<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
4973truncating it down to the size of the replacement area or zero extending it
4974up to that size.</p>
4975<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
4976are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
4977in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
4978to the <tt>%hi</tt>th bit.
4979<p>In reverse mode, a similar computation is made except that the bits are
4980reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
4981<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
4982<h5>Examples:</h5>
4983<pre>
4984 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
4985 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
4986 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
4987 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
4988 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
4989</pre>
4990</div>
4991
4992<!-- ======================================================================= -->
4993<div class="doc_subsection">
4994 <a name="int_debugger">Debugger Intrinsics</a>
4995</div>
4996
4997<div class="doc_text">
4998<p>
4999The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5000are described in the <a
5001href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5002Debugging</a> document.
5003</p>
5004</div>
5005
5006
5007<!-- ======================================================================= -->
5008<div class="doc_subsection">
5009 <a name="int_eh">Exception Handling Intrinsics</a>
5010</div>
5011
5012<div class="doc_text">
5013<p> The LLVM exception handling intrinsics (which all start with
5014<tt>llvm.eh.</tt> prefix), are described in the <a
5015href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5016Handling</a> document. </p>
5017</div>
5018
5019<!-- ======================================================================= -->
5020<div class="doc_subsection">
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5021 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5022</div>
5023
5024<div class="doc_text">
5025<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5026 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5027 the <tt>nest</tt> attribute, from a function. The result is a callable
5028 function pointer lacking the nest parameter - the caller does not need
5029 to provide a value for it. Instead, the value to use is stored in
5030 advance in a "trampoline", a block of memory usually allocated
5031 on the stack, which also contains code to splice the nest value into the
5032 argument list. This is used to implement the GCC nested function address
5033 extension.
5034</p>
5035<p>
5036 For example, if the function is
5037 <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]5038 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]5039<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5040 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5041 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5042 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5043 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5044</pre>
Bill Wendlinge262b4c2007-09-22 09:23:55 +0000[diff] [blame]5045 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5046 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5047</div>
5048
5049<!-- _______________________________________________________________________ -->
5050<div class="doc_subsubsection">
5051 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5052</div>
5053<div class="doc_text">
5054<h5>Syntax:</h5>
5055<pre>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5056declare 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]5057</pre>
5058<h5>Overview:</h5>
5059<p>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5060 This fills the memory pointed to by <tt>tramp</tt> with code
5061 and returns a function pointer suitable for executing it.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5062</p>
5063<h5>Arguments:</h5>
5064<p>
5065 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5066 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5067 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sands35012212007-08-22 23:39:54 +0000[diff] [blame]5068 intrinsic. Note that the size and the alignment are target-specific - LLVM
5069 currently provides no portable way of determining them, so a front-end that
5070 generates this intrinsic needs to have some target-specific knowledge.
5071 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]5072</p>
5073<h5>Semantics:</h5>
5074<p>
5075 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5076 dependent code, turning it into a function. A pointer to this function is
5077 returned, but needs to be bitcast to an
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5078 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands7407a9f2007-09-11 14:10:23 +0000[diff] [blame]5079 before being called. The new function's signature is the same as that of
5080 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5081 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5082 of pointer type. Calling the new function is equivalent to calling
5083 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5084 missing <tt>nest</tt> argument. If, after calling
5085 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5086 modified, then the effect of any later call to the returned function pointer is
5087 undefined.
Duncan Sands38947cd2007-07-27 12:58:54 +0000[diff] [blame]5088</p>
5089</div>
5090
5091<!-- ======================================================================= -->
5092<div class="doc_subsection">
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5093 <a name="int_general">General Intrinsics</a>
5094</div>
5095
5096<div class="doc_text">
5097<p> This class of intrinsics is designed to be generic and has
5098no specific purpose. </p>
5099</div>
5100
5101<!-- _______________________________________________________________________ -->
5102<div class="doc_subsubsection">
5103 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
5104</div>
5105
5106<div class="doc_text">
5107
5108<h5>Syntax:</h5>
5109<pre>
5110 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5111</pre>
5112
5113<h5>Overview:</h5>
5114
5115<p>
5116The '<tt>llvm.var.annotation</tt>' intrinsic
5117</p>
5118
5119<h5>Arguments:</h5>
5120
5121<p>
5122The first argument is a pointer to a value, the second is a pointer to a
5123global string, the third is a pointer to a global string which is the source
5124file name, and the last argument is the line number.
5125</p>
5126
5127<h5>Semantics:</h5>
5128
5129<p>
5130This intrinsic allows annotation of local variables with arbitrary strings.
5131This can be useful for special purpose optimizations that want to look for these
5132 annotations. These have no other defined use, they are ignored by code
5133 generation and optimization.
5134</div>
5135
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5136<!-- _______________________________________________________________________ -->
5137<div class="doc_subsubsection">
Tanya Lattnerc9869b12007-09-21 23:57:59 +0000[diff] [blame]5138 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5139</div>
5140
5141<div class="doc_text">
5142
5143<h5>Syntax:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5144<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
5145any integer bit width.
5146</p>
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5147<pre>
Tanya Lattner09161fe2007-09-22 00:03:01 +0000[diff] [blame]5148 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5149 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5150 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5151 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
5152 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]5153</pre>
5154
5155<h5>Overview:</h5>
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5156
5157<p>
5158The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5159</p>
5160
5161<h5>Arguments:</h5>
5162
5163<p>
5164The first argument is an integer value (result of some expression),
5165the second is a pointer to a global string, the third is a pointer to a global
5166string which is the source file name, and the last argument is the line number.
Tanya Lattnere545be72007-09-21 23:56:27 +0000[diff] [blame]5167It returns the value of the first argument.
Tanya Lattnerb306a9e2007-09-21 22:59:12 +0000[diff] [blame]5168</p>
5169
5170<h5>Semantics:</h5>
5171
5172<p>
5173This intrinsic allows annotations to be put on arbitrary expressions
5174with arbitrary strings. This can be useful for special purpose optimizations
5175that want to look for these annotations. These have no other defined use, they
5176are ignored by code generation and optimization.
5177</div>
Dan Gohmanf17a25c2007-07-18 16:29:46 +0000[diff] [blame]5178
5179<!-- *********************************************************************** -->
5180<hr>
5181<address>
5182 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
5183 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
5184 <a href="http://validator.w3.org/check/referer"><img
5185 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
5186
5187 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
5188 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
5189 Last modified: $Date$
5190</address>
5191</body>
5192</html>