blob: 7a97ff5d1d4159a6606cba055b93fee9ec050b50 [file] [log] [blame]
Misha Brukmanc501f552004-03-01 17:47:27 +00001<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
Misha Brukman76307852003-11-08 01:05:38 +00003<html>
4<head>
5 <title>LLVM Assembly Language Reference Manual</title>
Reid Spencercb84e432004-08-26 20:44:00 +00006 <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.">
Misha Brukman76307852003-11-08 01:05:38 +000010 <link rel="stylesheet" href="llvm.css" type="text/css">
11</head>
Chris Lattner757528b0b2004-05-23 21:06:01 +000012
Misha Brukman76307852003-11-08 01:05:38 +000013<body>
Chris Lattner757528b0b2004-05-23 21:06:01 +000014
Chris Lattner48b383b02003-11-25 01:02:51 +000015<div class="doc_title"> LLVM Language Reference Manual </div>
Chris Lattner2f7c9632001-06-06 20:29:01 +000016<ol>
Misha Brukman76307852003-11-08 01:05:38 +000017 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
Chris Lattner6af02f32004-12-09 16:11:40 +000020 <li><a href="#highlevel">High Level Structure</a>
21 <ol>
22 <li><a href="#modulestructure">Module Structure</a></li>
Chris Lattnerd79749a2004-12-09 16:36:40 +000023 <li><a href="#linkage">Linkage Types</a></li>
Chris Lattner0132aff2005-05-06 22:57:40 +000024 <li><a href="#callingconv">Calling Conventions</a></li>
Chris Lattner6af02f32004-12-09 16:11:40 +000025 <li><a href="#globalvars">Global Variables</a></li>
Chris Lattner91c15c42006-01-23 23:23:47 +000026 <li><a href="#functionstructure">Functions</a></li>
Anton Korobeynikov546ea7e2007-04-29 18:02:48 +000027 <li><a href="#aliasstructure">Aliases</a>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +000028 <li><a href="#paramattrs">Parameter Attributes</a></li>
Chris Lattner91c15c42006-01-23 23:23:47 +000029 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
Reid Spencer50c723a2007-02-19 23:54:10 +000030 <li><a href="#datalayout">Data Layout</a></li>
Chris Lattner6af02f32004-12-09 16:11:40 +000031 </ol>
32 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +000033 <li><a href="#typesystem">Type System</a>
34 <ol>
Robert Bocchino820bc75b2006-02-17 21:18:08 +000035 <li><a href="#t_primitive">Primitive Types</a>
Chris Lattner48b383b02003-11-25 01:02:51 +000036 <ol>
Misha Brukman76307852003-11-08 01:05:38 +000037 <li><a href="#t_classifications">Type Classifications</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000038 </ol>
39 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +000040 <li><a href="#t_derived">Derived Types</a>
41 <ol>
Chris Lattner48b383b02003-11-25 01:02:51 +000042 <li><a href="#t_array">Array Type</a></li>
Misha Brukman76307852003-11-08 01:05:38 +000043 <li><a href="#t_function">Function Type</a></li>
44 <li><a href="#t_pointer">Pointer Type</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000045 <li><a href="#t_struct">Structure Type</a></li>
Andrew Lenharth8df88e22006-12-08 17:13:00 +000046 <li><a href="#t_pstruct">Packed Structure Type</a></li>
Reid Spencer404a3252007-02-15 03:07:05 +000047 <li><a href="#t_vector">Vector Type</a></li>
Chris Lattner37b6b092005-04-25 17:34:15 +000048 <li><a href="#t_opaque">Opaque Type</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000049 </ol>
50 </li>
51 </ol>
52 </li>
Chris Lattner6af02f32004-12-09 16:11:40 +000053 <li><a href="#constants">Constants</a>
Chris Lattner74d3f822004-12-09 17:30:23 +000054 <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>
Chris Lattner48b383b02003-11-25 01:02:51 +000061 </li>
Chris Lattner98f013c2006-01-25 23:47:57 +000062 <li><a href="#othervalues">Other Values</a>
63 <ol>
64 <li><a href="#inlineasm">Inline Assembler Expressions</a>
65 </ol>
66 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +000067 <li><a href="#instref">Instruction Reference</a>
68 <ol>
69 <li><a href="#terminators">Terminator Instructions</a>
70 <ol>
Chris Lattner48b383b02003-11-25 01:02:51 +000071 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
72 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
Misha Brukman76307852003-11-08 01:05:38 +000073 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
74 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000075 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
Chris Lattner08b7d5b2004-10-16 18:04:13 +000076 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000077 </ol>
78 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +000079 <li><a href="#binaryops">Binary Operations</a>
80 <ol>
Chris Lattner48b383b02003-11-25 01:02:51 +000081 <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>
Reid Spencer7e80b0b2006-10-26 06:15:43 +000084 <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>
Reid Spencer7eb55b32006-11-02 01:53:59 +000087 <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>
Chris Lattner48b383b02003-11-25 01:02:51 +000090 </ol>
91 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +000092 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
93 <ol>
Reid Spencer2ab01932007-02-02 13:57:07 +000094 <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>
Misha Brukman76307852003-11-08 01:05:38 +000097 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000098 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
Misha Brukman76307852003-11-08 01:05:38 +000099 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000100 </ol>
101 </li>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000102 <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>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000107 </ol>
108 </li>
Chris Lattner6ab66722006-08-15 00:45:58 +0000109 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000110 <ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000111 <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>
Robert Bocchino820bc75b2006-02-17 21:18:08 +0000114 <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>
Chris Lattner48b383b02003-11-25 01:02:51 +0000117 </ol>
118 </li>
Reid Spencer97c5fa42006-11-08 01:18:52 +0000119 <li><a href="#convertops">Conversion Operations</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000120 <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>
Reid Spencer51b07252006-11-09 23:03:26 +0000126 <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>
Reid Spencerb7344ff2006-11-11 21:00:47 +0000130 <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>
Reid Spencer5b950642006-11-11 23:08:07 +0000132 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000133 </ol>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000134 <li><a href="#otherops">Other Operations</a>
135 <ol>
Reid Spencerc828a0e2006-11-18 21:50:54 +0000136 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
137 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000138 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000139 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000140 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
Chris Lattner33337472006-01-13 23:26:01 +0000141 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000142 </ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000143 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000144 </ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000145 </li>
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000146 <li><a href="#intrinsics">Intrinsic Functions</a>
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000147 <ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000148 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
149 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000150 <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>
Chris Lattner48b383b02003-11-25 01:02:51 +0000153 </ol>
154 </li>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000155 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
156 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000157 <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>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000160 </ol>
161 </li>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000162 <li><a href="#int_codegen">Code Generator Intrinsics</a>
163 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000164 <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>
John Criswellaa1c3c12004-04-09 16:43:20 +0000171 </ol>
172 </li>
Chris Lattnerfee11462004-02-12 17:01:32 +0000173 <li><a href="#int_libc">Standard C Library Intrinsics</a>
174 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000175 <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>
Chris Lattnerfee11462004-02-12 17:01:32 +0000180 </ol>
181 </li>
Nate Begeman0f223bb2006-01-13 23:26:38 +0000182 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000183 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000184 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
Chris Lattnerb748c672006-01-16 22:34:14 +0000185 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
186 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
187 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
Reid Spencer5bf54c82007-04-11 23:23:49 +0000188 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
189 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000190 </ol>
191 </li>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000192 <li><a href="#int_debugger">Debugger intrinsics</a></li>
Jim Laskey2211f492007-03-14 19:31:19 +0000193 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +0000194 <li><a href="#int_atomics">Atomic Operations and Synchronization Intrinsics</a>
195 <ol>
196 <li><a href="#int_lcs">'<tt>llvm.atomic.lcs.*</tt>' Intrinsic</a></li>
197 <li><a href="#int_ls">'<tt>llvm.atomic.ls.*</tt>' Intrinsic</a></li>
198 <li><a href="#int_las">'<tt>llvm.atomic.las.*</tt>' Intrinsic</a></li>
199 <li><a href="#int_lss">'<tt>llvm.atomic.lss.*</tt>' Intrinsic</a></li>
200 <li><a href="#int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a></li>
201 </ol>
202 </li>
Duncan Sands644f9172007-07-27 12:58:54 +0000203 <li><a href="#int_trampoline">Trampoline Intrinsics</a>
204 <ol>
205 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
206 <li><a href="#int_at">'<tt>llvm.adjust.trampoline</tt>' Intrinsic</a></li>
207 </ol>
208 </li>
Reid Spencer5b2cb0f2007-07-20 19:59:11 +0000209 <li><a href="#int_general">General intrinsics</a>
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000210 <ol>
Reid Spencer5b2cb0f2007-07-20 19:59:11 +0000211 <li><a href="#int_var_annotation">
212 <tt>llvm.var.annotation</tt>' Intrinsic</a></li>
213 </ol>
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000214 </li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000215 </ol>
216 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000217</ol>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000218
219<div class="doc_author">
220 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
221 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
Misha Brukman76307852003-11-08 01:05:38 +0000222</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000223
Chris Lattner2f7c9632001-06-06 20:29:01 +0000224<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +0000225<div class="doc_section"> <a name="abstract">Abstract </a></div>
226<!-- *********************************************************************** -->
Chris Lattner757528b0b2004-05-23 21:06:01 +0000227
Misha Brukman76307852003-11-08 01:05:38 +0000228<div class="doc_text">
Chris Lattner48b383b02003-11-25 01:02:51 +0000229<p>This document is a reference manual for the LLVM assembly language.
230LLVM is an SSA based representation that provides type safety,
231low-level operations, flexibility, and the capability of representing
232'all' high-level languages cleanly. It is the common code
233representation used throughout all phases of the LLVM compilation
234strategy.</p>
Misha Brukman76307852003-11-08 01:05:38 +0000235</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000236
Chris Lattner2f7c9632001-06-06 20:29:01 +0000237<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +0000238<div class="doc_section"> <a name="introduction">Introduction</a> </div>
239<!-- *********************************************************************** -->
Chris Lattner757528b0b2004-05-23 21:06:01 +0000240
Misha Brukman76307852003-11-08 01:05:38 +0000241<div class="doc_text">
Chris Lattner757528b0b2004-05-23 21:06:01 +0000242
Chris Lattner48b383b02003-11-25 01:02:51 +0000243<p>The LLVM code representation is designed to be used in three
Gabor Greifa54634a2007-07-06 22:07:22 +0000244different forms: as an in-memory compiler IR, as an on-disk bitcode
Chris Lattner48b383b02003-11-25 01:02:51 +0000245representation (suitable for fast loading by a Just-In-Time compiler),
246and as a human readable assembly language representation. This allows
247LLVM to provide a powerful intermediate representation for efficient
248compiler transformations and analysis, while providing a natural means
249to debug and visualize the transformations. The three different forms
250of LLVM are all equivalent. This document describes the human readable
251representation and notation.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000252
John Criswell4a3327e2005-05-13 22:25:59 +0000253<p>The LLVM representation aims to be light-weight and low-level
Chris Lattner48b383b02003-11-25 01:02:51 +0000254while being expressive, typed, and extensible at the same time. It
255aims to be a "universal IR" of sorts, by being at a low enough level
256that high-level ideas may be cleanly mapped to it (similar to how
257microprocessors are "universal IR's", allowing many source languages to
258be mapped to them). By providing type information, LLVM can be used as
259the target of optimizations: for example, through pointer analysis, it
260can be proven that a C automatic variable is never accessed outside of
261the current function... allowing it to be promoted to a simple SSA
262value instead of a memory location.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000263
Misha Brukman76307852003-11-08 01:05:38 +0000264</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000265
Chris Lattner2f7c9632001-06-06 20:29:01 +0000266<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +0000267<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000268
Misha Brukman76307852003-11-08 01:05:38 +0000269<div class="doc_text">
Chris Lattner757528b0b2004-05-23 21:06:01 +0000270
Chris Lattner48b383b02003-11-25 01:02:51 +0000271<p>It is important to note that this document describes 'well formed'
272LLVM assembly language. There is a difference between what the parser
273accepts and what is considered 'well formed'. For example, the
274following instruction is syntactically okay, but not well formed:</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000275
Bill Wendling3716c5d2007-05-29 09:04:49 +0000276<div class="doc_code">
Chris Lattner757528b0b2004-05-23 21:06:01 +0000277<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000278%x = <a href="#i_add">add</a> i32 1, %x
Chris Lattner757528b0b2004-05-23 21:06:01 +0000279</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000280</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000281
Chris Lattner48b383b02003-11-25 01:02:51 +0000282<p>...because the definition of <tt>%x</tt> does not dominate all of
283its uses. The LLVM infrastructure provides a verification pass that may
284be used to verify that an LLVM module is well formed. This pass is
John Criswell4a3327e2005-05-13 22:25:59 +0000285automatically run by the parser after parsing input assembly and by
Gabor Greifa54634a2007-07-06 22:07:22 +0000286the optimizer before it outputs bitcode. The violations pointed out
Chris Lattner48b383b02003-11-25 01:02:51 +0000287by the verifier pass indicate bugs in transformation passes or input to
288the parser.</p>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000289</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000290
Reid Spencer5b2cb0f2007-07-20 19:59:11 +0000291<!-- Describe the typesetting conventions here. -->
Chris Lattner757528b0b2004-05-23 21:06:01 +0000292
Chris Lattner2f7c9632001-06-06 20:29:01 +0000293<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +0000294<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000295<!-- *********************************************************************** -->
Chris Lattner757528b0b2004-05-23 21:06:01 +0000296
Misha Brukman76307852003-11-08 01:05:38 +0000297<div class="doc_text">
Chris Lattner757528b0b2004-05-23 21:06:01 +0000298
Reid Spencerb23b65f2007-08-07 14:34:28 +0000299 <p>LLVM identifiers come in two basic types: global and local. Global
300 identifiers (functions, global variables) begin with the @ character. Local
301 identifiers (register names, types) begin with the % character. Additionally,
302 there are three different formats for identifiers, for different purposes:
Chris Lattner757528b0b2004-05-23 21:06:01 +0000303
Chris Lattner2f7c9632001-06-06 20:29:01 +0000304<ol>
Reid Spencerb23b65f2007-08-07 14:34:28 +0000305 <li>Named values are represented as a string of characters with their prefix.
306 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
307 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
Chris Lattnerd79749a2004-12-09 16:36:40 +0000308 Identifiers which require other characters in their names can be surrounded
Reid Spencerb23b65f2007-08-07 14:34:28 +0000309 with quotes. In this way, anything except a <tt>&quot;</tt> character can
310 be used in a named value.</li>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000311
Reid Spencerb23b65f2007-08-07 14:34:28 +0000312 <li>Unnamed values are represented as an unsigned numeric value with their
313 prefix. For example, %12, @2, %44.</li>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000314
Reid Spencer8f08d802004-12-09 18:02:53 +0000315 <li>Constants, which are described in a <a href="#constants">section about
316 constants</a>, below.</li>
Misha Brukman76307852003-11-08 01:05:38 +0000317</ol>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000318
Reid Spencerb23b65f2007-08-07 14:34:28 +0000319<p>LLVM requires that values start with a prefix for two reasons: Compilers
Chris Lattnerd79749a2004-12-09 16:36:40 +0000320don't need to worry about name clashes with reserved words, and the set of
321reserved words may be expanded in the future without penalty. Additionally,
322unnamed identifiers allow a compiler to quickly come up with a temporary
323variable without having to avoid symbol table conflicts.</p>
324
Chris Lattner48b383b02003-11-25 01:02:51 +0000325<p>Reserved words in LLVM are very similar to reserved words in other
Reid Spencer5b950642006-11-11 23:08:07 +0000326languages. There are keywords for different opcodes
327('<tt><a href="#i_add">add</a></tt>',
328 '<tt><a href="#i_bitcast">bitcast</a></tt>',
329 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000330href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
Chris Lattnerd79749a2004-12-09 16:36:40 +0000331and others. These reserved words cannot conflict with variable names, because
Reid Spencerb23b65f2007-08-07 14:34:28 +0000332none of them start with a prefix character ('%' or '@').</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000333
334<p>Here is an example of LLVM code to multiply the integer variable
335'<tt>%X</tt>' by 8:</p>
336
Misha Brukman76307852003-11-08 01:05:38 +0000337<p>The easy way:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000338
Bill Wendling3716c5d2007-05-29 09:04:49 +0000339<div class="doc_code">
Chris Lattnerd79749a2004-12-09 16:36:40 +0000340<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000341%result = <a href="#i_mul">mul</a> i32 %X, 8
Chris Lattnerd79749a2004-12-09 16:36:40 +0000342</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000343</div>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000344
Misha Brukman76307852003-11-08 01:05:38 +0000345<p>After strength reduction:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000346
Bill Wendling3716c5d2007-05-29 09:04:49 +0000347<div class="doc_code">
Chris Lattnerd79749a2004-12-09 16:36:40 +0000348<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000349%result = <a href="#i_shl">shl</a> i32 %X, i8 3
Chris Lattnerd79749a2004-12-09 16:36:40 +0000350</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000351</div>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000352
Misha Brukman76307852003-11-08 01:05:38 +0000353<p>And the hard way:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000354
Bill Wendling3716c5d2007-05-29 09:04:49 +0000355<div class="doc_code">
Chris Lattnerd79749a2004-12-09 16:36:40 +0000356<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000357<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
358<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
359%result = <a href="#i_add">add</a> i32 %1, %1
Chris Lattnerd79749a2004-12-09 16:36:40 +0000360</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000361</div>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000362
Chris Lattner48b383b02003-11-25 01:02:51 +0000363<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
364important lexical features of LLVM:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000365
Chris Lattner2f7c9632001-06-06 20:29:01 +0000366<ol>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000367
368 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
369 line.</li>
370
371 <li>Unnamed temporaries are created when the result of a computation is not
372 assigned to a named value.</li>
373
Misha Brukman76307852003-11-08 01:05:38 +0000374 <li>Unnamed temporaries are numbered sequentially</li>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000375
Misha Brukman76307852003-11-08 01:05:38 +0000376</ol>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000377
John Criswell02fdc6f2005-05-12 16:52:32 +0000378<p>...and it also shows a convention that we follow in this document. When
Chris Lattnerd79749a2004-12-09 16:36:40 +0000379demonstrating instructions, we will follow an instruction with a comment that
380defines the type and name of value produced. Comments are shown in italic
381text.</p>
382
Misha Brukman76307852003-11-08 01:05:38 +0000383</div>
Chris Lattner6af02f32004-12-09 16:11:40 +0000384
385<!-- *********************************************************************** -->
386<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
387<!-- *********************************************************************** -->
388
389<!-- ======================================================================= -->
390<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
391</div>
392
393<div class="doc_text">
394
395<p>LLVM programs are composed of "Module"s, each of which is a
396translation unit of the input programs. Each module consists of
397functions, global variables, and symbol table entries. Modules may be
398combined together with the LLVM linker, which merges function (and
399global variable) definitions, resolves forward declarations, and merges
400symbol table entries. Here is an example of the "hello world" module:</p>
401
Bill Wendling3716c5d2007-05-29 09:04:49 +0000402<div class="doc_code">
Chris Lattner6af02f32004-12-09 16:11:40 +0000403<pre><i>; Declare the string constant as a global constant...</i>
Chris Lattner2b0bf4f2007-06-12 17:00:26 +0000404<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
405 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
Chris Lattner6af02f32004-12-09 16:11:40 +0000406
407<i>; External declaration of the puts function</i>
Chris Lattner2b0bf4f2007-06-12 17:00:26 +0000408<a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
Chris Lattner6af02f32004-12-09 16:11:40 +0000409
410<i>; Definition of main function</i>
Chris Lattner2b0bf4f2007-06-12 17:00:26 +0000411define i32 @main() { <i>; i32()* </i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000412 <i>; Convert [13x i8 ]* to i8 *...</i>
Chris Lattner6af02f32004-12-09 16:11:40 +0000413 %cast210 = <a
Chris Lattner2150cde2007-06-12 17:01:15 +0000414 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
Chris Lattner6af02f32004-12-09 16:11:40 +0000415
416 <i>; Call puts function to write out the string to stdout...</i>
417 <a
Chris Lattner2b0bf4f2007-06-12 17:00:26 +0000418 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
Chris Lattner6af02f32004-12-09 16:11:40 +0000419 <a
Bill Wendling3716c5d2007-05-29 09:04:49 +0000420 href="#i_ret">ret</a> i32 0<br>}<br>
421</pre>
422</div>
Chris Lattner6af02f32004-12-09 16:11:40 +0000423
424<p>This example is made up of a <a href="#globalvars">global variable</a>
425named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
426function, and a <a href="#functionstructure">function definition</a>
427for "<tt>main</tt>".</p>
428
Chris Lattnerd79749a2004-12-09 16:36:40 +0000429<p>In general, a module is made up of a list of global values,
430where both functions and global variables are global values. Global values are
431represented by a pointer to a memory location (in this case, a pointer to an
432array of char, and a pointer to a function), and have one of the following <a
433href="#linkage">linkage types</a>.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000434
Chris Lattnerd79749a2004-12-09 16:36:40 +0000435</div>
436
437<!-- ======================================================================= -->
438<div class="doc_subsection">
439 <a name="linkage">Linkage Types</a>
440</div>
441
442<div class="doc_text">
443
444<p>
445All Global Variables and Functions have one of the following types of linkage:
446</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000447
448<dl>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000449
Chris Lattner6af02f32004-12-09 16:11:40 +0000450 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000451
452 <dd>Global values with internal linkage are only directly accessible by
453 objects in the current module. In particular, linking code into a module with
454 an internal global value may cause the internal to be renamed as necessary to
455 avoid collisions. Because the symbol is internal to the module, all
456 references can be updated. This corresponds to the notion of the
Chris Lattnere20b4702007-01-14 06:51:48 +0000457 '<tt>static</tt>' keyword in C.
Chris Lattner6af02f32004-12-09 16:11:40 +0000458 </dd>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000459
Chris Lattner6af02f32004-12-09 16:11:40 +0000460 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000461
Chris Lattnere20b4702007-01-14 06:51:48 +0000462 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
463 the same name when linkage occurs. This is typically used to implement
464 inline functions, templates, or other code which must be generated in each
465 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
466 allowed to be discarded.
Chris Lattner6af02f32004-12-09 16:11:40 +0000467 </dd>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000468
Chris Lattner6af02f32004-12-09 16:11:40 +0000469 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000470
471 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
472 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
Chris Lattnere20b4702007-01-14 06:51:48 +0000473 used for globals that may be emitted in multiple translation units, but that
474 are not guaranteed to be emitted into every translation unit that uses them.
475 One example of this are common globals in C, such as "<tt>int X;</tt>" at
476 global scope.
Chris Lattner6af02f32004-12-09 16:11:40 +0000477 </dd>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000478
Chris Lattner6af02f32004-12-09 16:11:40 +0000479 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000480
481 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
482 pointer to array type. When two global variables with appending linkage are
483 linked together, the two global arrays are appended together. This is the
484 LLVM, typesafe, equivalent of having the system linker append together
485 "sections" with identical names when .o files are linked.
Chris Lattner6af02f32004-12-09 16:11:40 +0000486 </dd>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000487
Anton Korobeynikova0554d92007-01-12 19:20:47 +0000488 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
489 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
490 until linked, if not linked, the symbol becomes null instead of being an
491 undefined reference.
492 </dd>
Anton Korobeynikova0554d92007-01-12 19:20:47 +0000493
Chris Lattner6af02f32004-12-09 16:11:40 +0000494 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000495
496 <dd>If none of the above identifiers are used, the global is externally
497 visible, meaning that it participates in linkage and can be used to resolve
498 external symbol references.
Chris Lattner6af02f32004-12-09 16:11:40 +0000499 </dd>
Reid Spencer7972c472007-04-11 23:49:50 +0000500</dl>
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000501
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000502 <p>
503 The next two types of linkage are targeted for Microsoft Windows platform
504 only. They are designed to support importing (exporting) symbols from (to)
505 DLLs.
506 </p>
507
Anton Korobeynikova0554d92007-01-12 19:20:47 +0000508 <dl>
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000509 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
510
511 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
512 or variable via a global pointer to a pointer that is set up by the DLL
513 exporting the symbol. On Microsoft Windows targets, the pointer name is
514 formed by combining <code>_imp__</code> and the function or variable name.
515 </dd>
516
517 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
518
519 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
520 pointer to a pointer in a DLL, so that it can be referenced with the
521 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
522 name is formed by combining <code>_imp__</code> and the function or variable
523 name.
524 </dd>
525
Chris Lattner6af02f32004-12-09 16:11:40 +0000526</dl>
527
Anton Korobeynikova0554d92007-01-12 19:20:47 +0000528<p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
Chris Lattner6af02f32004-12-09 16:11:40 +0000529variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
530variable and was linked with this one, one of the two would be renamed,
531preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
532external (i.e., lacking any linkage declarations), they are accessible
Reid Spencer92c671e2007-01-05 00:59:10 +0000533outside of the current module.</p>
534<p>It is illegal for a function <i>declaration</i>
535to have any linkage type other than "externally visible", <tt>dllimport</tt>,
Anton Korobeynikova0554d92007-01-12 19:20:47 +0000536or <tt>extern_weak</tt>.</p>
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000537<p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
538linkages.
Chris Lattner6af02f32004-12-09 16:11:40 +0000539</div>
540
541<!-- ======================================================================= -->
542<div class="doc_subsection">
Chris Lattner0132aff2005-05-06 22:57:40 +0000543 <a name="callingconv">Calling Conventions</a>
544</div>
545
546<div class="doc_text">
547
548<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
549and <a href="#i_invoke">invokes</a> can all have an optional calling convention
550specified for the call. The calling convention of any pair of dynamic
551caller/callee must match, or the behavior of the program is undefined. The
552following calling conventions are supported by LLVM, and more may be added in
553the future:</p>
554
555<dl>
556 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
557
558 <dd>This calling convention (the default if no other calling convention is
559 specified) matches the target C calling conventions. This calling convention
John Criswell02fdc6f2005-05-12 16:52:32 +0000560 supports varargs function calls and tolerates some mismatch in the declared
Reid Spencer72ba4992006-12-31 21:30:18 +0000561 prototype and implemented declaration of the function (as does normal C).
Chris Lattner0132aff2005-05-06 22:57:40 +0000562 </dd>
563
564 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
565
566 <dd>This calling convention attempts to make calls as fast as possible
567 (e.g. by passing things in registers). This calling convention allows the
568 target to use whatever tricks it wants to produce fast code for the target,
Chris Lattnerc792eb32005-05-06 23:08:23 +0000569 without having to conform to an externally specified ABI. Implementations of
570 this convention should allow arbitrary tail call optimization to be supported.
571 This calling convention does not support varargs and requires the prototype of
572 all callees to exactly match the prototype of the function definition.
Chris Lattner0132aff2005-05-06 22:57:40 +0000573 </dd>
574
575 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
576
577 <dd>This calling convention attempts to make code in the caller as efficient
578 as possible under the assumption that the call is not commonly executed. As
579 such, these calls often preserve all registers so that the call does not break
580 any live ranges in the caller side. This calling convention does not support
581 varargs and requires the prototype of all callees to exactly match the
582 prototype of the function definition.
583 </dd>
584
Chris Lattner573f64e2005-05-07 01:46:40 +0000585 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
Chris Lattner0132aff2005-05-06 22:57:40 +0000586
587 <dd>Any calling convention may be specified by number, allowing
588 target-specific calling conventions to be used. Target specific calling
589 conventions start at 64.
590 </dd>
Chris Lattner573f64e2005-05-07 01:46:40 +0000591</dl>
Chris Lattner0132aff2005-05-06 22:57:40 +0000592
593<p>More calling conventions can be added/defined on an as-needed basis, to
594support pascal conventions or any other well-known target-independent
595convention.</p>
596
597</div>
598
599<!-- ======================================================================= -->
600<div class="doc_subsection">
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000601 <a name="visibility">Visibility Styles</a>
602</div>
603
604<div class="doc_text">
605
606<p>
607All Global Variables and Functions have one of the following visibility styles:
608</p>
609
610<dl>
611 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
612
613 <dd>On ELF, default visibility means that the declaration is visible to other
614 modules and, in shared libraries, means that the declared entity may be
615 overridden. On Darwin, default visibility means that the declaration is
616 visible to other modules. Default visibility corresponds to "external
617 linkage" in the language.
618 </dd>
619
620 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
621
622 <dd>Two declarations of an object with hidden visibility refer to the same
623 object if they are in the same shared object. Usually, hidden visibility
624 indicates that the symbol will not be placed into the dynamic symbol table,
625 so no other module (executable or shared library) can reference it
626 directly.
627 </dd>
628
Anton Korobeynikov39f3cff2007-04-29 18:35:00 +0000629 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
630
631 <dd>On ELF, protected visibility indicates that the symbol will be placed in
632 the dynamic symbol table, but that references within the defining module will
633 bind to the local symbol. That is, the symbol cannot be overridden by another
634 module.
635 </dd>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000636</dl>
637
638</div>
639
640<!-- ======================================================================= -->
641<div class="doc_subsection">
Chris Lattner6af02f32004-12-09 16:11:40 +0000642 <a name="globalvars">Global Variables</a>
643</div>
644
645<div class="doc_text">
646
Chris Lattner5d5aede2005-02-12 19:30:21 +0000647<p>Global variables define regions of memory allocated at compilation time
Chris Lattner662c8722005-11-12 00:45:07 +0000648instead of run-time. Global variables may optionally be initialized, may have
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000649an explicit section to be placed in, and may have an optional explicit alignment
650specified. A variable may be defined as "thread_local", which means that it
651will not be shared by threads (each thread will have a separated copy of the
652variable). A variable may be defined as a global "constant," which indicates
653that the contents of the variable will <b>never</b> be modified (enabling better
Chris Lattner5d5aede2005-02-12 19:30:21 +0000654optimization, allowing the global data to be placed in the read-only section of
655an executable, etc). Note that variables that need runtime initialization
John Criswell4c0cf7f2005-10-24 16:17:18 +0000656cannot be marked "constant" as there is a store to the variable.</p>
Chris Lattner5d5aede2005-02-12 19:30:21 +0000657
658<p>
659LLVM explicitly allows <em>declarations</em> of global variables to be marked
660constant, even if the final definition of the global is not. This capability
661can be used to enable slightly better optimization of the program, but requires
662the language definition to guarantee that optimizations based on the
663'constantness' are valid for the translation units that do not include the
664definition.
665</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000666
667<p>As SSA values, global variables define pointer values that are in
668scope (i.e. they dominate) all basic blocks in the program. Global
669variables always define a pointer to their "content" type because they
670describe a region of memory, and all memory objects in LLVM are
671accessed through pointers.</p>
672
Chris Lattner662c8722005-11-12 00:45:07 +0000673<p>LLVM allows an explicit section to be specified for globals. If the target
674supports it, it will emit globals to the section specified.</p>
675
Chris Lattner54611b42005-11-06 08:02:57 +0000676<p>An explicit alignment may be specified for a global. If not present, or if
677the alignment is set to zero, the alignment of the global is set by the target
678to whatever it feels convenient. If an explicit alignment is specified, the
679global is forced to have at least that much alignment. All alignments must be
680a power of 2.</p>
681
Chris Lattner5760c502007-01-14 00:27:09 +0000682<p>For example, the following defines a global with an initializer, section,
683 and alignment:</p>
684
Bill Wendling3716c5d2007-05-29 09:04:49 +0000685<div class="doc_code">
Chris Lattner5760c502007-01-14 00:27:09 +0000686<pre>
Chris Lattner0a2d0992007-07-13 20:01:46 +0000687@G = constant float 1.0, section "foo", align 4
Chris Lattner5760c502007-01-14 00:27:09 +0000688</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000689</div>
Chris Lattner5760c502007-01-14 00:27:09 +0000690
Chris Lattner6af02f32004-12-09 16:11:40 +0000691</div>
692
693
694<!-- ======================================================================= -->
695<div class="doc_subsection">
696 <a name="functionstructure">Functions</a>
697</div>
698
699<div class="doc_text">
700
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000701<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
702an optional <a href="#linkage">linkage type</a>, an optional
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000703<a href="#visibility">visibility style</a>, an optional
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000704<a href="#callingconv">calling convention</a>, a return type, an optional
705<a href="#paramattrs">parameter attribute</a> for the return type, a function
706name, a (possibly empty) argument list (each with optional
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000707<a href="#paramattrs">parameter attributes</a>), an optional section, an
708optional alignment, an opening curly brace, a list of basic blocks, and a
709closing curly brace.
710
711LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
712optional <a href="#linkage">linkage type</a>, an optional
713<a href="#visibility">visibility style</a>, an optional
714<a href="#callingconv">calling convention</a>, a return type, an optional
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000715<a href="#paramattrs">parameter attribute</a> for the return type, a function
716name, a possibly empty list of arguments, and an optional alignment.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000717
718<p>A function definition contains a list of basic blocks, forming the CFG for
719the function. Each basic block may optionally start with a label (giving the
720basic block a symbol table entry), contains a list of instructions, and ends
721with a <a href="#terminators">terminator</a> instruction (such as a branch or
722function return).</p>
723
Chris Lattnera59fb102007-06-08 16:52:14 +0000724<p>The first basic block in a function is special in two ways: it is immediately
Chris Lattner6af02f32004-12-09 16:11:40 +0000725executed on entrance to the function, and it is not allowed to have predecessor
726basic blocks (i.e. there can not be any branches to the entry block of a
727function). Because the block can have no predecessors, it also cannot have any
728<a href="#i_phi">PHI nodes</a>.</p>
729
Chris Lattner662c8722005-11-12 00:45:07 +0000730<p>LLVM allows an explicit section to be specified for functions. If the target
731supports it, it will emit functions to the section specified.</p>
732
Chris Lattner54611b42005-11-06 08:02:57 +0000733<p>An explicit alignment may be specified for a function. If not present, or if
734the alignment is set to zero, the alignment of the function is set by the target
735to whatever it feels convenient. If an explicit alignment is specified, the
736function is forced to have at least that much alignment. All alignments must be
737a power of 2.</p>
738
Chris Lattner6af02f32004-12-09 16:11:40 +0000739</div>
740
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000741
742<!-- ======================================================================= -->
743<div class="doc_subsection">
744 <a name="aliasstructure">Aliases</a>
745</div>
746<div class="doc_text">
747 <p>Aliases act as "second name" for the aliasee value (which can be either
Anton Korobeynikovb18f8f82007-04-28 13:45:00 +0000748 function or global variable or bitcast of global value). Aliases may have an
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000749 optional <a href="#linkage">linkage type</a>, and an
750 optional <a href="#visibility">visibility style</a>.</p>
751
752 <h5>Syntax:</h5>
753
Bill Wendling3716c5d2007-05-29 09:04:49 +0000754<div class="doc_code">
Bill Wendling2d8b9a82007-05-29 09:42:13 +0000755<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000756@&lt;Name&gt; = [Linkage] [Visibility] alias &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
Bill Wendling2d8b9a82007-05-29 09:42:13 +0000757</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000758</div>
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000759
760</div>
761
762
763
Chris Lattner91c15c42006-01-23 23:23:47 +0000764<!-- ======================================================================= -->
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000765<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
766<div class="doc_text">
767 <p>The return type and each parameter of a function type may have a set of
768 <i>parameter attributes</i> associated with them. Parameter attributes are
769 used to communicate additional information about the result or parameters of
770 a function. Parameter attributes are considered to be part of the function
771 type so two functions types that differ only by the parameter attributes
772 are different function types.</p>
773
Reid Spencercf7ebf52007-01-15 18:27:39 +0000774 <p>Parameter attributes are simple keywords that follow the type specified. If
775 multiple parameter attributes are needed, they are space separated. For
Bill Wendling3716c5d2007-05-29 09:04:49 +0000776 example:</p>
777
778<div class="doc_code">
779<pre>
Reid Spencer314e1cb2007-07-19 23:13:04 +0000780%someFunc = i16 (i8 signext %someParam) zeroext
781%someFunc = i16 (i8 zeroext %someParam) zeroext
Bill Wendling3716c5d2007-05-29 09:04:49 +0000782</pre>
783</div>
784
Anton Korobeynikova0554d92007-01-12 19:20:47 +0000785 <p>Note that the two function types above are unique because the parameter has
Reid Spencer314e1cb2007-07-19 23:13:04 +0000786 a different attribute (<tt>signext</tt> in the first one, <tt>zeroext</tt> in
787 the second). Also note that the attribute for the function result
788 (<tt>zeroext</tt>) comes immediately after the argument list.</p>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000789
Anton Korobeynikova0554d92007-01-12 19:20:47 +0000790 <p>Currently, only the following parameter attributes are defined:</p>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000791 <dl>
Reid Spencer314e1cb2007-07-19 23:13:04 +0000792 <dt><tt>zeroext</tt></dt>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000793 <dd>This indicates that the parameter should be zero extended just before
794 a call to this function.</dd>
Reid Spencer314e1cb2007-07-19 23:13:04 +0000795 <dt><tt>signext</tt></dt>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000796 <dd>This indicates that the parameter should be sign extended just before
797 a call to this function.</dd>
Anton Korobeynikove8166852007-01-28 14:30:45 +0000798 <dt><tt>inreg</tt></dt>
799 <dd>This indicates that the parameter should be placed in register (if
Anton Korobeynikove93c6e82007-01-28 15:27:21 +0000800 possible) during assembling function call. Support for this attribute is
801 target-specific</dd>
Anton Korobeynikove8166852007-01-28 14:30:45 +0000802 <dt><tt>sret</tt></dt>
Anton Korobeynikove93c6e82007-01-28 15:27:21 +0000803 <dd>This indicates that the parameter specifies the address of a structure
Reid Spencer05dbb9d2007-03-22 02:02:11 +0000804 that is the return value of the function in the source program.</dd>
Zhou Sheng2444a9a2007-06-05 05:28:26 +0000805 <dt><tt>noalias</tt></dt>
806 <dd>This indicates that the parameter not alias any other object or any
807 other "noalias" objects during the function call.
Reid Spencer9d1700e2007-03-22 02:18:56 +0000808 <dt><tt>noreturn</tt></dt>
809 <dd>This function attribute indicates that the function never returns. This
810 indicates to LLVM that every call to this function should be treated as if
811 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
Reid Spencer05dbb9d2007-03-22 02:02:11 +0000812 <dt><tt>nounwind</tt></dt>
813 <dd>This function attribute indicates that the function type does not use
814 the unwind instruction and does not allow stack unwinding to propagate
815 through it.</dd>
Duncan Sands27e91592007-07-27 19:57:41 +0000816 <dt><tt>nest</tt></dt>
817 <dd>This indicates that the parameter can be excised using the
818 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
Anton Korobeynikova0554d92007-01-12 19:20:47 +0000819 </dl>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000820
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000821</div>
822
823<!-- ======================================================================= -->
Chris Lattner91c15c42006-01-23 23:23:47 +0000824<div class="doc_subsection">
Chris Lattner93564892006-04-08 04:40:53 +0000825 <a name="moduleasm">Module-Level Inline Assembly</a>
Chris Lattner91c15c42006-01-23 23:23:47 +0000826</div>
827
828<div class="doc_text">
829<p>
830Modules may contain "module-level inline asm" blocks, which corresponds to the
831GCC "file scope inline asm" blocks. These blocks are internally concatenated by
832LLVM and treated as a single unit, but may be separated in the .ll file if
833desired. The syntax is very simple:
834</p>
835
Bill Wendling3716c5d2007-05-29 09:04:49 +0000836<div class="doc_code">
837<pre>
838module asm "inline asm code goes here"
839module asm "more can go here"
840</pre>
841</div>
Chris Lattner91c15c42006-01-23 23:23:47 +0000842
843<p>The strings can contain any character by escaping non-printable characters.
844 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
845 for the number.
846</p>
847
848<p>
849 The inline asm code is simply printed to the machine code .s file when
850 assembly code is generated.
851</p>
852</div>
Chris Lattner6af02f32004-12-09 16:11:40 +0000853
Reid Spencer50c723a2007-02-19 23:54:10 +0000854<!-- ======================================================================= -->
855<div class="doc_subsection">
856 <a name="datalayout">Data Layout</a>
857</div>
858
859<div class="doc_text">
860<p>A module may specify a target specific data layout string that specifies how
Reid Spencer7972c472007-04-11 23:49:50 +0000861data is to be laid out in memory. The syntax for the data layout is simply:</p>
862<pre> target datalayout = "<i>layout specification</i>"</pre>
863<p>The <i>layout specification</i> consists of a list of specifications
864separated by the minus sign character ('-'). Each specification starts with a
865letter and may include other information after the letter to define some
866aspect of the data layout. The specifications accepted are as follows: </p>
Reid Spencer50c723a2007-02-19 23:54:10 +0000867<dl>
868 <dt><tt>E</tt></dt>
869 <dd>Specifies that the target lays out data in big-endian form. That is, the
870 bits with the most significance have the lowest address location.</dd>
871 <dt><tt>e</tt></dt>
872 <dd>Specifies that hte target lays out data in little-endian form. That is,
873 the bits with the least significance have the lowest address location.</dd>
874 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
875 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
876 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
877 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
878 too.</dd>
879 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
880 <dd>This specifies the alignment for an integer type of a given bit
881 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
882 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
883 <dd>This specifies the alignment for a vector type of a given bit
884 <i>size</i>.</dd>
885 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
886 <dd>This specifies the alignment for a floating point type of a given bit
887 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
888 (double).</dd>
889 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
890 <dd>This specifies the alignment for an aggregate type of a given bit
891 <i>size</i>.</dd>
892</dl>
893<p>When constructing the data layout for a given target, LLVM starts with a
894default set of specifications which are then (possibly) overriden by the
895specifications in the <tt>datalayout</tt> keyword. The default specifications
896are given in this list:</p>
897<ul>
898 <li><tt>E</tt> - big endian</li>
899 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
900 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
901 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
902 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
903 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
904 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
905 alignment of 64-bits</li>
906 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
907 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
908 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
909 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
910 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
911</ul>
912<p>When llvm is determining the alignment for a given type, it uses the
913following rules:
914<ol>
915 <li>If the type sought is an exact match for one of the specifications, that
916 specification is used.</li>
917 <li>If no match is found, and the type sought is an integer type, then the
918 smallest integer type that is larger than the bitwidth of the sought type is
919 used. If none of the specifications are larger than the bitwidth then the the
920 largest integer type is used. For example, given the default specifications
921 above, the i7 type will use the alignment of i8 (next largest) while both
922 i65 and i256 will use the alignment of i64 (largest specified).</li>
923 <li>If no match is found, and the type sought is a vector type, then the
924 largest vector type that is smaller than the sought vector type will be used
925 as a fall back. This happens because <128 x double> can be implemented in
926 terms of 64 <2 x double>, for example.</li>
927</ol>
928</div>
Chris Lattner6af02f32004-12-09 16:11:40 +0000929
Chris Lattner2f7c9632001-06-06 20:29:01 +0000930<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +0000931<div class="doc_section"> <a name="typesystem">Type System</a> </div>
932<!-- *********************************************************************** -->
Chris Lattner6af02f32004-12-09 16:11:40 +0000933
Misha Brukman76307852003-11-08 01:05:38 +0000934<div class="doc_text">
Chris Lattner6af02f32004-12-09 16:11:40 +0000935
Misha Brukman76307852003-11-08 01:05:38 +0000936<p>The LLVM type system is one of the most important features of the
Chris Lattner48b383b02003-11-25 01:02:51 +0000937intermediate representation. Being typed enables a number of
938optimizations to be performed on the IR directly, without having to do
939extra analyses on the side before the transformation. A strong type
940system makes it easier to read the generated code and enables novel
941analyses and transformations that are not feasible to perform on normal
942three address code representations.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000943
944</div>
945
Chris Lattner2f7c9632001-06-06 20:29:01 +0000946<!-- ======================================================================= -->
Chris Lattner48b383b02003-11-25 01:02:51 +0000947<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +0000948<div class="doc_text">
John Criswell417228d2004-04-09 16:48:45 +0000949<p>The primitive types are the fundamental building blocks of the LLVM
Chris Lattner455fc8c2005-03-07 22:13:59 +0000950system. The current set of primitive types is as follows:</p>
Misha Brukmanc501f552004-03-01 17:47:27 +0000951
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000952<table class="layout">
953 <tr class="layout">
954 <td class="left">
955 <table>
Chris Lattner48b383b02003-11-25 01:02:51 +0000956 <tbody>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000957 <tr><th>Type</th><th>Description</th></tr>
Duncan Sands16f122e2007-03-30 12:22:09 +0000958 <tr><td><tt><a name="t_void">void</a></tt></td><td>No value</td></tr>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000959 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
Chris Lattner48b383b02003-11-25 01:02:51 +0000960 </tbody>
961 </table>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000962 </td>
963 <td class="right">
964 <table>
Chris Lattner48b383b02003-11-25 01:02:51 +0000965 <tbody>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000966 <tr><th>Type</th><th>Description</th></tr>
Reid Spencer138249b2007-05-16 18:44:01 +0000967 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000968 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
Chris Lattner48b383b02003-11-25 01:02:51 +0000969 </tbody>
970 </table>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000971 </td>
972 </tr>
Misha Brukman76307852003-11-08 01:05:38 +0000973</table>
Misha Brukman76307852003-11-08 01:05:38 +0000974</div>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000975
Chris Lattner2f7c9632001-06-06 20:29:01 +0000976<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +0000977<div class="doc_subsubsection"> <a name="t_classifications">Type
978Classifications</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +0000979<div class="doc_text">
Chris Lattner48b383b02003-11-25 01:02:51 +0000980<p>These different primitive types fall into a few useful
981classifications:</p>
Misha Brukmanc501f552004-03-01 17:47:27 +0000982
983<table border="1" cellspacing="0" cellpadding="4">
Chris Lattner48b383b02003-11-25 01:02:51 +0000984 <tbody>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000985 <tr><th>Classification</th><th>Types</th></tr>
Chris Lattner48b383b02003-11-25 01:02:51 +0000986 <tr>
Chris Lattner48b383b02003-11-25 01:02:51 +0000987 <td><a name="t_integer">integer</a></td>
Reid Spencer138249b2007-05-16 18:44:01 +0000988 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
Chris Lattner48b383b02003-11-25 01:02:51 +0000989 </tr>
990 <tr>
991 <td><a name="t_floating">floating point</a></td>
992 <td><tt>float, double</tt></td>
993 </tr>
994 <tr>
995 <td><a name="t_firstclass">first class</a></td>
Reid Spencer138249b2007-05-16 18:44:01 +0000996 <td><tt>i1, ..., float, double, <br/>
Reid Spencer404a3252007-02-15 03:07:05 +0000997 <a href="#t_pointer">pointer</a>,<a href="#t_vector">vector</a></tt>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000998 </td>
Chris Lattner48b383b02003-11-25 01:02:51 +0000999 </tr>
1000 </tbody>
Misha Brukman76307852003-11-08 01:05:38 +00001001</table>
Misha Brukmanc501f552004-03-01 17:47:27 +00001002
Chris Lattner48b383b02003-11-25 01:02:51 +00001003<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1004most important. Values of these types are the only ones which can be
1005produced by instructions, passed as arguments, or used as operands to
1006instructions. This means that all structures and arrays must be
1007manipulated either by pointer or by component.</p>
Misha Brukman76307852003-11-08 01:05:38 +00001008</div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001009
Chris Lattner2f7c9632001-06-06 20:29:01 +00001010<!-- ======================================================================= -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001011<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001012
Misha Brukman76307852003-11-08 01:05:38 +00001013<div class="doc_text">
Chris Lattner74d3f822004-12-09 17:30:23 +00001014
Chris Lattner48b383b02003-11-25 01:02:51 +00001015<p>The real power in LLVM comes from the derived types in the system.
1016This is what allows a programmer to represent arrays, functions,
1017pointers, and other useful types. Note that these derived types may be
1018recursive: For example, it is possible to have a two dimensional array.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001019
Misha Brukman76307852003-11-08 01:05:38 +00001020</div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001021
Chris Lattner2f7c9632001-06-06 20:29:01 +00001022<!-- _______________________________________________________________________ -->
Reid Spencer138249b2007-05-16 18:44:01 +00001023<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1024
1025<div class="doc_text">
1026
1027<h5>Overview:</h5>
1028<p>The integer type is a very simple derived type that simply specifies an
1029arbitrary bit width for the integer type desired. Any bit width from 1 bit to
10302^23-1 (about 8 million) can be specified.</p>
1031
1032<h5>Syntax:</h5>
1033
1034<pre>
1035 iN
1036</pre>
1037
1038<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1039value.</p>
1040
1041<h5>Examples:</h5>
1042<table class="layout">
1043 <tr class="layout">
1044 <td class="left">
1045 <tt>i1</tt><br/>
1046 <tt>i4</tt><br/>
1047 <tt>i8</tt><br/>
1048 <tt>i16</tt><br/>
1049 <tt>i32</tt><br/>
1050 <tt>i42</tt><br/>
1051 <tt>i64</tt><br/>
1052 <tt>i1942652</tt><br/>
1053 </td>
1054 <td class="left">
1055 A boolean integer of 1 bit<br/>
1056 A nibble sized integer of 4 bits.<br/>
1057 A byte sized integer of 8 bits.<br/>
1058 A half word sized integer of 16 bits.<br/>
1059 A word sized integer of 32 bits.<br/>
1060 An integer whose bit width is the answer. <br/>
1061 A double word sized integer of 64 bits.<br/>
1062 A really big integer of over 1 million bits.<br/>
1063 </td>
1064 </tr>
1065</table>
Bill Wendling3716c5d2007-05-29 09:04:49 +00001066</div>
Reid Spencer138249b2007-05-16 18:44:01 +00001067
1068<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001069<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001070
Misha Brukman76307852003-11-08 01:05:38 +00001071<div class="doc_text">
Chris Lattner74d3f822004-12-09 17:30:23 +00001072
Chris Lattner2f7c9632001-06-06 20:29:01 +00001073<h5>Overview:</h5>
Chris Lattner74d3f822004-12-09 17:30:23 +00001074
Misha Brukman76307852003-11-08 01:05:38 +00001075<p>The array type is a very simple derived type that arranges elements
Chris Lattner48b383b02003-11-25 01:02:51 +00001076sequentially in memory. The array type requires a size (number of
1077elements) and an underlying data type.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001078
Chris Lattner590645f2002-04-14 06:13:44 +00001079<h5>Syntax:</h5>
Chris Lattner74d3f822004-12-09 17:30:23 +00001080
1081<pre>
1082 [&lt;# elements&gt; x &lt;elementtype&gt;]
1083</pre>
1084
John Criswell02fdc6f2005-05-12 16:52:32 +00001085<p>The number of elements is a constant integer value; elementtype may
Chris Lattner48b383b02003-11-25 01:02:51 +00001086be any type with a size.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001087
Chris Lattner590645f2002-04-14 06:13:44 +00001088<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001089<table class="layout">
1090 <tr class="layout">
1091 <td class="left">
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001092 <tt>[40 x i32 ]</tt><br/>
1093 <tt>[41 x i32 ]</tt><br/>
Reid Spencer3e628eb92007-01-04 16:43:23 +00001094 <tt>[40 x i8]</tt><br/>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001095 </td>
1096 <td class="left">
Reid Spencer3e628eb92007-01-04 16:43:23 +00001097 Array of 40 32-bit integer values.<br/>
1098 Array of 41 32-bit integer values.<br/>
1099 Array of 40 8-bit integer values.<br/>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001100 </td>
1101 </tr>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001102</table>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001103<p>Here are some examples of multidimensional arrays:</p>
1104<table class="layout">
1105 <tr class="layout">
1106 <td class="left">
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001107 <tt>[3 x [4 x i32]]</tt><br/>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001108 <tt>[12 x [10 x float]]</tt><br/>
Reid Spencer3e628eb92007-01-04 16:43:23 +00001109 <tt>[2 x [3 x [4 x i16]]]</tt><br/>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001110 </td>
1111 <td class="left">
Reid Spencer3e628eb92007-01-04 16:43:23 +00001112 3x4 array of 32-bit integer values.<br/>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001113 12x10 array of single precision floating point values.<br/>
Reid Spencer3e628eb92007-01-04 16:43:23 +00001114 2x3x4 array of 16-bit integer values.<br/>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001115 </td>
1116 </tr>
1117</table>
Chris Lattnerc0ad71e2005-06-24 17:22:57 +00001118
John Criswell4c0cf7f2005-10-24 16:17:18 +00001119<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1120length array. Normally, accesses past the end of an array are undefined in
Chris Lattnerc0ad71e2005-06-24 17:22:57 +00001121LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1122As a special case, however, zero length arrays are recognized to be variable
1123length. This allows implementation of 'pascal style arrays' with the LLVM
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001124type "{ i32, [0 x float]}", for example.</p>
Chris Lattnerc0ad71e2005-06-24 17:22:57 +00001125
Misha Brukman76307852003-11-08 01:05:38 +00001126</div>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001127
Chris Lattner2f7c9632001-06-06 20:29:01 +00001128<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001129<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00001130<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00001131<h5>Overview:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001132<p>The function type can be thought of as a function signature. It
1133consists of a return type and a list of formal parameter types.
John Criswella0d50d22003-11-25 21:45:46 +00001134Function types are usually used to build virtual function tables
Chris Lattner48b383b02003-11-25 01:02:51 +00001135(which are structures of pointers to functions), for indirect function
1136calls, and when defining a function.</p>
John Criswella0d50d22003-11-25 21:45:46 +00001137<p>
1138The return type of a function type cannot be an aggregate type.
1139</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001140<h5>Syntax:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001141<pre> &lt;returntype&gt; (&lt;parameter list&gt;)<br></pre>
John Criswell4c0cf7f2005-10-24 16:17:18 +00001142<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
Misha Brukman20f9a622004-08-12 20:16:08 +00001143specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
Chris Lattner5ed60612003-09-03 00:41:47 +00001144which indicates that the function takes a variable number of arguments.
1145Variable argument functions can access their arguments with the <a
Chris Lattner48b383b02003-11-25 01:02:51 +00001146 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001147<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001148<table class="layout">
1149 <tr class="layout">
Reid Spencer58c08712006-12-31 07:18:34 +00001150 <td class="left"><tt>i32 (i32)</tt></td>
1151 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001152 </td>
Reid Spencer58c08712006-12-31 07:18:34 +00001153 </tr><tr class="layout">
Reid Spencer314e1cb2007-07-19 23:13:04 +00001154 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Reid Spencer655dcc62006-12-31 07:20:23 +00001155 </tt></td>
Reid Spencer58c08712006-12-31 07:18:34 +00001156 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1157 an <tt>i16</tt> that should be sign extended and a
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001158 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
Reid Spencer58c08712006-12-31 07:18:34 +00001159 <tt>float</tt>.
1160 </td>
1161 </tr><tr class="layout">
1162 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1163 <td class="left">A vararg function that takes at least one
Reid Spencer3e628eb92007-01-04 16:43:23 +00001164 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
Reid Spencer58c08712006-12-31 07:18:34 +00001165 which returns an integer. This is the signature for <tt>printf</tt> in
1166 LLVM.
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001167 </td>
1168 </tr>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001169</table>
Misha Brukmanc501f552004-03-01 17:47:27 +00001170
Misha Brukman76307852003-11-08 01:05:38 +00001171</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001172<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001173<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00001174<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00001175<h5>Overview:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001176<p>The structure type is used to represent a collection of data members
1177together in memory. The packing of the field types is defined to match
1178the ABI of the underlying processor. The elements of a structure may
1179be any type that has a size.</p>
1180<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1181and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1182field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1183instruction.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001184<h5>Syntax:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001185<pre> { &lt;type list&gt; }<br></pre>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001186<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001187<table class="layout">
1188 <tr class="layout">
Jeff Cohen5819f182007-04-22 01:17:39 +00001189 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1190 <td class="left">A triple of three <tt>i32</tt> values</td>
1191 </tr><tr class="layout">
1192 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1193 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1194 second element is a <a href="#t_pointer">pointer</a> to a
1195 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1196 an <tt>i32</tt>.</td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001197 </tr>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001198</table>
Misha Brukman76307852003-11-08 01:05:38 +00001199</div>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001200
Chris Lattner2f7c9632001-06-06 20:29:01 +00001201<!-- _______________________________________________________________________ -->
Andrew Lenharth8df88e22006-12-08 17:13:00 +00001202<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1203</div>
1204<div class="doc_text">
1205<h5>Overview:</h5>
1206<p>The packed structure type is used to represent a collection of data members
1207together in memory. There is no padding between fields. Further, the alignment
1208of a packed structure is 1 byte. The elements of a packed structure may
1209be any type that has a size.</p>
1210<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1211and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1212field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1213instruction.</p>
1214<h5>Syntax:</h5>
1215<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1216<h5>Examples:</h5>
1217<table class="layout">
1218 <tr class="layout">
Jeff Cohen5819f182007-04-22 01:17:39 +00001219 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1220 <td class="left">A triple of three <tt>i32</tt> values</td>
1221 </tr><tr class="layout">
1222 <td class="left"><tt>&lt;&nbsp;{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}&nbsp;&gt;</tt></td>
1223 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1224 second element is a <a href="#t_pointer">pointer</a> to a
1225 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1226 an <tt>i32</tt>.</td>
Andrew Lenharth8df88e22006-12-08 17:13:00 +00001227 </tr>
1228</table>
1229</div>
1230
1231<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001232<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00001233<div class="doc_text">
Chris Lattner590645f2002-04-14 06:13:44 +00001234<h5>Overview:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001235<p>As in many languages, the pointer type represents a pointer or
1236reference to another object, which must live in memory.</p>
Chris Lattner590645f2002-04-14 06:13:44 +00001237<h5>Syntax:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001238<pre> &lt;type&gt; *<br></pre>
Chris Lattner590645f2002-04-14 06:13:44 +00001239<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001240<table class="layout">
1241 <tr class="layout">
1242 <td class="left">
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001243 <tt>[4x i32]*</tt><br/>
1244 <tt>i32 (i32 *) *</tt><br/>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001245 </td>
1246 <td class="left">
1247 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001248 four <tt>i32</tt> values<br/>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001249 A <a href="#t_pointer">pointer</a> to a <a
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001250 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1251 <tt>i32</tt>.<br/>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001252 </td>
1253 </tr>
Misha Brukman76307852003-11-08 01:05:38 +00001254</table>
Misha Brukman76307852003-11-08 01:05:38 +00001255</div>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001256
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001257<!-- _______________________________________________________________________ -->
Reid Spencer404a3252007-02-15 03:07:05 +00001258<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00001259<div class="doc_text">
Chris Lattner37b6b092005-04-25 17:34:15 +00001260
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001261<h5>Overview:</h5>
Chris Lattner37b6b092005-04-25 17:34:15 +00001262
Reid Spencer404a3252007-02-15 03:07:05 +00001263<p>A vector type is a simple derived type that represents a vector
1264of elements. Vector types are used when multiple primitive data
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001265are operated in parallel using a single instruction (SIMD).
Reid Spencer404a3252007-02-15 03:07:05 +00001266A vector type requires a size (number of
Chris Lattner330ce692005-11-10 01:44:22 +00001267elements) and an underlying primitive data type. Vectors must have a power
Reid Spencer404a3252007-02-15 03:07:05 +00001268of two length (1, 2, 4, 8, 16 ...). Vector types are
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001269considered <a href="#t_firstclass">first class</a>.</p>
Chris Lattner37b6b092005-04-25 17:34:15 +00001270
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001271<h5>Syntax:</h5>
Chris Lattner37b6b092005-04-25 17:34:15 +00001272
1273<pre>
1274 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1275</pre>
1276
John Criswell4a3327e2005-05-13 22:25:59 +00001277<p>The number of elements is a constant integer value; elementtype may
Chris Lattnerc0f423a2007-01-15 01:54:13 +00001278be any integer or floating point type.</p>
Chris Lattner37b6b092005-04-25 17:34:15 +00001279
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001280<h5>Examples:</h5>
Chris Lattner37b6b092005-04-25 17:34:15 +00001281
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001282<table class="layout">
1283 <tr class="layout">
1284 <td class="left">
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001285 <tt>&lt;4 x i32&gt;</tt><br/>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001286 <tt>&lt;8 x float&gt;</tt><br/>
Reid Spencer3e628eb92007-01-04 16:43:23 +00001287 <tt>&lt;2 x i64&gt;</tt><br/>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001288 </td>
1289 <td class="left">
Reid Spencer404a3252007-02-15 03:07:05 +00001290 Vector of 4 32-bit integer values.<br/>
1291 Vector of 8 floating-point values.<br/>
1292 Vector of 2 64-bit integer values.<br/>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001293 </td>
1294 </tr>
1295</table>
Misha Brukman76307852003-11-08 01:05:38 +00001296</div>
1297
Chris Lattner37b6b092005-04-25 17:34:15 +00001298<!-- _______________________________________________________________________ -->
1299<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1300<div class="doc_text">
1301
1302<h5>Overview:</h5>
1303
1304<p>Opaque types are used to represent unknown types in the system. This
1305corresponds (for example) to the C notion of a foward declared structure type.
1306In LLVM, opaque types can eventually be resolved to any type (not just a
1307structure type).</p>
1308
1309<h5>Syntax:</h5>
1310
1311<pre>
1312 opaque
1313</pre>
1314
1315<h5>Examples:</h5>
1316
1317<table class="layout">
1318 <tr class="layout">
1319 <td class="left">
1320 <tt>opaque</tt>
1321 </td>
1322 <td class="left">
1323 An opaque type.<br/>
1324 </td>
1325 </tr>
1326</table>
1327</div>
1328
1329
Chris Lattner74d3f822004-12-09 17:30:23 +00001330<!-- *********************************************************************** -->
1331<div class="doc_section"> <a name="constants">Constants</a> </div>
1332<!-- *********************************************************************** -->
1333
1334<div class="doc_text">
1335
1336<p>LLVM has several different basic types of constants. This section describes
1337them all and their syntax.</p>
1338
1339</div>
1340
1341<!-- ======================================================================= -->
Reid Spencer8f08d802004-12-09 18:02:53 +00001342<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001343
1344<div class="doc_text">
1345
1346<dl>
1347 <dt><b>Boolean constants</b></dt>
1348
1349 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
Reid Spencer36a15422007-01-12 03:35:51 +00001350 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
Chris Lattner74d3f822004-12-09 17:30:23 +00001351 </dd>
1352
1353 <dt><b>Integer constants</b></dt>
1354
Reid Spencer8f08d802004-12-09 18:02:53 +00001355 <dd>Standard integers (such as '4') are constants of the <a
Reid Spencer3e628eb92007-01-04 16:43:23 +00001356 href="#t_integer">integer</a> type. Negative numbers may be used with
Chris Lattner74d3f822004-12-09 17:30:23 +00001357 integer types.
1358 </dd>
1359
1360 <dt><b>Floating point constants</b></dt>
1361
1362 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1363 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
Chris Lattner74d3f822004-12-09 17:30:23 +00001364 notation (see below). Floating point constants must have a <a
1365 href="#t_floating">floating point</a> type. </dd>
1366
1367 <dt><b>Null pointer constants</b></dt>
1368
John Criswelldfe6a862004-12-10 15:51:16 +00001369 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
Chris Lattner74d3f822004-12-09 17:30:23 +00001370 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1371
1372</dl>
1373
John Criswelldfe6a862004-12-10 15:51:16 +00001374<p>The one non-intuitive notation for constants is the optional hexadecimal form
Chris Lattner74d3f822004-12-09 17:30:23 +00001375of floating point constants. For example, the form '<tt>double
13760x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
13774.5e+15</tt>'. The only time hexadecimal floating point constants are required
Reid Spencer8f08d802004-12-09 18:02:53 +00001378(and the only time that they are generated by the disassembler) is when a
1379floating point constant must be emitted but it cannot be represented as a
1380decimal floating point number. For example, NaN's, infinities, and other
1381special values are represented in their IEEE hexadecimal format so that
1382assembly and disassembly do not cause any bits to change in the constants.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001383
1384</div>
1385
1386<!-- ======================================================================= -->
1387<div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1388</div>
1389
1390<div class="doc_text">
Chris Lattner455fc8c2005-03-07 22:13:59 +00001391<p>Aggregate constants arise from aggregation of simple constants
1392and smaller aggregate constants.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001393
1394<dl>
1395 <dt><b>Structure constants</b></dt>
1396
1397 <dd>Structure constants are represented with notation similar to structure
1398 type definitions (a comma separated list of elements, surrounded by braces
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001399 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* %G }</tt>",
Chris Lattner0a2d0992007-07-13 20:01:46 +00001400 where "<tt>%G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
Chris Lattner455fc8c2005-03-07 22:13:59 +00001401 must have <a href="#t_struct">structure type</a>, and the number and
Chris Lattner74d3f822004-12-09 17:30:23 +00001402 types of elements must match those specified by the type.
1403 </dd>
1404
1405 <dt><b>Array constants</b></dt>
1406
1407 <dd>Array constants are represented with notation similar to array type
1408 definitions (a comma separated list of elements, surrounded by square brackets
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001409 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
Chris Lattner74d3f822004-12-09 17:30:23 +00001410 constants must have <a href="#t_array">array type</a>, and the number and
1411 types of elements must match those specified by the type.
1412 </dd>
1413
Reid Spencer404a3252007-02-15 03:07:05 +00001414 <dt><b>Vector constants</b></dt>
Chris Lattner74d3f822004-12-09 17:30:23 +00001415
Reid Spencer404a3252007-02-15 03:07:05 +00001416 <dd>Vector constants are represented with notation similar to vector type
Chris Lattner74d3f822004-12-09 17:30:23 +00001417 definitions (a comma separated list of elements, surrounded by
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001418 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
Jeff Cohen5819f182007-04-22 01:17:39 +00001419 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
Reid Spencer404a3252007-02-15 03:07:05 +00001420 href="#t_vector">vector type</a>, and the number and types of elements must
Chris Lattner74d3f822004-12-09 17:30:23 +00001421 match those specified by the type.
1422 </dd>
1423
1424 <dt><b>Zero initialization</b></dt>
1425
1426 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1427 value to zero of <em>any</em> type, including scalar and aggregate types.
1428 This is often used to avoid having to print large zero initializers (e.g. for
John Criswell4c0cf7f2005-10-24 16:17:18 +00001429 large arrays) and is always exactly equivalent to using explicit zero
Chris Lattner74d3f822004-12-09 17:30:23 +00001430 initializers.
1431 </dd>
1432</dl>
1433
1434</div>
1435
1436<!-- ======================================================================= -->
1437<div class="doc_subsection">
1438 <a name="globalconstants">Global Variable and Function Addresses</a>
1439</div>
1440
1441<div class="doc_text">
1442
1443<p>The addresses of <a href="#globalvars">global variables</a> and <a
1444href="#functionstructure">functions</a> are always implicitly valid (link-time)
John Criswelldfe6a862004-12-10 15:51:16 +00001445constants. These constants are explicitly referenced when the <a
1446href="#identifiers">identifier for the global</a> is used and always have <a
Chris Lattner74d3f822004-12-09 17:30:23 +00001447href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1448file:</p>
1449
Bill Wendling3716c5d2007-05-29 09:04:49 +00001450<div class="doc_code">
Chris Lattner74d3f822004-12-09 17:30:23 +00001451<pre>
Chris Lattner00538a12007-06-06 18:28:13 +00001452@X = global i32 17
1453@Y = global i32 42
1454@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
Chris Lattner74d3f822004-12-09 17:30:23 +00001455</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00001456</div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001457
1458</div>
1459
1460<!-- ======================================================================= -->
Reid Spencer641f5c92004-12-09 18:13:12 +00001461<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001462<div class="doc_text">
Reid Spencer641f5c92004-12-09 18:13:12 +00001463 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
John Criswell4a3327e2005-05-13 22:25:59 +00001464 no specific value. Undefined values may be of any type and be used anywhere
Reid Spencer641f5c92004-12-09 18:13:12 +00001465 a constant is permitted.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001466
Reid Spencer641f5c92004-12-09 18:13:12 +00001467 <p>Undefined values indicate to the compiler that the program is well defined
1468 no matter what value is used, giving the compiler more freedom to optimize.
1469 </p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001470</div>
1471
1472<!-- ======================================================================= -->
1473<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1474</div>
1475
1476<div class="doc_text">
1477
1478<p>Constant expressions are used to allow expressions involving other constants
1479to be used as constants. Constant expressions may be of any <a
John Criswell4a3327e2005-05-13 22:25:59 +00001480href="#t_firstclass">first class</a> type and may involve any LLVM operation
Chris Lattner74d3f822004-12-09 17:30:23 +00001481that does not have side effects (e.g. load and call are not supported). The
1482following is the syntax for constant expressions:</p>
1483
1484<dl>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001485 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1486 <dd>Truncate a constant to another type. The bit size of CST must be larger
Chris Lattnerc0f423a2007-01-15 01:54:13 +00001487 than the bit size of TYPE. Both types must be integers.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00001488
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001489 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1490 <dd>Zero extend a constant to another type. The bit size of CST must be
Chris Lattnerc0f423a2007-01-15 01:54:13 +00001491 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001492
1493 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1494 <dd>Sign extend a constant to another type. The bit size of CST must be
Chris Lattnerc0f423a2007-01-15 01:54:13 +00001495 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001496
1497 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1498 <dd>Truncate a floating point constant to another floating point type. The
1499 size of CST must be larger than the size of TYPE. Both types must be
1500 floating point.</dd>
1501
1502 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1503 <dd>Floating point extend a constant to another type. The size of CST must be
1504 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1505
Reid Spencer753163d2007-07-31 14:40:14 +00001506 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001507 <dd>Convert a floating point constant to the corresponding unsigned integer
1508 constant. TYPE must be an integer type. CST must be floating point. If the
1509 value won't fit in the integer type, the results are undefined.</dd>
1510
Reid Spencer51b07252006-11-09 23:03:26 +00001511 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001512 <dd>Convert a floating point constant to the corresponding signed integer
1513 constant. TYPE must be an integer type. CST must be floating point. If the
1514 value won't fit in the integer type, the results are undefined.</dd>
1515
Reid Spencer51b07252006-11-09 23:03:26 +00001516 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001517 <dd>Convert an unsigned integer constant to the corresponding floating point
1518 constant. TYPE must be floating point. CST must be of integer type. If the
Jeff Cohenbeccb742007-04-22 14:56:37 +00001519 value won't fit in the floating point type, the results are undefined.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001520
Reid Spencer51b07252006-11-09 23:03:26 +00001521 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001522 <dd>Convert a signed integer constant to the corresponding floating point
1523 constant. TYPE must be floating point. CST must be of integer type. If the
Jeff Cohenbeccb742007-04-22 14:56:37 +00001524 value won't fit in the floating point type, the results are undefined.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001525
Reid Spencer5b950642006-11-11 23:08:07 +00001526 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1527 <dd>Convert a pointer typed constant to the corresponding integer constant
1528 TYPE must be an integer type. CST must be of pointer type. The CST value is
1529 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1530
1531 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1532 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1533 pointer type. CST must be of integer type. The CST value is zero extended,
1534 truncated, or unchanged to make it fit in a pointer size. This one is
1535 <i>really</i> dangerous!</dd>
1536
1537 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001538 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1539 identical (same number of bits). The conversion is done as if the CST value
1540 was stored to memory and read back as TYPE. In other words, no bits change
Reid Spencer5b950642006-11-11 23:08:07 +00001541 with this operator, just the type. This can be used for conversion of
Reid Spencer404a3252007-02-15 03:07:05 +00001542 vector types to any other type, as long as they have the same bit width. For
Reid Spencer5b950642006-11-11 23:08:07 +00001543 pointers it is only valid to cast to another pointer type.
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001544 </dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00001545
1546 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1547
1548 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1549 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1550 instruction, the index list may have zero or more indexes, which are required
1551 to make sense for the type of "CSTPTR".</dd>
1552
Robert Bocchino7e97a6d2006-01-10 19:31:34 +00001553 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1554
1555 <dd>Perform the <a href="#i_select">select operation</a> on
Reid Spencer9965ee72006-12-04 19:23:19 +00001556 constants.</dd>
1557
1558 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1559 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1560
1561 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1562 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
Robert Bocchino7e97a6d2006-01-10 19:31:34 +00001563
1564 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1565
1566 <dd>Perform the <a href="#i_extractelement">extractelement
1567 operation</a> on constants.
1568
Robert Bocchinof72fdfe2006-01-15 20:48:27 +00001569 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1570
1571 <dd>Perform the <a href="#i_insertelement">insertelement
Reid Spencer9965ee72006-12-04 19:23:19 +00001572 operation</a> on constants.</dd>
Robert Bocchinof72fdfe2006-01-15 20:48:27 +00001573
Chris Lattner016a0e52006-04-08 00:13:41 +00001574
1575 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1576
1577 <dd>Perform the <a href="#i_shufflevector">shufflevector
Reid Spencer9965ee72006-12-04 19:23:19 +00001578 operation</a> on constants.</dd>
Chris Lattner016a0e52006-04-08 00:13:41 +00001579
Chris Lattner74d3f822004-12-09 17:30:23 +00001580 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1581
Reid Spencer641f5c92004-12-09 18:13:12 +00001582 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1583 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
Chris Lattner74d3f822004-12-09 17:30:23 +00001584 binary</a> operations. The constraints on operands are the same as those for
1585 the corresponding instruction (e.g. no bitwise operations on floating point
John Criswell02fdc6f2005-05-12 16:52:32 +00001586 values are allowed).</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00001587</dl>
Chris Lattner74d3f822004-12-09 17:30:23 +00001588</div>
Chris Lattnerb1652612004-03-08 16:49:10 +00001589
Chris Lattner2f7c9632001-06-06 20:29:01 +00001590<!-- *********************************************************************** -->
Chris Lattner98f013c2006-01-25 23:47:57 +00001591<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1592<!-- *********************************************************************** -->
1593
1594<!-- ======================================================================= -->
1595<div class="doc_subsection">
1596<a name="inlineasm">Inline Assembler Expressions</a>
1597</div>
1598
1599<div class="doc_text">
1600
1601<p>
1602LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1603Module-Level Inline Assembly</a>) through the use of a special value. This
1604value represents the inline assembler as a string (containing the instructions
1605to emit), a list of operand constraints (stored as a string), and a flag that
1606indicates whether or not the inline asm expression has side effects. An example
1607inline assembler expression is:
1608</p>
1609
Bill Wendling3716c5d2007-05-29 09:04:49 +00001610<div class="doc_code">
Chris Lattner98f013c2006-01-25 23:47:57 +00001611<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00001612i32 (i32) asm "bswap $0", "=r,r"
Chris Lattner98f013c2006-01-25 23:47:57 +00001613</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00001614</div>
Chris Lattner98f013c2006-01-25 23:47:57 +00001615
1616<p>
1617Inline assembler expressions may <b>only</b> be used as the callee operand of
1618a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1619</p>
1620
Bill Wendling3716c5d2007-05-29 09:04:49 +00001621<div class="doc_code">
Chris Lattner98f013c2006-01-25 23:47:57 +00001622<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00001623%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
Chris Lattner98f013c2006-01-25 23:47:57 +00001624</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00001625</div>
Chris Lattner98f013c2006-01-25 23:47:57 +00001626
1627<p>
1628Inline asms with side effects not visible in the constraint list must be marked
1629as having side effects. This is done through the use of the
1630'<tt>sideeffect</tt>' keyword, like so:
1631</p>
1632
Bill Wendling3716c5d2007-05-29 09:04:49 +00001633<div class="doc_code">
Chris Lattner98f013c2006-01-25 23:47:57 +00001634<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00001635call void asm sideeffect "eieio", ""()
Chris Lattner98f013c2006-01-25 23:47:57 +00001636</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00001637</div>
Chris Lattner98f013c2006-01-25 23:47:57 +00001638
1639<p>TODO: The format of the asm and constraints string still need to be
1640documented here. Constraints on what can be done (e.g. duplication, moving, etc
1641need to be documented).
1642</p>
1643
1644</div>
1645
1646<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001647<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1648<!-- *********************************************************************** -->
Chris Lattner74d3f822004-12-09 17:30:23 +00001649
Misha Brukman76307852003-11-08 01:05:38 +00001650<div class="doc_text">
Chris Lattner74d3f822004-12-09 17:30:23 +00001651
Chris Lattner48b383b02003-11-25 01:02:51 +00001652<p>The LLVM instruction set consists of several different
1653classifications of instructions: <a href="#terminators">terminator
John Criswell4a3327e2005-05-13 22:25:59 +00001654instructions</a>, <a href="#binaryops">binary instructions</a>,
1655<a href="#bitwiseops">bitwise binary instructions</a>, <a
Chris Lattner48b383b02003-11-25 01:02:51 +00001656 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1657instructions</a>.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001658
Misha Brukman76307852003-11-08 01:05:38 +00001659</div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001660
Chris Lattner2f7c9632001-06-06 20:29:01 +00001661<!-- ======================================================================= -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001662<div class="doc_subsection"> <a name="terminators">Terminator
1663Instructions</a> </div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001664
Misha Brukman76307852003-11-08 01:05:38 +00001665<div class="doc_text">
Chris Lattner74d3f822004-12-09 17:30:23 +00001666
Chris Lattner48b383b02003-11-25 01:02:51 +00001667<p>As mentioned <a href="#functionstructure">previously</a>, every
1668basic block in a program ends with a "Terminator" instruction, which
1669indicates which block should be executed after the current block is
1670finished. These terminator instructions typically yield a '<tt>void</tt>'
1671value: they produce control flow, not values (the one exception being
1672the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
John Criswelldfe6a862004-12-10 15:51:16 +00001673<p>There are six different terminator instructions: the '<a
Chris Lattner48b383b02003-11-25 01:02:51 +00001674 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1675instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
Chris Lattner08b7d5b2004-10-16 18:04:13 +00001676the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1677 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1678 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001679
Misha Brukman76307852003-11-08 01:05:38 +00001680</div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001681
Chris Lattner2f7c9632001-06-06 20:29:01 +00001682<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001683<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1684Instruction</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00001685<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00001686<h5>Syntax:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001687<pre> ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
Chris Lattner590645f2002-04-14 06:13:44 +00001688 ret void <i>; Return from void function</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001689</pre>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001690<h5>Overview:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001691<p>The '<tt>ret</tt>' instruction is used to return control flow (and a
John Criswell4a3327e2005-05-13 22:25:59 +00001692value) from a function back to the caller.</p>
John Criswell417228d2004-04-09 16:48:45 +00001693<p>There are two forms of the '<tt>ret</tt>' instruction: one that
Chris Lattner48b383b02003-11-25 01:02:51 +00001694returns a value and then causes control flow, and one that just causes
1695control flow to occur.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001696<h5>Arguments:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001697<p>The '<tt>ret</tt>' instruction may return any '<a
1698 href="#t_firstclass">first class</a>' type. Notice that a function is
1699not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1700instruction inside of the function that returns a value that does not
1701match the return type of the function.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001702<h5>Semantics:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001703<p>When the '<tt>ret</tt>' instruction is executed, control flow
1704returns back to the calling function's context. If the caller is a "<a
John Criswell40db33f2004-06-25 15:16:57 +00001705 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
Chris Lattner48b383b02003-11-25 01:02:51 +00001706the instruction after the call. If the caller was an "<a
1707 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
John Criswell02fdc6f2005-05-12 16:52:32 +00001708at the beginning of the "normal" destination block. If the instruction
Chris Lattner48b383b02003-11-25 01:02:51 +00001709returns a value, that value shall set the call or invoke instruction's
1710return value.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001711<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001712<pre> ret i32 5 <i>; Return an integer value of 5</i>
Chris Lattner590645f2002-04-14 06:13:44 +00001713 ret void <i>; Return from a void function</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001714</pre>
Misha Brukman76307852003-11-08 01:05:38 +00001715</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001716<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001717<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00001718<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00001719<h5>Syntax:</h5>
Reid Spencer36a15422007-01-12 03:35:51 +00001720<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001721</pre>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001722<h5>Overview:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001723<p>The '<tt>br</tt>' instruction is used to cause control flow to
1724transfer to a different basic block in the current function. There are
1725two forms of this instruction, corresponding to a conditional branch
1726and an unconditional branch.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001727<h5>Arguments:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001728<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
Reid Spencer36a15422007-01-12 03:35:51 +00001729single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
Reid Spencer50c723a2007-02-19 23:54:10 +00001730unconditional form of the '<tt>br</tt>' instruction takes a single
1731'<tt>label</tt>' value as a target.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001732<h5>Semantics:</h5>
Reid Spencer36a15422007-01-12 03:35:51 +00001733<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
Chris Lattner48b383b02003-11-25 01:02:51 +00001734argument is evaluated. If the value is <tt>true</tt>, control flows
1735to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1736control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001737<h5>Example:</h5>
Reid Spencer36a15422007-01-12 03:35:51 +00001738<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
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001739 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
Misha Brukman76307852003-11-08 01:05:38 +00001740</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001741<!-- _______________________________________________________________________ -->
Chris Lattnercf96c6c2004-02-24 04:54:45 +00001742<div class="doc_subsubsection">
1743 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1744</div>
1745
Misha Brukman76307852003-11-08 01:05:38 +00001746<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00001747<h5>Syntax:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00001748
1749<pre>
1750 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
1751</pre>
1752
Chris Lattner2f7c9632001-06-06 20:29:01 +00001753<h5>Overview:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00001754
1755<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1756several different places. It is a generalization of the '<tt>br</tt>'
Misha Brukman76307852003-11-08 01:05:38 +00001757instruction, allowing a branch to occur to one of many possible
1758destinations.</p>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00001759
1760
Chris Lattner2f7c9632001-06-06 20:29:01 +00001761<h5>Arguments:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00001762
1763<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1764comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1765an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1766table is not allowed to contain duplicate constant entries.</p>
1767
Chris Lattner2f7c9632001-06-06 20:29:01 +00001768<h5>Semantics:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00001769
Chris Lattner48b383b02003-11-25 01:02:51 +00001770<p>The <tt>switch</tt> instruction specifies a table of values and
1771destinations. When the '<tt>switch</tt>' instruction is executed, this
John Criswellbcbb18c2004-06-25 16:05:06 +00001772table is searched for the given value. If the value is found, control flow is
1773transfered to the corresponding destination; otherwise, control flow is
1774transfered to the default destination.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001775
Chris Lattnercf96c6c2004-02-24 04:54:45 +00001776<h5>Implementation:</h5>
1777
1778<p>Depending on properties of the target machine and the particular
1779<tt>switch</tt> instruction, this instruction may be code generated in different
John Criswellbcbb18c2004-06-25 16:05:06 +00001780ways. For example, it could be generated as a series of chained conditional
1781branches or with a lookup table.</p>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00001782
1783<h5>Example:</h5>
1784
1785<pre>
1786 <i>; Emulate a conditional br instruction</i>
Reid Spencer36a15422007-01-12 03:35:51 +00001787 %Val = <a href="#i_zext">zext</a> i1 %value to i32
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001788 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
Chris Lattnercf96c6c2004-02-24 04:54:45 +00001789
1790 <i>; Emulate an unconditional br instruction</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001791 switch i32 0, label %dest [ ]
Chris Lattnercf96c6c2004-02-24 04:54:45 +00001792
1793 <i>; Implement a jump table:</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001794 switch i32 %val, label %otherwise [ i32 0, label %onzero
1795 i32 1, label %onone
1796 i32 2, label %ontwo ]
Chris Lattner2f7c9632001-06-06 20:29:01 +00001797</pre>
Misha Brukman76307852003-11-08 01:05:38 +00001798</div>
Chris Lattner0132aff2005-05-06 22:57:40 +00001799
Chris Lattner2f7c9632001-06-06 20:29:01 +00001800<!-- _______________________________________________________________________ -->
Chris Lattner0132aff2005-05-06 22:57:40 +00001801<div class="doc_subsubsection">
1802 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1803</div>
1804
Misha Brukman76307852003-11-08 01:05:38 +00001805<div class="doc_text">
Chris Lattner0132aff2005-05-06 22:57:40 +00001806
Chris Lattner2f7c9632001-06-06 20:29:01 +00001807<h5>Syntax:</h5>
Chris Lattner0132aff2005-05-06 22:57:40 +00001808
1809<pre>
1810 &lt;result&gt; = invoke [<a href="#callingconv">cconv</a>] &lt;ptr to function ty&gt; %&lt;function ptr val&gt;(&lt;function args&gt;)
Chris Lattner6b7a0082006-05-14 18:23:06 +00001811 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
Chris Lattner0132aff2005-05-06 22:57:40 +00001812</pre>
1813
Chris Lattnera8292f32002-05-06 22:08:29 +00001814<h5>Overview:</h5>
Chris Lattner0132aff2005-05-06 22:57:40 +00001815
1816<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1817function, with the possibility of control flow transfer to either the
John Criswell02fdc6f2005-05-12 16:52:32 +00001818'<tt>normal</tt>' label or the
1819'<tt>exception</tt>' label. If the callee function returns with the
Chris Lattner0132aff2005-05-06 22:57:40 +00001820"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1821"normal" label. If the callee (or any indirect callees) returns with the "<a
John Criswell02fdc6f2005-05-12 16:52:32 +00001822href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1823continued at the dynamically nearest "exception" label.</p>
Chris Lattner0132aff2005-05-06 22:57:40 +00001824
Chris Lattner2f7c9632001-06-06 20:29:01 +00001825<h5>Arguments:</h5>
Chris Lattner0132aff2005-05-06 22:57:40 +00001826
Misha Brukman76307852003-11-08 01:05:38 +00001827<p>This instruction requires several arguments:</p>
Chris Lattner0132aff2005-05-06 22:57:40 +00001828
Chris Lattner2f7c9632001-06-06 20:29:01 +00001829<ol>
Chris Lattner0132aff2005-05-06 22:57:40 +00001830 <li>
Duncan Sands16f122e2007-03-30 12:22:09 +00001831 The optional "cconv" marker indicates which <a href="#callingconv">calling
Chris Lattner0132aff2005-05-06 22:57:40 +00001832 convention</a> the call should use. If none is specified, the call defaults
1833 to using C calling conventions.
1834 </li>
1835 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1836 function value being invoked. In most cases, this is a direct function
1837 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1838 an arbitrary pointer to function value.
1839 </li>
1840
1841 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1842 function to be invoked. </li>
1843
1844 <li>'<tt>function args</tt>': argument list whose types match the function
1845 signature argument types. If the function signature indicates the function
1846 accepts a variable number of arguments, the extra arguments can be
1847 specified. </li>
1848
1849 <li>'<tt>normal label</tt>': the label reached when the called function
1850 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1851
1852 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1853 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1854
Chris Lattner2f7c9632001-06-06 20:29:01 +00001855</ol>
Chris Lattner0132aff2005-05-06 22:57:40 +00001856
Chris Lattner2f7c9632001-06-06 20:29:01 +00001857<h5>Semantics:</h5>
Chris Lattner0132aff2005-05-06 22:57:40 +00001858
Misha Brukman76307852003-11-08 01:05:38 +00001859<p>This instruction is designed to operate as a standard '<tt><a
Chris Lattner0132aff2005-05-06 22:57:40 +00001860href="#i_call">call</a></tt>' instruction in most regards. The primary
1861difference is that it establishes an association with a label, which is used by
1862the runtime library to unwind the stack.</p>
1863
1864<p>This instruction is used in languages with destructors to ensure that proper
1865cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1866exception. Additionally, this is important for implementation of
1867'<tt>catch</tt>' clauses in high-level languages that support them.</p>
1868
Chris Lattner2f7c9632001-06-06 20:29:01 +00001869<h5>Example:</h5>
Chris Lattner0132aff2005-05-06 22:57:40 +00001870<pre>
Jeff Cohen5819f182007-04-22 01:17:39 +00001871 %retval = invoke i32 %Test(i32 15) to label %Continue
1872 unwind label %TestCleanup <i>; {i32}:retval set</i>
1873 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1874 unwind label %TestCleanup <i>; {i32}:retval set</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001875</pre>
Misha Brukman76307852003-11-08 01:05:38 +00001876</div>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00001877
1878
Chris Lattner5ed60612003-09-03 00:41:47 +00001879<!-- _______________________________________________________________________ -->
Chris Lattner08b7d5b2004-10-16 18:04:13 +00001880
Chris Lattner48b383b02003-11-25 01:02:51 +00001881<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1882Instruction</a> </div>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00001883
Misha Brukman76307852003-11-08 01:05:38 +00001884<div class="doc_text">
Chris Lattner08b7d5b2004-10-16 18:04:13 +00001885
Chris Lattner5ed60612003-09-03 00:41:47 +00001886<h5>Syntax:</h5>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00001887<pre>
1888 unwind
1889</pre>
1890
Chris Lattner5ed60612003-09-03 00:41:47 +00001891<h5>Overview:</h5>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00001892
1893<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1894at the first callee in the dynamic call stack which used an <a
1895href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1896primarily used to implement exception handling.</p>
1897
Chris Lattner5ed60612003-09-03 00:41:47 +00001898<h5>Semantics:</h5>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00001899
1900<p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1901immediately halt. The dynamic call stack is then searched for the first <a
1902href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1903execution continues at the "exceptional" destination block specified by the
1904<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1905dynamic call chain, undefined behavior results.</p>
Misha Brukman76307852003-11-08 01:05:38 +00001906</div>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00001907
1908<!-- _______________________________________________________________________ -->
1909
1910<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1911Instruction</a> </div>
1912
1913<div class="doc_text">
1914
1915<h5>Syntax:</h5>
1916<pre>
1917 unreachable
1918</pre>
1919
1920<h5>Overview:</h5>
1921
1922<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1923instruction is used to inform the optimizer that a particular portion of the
1924code is not reachable. This can be used to indicate that the code after a
1925no-return function cannot be reached, and other facts.</p>
1926
1927<h5>Semantics:</h5>
1928
1929<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1930</div>
1931
1932
1933
Chris Lattner2f7c9632001-06-06 20:29:01 +00001934<!-- ======================================================================= -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001935<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00001936<div class="doc_text">
Chris Lattner48b383b02003-11-25 01:02:51 +00001937<p>Binary operators are used to do most of the computation in a
1938program. They require two operands, execute an operation on them, and
John Criswelldfe6a862004-12-10 15:51:16 +00001939produce a single value. The operands might represent
Reid Spencer404a3252007-02-15 03:07:05 +00001940multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001941The result value of a binary operator is not
Chris Lattner48b383b02003-11-25 01:02:51 +00001942necessarily the same type as its operands.</p>
Misha Brukman76307852003-11-08 01:05:38 +00001943<p>There are several different binary operators:</p>
Misha Brukman76307852003-11-08 01:05:38 +00001944</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001945<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001946<div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1947Instruction</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00001948<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00001949<h5>Syntax:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001950<pre> &lt;result&gt; = add &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001951</pre>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001952<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00001953<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001954<h5>Arguments:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00001955<p>The two arguments to the '<tt>add</tt>' instruction must be either <a
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001956 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
Reid Spencer404a3252007-02-15 03:07:05 +00001957 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001958Both arguments must have identical types.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001959<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00001960<p>The value produced is the integer or floating point sum of the two
1961operands.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001962<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001963<pre> &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001964</pre>
Misha Brukman76307852003-11-08 01:05:38 +00001965</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001966<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001967<div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1968Instruction</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00001969<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00001970<h5>Syntax:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001971<pre> &lt;result&gt; = sub &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001972</pre>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001973<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00001974<p>The '<tt>sub</tt>' instruction returns the difference of its two
1975operands.</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00001976<p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1977instruction present in most other intermediate representations.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001978<h5>Arguments:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00001979<p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
Chris Lattner48b383b02003-11-25 01:02:51 +00001980 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001981values.
Reid Spencer404a3252007-02-15 03:07:05 +00001982This instruction can also take <a href="#t_vector">vector</a> versions of the values.
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001983Both arguments must have identical types.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001984<h5>Semantics:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001985<p>The value produced is the integer or floating point difference of
1986the two operands.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001987<h5>Example:</h5>
Bill Wendling2d8b9a82007-05-29 09:42:13 +00001988<pre>
1989 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001990 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001991</pre>
Misha Brukman76307852003-11-08 01:05:38 +00001992</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001993<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001994<div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1995Instruction</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00001996<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00001997<h5>Syntax:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001998<pre> &lt;result&gt; = mul &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001999</pre>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002000<h5>Overview:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00002001<p>The '<tt>mul</tt>' instruction returns the product of its two
2002operands.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002003<h5>Arguments:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00002004<p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
Chris Lattner48b383b02003-11-25 01:02:51 +00002005 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
Chris Lattnerc8cb6952004-08-12 19:12:28 +00002006values.
Reid Spencer404a3252007-02-15 03:07:05 +00002007This instruction can also take <a href="#t_vector">vector</a> versions of the values.
Chris Lattnerc8cb6952004-08-12 19:12:28 +00002008Both arguments must have identical types.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002009<h5>Semantics:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00002010<p>The value produced is the integer or floating point product of the
Misha Brukman76307852003-11-08 01:05:38 +00002011two operands.</p>
Reid Spencer3e628eb92007-01-04 16:43:23 +00002012<p>Because the operands are the same width, the result of an integer
2013multiplication is the same whether the operands should be deemed unsigned or
2014signed.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002015<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002016<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002017</pre>
Misha Brukman76307852003-11-08 01:05:38 +00002018</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002019<!-- _______________________________________________________________________ -->
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002020<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2021</a></div>
2022<div class="doc_text">
2023<h5>Syntax:</h5>
2024<pre> &lt;result&gt; = udiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2025</pre>
2026<h5>Overview:</h5>
2027<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2028operands.</p>
2029<h5>Arguments:</h5>
2030<p>The two arguments to the '<tt>udiv</tt>' instruction must be
2031<a href="#t_integer">integer</a> values. Both arguments must have identical
Reid Spencer404a3252007-02-15 03:07:05 +00002032types. This instruction can also take <a href="#t_vector">vector</a> versions
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002033of the values in which case the elements must be integers.</p>
2034<h5>Semantics:</h5>
2035<p>The value produced is the unsigned integer quotient of the two operands. This
2036instruction always performs an unsigned division operation, regardless of
2037whether the arguments are unsigned or not.</p>
2038<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002039<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002040</pre>
2041</div>
2042<!-- _______________________________________________________________________ -->
2043<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2044</a> </div>
2045<div class="doc_text">
2046<h5>Syntax:</h5>
2047<pre> &lt;result&gt; = sdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2048</pre>
2049<h5>Overview:</h5>
2050<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2051operands.</p>
2052<h5>Arguments:</h5>
2053<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2054<a href="#t_integer">integer</a> values. Both arguments must have identical
Reid Spencer404a3252007-02-15 03:07:05 +00002055types. This instruction can also take <a href="#t_vector">vector</a> versions
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002056of the values in which case the elements must be integers.</p>
2057<h5>Semantics:</h5>
2058<p>The value produced is the signed integer quotient of the two operands. This
2059instruction always performs a signed division operation, regardless of whether
2060the arguments are signed or not.</p>
2061<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002062<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002063</pre>
2064</div>
2065<!-- _______________________________________________________________________ -->
2066<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
Chris Lattner48b383b02003-11-25 01:02:51 +00002067Instruction</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00002068<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00002069<h5>Syntax:</h5>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002070<pre> &lt;result&gt; = fdiv &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Chris Lattner48b383b02003-11-25 01:02:51 +00002071</pre>
2072<h5>Overview:</h5>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002073<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
Chris Lattner48b383b02003-11-25 01:02:51 +00002074operands.</p>
2075<h5>Arguments:</h5>
Jeff Cohen5819f182007-04-22 01:17:39 +00002076<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002077<a href="#t_floating">floating point</a> values. Both arguments must have
Reid Spencer404a3252007-02-15 03:07:05 +00002078identical types. This instruction can also take <a href="#t_vector">vector</a>
Jeff Cohen5819f182007-04-22 01:17:39 +00002079versions of floating point values.</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00002080<h5>Semantics:</h5>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002081<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00002082<h5>Example:</h5>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002083<pre> &lt;result&gt; = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
Chris Lattner48b383b02003-11-25 01:02:51 +00002084</pre>
2085</div>
2086<!-- _______________________________________________________________________ -->
Reid Spencer7eb55b32006-11-02 01:53:59 +00002087<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2088</div>
2089<div class="doc_text">
2090<h5>Syntax:</h5>
2091<pre> &lt;result&gt; = urem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2092</pre>
2093<h5>Overview:</h5>
2094<p>The '<tt>urem</tt>' instruction returns the remainder from the
2095unsigned division of its two arguments.</p>
2096<h5>Arguments:</h5>
2097<p>The two arguments to the '<tt>urem</tt>' instruction must be
2098<a href="#t_integer">integer</a> values. Both arguments must have identical
2099types.</p>
2100<h5>Semantics:</h5>
2101<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2102This instruction always performs an unsigned division to get the remainder,
2103regardless of whether the arguments are unsigned or not.</p>
2104<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002105<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
Reid Spencer7eb55b32006-11-02 01:53:59 +00002106</pre>
2107
2108</div>
2109<!-- _______________________________________________________________________ -->
2110<div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
Chris Lattner48b383b02003-11-25 01:02:51 +00002111Instruction</a> </div>
2112<div class="doc_text">
2113<h5>Syntax:</h5>
Reid Spencer7eb55b32006-11-02 01:53:59 +00002114<pre> &lt;result&gt; = srem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Chris Lattner48b383b02003-11-25 01:02:51 +00002115</pre>
2116<h5>Overview:</h5>
Reid Spencer7eb55b32006-11-02 01:53:59 +00002117<p>The '<tt>srem</tt>' instruction returns the remainder from the
2118signed division of its two operands.</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00002119<h5>Arguments:</h5>
Reid Spencer7eb55b32006-11-02 01:53:59 +00002120<p>The two arguments to the '<tt>srem</tt>' instruction must be
2121<a href="#t_integer">integer</a> values. Both arguments must have identical
2122types.</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00002123<h5>Semantics:</h5>
Reid Spencer7eb55b32006-11-02 01:53:59 +00002124<p>This instruction returns the <i>remainder</i> of a division (where the result
Reid Spencer806ad6a2007-03-24 22:23:39 +00002125has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2126operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2127a value. For more information about the difference, see <a
Chris Lattner48b383b02003-11-25 01:02:51 +00002128 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
Reid Spencer806ad6a2007-03-24 22:23:39 +00002129Math Forum</a>. For a table of how this is implemented in various languages,
Reid Spencerdb3b93b2007-03-24 22:40:44 +00002130please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
Reid Spencer806ad6a2007-03-24 22:23:39 +00002131Wikipedia: modulo operation</a>.</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00002132<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002133<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
Reid Spencer7eb55b32006-11-02 01:53:59 +00002134</pre>
2135
2136</div>
2137<!-- _______________________________________________________________________ -->
2138<div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2139Instruction</a> </div>
2140<div class="doc_text">
2141<h5>Syntax:</h5>
2142<pre> &lt;result&gt; = frem &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2143</pre>
2144<h5>Overview:</h5>
2145<p>The '<tt>frem</tt>' instruction returns the remainder from the
2146division of its two operands.</p>
2147<h5>Arguments:</h5>
2148<p>The two arguments to the '<tt>frem</tt>' instruction must be
2149<a href="#t_floating">floating point</a> values. Both arguments must have
2150identical types.</p>
2151<h5>Semantics:</h5>
2152<p>This instruction returns the <i>remainder</i> of a division.</p>
2153<h5>Example:</h5>
2154<pre> &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
Chris Lattner48b383b02003-11-25 01:02:51 +00002155</pre>
Misha Brukman76307852003-11-08 01:05:38 +00002156</div>
Robert Bocchino820bc75b2006-02-17 21:18:08 +00002157
Reid Spencer2ab01932007-02-02 13:57:07 +00002158<!-- ======================================================================= -->
2159<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2160Operations</a> </div>
2161<div class="doc_text">
2162<p>Bitwise binary operators are used to do various forms of
2163bit-twiddling in a program. They are generally very efficient
2164instructions and can commonly be strength reduced from other
2165instructions. They require two operands, execute an operation on them,
2166and produce a single value. The resulting value of the bitwise binary
2167operators is always the same type as its first operand.</p>
2168</div>
2169
Reid Spencer04e259b2007-01-31 21:39:12 +00002170<!-- _______________________________________________________________________ -->
2171<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2172Instruction</a> </div>
2173<div class="doc_text">
2174<h5>Syntax:</h5>
2175<pre> &lt;result&gt; = shl &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2176</pre>
2177<h5>Overview:</h5>
2178<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2179the left a specified number of bits.</p>
2180<h5>Arguments:</h5>
2181<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2182 href="#t_integer">integer</a> type.</p>
2183<h5>Semantics:</h5>
2184<p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
2185<h5>Example:</h5><pre>
2186 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2187 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2188 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2189</pre>
2190</div>
2191<!-- _______________________________________________________________________ -->
2192<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2193Instruction</a> </div>
2194<div class="doc_text">
2195<h5>Syntax:</h5>
2196<pre> &lt;result&gt; = lshr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2197</pre>
2198
2199<h5>Overview:</h5>
2200<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
Jeff Cohen5819f182007-04-22 01:17:39 +00002201operand shifted to the right a specified number of bits with zero fill.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +00002202
2203<h5>Arguments:</h5>
2204<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2205<a href="#t_integer">integer</a> type.</p>
2206
2207<h5>Semantics:</h5>
2208<p>This instruction always performs a logical shift right operation. The most
2209significant bits of the result will be filled with zero bits after the
2210shift.</p>
2211
2212<h5>Example:</h5>
2213<pre>
2214 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2215 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2216 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2217 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2218</pre>
2219</div>
2220
Reid Spencer2ab01932007-02-02 13:57:07 +00002221<!-- _______________________________________________________________________ -->
Reid Spencer04e259b2007-01-31 21:39:12 +00002222<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2223Instruction</a> </div>
2224<div class="doc_text">
2225
2226<h5>Syntax:</h5>
2227<pre> &lt;result&gt; = ashr &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
2228</pre>
2229
2230<h5>Overview:</h5>
2231<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
Jeff Cohen5819f182007-04-22 01:17:39 +00002232operand shifted to the right a specified number of bits with sign extension.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +00002233
2234<h5>Arguments:</h5>
2235<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2236<a href="#t_integer">integer</a> type.</p>
2237
2238<h5>Semantics:</h5>
2239<p>This instruction always performs an arithmetic shift right operation,
2240The most significant bits of the result will be filled with the sign bit
2241of <tt>var1</tt>.</p>
2242
2243<h5>Example:</h5>
2244<pre>
2245 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2246 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2247 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2248 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
2249</pre>
2250</div>
2251
Chris Lattner2f7c9632001-06-06 20:29:01 +00002252<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00002253<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2254Instruction</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00002255<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00002256<h5>Syntax:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00002257<pre> &lt;result&gt; = and &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002258</pre>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002259<h5>Overview:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00002260<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2261its two operands.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002262<h5>Arguments:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00002263<p>The two arguments to the '<tt>and</tt>' instruction must be <a
Chris Lattnerc0f423a2007-01-15 01:54:13 +00002264 href="#t_integer">integer</a> values. Both arguments must have
Chris Lattner48b383b02003-11-25 01:02:51 +00002265identical types.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002266<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00002267<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00002268<p> </p>
Misha Brukmanc501f552004-03-01 17:47:27 +00002269<div style="align: center">
Misha Brukman76307852003-11-08 01:05:38 +00002270<table border="1" cellspacing="0" cellpadding="4">
Chris Lattner48b383b02003-11-25 01:02:51 +00002271 <tbody>
2272 <tr>
2273 <td>In0</td>
2274 <td>In1</td>
2275 <td>Out</td>
2276 </tr>
2277 <tr>
2278 <td>0</td>
2279 <td>0</td>
2280 <td>0</td>
2281 </tr>
2282 <tr>
2283 <td>0</td>
2284 <td>1</td>
2285 <td>0</td>
2286 </tr>
2287 <tr>
2288 <td>1</td>
2289 <td>0</td>
2290 <td>0</td>
2291 </tr>
2292 <tr>
2293 <td>1</td>
2294 <td>1</td>
2295 <td>1</td>
2296 </tr>
2297 </tbody>
2298</table>
Misha Brukmanc501f552004-03-01 17:47:27 +00002299</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002300<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002301<pre> &lt;result&gt; = and i32 4, %var <i>; yields {i32}:result = 4 &amp; %var</i>
2302 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2303 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002304</pre>
Misha Brukman76307852003-11-08 01:05:38 +00002305</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002306<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00002307<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00002308<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00002309<h5>Syntax:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00002310<pre> &lt;result&gt; = or &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002311</pre>
Chris Lattner48b383b02003-11-25 01:02:51 +00002312<h5>Overview:</h5>
2313<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2314or of its two operands.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002315<h5>Arguments:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00002316<p>The two arguments to the '<tt>or</tt>' instruction must be <a
Chris Lattnerc0f423a2007-01-15 01:54:13 +00002317 href="#t_integer">integer</a> values. Both arguments must have
Chris Lattner48b383b02003-11-25 01:02:51 +00002318identical types.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002319<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00002320<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00002321<p> </p>
Misha Brukmanc501f552004-03-01 17:47:27 +00002322<div style="align: center">
Chris Lattner48b383b02003-11-25 01:02:51 +00002323<table border="1" cellspacing="0" cellpadding="4">
2324 <tbody>
2325 <tr>
2326 <td>In0</td>
2327 <td>In1</td>
2328 <td>Out</td>
2329 </tr>
2330 <tr>
2331 <td>0</td>
2332 <td>0</td>
2333 <td>0</td>
2334 </tr>
2335 <tr>
2336 <td>0</td>
2337 <td>1</td>
2338 <td>1</td>
2339 </tr>
2340 <tr>
2341 <td>1</td>
2342 <td>0</td>
2343 <td>1</td>
2344 </tr>
2345 <tr>
2346 <td>1</td>
2347 <td>1</td>
2348 <td>1</td>
2349 </tr>
2350 </tbody>
2351</table>
Misha Brukmanc501f552004-03-01 17:47:27 +00002352</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002353<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002354<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2355 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2356 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002357</pre>
Misha Brukman76307852003-11-08 01:05:38 +00002358</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002359<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00002360<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2361Instruction</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00002362<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00002363<h5>Syntax:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00002364<pre> &lt;result&gt; = xor &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {ty}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002365</pre>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002366<h5>Overview:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00002367<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2368or of its two operands. The <tt>xor</tt> is used to implement the
2369"one's complement" operation, which is the "~" operator in C.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002370<h5>Arguments:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00002371<p>The two arguments to the '<tt>xor</tt>' instruction must be <a
Chris Lattnerc0f423a2007-01-15 01:54:13 +00002372 href="#t_integer">integer</a> values. Both arguments must have
Chris Lattner48b383b02003-11-25 01:02:51 +00002373identical types.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002374<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00002375<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00002376<p> </p>
Misha Brukmanc501f552004-03-01 17:47:27 +00002377<div style="align: center">
Chris Lattner48b383b02003-11-25 01:02:51 +00002378<table border="1" cellspacing="0" cellpadding="4">
2379 <tbody>
2380 <tr>
2381 <td>In0</td>
2382 <td>In1</td>
2383 <td>Out</td>
2384 </tr>
2385 <tr>
2386 <td>0</td>
2387 <td>0</td>
2388 <td>0</td>
2389 </tr>
2390 <tr>
2391 <td>0</td>
2392 <td>1</td>
2393 <td>1</td>
2394 </tr>
2395 <tr>
2396 <td>1</td>
2397 <td>0</td>
2398 <td>1</td>
2399 </tr>
2400 <tr>
2401 <td>1</td>
2402 <td>1</td>
2403 <td>0</td>
2404 </tr>
2405 </tbody>
2406</table>
Misha Brukmanc501f552004-03-01 17:47:27 +00002407</div>
Chris Lattner48b383b02003-11-25 01:02:51 +00002408<p> </p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002409<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002410<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2411 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2412 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2413 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002414</pre>
Misha Brukman76307852003-11-08 01:05:38 +00002415</div>
Chris Lattner54611b42005-11-06 08:02:57 +00002416
Chris Lattner2f7c9632001-06-06 20:29:01 +00002417<!-- ======================================================================= -->
Chris Lattner54611b42005-11-06 08:02:57 +00002418<div class="doc_subsection">
Chris Lattnerce83bff2006-04-08 23:07:04 +00002419 <a name="vectorops">Vector Operations</a>
2420</div>
2421
2422<div class="doc_text">
2423
2424<p>LLVM supports several instructions to represent vector operations in a
Jeff Cohen5819f182007-04-22 01:17:39 +00002425target-independent manner. These instructions cover the element-access and
Chris Lattnerce83bff2006-04-08 23:07:04 +00002426vector-specific operations needed to process vectors effectively. While LLVM
2427does directly support these vector operations, many sophisticated algorithms
2428will want to use target-specific intrinsics to take full advantage of a specific
2429target.</p>
2430
2431</div>
2432
2433<!-- _______________________________________________________________________ -->
2434<div class="doc_subsubsection">
2435 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2436</div>
2437
2438<div class="doc_text">
2439
2440<h5>Syntax:</h5>
2441
2442<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002443 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
Chris Lattnerce83bff2006-04-08 23:07:04 +00002444</pre>
2445
2446<h5>Overview:</h5>
2447
2448<p>
2449The '<tt>extractelement</tt>' instruction extracts a single scalar
Reid Spencer404a3252007-02-15 03:07:05 +00002450element from a vector at a specified index.
Chris Lattnerce83bff2006-04-08 23:07:04 +00002451</p>
2452
2453
2454<h5>Arguments:</h5>
2455
2456<p>
2457The first operand of an '<tt>extractelement</tt>' instruction is a
Reid Spencer404a3252007-02-15 03:07:05 +00002458value of <a href="#t_vector">vector</a> type. The second operand is
Chris Lattnerce83bff2006-04-08 23:07:04 +00002459an index indicating the position from which to extract the element.
2460The index may be a variable.</p>
2461
2462<h5>Semantics:</h5>
2463
2464<p>
2465The result is a scalar of the same type as the element type of
2466<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2467<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2468results are undefined.
2469</p>
2470
2471<h5>Example:</h5>
2472
2473<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002474 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
Chris Lattnerce83bff2006-04-08 23:07:04 +00002475</pre>
2476</div>
2477
2478
2479<!-- _______________________________________________________________________ -->
2480<div class="doc_subsubsection">
2481 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2482</div>
2483
2484<div class="doc_text">
2485
2486<h5>Syntax:</h5>
2487
2488<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002489 &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>
Chris Lattnerce83bff2006-04-08 23:07:04 +00002490</pre>
2491
2492<h5>Overview:</h5>
2493
2494<p>
2495The '<tt>insertelement</tt>' instruction inserts a scalar
Reid Spencer404a3252007-02-15 03:07:05 +00002496element into a vector at a specified index.
Chris Lattnerce83bff2006-04-08 23:07:04 +00002497</p>
2498
2499
2500<h5>Arguments:</h5>
2501
2502<p>
2503The first operand of an '<tt>insertelement</tt>' instruction is a
Reid Spencer404a3252007-02-15 03:07:05 +00002504value of <a href="#t_vector">vector</a> type. The second operand is a
Chris Lattnerce83bff2006-04-08 23:07:04 +00002505scalar value whose type must equal the element type of the first
2506operand. The third operand is an index indicating the position at
2507which to insert the value. The index may be a variable.</p>
2508
2509<h5>Semantics:</h5>
2510
2511<p>
Reid Spencer404a3252007-02-15 03:07:05 +00002512The result is a vector of the same type as <tt>val</tt>. Its
Chris Lattnerce83bff2006-04-08 23:07:04 +00002513element values are those of <tt>val</tt> except at position
2514<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2515exceeds the length of <tt>val</tt>, the results are undefined.
2516</p>
2517
2518<h5>Example:</h5>
2519
2520<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002521 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
Chris Lattnerce83bff2006-04-08 23:07:04 +00002522</pre>
2523</div>
2524
2525<!-- _______________________________________________________________________ -->
2526<div class="doc_subsubsection">
2527 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2528</div>
2529
2530<div class="doc_text">
2531
2532<h5>Syntax:</h5>
2533
2534<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002535 &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>
Chris Lattnerce83bff2006-04-08 23:07:04 +00002536</pre>
2537
2538<h5>Overview:</h5>
2539
2540<p>
2541The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2542from two input vectors, returning a vector of the same type.
2543</p>
2544
2545<h5>Arguments:</h5>
2546
2547<p>
2548The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2549with types that match each other and types that match the result of the
2550instruction. The third argument is a shuffle mask, which has the same number
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002551of elements as the other vector type, but whose element type is always 'i32'.
Chris Lattnerce83bff2006-04-08 23:07:04 +00002552</p>
2553
2554<p>
2555The shuffle mask operand is required to be a constant vector with either
2556constant integer or undef values.
2557</p>
2558
2559<h5>Semantics:</h5>
2560
2561<p>
2562The elements of the two input vectors are numbered from left to right across
2563both of the vectors. The shuffle mask operand specifies, for each element of
2564the result vector, which element of the two input registers the result element
2565gets. The element selector may be undef (meaning "don't care") and the second
2566operand may be undef if performing a shuffle from only one vector.
2567</p>
2568
2569<h5>Example:</h5>
2570
2571<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002572 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
Jeff Cohen5819f182007-04-22 01:17:39 +00002573 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002574 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
2575 &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.
Chris Lattnerce83bff2006-04-08 23:07:04 +00002576</pre>
2577</div>
2578
Tanya Lattnerb138bbe2006-04-14 19:24:33 +00002579
Chris Lattnerce83bff2006-04-08 23:07:04 +00002580<!-- ======================================================================= -->
2581<div class="doc_subsection">
Chris Lattner6ab66722006-08-15 00:45:58 +00002582 <a name="memoryops">Memory Access and Addressing Operations</a>
Chris Lattner54611b42005-11-06 08:02:57 +00002583</div>
2584
Misha Brukman76307852003-11-08 01:05:38 +00002585<div class="doc_text">
Chris Lattner54611b42005-11-06 08:02:57 +00002586
Chris Lattner48b383b02003-11-25 01:02:51 +00002587<p>A key design point of an SSA-based representation is how it
2588represents memory. In LLVM, no memory locations are in SSA form, which
2589makes things very simple. This section describes how to read, write,
John Criswelldfe6a862004-12-10 15:51:16 +00002590allocate, and free memory in LLVM.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00002591
Misha Brukman76307852003-11-08 01:05:38 +00002592</div>
Chris Lattner54611b42005-11-06 08:02:57 +00002593
Chris Lattner2f7c9632001-06-06 20:29:01 +00002594<!-- _______________________________________________________________________ -->
Chris Lattner54611b42005-11-06 08:02:57 +00002595<div class="doc_subsubsection">
2596 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2597</div>
2598
Misha Brukman76307852003-11-08 01:05:38 +00002599<div class="doc_text">
Chris Lattner54611b42005-11-06 08:02:57 +00002600
Chris Lattner2f7c9632001-06-06 20:29:01 +00002601<h5>Syntax:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00002602
2603<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002604 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002605</pre>
Chris Lattner54611b42005-11-06 08:02:57 +00002606
Chris Lattner2f7c9632001-06-06 20:29:01 +00002607<h5>Overview:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00002608
Chris Lattner48b383b02003-11-25 01:02:51 +00002609<p>The '<tt>malloc</tt>' instruction allocates memory from the system
2610heap and returns a pointer to it.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00002611
Chris Lattner2f7c9632001-06-06 20:29:01 +00002612<h5>Arguments:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00002613
2614<p>The '<tt>malloc</tt>' instruction allocates
2615<tt>sizeof(&lt;type&gt;)*NumElements</tt>
John Criswella92e5862004-02-24 16:13:56 +00002616bytes of memory from the operating system and returns a pointer of the
Chris Lattner54611b42005-11-06 08:02:57 +00002617appropriate type to the program. If "NumElements" is specified, it is the
2618number of elements allocated. If an alignment is specified, the value result
2619of the allocation is guaranteed to be aligned to at least that boundary. If
2620not specified, or if zero, the target can choose to align the allocation on any
2621convenient boundary.</p>
2622
Misha Brukman76307852003-11-08 01:05:38 +00002623<p>'<tt>type</tt>' must be a sized type.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00002624
Chris Lattner2f7c9632001-06-06 20:29:01 +00002625<h5>Semantics:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00002626
Chris Lattner48b383b02003-11-25 01:02:51 +00002627<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2628a pointer is returned.</p>
Misha Brukman76307852003-11-08 01:05:38 +00002629
Chris Lattner54611b42005-11-06 08:02:57 +00002630<h5>Example:</h5>
2631
2632<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002633 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
Chris Lattner54611b42005-11-06 08:02:57 +00002634
Bill Wendling2d8b9a82007-05-29 09:42:13 +00002635 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2636 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2637 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2638 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2639 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002640</pre>
Misha Brukman76307852003-11-08 01:05:38 +00002641</div>
Chris Lattner54611b42005-11-06 08:02:57 +00002642
Chris Lattner2f7c9632001-06-06 20:29:01 +00002643<!-- _______________________________________________________________________ -->
Chris Lattner54611b42005-11-06 08:02:57 +00002644<div class="doc_subsubsection">
2645 <a name="i_free">'<tt>free</tt>' Instruction</a>
2646</div>
2647
Misha Brukman76307852003-11-08 01:05:38 +00002648<div class="doc_text">
Chris Lattner54611b42005-11-06 08:02:57 +00002649
Chris Lattner2f7c9632001-06-06 20:29:01 +00002650<h5>Syntax:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00002651
2652<pre>
2653 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002654</pre>
Chris Lattner54611b42005-11-06 08:02:57 +00002655
Chris Lattner2f7c9632001-06-06 20:29:01 +00002656<h5>Overview:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00002657
Chris Lattner48b383b02003-11-25 01:02:51 +00002658<p>The '<tt>free</tt>' instruction returns memory back to the unused
John Criswell4a3327e2005-05-13 22:25:59 +00002659memory heap to be reallocated in the future.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00002660
Chris Lattner2f7c9632001-06-06 20:29:01 +00002661<h5>Arguments:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00002662
Chris Lattner48b383b02003-11-25 01:02:51 +00002663<p>'<tt>value</tt>' shall be a pointer value that points to a value
2664that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2665instruction.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00002666
Chris Lattner2f7c9632001-06-06 20:29:01 +00002667<h5>Semantics:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00002668
John Criswelldfe6a862004-12-10 15:51:16 +00002669<p>Access to the memory pointed to by the pointer is no longer defined
Chris Lattner48b383b02003-11-25 01:02:51 +00002670after this instruction executes.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00002671
Chris Lattner2f7c9632001-06-06 20:29:01 +00002672<h5>Example:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00002673
2674<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002675 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2676 free [4 x i8]* %array
Chris Lattner2f7c9632001-06-06 20:29:01 +00002677</pre>
Misha Brukman76307852003-11-08 01:05:38 +00002678</div>
Chris Lattner54611b42005-11-06 08:02:57 +00002679
Chris Lattner2f7c9632001-06-06 20:29:01 +00002680<!-- _______________________________________________________________________ -->
Chris Lattner54611b42005-11-06 08:02:57 +00002681<div class="doc_subsubsection">
2682 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2683</div>
2684
Misha Brukman76307852003-11-08 01:05:38 +00002685<div class="doc_text">
Chris Lattner54611b42005-11-06 08:02:57 +00002686
Chris Lattner2f7c9632001-06-06 20:29:01 +00002687<h5>Syntax:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00002688
2689<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002690 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002691</pre>
Chris Lattner54611b42005-11-06 08:02:57 +00002692
Chris Lattner2f7c9632001-06-06 20:29:01 +00002693<h5>Overview:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00002694
Jeff Cohen5819f182007-04-22 01:17:39 +00002695<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
2696currently executing function, to be automatically released when this function
Chris Lattner48b383b02003-11-25 01:02:51 +00002697returns to its caller.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00002698
Chris Lattner2f7c9632001-06-06 20:29:01 +00002699<h5>Arguments:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00002700
John Criswelldfe6a862004-12-10 15:51:16 +00002701<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
Chris Lattner48b383b02003-11-25 01:02:51 +00002702bytes of memory on the runtime stack, returning a pointer of the
Chris Lattner54611b42005-11-06 08:02:57 +00002703appropriate type to the program. If "NumElements" is specified, it is the
2704number of elements allocated. If an alignment is specified, the value result
2705of the allocation is guaranteed to be aligned to at least that boundary. If
2706not specified, or if zero, the target can choose to align the allocation on any
2707convenient boundary.</p>
2708
Misha Brukman76307852003-11-08 01:05:38 +00002709<p>'<tt>type</tt>' may be any sized type.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00002710
Chris Lattner2f7c9632001-06-06 20:29:01 +00002711<h5>Semantics:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00002712
John Criswell4a3327e2005-05-13 22:25:59 +00002713<p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
Chris Lattner48b383b02003-11-25 01:02:51 +00002714memory is automatically released when the function returns. The '<tt>alloca</tt>'
2715instruction is commonly used to represent automatic variables that must
2716have an address available. When the function returns (either with the <tt><a
John Criswellc932bef2005-05-12 16:55:34 +00002717 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
Misha Brukman76307852003-11-08 01:05:38 +00002718instructions), the memory is reclaimed.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00002719
Chris Lattner2f7c9632001-06-06 20:29:01 +00002720<h5>Example:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00002721
2722<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002723 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
Bill Wendling3716c5d2007-05-29 09:04:49 +00002724 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2725 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002726 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002727</pre>
Misha Brukman76307852003-11-08 01:05:38 +00002728</div>
Chris Lattner54611b42005-11-06 08:02:57 +00002729
Chris Lattner2f7c9632001-06-06 20:29:01 +00002730<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00002731<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2732Instruction</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00002733<div class="doc_text">
Chris Lattner095735d2002-05-06 03:03:22 +00002734<h5>Syntax:</h5>
Christopher Lambbff50202007-04-21 08:16:25 +00002735<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>
Chris Lattner095735d2002-05-06 03:03:22 +00002736<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00002737<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
Chris Lattner095735d2002-05-06 03:03:22 +00002738<h5>Arguments:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00002739<p>The argument to the '<tt>load</tt>' instruction specifies the memory
John Criswell4c0cf7f2005-10-24 16:17:18 +00002740address from which to load. The pointer must point to a <a
Chris Lattner10ee9652004-06-03 22:57:15 +00002741 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
John Criswell4c0cf7f2005-10-24 16:17:18 +00002742marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
Chris Lattner48b383b02003-11-25 01:02:51 +00002743the number or order of execution of this <tt>load</tt> with other
2744volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2745instructions. </p>
Chris Lattner095735d2002-05-06 03:03:22 +00002746<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00002747<p>The location of memory pointed to is loaded.</p>
Chris Lattner095735d2002-05-06 03:03:22 +00002748<h5>Examples:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002749<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Chris Lattner48b383b02003-11-25 01:02:51 +00002750 <a
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002751 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2752 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
Chris Lattner095735d2002-05-06 03:03:22 +00002753</pre>
Misha Brukman76307852003-11-08 01:05:38 +00002754</div>
Chris Lattner095735d2002-05-06 03:03:22 +00002755<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00002756<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2757Instruction</a> </div>
Reid Spencera89fb182006-11-09 21:18:01 +00002758<div class="doc_text">
Chris Lattner095735d2002-05-06 03:03:22 +00002759<h5>Syntax:</h5>
Christopher Lambbff50202007-04-21 08:16:25 +00002760<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
2761 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
Chris Lattner095735d2002-05-06 03:03:22 +00002762</pre>
Chris Lattner095735d2002-05-06 03:03:22 +00002763<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00002764<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
Chris Lattner095735d2002-05-06 03:03:22 +00002765<h5>Arguments:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00002766<p>There are two arguments to the '<tt>store</tt>' instruction: a value
Jeff Cohen5819f182007-04-22 01:17:39 +00002767to store and an address at which to store it. The type of the '<tt>&lt;pointer&gt;</tt>'
Chris Lattner48b383b02003-11-25 01:02:51 +00002768operand must be a pointer to the type of the '<tt>&lt;value&gt;</tt>'
John Criswell4a3327e2005-05-13 22:25:59 +00002769operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
Chris Lattner48b383b02003-11-25 01:02:51 +00002770optimizer is not allowed to modify the number or order of execution of
2771this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2772 href="#i_store">store</a></tt> instructions.</p>
2773<h5>Semantics:</h5>
2774<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
2775at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.</p>
Chris Lattner095735d2002-05-06 03:03:22 +00002776<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002777<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Chris Lattner48b383b02003-11-25 01:02:51 +00002778 <a
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002779 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2780 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
Chris Lattner095735d2002-05-06 03:03:22 +00002781</pre>
Reid Spencer443460a2006-11-09 21:15:49 +00002782</div>
2783
Chris Lattner095735d2002-05-06 03:03:22 +00002784<!-- _______________________________________________________________________ -->
Chris Lattner33fd7022004-04-05 01:30:49 +00002785<div class="doc_subsubsection">
2786 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2787</div>
2788
Misha Brukman76307852003-11-08 01:05:38 +00002789<div class="doc_text">
Chris Lattner590645f2002-04-14 06:13:44 +00002790<h5>Syntax:</h5>
Chris Lattner33fd7022004-04-05 01:30:49 +00002791<pre>
2792 &lt;result&gt; = getelementptr &lt;ty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
2793</pre>
2794
Chris Lattner590645f2002-04-14 06:13:44 +00002795<h5>Overview:</h5>
Chris Lattner33fd7022004-04-05 01:30:49 +00002796
2797<p>
2798The '<tt>getelementptr</tt>' instruction is used to get the address of a
2799subelement of an aggregate data structure.</p>
2800
Chris Lattner590645f2002-04-14 06:13:44 +00002801<h5>Arguments:</h5>
Chris Lattner33fd7022004-04-05 01:30:49 +00002802
Reid Spencercee005c2006-12-04 21:29:24 +00002803<p>This instruction takes a list of integer operands that indicate what
Chris Lattner33fd7022004-04-05 01:30:49 +00002804elements of the aggregate object to index to. The actual types of the arguments
2805provided depend on the type of the first pointer argument. The
2806'<tt>getelementptr</tt>' instruction is used to index down through the type
John Criswell88190562005-05-16 16:17:45 +00002807levels of a structure or to a specific index in an array. When indexing into a
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002808structure, only <tt>i32</tt> integer constants are allowed. When indexing
Reid Spencercee005c2006-12-04 21:29:24 +00002809into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2810be sign extended to 64-bit values.</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00002811
Chris Lattner48b383b02003-11-25 01:02:51 +00002812<p>For example, let's consider a C code fragment and how it gets
2813compiled to LLVM:</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00002814
Bill Wendling3716c5d2007-05-29 09:04:49 +00002815<div class="doc_code">
Chris Lattner33fd7022004-04-05 01:30:49 +00002816<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00002817struct RT {
2818 char A;
Chris Lattnera446f1b2007-05-29 15:43:56 +00002819 int B[10][20];
Bill Wendling3716c5d2007-05-29 09:04:49 +00002820 char C;
2821};
2822struct ST {
Chris Lattnera446f1b2007-05-29 15:43:56 +00002823 int X;
Bill Wendling3716c5d2007-05-29 09:04:49 +00002824 double Y;
2825 struct RT Z;
2826};
Chris Lattner33fd7022004-04-05 01:30:49 +00002827
Chris Lattnera446f1b2007-05-29 15:43:56 +00002828int *foo(struct ST *s) {
Bill Wendling3716c5d2007-05-29 09:04:49 +00002829 return &amp;s[1].Z.B[5][13];
2830}
Chris Lattner33fd7022004-04-05 01:30:49 +00002831</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00002832</div>
Chris Lattner33fd7022004-04-05 01:30:49 +00002833
Misha Brukman76307852003-11-08 01:05:38 +00002834<p>The LLVM code generated by the GCC frontend is:</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00002835
Bill Wendling3716c5d2007-05-29 09:04:49 +00002836<div class="doc_code">
Chris Lattner33fd7022004-04-05 01:30:49 +00002837<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00002838%RT = type { i8 , [10 x [20 x i32]], i8 }
2839%ST = type { i32, double, %RT }
Chris Lattner33fd7022004-04-05 01:30:49 +00002840
Bill Wendling3716c5d2007-05-29 09:04:49 +00002841define i32* %foo(%ST* %s) {
2842entry:
2843 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
2844 ret i32* %reg
2845}
Chris Lattner33fd7022004-04-05 01:30:49 +00002846</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00002847</div>
Chris Lattner33fd7022004-04-05 01:30:49 +00002848
Chris Lattner590645f2002-04-14 06:13:44 +00002849<h5>Semantics:</h5>
Chris Lattner33fd7022004-04-05 01:30:49 +00002850
2851<p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
John Criswell4a3327e2005-05-13 22:25:59 +00002852on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
Reid Spencercee005c2006-12-04 21:29:24 +00002853and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
Reid Spencerc0312692006-12-03 16:53:48 +00002854<a href="#t_integer">integer</a> type but the value will always be sign extended
Jeff Cohen5819f182007-04-22 01:17:39 +00002855to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
Reid Spencerc0312692006-12-03 16:53:48 +00002856<b>constants</b>.</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00002857
Misha Brukman76307852003-11-08 01:05:38 +00002858<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002859type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
Chris Lattner33fd7022004-04-05 01:30:49 +00002860}</tt>' type, a structure. The second index indexes into the third element of
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002861the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
2862i8 }</tt>' type, another structure. The third index indexes into the second
2863element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
Chris Lattner33fd7022004-04-05 01:30:49 +00002864array. The two dimensions of the array are subscripted into, yielding an
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002865'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2866to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00002867
Chris Lattner48b383b02003-11-25 01:02:51 +00002868<p>Note that it is perfectly legal to index partially through a
2869structure, returning a pointer to an inner element. Because of this,
2870the LLVM code for the given testcase is equivalent to:</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00002871
2872<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002873 define i32* %foo(%ST* %s) {
2874 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
Jeff Cohen5819f182007-04-22 01:17:39 +00002875 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
2876 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002877 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
2878 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
2879 ret i32* %t5
Chris Lattner33fd7022004-04-05 01:30:49 +00002880 }
Chris Lattnera8292f32002-05-06 22:08:29 +00002881</pre>
Chris Lattnerc0ad71e2005-06-24 17:22:57 +00002882
2883<p>Note that it is undefined to access an array out of bounds: array and
2884pointer indexes must always be within the defined bounds of the array type.
2885The one exception for this rules is zero length arrays. These arrays are
2886defined to be accessible as variable length arrays, which requires access
2887beyond the zero'th element.</p>
2888
Chris Lattner6ab66722006-08-15 00:45:58 +00002889<p>The getelementptr instruction is often confusing. For some more insight
2890into how it works, see <a href="GetElementPtr.html">the getelementptr
2891FAQ</a>.</p>
2892
Chris Lattner590645f2002-04-14 06:13:44 +00002893<h5>Example:</h5>
Chris Lattnerc0ad71e2005-06-24 17:22:57 +00002894
Chris Lattner33fd7022004-04-05 01:30:49 +00002895<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002896 <i>; yields [12 x i8]*:aptr</i>
2897 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
Chris Lattner33fd7022004-04-05 01:30:49 +00002898</pre>
Chris Lattner33fd7022004-04-05 01:30:49 +00002899</div>
Reid Spencer443460a2006-11-09 21:15:49 +00002900
Chris Lattner2f7c9632001-06-06 20:29:01 +00002901<!-- ======================================================================= -->
Reid Spencer97c5fa42006-11-08 01:18:52 +00002902<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
Misha Brukman76307852003-11-08 01:05:38 +00002903</div>
Misha Brukman76307852003-11-08 01:05:38 +00002904<div class="doc_text">
Reid Spencer97c5fa42006-11-08 01:18:52 +00002905<p>The instructions in this category are the conversion instructions (casting)
2906which all take a single operand and a type. They perform various bit conversions
2907on the operand.</p>
Misha Brukman76307852003-11-08 01:05:38 +00002908</div>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00002909
Chris Lattnera8292f32002-05-06 22:08:29 +00002910<!-- _______________________________________________________________________ -->
Chris Lattnerb53c28d2004-03-12 05:50:16 +00002911<div class="doc_subsubsection">
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002912 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2913</div>
2914<div class="doc_text">
2915
2916<h5>Syntax:</h5>
2917<pre>
2918 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
2919</pre>
2920
2921<h5>Overview:</h5>
2922<p>
2923The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2924</p>
2925
2926<h5>Arguments:</h5>
2927<p>
2928The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2929be an <a href="#t_integer">integer</a> type, and a type that specifies the size
Chris Lattnerc0f423a2007-01-15 01:54:13 +00002930and type of the result, which must be an <a href="#t_integer">integer</a>
Reid Spencer51b07252006-11-09 23:03:26 +00002931type. The bit size of <tt>value</tt> must be larger than the bit size of
2932<tt>ty2</tt>. Equal sized types are not allowed.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002933
2934<h5>Semantics:</h5>
2935<p>
2936The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
Reid Spencer51b07252006-11-09 23:03:26 +00002937and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
2938larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
2939It will always truncate bits.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002940
2941<h5>Example:</h5>
2942<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002943 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
Reid Spencer36a15422007-01-12 03:35:51 +00002944 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
2945 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002946</pre>
2947</div>
2948
2949<!-- _______________________________________________________________________ -->
2950<div class="doc_subsubsection">
2951 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
2952</div>
2953<div class="doc_text">
2954
2955<h5>Syntax:</h5>
2956<pre>
2957 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
2958</pre>
2959
2960<h5>Overview:</h5>
2961<p>The '<tt>zext</tt>' instruction zero extends its operand to type
2962<tt>ty2</tt>.</p>
2963
2964
2965<h5>Arguments:</h5>
2966<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
Chris Lattnerc0f423a2007-01-15 01:54:13 +00002967<a href="#t_integer">integer</a> type, and a type to cast it to, which must
2968also be of <a href="#t_integer">integer</a> type. The bit size of the
Reid Spencer51b07252006-11-09 23:03:26 +00002969<tt>value</tt> must be smaller than the bit size of the destination type,
2970<tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002971
2972<h5>Semantics:</h5>
2973<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
Chris Lattnerc87f3df2007-05-24 19:13:27 +00002974bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002975
Reid Spencer07c9c682007-01-12 15:46:11 +00002976<p>When zero extending from i1, the result will always be either 0 or 1.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002977
2978<h5>Example:</h5>
2979<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002980 %X = zext i32 257 to i64 <i>; yields i64:257</i>
Reid Spencer36a15422007-01-12 03:35:51 +00002981 %Y = zext i1 true to i32 <i>; yields i32:1</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002982</pre>
2983</div>
2984
2985<!-- _______________________________________________________________________ -->
2986<div class="doc_subsubsection">
2987 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
2988</div>
2989<div class="doc_text">
2990
2991<h5>Syntax:</h5>
2992<pre>
2993 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
2994</pre>
2995
2996<h5>Overview:</h5>
2997<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
2998
2999<h5>Arguments:</h5>
3000<p>
3001The '<tt>sext</tt>' instruction takes a value to cast, which must be of
Chris Lattnerc0f423a2007-01-15 01:54:13 +00003002<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3003also be of <a href="#t_integer">integer</a> type. The bit size of the
Reid Spencer51b07252006-11-09 23:03:26 +00003004<tt>value</tt> must be smaller than the bit size of the destination type,
3005<tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003006
3007<h5>Semantics:</h5>
3008<p>
3009The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3010bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
Chris Lattnerc87f3df2007-05-24 19:13:27 +00003011the type <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003012
Reid Spencer36a15422007-01-12 03:35:51 +00003013<p>When sign extending from i1, the extension always results in -1 or 0.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003014
3015<h5>Example:</h5>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003016<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003017 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
Reid Spencer36a15422007-01-12 03:35:51 +00003018 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003019</pre>
3020</div>
3021
3022<!-- _______________________________________________________________________ -->
3023<div class="doc_subsubsection">
Reid Spencer2e2740d2006-11-09 21:48:10 +00003024 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3025</div>
3026
3027<div class="doc_text">
3028
3029<h5>Syntax:</h5>
3030
3031<pre>
3032 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3033</pre>
3034
3035<h5>Overview:</h5>
3036<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3037<tt>ty2</tt>.</p>
3038
3039
3040<h5>Arguments:</h5>
3041<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3042 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3043cast it to. The size of <tt>value</tt> must be larger than the size of
3044<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3045<i>no-op cast</i>.</p>
3046
3047<h5>Semantics:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00003048<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3049<a href="#t_floating">floating point</a> type to a smaller
3050<a href="#t_floating">floating point</a> type. If the value cannot fit within
3051the destination type, <tt>ty2</tt>, then the results are undefined.</p>
Reid Spencer2e2740d2006-11-09 21:48:10 +00003052
3053<h5>Example:</h5>
3054<pre>
3055 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3056 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3057</pre>
3058</div>
3059
3060<!-- _______________________________________________________________________ -->
3061<div class="doc_subsubsection">
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003062 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3063</div>
3064<div class="doc_text">
3065
3066<h5>Syntax:</h5>
3067<pre>
3068 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3069</pre>
3070
3071<h5>Overview:</h5>
3072<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3073floating point value.</p>
3074
3075<h5>Arguments:</h5>
3076<p>The '<tt>fpext</tt>' instruction takes a
3077<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
Reid Spencer51b07252006-11-09 23:03:26 +00003078and a <a href="#t_floating">floating point</a> type to cast it to. The source
3079type must be smaller than the destination type.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003080
3081<h5>Semantics:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00003082<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
Duncan Sands16f122e2007-03-30 12:22:09 +00003083<a href="#t_floating">floating point</a> type to a larger
3084<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
Reid Spencer51b07252006-11-09 23:03:26 +00003085used to make a <i>no-op cast</i> because it always changes bits. Use
Reid Spencer5b950642006-11-11 23:08:07 +00003086<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003087
3088<h5>Example:</h5>
3089<pre>
3090 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3091 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3092</pre>
3093</div>
3094
3095<!-- _______________________________________________________________________ -->
3096<div class="doc_subsubsection">
Reid Spencer2eadb532007-01-21 00:29:26 +00003097 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003098</div>
3099<div class="doc_text">
3100
3101<h5>Syntax:</h5>
3102<pre>
Reid Spencer753163d2007-07-31 14:40:14 +00003103 &lt;result&gt; = fptoui &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003104</pre>
3105
3106<h5>Overview:</h5>
Reid Spencer753163d2007-07-31 14:40:14 +00003107<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003108unsigned integer equivalent of type <tt>ty2</tt>.
3109</p>
3110
3111<h5>Arguments:</h5>
Reid Spencer753163d2007-07-31 14:40:14 +00003112<p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003113<a href="#t_floating">floating point</a> value, and a type to cast it to, which
Chris Lattnerc0f423a2007-01-15 01:54:13 +00003114must be an <a href="#t_integer">integer</a> type.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003115
3116<h5>Semantics:</h5>
Reid Spencer753163d2007-07-31 14:40:14 +00003117<p> The '<tt>fptoui</tt>' instruction converts its
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003118<a href="#t_floating">floating point</a> operand into the nearest (rounding
3119towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3120the results are undefined.</p>
3121
Reid Spencer36a15422007-01-12 03:35:51 +00003122<p>When converting to i1, the conversion is done as a comparison against
3123zero. If the <tt>value</tt> was zero, the i1 result will be <tt>false</tt>.
3124If the <tt>value</tt> was non-zero, the i1 result will be <tt>true</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003125
3126<h5>Example:</h5>
3127<pre>
Reid Spencer753163d2007-07-31 14:40:14 +00003128 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
3129 %Y = fptoui float 1.0E+300 to i1 <i>; yields i1:true</i>
3130 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003131</pre>
3132</div>
3133
3134<!-- _______________________________________________________________________ -->
3135<div class="doc_subsubsection">
Reid Spencer51b07252006-11-09 23:03:26 +00003136 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003137</div>
3138<div class="doc_text">
3139
3140<h5>Syntax:</h5>
3141<pre>
Reid Spencer51b07252006-11-09 23:03:26 +00003142 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003143</pre>
3144
3145<h5>Overview:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00003146<p>The '<tt>fptosi</tt>' instruction converts
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003147<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
Chris Lattnerb53c28d2004-03-12 05:50:16 +00003148</p>
3149
3150
Chris Lattnera8292f32002-05-06 22:08:29 +00003151<h5>Arguments:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00003152<p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003153<a href="#t_floating">floating point</a> value, and a type to cast it to, which
Chris Lattnerc0f423a2007-01-15 01:54:13 +00003154must also be an <a href="#t_integer">integer</a> type.</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00003155
Chris Lattnera8292f32002-05-06 22:08:29 +00003156<h5>Semantics:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00003157<p>The '<tt>fptosi</tt>' instruction converts its
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003158<a href="#t_floating">floating point</a> operand into the nearest (rounding
3159towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3160the results are undefined.</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00003161
Reid Spencer36a15422007-01-12 03:35:51 +00003162<p>When converting to i1, the conversion is done as a comparison against
3163zero. If the <tt>value</tt> was zero, the i1 result will be <tt>false</tt>.
3164If the <tt>value</tt> was non-zero, the i1 result will be <tt>true</tt>.</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00003165
Chris Lattner70de6632001-07-09 00:26:23 +00003166<h5>Example:</h5>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00003167<pre>
Reid Spencer36a15422007-01-12 03:35:51 +00003168 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
3169 %Y = fptosi float 1.0E-247 to i1 <i>; yields i1:true</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003170 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003171</pre>
3172</div>
3173
3174<!-- _______________________________________________________________________ -->
3175<div class="doc_subsubsection">
Reid Spencer51b07252006-11-09 23:03:26 +00003176 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003177</div>
3178<div class="doc_text">
3179
3180<h5>Syntax:</h5>
3181<pre>
Reid Spencer51b07252006-11-09 23:03:26 +00003182 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003183</pre>
3184
3185<h5>Overview:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00003186<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003187integer and converts that value to the <tt>ty2</tt> type.</p>
3188
3189
3190<h5>Arguments:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00003191<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be an
Chris Lattnerc0f423a2007-01-15 01:54:13 +00003192<a href="#t_integer">integer</a> value, and a type to cast it to, which must
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003193be a <a href="#t_floating">floating point</a> type.</p>
3194
3195<h5>Semantics:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00003196<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003197integer quantity and converts it to the corresponding floating point value. If
Jeff Cohenbeccb742007-04-22 14:56:37 +00003198the value cannot fit in the floating point value, the results are undefined.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003199
3200
3201<h5>Example:</h5>
3202<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003203 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
Jeff Cohen222a8a42007-04-29 01:07:00 +00003204 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003205</pre>
3206</div>
3207
3208<!-- _______________________________________________________________________ -->
3209<div class="doc_subsubsection">
Reid Spencer51b07252006-11-09 23:03:26 +00003210 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003211</div>
3212<div class="doc_text">
3213
3214<h5>Syntax:</h5>
3215<pre>
Reid Spencer51b07252006-11-09 23:03:26 +00003216 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003217</pre>
3218
3219<h5>Overview:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00003220<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003221integer and converts that value to the <tt>ty2</tt> type.</p>
3222
3223<h5>Arguments:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00003224<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be an
Chris Lattnerc0f423a2007-01-15 01:54:13 +00003225<a href="#t_integer">integer</a> value, and a type to cast it to, which must be
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003226a <a href="#t_floating">floating point</a> type.</p>
3227
3228<h5>Semantics:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00003229<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003230integer quantity and converts it to the corresponding floating point value. If
Jeff Cohenbeccb742007-04-22 14:56:37 +00003231the value cannot fit in the floating point value, the results are undefined.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003232
3233<h5>Example:</h5>
3234<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003235 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
Jeff Cohen222a8a42007-04-29 01:07:00 +00003236 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003237</pre>
3238</div>
3239
3240<!-- _______________________________________________________________________ -->
3241<div class="doc_subsubsection">
Reid Spencerb7344ff2006-11-11 21:00:47 +00003242 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3243</div>
3244<div class="doc_text">
3245
3246<h5>Syntax:</h5>
3247<pre>
3248 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3249</pre>
3250
3251<h5>Overview:</h5>
3252<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3253the integer type <tt>ty2</tt>.</p>
3254
3255<h5>Arguments:</h5>
3256<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
Duncan Sands16f122e2007-03-30 12:22:09 +00003257must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
Reid Spencerb7344ff2006-11-11 21:00:47 +00003258<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3259
3260<h5>Semantics:</h5>
3261<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3262<tt>ty2</tt> by interpreting the pointer value as an integer and either
3263truncating or zero extending that value to the size of the integer type. If
3264<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3265<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
Jeff Cohen222a8a42007-04-29 01:07:00 +00003266are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3267change.</p>
Reid Spencerb7344ff2006-11-11 21:00:47 +00003268
3269<h5>Example:</h5>
3270<pre>
Jeff Cohen222a8a42007-04-29 01:07:00 +00003271 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3272 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
Reid Spencerb7344ff2006-11-11 21:00:47 +00003273</pre>
3274</div>
3275
3276<!-- _______________________________________________________________________ -->
3277<div class="doc_subsubsection">
3278 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3279</div>
3280<div class="doc_text">
3281
3282<h5>Syntax:</h5>
3283<pre>
3284 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3285</pre>
3286
3287<h5>Overview:</h5>
3288<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3289a pointer type, <tt>ty2</tt>.</p>
3290
3291<h5>Arguments:</h5>
Duncan Sands16f122e2007-03-30 12:22:09 +00003292<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
Reid Spencerb7344ff2006-11-11 21:00:47 +00003293value to cast, and a type to cast it to, which must be a
Anton Korobeynikova0554d92007-01-12 19:20:47 +00003294<a href="#t_pointer">pointer</a> type.
Reid Spencerb7344ff2006-11-11 21:00:47 +00003295
3296<h5>Semantics:</h5>
3297<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3298<tt>ty2</tt> by applying either a zero extension or a truncation depending on
3299the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3300size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3301the size of a pointer then a zero extension is done. If they are the same size,
3302nothing is done (<i>no-op cast</i>).</p>
3303
3304<h5>Example:</h5>
3305<pre>
Jeff Cohen222a8a42007-04-29 01:07:00 +00003306 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3307 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3308 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
Reid Spencerb7344ff2006-11-11 21:00:47 +00003309</pre>
3310</div>
3311
3312<!-- _______________________________________________________________________ -->
3313<div class="doc_subsubsection">
Reid Spencer5b950642006-11-11 23:08:07 +00003314 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003315</div>
3316<div class="doc_text">
3317
3318<h5>Syntax:</h5>
3319<pre>
Reid Spencer5b950642006-11-11 23:08:07 +00003320 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003321</pre>
3322
3323<h5>Overview:</h5>
Reid Spencer5b950642006-11-11 23:08:07 +00003324<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003325<tt>ty2</tt> without changing any bits.</p>
3326
3327<h5>Arguments:</h5>
Reid Spencer5b950642006-11-11 23:08:07 +00003328<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003329a first class value, and a type to cast it to, which must also be a <a
3330 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
Reid Spencere3db84c2007-01-09 20:08:58 +00003331and the destination type, <tt>ty2</tt>, must be identical. If the source
3332type is a pointer, the destination type must also be a pointer.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003333
3334<h5>Semantics:</h5>
Reid Spencer5b950642006-11-11 23:08:07 +00003335<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
Reid Spencerb7344ff2006-11-11 21:00:47 +00003336<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3337this conversion. The conversion is done as if the <tt>value</tt> had been
3338stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3339converted to other pointer types with this instruction. To convert pointers to
3340other types, use the <a href="#i_inttoptr">inttoptr</a> or
3341<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003342
3343<h5>Example:</h5>
3344<pre>
Jeff Cohen222a8a42007-04-29 01:07:00 +00003345 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003346 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3347 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
Chris Lattner70de6632001-07-09 00:26:23 +00003348</pre>
Misha Brukman76307852003-11-08 01:05:38 +00003349</div>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00003350
Reid Spencer97c5fa42006-11-08 01:18:52 +00003351<!-- ======================================================================= -->
3352<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3353<div class="doc_text">
3354<p>The instructions in this category are the "miscellaneous"
3355instructions, which defy better classification.</p>
3356</div>
Reid Spencerc828a0e2006-11-18 21:50:54 +00003357
3358<!-- _______________________________________________________________________ -->
3359<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3360</div>
3361<div class="doc_text">
3362<h5>Syntax:</h5>
Jeff Cohen222a8a42007-04-29 01:07:00 +00003363<pre> &lt;result&gt; = icmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
Reid Spencerc828a0e2006-11-18 21:50:54 +00003364</pre>
3365<h5>Overview:</h5>
3366<p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3367of its two integer operands.</p>
3368<h5>Arguments:</h5>
3369<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
Jeff Cohen222a8a42007-04-29 01:07:00 +00003370the condition code indicating the kind of comparison to perform. It is not
3371a value, just a keyword. The possible condition code are:
Reid Spencerc828a0e2006-11-18 21:50:54 +00003372<ol>
3373 <li><tt>eq</tt>: equal</li>
3374 <li><tt>ne</tt>: not equal </li>
3375 <li><tt>ugt</tt>: unsigned greater than</li>
3376 <li><tt>uge</tt>: unsigned greater or equal</li>
3377 <li><tt>ult</tt>: unsigned less than</li>
3378 <li><tt>ule</tt>: unsigned less or equal</li>
3379 <li><tt>sgt</tt>: signed greater than</li>
3380 <li><tt>sge</tt>: signed greater or equal</li>
3381 <li><tt>slt</tt>: signed less than</li>
3382 <li><tt>sle</tt>: signed less or equal</li>
3383</ol>
Chris Lattnerc0f423a2007-01-15 01:54:13 +00003384<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
Reid Spencer784ef792007-01-04 05:19:58 +00003385<a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
Reid Spencerc828a0e2006-11-18 21:50:54 +00003386<h5>Semantics:</h5>
3387<p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3388the condition code given as <tt>cond</tt>. The comparison performed always
Reid Spencer36a15422007-01-12 03:35:51 +00003389yields a <a href="#t_primitive">i1</a> result, as follows:
Reid Spencerc828a0e2006-11-18 21:50:54 +00003390<ol>
3391 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3392 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3393 </li>
3394 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3395 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3396 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3397 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3398 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3399 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3400 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3401 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3402 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3403 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3404 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3405 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3406 <li><tt>sge</tt>: interprets the operands as signed values and yields
3407 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3408 <li><tt>slt</tt>: interprets the operands as signed values and yields
3409 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3410 <li><tt>sle</tt>: interprets the operands as signed values and yields
3411 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00003412</ol>
3413<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
Jeff Cohen222a8a42007-04-29 01:07:00 +00003414values are compared as if they were integers.</p>
Reid Spencerc828a0e2006-11-18 21:50:54 +00003415
3416<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003417<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
3418 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
3419 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
3420 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3421 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3422 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
Reid Spencerc828a0e2006-11-18 21:50:54 +00003423</pre>
3424</div>
3425
3426<!-- _______________________________________________________________________ -->
3427<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3428</div>
3429<div class="doc_text">
3430<h5>Syntax:</h5>
Jeff Cohen222a8a42007-04-29 01:07:00 +00003431<pre> &lt;result&gt; = fcmp &lt;cond&gt; &lt;ty&gt; &lt;var1&gt;, &lt;var2&gt; <i>; yields {i1}:result</i>
Reid Spencerc828a0e2006-11-18 21:50:54 +00003432</pre>
3433<h5>Overview:</h5>
3434<p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3435of its floating point operands.</p>
3436<h5>Arguments:</h5>
3437<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
Jeff Cohen222a8a42007-04-29 01:07:00 +00003438the condition code indicating the kind of comparison to perform. It is not
3439a value, just a keyword. The possible condition code are:
Reid Spencerc828a0e2006-11-18 21:50:54 +00003440<ol>
Reid Spencerf69acf32006-11-19 03:00:14 +00003441 <li><tt>false</tt>: no comparison, always returns false</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00003442 <li><tt>oeq</tt>: ordered and equal</li>
3443 <li><tt>ogt</tt>: ordered and greater than </li>
3444 <li><tt>oge</tt>: ordered and greater than or equal</li>
3445 <li><tt>olt</tt>: ordered and less than </li>
3446 <li><tt>ole</tt>: ordered and less than or equal</li>
3447 <li><tt>one</tt>: ordered and not equal</li>
3448 <li><tt>ord</tt>: ordered (no nans)</li>
3449 <li><tt>ueq</tt>: unordered or equal</li>
3450 <li><tt>ugt</tt>: unordered or greater than </li>
3451 <li><tt>uge</tt>: unordered or greater than or equal</li>
3452 <li><tt>ult</tt>: unordered or less than </li>
3453 <li><tt>ule</tt>: unordered or less than or equal</li>
3454 <li><tt>une</tt>: unordered or not equal</li>
3455 <li><tt>uno</tt>: unordered (either nans)</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00003456 <li><tt>true</tt>: no comparison, always returns true</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00003457</ol>
Jeff Cohen222a8a42007-04-29 01:07:00 +00003458<p><i>Ordered</i> means that neither operand is a QNAN while
Reid Spencer02e0d1d2006-12-06 07:08:07 +00003459<i>unordered</i> means that either operand may be a QNAN.</p>
Reid Spencer784ef792007-01-04 05:19:58 +00003460<p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3461<a href="#t_floating">floating point</a> typed. They must have identical
3462types.</p>
Reid Spencerc828a0e2006-11-18 21:50:54 +00003463<h5>Semantics:</h5>
3464<p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3465the condition code given as <tt>cond</tt>. The comparison performed always
Reid Spencer36a15422007-01-12 03:35:51 +00003466yields a <a href="#t_primitive">i1</a> result, as follows:
Reid Spencerc828a0e2006-11-18 21:50:54 +00003467<ol>
3468 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00003469 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Reid Spencerc828a0e2006-11-18 21:50:54 +00003470 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00003471 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Reid Spencerc828a0e2006-11-18 21:50:54 +00003472 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00003473 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Reid Spencerc828a0e2006-11-18 21:50:54 +00003474 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00003475 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Reid Spencerc828a0e2006-11-18 21:50:54 +00003476 <tt>var1</tt> is less than <tt>var2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00003477 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Reid Spencerc828a0e2006-11-18 21:50:54 +00003478 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00003479 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Reid Spencerc828a0e2006-11-18 21:50:54 +00003480 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00003481 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3482 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
Reid Spencerc828a0e2006-11-18 21:50:54 +00003483 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00003484 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
Reid Spencerc828a0e2006-11-18 21:50:54 +00003485 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00003486 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
Reid Spencerc828a0e2006-11-18 21:50:54 +00003487 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00003488 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
Reid Spencerc828a0e2006-11-18 21:50:54 +00003489 <tt>var1</tt> is less than <tt>var2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00003490 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
Reid Spencerc828a0e2006-11-18 21:50:54 +00003491 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00003492 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
Reid Spencerc828a0e2006-11-18 21:50:54 +00003493 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00003494 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00003495 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3496</ol>
Reid Spencerc828a0e2006-11-18 21:50:54 +00003497
3498<h5>Example:</h5>
3499<pre> &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3500 &lt;result&gt; = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3501 &lt;result&gt; = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3502 &lt;result&gt; = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3503</pre>
3504</div>
3505
Reid Spencer97c5fa42006-11-08 01:18:52 +00003506<!-- _______________________________________________________________________ -->
3507<div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3508Instruction</a> </div>
3509<div class="doc_text">
3510<h5>Syntax:</h5>
3511<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
3512<h5>Overview:</h5>
3513<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
3514the SSA graph representing the function.</p>
3515<h5>Arguments:</h5>
Jeff Cohen222a8a42007-04-29 01:07:00 +00003516<p>The type of the incoming values is specified with the first type
Reid Spencer97c5fa42006-11-08 01:18:52 +00003517field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3518as arguments, with one pair for each predecessor basic block of the
3519current block. Only values of <a href="#t_firstclass">first class</a>
3520type may be used as the value arguments to the PHI node. Only labels
3521may be used as the label arguments.</p>
3522<p>There must be no non-phi instructions between the start of a basic
3523block and the PHI instructions: i.e. PHI instructions must be first in
3524a basic block.</p>
3525<h5>Semantics:</h5>
Jeff Cohen222a8a42007-04-29 01:07:00 +00003526<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3527specified by the pair corresponding to the predecessor basic block that executed
3528just prior to the current block.</p>
Reid Spencer97c5fa42006-11-08 01:18:52 +00003529<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003530<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>
Reid Spencer97c5fa42006-11-08 01:18:52 +00003531</div>
3532
Chris Lattnerb53c28d2004-03-12 05:50:16 +00003533<!-- _______________________________________________________________________ -->
3534<div class="doc_subsubsection">
3535 <a name="i_select">'<tt>select</tt>' Instruction</a>
3536</div>
3537
3538<div class="doc_text">
3539
3540<h5>Syntax:</h5>
3541
3542<pre>
Reid Spencer36a15422007-01-12 03:35:51 +00003543 &lt;result&gt; = select i1 &lt;cond&gt;, &lt;ty&gt; &lt;val1&gt;, &lt;ty&gt; &lt;val2&gt; <i>; yields ty</i>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00003544</pre>
3545
3546<h5>Overview:</h5>
3547
3548<p>
3549The '<tt>select</tt>' instruction is used to choose one value based on a
3550condition, without branching.
3551</p>
3552
3553
3554<h5>Arguments:</h5>
3555
3556<p>
3557The '<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.
3558</p>
3559
3560<h5>Semantics:</h5>
3561
3562<p>
3563If the boolean condition evaluates to true, the instruction returns the first
John Criswell88190562005-05-16 16:17:45 +00003564value argument; otherwise, it returns the second value argument.
Chris Lattnerb53c28d2004-03-12 05:50:16 +00003565</p>
3566
3567<h5>Example:</h5>
3568
3569<pre>
Reid Spencer36a15422007-01-12 03:35:51 +00003570 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00003571</pre>
3572</div>
3573
Robert Bocchinof72fdfe2006-01-15 20:48:27 +00003574
3575<!-- _______________________________________________________________________ -->
3576<div class="doc_subsubsection">
Chris Lattnere23c1392005-05-06 05:47:36 +00003577 <a name="i_call">'<tt>call</tt>' Instruction</a>
3578</div>
3579
Misha Brukman76307852003-11-08 01:05:38 +00003580<div class="doc_text">
Chris Lattnere23c1392005-05-06 05:47:36 +00003581
Chris Lattner2f7c9632001-06-06 20:29:01 +00003582<h5>Syntax:</h5>
Chris Lattnere23c1392005-05-06 05:47:36 +00003583<pre>
Nick Lewyckya9b13d52007-09-08 13:57:50 +00003584 &lt;result&gt; = [tail] call [<a href="#callingconv">cconv</a>] &lt;ty&gt; [&lt;fnty&gt;*] &lt;fnptrval&gt;(&lt;param list&gt;)
Chris Lattnere23c1392005-05-06 05:47:36 +00003585</pre>
3586
Chris Lattner2f7c9632001-06-06 20:29:01 +00003587<h5>Overview:</h5>
Chris Lattnere23c1392005-05-06 05:47:36 +00003588
Misha Brukman76307852003-11-08 01:05:38 +00003589<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
Chris Lattnere23c1392005-05-06 05:47:36 +00003590
Chris Lattner2f7c9632001-06-06 20:29:01 +00003591<h5>Arguments:</h5>
Chris Lattnere23c1392005-05-06 05:47:36 +00003592
Misha Brukman76307852003-11-08 01:05:38 +00003593<p>This instruction requires several arguments:</p>
Chris Lattnere23c1392005-05-06 05:47:36 +00003594
Chris Lattnera8292f32002-05-06 22:08:29 +00003595<ol>
Chris Lattner48b383b02003-11-25 01:02:51 +00003596 <li>
Chris Lattner0132aff2005-05-06 22:57:40 +00003597 <p>The optional "tail" marker indicates whether the callee function accesses
3598 any allocas or varargs in the caller. If the "tail" marker is present, the
Chris Lattnere23c1392005-05-06 05:47:36 +00003599 function call is eligible for tail call optimization. Note that calls may
3600 be marked "tail" even if they do not occur before a <a
3601 href="#i_ret"><tt>ret</tt></a> instruction.
Chris Lattner48b383b02003-11-25 01:02:51 +00003602 </li>
3603 <li>
Duncan Sands16f122e2007-03-30 12:22:09 +00003604 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
Chris Lattner0132aff2005-05-06 22:57:40 +00003605 convention</a> the call should use. If none is specified, the call defaults
3606 to using C calling conventions.
3607 </li>
3608 <li>
Nick Lewyckya9b13d52007-09-08 13:57:50 +00003609 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
3610 the type of the return value. Functions that return no value are marked
3611 <tt><a href="#t_void">void</a></tt>.</p>
3612 </li>
3613 <li>
3614 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
3615 value being invoked. The argument types must match the types implied by
3616 this signature. This type can be omitted if the function is not varargs
3617 and if the function type does not return a pointer to a function.</p>
Chris Lattnere23c1392005-05-06 05:47:36 +00003618 </li>
3619 <li>
3620 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3621 be invoked. In most cases, this is a direct function invocation, but
3622 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
John Criswell88190562005-05-16 16:17:45 +00003623 to function value.</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00003624 </li>
3625 <li>
3626 <p>'<tt>function args</tt>': argument list whose types match the
Reid Spencerd845d162005-05-01 22:22:57 +00003627 function signature argument types. All arguments must be of
3628 <a href="#t_firstclass">first class</a> type. If the function signature
3629 indicates the function accepts a variable number of arguments, the extra
3630 arguments can be specified.</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00003631 </li>
Chris Lattnera8292f32002-05-06 22:08:29 +00003632</ol>
Chris Lattnere23c1392005-05-06 05:47:36 +00003633
Chris Lattner2f7c9632001-06-06 20:29:01 +00003634<h5>Semantics:</h5>
Chris Lattnere23c1392005-05-06 05:47:36 +00003635
Chris Lattner48b383b02003-11-25 01:02:51 +00003636<p>The '<tt>call</tt>' instruction is used to cause control flow to
3637transfer to a specified function, with its incoming arguments bound to
3638the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3639instruction in the called function, control flow continues with the
3640instruction after the function call, and the return value of the
3641function is bound to the result argument. This is a simpler case of
3642the <a href="#i_invoke">invoke</a> instruction.</p>
Chris Lattnere23c1392005-05-06 05:47:36 +00003643
Chris Lattner2f7c9632001-06-06 20:29:01 +00003644<h5>Example:</h5>
Chris Lattnere23c1392005-05-06 05:47:36 +00003645
3646<pre>
Nick Lewyckya9b13d52007-09-08 13:57:50 +00003647 %retval = call i32 @test(i32 %argc)
3648 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42);
3649 %X = tail call i32 @foo()
3650 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo()
3651 %Z = call void %foo(i8 97 signext)
Chris Lattnere23c1392005-05-06 05:47:36 +00003652</pre>
3653
Misha Brukman76307852003-11-08 01:05:38 +00003654</div>
Chris Lattner6a4a0492004-09-27 21:51:25 +00003655
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003656<!-- _______________________________________________________________________ -->
Chris Lattner6a4a0492004-09-27 21:51:25 +00003657<div class="doc_subsubsection">
Chris Lattner33337472006-01-13 23:26:01 +00003658 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
Chris Lattner6a4a0492004-09-27 21:51:25 +00003659</div>
3660
Misha Brukman76307852003-11-08 01:05:38 +00003661<div class="doc_text">
Chris Lattner6a4a0492004-09-27 21:51:25 +00003662
Chris Lattner26ca62e2003-10-18 05:51:36 +00003663<h5>Syntax:</h5>
Chris Lattner6a4a0492004-09-27 21:51:25 +00003664
3665<pre>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00003666 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
Chris Lattner6a4a0492004-09-27 21:51:25 +00003667</pre>
3668
Chris Lattner26ca62e2003-10-18 05:51:36 +00003669<h5>Overview:</h5>
Chris Lattner6a4a0492004-09-27 21:51:25 +00003670
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00003671<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
Chris Lattner6a4a0492004-09-27 21:51:25 +00003672the "variable argument" area of a function call. It is used to implement the
3673<tt>va_arg</tt> macro in C.</p>
3674
Chris Lattner26ca62e2003-10-18 05:51:36 +00003675<h5>Arguments:</h5>
Chris Lattner6a4a0492004-09-27 21:51:25 +00003676
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00003677<p>This instruction takes a <tt>va_list*</tt> value and the type of
3678the argument. It returns a value of the specified argument type and
Jeff Cohen222a8a42007-04-29 01:07:00 +00003679increments the <tt>va_list</tt> to point to the next argument. The
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00003680actual type of <tt>va_list</tt> is target specific.</p>
Chris Lattner6a4a0492004-09-27 21:51:25 +00003681
Chris Lattner26ca62e2003-10-18 05:51:36 +00003682<h5>Semantics:</h5>
Chris Lattner6a4a0492004-09-27 21:51:25 +00003683
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00003684<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3685type from the specified <tt>va_list</tt> and causes the
3686<tt>va_list</tt> to point to the next argument. For more information,
3687see the variable argument handling <a href="#int_varargs">Intrinsic
3688Functions</a>.</p>
Chris Lattner6a4a0492004-09-27 21:51:25 +00003689
3690<p>It is legal for this instruction to be called in a function which does not
3691take a variable number of arguments, for example, the <tt>vfprintf</tt>
Misha Brukman76307852003-11-08 01:05:38 +00003692function.</p>
Chris Lattner6a4a0492004-09-27 21:51:25 +00003693
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00003694<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
John Criswell88190562005-05-16 16:17:45 +00003695href="#intrinsics">intrinsic function</a> because it takes a type as an
Chris Lattner6a4a0492004-09-27 21:51:25 +00003696argument.</p>
3697
Chris Lattner26ca62e2003-10-18 05:51:36 +00003698<h5>Example:</h5>
Chris Lattner6a4a0492004-09-27 21:51:25 +00003699
3700<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3701
Misha Brukman76307852003-11-08 01:05:38 +00003702</div>
Chris Lattner941515c2004-01-06 05:31:32 +00003703
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003704<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +00003705<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3706<!-- *********************************************************************** -->
Chris Lattner941515c2004-01-06 05:31:32 +00003707
Misha Brukman76307852003-11-08 01:05:38 +00003708<div class="doc_text">
Chris Lattnerfee11462004-02-12 17:01:32 +00003709
3710<p>LLVM supports the notion of an "intrinsic function". These functions have
Reid Spencer4eefaab2007-04-01 08:04:23 +00003711well known names and semantics and are required to follow certain restrictions.
3712Overall, these intrinsics represent an extension mechanism for the LLVM
Jeff Cohen222a8a42007-04-29 01:07:00 +00003713language that does not require changing all of the transformations in LLVM when
Gabor Greifa54634a2007-07-06 22:07:22 +00003714adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
Chris Lattnerfee11462004-02-12 17:01:32 +00003715
John Criswell88190562005-05-16 16:17:45 +00003716<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
Jeff Cohen222a8a42007-04-29 01:07:00 +00003717prefix is reserved in LLVM for intrinsic names; thus, function names may not
3718begin with this prefix. Intrinsic functions must always be external functions:
3719you cannot define the body of intrinsic functions. Intrinsic functions may
3720only be used in call or invoke instructions: it is illegal to take the address
3721of an intrinsic function. Additionally, because intrinsic functions are part
3722of the LLVM language, it is required if any are added that they be documented
3723here.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +00003724
Chandler Carruth7132e002007-08-04 01:51:18 +00003725<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
3726a family of functions that perform the same operation but on different data
3727types. Because LLVM can represent over 8 million different integer types,
3728overloading is used commonly to allow an intrinsic function to operate on any
3729integer type. One or more of the argument types or the result type can be
3730overloaded to accept any integer type. Argument types may also be defined as
3731exactly matching a previous argument's type or the result type. This allows an
3732intrinsic function which accepts multiple arguments, but needs all of them to
3733be of the same type, to only be overloaded with respect to a single argument or
3734the result.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +00003735
Chandler Carruth7132e002007-08-04 01:51:18 +00003736<p>Overloaded intrinsics will have the names of its overloaded argument types
3737encoded into its function name, each preceded by a period. Only those types
3738which are overloaded result in a name suffix. Arguments whose type is matched
3739against another type do not. For example, the <tt>llvm.ctpop</tt> function can
3740take an integer of any width and returns an integer of exactly the same integer
3741width. This leads to a family of functions such as
3742<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
3743Only one type, the return type, is overloaded, and only one type suffix is
3744required. Because the argument's type is matched against the return type, it
3745does not require its own name suffix.</p>
Reid Spencer4eefaab2007-04-01 08:04:23 +00003746
3747<p>To learn how to add an intrinsic function, please see the
3748<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
Chris Lattnerfee11462004-02-12 17:01:32 +00003749</p>
3750
Misha Brukman76307852003-11-08 01:05:38 +00003751</div>
Chris Lattner941515c2004-01-06 05:31:32 +00003752
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003753<!-- ======================================================================= -->
Chris Lattner941515c2004-01-06 05:31:32 +00003754<div class="doc_subsection">
3755 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3756</div>
3757
Misha Brukman76307852003-11-08 01:05:38 +00003758<div class="doc_text">
Chris Lattner757528b0b2004-05-23 21:06:01 +00003759
Misha Brukman76307852003-11-08 01:05:38 +00003760<p>Variable argument support is defined in LLVM with the <a
Chris Lattner33337472006-01-13 23:26:01 +00003761 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
Chris Lattner48b383b02003-11-25 01:02:51 +00003762intrinsic functions. These functions are related to the similarly
3763named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00003764
Chris Lattner48b383b02003-11-25 01:02:51 +00003765<p>All of these functions operate on arguments that use a
3766target-specific value type "<tt>va_list</tt>". The LLVM assembly
3767language reference manual does not define what this type is, so all
Jeff Cohen222a8a42007-04-29 01:07:00 +00003768transformations should be prepared to handle these functions regardless of
3769the type used.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00003770
Chris Lattner30b868d2006-05-15 17:26:46 +00003771<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
Chris Lattner48b383b02003-11-25 01:02:51 +00003772instruction and the variable argument handling intrinsic functions are
3773used.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00003774
Bill Wendling3716c5d2007-05-29 09:04:49 +00003775<div class="doc_code">
Chris Lattnerfee11462004-02-12 17:01:32 +00003776<pre>
Anton Korobeynikov640bbe02007-03-21 23:58:04 +00003777define i32 @test(i32 %X, ...) {
Chris Lattnerfee11462004-02-12 17:01:32 +00003778 ; Initialize variable argument processing
Jeff Cohen222a8a42007-04-29 01:07:00 +00003779 %ap = alloca i8*
Chris Lattnerdb0790c2007-01-08 07:55:15 +00003780 %ap2 = bitcast i8** %ap to i8*
Anton Korobeynikov640bbe02007-03-21 23:58:04 +00003781 call void @llvm.va_start(i8* %ap2)
Chris Lattnerfee11462004-02-12 17:01:32 +00003782
3783 ; Read a single integer argument
Jeff Cohen222a8a42007-04-29 01:07:00 +00003784 %tmp = va_arg i8** %ap, i32
Chris Lattnerfee11462004-02-12 17:01:32 +00003785
3786 ; Demonstrate usage of llvm.va_copy and llvm.va_end
Jeff Cohen222a8a42007-04-29 01:07:00 +00003787 %aq = alloca i8*
Chris Lattnerdb0790c2007-01-08 07:55:15 +00003788 %aq2 = bitcast i8** %aq to i8*
Jeff Cohen222a8a42007-04-29 01:07:00 +00003789 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
Anton Korobeynikov640bbe02007-03-21 23:58:04 +00003790 call void @llvm.va_end(i8* %aq2)
Chris Lattnerfee11462004-02-12 17:01:32 +00003791
3792 ; Stop processing of arguments.
Anton Korobeynikov640bbe02007-03-21 23:58:04 +00003793 call void @llvm.va_end(i8* %ap2)
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003794 ret i32 %tmp
Chris Lattnerfee11462004-02-12 17:01:32 +00003795}
Anton Korobeynikov640bbe02007-03-21 23:58:04 +00003796
3797declare void @llvm.va_start(i8*)
3798declare void @llvm.va_copy(i8*, i8*)
3799declare void @llvm.va_end(i8*)
Chris Lattnerfee11462004-02-12 17:01:32 +00003800</pre>
Misha Brukman76307852003-11-08 01:05:38 +00003801</div>
Chris Lattner941515c2004-01-06 05:31:32 +00003802
Bill Wendling3716c5d2007-05-29 09:04:49 +00003803</div>
3804
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003805<!-- _______________________________________________________________________ -->
Chris Lattner941515c2004-01-06 05:31:32 +00003806<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00003807 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
Chris Lattner941515c2004-01-06 05:31:32 +00003808</div>
3809
3810
Misha Brukman76307852003-11-08 01:05:38 +00003811<div class="doc_text">
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003812<h5>Syntax:</h5>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00003813<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003814<h5>Overview:</h5>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00003815<P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3816<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
3817href="#i_va_arg">va_arg</a></tt>.</p>
3818
3819<h5>Arguments:</h5>
3820
3821<P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3822
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003823<h5>Semantics:</h5>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00003824
3825<P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3826macro available in C. In a target-dependent way, it initializes the
Jeff Cohen222a8a42007-04-29 01:07:00 +00003827<tt>va_list</tt> element to which the argument points, so that the next call to
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00003828<tt>va_arg</tt> will produce the first variable argument passed to the function.
3829Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
Jeff Cohen222a8a42007-04-29 01:07:00 +00003830last argument of the function as the compiler can figure that out.</p>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00003831
Misha Brukman76307852003-11-08 01:05:38 +00003832</div>
Chris Lattner941515c2004-01-06 05:31:32 +00003833
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003834<!-- _______________________________________________________________________ -->
Chris Lattner941515c2004-01-06 05:31:32 +00003835<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00003836 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
Chris Lattner941515c2004-01-06 05:31:32 +00003837</div>
3838
Misha Brukman76307852003-11-08 01:05:38 +00003839<div class="doc_text">
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003840<h5>Syntax:</h5>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00003841<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003842<h5>Overview:</h5>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00003843
Jeff Cohen222a8a42007-04-29 01:07:00 +00003844<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
Reid Spencer96a5f022007-04-04 02:42:35 +00003845which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
Chris Lattner48b383b02003-11-25 01:02:51 +00003846or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00003847
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003848<h5>Arguments:</h5>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00003849
Jeff Cohen222a8a42007-04-29 01:07:00 +00003850<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00003851
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003852<h5>Semantics:</h5>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00003853
Misha Brukman76307852003-11-08 01:05:38 +00003854<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
Jeff Cohen222a8a42007-04-29 01:07:00 +00003855macro available in C. In a target-dependent way, it destroys the
3856<tt>va_list</tt> element to which the argument points. Calls to <a
3857href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
3858<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
3859<tt>llvm.va_end</tt>.</p>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00003860
Misha Brukman76307852003-11-08 01:05:38 +00003861</div>
Chris Lattner941515c2004-01-06 05:31:32 +00003862
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003863<!-- _______________________________________________________________________ -->
Chris Lattner941515c2004-01-06 05:31:32 +00003864<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00003865 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
Chris Lattner941515c2004-01-06 05:31:32 +00003866</div>
3867
Misha Brukman76307852003-11-08 01:05:38 +00003868<div class="doc_text">
Chris Lattner757528b0b2004-05-23 21:06:01 +00003869
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003870<h5>Syntax:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00003871
3872<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00003873 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
Chris Lattner757528b0b2004-05-23 21:06:01 +00003874</pre>
3875
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003876<h5>Overview:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00003877
Jeff Cohen222a8a42007-04-29 01:07:00 +00003878<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
3879from the source argument list to the destination argument list.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00003880
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003881<h5>Arguments:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00003882
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00003883<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
Andrew Lenharth5305ea52005-06-22 20:38:11 +00003884The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00003885
Chris Lattner757528b0b2004-05-23 21:06:01 +00003886
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00003887<h5>Semantics:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00003888
Jeff Cohen222a8a42007-04-29 01:07:00 +00003889<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
3890macro available in C. In a target-dependent way, it copies the source
3891<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
3892intrinsic is necessary because the <tt><a href="#int_va_start">
3893llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
3894example, memory allocation.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00003895
Misha Brukman76307852003-11-08 01:05:38 +00003896</div>
Chris Lattner941515c2004-01-06 05:31:32 +00003897
Chris Lattnerfee11462004-02-12 17:01:32 +00003898<!-- ======================================================================= -->
3899<div class="doc_subsection">
Chris Lattner757528b0b2004-05-23 21:06:01 +00003900 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3901</div>
3902
3903<div class="doc_text">
3904
3905<p>
3906LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3907Collection</a> requires the implementation and generation of these intrinsics.
Reid Spencer96a5f022007-04-04 02:42:35 +00003908These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
Chris Lattner757528b0b2004-05-23 21:06:01 +00003909stack</a>, as well as garbage collector implementations that require <a
Reid Spencer96a5f022007-04-04 02:42:35 +00003910href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
Chris Lattner757528b0b2004-05-23 21:06:01 +00003911Front-ends for type-safe garbage collected languages should generate these
3912intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3913href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3914</p>
3915</div>
3916
3917<!-- _______________________________________________________________________ -->
3918<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00003919 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
Chris Lattner757528b0b2004-05-23 21:06:01 +00003920</div>
3921
3922<div class="doc_text">
3923
3924<h5>Syntax:</h5>
3925
3926<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00003927 declare void @llvm.gcroot(&lt;ty&gt;** %ptrloc, &lt;ty2&gt;* %metadata)
Chris Lattner757528b0b2004-05-23 21:06:01 +00003928</pre>
3929
3930<h5>Overview:</h5>
3931
John Criswelldfe6a862004-12-10 15:51:16 +00003932<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
Chris Lattner757528b0b2004-05-23 21:06:01 +00003933the code generator, and allows some metadata to be associated with it.</p>
3934
3935<h5>Arguments:</h5>
3936
3937<p>The first argument specifies the address of a stack object that contains the
3938root pointer. The second pointer (which must be either a constant or a global
3939value address) contains the meta-data to be associated with the root.</p>
3940
3941<h5>Semantics:</h5>
3942
3943<p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3944location. At compile-time, the code generator generates information to allow
3945the runtime to find the pointer at GC safe points.
3946</p>
3947
3948</div>
3949
3950
3951<!-- _______________________________________________________________________ -->
3952<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00003953 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
Chris Lattner757528b0b2004-05-23 21:06:01 +00003954</div>
3955
3956<div class="doc_text">
3957
3958<h5>Syntax:</h5>
3959
3960<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00003961 declare i8 * @llvm.gcread(i8 * %ObjPtr, i8 ** %Ptr)
Chris Lattner757528b0b2004-05-23 21:06:01 +00003962</pre>
3963
3964<h5>Overview:</h5>
3965
3966<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
3967locations, allowing garbage collector implementations that require read
3968barriers.</p>
3969
3970<h5>Arguments:</h5>
3971
Chris Lattnerf9228072006-03-14 20:02:51 +00003972<p>The second argument is the address to read from, which should be an address
3973allocated from the garbage collector. The first object is a pointer to the
3974start of the referenced object, if needed by the language runtime (otherwise
3975null).</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00003976
3977<h5>Semantics:</h5>
3978
3979<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
3980instruction, but may be replaced with substantially more complex code by the
3981garbage collector runtime, as needed.</p>
3982
3983</div>
3984
3985
3986<!-- _______________________________________________________________________ -->
3987<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00003988 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
Chris Lattner757528b0b2004-05-23 21:06:01 +00003989</div>
3990
3991<div class="doc_text">
3992
3993<h5>Syntax:</h5>
3994
3995<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00003996 declare void @llvm.gcwrite(i8 * %P1, i8 * %Obj, i8 ** %P2)
Chris Lattner757528b0b2004-05-23 21:06:01 +00003997</pre>
3998
3999<h5>Overview:</h5>
4000
4001<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4002locations, allowing garbage collector implementations that require write
4003barriers (such as generational or reference counting collectors).</p>
4004
4005<h5>Arguments:</h5>
4006
Chris Lattnerf9228072006-03-14 20:02:51 +00004007<p>The first argument is the reference to store, the second is the start of the
4008object to store it to, and the third is the address of the field of Obj to
4009store to. If the runtime does not require a pointer to the object, Obj may be
4010null.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00004011
4012<h5>Semantics:</h5>
4013
4014<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4015instruction, but may be replaced with substantially more complex code by the
4016garbage collector runtime, as needed.</p>
4017
4018</div>
4019
4020
4021
4022<!-- ======================================================================= -->
4023<div class="doc_subsection">
Chris Lattner3649c3a2004-02-14 04:08:35 +00004024 <a name="int_codegen">Code Generator Intrinsics</a>
4025</div>
4026
4027<div class="doc_text">
4028<p>
4029These intrinsics are provided by LLVM to expose special features that may only
4030be implemented with code generator support.
4031</p>
4032
4033</div>
4034
4035<!-- _______________________________________________________________________ -->
4036<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004037 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
Chris Lattner3649c3a2004-02-14 04:08:35 +00004038</div>
4039
4040<div class="doc_text">
4041
4042<h5>Syntax:</h5>
4043<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004044 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
Chris Lattner3649c3a2004-02-14 04:08:35 +00004045</pre>
4046
4047<h5>Overview:</h5>
4048
4049<p>
Chris Lattnerc1fb4262006-10-15 20:05:59 +00004050The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4051target-specific value indicating the return address of the current function
4052or one of its callers.
Chris Lattner3649c3a2004-02-14 04:08:35 +00004053</p>
4054
4055<h5>Arguments:</h5>
4056
4057<p>
4058The argument to this intrinsic indicates which function to return the address
4059for. Zero indicates the calling function, one indicates its caller, etc. The
4060argument is <b>required</b> to be a constant integer value.
4061</p>
4062
4063<h5>Semantics:</h5>
4064
4065<p>
4066The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4067the return address of the specified call frame, or zero if it cannot be
4068identified. The value returned by this intrinsic is likely to be incorrect or 0
4069for arguments other than zero, so it should only be used for debugging purposes.
4070</p>
4071
4072<p>
4073Note that calling this intrinsic does not prevent function inlining or other
Chris Lattner2e6eb5f2005-03-07 20:30:51 +00004074aggressive transformations, so the value returned may not be that of the obvious
Chris Lattner3649c3a2004-02-14 04:08:35 +00004075source-language caller.
4076</p>
4077</div>
4078
4079
4080<!-- _______________________________________________________________________ -->
4081<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004082 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
Chris Lattner3649c3a2004-02-14 04:08:35 +00004083</div>
4084
4085<div class="doc_text">
4086
4087<h5>Syntax:</h5>
4088<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004089 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Chris Lattner3649c3a2004-02-14 04:08:35 +00004090</pre>
4091
4092<h5>Overview:</h5>
4093
4094<p>
Chris Lattnerc1fb4262006-10-15 20:05:59 +00004095The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4096target-specific frame pointer value for the specified stack frame.
Chris Lattner3649c3a2004-02-14 04:08:35 +00004097</p>
4098
4099<h5>Arguments:</h5>
4100
4101<p>
4102The argument to this intrinsic indicates which function to return the frame
4103pointer for. Zero indicates the calling function, one indicates its caller,
4104etc. The argument is <b>required</b> to be a constant integer value.
4105</p>
4106
4107<h5>Semantics:</h5>
4108
4109<p>
4110The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4111the frame address of the specified call frame, or zero if it cannot be
4112identified. The value returned by this intrinsic is likely to be incorrect or 0
4113for arguments other than zero, so it should only be used for debugging purposes.
4114</p>
4115
4116<p>
4117Note that calling this intrinsic does not prevent function inlining or other
Chris Lattner2e6eb5f2005-03-07 20:30:51 +00004118aggressive transformations, so the value returned may not be that of the obvious
Chris Lattner3649c3a2004-02-14 04:08:35 +00004119source-language caller.
4120</p>
4121</div>
4122
Chris Lattnerc8a2c222005-02-28 19:24:19 +00004123<!-- _______________________________________________________________________ -->
4124<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004125 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
Chris Lattner2f0f0012006-01-13 02:03:13 +00004126</div>
4127
4128<div class="doc_text">
4129
4130<h5>Syntax:</h5>
4131<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004132 declare i8 *@llvm.stacksave()
Chris Lattner2f0f0012006-01-13 02:03:13 +00004133</pre>
4134
4135<h5>Overview:</h5>
4136
4137<p>
4138The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
Reid Spencer96a5f022007-04-04 02:42:35 +00004139the function stack, for use with <a href="#int_stackrestore">
Chris Lattner2f0f0012006-01-13 02:03:13 +00004140<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4141features like scoped automatic variable sized arrays in C99.
4142</p>
4143
4144<h5>Semantics:</h5>
4145
4146<p>
4147This intrinsic returns a opaque pointer value that can be passed to <a
Reid Spencer96a5f022007-04-04 02:42:35 +00004148href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
Chris Lattner2f0f0012006-01-13 02:03:13 +00004149<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4150<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4151state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4152practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4153that were allocated after the <tt>llvm.stacksave</tt> was executed.
4154</p>
4155
4156</div>
4157
4158<!-- _______________________________________________________________________ -->
4159<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004160 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
Chris Lattner2f0f0012006-01-13 02:03:13 +00004161</div>
4162
4163<div class="doc_text">
4164
4165<h5>Syntax:</h5>
4166<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004167 declare void @llvm.stackrestore(i8 * %ptr)
Chris Lattner2f0f0012006-01-13 02:03:13 +00004168</pre>
4169
4170<h5>Overview:</h5>
4171
4172<p>
4173The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4174the function stack to the state it was in when the corresponding <a
Reid Spencer96a5f022007-04-04 02:42:35 +00004175href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
Chris Lattner2f0f0012006-01-13 02:03:13 +00004176useful for implementing language features like scoped automatic variable sized
4177arrays in C99.
4178</p>
4179
4180<h5>Semantics:</h5>
4181
4182<p>
Reid Spencer96a5f022007-04-04 02:42:35 +00004183See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
Chris Lattner2f0f0012006-01-13 02:03:13 +00004184</p>
4185
4186</div>
4187
4188
4189<!-- _______________________________________________________________________ -->
4190<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004191 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
Chris Lattnerc8a2c222005-02-28 19:24:19 +00004192</div>
4193
4194<div class="doc_text">
4195
4196<h5>Syntax:</h5>
4197<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004198 declare void @llvm.prefetch(i8 * &lt;address&gt;,
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004199 i32 &lt;rw&gt;, i32 &lt;locality&gt;)
Chris Lattnerc8a2c222005-02-28 19:24:19 +00004200</pre>
4201
4202<h5>Overview:</h5>
4203
4204
4205<p>
4206The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
John Criswell88190562005-05-16 16:17:45 +00004207a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4208no
4209effect on the behavior of the program but can change its performance
Chris Lattnerff851072005-02-28 19:47:14 +00004210characteristics.
Chris Lattnerc8a2c222005-02-28 19:24:19 +00004211</p>
4212
4213<h5>Arguments:</h5>
4214
4215<p>
4216<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4217determining if the fetch should be for a read (0) or write (1), and
4218<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
Chris Lattnerd3e641c2005-03-07 20:31:38 +00004219locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
Chris Lattnerc8a2c222005-02-28 19:24:19 +00004220<tt>locality</tt> arguments must be constant integers.
4221</p>
4222
4223<h5>Semantics:</h5>
4224
4225<p>
4226This intrinsic does not modify the behavior of the program. In particular,
4227prefetches cannot trap and do not produce a value. On targets that support this
4228intrinsic, the prefetch can provide hints to the processor cache for better
4229performance.
4230</p>
4231
4232</div>
4233
Andrew Lenharthb4427912005-03-28 20:05:49 +00004234<!-- _______________________________________________________________________ -->
4235<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004236 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
Andrew Lenharthb4427912005-03-28 20:05:49 +00004237</div>
4238
4239<div class="doc_text">
4240
4241<h5>Syntax:</h5>
4242<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004243 declare void @llvm.pcmarker( i32 &lt;id&gt; )
Andrew Lenharthb4427912005-03-28 20:05:49 +00004244</pre>
4245
4246<h5>Overview:</h5>
4247
4248
4249<p>
John Criswell88190562005-05-16 16:17:45 +00004250The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4251(PC) in a region of
Andrew Lenharthb4427912005-03-28 20:05:49 +00004252code to simulators and other tools. The method is target specific, but it is
4253expected that the marker will use exported symbols to transmit the PC of the marker.
Jeff Cohendc6bfea2005-11-11 02:15:27 +00004254The marker makes no guarantees that it will remain with any specific instruction
Chris Lattnere64d41d2005-11-15 06:07:55 +00004255after optimizations. It is possible that the presence of a marker will inhibit
Chris Lattnerb40261e2006-03-24 07:16:10 +00004256optimizations. The intended use is to be inserted after optimizations to allow
John Criswell88190562005-05-16 16:17:45 +00004257correlations of simulation runs.
Andrew Lenharthb4427912005-03-28 20:05:49 +00004258</p>
4259
4260<h5>Arguments:</h5>
4261
4262<p>
4263<tt>id</tt> is a numerical id identifying the marker.
4264</p>
4265
4266<h5>Semantics:</h5>
4267
4268<p>
4269This intrinsic does not modify the behavior of the program. Backends that do not
4270support this intrinisic may ignore it.
4271</p>
4272
4273</div>
4274
Andrew Lenharth01aa5632005-11-11 16:47:30 +00004275<!-- _______________________________________________________________________ -->
4276<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004277 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
Andrew Lenharth01aa5632005-11-11 16:47:30 +00004278</div>
4279
4280<div class="doc_text">
4281
4282<h5>Syntax:</h5>
4283<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004284 declare i64 @llvm.readcyclecounter( )
Andrew Lenharth01aa5632005-11-11 16:47:30 +00004285</pre>
4286
4287<h5>Overview:</h5>
4288
4289
4290<p>
4291The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4292counter register (or similar low latency, high accuracy clocks) on those targets
4293that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4294As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4295should only be used for small timings.
4296</p>
4297
4298<h5>Semantics:</h5>
4299
4300<p>
4301When directly supported, reading the cycle counter should not modify any memory.
4302Implementations are allowed to either return a application specific value or a
4303system wide value. On backends without support, this is lowered to a constant 0.
4304</p>
4305
4306</div>
4307
Chris Lattner3649c3a2004-02-14 04:08:35 +00004308<!-- ======================================================================= -->
4309<div class="doc_subsection">
Chris Lattnerfee11462004-02-12 17:01:32 +00004310 <a name="int_libc">Standard C Library Intrinsics</a>
4311</div>
4312
4313<div class="doc_text">
4314<p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00004315LLVM provides intrinsics for a few important standard C library functions.
4316These intrinsics allow source-language front-ends to pass information about the
4317alignment of the pointer arguments to the code generator, providing opportunity
4318for more efficient code generation.
Chris Lattnerfee11462004-02-12 17:01:32 +00004319</p>
4320
4321</div>
4322
4323<!-- _______________________________________________________________________ -->
4324<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004325 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
Chris Lattnerfee11462004-02-12 17:01:32 +00004326</div>
4327
4328<div class="doc_text">
4329
4330<h5>Syntax:</h5>
4331<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004332 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004333 i32 &lt;len&gt;, i32 &lt;align&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004334 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004335 i64 &lt;len&gt;, i32 &lt;align&gt;)
Chris Lattnerfee11462004-02-12 17:01:32 +00004336</pre>
4337
4338<h5>Overview:</h5>
4339
4340<p>
Chris Lattner0c8b2592006-03-03 00:07:20 +00004341The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
Chris Lattnerfee11462004-02-12 17:01:32 +00004342location to the destination location.
4343</p>
4344
4345<p>
Chris Lattner0c8b2592006-03-03 00:07:20 +00004346Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4347intrinsics do not return a value, and takes an extra alignment argument.
Chris Lattnerfee11462004-02-12 17:01:32 +00004348</p>
4349
4350<h5>Arguments:</h5>
4351
4352<p>
4353The first argument is a pointer to the destination, the second is a pointer to
Chris Lattner0c8b2592006-03-03 00:07:20 +00004354the source. The third argument is an integer argument
Chris Lattnerfee11462004-02-12 17:01:32 +00004355specifying the number of bytes to copy, and the fourth argument is the alignment
4356of the source and destination locations.
4357</p>
4358
Chris Lattner4c67c482004-02-12 21:18:15 +00004359<p>
4360If the call to this intrinisic has an alignment value that is not 0 or 1, then
Chris Lattner5316e5d2006-03-04 00:02:10 +00004361the caller guarantees that both the source and destination pointers are aligned
4362to that boundary.
Chris Lattner4c67c482004-02-12 21:18:15 +00004363</p>
4364
Chris Lattnerfee11462004-02-12 17:01:32 +00004365<h5>Semantics:</h5>
4366
4367<p>
Chris Lattner0c8b2592006-03-03 00:07:20 +00004368The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
Chris Lattnerfee11462004-02-12 17:01:32 +00004369location to the destination location, which are not allowed to overlap. It
4370copies "len" bytes of memory over. If the argument is known to be aligned to
4371some boundary, this can be specified as the fourth argument, otherwise it should
4372be set to 0 or 1.
4373</p>
4374</div>
4375
4376
Chris Lattnerf30152e2004-02-12 18:10:10 +00004377<!-- _______________________________________________________________________ -->
4378<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004379 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
Chris Lattnerf30152e2004-02-12 18:10:10 +00004380</div>
4381
4382<div class="doc_text">
4383
4384<h5>Syntax:</h5>
4385<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004386 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004387 i32 &lt;len&gt;, i32 &lt;align&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004388 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004389 i64 &lt;len&gt;, i32 &lt;align&gt;)
Chris Lattnerf30152e2004-02-12 18:10:10 +00004390</pre>
4391
4392<h5>Overview:</h5>
4393
4394<p>
Chris Lattner0c8b2592006-03-03 00:07:20 +00004395The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4396location to the destination location. It is similar to the
4397'<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
Chris Lattnerf30152e2004-02-12 18:10:10 +00004398</p>
4399
4400<p>
Chris Lattner0c8b2592006-03-03 00:07:20 +00004401Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4402intrinsics do not return a value, and takes an extra alignment argument.
Chris Lattnerf30152e2004-02-12 18:10:10 +00004403</p>
4404
4405<h5>Arguments:</h5>
4406
4407<p>
4408The first argument is a pointer to the destination, the second is a pointer to
Chris Lattner0c8b2592006-03-03 00:07:20 +00004409the source. The third argument is an integer argument
Chris Lattnerf30152e2004-02-12 18:10:10 +00004410specifying the number of bytes to copy, and the fourth argument is the alignment
4411of the source and destination locations.
4412</p>
4413
Chris Lattner4c67c482004-02-12 21:18:15 +00004414<p>
4415If the call to this intrinisic has an alignment value that is not 0 or 1, then
Chris Lattner5316e5d2006-03-04 00:02:10 +00004416the caller guarantees that the source and destination pointers are aligned to
4417that boundary.
Chris Lattner4c67c482004-02-12 21:18:15 +00004418</p>
4419
Chris Lattnerf30152e2004-02-12 18:10:10 +00004420<h5>Semantics:</h5>
4421
4422<p>
Chris Lattner0c8b2592006-03-03 00:07:20 +00004423The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
Chris Lattnerf30152e2004-02-12 18:10:10 +00004424location to the destination location, which may overlap. It
4425copies "len" bytes of memory over. If the argument is known to be aligned to
4426some boundary, this can be specified as the fourth argument, otherwise it should
4427be set to 0 or 1.
4428</p>
4429</div>
4430
Chris Lattner941515c2004-01-06 05:31:32 +00004431
Chris Lattner3649c3a2004-02-14 04:08:35 +00004432<!-- _______________________________________________________________________ -->
4433<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004434 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
Chris Lattner3649c3a2004-02-14 04:08:35 +00004435</div>
4436
4437<div class="doc_text">
4438
4439<h5>Syntax:</h5>
4440<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004441 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004442 i32 &lt;len&gt;, i32 &lt;align&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004443 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004444 i64 &lt;len&gt;, i32 &lt;align&gt;)
Chris Lattner3649c3a2004-02-14 04:08:35 +00004445</pre>
4446
4447<h5>Overview:</h5>
4448
4449<p>
Chris Lattner0c8b2592006-03-03 00:07:20 +00004450The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
Chris Lattner3649c3a2004-02-14 04:08:35 +00004451byte value.
4452</p>
4453
4454<p>
4455Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4456does not return a value, and takes an extra alignment argument.
4457</p>
4458
4459<h5>Arguments:</h5>
4460
4461<p>
4462The first argument is a pointer to the destination to fill, the second is the
Chris Lattner0c8b2592006-03-03 00:07:20 +00004463byte value to fill it with, the third argument is an integer
Chris Lattner3649c3a2004-02-14 04:08:35 +00004464argument specifying the number of bytes to fill, and the fourth argument is the
4465known alignment of destination location.
4466</p>
4467
4468<p>
4469If the call to this intrinisic has an alignment value that is not 0 or 1, then
Chris Lattner5316e5d2006-03-04 00:02:10 +00004470the caller guarantees that the destination pointer is aligned to that boundary.
Chris Lattner3649c3a2004-02-14 04:08:35 +00004471</p>
4472
4473<h5>Semantics:</h5>
4474
4475<p>
Chris Lattner0c8b2592006-03-03 00:07:20 +00004476The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4477the
Chris Lattner3649c3a2004-02-14 04:08:35 +00004478destination location. If the argument is known to be aligned to some boundary,
4479this can be specified as the fourth argument, otherwise it should be set to 0 or
44801.
4481</p>
4482</div>
4483
4484
Chris Lattner3b4f4372004-06-11 02:28:03 +00004485<!-- _______________________________________________________________________ -->
4486<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004487 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
Chris Lattner8a8f2e52005-07-21 01:29:16 +00004488</div>
4489
4490<div class="doc_text">
4491
4492<h5>Syntax:</h5>
4493<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004494 declare float @llvm.sqrt.f32(float %Val)
4495 declare double @llvm.sqrt.f64(double %Val)
Chris Lattner8a8f2e52005-07-21 01:29:16 +00004496</pre>
4497
4498<h5>Overview:</h5>
4499
4500<p>
Reid Spencerb4f9a6f2006-01-16 21:12:35 +00004501The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Chris Lattner8a8f2e52005-07-21 01:29:16 +00004502returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
4503<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4504negative numbers (which allows for better optimization).
4505</p>
4506
4507<h5>Arguments:</h5>
4508
4509<p>
4510The argument and return value are floating point numbers of the same type.
4511</p>
4512
4513<h5>Semantics:</h5>
4514
4515<p>
Dan Gohman33988db2007-07-16 14:37:41 +00004516This function returns the sqrt of the specified operand if it is a nonnegative
Chris Lattner8a8f2e52005-07-21 01:29:16 +00004517floating point number.
4518</p>
4519</div>
4520
Chris Lattner33b73f92006-09-08 06:34:02 +00004521<!-- _______________________________________________________________________ -->
4522<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004523 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
Chris Lattner33b73f92006-09-08 06:34:02 +00004524</div>
4525
4526<div class="doc_text">
4527
4528<h5>Syntax:</h5>
4529<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004530 declare float @llvm.powi.f32(float %Val, i32 %power)
4531 declare double @llvm.powi.f64(double %Val, i32 %power)
Chris Lattner33b73f92006-09-08 06:34:02 +00004532</pre>
4533
4534<h5>Overview:</h5>
4535
4536<p>
4537The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4538specified (positive or negative) power. The order of evaluation of
4539multiplications is not defined.
4540</p>
4541
4542<h5>Arguments:</h5>
4543
4544<p>
4545The second argument is an integer power, and the first is a value to raise to
4546that power.
4547</p>
4548
4549<h5>Semantics:</h5>
4550
4551<p>
4552This function returns the first value raised to the second power with an
4553unspecified sequence of rounding operations.</p>
4554</div>
4555
4556
Andrew Lenharth1d463522005-05-03 18:01:48 +00004557<!-- ======================================================================= -->
4558<div class="doc_subsection">
Nate Begeman0f223bb2006-01-13 23:26:38 +00004559 <a name="int_manip">Bit Manipulation Intrinsics</a>
Andrew Lenharth1d463522005-05-03 18:01:48 +00004560</div>
4561
4562<div class="doc_text">
4563<p>
Nate Begeman0f223bb2006-01-13 23:26:38 +00004564LLVM provides intrinsics for a few important bit manipulation operations.
Andrew Lenharth1d463522005-05-03 18:01:48 +00004565These allow efficient code generation for some algorithms.
4566</p>
4567
4568</div>
4569
4570<!-- _______________________________________________________________________ -->
4571<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004572 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
Nate Begeman0f223bb2006-01-13 23:26:38 +00004573</div>
4574
4575<div class="doc_text">
4576
4577<h5>Syntax:</h5>
Reid Spencer4eefaab2007-04-01 08:04:23 +00004578<p>This is an overloaded intrinsic function. You can use bswap on any integer
Chandler Carruth7132e002007-08-04 01:51:18 +00004579type that is an even number of bytes (i.e. BitWidth % 16 == 0).
Nate Begeman0f223bb2006-01-13 23:26:38 +00004580<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +00004581 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
4582 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
4583 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Nate Begeman0f223bb2006-01-13 23:26:38 +00004584</pre>
4585
4586<h5>Overview:</h5>
4587
4588<p>
Reid Spencerf361c4f2007-04-02 02:25:19 +00004589The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
Reid Spencer4eefaab2007-04-01 08:04:23 +00004590values with an even number of bytes (positive multiple of 16 bits). These are
4591useful for performing operations on data that is not in the target's native
4592byte order.
Nate Begeman0f223bb2006-01-13 23:26:38 +00004593</p>
4594
4595<h5>Semantics:</h5>
4596
4597<p>
Chandler Carruth7132e002007-08-04 01:51:18 +00004598The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004599and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4600intrinsic returns an i32 value that has the four bytes of the input i32
4601swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
Chandler Carruth7132e002007-08-04 01:51:18 +00004602i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
4603<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Reid Spencer4eefaab2007-04-01 08:04:23 +00004604additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
Nate Begeman0f223bb2006-01-13 23:26:38 +00004605</p>
4606
4607</div>
4608
4609<!-- _______________________________________________________________________ -->
4610<div class="doc_subsubsection">
Reid Spencerb4f9a6f2006-01-16 21:12:35 +00004611 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
Andrew Lenharth1d463522005-05-03 18:01:48 +00004612</div>
4613
4614<div class="doc_text">
4615
4616<h5>Syntax:</h5>
Reid Spencer4eefaab2007-04-01 08:04:23 +00004617<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4618width. Not all targets support all bit widths however.
Andrew Lenharth1d463522005-05-03 18:01:48 +00004619<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +00004620 declare i8 @llvm.ctpop.i8 (i8 &lt;src&gt;)
4621 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004622 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carruth7132e002007-08-04 01:51:18 +00004623 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
4624 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Andrew Lenharth1d463522005-05-03 18:01:48 +00004625</pre>
4626
4627<h5>Overview:</h5>
4628
4629<p>
Chris Lattner069b5bd2006-01-16 22:38:59 +00004630The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4631value.
Andrew Lenharth1d463522005-05-03 18:01:48 +00004632</p>
4633
4634<h5>Arguments:</h5>
4635
4636<p>
Chris Lattner573f64e2005-05-07 01:46:40 +00004637The only argument is the value to be counted. The argument may be of any
Reid Spencer3e628eb92007-01-04 16:43:23 +00004638integer type. The return type must match the argument type.
Andrew Lenharth1d463522005-05-03 18:01:48 +00004639</p>
4640
4641<h5>Semantics:</h5>
4642
4643<p>
4644The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4645</p>
4646</div>
4647
4648<!-- _______________________________________________________________________ -->
4649<div class="doc_subsubsection">
Chris Lattnerb748c672006-01-16 22:34:14 +00004650 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
Andrew Lenharth1d463522005-05-03 18:01:48 +00004651</div>
4652
4653<div class="doc_text">
4654
4655<h5>Syntax:</h5>
Reid Spencer4eefaab2007-04-01 08:04:23 +00004656<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4657integer bit width. Not all targets support all bit widths however.
Andrew Lenharth1d463522005-05-03 18:01:48 +00004658<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +00004659 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
4660 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004661 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carruth7132e002007-08-04 01:51:18 +00004662 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
4663 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Andrew Lenharth1d463522005-05-03 18:01:48 +00004664</pre>
4665
4666<h5>Overview:</h5>
4667
4668<p>
Reid Spencerb4f9a6f2006-01-16 21:12:35 +00004669The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4670leading zeros in a variable.
Andrew Lenharth1d463522005-05-03 18:01:48 +00004671</p>
4672
4673<h5>Arguments:</h5>
4674
4675<p>
Chris Lattner573f64e2005-05-07 01:46:40 +00004676The only argument is the value to be counted. The argument may be of any
Reid Spencer3e628eb92007-01-04 16:43:23 +00004677integer type. The return type must match the argument type.
Andrew Lenharth1d463522005-05-03 18:01:48 +00004678</p>
4679
4680<h5>Semantics:</h5>
4681
4682<p>
Chris Lattnerefa20fa2005-05-15 19:39:26 +00004683The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4684in a variable. If the src == 0 then the result is the size in bits of the type
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004685of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
Andrew Lenharth1d463522005-05-03 18:01:48 +00004686</p>
4687</div>
Chris Lattner3b4f4372004-06-11 02:28:03 +00004688
4689
Chris Lattnerefa20fa2005-05-15 19:39:26 +00004690
4691<!-- _______________________________________________________________________ -->
4692<div class="doc_subsubsection">
Chris Lattnerb748c672006-01-16 22:34:14 +00004693 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
Chris Lattnerefa20fa2005-05-15 19:39:26 +00004694</div>
4695
4696<div class="doc_text">
4697
4698<h5>Syntax:</h5>
Reid Spencer4eefaab2007-04-01 08:04:23 +00004699<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
4700integer bit width. Not all targets support all bit widths however.
Chris Lattnerefa20fa2005-05-15 19:39:26 +00004701<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +00004702 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
4703 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004704 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carruth7132e002007-08-04 01:51:18 +00004705 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
4706 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Chris Lattnerefa20fa2005-05-15 19:39:26 +00004707</pre>
4708
4709<h5>Overview:</h5>
4710
4711<p>
Reid Spencerb4f9a6f2006-01-16 21:12:35 +00004712The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4713trailing zeros.
Chris Lattnerefa20fa2005-05-15 19:39:26 +00004714</p>
4715
4716<h5>Arguments:</h5>
4717
4718<p>
4719The only argument is the value to be counted. The argument may be of any
Reid Spencer3e628eb92007-01-04 16:43:23 +00004720integer type. The return type must match the argument type.
Chris Lattnerefa20fa2005-05-15 19:39:26 +00004721</p>
4722
4723<h5>Semantics:</h5>
4724
4725<p>
4726The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4727in a variable. If the src == 0 then the result is the size in bits of the type
4728of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4729</p>
4730</div>
4731
Reid Spencer8a5799f2007-04-01 08:27:01 +00004732<!-- _______________________________________________________________________ -->
4733<div class="doc_subsubsection">
Reid Spencerea2945e2007-04-10 02:51:31 +00004734 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
Reid Spencer8bc7d952007-04-01 19:00:37 +00004735</div>
4736
4737<div class="doc_text">
4738
4739<h5>Syntax:</h5>
Reid Spencerea2945e2007-04-10 02:51:31 +00004740<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
Reid Spencer8bc7d952007-04-01 19:00:37 +00004741on any integer bit width.
4742<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +00004743 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
4744 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Reid Spencer8bc7d952007-04-01 19:00:37 +00004745</pre>
4746
4747<h5>Overview:</h5>
Reid Spencerea2945e2007-04-10 02:51:31 +00004748<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
Reid Spencer8bc7d952007-04-01 19:00:37 +00004749range of bits from an integer value and returns them in the same bit width as
4750the original value.</p>
4751
4752<h5>Arguments:</h5>
4753<p>The first argument, <tt>%val</tt> and the result may be integer types of
4754any bit width but they must have the same bit width. The second and third
Reid Spencer96a5f022007-04-04 02:42:35 +00004755arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
Reid Spencer8bc7d952007-04-01 19:00:37 +00004756
4757<h5>Semantics:</h5>
Reid Spencerea2945e2007-04-10 02:51:31 +00004758<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
Reid Spencer96a5f022007-04-04 02:42:35 +00004759of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
4760<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
4761operates in forward mode.</p>
4762<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
4763right by <tt>%loBit</tt> bits and then ANDing it with a mask with
Reid Spencer8bc7d952007-04-01 19:00:37 +00004764only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
4765<ol>
4766 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
4767 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
4768 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
4769 to determine the number of bits to retain.</li>
4770 <li>A mask of the retained bits is created by shifting a -1 value.</li>
4771 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
4772</ol>
Reid Spencer70845c02007-05-14 16:14:57 +00004773<p>In reverse mode, a similar computation is made except that the bits are
4774returned in the reverse order. So, for example, if <tt>X</tt> has the value
4775<tt>i16 0x0ACF (101011001111)</tt> and we apply
4776<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
4777<tt>i16 0x0026 (000000100110)</tt>.</p>
Reid Spencer8bc7d952007-04-01 19:00:37 +00004778</div>
4779
Reid Spencer5bf54c82007-04-11 23:23:49 +00004780<div class="doc_subsubsection">
4781 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
4782</div>
4783
4784<div class="doc_text">
4785
4786<h5>Syntax:</h5>
4787<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
4788on any integer bit width.
4789<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +00004790 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
4791 declare i29 @llvm.part.set.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
Reid Spencer5bf54c82007-04-11 23:23:49 +00004792</pre>
4793
4794<h5>Overview:</h5>
4795<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
4796of bits in an integer value with another integer value. It returns the integer
4797with the replaced bits.</p>
4798
4799<h5>Arguments:</h5>
4800<p>The first argument, <tt>%val</tt> and the result may be integer types of
4801any bit width but they must have the same bit width. <tt>%val</tt> is the value
4802whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
4803integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
4804type since they specify only a bit index.</p>
4805
4806<h5>Semantics:</h5>
4807<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
4808of operation: forwards and reverse. If <tt>%lo</tt> is greater than
4809<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
4810operates in forward mode.</p>
4811<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
4812truncating it down to the size of the replacement area or zero extending it
4813up to that size.</p>
4814<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
4815are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
4816in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
4817to the <tt>%hi</tt>th bit.
Reid Spencer146281c2007-05-14 16:50:20 +00004818<p>In reverse mode, a similar computation is made except that the bits are
4819reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
4820<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
Reid Spencer5bf54c82007-04-11 23:23:49 +00004821<h5>Examples:</h5>
4822<pre>
Reid Spencerc70afc32007-04-12 01:03:03 +00004823 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
Reid Spencer146281c2007-05-14 16:50:20 +00004824 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
4825 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
4826 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
Reid Spencerc70afc32007-04-12 01:03:03 +00004827 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
Reid Spencer7972c472007-04-11 23:49:50 +00004828</pre>
Reid Spencer5bf54c82007-04-11 23:23:49 +00004829</div>
4830
Chris Lattner941515c2004-01-06 05:31:32 +00004831<!-- ======================================================================= -->
4832<div class="doc_subsection">
4833 <a name="int_debugger">Debugger Intrinsics</a>
4834</div>
4835
4836<div class="doc_text">
4837<p>
4838The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
4839are described in the <a
4840href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
4841Debugging</a> document.
4842</p>
4843</div>
4844
4845
Jim Laskey2211f492007-03-14 19:31:19 +00004846<!-- ======================================================================= -->
4847<div class="doc_subsection">
4848 <a name="int_eh">Exception Handling Intrinsics</a>
4849</div>
4850
4851<div class="doc_text">
4852<p> The LLVM exception handling intrinsics (which all start with
4853<tt>llvm.eh.</tt> prefix), are described in the <a
4854href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
4855Handling</a> document. </p>
4856</div>
4857
Tanya Lattnercb1b9602007-06-15 20:50:54 +00004858<!-- ======================================================================= -->
4859<div class="doc_subsection">
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00004860 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
4861</div>
4862
4863<div class="doc_text">
4864<p>
4865 These intrinsic functions expand the "universal IR" of LLVM to represent
4866 hardware constructs for atomic operations and memory synchronization. This
4867 provides an interface to the hardware, not an interface to the programmer. It
4868 is aimed at a low enough level to allow any programming models or APIs which
4869 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
Chandler Carruthfd4184352007-07-20 20:14:52 +00004870 hardware behavior. Just as hardware provides a "universal IR" for source
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00004871 languages, it also provides a starting point for developing a "universal"
4872 atomic operation and synchronization IR.
4873</p>
4874<p>
4875 These do <em>not</em> form an API such as high-level threading libraries,
4876 software transaction memory systems, atomic primitives, and intrinsic
Reid Spencer5b2cb0f2007-07-20 19:59:11 +00004877 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00004878 application libraries. The hardware interface provided by LLVM should allow
4879 a clean implementation of all of these APIs and parallel programming models.
4880 No one model or paradigm should be selected above others unless the hardware
4881 itself ubiquitously does so.
4882</p>
4883</div>
4884
4885<!-- _______________________________________________________________________ -->
4886<div class="doc_subsubsection">
4887 <a name="int_lcs">'<tt>llvm.atomic.lcs.*</tt>' Intrinsic</a>
4888</div>
4889<div class="doc_text">
4890<h5>Syntax:</h5>
4891<p>
4892 This is an overloaded intrinsic. You can use <tt>llvm.atomic.lcs</tt> on any
Reid Spencer5b2cb0f2007-07-20 19:59:11 +00004893 integer bit width. Not all targets support all bit widths however.</p>
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00004894<pre>
4895declare i8 @llvm.atomic.lcs.i8.i8p.i8.i8( i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt; )
4896declare i16 @llvm.atomic.lcs.i16.i16p.i16.i16( i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt; )
4897declare i32 @llvm.atomic.lcs.i32.i32p.i32.i32( i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt; )
4898declare i64 @llvm.atomic.lcs.i64.i64p.i64.i64( i64* &lt;ptr&gt;, i64 &lt;cmp&gt;, i64 &lt;val&gt; )
4899</pre>
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00004900<h5>Overview:</h5>
4901<p>
Chandler Carruthfd4184352007-07-20 20:14:52 +00004902 This loads a value in memory and compares it to a given value. If they are
4903 equal, it stores a new value into the memory.
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00004904</p>
4905<h5>Arguments:</h5>
4906<p>
4907 The <tt>llvm.atomic.lcs</tt> intrinsic takes three arguments. The result as
4908 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
4909 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
4910 this integer type. While any bit width integer may be used, targets may only
4911 lower representations they support in hardware.
4912</p>
4913<h5>Semantics:</h5>
4914<p>
4915 This entire intrinsic must be executed atomically. It first loads the value
Chandler Carruthfd4184352007-07-20 20:14:52 +00004916 in memory pointed to by <tt>ptr</tt> and compares it with the value
4917 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
4918 loaded value is yielded in all cases. This provides the equivalent of an
4919 atomic compare-and-swap operation within the SSA framework.
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00004920</p>
4921<h5>Examples:</h5>
4922<pre>
4923%ptr = malloc i32
4924 store i32 4, %ptr
4925
4926%val1 = add i32 4, 4
4927%result1 = call i32 @llvm.atomic.lcs( i32* %ptr, i32 4, %val1 )
4928 <i>; yields {i32}:result1 = 4</i>
4929%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
4930%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
4931
4932%val2 = add i32 1, 1
4933%result2 = call i32 @llvm.atomic.lcs( i32* %ptr, i32 5, %val2 )
4934 <i>; yields {i32}:result2 = 8</i>
4935%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
4936%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
4937</pre>
4938</div>
4939
4940<!-- _______________________________________________________________________ -->
4941<div class="doc_subsubsection">
4942 <a name="int_ls">'<tt>llvm.atomic.ls.*</tt>' Intrinsic</a>
4943</div>
4944<div class="doc_text">
4945<h5>Syntax:</h5>
4946<p>
4947 This is an overloaded intrinsic. You can use <tt>llvm.atomic.ls</tt> on any
Reid Spencer5b2cb0f2007-07-20 19:59:11 +00004948 integer bit width. Not all targets support all bit widths however.</p>
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00004949<pre>
4950declare i8 @llvm.atomic.ls.i8.i8p.i8( i8* &lt;ptr&gt;, i8 &lt;val&gt; )
4951declare i16 @llvm.atomic.ls.i16.i16p.i16( i16* &lt;ptr&gt;, i16 &lt;val&gt; )
4952declare i32 @llvm.atomic.ls.i32.i32p.i32( i32* &lt;ptr&gt;, i32 &lt;val&gt; )
4953declare i64 @llvm.atomic.ls.i64.i64p.i64( i64* &lt;ptr&gt;, i64 &lt;val&gt; )
4954</pre>
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00004955<h5>Overview:</h5>
4956<p>
Chandler Carruthfd4184352007-07-20 20:14:52 +00004957 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
4958 the value from memory. It then stores the value in <tt>val</tt> in the memory
4959 at <tt>ptr</tt>.
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00004960</p>
4961<h5>Arguments:</h5>
4962<p>
4963 The <tt>llvm.atomic.ls</tt> intrinsic takes two arguments. Both the
4964 <tt>val</tt> argument and the result must be integers of the same bit width.
4965 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
4966 integer type. The targets may only lower integer representations they
4967 support.
4968</p>
4969<h5>Semantics:</h5>
4970<p>
4971 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
4972 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
4973 equivalent of an atomic swap operation within the SSA framework.
4974</p>
4975<h5>Examples:</h5>
4976<pre>
4977%ptr = malloc i32
4978 store i32 4, %ptr
4979
4980%val1 = add i32 4, 4
4981%result1 = call i32 @llvm.atomic.ls( i32* %ptr, i32 %val1 )
4982 <i>; yields {i32}:result1 = 4</i>
4983%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
4984%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
4985
4986%val2 = add i32 1, 1
4987%result2 = call i32 @llvm.atomic.ls( i32* %ptr, i32 %val2 )
4988 <i>; yields {i32}:result2 = 8</i>
4989%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
4990%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
4991</pre>
4992 </div>
4993
4994<!-- _______________________________________________________________________ -->
4995<div class="doc_subsubsection">
4996 <a name="int_las">'<tt>llvm.atomic.las.*</tt>' Intrinsic</a>
4997</div>
4998<div class="doc_text">
4999<h5>Syntax:</h5>
5000<p>
5001 This is an overloaded intrinsic. You can use <tt>llvm.atomic.las</tt> on any
Reid Spencer5b2cb0f2007-07-20 19:59:11 +00005002 integer bit width. Not all targets support all bit widths however.</p>
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00005003<pre>
5004declare i8 @llvm.atomic.las.i8.i8p.i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
5005declare i16 @llvm.atomic.las.i16.i16p.i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
5006declare i32 @llvm.atomic.las.i32.i32p.i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
5007declare i64 @llvm.atomic.las.i64.i64p.i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
5008</pre>
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00005009<h5>Overview:</h5>
5010<p>
Chandler Carruthfd4184352007-07-20 20:14:52 +00005011 This intrinsic adds <tt>delta</tt> to the value stored in memory at
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00005012 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
5013</p>
5014<h5>Arguments:</h5>
5015<p>
5016 The intrinsic takes two arguments, the first a pointer to an integer value
5017 and the second an integer value. The result is also an integer value. These
5018 integer types can have any bit width, but they must all have the same bit
5019 width. The targets may only lower integer representations they support.
5020</p>
5021<h5>Semantics:</h5>
5022<p>
5023 This intrinsic does a series of operations atomically. It first loads the
5024 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
5025 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
5026</p>
5027<h5>Examples:</h5>
5028<pre>
5029%ptr = malloc i32
5030 store i32 4, %ptr
5031%result1 = call i32 @llvm.atomic.las( i32* %ptr, i32 4 )
5032 <i>; yields {i32}:result1 = 4</i>
5033%result2 = call i32 @llvm.atomic.las( i32* %ptr, i32 2 )
5034 <i>; yields {i32}:result2 = 8</i>
5035%result3 = call i32 @llvm.atomic.las( i32* %ptr, i32 5 )
5036 <i>; yields {i32}:result3 = 10</i>
5037%memval = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
5038</pre>
5039</div>
5040
5041<!-- _______________________________________________________________________ -->
5042<div class="doc_subsubsection">
5043 <a name="int_lss">'<tt>llvm.atomic.lss.*</tt>' Intrinsic</a>
5044</div>
5045<div class="doc_text">
5046<h5>Syntax:</h5>
5047<p>
5048 This is an overloaded intrinsic. You can use <tt>llvm.atomic.lss</tt> on any
Reid Spencer5b2cb0f2007-07-20 19:59:11 +00005049 integer bit width. Not all targets support all bit widths however.</p>
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00005050<pre>
5051declare i8 @llvm.atomic.lss.i8.i8.i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
5052declare i16 @llvm.atomic.lss.i16.i16.i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
5053declare i32 @llvm.atomic.lss.i32.i32.i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
5054declare i64 @llvm.atomic.lss.i64.i64.i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
5055</pre>
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00005056<h5>Overview:</h5>
5057<p>
Chandler Carruthfd4184352007-07-20 20:14:52 +00005058 This intrinsic subtracts <tt>delta</tt> from the value stored in memory at
5059 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00005060</p>
5061<h5>Arguments:</h5>
5062<p>
5063 The intrinsic takes two arguments, the first a pointer to an integer value
5064 and the second an integer value. The result is also an integer value. These
5065 integer types can have any bit width, but they must all have the same bit
5066 width. The targets may only lower integer representations they support.
5067</p>
5068<h5>Semantics:</h5>
5069<p>
5070 This intrinsic does a series of operations atomically. It first loads the
5071 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>,
5072 stores the result to <tt>ptr</tt>. It yields the original value stored
5073 at <tt>ptr</tt>.
5074</p>
5075<h5>Examples:</h5>
5076<pre>
5077%ptr = malloc i32
5078 store i32 32, %ptr
5079%result1 = call i32 @llvm.atomic.lss( i32* %ptr, i32 4 )
5080 <i>; yields {i32}:result1 = 32</i>
5081%result2 = call i32 @llvm.atomic.lss( i32* %ptr, i32 2 )
5082 <i>; yields {i32}:result2 = 28</i>
5083%result3 = call i32 @llvm.atomic.lss( i32* %ptr, i32 5 )
5084 <i>; yields {i32}:result3 = 26</i>
5085%memval = load i32* %ptr <i>; yields {i32}:memval1 = 21</i>
5086</pre>
5087</div>
5088
5089<!-- _______________________________________________________________________ -->
5090<div class="doc_subsubsection">
5091 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
5092</div>
5093<div class="doc_text">
5094<h5>Syntax:</h5>
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00005095<pre>
5096declare void @llvm.memory.barrier( i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt; )
5097</pre>
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00005098<h5>Overview:</h5>
5099<p>
5100 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
5101 specific pairs of memory access types.
5102</p>
5103<h5>Arguments:</h5>
5104<p>
5105 The <tt>llvm.memory.barrier</tt> intrinsic requires four boolean arguments.
5106 Each argument enables a specific barrier as listed below.
Reid Spencer65a72e32007-07-20 20:03:33 +00005107</p>
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00005108 <ul>
5109 <li><tt>ll</tt>: load-load barrier</li>
5110 <li><tt>ls</tt>: load-store barrier</li>
5111 <li><tt>sl</tt>: store-load barrier</li>
5112 <li><tt>ss</tt>: store-store barrier</li>
5113 </ul>
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00005114<h5>Semantics:</h5>
5115<p>
5116 This intrinsic causes the system to enforce some ordering constraints upon
5117 the loads and stores of the program. This barrier does not indicate
5118 <em>when</em> any events will occur, it only enforces an <em>order</em> in
5119 which they occur. For any of the specified pairs of load and store operations
5120 (f.ex. load-load, or store-load), all of the first operations preceding the
5121 barrier will complete before any of the second operations succeeding the
5122 barrier begin. Specifically the semantics for each pairing is as follows:
Reid Spencer65a72e32007-07-20 20:03:33 +00005123</p>
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00005124 <ul>
5125 <li><tt>ll</tt>: All loads before the barrier must complete before any load
5126 after the barrier begins.</li>
5127 <li><tt>ls</tt>: All loads before the barrier must complete before any
5128 store after the barrier begins.</li>
5129 <li><tt>ss</tt>: All stores before the barrier must complete before any
5130 store after the barrier begins.</li>
5131 <li><tt>sl</tt>: All stores before the barrier must complete before any
5132 load after the barrier begins.</li>
5133 </ul>
Reid Spencer65a72e32007-07-20 20:03:33 +00005134<p>
Chandler Carruth6a5f6bf62007-07-20 19:34:37 +00005135 These semantics are applied with a logical "and" behavior when more than one
5136 is enabled in a single memory barrier intrinsic.
5137</p>
5138<h5>Example:</h5>
5139<pre>
5140%ptr = malloc i32
5141 store i32 4, %ptr
5142
5143%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
5144 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
5145 <i>; guarantee the above finishes</i>
5146 store i32 8, %ptr <i>; before this begins</i>
5147</pre>
5148</div>
5149
5150<!-- ======================================================================= -->
5151<div class="doc_subsection">
Duncan Sands644f9172007-07-27 12:58:54 +00005152 <a name="int_trampoline">Trampoline Intrinsics</a>
5153</div>
5154
5155<div class="doc_text">
5156<p>
5157 These intrinsics make it possible to excise one parameter, marked with
5158 the <tt>nest</tt> attribute, from a function. The result is a callable
5159 function pointer lacking the nest parameter - the caller does not need
5160 to provide a value for it. Instead, the value to use is stored in
5161 advance in a "trampoline", a block of memory usually allocated
5162 on the stack, which also contains code to splice the nest value into the
5163 argument list. This is used to implement the GCC nested function address
5164 extension.
5165</p>
5166<p>
5167 For example, if the function is
5168 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
5169 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:
5170<pre>
5171 %tramp1 = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5172 %tramp = getelementptr [10 x i8]* %tramp1, i32 0, i32 0
5173 call void @llvm.init.trampoline( i8* %tramp, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5174 %adj = call i8* @llvm.adjust.trampoline( i8* %tramp )
5175 %fp = bitcast i8* %adj to i32 (i32, i32)*
5176</pre>
5177 The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent to
5178 <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.
5179</p>
Duncan Sands644f9172007-07-27 12:58:54 +00005180</div>
5181
5182<!-- _______________________________________________________________________ -->
5183<div class="doc_subsubsection">
5184 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5185</div>
5186<div class="doc_text">
5187<h5>Syntax:</h5>
5188<pre>
5189declare void @llvm.init.trampoline(i8* &lt;tramp&gt;, i8* &lt;func&gt;, i8* &lt;nval&gt;)
5190</pre>
5191<h5>Overview:</h5>
5192<p>
5193 This initializes the memory pointed to by <tt>tramp</tt> as a trampoline.
5194</p>
5195<h5>Arguments:</h5>
5196<p>
5197 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5198 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5199 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sandsf2bcd372007-08-22 23:39:54 +00005200 intrinsic. Note that the size and the alignment are target-specific - LLVM
5201 currently provides no portable way of determining them, so a front-end that
5202 generates this intrinsic needs to have some target-specific knowledge.
5203 The <tt>func</tt> argument must hold a function bitcast to an <tt>i8*</tt>.
Duncan Sands644f9172007-07-27 12:58:54 +00005204</p>
5205<h5>Semantics:</h5>
5206<p>
5207 The block of memory pointed to by <tt>tramp</tt> is filled with target
5208 dependent code, turning it into a function.
5209 The new function's signature is the same as that of <tt>func</tt> with
5210 any arguments marked with the <tt>nest</tt> attribute removed. At most
5211 one such <tt>nest</tt> argument is allowed, and it must be of pointer
5212 type. Calling the new function is equivalent to calling <tt>func</tt>
5213 with the same argument list, but with <tt>nval</tt> used for the missing
5214 <tt>nest</tt> argument.
5215</p>
5216</div>
5217
5218<!-- _______________________________________________________________________ -->
5219<div class="doc_subsubsection">
5220 <a name="int_at">'<tt>llvm.adjust.trampoline</tt>' Intrinsic</a>
5221</div>
5222<div class="doc_text">
5223<h5>Syntax:</h5>
5224<pre>
5225declare i8* @llvm.adjust.trampoline(i8* &lt;tramp&gt;)
5226</pre>
5227<h5>Overview:</h5>
5228<p>
5229 This intrinsic returns a function pointer suitable for executing
5230 the trampoline code pointed to by <tt>tramp</tt>.
5231</p>
5232<h5>Arguments:</h5>
5233<p>
5234 The <tt>llvm.adjust.trampoline</tt> takes one argument, a pointer to a
5235 trampoline initialized by the
5236 <a href="#int_it">'<tt>llvm.init.trampoline</tt>' intrinsic</a>.
5237</p>
5238<h5>Semantics:</h5>
5239<p>
5240 A function pointer that can be used to execute the trampoline code in
5241 <tt>tramp</tt> is returned. The returned value should be bitcast to an
5242 <a href="#int_trampoline">appropriate function pointer type</a>
5243 before being called.
5244</p>
5245</div>
5246
5247<!-- ======================================================================= -->
5248<div class="doc_subsection">
Tanya Lattnercb1b9602007-06-15 20:50:54 +00005249 <a name="int_general">General Intrinsics</a>
5250</div>
5251
5252<div class="doc_text">
5253<p> This class of intrinsics is designed to be generic and has
5254no specific purpose. </p>
5255</div>
5256
5257<!-- _______________________________________________________________________ -->
5258<div class="doc_subsubsection">
5259 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
5260</div>
5261
5262<div class="doc_text">
5263
5264<h5>Syntax:</h5>
5265<pre>
Tanya Lattnerbed1d4d2007-06-18 23:42:37 +00005266 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
Tanya Lattnercb1b9602007-06-15 20:50:54 +00005267</pre>
5268
5269<h5>Overview:</h5>
5270
5271<p>
5272The '<tt>llvm.var.annotation</tt>' intrinsic
5273</p>
5274
5275<h5>Arguments:</h5>
5276
5277<p>
Tanya Lattnerbed1d4d2007-06-18 23:42:37 +00005278The first argument is a pointer to a value, the second is a pointer to a
5279global string, the third is a pointer to a global string which is the source
5280file name, and the last argument is the line number.
Tanya Lattnercb1b9602007-06-15 20:50:54 +00005281</p>
5282
5283<h5>Semantics:</h5>
5284
5285<p>
5286This intrinsic allows annotation of local variables with arbitrary strings.
5287This can be useful for special purpose optimizations that want to look for these
5288 annotations. These have no other defined use, they are ignored by code
5289 generation and optimization.
5290</div>
5291
Jim Laskey2211f492007-03-14 19:31:19 +00005292
Chris Lattner2f7c9632001-06-06 20:29:01 +00005293<!-- *********************************************************************** -->
Chris Lattner2f7c9632001-06-06 20:29:01 +00005294<hr>
Misha Brukmanc501f552004-03-01 17:47:27 +00005295<address>
5296 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
5297 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
5298 <a href="http://validator.w3.org/check/referer"><img
5299 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
5300
5301 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
Reid Spencerca058542006-03-14 05:39:39 +00005302 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
Misha Brukmanc501f552004-03-01 17:47:27 +00005303 Last modified: $Date$
5304</address>
Misha Brukman76307852003-11-08 01:05:38 +00005305</body>
5306</html>