blob: 225173e3d301aed87ce0b455fccd162c5f6f3cfa [file] [log] [blame]
Misha Brukmandaa4cb02004-03-01 17:47:27 +00001<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
Misha Brukman9d0919f2003-11-08 01:05:38 +00003<html>
4<head>
5 <title>LLVM Assembly Language Reference Manual</title>
Reid Spencer3921c742004-08-26 20:44:00 +00006 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
Eric Christopher6c7e8a02009-12-05 02:46:03 +00008 <meta name="description"
Reid Spencer3921c742004-08-26 20:44:00 +00009 content="LLVM Assembly Language Reference Manual.">
Misha Brukman9d0919f2003-11-08 01:05:38 +000010 <link rel="stylesheet" href="llvm.css" type="text/css">
11</head>
Chris Lattnerd7923912004-05-23 21:06:01 +000012
Misha Brukman9d0919f2003-11-08 01:05:38 +000013<body>
Chris Lattnerd7923912004-05-23 21:06:01 +000014
NAKAMURA Takumi05d02652011-04-18 23:59:50 +000015<h1>LLVM Language Reference Manual</h1>
Chris Lattner00950542001-06-06 20:29:01 +000016<ol>
Misha Brukman9d0919f2003-11-08 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 Lattnerfa730212004-12-09 16:11:40 +000020 <li><a href="#highlevel">High Level Structure</a>
21 <ol>
22 <li><a href="#modulestructure">Module Structure</a></li>
Bill Wendling3d10a5a2009-07-20 01:03:30 +000023 <li><a href="#linkage">Linkage Types</a>
24 <ol>
Bill Wendling987e7eb2009-07-20 02:41:50 +000025 <li><a href="#linkage_private">'<tt>private</tt>' Linkage</a></li>
26 <li><a href="#linkage_linker_private">'<tt>linker_private</tt>' Linkage</a></li>
Bill Wendling5e721d72010-07-01 21:55:59 +000027 <li><a href="#linkage_linker_private_weak">'<tt>linker_private_weak</tt>' Linkage</a></li>
Bill Wendling55ae5152010-08-20 22:05:50 +000028 <li><a href="#linkage_linker_private_weak_def_auto">'<tt>linker_private_weak_def_auto</tt>' Linkage</a></li>
Bill Wendling987e7eb2009-07-20 02:41:50 +000029 <li><a href="#linkage_internal">'<tt>internal</tt>' Linkage</a></li>
30 <li><a href="#linkage_available_externally">'<tt>available_externally</tt>' Linkage</a></li>
31 <li><a href="#linkage_linkonce">'<tt>linkonce</tt>' Linkage</a></li>
32 <li><a href="#linkage_common">'<tt>common</tt>' Linkage</a></li>
33 <li><a href="#linkage_weak">'<tt>weak</tt>' Linkage</a></li>
34 <li><a href="#linkage_appending">'<tt>appending</tt>' Linkage</a></li>
35 <li><a href="#linkage_externweak">'<tt>extern_weak</tt>' Linkage</a></li>
Chris Lattner5a2d8752009-10-10 18:26:06 +000036 <li><a href="#linkage_linkonce_odr">'<tt>linkonce_odr</tt>' Linkage</a></li>
Bill Wendling987e7eb2009-07-20 02:41:50 +000037 <li><a href="#linkage_weak">'<tt>weak_odr</tt>' Linkage</a></li>
38 <li><a href="#linkage_external">'<tt>externally visible</tt>' Linkage</a></li>
39 <li><a href="#linkage_dllimport">'<tt>dllimport</tt>' Linkage</a></li>
40 <li><a href="#linkage_dllexport">'<tt>dllexport</tt>' Linkage</a></li>
Bill Wendling3d10a5a2009-07-20 01:03:30 +000041 </ol>
42 </li>
Chris Lattnerbad10ee2005-05-06 22:57:40 +000043 <li><a href="#callingconv">Calling Conventions</a></li>
Chris Lattnere7886e42009-01-11 20:53:49 +000044 <li><a href="#namedtypes">Named Types</a></li>
Chris Lattnerfa730212004-12-09 16:11:40 +000045 <li><a href="#globalvars">Global Variables</a></li>
Chris Lattner4e9aba72006-01-23 23:23:47 +000046 <li><a href="#functionstructure">Functions</a></li>
Dan Gohman0e451ce2008-10-14 16:51:45 +000047 <li><a href="#aliasstructure">Aliases</a></li>
Devang Patelcd1fd252010-01-11 19:35:55 +000048 <li><a href="#namedmetadatastructure">Named Metadata</a></li>
Reid Spencerca86e162006-12-31 07:07:53 +000049 <li><a href="#paramattrs">Parameter Attributes</a></li>
Devang Patel2c9c3e72008-09-26 23:51:19 +000050 <li><a href="#fnattrs">Function Attributes</a></li>
Gordon Henriksen80a75bf2007-12-10 03:18:06 +000051 <li><a href="#gc">Garbage Collector Names</a></li>
Chris Lattner4e9aba72006-01-23 23:23:47 +000052 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
Reid Spencerde151942007-02-19 23:54:10 +000053 <li><a href="#datalayout">Data Layout</a></li>
Dan Gohman556ca272009-07-27 18:07:55 +000054 <li><a href="#pointeraliasing">Pointer Aliasing Rules</a></li>
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +000055 <li><a href="#volatile">Volatile Memory Accesses</a></li>
Eli Friedman5b60e1b2011-07-20 21:35:53 +000056 <li><a href="#memmodel">Memory Model for Concurrent Operations</a></li>
Chris Lattnerfa730212004-12-09 16:11:40 +000057 </ol>
58 </li>
Chris Lattner00950542001-06-06 20:29:01 +000059 <li><a href="#typesystem">Type System</a>
60 <ol>
Chris Lattner4f69f462008-01-04 04:32:38 +000061 <li><a href="#t_classifications">Type Classifications</a></li>
Eric Christopher6c7e8a02009-12-05 02:46:03 +000062 <li><a href="#t_primitive">Primitive Types</a>
Chris Lattner261efe92003-11-25 01:02:51 +000063 <ol>
Nick Lewyckyec38da42009-09-27 00:45:11 +000064 <li><a href="#t_integer">Integer Type</a></li>
Chris Lattner4f69f462008-01-04 04:32:38 +000065 <li><a href="#t_floating">Floating Point Types</a></li>
Dale Johannesen21fe99b2010-10-01 00:48:59 +000066 <li><a href="#t_x86mmx">X86mmx Type</a></li>
Chris Lattner4f69f462008-01-04 04:32:38 +000067 <li><a href="#t_void">Void Type</a></li>
68 <li><a href="#t_label">Label Type</a></li>
Nick Lewycky7a0370f2009-05-30 05:06:04 +000069 <li><a href="#t_metadata">Metadata Type</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +000070 </ol>
71 </li>
Chris Lattner00950542001-06-06 20:29:01 +000072 <li><a href="#t_derived">Derived Types</a>
73 <ol>
Chris Lattnerfdfeb692010-02-12 20:49:41 +000074 <li><a href="#t_aggregate">Aggregate Types</a>
75 <ol>
76 <li><a href="#t_array">Array Type</a></li>
77 <li><a href="#t_struct">Structure Type</a></li>
Chris Lattner1afcace2011-07-09 17:41:24 +000078 <li><a href="#t_opaque">Opaque Type</a></li>
Chris Lattnerfdfeb692010-02-12 20:49:41 +000079 <li><a href="#t_vector">Vector Type</a></li>
80 </ol>
81 </li>
Misha Brukman9d0919f2003-11-08 01:05:38 +000082 <li><a href="#t_function">Function Type</a></li>
83 <li><a href="#t_pointer">Pointer Type</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +000084 </ol>
85 </li>
86 </ol>
87 </li>
Chris Lattnerfa730212004-12-09 16:11:40 +000088 <li><a href="#constants">Constants</a>
Chris Lattnerc3f59762004-12-09 17:30:23 +000089 <ol>
Dan Gohman0e451ce2008-10-14 16:51:45 +000090 <li><a href="#simpleconstants">Simple Constants</a></li>
Chris Lattner70882792009-02-28 18:32:25 +000091 <li><a href="#complexconstants">Complex Constants</a></li>
Dan Gohman0e451ce2008-10-14 16:51:45 +000092 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
93 <li><a href="#undefvalues">Undefined Values</a></li>
Dan Gohmanfff6c532010-04-22 23:14:21 +000094 <li><a href="#trapvalues">Trap Values</a></li>
Chris Lattnerf9d078e2009-10-27 21:19:13 +000095 <li><a href="#blockaddress">Addresses of Basic Blocks</a></li>
Dan Gohman0e451ce2008-10-14 16:51:45 +000096 <li><a href="#constantexprs">Constant Expressions</a></li>
Chris Lattnerc3f59762004-12-09 17:30:23 +000097 </ol>
Chris Lattner261efe92003-11-25 01:02:51 +000098 </li>
Chris Lattnere87d6532006-01-25 23:47:57 +000099 <li><a href="#othervalues">Other Values</a>
100 <ol>
Dan Gohman0e451ce2008-10-14 16:51:45 +0000101 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
Devang Patelcd1fd252010-01-11 19:35:55 +0000102 <li><a href="#metadata">Metadata Nodes and Metadata Strings</a></li>
Chris Lattnere87d6532006-01-25 23:47:57 +0000103 </ol>
104 </li>
Chris Lattner857755c2009-07-20 05:55:19 +0000105 <li><a href="#intrinsic_globals">Intrinsic Global Variables</a>
106 <ol>
107 <li><a href="#intg_used">The '<tt>llvm.used</tt>' Global Variable</a></li>
Chris Lattner401e10c2009-07-20 06:14:25 +0000108 <li><a href="#intg_compiler_used">The '<tt>llvm.compiler.used</tt>'
109 Global Variable</a></li>
Chris Lattner857755c2009-07-20 05:55:19 +0000110 <li><a href="#intg_global_ctors">The '<tt>llvm.global_ctors</tt>'
111 Global Variable</a></li>
112 <li><a href="#intg_global_dtors">The '<tt>llvm.global_dtors</tt>'
113 Global Variable</a></li>
114 </ol>
115 </li>
Chris Lattner00950542001-06-06 20:29:01 +0000116 <li><a href="#instref">Instruction Reference</a>
117 <ol>
118 <li><a href="#terminators">Terminator Instructions</a>
119 <ol>
Chris Lattner261efe92003-11-25 01:02:51 +0000120 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
121 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000122 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
Chris Lattnerab21db72009-10-28 00:19:10 +0000123 <li><a href="#i_indirectbr">'<tt>indirectbr</tt>' Instruction</a></li>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000124 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000125 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
Chris Lattner35eca582004-10-16 18:04:13 +0000126 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000127 </ol>
128 </li>
Chris Lattner00950542001-06-06 20:29:01 +0000129 <li><a href="#binaryops">Binary Operations</a>
130 <ol>
Chris Lattner261efe92003-11-25 01:02:51 +0000131 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
Dan Gohmanae3a0be2009-06-04 22:49:04 +0000132 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000133 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
Dan Gohmanae3a0be2009-06-04 22:49:04 +0000134 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000135 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
Dan Gohmanae3a0be2009-06-04 22:49:04 +0000136 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
Reid Spencer1628cec2006-10-26 06:15:43 +0000137 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
138 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
139 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
Reid Spencer0a783f72006-11-02 01:53:59 +0000140 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
141 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
142 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000143 </ol>
144 </li>
Chris Lattner00950542001-06-06 20:29:01 +0000145 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
146 <ol>
Reid Spencer8e11bf82007-02-02 13:57:07 +0000147 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
148 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
149 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000150 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000151 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000152 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000153 </ol>
154 </li>
Chris Lattner3df241e2006-04-08 23:07:04 +0000155 <li><a href="#vectorops">Vector Operations</a>
156 <ol>
157 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
158 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
159 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
Chris Lattner3df241e2006-04-08 23:07:04 +0000160 </ol>
161 </li>
Dan Gohmana334d5f2008-05-12 23:51:09 +0000162 <li><a href="#aggregateops">Aggregate Operations</a>
163 <ol>
164 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
165 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
166 </ol>
167 </li>
Chris Lattner884a9702006-08-15 00:45:58 +0000168 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
Chris Lattner00950542001-06-06 20:29:01 +0000169 <ol>
Chris Lattner261efe92003-11-25 01:02:51 +0000170 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
Robert Bocchino7b81c752006-02-17 21:18:08 +0000171 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
172 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
173 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000174 </ol>
175 </li>
Reid Spencer2fd21e62006-11-08 01:18:52 +0000176 <li><a href="#convertops">Conversion Operations</a>
Reid Spencer9dee3ac2006-11-08 01:11:31 +0000177 <ol>
178 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
179 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
180 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
181 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
182 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
Reid Spencerd4448792006-11-09 23:03:26 +0000183 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
184 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
185 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
186 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
Reid Spencer72679252006-11-11 21:00:47 +0000187 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
188 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
Reid Spencer5c0ef472006-11-11 23:08:07 +0000189 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
Reid Spencer9dee3ac2006-11-08 01:11:31 +0000190 </ol>
Dan Gohman0e451ce2008-10-14 16:51:45 +0000191 </li>
Chris Lattner00950542001-06-06 20:29:01 +0000192 <li><a href="#otherops">Other Operations</a>
193 <ol>
Reid Spencerf3a70a62006-11-18 21:50:54 +0000194 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
195 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000196 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
Chris Lattnercc37aae2004-03-12 05:50:16 +0000197 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000198 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
Chris Lattnerfb6977d2006-01-13 23:26:01 +0000199 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
Chris Lattner00950542001-06-06 20:29:01 +0000200 </ol>
Chris Lattner261efe92003-11-25 01:02:51 +0000201 </li>
Chris Lattner00950542001-06-06 20:29:01 +0000202 </ol>
Chris Lattner261efe92003-11-25 01:02:51 +0000203 </li>
Chris Lattnerd9ad5b32003-05-08 04:57:36 +0000204 <li><a href="#intrinsics">Intrinsic Functions</a>
Chris Lattnerd9ad5b32003-05-08 04:57:36 +0000205 <ol>
Chris Lattner261efe92003-11-25 01:02:51 +0000206 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
207 <ol>
Reid Spencera3e435f2007-04-04 02:42:35 +0000208 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
209 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
210 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000211 </ol>
212 </li>
Chris Lattnerd7923912004-05-23 21:06:01 +0000213 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
214 <ol>
Reid Spencera3e435f2007-04-04 02:42:35 +0000215 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
216 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
217 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
Chris Lattnerd7923912004-05-23 21:06:01 +0000218 </ol>
219 </li>
Chris Lattner10610642004-02-14 04:08:35 +0000220 <li><a href="#int_codegen">Code Generator Intrinsics</a>
221 <ol>
Reid Spencera3e435f2007-04-04 02:42:35 +0000222 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
223 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
224 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
225 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
226 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
227 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
Dan Gohman31f1af12010-05-26 21:56:15 +0000228 <li><a href="#int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
John Criswell7123e272004-04-09 16:43:20 +0000229 </ol>
230 </li>
Chris Lattner33aec9e2004-02-12 17:01:32 +0000231 <li><a href="#int_libc">Standard C Library Intrinsics</a>
232 <ol>
Reid Spencera3e435f2007-04-04 02:42:35 +0000233 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
234 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
235 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
236 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
237 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohman91c284c2007-10-15 20:30:11 +0000238 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
239 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
240 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Dan Gohman08b280b2011-05-27 00:36:31 +0000241 <li><a href="#int_exp">'<tt>llvm.exp.*</tt>' Intrinsic</a></li>
242 <li><a href="#int_log">'<tt>llvm.log.*</tt>' Intrinsic</a></li>
Cameron Zwarich33390842011-07-08 21:39:21 +0000243 <li><a href="#int_fma">'<tt>llvm.fma.*</tt>' Intrinsic</a></li>
Chris Lattner33aec9e2004-02-12 17:01:32 +0000244 </ol>
245 </li>
Nate Begeman7e36c472006-01-13 23:26:38 +0000246 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
Andrew Lenharthec370fd2005-05-03 18:01:48 +0000247 <ol>
Reid Spencera3e435f2007-04-04 02:42:35 +0000248 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
Chris Lattner8a886be2006-01-16 22:34:14 +0000249 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
250 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
251 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
Andrew Lenharthec370fd2005-05-03 18:01:48 +0000252 </ol>
253 </li>
Bill Wendlingac1df8e2009-02-08 01:40:31 +0000254 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
255 <ol>
Bill Wendlingda01af72009-02-08 04:04:40 +0000256 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
257 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
258 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
259 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
260 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
Bill Wendling41b485c2009-02-08 23:00:09 +0000261 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
Bill Wendlingac1df8e2009-02-08 01:40:31 +0000262 </ol>
263 </li>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +0000264 <li><a href="#int_fp16">Half Precision Floating Point Intrinsics</a>
265 <ol>
Chris Lattner82c3dc62010-03-14 23:03:31 +0000266 <li><a href="#int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a></li>
267 <li><a href="#int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a></li>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +0000268 </ol>
269 </li>
Chris Lattnerd7923912004-05-23 21:06:01 +0000270 <li><a href="#int_debugger">Debugger intrinsics</a></li>
Jim Laskeydd4ef1b2007-03-14 19:31:19 +0000271 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sandsf7331b32007-09-11 14:10:23 +0000272 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands36397f52007-07-27 12:58:54 +0000273 <ol>
274 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands36397f52007-07-27 12:58:54 +0000275 </ol>
276 </li>
Bill Wendling3c44f5b2008-11-18 22:10:53 +0000277 <li><a href="#int_atomics">Atomic intrinsics</a>
278 <ol>
279 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
280 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
281 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
282 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
283 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
284 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
285 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
286 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
287 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
288 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
289 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
290 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
291 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
292 </ol>
293 </li>
Nick Lewyckycc271862009-10-13 07:03:23 +0000294 <li><a href="#int_memorymarkers">Memory Use Markers</a>
295 <ol>
296 <li><a href="#int_lifetime_start"><tt>llvm.lifetime.start</tt></a></li>
297 <li><a href="#int_lifetime_end"><tt>llvm.lifetime.end</tt></a></li>
298 <li><a href="#int_invariant_start"><tt>llvm.invariant.start</tt></a></li>
299 <li><a href="#int_invariant_end"><tt>llvm.invariant.end</tt></a></li>
300 </ol>
301 </li>
Reid Spencer20677642007-07-20 19:59:11 +0000302 <li><a href="#int_general">General intrinsics</a>
Tanya Lattner6d806e92007-06-15 20:50:54 +0000303 <ol>
Reid Spencer20677642007-07-20 19:59:11 +0000304 <li><a href="#int_var_annotation">
Bill Wendling69e4adb2008-11-19 05:56:17 +0000305 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Tanya Lattnerb6367882007-09-21 22:59:12 +0000306 <li><a href="#int_annotation">
Bill Wendling69e4adb2008-11-19 05:56:17 +0000307 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Anton Korobeynikov4cb86182008-01-15 22:31:34 +0000308 <li><a href="#int_trap">
Bill Wendling69e4adb2008-11-19 05:56:17 +0000309 '<tt>llvm.trap</tt>' Intrinsic</a></li>
310 <li><a href="#int_stackprotector">
311 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
Eric Christopher0e671492009-11-30 08:03:53 +0000312 <li><a href="#int_objectsize">
313 '<tt>llvm.objectsize</tt>' Intrinsic</a></li>
Tanya Lattnerb6367882007-09-21 22:59:12 +0000314 </ol>
Tanya Lattner6d806e92007-06-15 20:50:54 +0000315 </li>
Chris Lattner261efe92003-11-25 01:02:51 +0000316 </ol>
317 </li>
Chris Lattner00950542001-06-06 20:29:01 +0000318</ol>
Chris Lattnerd7923912004-05-23 21:06:01 +0000319
320<div class="doc_author">
321 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
322 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000323</div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000324
Chris Lattner00950542001-06-06 20:29:01 +0000325<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000326<h2><a name="abstract">Abstract</a></h2>
Chris Lattner261efe92003-11-25 01:02:51 +0000327<!-- *********************************************************************** -->
Chris Lattnerd7923912004-05-23 21:06:01 +0000328
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000329<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000330
331<p>This document is a reference manual for the LLVM assembly language. LLVM is
332 a Static Single Assignment (SSA) based representation that provides type
333 safety, low-level operations, flexibility, and the capability of representing
334 'all' high-level languages cleanly. It is the common code representation
335 used throughout all phases of the LLVM compilation strategy.</p>
336
Misha Brukman9d0919f2003-11-08 01:05:38 +0000337</div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000338
Chris Lattner00950542001-06-06 20:29:01 +0000339<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000340<h2><a name="introduction">Introduction</a></h2>
Chris Lattner261efe92003-11-25 01:02:51 +0000341<!-- *********************************************************************** -->
Chris Lattnerd7923912004-05-23 21:06:01 +0000342
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000343<div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000344
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000345<p>The LLVM code representation is designed to be used in three different forms:
346 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
347 for fast loading by a Just-In-Time compiler), and as a human readable
348 assembly language representation. This allows LLVM to provide a powerful
349 intermediate representation for efficient compiler transformations and
350 analysis, while providing a natural means to debug and visualize the
351 transformations. The three different forms of LLVM are all equivalent. This
352 document describes the human readable representation and notation.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +0000353
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000354<p>The LLVM representation aims to be light-weight and low-level while being
355 expressive, typed, and extensible at the same time. It aims to be a
356 "universal IR" of sorts, by being at a low enough level that high-level ideas
357 may be cleanly mapped to it (similar to how microprocessors are "universal
358 IR's", allowing many source languages to be mapped to them). By providing
359 type information, LLVM can be used as the target of optimizations: for
360 example, through pointer analysis, it can be proven that a C automatic
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000361 variable is never accessed outside of the current function, allowing it to
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000362 be promoted to a simple SSA value instead of a memory location.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +0000363
Chris Lattner00950542001-06-06 20:29:01 +0000364<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000365<h4>
366 <a name="wellformed">Well-Formedness</a>
367</h4>
Chris Lattnerd7923912004-05-23 21:06:01 +0000368
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000369<div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000370
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000371<p>It is important to note that this document describes 'well formed' LLVM
372 assembly language. There is a difference between what the parser accepts and
373 what is considered 'well formed'. For example, the following instruction is
374 syntactically okay, but not well formed:</p>
Chris Lattnerd7923912004-05-23 21:06:01 +0000375
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000376<pre class="doc_code">
Bill Wendling2f7a8b02007-05-29 09:04:49 +0000377%x = <a href="#i_add">add</a> i32 1, %x
Chris Lattnerd7923912004-05-23 21:06:01 +0000378</pre>
379
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000380<p>because the definition of <tt>%x</tt> does not dominate all of its uses. The
381 LLVM infrastructure provides a verification pass that may be used to verify
382 that an LLVM module is well formed. This pass is automatically run by the
383 parser after parsing input assembly and by the optimizer before it outputs
384 bitcode. The violations pointed out by the verifier pass indicate bugs in
385 transformation passes or input to the parser.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000386
Bill Wendling2f7a8b02007-05-29 09:04:49 +0000387</div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000388
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000389</div>
390
Chris Lattnercc689392007-10-03 17:34:29 +0000391<!-- Describe the typesetting conventions here. -->
Chris Lattnerd7923912004-05-23 21:06:01 +0000392
Chris Lattner00950542001-06-06 20:29:01 +0000393<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000394<h2><a name="identifiers">Identifiers</a></h2>
Chris Lattner00950542001-06-06 20:29:01 +0000395<!-- *********************************************************************** -->
Chris Lattnerd7923912004-05-23 21:06:01 +0000396
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000397<div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000398
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000399<p>LLVM identifiers come in two basic types: global and local. Global
400 identifiers (functions, global variables) begin with the <tt>'@'</tt>
401 character. Local identifiers (register names, types) begin with
402 the <tt>'%'</tt> character. Additionally, there are three different formats
403 for identifiers, for different purposes:</p>
Chris Lattnerd7923912004-05-23 21:06:01 +0000404
Chris Lattner00950542001-06-06 20:29:01 +0000405<ol>
Reid Spencer2c452282007-08-07 14:34:28 +0000406 <li>Named values are represented as a string of characters with their prefix.
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000407 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
408 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
409 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
410 other characters in their names can be surrounded with quotes. Special
411 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
412 ASCII code for the character in hexadecimal. In this way, any character
413 can be used in a name value, even quotes themselves.</li>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000414
Reid Spencer2c452282007-08-07 14:34:28 +0000415 <li>Unnamed values are represented as an unsigned numeric value with their
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000416 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000417
Reid Spencercc16dc32004-12-09 18:02:53 +0000418 <li>Constants, which are described in a <a href="#constants">section about
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000419 constants</a>, below.</li>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000420</ol>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000421
Reid Spencer2c452282007-08-07 14:34:28 +0000422<p>LLVM requires that values start with a prefix for two reasons: Compilers
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000423 don't need to worry about name clashes with reserved words, and the set of
424 reserved words may be expanded in the future without penalty. Additionally,
425 unnamed identifiers allow a compiler to quickly come up with a temporary
426 variable without having to avoid symbol table conflicts.</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000427
Chris Lattner261efe92003-11-25 01:02:51 +0000428<p>Reserved words in LLVM are very similar to reserved words in other
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000429 languages. There are keywords for different opcodes
430 ('<tt><a href="#i_add">add</a></tt>',
431 '<tt><a href="#i_bitcast">bitcast</a></tt>',
432 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
433 ('<tt><a href="#t_void">void</a></tt>',
434 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
435 reserved words cannot conflict with variable names, because none of them
436 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000437
438<p>Here is an example of LLVM code to multiply the integer variable
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000439 '<tt>%X</tt>' by 8:</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000440
Misha Brukman9d0919f2003-11-08 01:05:38 +0000441<p>The easy way:</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000442
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000443<pre class="doc_code">
Bill Wendling2f7a8b02007-05-29 09:04:49 +0000444%result = <a href="#i_mul">mul</a> i32 %X, 8
Chris Lattnere5d947b2004-12-09 16:36:40 +0000445</pre>
446
Misha Brukman9d0919f2003-11-08 01:05:38 +0000447<p>After strength reduction:</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000448
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000449<pre class="doc_code">
Bill Wendling2f7a8b02007-05-29 09:04:49 +0000450%result = <a href="#i_shl">shl</a> i32 %X, i8 3
Chris Lattnere5d947b2004-12-09 16:36:40 +0000451</pre>
452
Misha Brukman9d0919f2003-11-08 01:05:38 +0000453<p>And the hard way:</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000454
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000455<pre class="doc_code">
Gabor Greifec58f752009-10-28 13:05:07 +0000456%0 = <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
457%1 = <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
Bill Wendling2f7a8b02007-05-29 09:04:49 +0000458%result = <a href="#i_add">add</a> i32 %1, %1
Chris Lattnere5d947b2004-12-09 16:36:40 +0000459</pre>
460
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000461<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
462 lexical features of LLVM:</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000463
Chris Lattner00950542001-06-06 20:29:01 +0000464<ol>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000465 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000466 line.</li>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000467
468 <li>Unnamed temporaries are created when the result of a computation is not
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000469 assigned to a named value.</li>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000470
Misha Brukman9d0919f2003-11-08 01:05:38 +0000471 <li>Unnamed temporaries are numbered sequentially</li>
472</ol>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000473
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000474<p>It also shows a convention that we follow in this document. When
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000475 demonstrating instructions, we will follow an instruction with a comment that
476 defines the type and name of value produced. Comments are shown in italic
477 text.</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000478
Misha Brukman9d0919f2003-11-08 01:05:38 +0000479</div>
Chris Lattnerfa730212004-12-09 16:11:40 +0000480
481<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000482<h2><a name="highlevel">High Level Structure</a></h2>
Chris Lattnerfa730212004-12-09 16:11:40 +0000483<!-- *********************************************************************** -->
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000484<div>
Chris Lattnerfa730212004-12-09 16:11:40 +0000485<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000486<h3>
487 <a name="modulestructure">Module Structure</a>
488</h3>
Chris Lattnerfa730212004-12-09 16:11:40 +0000489
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000490<div>
Chris Lattnerfa730212004-12-09 16:11:40 +0000491
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000492<p>LLVM programs are composed of "Module"s, each of which is a translation unit
493 of the input programs. Each module consists of functions, global variables,
494 and symbol table entries. Modules may be combined together with the LLVM
495 linker, which merges function (and global variable) definitions, resolves
496 forward declarations, and merges symbol table entries. Here is an example of
497 the "hello world" module:</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000498
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000499<pre class="doc_code">
Chris Lattner63e4ccb2010-08-17 17:13:42 +0000500<i>; Declare the string constant as a global constant.</i>&nbsp;
Nick Lewyckydb9cd762011-01-29 01:09:53 +0000501<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a>&nbsp;<a href="#globalvars">constant</a>&nbsp;<a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>&nbsp;
Chris Lattnerfa730212004-12-09 16:11:40 +0000502
Chris Lattner63e4ccb2010-08-17 17:13:42 +0000503<i>; External declaration of the puts function</i>&nbsp;
504<a href="#functionstructure">declare</a> i32 @puts(i8*) <i>; i32 (i8*)* </i>&nbsp;
Chris Lattnerfa730212004-12-09 16:11:40 +0000505
506<i>; Definition of main function</i>
Chris Lattner63e4ccb2010-08-17 17:13:42 +0000507define i32 @main() { <i>; i32()* </i>&nbsp;
508 <i>; Convert [13 x i8]* to i8 *...</i>&nbsp;
509 %cast210 = <a href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8*</i>&nbsp;
Chris Lattnerfa730212004-12-09 16:11:40 +0000510
Chris Lattner63e4ccb2010-08-17 17:13:42 +0000511 <i>; Call puts function to write out the string to stdout.</i>&nbsp;
512 <a href="#i_call">call</a> i32 @puts(i8* %cast210) <i>; i32</i>&nbsp;
513 <a href="#i_ret">ret</a> i32 0&nbsp;
514}
Devang Patelcd1fd252010-01-11 19:35:55 +0000515
516<i>; Named metadata</i>
517!1 = metadata !{i32 41}
518!foo = !{!1, null}
Bill Wendling2f7a8b02007-05-29 09:04:49 +0000519</pre>
Chris Lattnerfa730212004-12-09 16:11:40 +0000520
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000521<p>This example is made up of a <a href="#globalvars">global variable</a> named
Devang Patelcd1fd252010-01-11 19:35:55 +0000522 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function,
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000523 a <a href="#functionstructure">function definition</a> for
Devang Patelcd1fd252010-01-11 19:35:55 +0000524 "<tt>main</tt>" and <a href="#namedmetadatastructure">named metadata</a>
525 "<tt>foo"</tt>.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000526
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000527<p>In general, a module is made up of a list of global values, where both
528 functions and global variables are global values. Global values are
529 represented by a pointer to a memory location (in this case, a pointer to an
530 array of char, and a pointer to a function), and have one of the
531 following <a href="#linkage">linkage types</a>.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000532
Chris Lattnere5d947b2004-12-09 16:36:40 +0000533</div>
534
535<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000536<h3>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000537 <a name="linkage">Linkage Types</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000538</h3>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000539
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000540<div>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000541
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000542<p>All Global Variables and Functions have one of the following types of
543 linkage:</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000544
545<dl>
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000546 <dt><tt><b><a name="linkage_private">private</a></b></tt></dt>
Bill Wendling5e721d72010-07-01 21:55:59 +0000547 <dd>Global values with "<tt>private</tt>" linkage are only directly accessible
548 by objects in the current module. In particular, linking code into a
549 module with an private global value may cause the private to be renamed as
550 necessary to avoid collisions. Because the symbol is private to the
551 module, all references can be updated. This doesn't show up in any symbol
552 table in the object file.</dd>
Rafael Espindolabb46f522009-01-15 20:18:42 +0000553
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000554 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt></dt>
Bill Wendling5e721d72010-07-01 21:55:59 +0000555 <dd>Similar to <tt>private</tt>, but the symbol is passed through the
556 assembler and evaluated by the linker. Unlike normal strong symbols, they
557 are removed by the linker from the final linked image (executable or
558 dynamic library).</dd>
559
560 <dt><tt><b><a name="linkage_linker_private_weak">linker_private_weak</a></b></tt></dt>
561 <dd>Similar to "<tt>linker_private</tt>", but the symbol is weak. Note that
562 <tt>linker_private_weak</tt> symbols are subject to coalescing by the
563 linker. The symbols are removed by the linker from the final linked image
564 (executable or dynamic library).</dd>
Bill Wendling3d10a5a2009-07-20 01:03:30 +0000565
Bill Wendling55ae5152010-08-20 22:05:50 +0000566 <dt><tt><b><a name="linkage_linker_private_weak_def_auto">linker_private_weak_def_auto</a></b></tt></dt>
567 <dd>Similar to "<tt>linker_private_weak</tt>", but it's known that the address
568 of the object is not taken. For instance, functions that had an inline
569 definition, but the compiler decided not to inline it. Note,
570 unlike <tt>linker_private</tt> and <tt>linker_private_weak</tt>,
571 <tt>linker_private_weak_def_auto</tt> may have only <tt>default</tt>
572 visibility. The symbols are removed by the linker from the final linked
573 image (executable or dynamic library).</dd>
574
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000575 <dt><tt><b><a name="linkage_internal">internal</a></b></tt></dt>
Bill Wendling07d31772010-06-29 22:34:52 +0000576 <dd>Similar to private, but the value shows as a local symbol
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000577 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
578 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000579
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000580 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt></dt>
Chris Lattner266c7bb2009-04-13 05:44:34 +0000581 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000582 into the object file corresponding to the LLVM module. They exist to
583 allow inlining and other optimizations to take place given knowledge of
584 the definition of the global, which is known to be somewhere outside the
585 module. Globals with <tt>available_externally</tt> linkage are allowed to
586 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
587 This linkage type is only allowed on definitions, not declarations.</dd>
Chris Lattner266c7bb2009-04-13 05:44:34 +0000588
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000589 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt></dt>
Chris Lattner4887bd82007-01-14 06:51:48 +0000590 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
Chris Lattner873187c2010-01-09 19:15:14 +0000591 the same name when linkage occurs. This can be used to implement
592 some forms of inline functions, templates, or other code which must be
593 generated in each translation unit that uses it, but where the body may
594 be overridden with a more definitive definition later. Unreferenced
595 <tt>linkonce</tt> globals are allowed to be discarded. Note that
596 <tt>linkonce</tt> linkage does not actually allow the optimizer to
597 inline the body of this function into callers because it doesn't know if
598 this definition of the function is the definitive definition within the
599 program or whether it will be overridden by a stronger definition.
600 To enable inlining and other optimizations, use "<tt>linkonce_odr</tt>"
601 linkage.</dd>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000602
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000603 <dt><tt><b><a name="linkage_weak">weak</a></b></tt></dt>
Chris Lattner26d054d2009-08-05 05:21:07 +0000604 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
605 <tt>linkonce</tt> linkage, except that unreferenced globals with
606 <tt>weak</tt> linkage may not be discarded. This is used for globals that
607 are declared "weak" in C source code.</dd>
608
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000609 <dt><tt><b><a name="linkage_common">common</a></b></tt></dt>
Chris Lattner26d054d2009-08-05 05:21:07 +0000610 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
611 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
612 global scope.
613 Symbols with "<tt>common</tt>" linkage are merged in the same way as
614 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
Chris Lattnercd81f5d2009-08-05 05:41:44 +0000615 <tt>common</tt> symbols may not have an explicit section,
Eric Christopher6c7e8a02009-12-05 02:46:03 +0000616 must have a zero initializer, and may not be marked '<a
Chris Lattnercd81f5d2009-08-05 05:41:44 +0000617 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
618 have common linkage.</dd>
Chris Lattner26d054d2009-08-05 05:21:07 +0000619
Chris Lattnere5d947b2004-12-09 16:36:40 +0000620
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000621 <dt><tt><b><a name="linkage_appending">appending</a></b></tt></dt>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000622 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000623 pointer to array type. When two global variables with appending linkage
624 are linked together, the two global arrays are appended together. This is
625 the LLVM, typesafe, equivalent of having the system linker append together
626 "sections" with identical names when .o files are linked.</dd>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000627
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000628 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000629 <dd>The semantics of this linkage follow the ELF object file model: the symbol
630 is weak until linked, if not linked, the symbol becomes null instead of
631 being an undefined reference.</dd>
Anton Korobeynikov7f705592007-01-12 19:20:47 +0000632
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000633 <dt><tt><b><a name="linkage_linkonce_odr">linkonce_odr</a></b></tt></dt>
634 <dt><tt><b><a name="linkage_weak_odr">weak_odr</a></b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000635 <dd>Some languages allow differing globals to be merged, such as two functions
636 with different semantics. Other languages, such as <tt>C++</tt>, ensure
Bill Wendling5e721d72010-07-01 21:55:59 +0000637 that only equivalent globals are ever merged (the "one definition rule"
638 &mdash; "ODR"). Such languages can use the <tt>linkonce_odr</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000639 and <tt>weak_odr</tt> linkage types to indicate that the global will only
640 be merged with equivalent globals. These linkage types are otherwise the
641 same as their non-<tt>odr</tt> versions.</dd>
Duncan Sands667d4b82009-03-07 15:45:40 +0000642
Chris Lattnerfa730212004-12-09 16:11:40 +0000643 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000644 <dd>If none of the above identifiers are used, the global is externally
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000645 visible, meaning that it participates in linkage and can be used to
646 resolve external symbol references.</dd>
Reid Spencerc8910842007-04-11 23:49:50 +0000647</dl>
Anton Korobeynikovb74ed072006-09-14 18:23:27 +0000648
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000649<p>The next two types of linkage are targeted for Microsoft Windows platform
650 only. They are designed to support importing (exporting) symbols from (to)
651 DLLs (Dynamic Link Libraries).</p>
Anton Korobeynikovb74ed072006-09-14 18:23:27 +0000652
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000653<dl>
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000654 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt></dt>
Anton Korobeynikovb74ed072006-09-14 18:23:27 +0000655 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000656 or variable via a global pointer to a pointer that is set up by the DLL
657 exporting the symbol. On Microsoft Windows targets, the pointer name is
658 formed by combining <code>__imp_</code> and the function or variable
659 name.</dd>
Anton Korobeynikovb74ed072006-09-14 18:23:27 +0000660
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000661 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt></dt>
Anton Korobeynikovb74ed072006-09-14 18:23:27 +0000662 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000663 pointer to a pointer in a DLL, so that it can be referenced with the
664 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
665 name is formed by combining <code>__imp_</code> and the function or
666 variable name.</dd>
Chris Lattnerfa730212004-12-09 16:11:40 +0000667</dl>
668
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000669<p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
670 another module defined a "<tt>.LC0</tt>" variable and was linked with this
671 one, one of the two would be renamed, preventing a collision. Since
672 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
673 declarations), they are accessible outside of the current module.</p>
674
675<p>It is illegal for a function <i>declaration</i> to have any linkage type
676 other than "externally visible", <tt>dllimport</tt>
677 or <tt>extern_weak</tt>.</p>
678
Duncan Sands667d4b82009-03-07 15:45:40 +0000679<p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000680 or <tt>weak_odr</tt> linkages.</p>
681
Chris Lattnerfa730212004-12-09 16:11:40 +0000682</div>
683
684<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000685<h3>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000686 <a name="callingconv">Calling Conventions</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000687</h3>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000688
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000689<div>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000690
691<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000692 and <a href="#i_invoke">invokes</a> can all have an optional calling
693 convention specified for the call. The calling convention of any pair of
694 dynamic caller/callee must match, or the behavior of the program is
695 undefined. The following calling conventions are supported by LLVM, and more
696 may be added in the future:</p>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000697
698<dl>
699 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000700 <dd>This calling convention (the default if no other calling convention is
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000701 specified) matches the target C calling conventions. This calling
702 convention supports varargs function calls and tolerates some mismatch in
703 the declared prototype and implemented declaration of the function (as
704 does normal C).</dd>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000705
706 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000707 <dd>This calling convention attempts to make calls as fast as possible
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000708 (e.g. by passing things in registers). This calling convention allows the
709 target to use whatever tricks it wants to produce fast code for the
710 target, without having to conform to an externally specified ABI
Jeffrey Yasskin95fa80a2010-01-09 19:44:16 +0000711 (Application Binary Interface).
712 <a href="CodeGenerator.html#tailcallopt">Tail calls can only be optimized
Chris Lattner29689432010-03-11 00:22:57 +0000713 when this or the GHC convention is used.</a> This calling convention
714 does not support varargs and requires the prototype of all callees to
715 exactly match the prototype of the function definition.</dd>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000716
717 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000718 <dd>This calling convention attempts to make code in the caller as efficient
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000719 as possible under the assumption that the call is not commonly executed.
720 As such, these calls often preserve all registers so that the call does
721 not break any live ranges in the caller side. This calling convention
722 does not support varargs and requires the prototype of all callees to
723 exactly match the prototype of the function definition.</dd>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000724
Chris Lattner29689432010-03-11 00:22:57 +0000725 <dt><b>"<tt>cc <em>10</em></tt>" - GHC convention</b>:</dt>
726 <dd>This calling convention has been implemented specifically for use by the
727 <a href="http://www.haskell.org/ghc">Glasgow Haskell Compiler (GHC)</a>.
728 It passes everything in registers, going to extremes to achieve this by
729 disabling callee save registers. This calling convention should not be
730 used lightly but only for specific situations such as an alternative to
731 the <em>register pinning</em> performance technique often used when
732 implementing functional programming languages.At the moment only X86
733 supports this convention and it has the following limitations:
734 <ul>
735 <li>On <em>X86-32</em> only supports up to 4 bit type parameters. No
736 floating point types are supported.</li>
737 <li>On <em>X86-64</em> only supports up to 10 bit type parameters and
738 6 floating point parameters.</li>
739 </ul>
740 This calling convention supports
741 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> but
742 requires both the caller and callee are using it.
743 </dd>
744
Chris Lattnercfe6b372005-05-07 01:46:40 +0000745 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000746 <dd>Any calling convention may be specified by number, allowing
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000747 target-specific calling conventions to be used. Target specific calling
748 conventions start at 64.</dd>
Chris Lattnercfe6b372005-05-07 01:46:40 +0000749</dl>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000750
751<p>More calling conventions can be added/defined on an as-needed basis, to
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000752 support Pascal conventions or any other well-known target-independent
753 convention.</p>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000754
755</div>
756
757<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000758<h3>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000759 <a name="visibility">Visibility Styles</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000760</h3>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000761
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000762<div>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000763
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000764<p>All Global Variables and Functions have one of the following visibility
765 styles:</p>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000766
767<dl>
768 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
Chris Lattnerd3eda892008-08-05 18:29:16 +0000769 <dd>On targets that use the ELF object file format, default visibility means
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000770 that the declaration is visible to other modules and, in shared libraries,
771 means that the declared entity may be overridden. On Darwin, default
772 visibility means that the declaration is visible to other modules. Default
773 visibility corresponds to "external linkage" in the language.</dd>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000774
775 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000776 <dd>Two declarations of an object with hidden visibility refer to the same
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000777 object if they are in the same shared object. Usually, hidden visibility
778 indicates that the symbol will not be placed into the dynamic symbol
779 table, so no other module (executable or shared library) can reference it
780 directly.</dd>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000781
Anton Korobeynikov6f9896f2007-04-29 18:35:00 +0000782 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
Anton Korobeynikov6f9896f2007-04-29 18:35:00 +0000783 <dd>On ELF, protected visibility indicates that the symbol will be placed in
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000784 the dynamic symbol table, but that references within the defining module
785 will bind to the local symbol. That is, the symbol cannot be overridden by
786 another module.</dd>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000787</dl>
788
789</div>
790
791<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000792<h3>
Chris Lattnere7886e42009-01-11 20:53:49 +0000793 <a name="namedtypes">Named Types</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000794</h3>
Chris Lattnere7886e42009-01-11 20:53:49 +0000795
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000796<div>
Chris Lattnere7886e42009-01-11 20:53:49 +0000797
798<p>LLVM IR allows you to specify name aliases for certain types. This can make
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000799 it easier to read the IR and make the IR more condensed (particularly when
800 recursive types are involved). An example of a name specification is:</p>
Chris Lattnere7886e42009-01-11 20:53:49 +0000801
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000802<pre class="doc_code">
Chris Lattnere7886e42009-01-11 20:53:49 +0000803%mytype = type { %mytype*, i32 }
804</pre>
Chris Lattnere7886e42009-01-11 20:53:49 +0000805
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000806<p>You may give a name to any <a href="#typesystem">type</a> except
Chris Lattnerdc65f222010-08-17 23:26:04 +0000807 "<a href="#t_void">void</a>". Type name aliases may be used anywhere a type
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000808 is expected with the syntax "%mytype".</p>
Chris Lattnere7886e42009-01-11 20:53:49 +0000809
810<p>Note that type names are aliases for the structural type that they indicate,
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000811 and that you can therefore specify multiple names for the same type. This
812 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
813 uses structural typing, the name is not part of the type. When printing out
814 LLVM IR, the printer will pick <em>one name</em> to render all types of a
815 particular shape. This means that if you have code where two different
816 source types end up having the same LLVM type, that the dumper will sometimes
817 print the "wrong" or unexpected type. This is an important design point and
818 isn't going to change.</p>
Chris Lattnere7886e42009-01-11 20:53:49 +0000819
820</div>
821
Chris Lattnere7886e42009-01-11 20:53:49 +0000822<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000823<h3>
Chris Lattnerfa730212004-12-09 16:11:40 +0000824 <a name="globalvars">Global Variables</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000825</h3>
Chris Lattnerfa730212004-12-09 16:11:40 +0000826
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000827<div>
Chris Lattnerfa730212004-12-09 16:11:40 +0000828
Chris Lattner3689a342005-02-12 19:30:21 +0000829<p>Global variables define regions of memory allocated at compilation time
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000830 instead of run-time. Global variables may optionally be initialized, may
831 have an explicit section to be placed in, and may have an optional explicit
832 alignment specified. A variable may be defined as "thread_local", which
833 means that it will not be shared by threads (each thread will have a
834 separated copy of the variable). A variable may be defined as a global
835 "constant," which indicates that the contents of the variable
836 will <b>never</b> be modified (enabling better optimization, allowing the
837 global data to be placed in the read-only section of an executable, etc).
838 Note that variables that need runtime initialization cannot be marked
839 "constant" as there is a store to the variable.</p>
Chris Lattner3689a342005-02-12 19:30:21 +0000840
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000841<p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
842 constant, even if the final definition of the global is not. This capability
843 can be used to enable slightly better optimization of the program, but
844 requires the language definition to guarantee that optimizations based on the
845 'constantness' are valid for the translation units that do not include the
846 definition.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000847
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000848<p>As SSA values, global variables define pointer values that are in scope
849 (i.e. they dominate) all basic blocks in the program. Global variables
850 always define a pointer to their "content" type because they describe a
851 region of memory, and all memory objects in LLVM are accessed through
852 pointers.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000853
Rafael Espindolabea46262011-01-08 16:42:36 +0000854<p>Global variables can be marked with <tt>unnamed_addr</tt> which indicates
855 that the address is not significant, only the content. Constants marked
Rafael Espindolaa5eaa862011-01-15 08:20:57 +0000856 like this can be merged with other constants if they have the same
857 initializer. Note that a constant with significant address <em>can</em>
858 be merged with a <tt>unnamed_addr</tt> constant, the result being a
859 constant whose address is significant.</p>
Rafael Espindolabea46262011-01-08 16:42:36 +0000860
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000861<p>A global variable may be declared to reside in a target-specific numbered
862 address space. For targets that support them, address spaces may affect how
863 optimizations are performed and/or what target instructions are used to
864 access the variable. The default address space is zero. The address space
865 qualifier must precede any other attributes.</p>
Christopher Lamb284d9922007-12-11 09:31:00 +0000866
Chris Lattner88f6c462005-11-12 00:45:07 +0000867<p>LLVM allows an explicit section to be specified for globals. If the target
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000868 supports it, it will emit globals to the section specified.</p>
Chris Lattner88f6c462005-11-12 00:45:07 +0000869
Chris Lattnerce99fa92010-04-28 00:13:42 +0000870<p>An explicit alignment may be specified for a global, which must be a power
871 of 2. If not present, or if the alignment is set to zero, the alignment of
872 the global is set by the target to whatever it feels convenient. If an
873 explicit alignment is specified, the global is forced to have exactly that
Chris Lattner2d4b8ee2010-04-28 00:31:12 +0000874 alignment. Targets and optimizers are not allowed to over-align the global
875 if the global has an assigned section. In this case, the extra alignment
876 could be observable: for example, code could assume that the globals are
877 densely packed in their section and try to iterate over them as an array,
878 alignment padding would break this iteration.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +0000879
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000880<p>For example, the following defines a global in a numbered address space with
881 an initializer, section, and alignment:</p>
Chris Lattner68027ea2007-01-14 00:27:09 +0000882
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000883<pre class="doc_code">
Dan Gohman398873c2009-01-11 00:40:00 +0000884@G = addrspace(5) constant float 1.0, section "foo", align 4
Chris Lattner68027ea2007-01-14 00:27:09 +0000885</pre>
886
Chris Lattnerfa730212004-12-09 16:11:40 +0000887</div>
888
889
890<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000891<h3>
Chris Lattnerfa730212004-12-09 16:11:40 +0000892 <a name="functionstructure">Functions</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000893</h3>
Chris Lattnerfa730212004-12-09 16:11:40 +0000894
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000895<div>
Chris Lattnerfa730212004-12-09 16:11:40 +0000896
Dan Gohmanb55a1ee2010-03-01 17:41:39 +0000897<p>LLVM function definitions consist of the "<tt>define</tt>" keyword, an
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000898 optional <a href="#linkage">linkage type</a>, an optional
899 <a href="#visibility">visibility style</a>, an optional
Rafael Espindolabea46262011-01-08 16:42:36 +0000900 <a href="#callingconv">calling convention</a>,
901 an optional <tt>unnamed_addr</tt> attribute, a return type, an optional
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000902 <a href="#paramattrs">parameter attribute</a> for the return type, a function
903 name, a (possibly empty) argument list (each with optional
904 <a href="#paramattrs">parameter attributes</a>), optional
905 <a href="#fnattrs">function attributes</a>, an optional section, an optional
906 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
907 curly brace, a list of basic blocks, and a closing curly brace.</p>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000908
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000909<p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
910 optional <a href="#linkage">linkage type</a>, an optional
Eric Christopher6c7e8a02009-12-05 02:46:03 +0000911 <a href="#visibility">visibility style</a>, an optional
Rafael Espindolabea46262011-01-08 16:42:36 +0000912 <a href="#callingconv">calling convention</a>,
913 an optional <tt>unnamed_addr</tt> attribute, a return type, an optional
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000914 <a href="#paramattrs">parameter attribute</a> for the return type, a function
915 name, a possibly empty list of arguments, an optional alignment, and an
916 optional <a href="#gc">garbage collector name</a>.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000917
Chris Lattnerd3eda892008-08-05 18:29:16 +0000918<p>A function definition contains a list of basic blocks, forming the CFG
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000919 (Control Flow Graph) for the function. Each basic block may optionally start
920 with a label (giving the basic block a symbol table entry), contains a list
921 of instructions, and ends with a <a href="#terminators">terminator</a>
922 instruction (such as a branch or function return).</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000923
Chris Lattner4a3c9012007-06-08 16:52:14 +0000924<p>The first basic block in a function is special in two ways: it is immediately
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000925 executed on entrance to the function, and it is not allowed to have
926 predecessor basic blocks (i.e. there can not be any branches to the entry
927 block of a function). Because the block can have no predecessors, it also
928 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000929
Chris Lattner88f6c462005-11-12 00:45:07 +0000930<p>LLVM allows an explicit section to be specified for functions. If the target
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000931 supports it, it will emit functions to the section specified.</p>
Chris Lattner88f6c462005-11-12 00:45:07 +0000932
Chris Lattner2cbdc452005-11-06 08:02:57 +0000933<p>An explicit alignment may be specified for a function. If not present, or if
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000934 the alignment is set to zero, the alignment of the function is set by the
935 target to whatever it feels convenient. If an explicit alignment is
936 specified, the function is forced to have at least that much alignment. All
937 alignments must be a power of 2.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +0000938
Rafael Espindolabea46262011-01-08 16:42:36 +0000939<p>If the <tt>unnamed_addr</tt> attribute is given, the address is know to not
940 be significant and two identical functions can be merged</p>.
941
Bill Wendlingc39e3e02009-07-20 02:39:26 +0000942<h5>Syntax:</h5>
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000943<pre class="doc_code">
Chris Lattner50ad45c2008-10-13 16:55:18 +0000944define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000945 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
946 &lt;ResultType&gt; @&lt;FunctionName&gt; ([argument list])
947 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
948 [<a href="#gc">gc</a>] { ... }
949</pre>
Devang Patel307e8ab2008-10-07 17:48:33 +0000950
Chris Lattnerfa730212004-12-09 16:11:40 +0000951</div>
952
Anton Korobeynikov8b0a8c82007-04-25 14:27:10 +0000953<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000954<h3>
Anton Korobeynikov8b0a8c82007-04-25 14:27:10 +0000955 <a name="aliasstructure">Aliases</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000956</h3>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000957
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000958<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000959
960<p>Aliases act as "second name" for the aliasee value (which can be either
961 function, global variable, another alias or bitcast of global value). Aliases
962 may have an optional <a href="#linkage">linkage type</a>, and an
963 optional <a href="#visibility">visibility style</a>.</p>
Anton Korobeynikov8b0a8c82007-04-25 14:27:10 +0000964
Bill Wendlingc39e3e02009-07-20 02:39:26 +0000965<h5>Syntax:</h5>
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000966<pre class="doc_code">
Duncan Sands0b23ac12008-09-12 20:48:21 +0000967@&lt;Name&gt; = alias [Linkage] [Visibility] &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
Bill Wendlingaac388b2007-05-29 09:42:13 +0000968</pre>
Anton Korobeynikov8b0a8c82007-04-25 14:27:10 +0000969
970</div>
971
Chris Lattner4e9aba72006-01-23 23:23:47 +0000972<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000973<h3>
Devang Patelcd1fd252010-01-11 19:35:55 +0000974 <a name="namedmetadatastructure">Named Metadata</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000975</h3>
Devang Patelcd1fd252010-01-11 19:35:55 +0000976
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000977<div>
Devang Patelcd1fd252010-01-11 19:35:55 +0000978
Chris Lattnere6a5ddd2010-01-15 21:50:19 +0000979<p>Named metadata is a collection of metadata. <a href="#metadata">Metadata
Dan Gohman872814a2010-07-21 18:54:18 +0000980 nodes</a> (but not metadata strings) are the only valid operands for
Chris Lattnere6a5ddd2010-01-15 21:50:19 +0000981 a named metadata.</p>
Devang Patelcd1fd252010-01-11 19:35:55 +0000982
983<h5>Syntax:</h5>
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000984<pre class="doc_code">
Dan Gohman872814a2010-07-21 18:54:18 +0000985; Some unnamed metadata nodes, which are referenced by the named metadata.
986!0 = metadata !{metadata !"zero"}
Devang Patelcd1fd252010-01-11 19:35:55 +0000987!1 = metadata !{metadata !"one"}
Dan Gohman872814a2010-07-21 18:54:18 +0000988!2 = metadata !{metadata !"two"}
Dan Gohman1005bc52010-07-13 19:48:13 +0000989; A named metadata.
Dan Gohman872814a2010-07-21 18:54:18 +0000990!name = !{!0, !1, !2}
Devang Patelcd1fd252010-01-11 19:35:55 +0000991</pre>
Devang Patelcd1fd252010-01-11 19:35:55 +0000992
993</div>
994
995<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000996<h3>
997 <a name="paramattrs">Parameter Attributes</a>
998</h3>
Reid Spencerca86e162006-12-31 07:07:53 +0000999
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001000<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001001
1002<p>The return type and each parameter of a function type may have a set of
1003 <i>parameter attributes</i> associated with them. Parameter attributes are
1004 used to communicate additional information about the result or parameters of
1005 a function. Parameter attributes are considered to be part of the function,
1006 not of the function type, so functions with different parameter attributes
1007 can have the same function type.</p>
1008
1009<p>Parameter attributes are simple keywords that follow the type specified. If
1010 multiple parameter attributes are needed, they are space separated. For
1011 example:</p>
Bill Wendling2f7a8b02007-05-29 09:04:49 +00001012
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00001013<pre class="doc_code">
Nick Lewyckyb6a7d252009-02-15 23:06:14 +00001014declare i32 @printf(i8* noalias nocapture, ...)
Chris Lattner66d922c2008-10-04 18:33:34 +00001015declare i32 @atoi(i8 zeroext)
1016declare signext i8 @returns_signed_char()
Bill Wendling2f7a8b02007-05-29 09:04:49 +00001017</pre>
Bill Wendling2f7a8b02007-05-29 09:04:49 +00001018
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001019<p>Note that any attributes for the function result (<tt>nounwind</tt>,
1020 <tt>readonly</tt>) come immediately after the argument list.</p>
Reid Spencerca86e162006-12-31 07:07:53 +00001021
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001022<p>Currently, only the following parameter attributes are defined:</p>
Chris Lattner47507de2008-01-11 06:20:47 +00001023
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001024<dl>
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001025 <dt><tt><b>zeroext</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001026 <dd>This indicates to the code generator that the parameter or return value
Cameron Zwarichebe81732011-03-16 22:20:18 +00001027 should be zero-extended to the extent required by the target's ABI (which
1028 is usually 32-bits, but is 8-bits for a i1 on x86-64) by the caller (for a
1029 parameter) or the callee (for a return value).</dd>
Chris Lattner47507de2008-01-11 06:20:47 +00001030
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001031 <dt><tt><b>signext</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001032 <dd>This indicates to the code generator that the parameter or return value
Cameron Zwarich9e69ff92011-03-17 14:21:58 +00001033 should be sign-extended to the extent required by the target's ABI (which
1034 is usually 32-bits) by the caller (for a parameter) or the callee (for a
1035 return value).</dd>
Chris Lattner47507de2008-01-11 06:20:47 +00001036
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001037 <dt><tt><b>inreg</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001038 <dd>This indicates that this parameter or return value should be treated in a
1039 special target-dependent fashion during while emitting code for a function
1040 call or return (usually, by putting it in a register as opposed to memory,
1041 though some targets use it to distinguish between two different kinds of
1042 registers). Use of this attribute is target-specific.</dd>
Chris Lattner47507de2008-01-11 06:20:47 +00001043
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001044 <dt><tt><b><a name="byval">byval</a></b></tt></dt>
Chris Lattnera6fd81d2010-11-20 23:49:06 +00001045 <dd><p>This indicates that the pointer parameter should really be passed by
1046 value to the function. The attribute implies that a hidden copy of the
1047 pointee
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001048 is made between the caller and the callee, so the callee is unable to
1049 modify the value in the callee. This attribute is only valid on LLVM
1050 pointer arguments. It is generally used to pass structs and arrays by
1051 value, but is also valid on pointers to scalars. The copy is considered
1052 to belong to the caller not the callee (for example,
1053 <tt><a href="#readonly">readonly</a></tt> functions should not write to
1054 <tt>byval</tt> parameters). This is not a valid attribute for return
Chris Lattnera6fd81d2010-11-20 23:49:06 +00001055 values.</p>
1056
1057 <p>The byval attribute also supports specifying an alignment with
1058 the align attribute. It indicates the alignment of the stack slot to
1059 form and the known alignment of the pointer specified to the call site. If
1060 the alignment is not specified, then the code generator makes a
1061 target-specific assumption.</p></dd>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001062
Dan Gohmanff235352010-07-02 23:18:08 +00001063 <dt><tt><b><a name="sret">sret</a></b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001064 <dd>This indicates that the pointer parameter specifies the address of a
1065 structure that is the return value of the function in the source program.
1066 This pointer must be guaranteed by the caller to be valid: loads and
1067 stores to the structure may be assumed by the callee to not to trap. This
1068 may only be applied to the first parameter. This is not a valid attribute
1069 for return values. </dd>
1070
Dan Gohmanff235352010-07-02 23:18:08 +00001071 <dt><tt><b><a name="noalias">noalias</a></b></tt></dt>
Dan Gohman1e109622010-07-02 18:41:32 +00001072 <dd>This indicates that pointer values
1073 <a href="#pointeraliasing"><i>based</i></a> on the argument or return
Dan Gohmanefca7f92010-07-02 23:46:54 +00001074 value do not alias pointer values which are not <i>based</i> on it,
1075 ignoring certain "irrelevant" dependencies.
1076 For a call to the parent function, dependencies between memory
1077 references from before or after the call and from those during the call
1078 are "irrelevant" to the <tt>noalias</tt> keyword for the arguments and
1079 return value used in that call.
Dan Gohman1e109622010-07-02 18:41:32 +00001080 The caller shares the responsibility with the callee for ensuring that
1081 these requirements are met.
1082 For further details, please see the discussion of the NoAlias response in
Dan Gohmanff70fe42010-07-06 15:26:33 +00001083 <a href="AliasAnalysis.html#MustMayNo">alias analysis</a>.<br>
1084<br>
John McCall191d4ee2010-07-06 21:07:14 +00001085 Note that this definition of <tt>noalias</tt> is intentionally
1086 similar to the definition of <tt>restrict</tt> in C99 for function
Chris Lattner211244a2010-07-06 20:51:35 +00001087 arguments, though it is slightly weaker.
Dan Gohmanff70fe42010-07-06 15:26:33 +00001088<br>
1089 For function return values, C99's <tt>restrict</tt> is not meaningful,
1090 while LLVM's <tt>noalias</tt> is.
1091 </dd>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001092
Dan Gohmanff235352010-07-02 23:18:08 +00001093 <dt><tt><b><a name="nocapture">nocapture</a></b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001094 <dd>This indicates that the callee does not make any copies of the pointer
1095 that outlive the callee itself. This is not a valid attribute for return
1096 values.</dd>
1097
Dan Gohmanff235352010-07-02 23:18:08 +00001098 <dt><tt><b><a name="nest">nest</a></b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001099 <dd>This indicates that the pointer parameter can be excised using the
1100 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
1101 attribute for return values.</dd>
1102</dl>
Reid Spencerca86e162006-12-31 07:07:53 +00001103
Reid Spencerca86e162006-12-31 07:07:53 +00001104</div>
1105
1106<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001107<h3>
Gordon Henriksen80a75bf2007-12-10 03:18:06 +00001108 <a name="gc">Garbage Collector Names</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001109</h3>
Gordon Henriksen80a75bf2007-12-10 03:18:06 +00001110
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001111<div>
Gordon Henriksen80a75bf2007-12-10 03:18:06 +00001112
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001113<p>Each function may specify a garbage collector name, which is simply a
1114 string:</p>
1115
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00001116<pre class="doc_code">
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001117define void @f() gc "name" { ... }
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001118</pre>
Gordon Henriksen80a75bf2007-12-10 03:18:06 +00001119
1120<p>The compiler declares the supported values of <i>name</i>. Specifying a
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001121 collector which will cause the compiler to alter its output in order to
1122 support the named garbage collection algorithm.</p>
1123
Gordon Henriksen80a75bf2007-12-10 03:18:06 +00001124</div>
1125
1126<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001127<h3>
Devang Patel2c9c3e72008-09-26 23:51:19 +00001128 <a name="fnattrs">Function Attributes</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001129</h3>
Devang Patelf8b94812008-09-04 23:05:13 +00001130
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001131<div>
Devang Patel2c9c3e72008-09-26 23:51:19 +00001132
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001133<p>Function attributes are set to communicate additional information about a
1134 function. Function attributes are considered to be part of the function, not
1135 of the function type, so functions with different parameter attributes can
1136 have the same function type.</p>
Devang Patel2c9c3e72008-09-26 23:51:19 +00001137
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001138<p>Function attributes are simple keywords that follow the type specified. If
1139 multiple attributes are needed, they are space separated. For example:</p>
Devang Patelf8b94812008-09-04 23:05:13 +00001140
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00001141<pre class="doc_code">
Devang Patel2c9c3e72008-09-26 23:51:19 +00001142define void @f() noinline { ... }
1143define void @f() alwaysinline { ... }
1144define void @f() alwaysinline optsize { ... }
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001145define void @f() optsize { ... }
Bill Wendlinge36dccc2008-09-07 10:26:33 +00001146</pre>
Devang Patelf8b94812008-09-04 23:05:13 +00001147
Bill Wendlinge36dccc2008-09-07 10:26:33 +00001148<dl>
Charles Davis1e063d12010-02-12 00:31:15 +00001149 <dt><tt><b>alignstack(&lt;<em>n</em>&gt;)</b></tt></dt>
1150 <dd>This attribute indicates that, when emitting the prologue and epilogue,
1151 the backend should forcibly align the stack pointer. Specify the
1152 desired alignment, which must be a power of two, in parentheses.
1153
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001154 <dt><tt><b>alwaysinline</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001155 <dd>This attribute indicates that the inliner should attempt to inline this
1156 function into callers whenever possible, ignoring any active inlining size
1157 threshold for this caller.</dd>
Bill Wendlinge36dccc2008-09-07 10:26:33 +00001158
Charles Davis970bfcc2010-10-25 15:37:09 +00001159 <dt><tt><b>hotpatch</b></tt></dt>
Charles Davis6f12e292010-10-25 16:29:03 +00001160 <dd>This attribute indicates that the function should be 'hotpatchable',
Charles Davis0076d202010-10-25 19:07:39 +00001161 meaning the function can be patched and/or hooked even while it is
1162 loaded into memory. On x86, the function prologue will be preceded
1163 by six bytes of padding and will begin with a two-byte instruction.
1164 Most of the functions in the Windows system DLLs in Windows XP SP2 or
1165 higher were compiled in this fashion.</dd>
Charles Davis970bfcc2010-10-25 15:37:09 +00001166
Dan Gohman129bd562011-06-16 16:03:13 +00001167 <dt><tt><b>nonlazybind</b></tt></dt>
1168 <dd>This attribute suppresses lazy symbol binding for the function. This
1169 may make calls to the function faster, at the cost of extra program
1170 startup time if the function is not called during program startup.</dd>
1171
Jakob Stoklund Olesen570a4a52010-02-06 01:16:28 +00001172 <dt><tt><b>inlinehint</b></tt></dt>
1173 <dd>This attribute indicates that the source code contained a hint that inlining
1174 this function is desirable (such as the "inline" keyword in C/C++). It
1175 is just a hint; it imposes no requirements on the inliner.</dd>
1176
Nick Lewycky76ec37a2010-07-06 18:24:09 +00001177 <dt><tt><b>naked</b></tt></dt>
1178 <dd>This attribute disables prologue / epilogue emission for the function.
1179 This can have very system-specific consequences.</dd>
1180
1181 <dt><tt><b>noimplicitfloat</b></tt></dt>
1182 <dd>This attributes disables implicit floating point instructions.</dd>
1183
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001184 <dt><tt><b>noinline</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001185 <dd>This attribute indicates that the inliner should never inline this
1186 function in any situation. This attribute may not be used together with
1187 the <tt>alwaysinline</tt> attribute.</dd>
Devang Patel2c9c3e72008-09-26 23:51:19 +00001188
Nick Lewycky76ec37a2010-07-06 18:24:09 +00001189 <dt><tt><b>noredzone</b></tt></dt>
1190 <dd>This attribute indicates that the code generator should not use a red
1191 zone, even if the target-specific ABI normally permits it.</dd>
Devang Patel2c9c3e72008-09-26 23:51:19 +00001192
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001193 <dt><tt><b>noreturn</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001194 <dd>This function attribute indicates that the function never returns
1195 normally. This produces undefined behavior at runtime if the function
1196 ever does dynamically return.</dd>
Bill Wendling31359ba2008-11-13 01:02:51 +00001197
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001198 <dt><tt><b>nounwind</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001199 <dd>This function attribute indicates that the function never returns with an
1200 unwind or exceptional control flow. If the function does unwind, its
1201 runtime behavior is undefined.</dd>
Bill Wendlingfbaa7ed2008-11-26 19:07:40 +00001202
Nick Lewycky76ec37a2010-07-06 18:24:09 +00001203 <dt><tt><b>optsize</b></tt></dt>
1204 <dd>This attribute suggests that optimization passes and code generator passes
1205 make choices that keep the code size of this function low, and otherwise
1206 do optimizations specifically to reduce code size.</dd>
1207
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001208 <dt><tt><b>readnone</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001209 <dd>This attribute indicates that the function computes its result (or decides
1210 to unwind an exception) based strictly on its arguments, without
1211 dereferencing any pointer arguments or otherwise accessing any mutable
1212 state (e.g. memory, control registers, etc) visible to caller functions.
1213 It does not write through any pointer arguments
1214 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1215 changes any state visible to callers. This means that it cannot unwind
1216 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1217 could use the <tt>unwind</tt> instruction.</dd>
Devang Patel5d96fda2009-06-12 19:45:19 +00001218
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001219 <dt><tt><b><a name="readonly">readonly</a></b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001220 <dd>This attribute indicates that the function does not write through any
1221 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1222 arguments) or otherwise modify any state (e.g. memory, control registers,
1223 etc) visible to caller functions. It may dereference pointer arguments
1224 and read state that may be set in the caller. A readonly function always
1225 returns the same value (or unwinds an exception identically) when called
1226 with the same set of arguments and global state. It cannot unwind an
1227 exception by calling the <tt>C++</tt> exception throwing methods, but may
1228 use the <tt>unwind</tt> instruction.</dd>
Anton Korobeynikovc5ec8a72009-07-17 18:07:26 +00001229
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001230 <dt><tt><b><a name="ssp">ssp</a></b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001231 <dd>This attribute indicates that the function should emit a stack smashing
1232 protector. It is in the form of a "canary"&mdash;a random value placed on
1233 the stack before the local variables that's checked upon return from the
1234 function to see if it has been overwritten. A heuristic is used to
1235 determine if a function needs stack protectors or not.<br>
1236<br>
1237 If a function that has an <tt>ssp</tt> attribute is inlined into a
1238 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1239 function will have an <tt>ssp</tt> attribute.</dd>
1240
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001241 <dt><tt><b>sspreq</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001242 <dd>This attribute indicates that the function should <em>always</em> emit a
1243 stack smashing protector. This overrides
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001244 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1245<br>
1246 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1247 function that doesn't have an <tt>sspreq</tt> attribute or which has
1248 an <tt>ssp</tt> attribute, then the resulting function will have
1249 an <tt>sspreq</tt> attribute.</dd>
Bill Wendlinge36dccc2008-09-07 10:26:33 +00001250</dl>
1251
Devang Patelf8b94812008-09-04 23:05:13 +00001252</div>
1253
1254<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001255<h3>
Chris Lattner1eeeb0c2006-04-08 04:40:53 +00001256 <a name="moduleasm">Module-Level Inline Assembly</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001257</h3>
Chris Lattner4e9aba72006-01-23 23:23:47 +00001258
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001259<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001260
1261<p>Modules may contain "module-level inline asm" blocks, which corresponds to
1262 the GCC "file scope inline asm" blocks. These blocks are internally
1263 concatenated by LLVM and treated as a single unit, but may be separated in
1264 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
Chris Lattner4e9aba72006-01-23 23:23:47 +00001265
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00001266<pre class="doc_code">
Bill Wendling2f7a8b02007-05-29 09:04:49 +00001267module asm "inline asm code goes here"
1268module asm "more can go here"
1269</pre>
Chris Lattner4e9aba72006-01-23 23:23:47 +00001270
1271<p>The strings can contain any character by escaping non-printable characters.
1272 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001273 for the number.</p>
Chris Lattner4e9aba72006-01-23 23:23:47 +00001274
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001275<p>The inline asm code is simply printed to the machine code .s file when
1276 assembly code is generated.</p>
1277
Chris Lattner4e9aba72006-01-23 23:23:47 +00001278</div>
Chris Lattnerfa730212004-12-09 16:11:40 +00001279
Reid Spencerde151942007-02-19 23:54:10 +00001280<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001281<h3>
Reid Spencerde151942007-02-19 23:54:10 +00001282 <a name="datalayout">Data Layout</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001283</h3>
Reid Spencerde151942007-02-19 23:54:10 +00001284
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001285<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001286
Reid Spencerde151942007-02-19 23:54:10 +00001287<p>A module may specify a target specific data layout string that specifies how
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001288 data is to be laid out in memory. The syntax for the data layout is
1289 simply:</p>
1290
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00001291<pre class="doc_code">
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001292target datalayout = "<i>layout specification</i>"
1293</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001294
1295<p>The <i>layout specification</i> consists of a list of specifications
1296 separated by the minus sign character ('-'). Each specification starts with
1297 a letter and may include other information after the letter to define some
1298 aspect of the data layout. The specifications accepted are as follows:</p>
1299
Reid Spencerde151942007-02-19 23:54:10 +00001300<dl>
1301 <dt><tt>E</tt></dt>
1302 <dd>Specifies that the target lays out data in big-endian form. That is, the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001303 bits with the most significance have the lowest address location.</dd>
1304
Reid Spencerde151942007-02-19 23:54:10 +00001305 <dt><tt>e</tt></dt>
Chris Lattnerd3eda892008-08-05 18:29:16 +00001306 <dd>Specifies that the target lays out data in little-endian form. That is,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001307 the bits with the least significance have the lowest address
1308 location.</dd>
1309
Reid Spencerde151942007-02-19 23:54:10 +00001310 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00001311 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001312 <i>preferred</i> alignments. All sizes are in bits. Specifying
1313 the <i>pref</i> alignment is optional. If omitted, the
1314 preceding <tt>:</tt> should be omitted too.</dd>
1315
Reid Spencerde151942007-02-19 23:54:10 +00001316 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1317 <dd>This specifies the alignment for an integer type of a given bit
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001318 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1319
Reid Spencerde151942007-02-19 23:54:10 +00001320 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00001321 <dd>This specifies the alignment for a vector type of a given bit
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001322 <i>size</i>.</dd>
1323
Reid Spencerde151942007-02-19 23:54:10 +00001324 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00001325 <dd>This specifies the alignment for a floating point type of a given bit
Dale Johannesen9d8d2212010-05-28 18:54:47 +00001326 <i>size</i>. Only values of <i>size</i> that are supported by the target
1327 will work. 32 (float) and 64 (double) are supported on all targets;
1328 80 or 128 (different flavors of long double) are also supported on some
1329 targets.
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001330
Reid Spencerde151942007-02-19 23:54:10 +00001331 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1332 <dd>This specifies the alignment for an aggregate type of a given bit
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001333 <i>size</i>.</dd>
1334
Daniel Dunbar87bde0b2009-06-08 22:17:53 +00001335 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1336 <dd>This specifies the alignment for a stack object of a given bit
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001337 <i>size</i>.</dd>
Chris Lattnere82bdc42009-11-07 09:35:34 +00001338
1339 <dt><tt>n<i>size1</i>:<i>size2</i>:<i>size3</i>...</tt></dt>
1340 <dd>This specifies a set of native integer widths for the target CPU
1341 in bits. For example, it might contain "n32" for 32-bit PowerPC,
1342 "n32:64" for PowerPC 64, or "n8:16:32:64" for X86-64. Elements of
Eric Christopher6c7e8a02009-12-05 02:46:03 +00001343 this set are considered to support most general arithmetic
Chris Lattnere82bdc42009-11-07 09:35:34 +00001344 operations efficiently.</dd>
Reid Spencerde151942007-02-19 23:54:10 +00001345</dl>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001346
Reid Spencerde151942007-02-19 23:54:10 +00001347<p>When constructing the data layout for a given target, LLVM starts with a
Dan Gohman1c70c002010-04-28 00:36:01 +00001348 default set of specifications which are then (possibly) overridden by the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001349 specifications in the <tt>datalayout</tt> keyword. The default specifications
1350 are given in this list:</p>
1351
Reid Spencerde151942007-02-19 23:54:10 +00001352<ul>
1353 <li><tt>E</tt> - big endian</li>
Dan Gohmanfdf2e8c2010-02-23 02:44:03 +00001354 <li><tt>p:64:64:64</tt> - 64-bit pointers with 64-bit alignment</li>
Reid Spencerde151942007-02-19 23:54:10 +00001355 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1356 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1357 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1358 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
Chris Lattnerd3eda892008-08-05 18:29:16 +00001359 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
Reid Spencerde151942007-02-19 23:54:10 +00001360 alignment of 64-bits</li>
1361 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1362 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1363 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1364 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1365 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
Daniel Dunbar87bde0b2009-06-08 22:17:53 +00001366 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
Reid Spencerde151942007-02-19 23:54:10 +00001367</ul>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001368
1369<p>When LLVM is determining the alignment for a given type, it uses the
1370 following rules:</p>
1371
Reid Spencerde151942007-02-19 23:54:10 +00001372<ol>
1373 <li>If the type sought is an exact match for one of the specifications, that
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001374 specification is used.</li>
1375
Reid Spencerde151942007-02-19 23:54:10 +00001376 <li>If no match is found, and the type sought is an integer type, then the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001377 smallest integer type that is larger than the bitwidth of the sought type
1378 is used. If none of the specifications are larger than the bitwidth then
1379 the the largest integer type is used. For example, given the default
1380 specifications above, the i7 type will use the alignment of i8 (next
1381 largest) while both i65 and i256 will use the alignment of i64 (largest
1382 specified).</li>
1383
Reid Spencerde151942007-02-19 23:54:10 +00001384 <li>If no match is found, and the type sought is a vector type, then the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001385 largest vector type that is smaller than the sought vector type will be
1386 used as a fall back. This happens because &lt;128 x double&gt; can be
1387 implemented in terms of 64 &lt;2 x double&gt;, for example.</li>
Reid Spencerde151942007-02-19 23:54:10 +00001388</ol>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001389
Reid Spencerde151942007-02-19 23:54:10 +00001390</div>
Chris Lattnerfa730212004-12-09 16:11:40 +00001391
Dan Gohman556ca272009-07-27 18:07:55 +00001392<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001393<h3>
Dan Gohman556ca272009-07-27 18:07:55 +00001394 <a name="pointeraliasing">Pointer Aliasing Rules</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001395</h3>
Dan Gohman556ca272009-07-27 18:07:55 +00001396
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001397<div>
Dan Gohman556ca272009-07-27 18:07:55 +00001398
Andreas Bolka55e459a2009-07-29 00:02:05 +00001399<p>Any memory access must be done through a pointer value associated
Andreas Bolka99a82052009-07-27 20:37:10 +00001400with an address range of the memory access, otherwise the behavior
Dan Gohman556ca272009-07-27 18:07:55 +00001401is undefined. Pointer values are associated with address ranges
1402according to the following rules:</p>
1403
1404<ul>
Dan Gohman1e109622010-07-02 18:41:32 +00001405 <li>A pointer value is associated with the addresses associated with
1406 any value it is <i>based</i> on.
Andreas Bolka55e459a2009-07-29 00:02:05 +00001407 <li>An address of a global variable is associated with the address
Dan Gohman556ca272009-07-27 18:07:55 +00001408 range of the variable's storage.</li>
1409 <li>The result value of an allocation instruction is associated with
1410 the address range of the allocated storage.</li>
1411 <li>A null pointer in the default address-space is associated with
Andreas Bolka55e459a2009-07-29 00:02:05 +00001412 no address.</li>
Dan Gohman556ca272009-07-27 18:07:55 +00001413 <li>An integer constant other than zero or a pointer value returned
1414 from a function not defined within LLVM may be associated with address
1415 ranges allocated through mechanisms other than those provided by
Andreas Bolka55e459a2009-07-29 00:02:05 +00001416 LLVM. Such ranges shall not overlap with any ranges of addresses
Dan Gohman556ca272009-07-27 18:07:55 +00001417 allocated by mechanisms provided by LLVM.</li>
Dan Gohman1e109622010-07-02 18:41:32 +00001418</ul>
1419
1420<p>A pointer value is <i>based</i> on another pointer value according
1421 to the following rules:</p>
1422
1423<ul>
1424 <li>A pointer value formed from a
1425 <tt><a href="#i_getelementptr">getelementptr</a></tt> operation
1426 is <i>based</i> on the first operand of the <tt>getelementptr</tt>.</li>
1427 <li>The result value of a
1428 <tt><a href="#i_bitcast">bitcast</a></tt> is <i>based</i> on the operand
1429 of the <tt>bitcast</tt>.</li>
1430 <li>A pointer value formed by an
1431 <tt><a href="#i_inttoptr">inttoptr</a></tt> is <i>based</i> on all
1432 pointer values that contribute (directly or indirectly) to the
1433 computation of the pointer's value.</li>
1434 <li>The "<i>based</i> on" relationship is transitive.</li>
1435</ul>
1436
1437<p>Note that this definition of <i>"based"</i> is intentionally
1438 similar to the definition of <i>"based"</i> in C99, though it is
1439 slightly weaker.</p>
Dan Gohman556ca272009-07-27 18:07:55 +00001440
1441<p>LLVM IR does not associate types with memory. The result type of a
Andreas Bolka55e459a2009-07-29 00:02:05 +00001442<tt><a href="#i_load">load</a></tt> merely indicates the size and
1443alignment of the memory from which to load, as well as the
Dan Gohmanc22c0f32010-06-17 19:23:50 +00001444interpretation of the value. The first operand type of a
Andreas Bolka55e459a2009-07-29 00:02:05 +00001445<tt><a href="#i_store">store</a></tt> similarly only indicates the size
1446and alignment of the store.</p>
Dan Gohman556ca272009-07-27 18:07:55 +00001447
1448<p>Consequently, type-based alias analysis, aka TBAA, aka
1449<tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1450LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1451additional information which specialized optimization passes may use
1452to implement type-based alias analysis.</p>
1453
1454</div>
1455
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00001456<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001457<h3>
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00001458 <a name="volatile">Volatile Memory Accesses</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001459</h3>
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00001460
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001461<div>
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00001462
1463<p>Certain memory accesses, such as <a href="#i_load"><tt>load</tt></a>s, <a
1464href="#i_store"><tt>store</tt></a>s, and <a
1465href="#int_memcpy"><tt>llvm.memcpy</tt></a>s may be marked <tt>volatile</tt>.
1466The optimizers must not change the number of volatile operations or change their
1467order of execution relative to other volatile operations. The optimizers
1468<i>may</i> change the order of volatile operations relative to non-volatile
1469operations. This is not Java's "volatile" and has no cross-thread
1470synchronization behavior.</p>
1471
1472</div>
1473
Eli Friedman5b60e1b2011-07-20 21:35:53 +00001474<!-- ======================================================================= -->
1475<h3>
1476 <a name="memmodel">Memory Model for Concurrent Operations</a>
1477</h3>
1478
1479<div>
1480
1481<p>The LLVM IR does not define any way to start parallel threads of execution
1482or to register signal handlers. Nonetheless, there are platform-specific
1483ways to create them, and we define LLVM IR's behavior in their presence. This
1484model is inspired by the C++0x memory model.</p>
1485
1486<p>We define a <i>happens-before</i> partial order as the least partial order
1487that</p>
1488<ul>
1489 <li>Is a superset of single-thread program order, and</li>
1490 <li>When a <i>synchronizes-with</i> <tt>b</tt>, includes an edge from
1491 <tt>a</tt> to <tt>b</tt>. <i>Synchronizes-with</i> pairs are introduced
1492 by platform-specific techniques, like pthread locks, thread
1493 creation, thread joining, etc., and by the atomic operations described
1494 in the <a href="#int_atomics">Atomic intrinsics</a> section.</li>
1495</ul>
1496
1497<p>Note that program order does not introduce <i>happens-before</i> edges
1498between a thread and signals executing inside that thread.</p>
1499
1500<p>Every (defined) read operation (load instructions, memcpy, atomic
1501loads/read-modify-writes, etc.) <var>R</var> reads a series of bytes written by
1502(defined) write operations (store instructions, atomic
Eli Friedman118973a2011-07-22 03:04:45 +00001503stores/read-modify-writes, memcpy, etc.). For the purposes of this section,
1504initialized globals are considered to have a write of the initializer which is
1505atomic and happens before any other read or write of the memory in question.
1506For each byte of a read <var>R</var>, <var>R<sub>byte</sub></var> may see
1507any write to the same byte, except:</p>
Eli Friedman5b60e1b2011-07-20 21:35:53 +00001508
1509<ul>
1510 <li>If <var>write<sub>1</sub></var> happens before
1511 <var>write<sub>2</sub></var>, and <var>write<sub>2</sub></var> happens
1512 before <var>R<sub>byte</sub></var>, then <var>R<sub>byte</sub></var>
Eli Friedman118973a2011-07-22 03:04:45 +00001513 does not see <var>write<sub>1</sub></var>.
Eli Friedman5b60e1b2011-07-20 21:35:53 +00001514 <li>If <var>R<sub>byte</sub></var> happens before <var>write<sub>3</var>,
Eli Friedman118973a2011-07-22 03:04:45 +00001515 then <var>R<sub>byte</sub></var> does not see
Eli Friedman5b60e1b2011-07-20 21:35:53 +00001516 <var>write<sub>3</sub></var>.
1517</ul>
1518
1519<p>Given that definition, <var>R<sub>byte</sub></var> is defined as follows:
1520<ul>
1521 <li>If there is no write to the same byte that happens before
1522 <var>R<sub>byte</sub></var>, <var>R<sub>byte</sub></var> returns
1523 <tt>undef</tt> for that byte.
Eli Friedman118973a2011-07-22 03:04:45 +00001524 <li>Otherwise, if <var>R<sub>byte</sub></var> may see exactly one write,
Eli Friedman5b60e1b2011-07-20 21:35:53 +00001525 <var>R<sub>byte</sub></var> returns the value written by that
1526 write.</li>
Eli Friedman118973a2011-07-22 03:04:45 +00001527 <li>Otherwise, if <var>R</var> is atomic, and all the writes
1528 <var>R<sub>byte</sub></var> may see are atomic, it chooses one of the
1529 values written. See the <a href="#int_atomics">Atomic intrinsics</a>
1530 section for additional guarantees on how the choice is made.
Eli Friedman5b60e1b2011-07-20 21:35:53 +00001531 <li>Otherwise <var>R<sub>byte</sub></var> returns <tt>undef</tt>.</li>
1532</ul>
1533
1534<p><var>R</var> returns the value composed of the series of bytes it read.
1535This implies that some bytes within the value may be <tt>undef</tt>
1536<b>without</b> the entire value being <tt>undef</tt>. Note that this only
1537defines the semantics of the operation; it doesn't mean that targets will
1538emit more than one instruction to read the series of bytes.</p>
1539
1540<p>Note that in cases where none of the atomic intrinsics are used, this model
1541places only one restriction on IR transformations on top of what is required
1542for single-threaded execution: introducing a store to a byte which might not
Eli Friedman118973a2011-07-22 03:04:45 +00001543otherwise be stored to can introduce undefined behavior. (Specifically, in
1544the case where another thread might write to and read from an address,
1545introducing a store can change a load that may see exactly one write into
1546a load that may see multiple writes.)</p>
Eli Friedman5b60e1b2011-07-20 21:35:53 +00001547
1548<!-- FIXME: This model assumes all targets where concurrency is relevant have
1549a byte-size store which doesn't affect adjacent bytes. As far as I can tell,
1550none of the backends currently in the tree fall into this category; however,
1551there might be targets which care. If there are, we want a paragraph
1552like the following:
1553
1554Targets may specify that stores narrower than a certain width are not
1555available; on such a target, for the purposes of this model, treat any
1556non-atomic write with an alignment or width less than the minimum width
1557as if it writes to the relevant surrounding bytes.
1558-->
1559
1560</div>
1561
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001562</div>
1563
Chris Lattner00950542001-06-06 20:29:01 +00001564<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001565<h2><a name="typesystem">Type System</a></h2>
Chris Lattner261efe92003-11-25 01:02:51 +00001566<!-- *********************************************************************** -->
Chris Lattnerfa730212004-12-09 16:11:40 +00001567
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001568<div>
Chris Lattnerfa730212004-12-09 16:11:40 +00001569
Misha Brukman9d0919f2003-11-08 01:05:38 +00001570<p>The LLVM type system is one of the most important features of the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001571 intermediate representation. Being typed enables a number of optimizations
1572 to be performed on the intermediate representation directly, without having
1573 to do extra analyses on the side before the transformation. A strong type
1574 system makes it easier to read the generated code and enables novel analyses
1575 and transformations that are not feasible to perform on normal three address
1576 code representations.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +00001577
Chris Lattner00950542001-06-06 20:29:01 +00001578<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001579<h3>
1580 <a name="t_classifications">Type Classifications</a>
1581</h3>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001582
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001583<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001584
1585<p>The types fall into a few useful classifications:</p>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00001586
1587<table border="1" cellspacing="0" cellpadding="4">
Chris Lattner261efe92003-11-25 01:02:51 +00001588 <tbody>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001589 <tr><th>Classification</th><th>Types</th></tr>
Chris Lattner261efe92003-11-25 01:02:51 +00001590 <tr>
Chris Lattner4f69f462008-01-04 04:32:38 +00001591 <td><a href="#t_integer">integer</a></td>
Reid Spencer2b916312007-05-16 18:44:01 +00001592 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
Chris Lattner261efe92003-11-25 01:02:51 +00001593 </tr>
1594 <tr>
Chris Lattner4f69f462008-01-04 04:32:38 +00001595 <td><a href="#t_floating">floating point</a></td>
1596 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Chris Lattner261efe92003-11-25 01:02:51 +00001597 </tr>
1598 <tr>
1599 <td><a name="t_firstclass">first class</a></td>
Chris Lattner4f69f462008-01-04 04:32:38 +00001600 <td><a href="#t_integer">integer</a>,
1601 <a href="#t_floating">floating point</a>,
1602 <a href="#t_pointer">pointer</a>,
Dan Gohman0066db62008-06-18 18:42:13 +00001603 <a href="#t_vector">vector</a>,
Dan Gohmana334d5f2008-05-12 23:51:09 +00001604 <a href="#t_struct">structure</a>,
1605 <a href="#t_array">array</a>,
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001606 <a href="#t_label">label</a>,
1607 <a href="#t_metadata">metadata</a>.
Reid Spencerca86e162006-12-31 07:07:53 +00001608 </td>
Chris Lattner261efe92003-11-25 01:02:51 +00001609 </tr>
Chris Lattner4f69f462008-01-04 04:32:38 +00001610 <tr>
1611 <td><a href="#t_primitive">primitive</a></td>
1612 <td><a href="#t_label">label</a>,
1613 <a href="#t_void">void</a>,
Tobias Grosser05387292010-12-28 20:29:31 +00001614 <a href="#t_integer">integer</a>,
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001615 <a href="#t_floating">floating point</a>,
Dale Johannesen21fe99b2010-10-01 00:48:59 +00001616 <a href="#t_x86mmx">x86mmx</a>,
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001617 <a href="#t_metadata">metadata</a>.</td>
Chris Lattner4f69f462008-01-04 04:32:38 +00001618 </tr>
1619 <tr>
1620 <td><a href="#t_derived">derived</a></td>
Chris Lattnerfdfeb692010-02-12 20:49:41 +00001621 <td><a href="#t_array">array</a>,
Chris Lattner4f69f462008-01-04 04:32:38 +00001622 <a href="#t_function">function</a>,
1623 <a href="#t_pointer">pointer</a>,
1624 <a href="#t_struct">structure</a>,
Chris Lattner4f69f462008-01-04 04:32:38 +00001625 <a href="#t_vector">vector</a>,
1626 <a href="#t_opaque">opaque</a>.
Dan Gohman01ac1012008-10-14 16:32:04 +00001627 </td>
Chris Lattner4f69f462008-01-04 04:32:38 +00001628 </tr>
Chris Lattner261efe92003-11-25 01:02:51 +00001629 </tbody>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001630</table>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00001631
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001632<p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1633 important. Values of these types are the only ones which can be produced by
Nick Lewyckyec38da42009-09-27 00:45:11 +00001634 instructions.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001635
Misha Brukman9d0919f2003-11-08 01:05:38 +00001636</div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001637
Chris Lattner00950542001-06-06 20:29:01 +00001638<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001639<h3>
1640 <a name="t_primitive">Primitive Types</a>
1641</h3>
Chris Lattner8f8c7b72008-01-04 04:34:14 +00001642
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001643<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001644
Chris Lattner4f69f462008-01-04 04:32:38 +00001645<p>The primitive types are the fundamental building blocks of the LLVM
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001646 system.</p>
Chris Lattner4f69f462008-01-04 04:32:38 +00001647
1648<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001649<h4>
1650 <a name="t_integer">Integer Type</a>
1651</h4>
Nick Lewyckyec38da42009-09-27 00:45:11 +00001652
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001653<div>
Nick Lewyckyec38da42009-09-27 00:45:11 +00001654
1655<h5>Overview:</h5>
1656<p>The integer type is a very simple type that simply specifies an arbitrary
1657 bit width for the integer type desired. Any bit width from 1 bit to
1658 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
1659
1660<h5>Syntax:</h5>
1661<pre>
1662 iN
1663</pre>
1664
1665<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1666 value.</p>
1667
1668<h5>Examples:</h5>
1669<table class="layout">
1670 <tr class="layout">
1671 <td class="left"><tt>i1</tt></td>
1672 <td class="left">a single-bit integer.</td>
1673 </tr>
1674 <tr class="layout">
1675 <td class="left"><tt>i32</tt></td>
1676 <td class="left">a 32-bit integer.</td>
1677 </tr>
1678 <tr class="layout">
1679 <td class="left"><tt>i1942652</tt></td>
1680 <td class="left">a really big integer of over 1 million bits.</td>
1681 </tr>
1682</table>
1683
Nick Lewyckyec38da42009-09-27 00:45:11 +00001684</div>
1685
1686<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001687<h4>
1688 <a name="t_floating">Floating Point Types</a>
1689</h4>
Chris Lattner4f69f462008-01-04 04:32:38 +00001690
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001691<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001692
1693<table>
1694 <tbody>
1695 <tr><th>Type</th><th>Description</th></tr>
1696 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1697 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1698 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1699 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1700 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1701 </tbody>
1702</table>
1703
Chris Lattner4f69f462008-01-04 04:32:38 +00001704</div>
1705
1706<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001707<h4>
1708 <a name="t_x86mmx">X86mmx Type</a>
1709</h4>
Dale Johannesen21fe99b2010-10-01 00:48:59 +00001710
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001711<div>
Dale Johannesen21fe99b2010-10-01 00:48:59 +00001712
1713<h5>Overview:</h5>
1714<p>The x86mmx type represents a value held in an MMX register on an x86 machine. The operations allowed on it are quite limited: parameters and return values, load and store, and bitcast. User-specified MMX instructions are represented as intrinsic or asm calls with arguments and/or results of this type. There are no arrays, vectors or constants of this type.</p>
1715
1716<h5>Syntax:</h5>
1717<pre>
Dale Johannesen473a8c82010-10-01 01:07:02 +00001718 x86mmx
Dale Johannesen21fe99b2010-10-01 00:48:59 +00001719</pre>
1720
1721</div>
1722
1723<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001724<h4>
1725 <a name="t_void">Void Type</a>
1726</h4>
Chris Lattner4f69f462008-01-04 04:32:38 +00001727
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001728<div>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001729
Chris Lattner4f69f462008-01-04 04:32:38 +00001730<h5>Overview:</h5>
1731<p>The void type does not represent any value and has no size.</p>
1732
1733<h5>Syntax:</h5>
Chris Lattner4f69f462008-01-04 04:32:38 +00001734<pre>
1735 void
1736</pre>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001737
Chris Lattner4f69f462008-01-04 04:32:38 +00001738</div>
1739
1740<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001741<h4>
1742 <a name="t_label">Label Type</a>
1743</h4>
Chris Lattner4f69f462008-01-04 04:32:38 +00001744
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001745<div>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001746
Chris Lattner4f69f462008-01-04 04:32:38 +00001747<h5>Overview:</h5>
1748<p>The label type represents code labels.</p>
1749
1750<h5>Syntax:</h5>
Chris Lattner4f69f462008-01-04 04:32:38 +00001751<pre>
1752 label
1753</pre>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001754
Chris Lattner4f69f462008-01-04 04:32:38 +00001755</div>
1756
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001757<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001758<h4>
1759 <a name="t_metadata">Metadata Type</a>
1760</h4>
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001761
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001762<div>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001763
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001764<h5>Overview:</h5>
Nick Lewyckyc261df92009-09-27 23:27:42 +00001765<p>The metadata type represents embedded metadata. No derived types may be
1766 created from metadata except for <a href="#t_function">function</a>
1767 arguments.
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001768
1769<h5>Syntax:</h5>
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001770<pre>
1771 metadata
1772</pre>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001773
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001774</div>
1775
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001776</div>
Chris Lattner4f69f462008-01-04 04:32:38 +00001777
1778<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001779<h3>
1780 <a name="t_derived">Derived Types</a>
1781</h3>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001782
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001783<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001784
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001785<p>The real power in LLVM comes from the derived types in the system. This is
1786 what allows a programmer to represent arrays, functions, pointers, and other
Nick Lewyckyec38da42009-09-27 00:45:11 +00001787 useful types. Each of these types contain one or more element types which
1788 may be a primitive type, or another derived type. For example, it is
1789 possible to have a two dimensional array, using an array as the element type
1790 of another array.</p>
Dan Gohmand8791e52009-01-24 15:58:40 +00001791
Chris Lattner1afcace2011-07-09 17:41:24 +00001792</div>
1793
1794
Chris Lattnerfdfeb692010-02-12 20:49:41 +00001795<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001796<h4>
1797 <a name="t_aggregate">Aggregate Types</a>
1798</h4>
Chris Lattnerfdfeb692010-02-12 20:49:41 +00001799
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001800<div>
Chris Lattnerfdfeb692010-02-12 20:49:41 +00001801
1802<p>Aggregate Types are a subset of derived types that can contain multiple
1803 member types. <a href="#t_array">Arrays</a>,
Chris Lattner61c70e92010-08-28 04:09:24 +00001804 <a href="#t_struct">structs</a>, and <a href="#t_vector">vectors</a> are
1805 aggregate types.</p>
Chris Lattnerfdfeb692010-02-12 20:49:41 +00001806
1807</div>
1808
Reid Spencer2b916312007-05-16 18:44:01 +00001809<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001810<h4>
1811 <a name="t_array">Array Type</a>
1812</h4>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001813
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001814<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001815
Chris Lattner00950542001-06-06 20:29:01 +00001816<h5>Overview:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001817<p>The array type is a very simple derived type that arranges elements
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001818 sequentially in memory. The array type requires a size (number of elements)
1819 and an underlying data type.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001820
Chris Lattner7faa8832002-04-14 06:13:44 +00001821<h5>Syntax:</h5>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001822<pre>
1823 [&lt;# elements&gt; x &lt;elementtype&gt;]
1824</pre>
1825
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001826<p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1827 be any type with a size.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001828
Chris Lattner7faa8832002-04-14 06:13:44 +00001829<h5>Examples:</h5>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001830<table class="layout">
1831 <tr class="layout">
Chris Lattner23ff1f92007-12-19 05:04:11 +00001832 <td class="left"><tt>[40 x i32]</tt></td>
1833 <td class="left">Array of 40 32-bit integer values.</td>
1834 </tr>
1835 <tr class="layout">
1836 <td class="left"><tt>[41 x i32]</tt></td>
1837 <td class="left">Array of 41 32-bit integer values.</td>
1838 </tr>
1839 <tr class="layout">
1840 <td class="left"><tt>[4 x i8]</tt></td>
1841 <td class="left">Array of 4 8-bit integer values.</td>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001842 </tr>
Chris Lattner00950542001-06-06 20:29:01 +00001843</table>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001844<p>Here are some examples of multidimensional arrays:</p>
1845<table class="layout">
1846 <tr class="layout">
Chris Lattner23ff1f92007-12-19 05:04:11 +00001847 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1848 <td class="left">3x4 array of 32-bit integer values.</td>
1849 </tr>
1850 <tr class="layout">
1851 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1852 <td class="left">12x10 array of single precision floating point values.</td>
1853 </tr>
1854 <tr class="layout">
1855 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1856 <td class="left">2x3x4 array of 16-bit integer values.</td>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001857 </tr>
1858</table>
Chris Lattnere67a9512005-06-24 17:22:57 +00001859
Dan Gohman7657f6b2009-11-09 19:01:53 +00001860<p>There is no restriction on indexing beyond the end of the array implied by
1861 a static type (though there are restrictions on indexing beyond the bounds
1862 of an allocated object in some cases). This means that single-dimension
1863 'variable sized array' addressing can be implemented in LLVM with a zero
1864 length array type. An implementation of 'pascal style arrays' in LLVM could
1865 use the type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
Chris Lattnere67a9512005-06-24 17:22:57 +00001866
Misha Brukman9d0919f2003-11-08 01:05:38 +00001867</div>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001868
Chris Lattner00950542001-06-06 20:29:01 +00001869<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001870<h4>
1871 <a name="t_function">Function Type</a>
1872</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001873
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001874<div>
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00001875
Chris Lattner00950542001-06-06 20:29:01 +00001876<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001877<p>The function type can be thought of as a function signature. It consists of
1878 a return type and a list of formal parameter types. The return type of a
Chris Lattner61c70e92010-08-28 04:09:24 +00001879 function type is a first class type or a void type.</p>
Devang Patelc3fc6df2008-03-10 20:49:15 +00001880
Chris Lattner00950542001-06-06 20:29:01 +00001881<h5>Syntax:</h5>
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00001882<pre>
Nick Lewycky51386942009-09-27 07:55:32 +00001883 &lt;returntype&gt; (&lt;parameter list&gt;)
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00001884</pre>
1885
John Criswell0ec250c2005-10-24 16:17:18 +00001886<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001887 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1888 which indicates that the function takes a variable number of arguments.
1889 Variable argument functions can access their arguments with
1890 the <a href="#int_varargs">variable argument handling intrinsic</a>
Chris Lattner0724fbd2010-03-02 06:36:51 +00001891 functions. '<tt>&lt;returntype&gt;</tt>' is any type except
Nick Lewyckyc261df92009-09-27 23:27:42 +00001892 <a href="#t_label">label</a>.</p>
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00001893
Chris Lattner00950542001-06-06 20:29:01 +00001894<h5>Examples:</h5>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001895<table class="layout">
1896 <tr class="layout">
Reid Spencer92f82302006-12-31 07:18:34 +00001897 <td class="left"><tt>i32 (i32)</tt></td>
1898 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001899 </td>
Reid Spencer92f82302006-12-31 07:18:34 +00001900 </tr><tr class="layout">
Chris Lattner0724fbd2010-03-02 06:36:51 +00001901 <td class="left"><tt>float&nbsp;(i16,&nbsp;i32&nbsp;*)&nbsp;*
Reid Spencerf17a0b72006-12-31 07:20:23 +00001902 </tt></td>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00001903 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
Chris Lattner0724fbd2010-03-02 06:36:51 +00001904 an <tt>i16</tt> and a <a href="#t_pointer">pointer</a> to <tt>i32</tt>,
1905 returning <tt>float</tt>.
Reid Spencer92f82302006-12-31 07:18:34 +00001906 </td>
1907 </tr><tr class="layout">
1908 <td class="left"><tt>i32 (i8*, ...)</tt></td>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00001909 <td class="left">A vararg function that takes at least one
1910 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1911 which returns an integer. This is the signature for <tt>printf</tt> in
Reid Spencer92f82302006-12-31 07:18:34 +00001912 LLVM.
Reid Spencerd3f876c2004-11-01 08:19:36 +00001913 </td>
Devang Patela582f402008-03-24 05:35:41 +00001914 </tr><tr class="layout">
1915 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Nick Lewycky51386942009-09-27 07:55:32 +00001916 <td class="left">A function taking an <tt>i32</tt>, returning a
1917 <a href="#t_struct">structure</a> containing two <tt>i32</tt> values
Devang Patela582f402008-03-24 05:35:41 +00001918 </td>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001919 </tr>
Chris Lattner00950542001-06-06 20:29:01 +00001920</table>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00001921
Misha Brukman9d0919f2003-11-08 01:05:38 +00001922</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001923
Chris Lattner00950542001-06-06 20:29:01 +00001924<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001925<h4>
1926 <a name="t_struct">Structure Type</a>
1927</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001928
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001929<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001930
Chris Lattner00950542001-06-06 20:29:01 +00001931<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001932<p>The structure type is used to represent a collection of data members together
Chris Lattner1afcace2011-07-09 17:41:24 +00001933 in memory. The elements of a structure may be any type that has a size.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001934
Jeffrey Yasskin7a088cf2010-01-11 19:19:26 +00001935<p>Structures in memory are accessed using '<tt><a href="#i_load">load</a></tt>'
1936 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field
1937 with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
1938 Structures in registers are accessed using the
1939 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' and
1940 '<tt><a href="#i_insertvalue">insertvalue</a></tt>' instructions.</p>
Chris Lattner1afcace2011-07-09 17:41:24 +00001941
1942<p>Structures may optionally be "packed" structures, which indicate that the
1943 alignment of the struct is one byte, and that there is no padding between
1944 the elements. In non-packed structs, padding between field types is defined
1945 by the target data string to match the underlying processor.</p>
1946
1947<p>Structures can either be "anonymous" or "named". An anonymous structure is
1948 defined inline with other types (e.g. <tt>{i32, i32}*</tt>) and a named types
1949 are always defined at the top level with a name. Anonmyous types are uniqued
1950 by their contents and can never be recursive since there is no way to write
1951 one. Named types can be recursive.
1952</p>
1953
Chris Lattner00950542001-06-06 20:29:01 +00001954<h5>Syntax:</h5>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001955<pre>
Chris Lattner1afcace2011-07-09 17:41:24 +00001956 %T1 = type { &lt;type list&gt; } <i>; Named normal struct type</i>
1957 %T2 = type &lt;{ &lt;type list&gt; }&gt; <i>; Named packed struct type</i>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001958</pre>
Chris Lattner1afcace2011-07-09 17:41:24 +00001959
Chris Lattner00950542001-06-06 20:29:01 +00001960<h5>Examples:</h5>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001961<table class="layout">
1962 <tr class="layout">
Jeff Cohen6f1cc772007-04-22 01:17:39 +00001963 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1964 <td class="left">A triple of three <tt>i32</tt> values</td>
Chris Lattner1afcace2011-07-09 17:41:24 +00001965 </tr>
1966 <tr class="layout">
Jeff Cohen6f1cc772007-04-22 01:17:39 +00001967 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1968 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1969 second element is a <a href="#t_pointer">pointer</a> to a
1970 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1971 an <tt>i32</tt>.</td>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001972 </tr>
Chris Lattner1afcace2011-07-09 17:41:24 +00001973 <tr class="layout">
1974 <td class="left"><tt>&lt;{ i8, i32 }&gt;</tt></td>
1975 <td class="left">A packed struct known to be 5 bytes in size.</td>
1976 </tr>
Chris Lattner00950542001-06-06 20:29:01 +00001977</table>
Dan Gohmand8791e52009-01-24 15:58:40 +00001978
Misha Brukman9d0919f2003-11-08 01:05:38 +00001979</div>
Chris Lattner1afcace2011-07-09 17:41:24 +00001980
Chris Lattner00950542001-06-06 20:29:01 +00001981<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001982<h4>
Chris Lattner1afcace2011-07-09 17:41:24 +00001983 <a name="t_opaque">Opaque Type</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001984</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001985
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001986<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001987
Andrew Lenharth75e10682006-12-08 17:13:00 +00001988<h5>Overview:</h5>
Chris Lattner1afcace2011-07-09 17:41:24 +00001989<p>Opaque types are used to represent named structure types that do not have a
1990 body specified. This corresponds (for example) to the C notion of a forward
1991 declared structure.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001992
Andrew Lenharth75e10682006-12-08 17:13:00 +00001993<h5>Syntax:</h5>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001994<pre>
Chris Lattner1afcace2011-07-09 17:41:24 +00001995 %X = type opaque
1996 %52 = type opaque
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001997</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001998
Andrew Lenharth75e10682006-12-08 17:13:00 +00001999<h5>Examples:</h5>
2000<table class="layout">
2001 <tr class="layout">
Chris Lattner1afcace2011-07-09 17:41:24 +00002002 <td class="left"><tt>opaque</tt></td>
2003 <td class="left">An opaque type.</td>
Andrew Lenharth75e10682006-12-08 17:13:00 +00002004 </tr>
2005</table>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002006
Andrew Lenharth75e10682006-12-08 17:13:00 +00002007</div>
2008
Chris Lattner1afcace2011-07-09 17:41:24 +00002009
2010
Andrew Lenharth75e10682006-12-08 17:13:00 +00002011<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002012<h4>
2013 <a name="t_pointer">Pointer Type</a>
2014</h4>
Chris Lattner0fd4a272009-02-08 19:53:29 +00002015
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002016<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002017
2018<h5>Overview:</h5>
Dan Gohmanff3ef322010-02-25 16:50:07 +00002019<p>The pointer type is used to specify memory locations.
2020 Pointers are commonly used to reference objects in memory.</p>
2021
2022<p>Pointer types may have an optional address space attribute defining the
2023 numbered address space where the pointed-to object resides. The default
2024 address space is number zero. The semantics of non-zero address
2025 spaces are target-specific.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002026
2027<p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
2028 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
Chris Lattner0fd4a272009-02-08 19:53:29 +00002029
Chris Lattner7faa8832002-04-14 06:13:44 +00002030<h5>Syntax:</h5>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00002031<pre>
2032 &lt;type&gt; *
2033</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002034
Chris Lattner7faa8832002-04-14 06:13:44 +00002035<h5>Examples:</h5>
Reid Spencerd3f876c2004-11-01 08:19:36 +00002036<table class="layout">
2037 <tr class="layout">
Dan Gohman2a08c532009-01-04 23:44:43 +00002038 <td class="left"><tt>[4 x i32]*</tt></td>
Chris Lattner23ff1f92007-12-19 05:04:11 +00002039 <td class="left">A <a href="#t_pointer">pointer</a> to <a
2040 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
2041 </tr>
2042 <tr class="layout">
Dan Gohmanfe47aae2010-05-28 17:13:49 +00002043 <td class="left"><tt>i32 (i32*) *</tt></td>
Chris Lattner23ff1f92007-12-19 05:04:11 +00002044 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Reid Spencerca86e162006-12-31 07:07:53 +00002045 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner23ff1f92007-12-19 05:04:11 +00002046 <tt>i32</tt>.</td>
2047 </tr>
2048 <tr class="layout">
2049 <td class="left"><tt>i32 addrspace(5)*</tt></td>
2050 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
2051 that resides in address space #5.</td>
Reid Spencerd3f876c2004-11-01 08:19:36 +00002052 </tr>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002053</table>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002054
Misha Brukman9d0919f2003-11-08 01:05:38 +00002055</div>
Reid Spencerd3f876c2004-11-01 08:19:36 +00002056
Chris Lattnera58561b2004-08-12 19:12:28 +00002057<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002058<h4>
2059 <a name="t_vector">Vector Type</a>
2060</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002061
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002062<div>
Chris Lattner69c11bb2005-04-25 17:34:15 +00002063
Chris Lattnera58561b2004-08-12 19:12:28 +00002064<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002065<p>A vector type is a simple derived type that represents a vector of elements.
2066 Vector types are used when multiple primitive data are operated in parallel
2067 using a single instruction (SIMD). A vector type requires a size (number of
Duncan Sandsd40d14e2009-11-27 13:38:03 +00002068 elements) and an underlying primitive data type. Vector types are considered
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002069 <a href="#t_firstclass">first class</a>.</p>
Chris Lattner69c11bb2005-04-25 17:34:15 +00002070
Chris Lattnera58561b2004-08-12 19:12:28 +00002071<h5>Syntax:</h5>
Chris Lattner69c11bb2005-04-25 17:34:15 +00002072<pre>
2073 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
2074</pre>
2075
Chris Lattner7d2e7be2010-10-10 18:20:35 +00002076<p>The number of elements is a constant integer value larger than 0; elementtype
2077 may be any integer or floating point type. Vectors of size zero are not
2078 allowed, and pointers are not allowed as the element type.</p>
Chris Lattner69c11bb2005-04-25 17:34:15 +00002079
Chris Lattnera58561b2004-08-12 19:12:28 +00002080<h5>Examples:</h5>
Reid Spencerd3f876c2004-11-01 08:19:36 +00002081<table class="layout">
2082 <tr class="layout">
Chris Lattner23ff1f92007-12-19 05:04:11 +00002083 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
2084 <td class="left">Vector of 4 32-bit integer values.</td>
2085 </tr>
2086 <tr class="layout">
2087 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
2088 <td class="left">Vector of 8 32-bit floating-point values.</td>
2089 </tr>
2090 <tr class="layout">
2091 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
2092 <td class="left">Vector of 2 64-bit integer values.</td>
Reid Spencerd3f876c2004-11-01 08:19:36 +00002093 </tr>
2094</table>
Dan Gohmand8791e52009-01-24 15:58:40 +00002095
Misha Brukman9d0919f2003-11-08 01:05:38 +00002096</div>
2097
Chris Lattnerc3f59762004-12-09 17:30:23 +00002098<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002099<h2><a name="constants">Constants</a></h2>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002100<!-- *********************************************************************** -->
2101
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002102<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002103
2104<p>LLVM has several different basic types of constants. This section describes
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002105 them all and their syntax.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002106
Chris Lattnerc3f59762004-12-09 17:30:23 +00002107<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002108<h3>
2109 <a name="simpleconstants">Simple Constants</a>
2110</h3>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002111
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002112<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002113
2114<dl>
2115 <dt><b>Boolean constants</b></dt>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002116 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
Nick Lewyckyec38da42009-09-27 00:45:11 +00002117 constants of the <tt><a href="#t_integer">i1</a></tt> type.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002118
2119 <dt><b>Integer constants</b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002120 <dd>Standard integers (such as '4') are constants of
2121 the <a href="#t_integer">integer</a> type. Negative numbers may be used
2122 with integer types.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002123
2124 <dt><b>Floating point constants</b></dt>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002125 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002126 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
2127 notation (see below). The assembler requires the exact decimal value of a
2128 floating-point constant. For example, the assembler accepts 1.25 but
2129 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
2130 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002131
2132 <dt><b>Null pointer constants</b></dt>
John Criswell9e2485c2004-12-10 15:51:16 +00002133 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002134 and must be of <a href="#t_pointer">pointer type</a>.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002135</dl>
2136
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002137<p>The one non-intuitive notation for constants is the hexadecimal form of
2138 floating point constants. For example, the form '<tt>double
2139 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
2140 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
2141 constants are required (and the only time that they are generated by the
2142 disassembler) is when a floating point constant must be emitted but it cannot
2143 be represented as a decimal floating point number in a reasonable number of
2144 digits. For example, NaN's, infinities, and other special values are
2145 represented in their IEEE hexadecimal format so that assembly and disassembly
2146 do not cause any bits to change in the constants.</p>
2147
Dale Johannesenbd5e5a82009-02-11 22:14:51 +00002148<p>When using the hexadecimal form, constants of types float and double are
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002149 represented using the 16-digit form shown above (which matches the IEEE754
2150 representation for double); float values must, however, be exactly
2151 representable as IEE754 single precision. Hexadecimal format is always used
2152 for long double, and there are three forms of long double. The 80-bit format
2153 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
2154 The 128-bit format used by PowerPC (two adjacent doubles) is represented
2155 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
2156 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
2157 currently supported target uses this format. Long doubles will only work if
2158 they match the long double format on your target. All hexadecimal formats
2159 are big-endian (sign bit at the left).</p>
2160
Dale Johannesen21fe99b2010-10-01 00:48:59 +00002161<p>There are no constants of type x86mmx.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002162</div>
2163
2164<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002165<h3>
Bill Wendlingd9fe2982009-07-20 02:32:41 +00002166<a name="aggregateconstants"></a> <!-- old anchor -->
2167<a name="complexconstants">Complex Constants</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002168</h3>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002169
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002170<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002171
Chris Lattner70882792009-02-28 18:32:25 +00002172<p>Complex constants are a (potentially recursive) combination of simple
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002173 constants and smaller complex constants.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002174
2175<dl>
2176 <dt><b>Structure constants</b></dt>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002177 <dd>Structure constants are represented with notation similar to structure
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002178 type definitions (a comma separated list of elements, surrounded by braces
2179 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
2180 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
2181 Structure constants must have <a href="#t_struct">structure type</a>, and
2182 the number and types of elements must match those specified by the
2183 type.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002184
2185 <dt><b>Array constants</b></dt>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002186 <dd>Array constants are represented with notation similar to array type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002187 definitions (a comma separated list of elements, surrounded by square
2188 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
2189 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
2190 the number and types of elements must match those specified by the
2191 type.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002192
Reid Spencer485bad12007-02-15 03:07:05 +00002193 <dt><b>Vector constants</b></dt>
Reid Spencer485bad12007-02-15 03:07:05 +00002194 <dd>Vector constants are represented with notation similar to vector type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002195 definitions (a comma separated list of elements, surrounded by
2196 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32
2197 42, i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must
2198 have <a href="#t_vector">vector type</a>, and the number and types of
2199 elements must match those specified by the type.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002200
2201 <dt><b>Zero initialization</b></dt>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002202 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
Chris Lattnerfdfeb692010-02-12 20:49:41 +00002203 value to zero of <em>any</em> type, including scalar and
2204 <a href="#t_aggregate">aggregate</a> types.
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002205 This is often used to avoid having to print large zero initializers
2206 (e.g. for large arrays) and is always exactly equivalent to using explicit
2207 zero initializers.</dd>
Nick Lewycky21cc4462009-04-04 07:22:01 +00002208
2209 <dt><b>Metadata node</b></dt>
Nick Lewycky1e8c7a62009-05-30 16:08:30 +00002210 <dd>A metadata node is a structure-like constant with
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002211 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
2212 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
2213 be interpreted as part of the instruction stream, metadata is a place to
2214 attach additional information such as debug info.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002215</dl>
2216
2217</div>
2218
2219<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002220<h3>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002221 <a name="globalconstants">Global Variable and Function Addresses</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002222</h3>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002223
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002224<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002225
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002226<p>The addresses of <a href="#globalvars">global variables</a>
2227 and <a href="#functionstructure">functions</a> are always implicitly valid
2228 (link-time) constants. These constants are explicitly referenced when
2229 the <a href="#identifiers">identifier for the global</a> is used and always
2230 have <a href="#t_pointer">pointer</a> type. For example, the following is a
2231 legal LLVM file:</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002232
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002233<pre class="doc_code">
Chris Lattnera18a4242007-06-06 18:28:13 +00002234@X = global i32 17
2235@Y = global i32 42
2236@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
Chris Lattnerc3f59762004-12-09 17:30:23 +00002237</pre>
2238
2239</div>
2240
2241<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002242<h3>
2243 <a name="undefvalues">Undefined Values</a>
2244</h3>
2245
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002246<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002247
Chris Lattner48a109c2009-09-07 22:52:39 +00002248<p>The string '<tt>undef</tt>' can be used anywhere a constant is expected, and
Benjamin Kramer8040cd32009-10-12 14:46:08 +00002249 indicates that the user of the value may receive an unspecified bit-pattern.
Bill Wendling1b383ba2010-10-27 01:07:41 +00002250 Undefined values may be of any type (other than '<tt>label</tt>'
2251 or '<tt>void</tt>') and be used anywhere a constant is permitted.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002252
Chris Lattnerc608cb12009-09-11 01:49:31 +00002253<p>Undefined values are useful because they indicate to the compiler that the
Chris Lattner48a109c2009-09-07 22:52:39 +00002254 program is well defined no matter what value is used. This gives the
2255 compiler more freedom to optimize. Here are some examples of (potentially
2256 surprising) transformations that are valid (in pseudo IR):</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002257
Chris Lattner48a109c2009-09-07 22:52:39 +00002258
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002259<pre class="doc_code">
Chris Lattner48a109c2009-09-07 22:52:39 +00002260 %A = add %X, undef
2261 %B = sub %X, undef
2262 %C = xor %X, undef
2263Safe:
2264 %A = undef
2265 %B = undef
2266 %C = undef
2267</pre>
Chris Lattner48a109c2009-09-07 22:52:39 +00002268
2269<p>This is safe because all of the output bits are affected by the undef bits.
Bill Wendling1b383ba2010-10-27 01:07:41 +00002270 Any output bit can have a zero or one depending on the input bits.</p>
Chris Lattner48a109c2009-09-07 22:52:39 +00002271
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002272<pre class="doc_code">
Chris Lattner48a109c2009-09-07 22:52:39 +00002273 %A = or %X, undef
2274 %B = and %X, undef
2275Safe:
2276 %A = -1
2277 %B = 0
2278Unsafe:
2279 %A = undef
2280 %B = undef
2281</pre>
Chris Lattner48a109c2009-09-07 22:52:39 +00002282
2283<p>These logical operations have bits that are not always affected by the input.
Bill Wendling1b383ba2010-10-27 01:07:41 +00002284 For example, if <tt>%X</tt> has a zero bit, then the output of the
2285 '<tt>and</tt>' operation will always be a zero for that bit, no matter what
2286 the corresponding bit from the '<tt>undef</tt>' is. As such, it is unsafe to
2287 optimize or assume that the result of the '<tt>and</tt>' is '<tt>undef</tt>'.
2288 However, it is safe to assume that all bits of the '<tt>undef</tt>' could be
2289 0, and optimize the '<tt>and</tt>' to 0. Likewise, it is safe to assume that
2290 all the bits of the '<tt>undef</tt>' operand to the '<tt>or</tt>' could be
2291 set, allowing the '<tt>or</tt>' to be folded to -1.</p>
Chris Lattner48a109c2009-09-07 22:52:39 +00002292
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002293<pre class="doc_code">
Chris Lattner48a109c2009-09-07 22:52:39 +00002294 %A = select undef, %X, %Y
2295 %B = select undef, 42, %Y
2296 %C = select %X, %Y, undef
2297Safe:
2298 %A = %X (or %Y)
2299 %B = 42 (or %Y)
2300 %C = %Y
2301Unsafe:
2302 %A = undef
2303 %B = undef
2304 %C = undef
2305</pre>
Chris Lattner48a109c2009-09-07 22:52:39 +00002306
Bill Wendling1b383ba2010-10-27 01:07:41 +00002307<p>This set of examples shows that undefined '<tt>select</tt>' (and conditional
2308 branch) conditions can go <em>either way</em>, but they have to come from one
2309 of the two operands. In the <tt>%A</tt> example, if <tt>%X</tt> and
2310 <tt>%Y</tt> were both known to have a clear low bit, then <tt>%A</tt> would
2311 have to have a cleared low bit. However, in the <tt>%C</tt> example, the
2312 optimizer is allowed to assume that the '<tt>undef</tt>' operand could be the
2313 same as <tt>%Y</tt>, allowing the whole '<tt>select</tt>' to be
2314 eliminated.</p>
Chris Lattner48a109c2009-09-07 22:52:39 +00002315
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002316<pre class="doc_code">
Chris Lattner48a109c2009-09-07 22:52:39 +00002317 %A = xor undef, undef
Eric Christopher6c7e8a02009-12-05 02:46:03 +00002318
Chris Lattner48a109c2009-09-07 22:52:39 +00002319 %B = undef
2320 %C = xor %B, %B
2321
2322 %D = undef
2323 %E = icmp lt %D, 4
2324 %F = icmp gte %D, 4
2325
2326Safe:
2327 %A = undef
2328 %B = undef
2329 %C = undef
2330 %D = undef
2331 %E = undef
2332 %F = undef
2333</pre>
Chris Lattner48a109c2009-09-07 22:52:39 +00002334
Bill Wendling1b383ba2010-10-27 01:07:41 +00002335<p>This example points out that two '<tt>undef</tt>' operands are not
2336 necessarily the same. This can be surprising to people (and also matches C
2337 semantics) where they assume that "<tt>X^X</tt>" is always zero, even
2338 if <tt>X</tt> is undefined. This isn't true for a number of reasons, but the
2339 short answer is that an '<tt>undef</tt>' "variable" can arbitrarily change
2340 its value over its "live range". This is true because the variable doesn't
2341 actually <em>have a live range</em>. Instead, the value is logically read
2342 from arbitrary registers that happen to be around when needed, so the value
2343 is not necessarily consistent over time. In fact, <tt>%A</tt> and <tt>%C</tt>
2344 need to have the same semantics or the core LLVM "replace all uses with"
2345 concept would not hold.</p>
Chris Lattner6e9057b2009-09-07 23:33:52 +00002346
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002347<pre class="doc_code">
Chris Lattner6e9057b2009-09-07 23:33:52 +00002348 %A = fdiv undef, %X
2349 %B = fdiv %X, undef
2350Safe:
2351 %A = undef
2352b: unreachable
2353</pre>
Chris Lattner6e9057b2009-09-07 23:33:52 +00002354
2355<p>These examples show the crucial difference between an <em>undefined
Bill Wendling1b383ba2010-10-27 01:07:41 +00002356 value</em> and <em>undefined behavior</em>. An undefined value (like
2357 '<tt>undef</tt>') is allowed to have an arbitrary bit-pattern. This means that
2358 the <tt>%A</tt> operation can be constant folded to '<tt>undef</tt>', because
2359 the '<tt>undef</tt>' could be an SNaN, and <tt>fdiv</tt> is not (currently)
2360 defined on SNaN's. However, in the second example, we can make a more
2361 aggressive assumption: because the <tt>undef</tt> is allowed to be an
2362 arbitrary value, we are allowed to assume that it could be zero. Since a
2363 divide by zero has <em>undefined behavior</em>, we are allowed to assume that
2364 the operation does not execute at all. This allows us to delete the divide and
2365 all code after it. Because the undefined operation "can't happen", the
2366 optimizer can assume that it occurs in dead code.</p>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00002367
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002368<pre class="doc_code">
Chris Lattner6e9057b2009-09-07 23:33:52 +00002369a: store undef -> %X
2370b: store %X -> undef
2371Safe:
2372a: &lt;deleted&gt;
2373b: unreachable
2374</pre>
Chris Lattner6e9057b2009-09-07 23:33:52 +00002375
Bill Wendling1b383ba2010-10-27 01:07:41 +00002376<p>These examples reiterate the <tt>fdiv</tt> example: a store <em>of</em> an
2377 undefined value can be assumed to not have any effect; we can assume that the
2378 value is overwritten with bits that happen to match what was already there.
2379 However, a store <em>to</em> an undefined location could clobber arbitrary
2380 memory, therefore, it has undefined behavior.</p>
Chris Lattner6e9057b2009-09-07 23:33:52 +00002381
Chris Lattnerc3f59762004-12-09 17:30:23 +00002382</div>
2383
2384<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002385<h3>
2386 <a name="trapvalues">Trap Values</a>
2387</h3>
2388
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002389<div>
Dan Gohmanfff6c532010-04-22 23:14:21 +00002390
Dan Gohmanc68ce062010-04-26 20:21:21 +00002391<p>Trap values are similar to <a href="#undefvalues">undef values</a>, however
Dan Gohmanfff6c532010-04-22 23:14:21 +00002392 instead of representing an unspecified bit pattern, they represent the
2393 fact that an instruction or constant expression which cannot evoke side
2394 effects has nevertheless detected a condition which results in undefined
Dan Gohmanc68ce062010-04-26 20:21:21 +00002395 behavior.</p>
Dan Gohmanfff6c532010-04-22 23:14:21 +00002396
Dan Gohman34b3d992010-04-28 00:49:41 +00002397<p>There is currently no way of representing a trap value in the IR; they
Dan Gohman855abed2010-05-03 14:51:43 +00002398 only exist when produced by operations such as
Dan Gohman34b3d992010-04-28 00:49:41 +00002399 <a href="#i_add"><tt>add</tt></a> with the <tt>nsw</tt> flag.</p>
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002400
Dan Gohman34b3d992010-04-28 00:49:41 +00002401<p>Trap value behavior is defined in terms of value <i>dependence</i>:</p>
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002402
Dan Gohman34b3d992010-04-28 00:49:41 +00002403<ul>
2404<li>Values other than <a href="#i_phi"><tt>phi</tt></a> nodes depend on
2405 their operands.</li>
2406
2407<li><a href="#i_phi"><tt>Phi</tt></a> nodes depend on the operand corresponding
2408 to their dynamic predecessor basic block.</li>
2409
2410<li>Function arguments depend on the corresponding actual argument values in
2411 the dynamic callers of their functions.</li>
2412
2413<li><a href="#i_call"><tt>Call</tt></a> instructions depend on the
2414 <a href="#i_ret"><tt>ret</tt></a> instructions that dynamically transfer
2415 control back to them.</li>
2416
Dan Gohmanb5328162010-05-03 14:55:22 +00002417<li><a href="#i_invoke"><tt>Invoke</tt></a> instructions depend on the
2418 <a href="#i_ret"><tt>ret</tt></a>, <a href="#i_unwind"><tt>unwind</tt></a>,
2419 or exception-throwing call instructions that dynamically transfer control
2420 back to them.</li>
2421
Dan Gohman34b3d992010-04-28 00:49:41 +00002422<li>Non-volatile loads and stores depend on the most recent stores to all of the
2423 referenced memory addresses, following the order in the IR
2424 (including loads and stores implied by intrinsics such as
2425 <a href="#int_memcpy"><tt>@llvm.memcpy</tt></a>.)</li>
2426
Dan Gohman7c24ff12010-05-03 14:59:34 +00002427<!-- TODO: In the case of multiple threads, this only applies if the store
2428 "happens-before" the load or store. -->
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002429
Dan Gohman34b3d992010-04-28 00:49:41 +00002430<!-- TODO: floating-point exception state -->
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002431
Dan Gohman34b3d992010-04-28 00:49:41 +00002432<li>An instruction with externally visible side effects depends on the most
2433 recent preceding instruction with externally visible side effects, following
Dan Gohmanff70fe42010-07-06 15:26:33 +00002434 the order in the IR. (This includes
2435 <a href="#volatile">volatile operations</a>.)</li>
Dan Gohmanfff6c532010-04-22 23:14:21 +00002436
Dan Gohmanb5328162010-05-03 14:55:22 +00002437<li>An instruction <i>control-depends</i> on a
2438 <a href="#terminators">terminator instruction</a>
2439 if the terminator instruction has multiple successors and the instruction
2440 is always executed when control transfers to one of the successors, and
Chris Lattner7a2bdde2011-04-15 05:18:47 +00002441 may not be executed when control is transferred to another.</li>
Dan Gohman34b3d992010-04-28 00:49:41 +00002442
Dan Gohmanca4cac42011-04-12 23:05:59 +00002443<li>Additionally, an instruction also <i>control-depends</i> on a terminator
2444 instruction if the set of instructions it otherwise depends on would be
Chris Lattner7a2bdde2011-04-15 05:18:47 +00002445 different if the terminator had transferred control to a different
Dan Gohmanca4cac42011-04-12 23:05:59 +00002446 successor.</li>
2447
Dan Gohman34b3d992010-04-28 00:49:41 +00002448<li>Dependence is transitive.</li>
2449
2450</ul>
Dan Gohman34b3d992010-04-28 00:49:41 +00002451
2452<p>Whenever a trap value is generated, all values which depend on it evaluate
2453 to trap. If they have side effects, the evoke their side effects as if each
2454 operand with a trap value were undef. If they have externally-visible side
2455 effects, the behavior is undefined.</p>
2456
2457<p>Here are some examples:</p>
Dan Gohmanc30f6e12010-04-26 20:54:53 +00002458
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002459<pre class="doc_code">
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002460entry:
2461 %trap = sub nuw i32 0, 1 ; Results in a trap value.
Dan Gohman34b3d992010-04-28 00:49:41 +00002462 %still_trap = and i32 %trap, 0 ; Whereas (and i32 undef, 0) would return 0.
2463 %trap_yet_again = getelementptr i32* @h, i32 %still_trap
2464 store i32 0, i32* %trap_yet_again ; undefined behavior
2465
2466 store i32 %trap, i32* @g ; Trap value conceptually stored to memory.
2467 %trap2 = load i32* @g ; Returns a trap value, not just undef.
2468
2469 volatile store i32 %trap, i32* @g ; External observation; undefined behavior.
2470
2471 %narrowaddr = bitcast i32* @g to i16*
2472 %wideaddr = bitcast i32* @g to i64*
Nick Lewycky64f9fb12011-05-16 19:29:30 +00002473 %trap3 = load i16* %narrowaddr ; Returns a trap value.
2474 %trap4 = load i64* %wideaddr ; Returns a trap value.
Dan Gohman34b3d992010-04-28 00:49:41 +00002475
Nick Lewycky64f9fb12011-05-16 19:29:30 +00002476 %cmp = icmp slt i32 %trap, 0 ; Returns a trap value.
2477 br i1 %cmp, label %true, label %end ; Branch to either destination.
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002478
2479true:
Dan Gohman34b3d992010-04-28 00:49:41 +00002480 volatile store i32 0, i32* @g ; This is control-dependent on %cmp, so
2481 ; it has undefined behavior.
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002482 br label %end
2483
2484end:
2485 %p = phi i32 [ 0, %entry ], [ 1, %true ]
2486 ; Both edges into this PHI are
2487 ; control-dependent on %cmp, so this
Dan Gohman34b3d992010-04-28 00:49:41 +00002488 ; always results in a trap value.
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002489
Dan Gohmanca4cac42011-04-12 23:05:59 +00002490 volatile store i32 0, i32* @g ; This would depend on the store in %true
2491 ; if %cmp is true, or the store in %entry
2492 ; otherwise, so this is undefined behavior.
2493
Nick Lewycky64f9fb12011-05-16 19:29:30 +00002494 br i1 %cmp, label %second_true, label %second_end
Dan Gohmanca4cac42011-04-12 23:05:59 +00002495 ; The same branch again, but this time the
2496 ; true block doesn't have side effects.
2497
2498second_true:
2499 ; No side effects!
Nick Lewycky64f9fb12011-05-16 19:29:30 +00002500 ret void
Dan Gohmanca4cac42011-04-12 23:05:59 +00002501
2502second_end:
2503 volatile store i32 0, i32* @g ; This time, the instruction always depends
2504 ; on the store in %end. Also, it is
2505 ; control-equivalent to %end, so this is
Nick Lewycky64f9fb12011-05-16 19:29:30 +00002506 ; well-defined (again, ignoring earlier
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002507 ; undefined behavior in this example).
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002508</pre>
Dan Gohmanfff6c532010-04-22 23:14:21 +00002509
Dan Gohmanfff6c532010-04-22 23:14:21 +00002510</div>
2511
2512<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002513<h3>
2514 <a name="blockaddress">Addresses of Basic Blocks</a>
2515</h3>
2516
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002517<div>
Chris Lattnerc6f44362009-10-27 21:01:34 +00002518
Chris Lattnercdfc9402009-11-01 01:27:45 +00002519<p><b><tt>blockaddress(@function, %block)</tt></b></p>
Chris Lattnerc6f44362009-10-27 21:01:34 +00002520
2521<p>The '<tt>blockaddress</tt>' constant computes the address of the specified
Chris Lattner2dfdf2a2009-10-27 21:49:40 +00002522 basic block in the specified function, and always has an i8* type. Taking
Chris Lattnercdfc9402009-11-01 01:27:45 +00002523 the address of the entry block is illegal.</p>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00002524
Chris Lattnerc6f44362009-10-27 21:01:34 +00002525<p>This value only has defined behavior when used as an operand to the
Bill Wendling1b383ba2010-10-27 01:07:41 +00002526 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>' instruction, or for
2527 comparisons against null. Pointer equality tests between labels addresses
2528 results in undefined behavior &mdash; though, again, comparison against null
2529 is ok, and no label is equal to the null pointer. This may be passed around
2530 as an opaque pointer sized value as long as the bits are not inspected. This
2531 allows <tt>ptrtoint</tt> and arithmetic to be performed on these values so
2532 long as the original value is reconstituted before the <tt>indirectbr</tt>
2533 instruction.</p>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00002534
Bill Wendling1b383ba2010-10-27 01:07:41 +00002535<p>Finally, some targets may provide defined semantics when using the value as
2536 the operand to an inline assembly, but that is target specific.</p>
Chris Lattnerc6f44362009-10-27 21:01:34 +00002537
2538</div>
2539
2540
2541<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002542<h3>
2543 <a name="constantexprs">Constant Expressions</a>
2544</h3>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002545
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002546<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002547
2548<p>Constant expressions are used to allow expressions involving other constants
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002549 to be used as constants. Constant expressions may be of
2550 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2551 operation that does not have side effects (e.g. load and call are not
Bill Wendling1b383ba2010-10-27 01:07:41 +00002552 supported). The following is the syntax for constant expressions:</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002553
2554<dl>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002555 <dt><b><tt>trunc (CST to TYPE)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002556 <dd>Truncate a constant to another type. The bit size of CST must be larger
2557 than the bit size of TYPE. Both types must be integers.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002558
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002559 <dt><b><tt>zext (CST to TYPE)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002560 <dd>Zero extend a constant to another type. The bit size of CST must be
Duncan Sands28afd432010-07-13 12:06:14 +00002561 smaller than the bit size of TYPE. Both types must be integers.</dd>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002562
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002563 <dt><b><tt>sext (CST to TYPE)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002564 <dd>Sign extend a constant to another type. The bit size of CST must be
Duncan Sands28afd432010-07-13 12:06:14 +00002565 smaller than the bit size of TYPE. Both types must be integers.</dd>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002566
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002567 <dt><b><tt>fptrunc (CST to TYPE)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002568 <dd>Truncate a floating point constant to another floating point type. The
2569 size of CST must be larger than the size of TYPE. Both types must be
2570 floating point.</dd>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002571
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002572 <dt><b><tt>fpext (CST to TYPE)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002573 <dd>Floating point extend a constant to another type. The size of CST must be
2574 smaller or equal to the size of TYPE. Both types must be floating
2575 point.</dd>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002576
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002577 <dt><b><tt>fptoui (CST to TYPE)</tt></b></dt>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002578 <dd>Convert a floating point constant to the corresponding unsigned integer
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002579 constant. TYPE must be a scalar or vector integer type. CST must be of
2580 scalar or vector floating point type. Both CST and TYPE must be scalars,
2581 or vectors of the same number of elements. If the value won't fit in the
2582 integer type, the results are undefined.</dd>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002583
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002584 <dt><b><tt>fptosi (CST to TYPE)</tt></b></dt>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002585 <dd>Convert a floating point constant to the corresponding signed integer
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002586 constant. TYPE must be a scalar or vector integer type. CST must be of
2587 scalar or vector floating point type. Both CST and TYPE must be scalars,
2588 or vectors of the same number of elements. If the value won't fit in the
2589 integer type, the results are undefined.</dd>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002590
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002591 <dt><b><tt>uitofp (CST to TYPE)</tt></b></dt>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002592 <dd>Convert an unsigned integer constant to the corresponding floating point
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002593 constant. TYPE must be a scalar or vector floating point type. CST must be
2594 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2595 vectors of the same number of elements. If the value won't fit in the
2596 floating point type, the results are undefined.</dd>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002597
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002598 <dt><b><tt>sitofp (CST to TYPE)</tt></b></dt>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002599 <dd>Convert a signed integer constant to the corresponding floating point
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002600 constant. TYPE must be a scalar or vector floating point type. CST must be
2601 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2602 vectors of the same number of elements. If the value won't fit in the
2603 floating point type, the results are undefined.</dd>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002604
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002605 <dt><b><tt>ptrtoint (CST to TYPE)</tt></b></dt>
Reid Spencer5c0ef472006-11-11 23:08:07 +00002606 <dd>Convert a pointer typed constant to the corresponding integer constant
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002607 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2608 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2609 make it fit in <tt>TYPE</tt>.</dd>
Reid Spencer5c0ef472006-11-11 23:08:07 +00002610
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002611 <dt><b><tt>inttoptr (CST to TYPE)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002612 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2613 type. CST must be of integer type. The CST value is zero extended,
2614 truncated, or unchanged to make it fit in a pointer size. This one is
2615 <i>really</i> dangerous!</dd>
Reid Spencer5c0ef472006-11-11 23:08:07 +00002616
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002617 <dt><b><tt>bitcast (CST to TYPE)</tt></b></dt>
Chris Lattner03bbad62009-02-28 18:27:03 +00002618 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2619 are the same as those for the <a href="#i_bitcast">bitcast
2620 instruction</a>.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002621
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002622 <dt><b><tt>getelementptr (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2623 <dt><b><tt>getelementptr inbounds (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002624 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002625 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2626 instruction, the index list may have zero or more indexes, which are
2627 required to make sense for the type of "CSTPTR".</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002628
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002629 <dt><b><tt>select (COND, VAL1, VAL2)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002630 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
Reid Spencer01c42592006-12-04 19:23:19 +00002631
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002632 <dt><b><tt>icmp COND (VAL1, VAL2)</tt></b></dt>
Reid Spencer01c42592006-12-04 19:23:19 +00002633 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2634
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002635 <dt><b><tt>fcmp COND (VAL1, VAL2)</tt></b></dt>
Reid Spencer01c42592006-12-04 19:23:19 +00002636 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
Robert Bocchino9fbe1452006-01-10 19:31:34 +00002637
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002638 <dt><b><tt>extractelement (VAL, IDX)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002639 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2640 constants.</dd>
Robert Bocchino9fbe1452006-01-10 19:31:34 +00002641
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002642 <dt><b><tt>insertelement (VAL, ELT, IDX)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002643 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2644 constants.</dd>
Chris Lattnerc1989542006-04-08 00:13:41 +00002645
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002646 <dt><b><tt>shufflevector (VEC1, VEC2, IDXMASK)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002647 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2648 constants.</dd>
Chris Lattnerc1989542006-04-08 00:13:41 +00002649
Nick Lewycky9e130ce2010-05-29 06:44:15 +00002650 <dt><b><tt>extractvalue (VAL, IDX0, IDX1, ...)</tt></b></dt>
2651 <dd>Perform the <a href="#i_extractvalue">extractvalue operation</a> on
2652 constants. The index list is interpreted in a similar manner as indices in
2653 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2654 index value must be specified.</dd>
2655
2656 <dt><b><tt>insertvalue (VAL, ELT, IDX0, IDX1, ...)</tt></b></dt>
2657 <dd>Perform the <a href="#i_insertvalue">insertvalue operation</a> on
2658 constants. The index list is interpreted in a similar manner as indices in
2659 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2660 index value must be specified.</dd>
2661
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002662 <dt><b><tt>OPCODE (LHS, RHS)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002663 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2664 be any of the <a href="#binaryops">binary</a>
2665 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2666 on operands are the same as those for the corresponding instruction
2667 (e.g. no bitwise operations on floating point values are allowed).</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002668</dl>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002669
Chris Lattnerc3f59762004-12-09 17:30:23 +00002670</div>
Chris Lattner9ee5d222004-03-08 16:49:10 +00002671
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002672</div>
2673
Chris Lattner00950542001-06-06 20:29:01 +00002674<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002675<h2><a name="othervalues">Other Values</a></h2>
Chris Lattnere87d6532006-01-25 23:47:57 +00002676<!-- *********************************************************************** -->
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002677<div>
Chris Lattnere87d6532006-01-25 23:47:57 +00002678<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002679<h3>
Chris Lattnere87d6532006-01-25 23:47:57 +00002680<a name="inlineasm">Inline Assembler Expressions</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002681</h3>
Chris Lattnere87d6532006-01-25 23:47:57 +00002682
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002683<div>
Chris Lattnere87d6532006-01-25 23:47:57 +00002684
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002685<p>LLVM supports inline assembler expressions (as opposed
2686 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2687 a special value. This value represents the inline assembler as a string
2688 (containing the instructions to emit), a list of operand constraints (stored
Dale Johannesen09fed252009-10-13 21:56:55 +00002689 as a string), a flag that indicates whether or not the inline asm
Dale Johannesen8ba2d5b2009-10-21 23:28:00 +00002690 expression has side effects, and a flag indicating whether the function
2691 containing the asm needs to align its stack conservatively. An example
2692 inline assembler expression is:</p>
Chris Lattnere87d6532006-01-25 23:47:57 +00002693
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002694<pre class="doc_code">
Bill Wendling2f7a8b02007-05-29 09:04:49 +00002695i32 (i32) asm "bswap $0", "=r,r"
Chris Lattnere87d6532006-01-25 23:47:57 +00002696</pre>
2697
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002698<p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2699 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2700 have:</p>
Chris Lattnere87d6532006-01-25 23:47:57 +00002701
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002702<pre class="doc_code">
Bill Wendling2f7a8b02007-05-29 09:04:49 +00002703%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
Chris Lattnere87d6532006-01-25 23:47:57 +00002704</pre>
2705
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002706<p>Inline asms with side effects not visible in the constraint list must be
2707 marked as having side effects. This is done through the use of the
2708 '<tt>sideeffect</tt>' keyword, like so:</p>
Chris Lattnere87d6532006-01-25 23:47:57 +00002709
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002710<pre class="doc_code">
Bill Wendling2f7a8b02007-05-29 09:04:49 +00002711call void asm sideeffect "eieio", ""()
Chris Lattnere87d6532006-01-25 23:47:57 +00002712</pre>
2713
Dale Johannesen8ba2d5b2009-10-21 23:28:00 +00002714<p>In some cases inline asms will contain code that will not work unless the
2715 stack is aligned in some way, such as calls or SSE instructions on x86,
2716 yet will not contain code that does that alignment within the asm.
2717 The compiler should make conservative assumptions about what the asm might
2718 contain and should generate its usual stack alignment code in the prologue
2719 if the '<tt>alignstack</tt>' keyword is present:</p>
Dale Johannesen09fed252009-10-13 21:56:55 +00002720
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002721<pre class="doc_code">
Dale Johannesen8ba2d5b2009-10-21 23:28:00 +00002722call void asm alignstack "eieio", ""()
Dale Johannesen09fed252009-10-13 21:56:55 +00002723</pre>
Dale Johannesen09fed252009-10-13 21:56:55 +00002724
2725<p>If both keywords appear the '<tt>sideeffect</tt>' keyword must come
2726 first.</p>
2727
Chris Lattnere87d6532006-01-25 23:47:57 +00002728<p>TODO: The format of the asm and constraints string still need to be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002729 documented here. Constraints on what can be done (e.g. duplication, moving,
2730 etc need to be documented). This is probably best done by reference to
2731 another document that covers inline asm from a holistic perspective.</p>
Chris Lattnercf9a4152010-04-07 05:38:05 +00002732
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002733<h4>
Chris Lattnercf9a4152010-04-07 05:38:05 +00002734<a name="inlineasm_md">Inline Asm Metadata</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002735</h4>
Chris Lattnercf9a4152010-04-07 05:38:05 +00002736
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002737<div>
Chris Lattnercf9a4152010-04-07 05:38:05 +00002738
2739<p>The call instructions that wrap inline asm nodes may have a "!srcloc" MDNode
Chris Lattnerce1b9ad2010-11-17 08:20:42 +00002740 attached to it that contains a list of constant integers. If present, the
2741 code generator will use the integer as the location cookie value when report
Chris Lattnercf9a4152010-04-07 05:38:05 +00002742 errors through the LLVMContext error reporting mechanisms. This allows a
Dan Gohman1c70c002010-04-28 00:36:01 +00002743 front-end to correlate backend errors that occur with inline asm back to the
Chris Lattnercf9a4152010-04-07 05:38:05 +00002744 source code that produced it. For example:</p>
2745
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002746<pre class="doc_code">
Chris Lattnercf9a4152010-04-07 05:38:05 +00002747call void asm sideeffect "something bad", ""()<b>, !srcloc !42</b>
2748...
2749!42 = !{ i32 1234567 }
2750</pre>
Chris Lattnercf9a4152010-04-07 05:38:05 +00002751
2752<p>It is up to the front-end to make sense of the magic numbers it places in the
Chris Lattnerce1b9ad2010-11-17 08:20:42 +00002753 IR. If the MDNode contains multiple constants, the code generator will use
2754 the one that corresponds to the line of the asm that the error occurs on.</p>
Chris Lattnere87d6532006-01-25 23:47:57 +00002755
2756</div>
2757
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002758</div>
2759
Chris Lattnere6a5ddd2010-01-15 21:50:19 +00002760<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002761<h3>
2762 <a name="metadata">Metadata Nodes and Metadata Strings</a>
2763</h3>
Chris Lattnere6a5ddd2010-01-15 21:50:19 +00002764
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002765<div>
Chris Lattnere6a5ddd2010-01-15 21:50:19 +00002766
2767<p>LLVM IR allows metadata to be attached to instructions in the program that
2768 can convey extra information about the code to the optimizers and code
2769 generator. One example application of metadata is source-level debug
2770 information. There are two metadata primitives: strings and nodes. All
2771 metadata has the <tt>metadata</tt> type and is identified in syntax by a
2772 preceding exclamation point ('<tt>!</tt>').</p>
2773
2774<p>A metadata string is a string surrounded by double quotes. It can contain
2775 any character by escaping non-printable characters with "\xx" where "xx" is
2776 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
2777
2778<p>Metadata nodes are represented with notation similar to structure constants
2779 (a comma separated list of elements, surrounded by braces and preceded by an
2780 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2781 10}</tt>". Metadata nodes can have any values as their operand.</p>
2782
2783<p>A <a href="#namedmetadatastructure">named metadata</a> is a collection of
2784 metadata nodes, which can be looked up in the module symbol table. For
2785 example: "<tt>!foo = metadata !{!4, !3}</tt>".
2786
Devang Patele1d50cd2010-03-04 23:44:48 +00002787<p>Metadata can be used as function arguments. Here <tt>llvm.dbg.value</tt>
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002788 function is using two metadata arguments.</p>
Devang Patele1d50cd2010-03-04 23:44:48 +00002789
Bill Wendling9ff5de92011-03-02 02:17:11 +00002790<div class="doc_code">
2791<pre>
2792call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
2793</pre>
2794</div>
Devang Patele1d50cd2010-03-04 23:44:48 +00002795
2796<p>Metadata can be attached with an instruction. Here metadata <tt>!21</tt> is
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002797 attached with <tt>add</tt> instruction using <tt>!dbg</tt> identifier.</p>
Devang Patele1d50cd2010-03-04 23:44:48 +00002798
Bill Wendling9ff5de92011-03-02 02:17:11 +00002799<div class="doc_code">
2800<pre>
2801%indvar.next = add i64 %indvar, 1, !dbg !21
2802</pre>
2803</div>
2804
Chris Lattnere6a5ddd2010-01-15 21:50:19 +00002805</div>
2806
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002807</div>
Chris Lattner857755c2009-07-20 05:55:19 +00002808
2809<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002810<h2>
Chris Lattner857755c2009-07-20 05:55:19 +00002811 <a name="intrinsic_globals">Intrinsic Global Variables</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002812</h2>
Chris Lattner857755c2009-07-20 05:55:19 +00002813<!-- *********************************************************************** -->
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002814<div>
Chris Lattner857755c2009-07-20 05:55:19 +00002815<p>LLVM has a number of "magic" global variables that contain data that affect
2816code generation or other IR semantics. These are documented here. All globals
Chris Lattner401e10c2009-07-20 06:14:25 +00002817of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2818section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2819by LLVM.</p>
Chris Lattner857755c2009-07-20 05:55:19 +00002820
2821<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002822<h3>
Chris Lattner857755c2009-07-20 05:55:19 +00002823<a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002824</h3>
Chris Lattner857755c2009-07-20 05:55:19 +00002825
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002826<div>
Chris Lattner857755c2009-07-20 05:55:19 +00002827
2828<p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2829href="#linkage_appending">appending linkage</a>. This array contains a list of
2830pointers to global variables and functions which may optionally have a pointer
2831cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2832
2833<pre>
2834 @X = global i8 4
2835 @Y = global i32 123
2836
2837 @llvm.used = appending global [2 x i8*] [
2838 i8* @X,
2839 i8* bitcast (i32* @Y to i8*)
2840 ], section "llvm.metadata"
2841</pre>
2842
2843<p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2844compiler, assembler, and linker are required to treat the symbol as if there is
2845a reference to the global that it cannot see. For example, if a variable has
2846internal linkage and no references other than that from the <tt>@llvm.used</tt>
2847list, it cannot be deleted. This is commonly used to represent references from
2848inline asms and other things the compiler cannot "see", and corresponds to
2849"attribute((used))" in GNU C.</p>
2850
2851<p>On some targets, the code generator must emit a directive to the assembler or
2852object file to prevent the assembler and linker from molesting the symbol.</p>
2853
2854</div>
2855
2856<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002857<h3>
2858 <a name="intg_compiler_used">
2859 The '<tt>llvm.compiler.used</tt>' Global Variable
2860 </a>
2861</h3>
Chris Lattner401e10c2009-07-20 06:14:25 +00002862
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002863<div>
Chris Lattner401e10c2009-07-20 06:14:25 +00002864
2865<p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2866<tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2867touching the symbol. On targets that support it, this allows an intelligent
2868linker to optimize references to the symbol without being impeded as it would be
2869by <tt>@llvm.used</tt>.</p>
2870
2871<p>This is a rare construct that should only be used in rare circumstances, and
2872should not be exposed to source languages.</p>
2873
2874</div>
2875
2876<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002877<h3>
Chris Lattner857755c2009-07-20 05:55:19 +00002878<a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002879</h3>
Chris Lattner857755c2009-07-20 05:55:19 +00002880
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002881<div>
David Chisnalle31e9962010-04-30 19:23:49 +00002882<pre>
2883%0 = type { i32, void ()* }
David Chisnall27195a52010-04-30 19:27:35 +00002884@llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor }]
David Chisnalle31e9962010-04-30 19:23:49 +00002885</pre>
2886<p>The <tt>@llvm.global_ctors</tt> array contains a list of constructor functions and associated priorities. The functions referenced by this array will be called in ascending order of priority (i.e. lowest first) when the module is loaded. The order of functions with the same priority is not defined.
2887</p>
Chris Lattner857755c2009-07-20 05:55:19 +00002888
2889</div>
2890
2891<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002892<h3>
Chris Lattner857755c2009-07-20 05:55:19 +00002893<a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002894</h3>
Chris Lattner857755c2009-07-20 05:55:19 +00002895
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002896<div>
David Chisnalle31e9962010-04-30 19:23:49 +00002897<pre>
2898%0 = type { i32, void ()* }
David Chisnall27195a52010-04-30 19:27:35 +00002899@llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor }]
David Chisnalle31e9962010-04-30 19:23:49 +00002900</pre>
Chris Lattner857755c2009-07-20 05:55:19 +00002901
David Chisnalle31e9962010-04-30 19:23:49 +00002902<p>The <tt>@llvm.global_dtors</tt> array contains a list of destructor functions and associated priorities. The functions referenced by this array will be called in descending order of priority (i.e. highest first) when the module is loaded. The order of functions with the same priority is not defined.
2903</p>
Chris Lattner857755c2009-07-20 05:55:19 +00002904
2905</div>
2906
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002907</div>
Chris Lattner857755c2009-07-20 05:55:19 +00002908
Chris Lattnere87d6532006-01-25 23:47:57 +00002909<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002910<h2><a name="instref">Instruction Reference</a></h2>
Chris Lattner261efe92003-11-25 01:02:51 +00002911<!-- *********************************************************************** -->
Chris Lattnerc3f59762004-12-09 17:30:23 +00002912
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002913<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002914
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002915<p>The LLVM instruction set consists of several different classifications of
2916 instructions: <a href="#terminators">terminator
2917 instructions</a>, <a href="#binaryops">binary instructions</a>,
2918 <a href="#bitwiseops">bitwise binary instructions</a>,
2919 <a href="#memoryops">memory instructions</a>, and
2920 <a href="#otherops">other instructions</a>.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002921
Chris Lattner00950542001-06-06 20:29:01 +00002922<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002923<h3>
2924 <a name="terminators">Terminator Instructions</a>
2925</h3>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002926
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002927<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002928
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002929<p>As mentioned <a href="#functionstructure">previously</a>, every basic block
2930 in a program ends with a "Terminator" instruction, which indicates which
2931 block should be executed after the current block is finished. These
2932 terminator instructions typically yield a '<tt>void</tt>' value: they produce
2933 control flow, not values (the one exception being the
2934 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2935
Duncan Sands83821c82010-04-15 20:35:54 +00002936<p>There are seven different terminator instructions: the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002937 '<a href="#i_ret"><tt>ret</tt></a>' instruction, the
2938 '<a href="#i_br"><tt>br</tt></a>' instruction, the
2939 '<a href="#i_switch"><tt>switch</tt></a>' instruction, the
Bill Wendling21c346e2009-11-02 00:25:26 +00002940 '<a href="#i_indirectbr">'<tt>indirectbr</tt></a>' Instruction, the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002941 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the
2942 '<a href="#i_unwind"><tt>unwind</tt></a>' instruction, and the
2943 '<a href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002944
Chris Lattner00950542001-06-06 20:29:01 +00002945<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002946<h4>
2947 <a name="i_ret">'<tt>ret</tt>' Instruction</a>
2948</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002949
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002950<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002951
Chris Lattner00950542001-06-06 20:29:01 +00002952<h5>Syntax:</h5>
Dan Gohmanb1e6b962008-10-04 19:00:07 +00002953<pre>
2954 ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
Chris Lattner7faa8832002-04-14 06:13:44 +00002955 ret void <i>; Return from void function</i>
Chris Lattner00950542001-06-06 20:29:01 +00002956</pre>
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00002957
Chris Lattner00950542001-06-06 20:29:01 +00002958<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002959<p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
2960 a value) from a function back to the caller.</p>
2961
2962<p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
2963 value and then causes control flow, and one that just causes control flow to
2964 occur.</p>
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00002965
Chris Lattner00950542001-06-06 20:29:01 +00002966<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002967<p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
2968 return value. The type of the return value must be a
2969 '<a href="#t_firstclass">first class</a>' type.</p>
Dan Gohmanb1e6b962008-10-04 19:00:07 +00002970
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002971<p>A function is not <a href="#wellformed">well formed</a> if it it has a
2972 non-void return type and contains a '<tt>ret</tt>' instruction with no return
2973 value or a return value with a type that does not match its type, or if it
2974 has a void return type and contains a '<tt>ret</tt>' instruction with a
2975 return value.</p>
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00002976
Chris Lattner00950542001-06-06 20:29:01 +00002977<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002978<p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
2979 the calling function's context. If the caller is a
2980 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
2981 instruction after the call. If the caller was an
2982 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
2983 the beginning of the "normal" destination block. If the instruction returns
2984 a value, that value shall set the call or invoke instruction's return
2985 value.</p>
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00002986
Chris Lattner00950542001-06-06 20:29:01 +00002987<h5>Example:</h5>
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00002988<pre>
2989 ret i32 5 <i>; Return an integer value of 5</i>
Chris Lattner7faa8832002-04-14 06:13:44 +00002990 ret void <i>; Return from a void function</i>
Bill Wendling0a4bbbf2009-02-28 22:12:54 +00002991 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
Chris Lattner00950542001-06-06 20:29:01 +00002992</pre>
Dan Gohmanf3e60bd2009-01-12 23:12:39 +00002993
Misha Brukman9d0919f2003-11-08 01:05:38 +00002994</div>
Chris Lattner00950542001-06-06 20:29:01 +00002995<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002996<h4>
2997 <a name="i_br">'<tt>br</tt>' Instruction</a>
2998</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002999
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003000<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003001
Chris Lattner00950542001-06-06 20:29:01 +00003002<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003003<pre>
3004 br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
Chris Lattner00950542001-06-06 20:29:01 +00003005</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003006
Chris Lattner00950542001-06-06 20:29:01 +00003007<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003008<p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
3009 different basic block in the current function. There are two forms of this
3010 instruction, corresponding to a conditional branch and an unconditional
3011 branch.</p>
3012
Chris Lattner00950542001-06-06 20:29:01 +00003013<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003014<p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
3015 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
3016 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
3017 target.</p>
3018
Chris Lattner00950542001-06-06 20:29:01 +00003019<h5>Semantics:</h5>
Reid Spencerc78f3372007-01-12 03:35:51 +00003020<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003021 argument is evaluated. If the value is <tt>true</tt>, control flows to the
3022 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
3023 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
3024
Chris Lattner00950542001-06-06 20:29:01 +00003025<h5>Example:</h5>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00003026<pre>
3027Test:
3028 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
3029 br i1 %cond, label %IfEqual, label %IfUnequal
3030IfEqual:
3031 <a href="#i_ret">ret</a> i32 1
3032IfUnequal:
3033 <a href="#i_ret">ret</a> i32 0
3034</pre>
3035
Misha Brukman9d0919f2003-11-08 01:05:38 +00003036</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003037
Chris Lattner00950542001-06-06 20:29:01 +00003038<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003039<h4>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003040 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003041</h4>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003042
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003043<div>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003044
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003045<h5>Syntax:</h5>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003046<pre>
3047 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
3048</pre>
3049
Chris Lattner00950542001-06-06 20:29:01 +00003050<h5>Overview:</h5>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003051<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003052 several different places. It is a generalization of the '<tt>br</tt>'
3053 instruction, allowing a branch to occur to one of many possible
3054 destinations.</p>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003055
Chris Lattner00950542001-06-06 20:29:01 +00003056<h5>Arguments:</h5>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003057<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003058 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
3059 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
3060 The table is not allowed to contain duplicate constant entries.</p>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003061
Chris Lattner00950542001-06-06 20:29:01 +00003062<h5>Semantics:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00003063<p>The <tt>switch</tt> instruction specifies a table of values and
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003064 destinations. When the '<tt>switch</tt>' instruction is executed, this table
3065 is searched for the given value. If the value is found, control flow is
Benjamin Kramer8040cd32009-10-12 14:46:08 +00003066 transferred to the corresponding destination; otherwise, control flow is
3067 transferred to the default destination.</p>
Chris Lattner00950542001-06-06 20:29:01 +00003068
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003069<h5>Implementation:</h5>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003070<p>Depending on properties of the target machine and the particular
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003071 <tt>switch</tt> instruction, this instruction may be code generated in
3072 different ways. For example, it could be generated as a series of chained
3073 conditional branches or with a lookup table.</p>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003074
3075<h5>Example:</h5>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003076<pre>
3077 <i>; Emulate a conditional br instruction</i>
Reid Spencerc78f3372007-01-12 03:35:51 +00003078 %Val = <a href="#i_zext">zext</a> i1 %value to i32
Dan Gohman2a08c532009-01-04 23:44:43 +00003079 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003080
3081 <i>; Emulate an unconditional br instruction</i>
Reid Spencerca86e162006-12-31 07:07:53 +00003082 switch i32 0, label %dest [ ]
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003083
3084 <i>; Implement a jump table:</i>
Dan Gohman2a08c532009-01-04 23:44:43 +00003085 switch i32 %val, label %otherwise [ i32 0, label %onzero
3086 i32 1, label %onone
3087 i32 2, label %ontwo ]
Chris Lattner00950542001-06-06 20:29:01 +00003088</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003089
Misha Brukman9d0919f2003-11-08 01:05:38 +00003090</div>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003091
Chris Lattnerf9be95f2009-10-27 19:13:16 +00003092
3093<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003094<h4>
Chris Lattnerab21db72009-10-28 00:19:10 +00003095 <a name="i_indirectbr">'<tt>indirectbr</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003096</h4>
Chris Lattnerf9be95f2009-10-27 19:13:16 +00003097
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003098<div>
Chris Lattnerf9be95f2009-10-27 19:13:16 +00003099
3100<h5>Syntax:</h5>
3101<pre>
Chris Lattnerab21db72009-10-28 00:19:10 +00003102 indirectbr &lt;somety&gt;* &lt;address&gt;, [ label &lt;dest1&gt;, label &lt;dest2&gt;, ... ]
Chris Lattnerf9be95f2009-10-27 19:13:16 +00003103</pre>
3104
3105<h5>Overview:</h5>
3106
Chris Lattnerab21db72009-10-28 00:19:10 +00003107<p>The '<tt>indirectbr</tt>' instruction implements an indirect branch to a label
Chris Lattnerf9be95f2009-10-27 19:13:16 +00003108 within the current function, whose address is specified by
Chris Lattnerc6f44362009-10-27 21:01:34 +00003109 "<tt>address</tt>". Address must be derived from a <a
3110 href="#blockaddress">blockaddress</a> constant.</p>
Chris Lattnerf9be95f2009-10-27 19:13:16 +00003111
3112<h5>Arguments:</h5>
3113
3114<p>The '<tt>address</tt>' argument is the address of the label to jump to. The
3115 rest of the arguments indicate the full set of possible destinations that the
3116 address may point to. Blocks are allowed to occur multiple times in the
3117 destination list, though this isn't particularly useful.</p>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003118
Chris Lattnerf9be95f2009-10-27 19:13:16 +00003119<p>This destination list is required so that dataflow analysis has an accurate
3120 understanding of the CFG.</p>
3121
3122<h5>Semantics:</h5>
3123
3124<p>Control transfers to the block specified in the address argument. All
3125 possible destination blocks must be listed in the label list, otherwise this
3126 instruction has undefined behavior. This implies that jumps to labels
3127 defined in other functions have undefined behavior as well.</p>
3128
3129<h5>Implementation:</h5>
3130
3131<p>This is typically implemented with a jump through a register.</p>
3132
3133<h5>Example:</h5>
3134<pre>
Chris Lattnerab21db72009-10-28 00:19:10 +00003135 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
Chris Lattnerf9be95f2009-10-27 19:13:16 +00003136</pre>
3137
3138</div>
3139
3140
Chris Lattner00950542001-06-06 20:29:01 +00003141<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003142<h4>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003143 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003144</h4>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003145
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003146<div>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003147
Chris Lattner00950542001-06-06 20:29:01 +00003148<h5>Syntax:</h5>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003149<pre>
Devang Patel307e8ab2008-10-07 17:48:33 +00003150 &lt;result&gt; = invoke [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>] &lt;ptr to function ty&gt; &lt;function ptr val&gt;(&lt;function args&gt;) [<a href="#fnattrs">fn attrs</a>]
Chris Lattner76b8a332006-05-14 18:23:06 +00003151 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003152</pre>
3153
Chris Lattner6536cfe2002-05-06 22:08:29 +00003154<h5>Overview:</h5>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003155<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003156 function, with the possibility of control flow transfer to either the
3157 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
3158 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
3159 control flow will return to the "normal" label. If the callee (or any
3160 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
3161 instruction, control is interrupted and continued at the dynamically nearest
3162 "exception" label.</p>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003163
Chris Lattner00950542001-06-06 20:29:01 +00003164<h5>Arguments:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00003165<p>This instruction requires several arguments:</p>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003166
Chris Lattner00950542001-06-06 20:29:01 +00003167<ol>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003168 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
3169 convention</a> the call should use. If none is specified, the call
3170 defaults to using C calling conventions.</li>
Devang Patelf642f472008-10-06 18:50:38 +00003171
3172 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003173 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
3174 '<tt>inreg</tt>' attributes are valid here.</li>
Devang Patelf642f472008-10-06 18:50:38 +00003175
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003176 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003177 function value being invoked. In most cases, this is a direct function
3178 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
3179 off an arbitrary pointer to function value.</li>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003180
3181 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003182 function to be invoked. </li>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003183
3184 <li>'<tt>function args</tt>': argument list whose types match the function
Chris Lattner0724fbd2010-03-02 06:36:51 +00003185 signature argument types and parameter attributes. All arguments must be
3186 of <a href="#t_firstclass">first class</a> type. If the function
3187 signature indicates the function accepts a variable number of arguments,
3188 the extra arguments can be specified.</li>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003189
3190 <li>'<tt>normal label</tt>': the label reached when the called function
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003191 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003192
3193 <li>'<tt>exception label</tt>': the label reached when a callee returns with
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003194 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003195
Devang Patel307e8ab2008-10-07 17:48:33 +00003196 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003197 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
3198 '<tt>readnone</tt>' attributes are valid here.</li>
Chris Lattner00950542001-06-06 20:29:01 +00003199</ol>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003200
Chris Lattner00950542001-06-06 20:29:01 +00003201<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003202<p>This instruction is designed to operate as a standard
3203 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
3204 primary difference is that it establishes an association with a label, which
3205 is used by the runtime library to unwind the stack.</p>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003206
3207<p>This instruction is used in languages with destructors to ensure that proper
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003208 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
3209 exception. Additionally, this is important for implementation of
3210 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003211
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003212<p>For the purposes of the SSA form, the definition of the value returned by the
3213 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
3214 block to the "normal" label. If the callee unwinds then no return value is
3215 available.</p>
Dan Gohmanf96a4992009-05-22 21:47:08 +00003216
Chris Lattnerdf7a6802010-01-15 18:08:37 +00003217<p>Note that the code generator does not yet completely support unwind, and
3218that the invoke/unwind semantics are likely to change in future versions.</p>
3219
Chris Lattner00950542001-06-06 20:29:01 +00003220<h5>Example:</h5>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003221<pre>
Nick Lewyckyd703f652008-03-16 07:18:12 +00003222 %retval = invoke i32 @Test(i32 15) to label %Continue
Jeff Cohen6f1cc772007-04-22 01:17:39 +00003223 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewyckyd703f652008-03-16 07:18:12 +00003224 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Jeff Cohen6f1cc772007-04-22 01:17:39 +00003225 unwind label %TestCleanup <i>; {i32}:retval set</i>
Chris Lattner00950542001-06-06 20:29:01 +00003226</pre>
Chris Lattner35eca582004-10-16 18:04:13 +00003227
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003228</div>
Chris Lattner35eca582004-10-16 18:04:13 +00003229
Chris Lattner27f71f22003-09-03 00:41:47 +00003230<!-- _______________________________________________________________________ -->
Chris Lattner35eca582004-10-16 18:04:13 +00003231
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003232<h4>
3233 <a name="i_unwind">'<tt>unwind</tt>' Instruction</a>
3234</h4>
Chris Lattner35eca582004-10-16 18:04:13 +00003235
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003236<div>
Chris Lattner35eca582004-10-16 18:04:13 +00003237
Chris Lattner27f71f22003-09-03 00:41:47 +00003238<h5>Syntax:</h5>
Chris Lattner35eca582004-10-16 18:04:13 +00003239<pre>
3240 unwind
3241</pre>
3242
Chris Lattner27f71f22003-09-03 00:41:47 +00003243<h5>Overview:</h5>
Chris Lattner35eca582004-10-16 18:04:13 +00003244<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003245 at the first callee in the dynamic call stack which used
3246 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
3247 This is primarily used to implement exception handling.</p>
Chris Lattner35eca582004-10-16 18:04:13 +00003248
Chris Lattner27f71f22003-09-03 00:41:47 +00003249<h5>Semantics:</h5>
Chris Lattner72ed2002008-04-19 21:01:16 +00003250<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003251 immediately halt. The dynamic call stack is then searched for the
3252 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
3253 Once found, execution continues at the "exceptional" destination block
3254 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
3255 instruction in the dynamic call chain, undefined behavior results.</p>
3256
Chris Lattnerdf7a6802010-01-15 18:08:37 +00003257<p>Note that the code generator does not yet completely support unwind, and
3258that the invoke/unwind semantics are likely to change in future versions.</p>
3259
Misha Brukman9d0919f2003-11-08 01:05:38 +00003260</div>
Chris Lattner35eca582004-10-16 18:04:13 +00003261
3262<!-- _______________________________________________________________________ -->
3263
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003264<h4>
3265 <a name="i_unreachable">'<tt>unreachable</tt>' Instruction</a>
3266</h4>
Chris Lattner35eca582004-10-16 18:04:13 +00003267
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003268<div>
Chris Lattner35eca582004-10-16 18:04:13 +00003269
3270<h5>Syntax:</h5>
3271<pre>
3272 unreachable
3273</pre>
3274
3275<h5>Overview:</h5>
Chris Lattner35eca582004-10-16 18:04:13 +00003276<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003277 instruction is used to inform the optimizer that a particular portion of the
3278 code is not reachable. This can be used to indicate that the code after a
3279 no-return function cannot be reached, and other facts.</p>
Chris Lattner35eca582004-10-16 18:04:13 +00003280
3281<h5>Semantics:</h5>
Chris Lattner35eca582004-10-16 18:04:13 +00003282<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003283
Chris Lattner35eca582004-10-16 18:04:13 +00003284</div>
3285
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003286</div>
3287
Chris Lattner00950542001-06-06 20:29:01 +00003288<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003289<h3>
3290 <a name="binaryops">Binary Operations</a>
3291</h3>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003292
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003293<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003294
3295<p>Binary operators are used to do most of the computation in a program. They
3296 require two operands of the same type, execute an operation on them, and
3297 produce a single value. The operands might represent multiple data, as is
3298 the case with the <a href="#t_vector">vector</a> data type. The result value
3299 has the same type as its operands.</p>
3300
Misha Brukman9d0919f2003-11-08 01:05:38 +00003301<p>There are several different binary operators:</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003302
Chris Lattner00950542001-06-06 20:29:01 +00003303<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003304<h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003305 <a name="i_add">'<tt>add</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003306</h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003307
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003308<div>
Chris Lattner5568e942008-05-20 20:48:21 +00003309
Chris Lattner00950542001-06-06 20:29:01 +00003310<h5>Syntax:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003311<pre>
Dan Gohmancbb38f22009-07-20 22:41:19 +00003312 &lt;result&gt; = add &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanfdfca792009-09-02 17:31:42 +00003313 &lt;result&gt; = add nuw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3314 &lt;result&gt; = add nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3315 &lt;result&gt; = add nuw nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00003316</pre>
Chris Lattner5568e942008-05-20 20:48:21 +00003317
Chris Lattner00950542001-06-06 20:29:01 +00003318<h5>Overview:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00003319<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003320
Chris Lattner00950542001-06-06 20:29:01 +00003321<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003322<p>The two arguments to the '<tt>add</tt>' instruction must
3323 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3324 integer values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003325
Chris Lattner00950542001-06-06 20:29:01 +00003326<h5>Semantics:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003327<p>The value produced is the integer sum of the two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003328
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003329<p>If the sum has unsigned overflow, the result returned is the mathematical
3330 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003331
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003332<p>Because LLVM integers use a two's complement representation, this instruction
3333 is appropriate for both signed and unsigned integers.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003334
Dan Gohman08d012e2009-07-22 22:44:56 +00003335<p><tt>nuw</tt> and <tt>nsw</tt> stand for &quot;No Unsigned Wrap&quot;
3336 and &quot;No Signed Wrap&quot;, respectively. If the <tt>nuw</tt> and/or
3337 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
Dan Gohmanfff6c532010-04-22 23:14:21 +00003338 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3339 respectively, occurs.</p>
Dan Gohmancbb38f22009-07-20 22:41:19 +00003340
Chris Lattner00950542001-06-06 20:29:01 +00003341<h5>Example:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003342<pre>
3343 &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Chris Lattner00950542001-06-06 20:29:01 +00003344</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003345
Misha Brukman9d0919f2003-11-08 01:05:38 +00003346</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003347
Chris Lattner00950542001-06-06 20:29:01 +00003348<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003349<h4>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003350 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003351</h4>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003352
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003353<div>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003354
3355<h5>Syntax:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003356<pre>
3357 &lt;result&gt; = fadd &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3358</pre>
3359
3360<h5>Overview:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003361<p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
3362
3363<h5>Arguments:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003364<p>The two arguments to the '<tt>fadd</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003365 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3366 floating point values. Both arguments must have identical types.</p>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003367
3368<h5>Semantics:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003369<p>The value produced is the floating point sum of the two operands.</p>
3370
3371<h5>Example:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003372<pre>
3373 &lt;result&gt; = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
3374</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003375
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003376</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003377
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003378<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003379<h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003380 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003381</h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003382
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003383<div>
Chris Lattner5568e942008-05-20 20:48:21 +00003384
Chris Lattner00950542001-06-06 20:29:01 +00003385<h5>Syntax:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003386<pre>
Dan Gohman08d012e2009-07-22 22:44:56 +00003387 &lt;result&gt; = sub &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanfdfca792009-09-02 17:31:42 +00003388 &lt;result&gt; = sub nuw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3389 &lt;result&gt; = sub nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3390 &lt;result&gt; = sub nuw nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00003391</pre>
Chris Lattner5568e942008-05-20 20:48:21 +00003392
Chris Lattner00950542001-06-06 20:29:01 +00003393<h5>Overview:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00003394<p>The '<tt>sub</tt>' instruction returns the difference of its two
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003395 operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003396
3397<p>Note that the '<tt>sub</tt>' instruction is used to represent the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003398 '<tt>neg</tt>' instruction present in most other intermediate
3399 representations.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003400
Chris Lattner00950542001-06-06 20:29:01 +00003401<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003402<p>The two arguments to the '<tt>sub</tt>' instruction must
3403 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3404 integer values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003405
Chris Lattner00950542001-06-06 20:29:01 +00003406<h5>Semantics:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003407<p>The value produced is the integer difference of the two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003408
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003409<p>If the difference has unsigned overflow, the result returned is the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003410 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
3411 result.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003412
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003413<p>Because LLVM integers use a two's complement representation, this instruction
3414 is appropriate for both signed and unsigned integers.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003415
Dan Gohman08d012e2009-07-22 22:44:56 +00003416<p><tt>nuw</tt> and <tt>nsw</tt> stand for &quot;No Unsigned Wrap&quot;
3417 and &quot;No Signed Wrap&quot;, respectively. If the <tt>nuw</tt> and/or
3418 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
Dan Gohmanfff6c532010-04-22 23:14:21 +00003419 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3420 respectively, occurs.</p>
Dan Gohmancbb38f22009-07-20 22:41:19 +00003421
Chris Lattner00950542001-06-06 20:29:01 +00003422<h5>Example:</h5>
Bill Wendlingaac388b2007-05-29 09:42:13 +00003423<pre>
3424 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
Reid Spencerca86e162006-12-31 07:07:53 +00003425 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
Chris Lattner00950542001-06-06 20:29:01 +00003426</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003427
Misha Brukman9d0919f2003-11-08 01:05:38 +00003428</div>
Chris Lattner5568e942008-05-20 20:48:21 +00003429
Chris Lattner00950542001-06-06 20:29:01 +00003430<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003431<h4>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003432 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003433</h4>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003434
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003435<div>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003436
3437<h5>Syntax:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003438<pre>
3439 &lt;result&gt; = fsub &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3440</pre>
3441
3442<h5>Overview:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003443<p>The '<tt>fsub</tt>' instruction returns the difference of its two
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003444 operands.</p>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003445
3446<p>Note that the '<tt>fsub</tt>' instruction is used to represent the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003447 '<tt>fneg</tt>' instruction present in most other intermediate
3448 representations.</p>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003449
3450<h5>Arguments:</h5>
Bill Wendlingd9fe2982009-07-20 02:32:41 +00003451<p>The two arguments to the '<tt>fsub</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003452 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3453 floating point values. Both arguments must have identical types.</p>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003454
3455<h5>Semantics:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003456<p>The value produced is the floating point difference of the two operands.</p>
3457
3458<h5>Example:</h5>
3459<pre>
3460 &lt;result&gt; = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
3461 &lt;result&gt; = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
3462</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003463
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003464</div>
3465
3466<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003467<h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003468 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003469</h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003470
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003471<div>
Chris Lattner5568e942008-05-20 20:48:21 +00003472
Chris Lattner00950542001-06-06 20:29:01 +00003473<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003474<pre>
Dan Gohman08d012e2009-07-22 22:44:56 +00003475 &lt;result&gt; = mul &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanfdfca792009-09-02 17:31:42 +00003476 &lt;result&gt; = mul nuw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3477 &lt;result&gt; = mul nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3478 &lt;result&gt; = mul nuw nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00003479</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003480
Chris Lattner00950542001-06-06 20:29:01 +00003481<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003482<p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003483
Chris Lattner00950542001-06-06 20:29:01 +00003484<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003485<p>The two arguments to the '<tt>mul</tt>' instruction must
3486 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3487 integer values. Both arguments must have identical types.</p>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003488
Chris Lattner00950542001-06-06 20:29:01 +00003489<h5>Semantics:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003490<p>The value produced is the integer product of the two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003491
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003492<p>If the result of the multiplication has unsigned overflow, the result
3493 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
3494 width of the result.</p>
3495
3496<p>Because LLVM integers use a two's complement representation, and the result
3497 is the same width as the operands, this instruction returns the correct
3498 result for both signed and unsigned integers. If a full product
3499 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
3500 be sign-extended or zero-extended as appropriate to the width of the full
3501 product.</p>
3502
Dan Gohman08d012e2009-07-22 22:44:56 +00003503<p><tt>nuw</tt> and <tt>nsw</tt> stand for &quot;No Unsigned Wrap&quot;
3504 and &quot;No Signed Wrap&quot;, respectively. If the <tt>nuw</tt> and/or
3505 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
Dan Gohmanfff6c532010-04-22 23:14:21 +00003506 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3507 respectively, occurs.</p>
Dan Gohmancbb38f22009-07-20 22:41:19 +00003508
Chris Lattner00950542001-06-06 20:29:01 +00003509<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003510<pre>
3511 &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
Chris Lattner00950542001-06-06 20:29:01 +00003512</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003513
Misha Brukman9d0919f2003-11-08 01:05:38 +00003514</div>
Chris Lattner5568e942008-05-20 20:48:21 +00003515
Chris Lattner00950542001-06-06 20:29:01 +00003516<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003517<h4>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003518 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003519</h4>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003520
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003521<div>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003522
3523<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003524<pre>
3525 &lt;result&gt; = fmul &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003526</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003527
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003528<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003529<p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003530
3531<h5>Arguments:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003532<p>The two arguments to the '<tt>fmul</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003533 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3534 floating point values. Both arguments must have identical types.</p>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003535
3536<h5>Semantics:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003537<p>The value produced is the floating point product of the two operands.</p>
3538
3539<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003540<pre>
3541 &lt;result&gt; = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003542</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003543
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003544</div>
3545
3546<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003547<h4>
3548 <a name="i_udiv">'<tt>udiv</tt>' Instruction</a>
3549</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003550
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003551<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003552
Reid Spencer1628cec2006-10-26 06:15:43 +00003553<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003554<pre>
Chris Lattner35bda892011-02-06 21:44:57 +00003555 &lt;result&gt; = udiv &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3556 &lt;result&gt; = udiv exact &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer1628cec2006-10-26 06:15:43 +00003557</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003558
Reid Spencer1628cec2006-10-26 06:15:43 +00003559<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003560<p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003561
Reid Spencer1628cec2006-10-26 06:15:43 +00003562<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003563<p>The two arguments to the '<tt>udiv</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003564 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3565 values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003566
Reid Spencer1628cec2006-10-26 06:15:43 +00003567<h5>Semantics:</h5>
Chris Lattner5ec89832008-01-28 00:36:27 +00003568<p>The value produced is the unsigned integer quotient of the two operands.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003569
Chris Lattner5ec89832008-01-28 00:36:27 +00003570<p>Note that unsigned integer division and signed integer division are distinct
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003571 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
3572
Chris Lattner5ec89832008-01-28 00:36:27 +00003573<p>Division by zero leads to undefined behavior.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003574
Chris Lattner35bda892011-02-06 21:44:57 +00003575<p>If the <tt>exact</tt> keyword is present, the result value of the
3576 <tt>udiv</tt> is a <a href="#trapvalues">trap value</a> if %op1 is not a
3577 multiple of %op2 (as such, "((a udiv exact b) mul b) == a").</p>
3578
3579
Reid Spencer1628cec2006-10-26 06:15:43 +00003580<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003581<pre>
3582 &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
Reid Spencer1628cec2006-10-26 06:15:43 +00003583</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003584
Reid Spencer1628cec2006-10-26 06:15:43 +00003585</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003586
Reid Spencer1628cec2006-10-26 06:15:43 +00003587<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003588<h4>
3589 <a name="i_sdiv">'<tt>sdiv</tt>' Instruction</a>
3590</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003591
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003592<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003593
Reid Spencer1628cec2006-10-26 06:15:43 +00003594<h5>Syntax:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003595<pre>
Dan Gohmancbb38f22009-07-20 22:41:19 +00003596 &lt;result&gt; = sdiv &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmanfdfca792009-09-02 17:31:42 +00003597 &lt;result&gt; = sdiv exact &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer1628cec2006-10-26 06:15:43 +00003598</pre>
Chris Lattner5568e942008-05-20 20:48:21 +00003599
Reid Spencer1628cec2006-10-26 06:15:43 +00003600<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003601<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003602
Reid Spencer1628cec2006-10-26 06:15:43 +00003603<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003604<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003605 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3606 values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003607
Reid Spencer1628cec2006-10-26 06:15:43 +00003608<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003609<p>The value produced is the signed integer quotient of the two operands rounded
3610 towards zero.</p>
3611
Chris Lattner5ec89832008-01-28 00:36:27 +00003612<p>Note that signed integer division and unsigned integer division are distinct
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003613 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
3614
Chris Lattner5ec89832008-01-28 00:36:27 +00003615<p>Division by zero leads to undefined behavior. Overflow also leads to
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003616 undefined behavior; this is a rare case, but can occur, for example, by doing
3617 a 32-bit division of -2147483648 by -1.</p>
3618
Dan Gohman9c5beed2009-07-22 00:04:19 +00003619<p>If the <tt>exact</tt> keyword is present, the result value of the
Dan Gohman27ef9972010-04-23 15:23:32 +00003620 <tt>sdiv</tt> is a <a href="#trapvalues">trap value</a> if the result would
Dan Gohman38da9272010-07-11 00:08:34 +00003621 be rounded.</p>
Dan Gohmancbb38f22009-07-20 22:41:19 +00003622
Reid Spencer1628cec2006-10-26 06:15:43 +00003623<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003624<pre>
3625 &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
Reid Spencer1628cec2006-10-26 06:15:43 +00003626</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003627
Reid Spencer1628cec2006-10-26 06:15:43 +00003628</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003629
Reid Spencer1628cec2006-10-26 06:15:43 +00003630<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003631<h4>
3632 <a name="i_fdiv">'<tt>fdiv</tt>' Instruction</a>
3633</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003634
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003635<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003636
Chris Lattner00950542001-06-06 20:29:01 +00003637<h5>Syntax:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003638<pre>
Gabor Greiffb224a22008-08-07 21:46:00 +00003639 &lt;result&gt; = fdiv &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner261efe92003-11-25 01:02:51 +00003640</pre>
Chris Lattner5568e942008-05-20 20:48:21 +00003641
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003642<h5>Overview:</h5>
3643<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003644
Chris Lattner261efe92003-11-25 01:02:51 +00003645<h5>Arguments:</h5>
Jeff Cohen6f1cc772007-04-22 01:17:39 +00003646<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003647 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3648 floating point values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003649
Chris Lattner261efe92003-11-25 01:02:51 +00003650<h5>Semantics:</h5>
Reid Spencer1628cec2006-10-26 06:15:43 +00003651<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003652
Chris Lattner261efe92003-11-25 01:02:51 +00003653<h5>Example:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003654<pre>
3655 &lt;result&gt; = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
Chris Lattner261efe92003-11-25 01:02:51 +00003656</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003657
Chris Lattner261efe92003-11-25 01:02:51 +00003658</div>
Chris Lattner5568e942008-05-20 20:48:21 +00003659
Chris Lattner261efe92003-11-25 01:02:51 +00003660<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003661<h4>
3662 <a name="i_urem">'<tt>urem</tt>' Instruction</a>
3663</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003664
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003665<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003666
Reid Spencer0a783f72006-11-02 01:53:59 +00003667<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003668<pre>
3669 &lt;result&gt; = urem &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer0a783f72006-11-02 01:53:59 +00003670</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003671
Reid Spencer0a783f72006-11-02 01:53:59 +00003672<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003673<p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3674 division of its two arguments.</p>
3675
Reid Spencer0a783f72006-11-02 01:53:59 +00003676<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003677<p>The two arguments to the '<tt>urem</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003678 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3679 values. Both arguments must have identical types.</p>
3680
Reid Spencer0a783f72006-11-02 01:53:59 +00003681<h5>Semantics:</h5>
3682<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003683 This instruction always performs an unsigned division to get the
3684 remainder.</p>
3685
Chris Lattner5ec89832008-01-28 00:36:27 +00003686<p>Note that unsigned integer remainder and signed integer remainder are
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003687 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3688
Chris Lattner5ec89832008-01-28 00:36:27 +00003689<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003690
Reid Spencer0a783f72006-11-02 01:53:59 +00003691<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003692<pre>
3693 &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
Reid Spencer0a783f72006-11-02 01:53:59 +00003694</pre>
3695
3696</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003697
Reid Spencer0a783f72006-11-02 01:53:59 +00003698<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003699<h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003700 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003701</h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003702
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003703<div>
Chris Lattner5568e942008-05-20 20:48:21 +00003704
Chris Lattner261efe92003-11-25 01:02:51 +00003705<h5>Syntax:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003706<pre>
Gabor Greiffb224a22008-08-07 21:46:00 +00003707 &lt;result&gt; = srem &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner261efe92003-11-25 01:02:51 +00003708</pre>
Chris Lattner5568e942008-05-20 20:48:21 +00003709
Chris Lattner261efe92003-11-25 01:02:51 +00003710<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003711<p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3712 division of its two operands. This instruction can also take
3713 <a href="#t_vector">vector</a> versions of the values in which case the
3714 elements must be integers.</p>
Chris Lattnerc7d3ab32008-01-04 04:33:49 +00003715
Chris Lattner261efe92003-11-25 01:02:51 +00003716<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003717<p>The two arguments to the '<tt>srem</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003718 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3719 values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003720
Chris Lattner261efe92003-11-25 01:02:51 +00003721<h5>Semantics:</h5>
Reid Spencer0a783f72006-11-02 01:53:59 +00003722<p>This instruction returns the <i>remainder</i> of a division (where the result
Duncan Sandsdea3a5e2011-03-07 09:12:24 +00003723 is either zero or has the same sign as the dividend, <tt>op1</tt>), not the
3724 <i>modulo</i> operator (where the result is either zero or has the same sign
3725 as the divisor, <tt>op2</tt>) of a value.
3726 For more information about the difference,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003727 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3728 Math Forum</a>. For a table of how this is implemented in various languages,
3729 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3730 Wikipedia: modulo operation</a>.</p>
3731
Chris Lattner5ec89832008-01-28 00:36:27 +00003732<p>Note that signed integer remainder and unsigned integer remainder are
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003733 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3734
Chris Lattner5ec89832008-01-28 00:36:27 +00003735<p>Taking the remainder of a division by zero leads to undefined behavior.
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003736 Overflow also leads to undefined behavior; this is a rare case, but can
3737 occur, for example, by taking the remainder of a 32-bit division of
3738 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3739 lets srem be implemented using instructions that return both the result of
3740 the division and the remainder.)</p>
3741
Chris Lattner261efe92003-11-25 01:02:51 +00003742<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003743<pre>
3744 &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
Reid Spencer0a783f72006-11-02 01:53:59 +00003745</pre>
3746
3747</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003748
Reid Spencer0a783f72006-11-02 01:53:59 +00003749<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003750<h4>
3751 <a name="i_frem">'<tt>frem</tt>' Instruction</a>
3752</h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003753
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003754<div>
Chris Lattner5568e942008-05-20 20:48:21 +00003755
Reid Spencer0a783f72006-11-02 01:53:59 +00003756<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003757<pre>
3758 &lt;result&gt; = frem &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer0a783f72006-11-02 01:53:59 +00003759</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003760
Reid Spencer0a783f72006-11-02 01:53:59 +00003761<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003762<p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3763 its two operands.</p>
3764
Reid Spencer0a783f72006-11-02 01:53:59 +00003765<h5>Arguments:</h5>
3766<p>The two arguments to the '<tt>frem</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003767 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3768 floating point values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003769
Reid Spencer0a783f72006-11-02 01:53:59 +00003770<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003771<p>This instruction returns the <i>remainder</i> of a division. The remainder
3772 has the same sign as the dividend.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003773
Reid Spencer0a783f72006-11-02 01:53:59 +00003774<h5>Example:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003775<pre>
3776 &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
Chris Lattner261efe92003-11-25 01:02:51 +00003777</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003778
Misha Brukman9d0919f2003-11-08 01:05:38 +00003779</div>
Robert Bocchino7b81c752006-02-17 21:18:08 +00003780
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003781</div>
3782
Reid Spencer8e11bf82007-02-02 13:57:07 +00003783<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003784<h3>
3785 <a name="bitwiseops">Bitwise Binary Operations</a>
3786</h3>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003787
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003788<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003789
3790<p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3791 program. They are generally very efficient instructions and can commonly be
3792 strength reduced from other instructions. They require two operands of the
3793 same type, execute an operation on them, and produce a single value. The
3794 resulting value is the same type as its operands.</p>
3795
Reid Spencer569f2fa2007-01-31 21:39:12 +00003796<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003797<h4>
3798 <a name="i_shl">'<tt>shl</tt>' Instruction</a>
3799</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003800
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003801<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003802
Reid Spencer569f2fa2007-01-31 21:39:12 +00003803<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003804<pre>
Chris Lattnerf067d582011-02-07 16:40:21 +00003805 &lt;result&gt; = shl &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3806 &lt;result&gt; = shl nuw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3807 &lt;result&gt; = shl nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3808 &lt;result&gt; = shl nuw nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003809</pre>
Chris Lattner6ccc2d52007-10-03 21:01:14 +00003810
Reid Spencer569f2fa2007-01-31 21:39:12 +00003811<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003812<p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3813 a specified number of bits.</p>
Chris Lattner6ccc2d52007-10-03 21:01:14 +00003814
Reid Spencer569f2fa2007-01-31 21:39:12 +00003815<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003816<p>Both arguments to the '<tt>shl</tt>' instruction must be the
3817 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3818 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003819
Reid Spencer569f2fa2007-01-31 21:39:12 +00003820<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003821<p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3822 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3823 is (statically or dynamically) negative or equal to or larger than the number
3824 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3825 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3826 shift amount in <tt>op2</tt>.</p>
Chris Lattner6ccc2d52007-10-03 21:01:14 +00003827
Chris Lattnerf067d582011-02-07 16:40:21 +00003828<p>If the <tt>nuw</tt> keyword is present, then the shift produces a
3829 <a href="#trapvalues">trap value</a> if it shifts out any non-zero bits. If
Chris Lattner66298c12011-02-09 16:44:44 +00003830 the <tt>nsw</tt> keyword is present, then the shift produces a
Chris Lattnerf067d582011-02-07 16:40:21 +00003831 <a href="#trapvalues">trap value</a> if it shifts out any bits that disagree
3832 with the resultant sign bit. As such, NUW/NSW have the same semantics as
3833 they would if the shift were expressed as a mul instruction with the same
3834 nsw/nuw bits in (mul %op1, (shl 1, %op2)).</p>
3835
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003836<h5>Example:</h5>
3837<pre>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003838 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
3839 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
3840 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattner6ccc2d52007-10-03 21:01:14 +00003841 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Mon P Wange9f10152008-12-09 05:46:39 +00003842 &lt;result&gt; = shl &lt;2 x i32&gt; &lt; i32 1, i32 1&gt;, &lt; i32 1, i32 2&gt; <i>; yields: result=&lt;2 x i32&gt; &lt; i32 2, i32 4&gt;</i>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003843</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003844
Reid Spencer569f2fa2007-01-31 21:39:12 +00003845</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003846
Reid Spencer569f2fa2007-01-31 21:39:12 +00003847<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003848<h4>
3849 <a name="i_lshr">'<tt>lshr</tt>' Instruction</a>
3850</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003851
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003852<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003853
Reid Spencer569f2fa2007-01-31 21:39:12 +00003854<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003855<pre>
Chris Lattnerf067d582011-02-07 16:40:21 +00003856 &lt;result&gt; = lshr &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3857 &lt;result&gt; = lshr exact &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003858</pre>
3859
3860<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003861<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
3862 operand shifted to the right a specified number of bits with zero fill.</p>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003863
3864<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003865<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003866 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3867 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003868
3869<h5>Semantics:</h5>
3870<p>This instruction always performs a logical shift right operation. The most
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003871 significant bits of the result will be filled with zero bits after the shift.
3872 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
3873 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3874 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3875 shift amount in <tt>op2</tt>.</p>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003876
Chris Lattnerf067d582011-02-07 16:40:21 +00003877<p>If the <tt>exact</tt> keyword is present, the result value of the
3878 <tt>lshr</tt> is a <a href="#trapvalues">trap value</a> if any of the bits
3879 shifted out are non-zero.</p>
3880
3881
Reid Spencer569f2fa2007-01-31 21:39:12 +00003882<h5>Example:</h5>
3883<pre>
3884 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
3885 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
3886 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
3887 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattner6ccc2d52007-10-03 21:01:14 +00003888 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Mon P Wange9f10152008-12-09 05:46:39 +00003889 &lt;result&gt; = lshr &lt;2 x i32&gt; &lt; i32 -2, i32 4&gt;, &lt; i32 1, i32 2&gt; <i>; yields: result=&lt;2 x i32&gt; &lt; i32 0x7FFFFFFF, i32 1&gt;</i>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003890</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003891
Reid Spencer569f2fa2007-01-31 21:39:12 +00003892</div>
3893
Reid Spencer8e11bf82007-02-02 13:57:07 +00003894<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003895<h4>
3896 <a name="i_ashr">'<tt>ashr</tt>' Instruction</a>
3897</h4>
3898
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003899<div>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003900
3901<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003902<pre>
Chris Lattnerf067d582011-02-07 16:40:21 +00003903 &lt;result&gt; = ashr &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3904 &lt;result&gt; = ashr exact &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003905</pre>
3906
3907<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003908<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3909 operand shifted to the right a specified number of bits with sign
3910 extension.</p>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003911
3912<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003913<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003914 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3915 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003916
3917<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003918<p>This instruction always performs an arithmetic shift right operation, The
3919 most significant bits of the result will be filled with the sign bit
3920 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3921 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
3922 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
3923 the corresponding shift amount in <tt>op2</tt>.</p>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003924
Chris Lattnerf067d582011-02-07 16:40:21 +00003925<p>If the <tt>exact</tt> keyword is present, the result value of the
3926 <tt>ashr</tt> is a <a href="#trapvalues">trap value</a> if any of the bits
3927 shifted out are non-zero.</p>
3928
Reid Spencer569f2fa2007-01-31 21:39:12 +00003929<h5>Example:</h5>
3930<pre>
3931 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3932 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3933 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3934 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattner6ccc2d52007-10-03 21:01:14 +00003935 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Mon P Wange9f10152008-12-09 05:46:39 +00003936 &lt;result&gt; = ashr &lt;2 x i32&gt; &lt; i32 -2, i32 4&gt;, &lt; i32 1, i32 3&gt; <i>; yields: result=&lt;2 x i32&gt; &lt; i32 -1, i32 0&gt;</i>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003937</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003938
Reid Spencer569f2fa2007-01-31 21:39:12 +00003939</div>
3940
Chris Lattner00950542001-06-06 20:29:01 +00003941<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003942<h4>
3943 <a name="i_and">'<tt>and</tt>' Instruction</a>
3944</h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003945
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003946<div>
Chris Lattner5568e942008-05-20 20:48:21 +00003947
Chris Lattner00950542001-06-06 20:29:01 +00003948<h5>Syntax:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003949<pre>
Gabor Greiffb224a22008-08-07 21:46:00 +00003950 &lt;result&gt; = and &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00003951</pre>
Chris Lattner5568e942008-05-20 20:48:21 +00003952
Chris Lattner00950542001-06-06 20:29:01 +00003953<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003954<p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
3955 operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003956
Chris Lattner00950542001-06-06 20:29:01 +00003957<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003958<p>The two arguments to the '<tt>and</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003959 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3960 values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003961
Chris Lattner00950542001-06-06 20:29:01 +00003962<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00003963<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003964
Misha Brukman9d0919f2003-11-08 01:05:38 +00003965<table border="1" cellspacing="0" cellpadding="4">
Chris Lattner261efe92003-11-25 01:02:51 +00003966 <tbody>
3967 <tr>
3968 <td>In0</td>
3969 <td>In1</td>
3970 <td>Out</td>
3971 </tr>
3972 <tr>
3973 <td>0</td>
3974 <td>0</td>
3975 <td>0</td>
3976 </tr>
3977 <tr>
3978 <td>0</td>
3979 <td>1</td>
3980 <td>0</td>
3981 </tr>
3982 <tr>
3983 <td>1</td>
3984 <td>0</td>
3985 <td>0</td>
3986 </tr>
3987 <tr>
3988 <td>1</td>
3989 <td>1</td>
3990 <td>1</td>
3991 </tr>
3992 </tbody>
3993</table>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003994
Chris Lattner00950542001-06-06 20:29:01 +00003995<h5>Example:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003996<pre>
3997 &lt;result&gt; = and i32 4, %var <i>; yields {i32}:result = 4 &amp; %var</i>
Reid Spencerca86e162006-12-31 07:07:53 +00003998 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
3999 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
Chris Lattner00950542001-06-06 20:29:01 +00004000</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00004001</div>
Chris Lattner00950542001-06-06 20:29:01 +00004002<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004003<h4>
4004 <a name="i_or">'<tt>or</tt>' Instruction</a>
4005</h4>
Chris Lattner5568e942008-05-20 20:48:21 +00004006
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004007<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004008
4009<h5>Syntax:</h5>
4010<pre>
4011 &lt;result&gt; = or &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
4012</pre>
4013
4014<h5>Overview:</h5>
4015<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
4016 two operands.</p>
4017
4018<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004019<p>The two arguments to the '<tt>or</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004020 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4021 values. Both arguments must have identical types.</p>
4022
Chris Lattner00950542001-06-06 20:29:01 +00004023<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00004024<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004025
Chris Lattner261efe92003-11-25 01:02:51 +00004026<table border="1" cellspacing="0" cellpadding="4">
4027 <tbody>
4028 <tr>
4029 <td>In0</td>
4030 <td>In1</td>
4031 <td>Out</td>
4032 </tr>
4033 <tr>
4034 <td>0</td>
4035 <td>0</td>
4036 <td>0</td>
4037 </tr>
4038 <tr>
4039 <td>0</td>
4040 <td>1</td>
4041 <td>1</td>
4042 </tr>
4043 <tr>
4044 <td>1</td>
4045 <td>0</td>
4046 <td>1</td>
4047 </tr>
4048 <tr>
4049 <td>1</td>
4050 <td>1</td>
4051 <td>1</td>
4052 </tr>
4053 </tbody>
4054</table>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004055
Chris Lattner00950542001-06-06 20:29:01 +00004056<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004057<pre>
4058 &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
Reid Spencerca86e162006-12-31 07:07:53 +00004059 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
4060 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
Chris Lattner00950542001-06-06 20:29:01 +00004061</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004062
Misha Brukman9d0919f2003-11-08 01:05:38 +00004063</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004064
Chris Lattner00950542001-06-06 20:29:01 +00004065<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004066<h4>
4067 <a name="i_xor">'<tt>xor</tt>' Instruction</a>
4068</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004069
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004070<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004071
Chris Lattner00950542001-06-06 20:29:01 +00004072<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004073<pre>
4074 &lt;result&gt; = xor &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00004075</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004076
Chris Lattner00950542001-06-06 20:29:01 +00004077<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004078<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
4079 its two operands. The <tt>xor</tt> is used to implement the "one's
4080 complement" operation, which is the "~" operator in C.</p>
4081
Chris Lattner00950542001-06-06 20:29:01 +00004082<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004083<p>The two arguments to the '<tt>xor</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004084 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4085 values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00004086
Chris Lattner00950542001-06-06 20:29:01 +00004087<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00004088<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004089
Chris Lattner261efe92003-11-25 01:02:51 +00004090<table border="1" cellspacing="0" cellpadding="4">
4091 <tbody>
4092 <tr>
4093 <td>In0</td>
4094 <td>In1</td>
4095 <td>Out</td>
4096 </tr>
4097 <tr>
4098 <td>0</td>
4099 <td>0</td>
4100 <td>0</td>
4101 </tr>
4102 <tr>
4103 <td>0</td>
4104 <td>1</td>
4105 <td>1</td>
4106 </tr>
4107 <tr>
4108 <td>1</td>
4109 <td>0</td>
4110 <td>1</td>
4111 </tr>
4112 <tr>
4113 <td>1</td>
4114 <td>1</td>
4115 <td>0</td>
4116 </tr>
4117 </tbody>
4118</table>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004119
Chris Lattner00950542001-06-06 20:29:01 +00004120<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004121<pre>
4122 &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
Reid Spencerca86e162006-12-31 07:07:53 +00004123 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
4124 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
4125 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
Chris Lattner00950542001-06-06 20:29:01 +00004126</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004127
Misha Brukman9d0919f2003-11-08 01:05:38 +00004128</div>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004129
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004130</div>
4131
Chris Lattner00950542001-06-06 20:29:01 +00004132<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004133<h3>
Chris Lattner3df241e2006-04-08 23:07:04 +00004134 <a name="vectorops">Vector Operations</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004135</h3>
Chris Lattner3df241e2006-04-08 23:07:04 +00004136
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004137<div>
Chris Lattner3df241e2006-04-08 23:07:04 +00004138
4139<p>LLVM supports several instructions to represent vector operations in a
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004140 target-independent manner. These instructions cover the element-access and
4141 vector-specific operations needed to process vectors effectively. While LLVM
4142 does directly support these vector operations, many sophisticated algorithms
4143 will want to use target-specific intrinsics to take full advantage of a
4144 specific target.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004145
Chris Lattner3df241e2006-04-08 23:07:04 +00004146<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004147<h4>
Chris Lattner3df241e2006-04-08 23:07:04 +00004148 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004149</h4>
Chris Lattner3df241e2006-04-08 23:07:04 +00004150
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004151<div>
Chris Lattner3df241e2006-04-08 23:07:04 +00004152
4153<h5>Syntax:</h5>
Chris Lattner3df241e2006-04-08 23:07:04 +00004154<pre>
Reid Spencerca86e162006-12-31 07:07:53 +00004155 &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 Lattner3df241e2006-04-08 23:07:04 +00004156</pre>
4157
4158<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004159<p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
4160 from a vector at a specified index.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004161
4162
4163<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004164<p>The first operand of an '<tt>extractelement</tt>' instruction is a value
4165 of <a href="#t_vector">vector</a> type. The second operand is an index
4166 indicating the position from which to extract the element. The index may be
4167 a variable.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004168
4169<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004170<p>The result is a scalar of the same type as the element type of
4171 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
4172 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4173 results are undefined.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004174
4175<h5>Example:</h5>
Chris Lattner3df241e2006-04-08 23:07:04 +00004176<pre>
Gabor Greifa5b6f452009-10-28 13:14:50 +00004177 &lt;result&gt; = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
Chris Lattner3df241e2006-04-08 23:07:04 +00004178</pre>
Chris Lattner3df241e2006-04-08 23:07:04 +00004179
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004180</div>
Chris Lattner3df241e2006-04-08 23:07:04 +00004181
4182<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004183<h4>
Chris Lattner3df241e2006-04-08 23:07:04 +00004184 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004185</h4>
Chris Lattner3df241e2006-04-08 23:07:04 +00004186
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004187<div>
Chris Lattner3df241e2006-04-08 23:07:04 +00004188
4189<h5>Syntax:</h5>
Chris Lattner3df241e2006-04-08 23:07:04 +00004190<pre>
Dan Gohmanf3480b92008-05-12 23:38:42 +00004191 &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 Lattner3df241e2006-04-08 23:07:04 +00004192</pre>
4193
4194<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004195<p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
4196 vector at a specified index.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004197
4198<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004199<p>The first operand of an '<tt>insertelement</tt>' instruction is a value
4200 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
4201 whose type must equal the element type of the first operand. The third
4202 operand is an index indicating the position at which to insert the value.
4203 The index may be a variable.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004204
4205<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004206<p>The result is a vector of the same type as <tt>val</tt>. Its element values
4207 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
4208 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4209 results are undefined.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004210
4211<h5>Example:</h5>
Chris Lattner3df241e2006-04-08 23:07:04 +00004212<pre>
Gabor Greifa5b6f452009-10-28 13:14:50 +00004213 &lt;result&gt; = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
Chris Lattner3df241e2006-04-08 23:07:04 +00004214</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004215
Chris Lattner3df241e2006-04-08 23:07:04 +00004216</div>
4217
4218<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004219<h4>
Chris Lattner3df241e2006-04-08 23:07:04 +00004220 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004221</h4>
Chris Lattner3df241e2006-04-08 23:07:04 +00004222
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004223<div>
Chris Lattner3df241e2006-04-08 23:07:04 +00004224
4225<h5>Syntax:</h5>
Chris Lattner3df241e2006-04-08 23:07:04 +00004226<pre>
Mon P Wangaeb06d22008-11-10 04:46:22 +00004227 &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;m x i32&gt; &lt;mask&gt; <i>; yields &lt;m x &lt;ty&gt;&gt;</i>
Chris Lattner3df241e2006-04-08 23:07:04 +00004228</pre>
4229
4230<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004231<p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
4232 from two input vectors, returning a vector with the same element type as the
4233 input and length that is the same as the shuffle mask.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004234
4235<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004236<p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
4237 with types that match each other. The third argument is a shuffle mask whose
4238 element type is always 'i32'. The result of the instruction is a vector
4239 whose length is the same as the shuffle mask and whose element type is the
4240 same as the element type of the first two operands.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004241
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004242<p>The shuffle mask operand is required to be a constant vector with either
4243 constant integer or undef values.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004244
4245<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004246<p>The elements of the two input vectors are numbered from left to right across
4247 both of the vectors. The shuffle mask operand specifies, for each element of
4248 the result vector, which element of the two input vectors the result element
4249 gets. The element selector may be undef (meaning "don't care") and the
4250 second operand may be undef if performing a shuffle from only one vector.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004251
4252<h5>Example:</h5>
Chris Lattner3df241e2006-04-08 23:07:04 +00004253<pre>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004254 &lt;result&gt; = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
Jeff Cohen6f1cc772007-04-22 01:17:39 +00004255 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004256 &lt;result&gt; = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
Reid Spencerca86e162006-12-31 07:07:53 +00004257 &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.
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004258 &lt;result&gt; = shufflevector &lt;8 x i32&gt; %v1, &lt;8 x i32&gt; undef,
Mon P Wangaeb06d22008-11-10 04:46:22 +00004259 &lt;4 x i32&gt; &lt;i32 0, i32 1, i32 2, i32 3&gt; <i>; yields &lt;4 x i32&gt;</i>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004260 &lt;result&gt; = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
Mon P Wangaeb06d22008-11-10 04:46:22 +00004261 &lt;8 x i32&gt; &lt;i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7 &gt; <i>; yields &lt;8 x i32&gt;</i>
Chris Lattner3df241e2006-04-08 23:07:04 +00004262</pre>
Chris Lattner3df241e2006-04-08 23:07:04 +00004263
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004264</div>
Tanya Lattner09474292006-04-14 19:24:33 +00004265
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004266</div>
4267
Chris Lattner3df241e2006-04-08 23:07:04 +00004268<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004269<h3>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004270 <a name="aggregateops">Aggregate Operations</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004271</h3>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004272
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004273<div>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004274
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004275<p>LLVM supports several instructions for working with
4276 <a href="#t_aggregate">aggregate</a> values.</p>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004277
Dan Gohmana334d5f2008-05-12 23:51:09 +00004278<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004279<h4>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004280 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004281</h4>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004282
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004283<div>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004284
4285<h5>Syntax:</h5>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004286<pre>
4287 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
4288</pre>
4289
4290<h5>Overview:</h5>
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004291<p>The '<tt>extractvalue</tt>' instruction extracts the value of a member field
4292 from an <a href="#t_aggregate">aggregate</a> value.</p>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004293
4294<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004295<p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
Chris Lattner61c70e92010-08-28 04:09:24 +00004296 of <a href="#t_struct">struct</a> or
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004297 <a href="#t_array">array</a> type. The operands are constant indices to
4298 specify which value to extract in a similar manner as indices in a
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004299 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
Frits van Bommel13242892010-12-05 20:54:38 +00004300 <p>The major differences to <tt>getelementptr</tt> indexing are:</p>
4301 <ul>
4302 <li>Since the value being indexed is not a pointer, the first index is
4303 omitted and assumed to be zero.</li>
4304 <li>At least one index must be specified.</li>
4305 <li>Not only struct indices but also array indices must be in
4306 bounds.</li>
4307 </ul>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004308
4309<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004310<p>The result is the value at the position in the aggregate specified by the
4311 index operands.</p>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004312
4313<h5>Example:</h5>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004314<pre>
Gabor Greifa5b6f452009-10-28 13:14:50 +00004315 &lt;result&gt; = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004316</pre>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004317
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004318</div>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004319
4320<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004321<h4>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004322 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004323</h4>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004324
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004325<div>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004326
4327<h5>Syntax:</h5>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004328<pre>
Chris Lattner8645d1a2011-05-22 07:18:08 +00004329 &lt;result&gt; = insertvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;ty&gt; &lt;elt&gt;, &lt;idx&gt;{, <idx>}* <i>; yields &lt;aggregate type&gt;</i>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004330</pre>
4331
4332<h5>Overview:</h5>
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004333<p>The '<tt>insertvalue</tt>' instruction inserts a value into a member field
4334 in an <a href="#t_aggregate">aggregate</a> value.</p>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004335
4336<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004337<p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
Chris Lattner61c70e92010-08-28 04:09:24 +00004338 of <a href="#t_struct">struct</a> or
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004339 <a href="#t_array">array</a> type. The second operand is a first-class
4340 value to insert. The following operands are constant indices indicating
4341 the position at which to insert the value in a similar manner as indices in a
Frits van Bommel13242892010-12-05 20:54:38 +00004342 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' instruction. The
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004343 value to insert must have the same type as the value identified by the
4344 indices.</p>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004345
4346<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004347<p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
4348 that of <tt>val</tt> except that the value at the position specified by the
4349 indices is that of <tt>elt</tt>.</p>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004350
4351<h5>Example:</h5>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004352<pre>
Chris Lattner8645d1a2011-05-22 07:18:08 +00004353 %agg1 = insertvalue {i32, float} undef, i32 1, 0 <i>; yields {i32 1, float undef}</i>
4354 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 <i>; yields {i32 1, float %val}</i>
4355 %agg3 = insertvalue {i32, {float}} %agg1, float %val, 1, 0 <i>; yields {i32 1, float %val}</i>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004356</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004357
Dan Gohmana334d5f2008-05-12 23:51:09 +00004358</div>
4359
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004360</div>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004361
4362<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004363<h3>
Chris Lattner884a9702006-08-15 00:45:58 +00004364 <a name="memoryops">Memory Access and Addressing Operations</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004365</h3>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004366
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004367<div>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004368
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004369<p>A key design point of an SSA-based representation is how it represents
4370 memory. In LLVM, no memory locations are in SSA form, which makes things
Victor Hernandez2fee2942009-10-26 23:44:29 +00004371 very simple. This section describes how to read, write, and allocate
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004372 memory in LLVM.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004373
Chris Lattner00950542001-06-06 20:29:01 +00004374<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004375<h4>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004376 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004377</h4>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004378
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004379<div>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004380
Chris Lattner00950542001-06-06 20:29:01 +00004381<h5>Syntax:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004382<pre>
Dan Gohmanf75a7d32010-05-28 01:14:11 +00004383 &lt;result&gt; = alloca &lt;type&gt;[, &lt;ty&gt; &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
Chris Lattner00950542001-06-06 20:29:01 +00004384</pre>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004385
Chris Lattner00950542001-06-06 20:29:01 +00004386<h5>Overview:</h5>
Jeff Cohen6f1cc772007-04-22 01:17:39 +00004387<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004388 currently executing function, to be automatically released when this function
4389 returns to its caller. The object is always allocated in the generic address
4390 space (address space zero).</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004391
Chris Lattner00950542001-06-06 20:29:01 +00004392<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004393<p>The '<tt>alloca</tt>' instruction
4394 allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt> bytes of memory on the
4395 runtime stack, returning a pointer of the appropriate type to the program.
4396 If "NumElements" is specified, it is the number of elements allocated,
4397 otherwise "NumElements" is defaulted to be one. If a constant alignment is
4398 specified, the value result of the allocation is guaranteed to be aligned to
4399 at least that boundary. If not specified, or if zero, the target can choose
4400 to align the allocation on any convenient boundary compatible with the
4401 type.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004402
Misha Brukman9d0919f2003-11-08 01:05:38 +00004403<p>'<tt>type</tt>' may be any sized type.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004404
Chris Lattner00950542001-06-06 20:29:01 +00004405<h5>Semantics:</h5>
Bill Wendling871eb0a2009-05-08 20:49:29 +00004406<p>Memory is allocated; a pointer is returned. The operation is undefined if
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004407 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
4408 memory is automatically released when the function returns. The
4409 '<tt>alloca</tt>' instruction is commonly used to represent automatic
4410 variables that must have an address available. When the function returns
4411 (either with the <tt><a href="#i_ret">ret</a></tt>
4412 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
4413 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004414
Chris Lattner00950542001-06-06 20:29:01 +00004415<h5>Example:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004416<pre>
Dan Gohman81e21672009-01-04 23:49:44 +00004417 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
4418 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
4419 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
4420 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
Chris Lattner00950542001-06-06 20:29:01 +00004421</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004422
Misha Brukman9d0919f2003-11-08 01:05:38 +00004423</div>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004424
Chris Lattner00950542001-06-06 20:29:01 +00004425<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004426<h4>
4427 <a name="i_load">'<tt>load</tt>' Instruction</a>
4428</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004429
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004430<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004431
Chris Lattner2b7d3202002-05-06 03:03:22 +00004432<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004433<pre>
Bill Wendling7c78dbb2010-02-25 21:23:24 +00004434 &lt;result&gt; = load &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;][, !nontemporal !&lt;index&gt;]
4435 &lt;result&gt; = volatile load &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;][, !nontemporal !&lt;index&gt;]
4436 !&lt;index&gt; = !{ i32 1 }
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004437</pre>
4438
Chris Lattner2b7d3202002-05-06 03:03:22 +00004439<h5>Overview:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00004440<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004441
Chris Lattner2b7d3202002-05-06 03:03:22 +00004442<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004443<p>The argument to the '<tt>load</tt>' instruction specifies the memory address
4444 from which to load. The pointer must point to
4445 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
4446 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00004447 number or order of execution of this <tt>load</tt> with other <a
4448 href="#volatile">volatile operations</a>.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004449
Bill Wendling7c78dbb2010-02-25 21:23:24 +00004450<p>The optional constant <tt>align</tt> argument specifies the alignment of the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004451 operation (that is, the alignment of the memory address). A value of 0 or an
Bill Wendling7c78dbb2010-02-25 21:23:24 +00004452 omitted <tt>align</tt> argument means that the operation has the preferential
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004453 alignment for the target. It is the responsibility of the code emitter to
4454 ensure that the alignment information is correct. Overestimating the
Bill Wendling7c78dbb2010-02-25 21:23:24 +00004455 alignment results in undefined behavior. Underestimating the alignment may
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004456 produce less efficient code. An alignment of 1 is always safe.</p>
4457
Bill Wendling7c78dbb2010-02-25 21:23:24 +00004458<p>The optional <tt>!nontemporal</tt> metadata must reference a single
4459 metatadata name &lt;index&gt; corresponding to a metadata node with
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00004460 one <tt>i32</tt> entry of value 1. The existence of
Bill Wendling7c78dbb2010-02-25 21:23:24 +00004461 the <tt>!nontemporal</tt> metatadata on the instruction tells the optimizer
4462 and code generator that this load is not expected to be reused in the cache.
4463 The code generator may select special instructions to save cache bandwidth,
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00004464 such as the <tt>MOVNT</tt> instruction on x86.</p>
David Greene8939b0d2010-02-16 20:50:18 +00004465
Chris Lattner2b7d3202002-05-06 03:03:22 +00004466<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004467<p>The location of memory pointed to is loaded. If the value being loaded is of
4468 scalar type then the number of bytes read does not exceed the minimum number
4469 of bytes needed to hold all bits of the type. For example, loading an
4470 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
4471 <tt>i20</tt> with a size that is not an integral number of bytes, the result
4472 is undefined if the value was not originally written using a store of the
4473 same type.</p>
4474
Chris Lattner2b7d3202002-05-06 03:03:22 +00004475<h5>Examples:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004476<pre>
4477 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4478 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
Reid Spencerca86e162006-12-31 07:07:53 +00004479 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
Chris Lattner2b7d3202002-05-06 03:03:22 +00004480</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004481
Misha Brukman9d0919f2003-11-08 01:05:38 +00004482</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004483
Chris Lattner2b7d3202002-05-06 03:03:22 +00004484<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004485<h4>
4486 <a name="i_store">'<tt>store</tt>' Instruction</a>
4487</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004488
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004489<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004490
Chris Lattner2b7d3202002-05-06 03:03:22 +00004491<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004492<pre>
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00004493 store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;][, !nontemporal !&lt;index&gt;] <i>; yields {void}</i>
4494 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;][, !nontemporal !&lt;index&gt;] <i>; yields {void}</i>
Chris Lattner2b7d3202002-05-06 03:03:22 +00004495</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004496
Chris Lattner2b7d3202002-05-06 03:03:22 +00004497<h5>Overview:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00004498<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004499
Chris Lattner2b7d3202002-05-06 03:03:22 +00004500<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004501<p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
4502 and an address at which to store it. The type of the
4503 '<tt>&lt;pointer&gt;</tt>' operand must be a pointer to
4504 the <a href="#t_firstclass">first class</a> type of the
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00004505 '<tt>&lt;value&gt;</tt>' operand. If the <tt>store</tt> is marked as
4506 <tt>volatile</tt>, then the optimizer is not allowed to modify the number or
4507 order of execution of this <tt>store</tt> with other <a
4508 href="#volatile">volatile operations</a>.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004509
4510<p>The optional constant "align" argument specifies the alignment of the
4511 operation (that is, the alignment of the memory address). A value of 0 or an
4512 omitted "align" argument means that the operation has the preferential
4513 alignment for the target. It is the responsibility of the code emitter to
4514 ensure that the alignment information is correct. Overestimating the
4515 alignment results in an undefined behavior. Underestimating the alignment may
4516 produce less efficient code. An alignment of 1 is always safe.</p>
4517
David Greene8939b0d2010-02-16 20:50:18 +00004518<p>The optional !nontemporal metadata must reference a single metatadata
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00004519 name &lt;index&gt; corresponding to a metadata node with one i32 entry of
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00004520 value 1. The existence of the !nontemporal metatadata on the
David Greene8939b0d2010-02-16 20:50:18 +00004521 instruction tells the optimizer and code generator that this load is
4522 not expected to be reused in the cache. The code generator may
4523 select special instructions to save cache bandwidth, such as the
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00004524 MOVNT instruction on x86.</p>
David Greene8939b0d2010-02-16 20:50:18 +00004525
4526
Chris Lattner261efe92003-11-25 01:02:51 +00004527<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004528<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>' at the
4529 location specified by the '<tt>&lt;pointer&gt;</tt>' operand. If
4530 '<tt>&lt;value&gt;</tt>' is of scalar type then the number of bytes written
4531 does not exceed the minimum number of bytes needed to hold all bits of the
4532 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
4533 writing a value of a type like <tt>i20</tt> with a size that is not an
4534 integral number of bytes, it is unspecified what happens to the extra bits
4535 that do not belong to the type, but they will typically be overwritten.</p>
4536
Chris Lattner2b7d3202002-05-06 03:03:22 +00004537<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004538<pre>
4539 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling8c6c72d2007-10-22 05:10:05 +00004540 store i32 3, i32* %ptr <i>; yields {void}</i>
4541 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
Chris Lattner2b7d3202002-05-06 03:03:22 +00004542</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004543
Reid Spencer47ce1792006-11-09 21:15:49 +00004544</div>
4545
Chris Lattner2b7d3202002-05-06 03:03:22 +00004546<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004547<h4>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004548 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004549</h4>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004550
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004551<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004552
Chris Lattner7faa8832002-04-14 06:13:44 +00004553<h5>Syntax:</h5>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004554<pre>
Matthijs Kooijmane49d0bc2008-10-13 13:44:15 +00004555 &lt;result&gt; = getelementptr &lt;pty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
Dan Gohmandd8004d2009-07-27 21:53:46 +00004556 &lt;result&gt; = getelementptr inbounds &lt;pty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004557</pre>
4558
Chris Lattner7faa8832002-04-14 06:13:44 +00004559<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004560<p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004561 subelement of an <a href="#t_aggregate">aggregate</a> data structure.
4562 It performs address calculation only and does not access memory.</p>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004563
Chris Lattner7faa8832002-04-14 06:13:44 +00004564<h5>Arguments:</h5>
Matthijs Kooijmane49d0bc2008-10-13 13:44:15 +00004565<p>The first argument is always a pointer, and forms the basis of the
Chris Lattnerc8eef442009-07-29 06:44:13 +00004566 calculation. The remaining arguments are indices that indicate which of the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004567 elements of the aggregate object are indexed. The interpretation of each
4568 index is dependent on the type being indexed into. The first index always
4569 indexes the pointer value given as the first argument, the second index
4570 indexes a value of the type pointed to (not necessarily the value directly
4571 pointed to, since the first index can be non-zero), etc. The first type
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004572 indexed into must be a pointer value, subsequent types can be arrays,
Chris Lattner61c70e92010-08-28 04:09:24 +00004573 vectors, and structs. Note that subsequent types being indexed into
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004574 can never be pointers, since that would require loading the pointer before
4575 continuing calculation.</p>
Matthijs Kooijmane49d0bc2008-10-13 13:44:15 +00004576
4577<p>The type of each index argument depends on the type it is indexing into.
Chris Lattner61c70e92010-08-28 04:09:24 +00004578 When indexing into a (optionally packed) structure, only <tt>i32</tt>
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004579 integer <b>constants</b> are allowed. When indexing into an array, pointer
4580 or vector, integers of any width are allowed, and they are not required to be
Chris Lattnerc8eef442009-07-29 06:44:13 +00004581 constant.</p>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004582
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004583<p>For example, let's consider a C code fragment and how it gets compiled to
4584 LLVM:</p>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004585
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00004586<pre class="doc_code">
Bill Wendling2f7a8b02007-05-29 09:04:49 +00004587struct RT {
4588 char A;
Chris Lattnercabc8462007-05-29 15:43:56 +00004589 int B[10][20];
Bill Wendling2f7a8b02007-05-29 09:04:49 +00004590 char C;
4591};
4592struct ST {
Chris Lattnercabc8462007-05-29 15:43:56 +00004593 int X;
Bill Wendling2f7a8b02007-05-29 09:04:49 +00004594 double Y;
4595 struct RT Z;
4596};
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004597
Chris Lattnercabc8462007-05-29 15:43:56 +00004598int *foo(struct ST *s) {
Bill Wendling2f7a8b02007-05-29 09:04:49 +00004599 return &amp;s[1].Z.B[5][13];
4600}
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004601</pre>
4602
Misha Brukman9d0919f2003-11-08 01:05:38 +00004603<p>The LLVM code generated by the GCC frontend is:</p>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004604
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00004605<pre class="doc_code">
Chris Lattnere7886e42009-01-11 20:53:49 +00004606%RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
4607%ST = <a href="#namedtypes">type</a> { i32, double, %RT }
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004608
Dan Gohman4df605b2009-07-25 02:23:48 +00004609define i32* @foo(%ST* %s) {
Bill Wendling2f7a8b02007-05-29 09:04:49 +00004610entry:
4611 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
4612 ret i32* %reg
4613}
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004614</pre>
4615
Chris Lattner7faa8832002-04-14 06:13:44 +00004616<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00004617<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004618 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
4619 }</tt>' type, a structure. The second index indexes into the third element
4620 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
4621 i8 }</tt>' type, another structure. The third index indexes into the second
4622 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
4623 array. The two dimensions of the array are subscripted into, yielding an
4624 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
4625 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004626
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004627<p>Note that it is perfectly legal to index partially through a structure,
4628 returning a pointer to an inner element. Because of this, the LLVM code for
4629 the given testcase is equivalent to:</p>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004630
4631<pre>
Dan Gohman4df605b2009-07-25 02:23:48 +00004632 define i32* @foo(%ST* %s) {
Reid Spencerca86e162006-12-31 07:07:53 +00004633 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
Jeff Cohen6f1cc772007-04-22 01:17:39 +00004634 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
4635 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
Reid Spencerca86e162006-12-31 07:07:53 +00004636 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
4637 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
4638 ret i32* %t5
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004639 }
Chris Lattner6536cfe2002-05-06 22:08:29 +00004640</pre>
Chris Lattnere67a9512005-06-24 17:22:57 +00004641
Dan Gohmandd8004d2009-07-27 21:53:46 +00004642<p>If the <tt>inbounds</tt> keyword is present, the result value of the
Dan Gohman27ef9972010-04-23 15:23:32 +00004643 <tt>getelementptr</tt> is a <a href="#trapvalues">trap value</a> if the
4644 base pointer is not an <i>in bounds</i> address of an allocated object,
4645 or if any of the addresses that would be formed by successive addition of
4646 the offsets implied by the indices to the base address with infinitely
4647 precise arithmetic are not an <i>in bounds</i> address of that allocated
4648 object. The <i>in bounds</i> addresses for an allocated object are all
4649 the addresses that point into the object, plus the address one byte past
4650 the end.</p>
Dan Gohmandd8004d2009-07-27 21:53:46 +00004651
4652<p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
4653 the base address with silently-wrapping two's complement arithmetic, and
4654 the result value of the <tt>getelementptr</tt> may be outside the object
4655 pointed to by the base pointer. The result value may not necessarily be
4656 used to access memory though, even if it happens to point into allocated
4657 storage. See the <a href="#pointeraliasing">Pointer Aliasing Rules</a>
4658 section for more information.</p>
4659
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004660<p>The getelementptr instruction is often confusing. For some more insight into
4661 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
Chris Lattner884a9702006-08-15 00:45:58 +00004662
Chris Lattner7faa8832002-04-14 06:13:44 +00004663<h5>Example:</h5>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004664<pre>
Reid Spencerca86e162006-12-31 07:07:53 +00004665 <i>; yields [12 x i8]*:aptr</i>
Matthijs Kooijmane49d0bc2008-10-13 13:44:15 +00004666 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
4667 <i>; yields i8*:vptr</i>
Dan Gohman0e451ce2008-10-14 16:51:45 +00004668 %vptr = getelementptr {i32, &lt;2 x i8&gt;}* %svptr, i64 0, i32 1, i32 1
Matthijs Kooijmane49d0bc2008-10-13 13:44:15 +00004669 <i>; yields i8*:eptr</i>
4670 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
Sanjiv Gupta9f805c22009-04-25 07:27:44 +00004671 <i>; yields i32*:iptr</i>
Sanjiv Gupta16ffa802009-04-24 16:38:13 +00004672 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004673</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004674
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004675</div>
Reid Spencer47ce1792006-11-09 21:15:49 +00004676
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004677</div>
4678
Chris Lattner00950542001-06-06 20:29:01 +00004679<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004680<h3>
4681 <a name="convertops">Conversion Operations</a>
4682</h3>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004683
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004684<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004685
Reid Spencer2fd21e62006-11-08 01:18:52 +00004686<p>The instructions in this category are the conversion instructions (casting)
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004687 which all take a single operand and a type. They perform various bit
4688 conversions on the operand.</p>
4689
Chris Lattner6536cfe2002-05-06 22:08:29 +00004690<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004691<h4>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004692 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004693</h4>
4694
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004695<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004696
4697<h5>Syntax:</h5>
4698<pre>
4699 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4700</pre>
4701
4702<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004703<p>The '<tt>trunc</tt>' instruction truncates its operand to the
4704 type <tt>ty2</tt>.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004705
4706<h5>Arguments:</h5>
Nadav Rotem8c20ec52011-02-24 21:01:34 +00004707<p>The '<tt>trunc</tt>' instruction takes a value to trunc, and a type to trunc it to.
4708 Both types must be of <a href="#t_integer">integer</a> types, or vectors
4709 of the same number of integers.
4710 The bit size of the <tt>value</tt> must be larger than
4711 the bit size of the destination type, <tt>ty2</tt>.
4712 Equal sized types are not allowed.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004713
4714<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004715<p>The '<tt>trunc</tt>' instruction truncates the high order bits
4716 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
4717 source size must be larger than the destination size, <tt>trunc</tt> cannot
4718 be a <i>no-op cast</i>. It will always truncate bits.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004719
4720<h5>Example:</h5>
4721<pre>
Nadav Rotem8c20ec52011-02-24 21:01:34 +00004722 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
4723 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
4724 %Z = trunc i32 122 to i1 <i>; yields i1:false</i>
4725 %W = trunc &lt;2 x i16&gt; &lt;i16 8, i16 7&gt; to &lt;2 x i8&gt; <i>; yields &lt;i8 8, i8 7&gt;</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004726</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004727
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004728</div>
4729
4730<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004731<h4>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004732 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004733</h4>
4734
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004735<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004736
4737<h5>Syntax:</h5>
4738<pre>
4739 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4740</pre>
4741
4742<h5>Overview:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004743<p>The '<tt>zext</tt>' instruction zero extends its operand to type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004744 <tt>ty2</tt>.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004745
4746
4747<h5>Arguments:</h5>
Nadav Rotemed9b9342011-02-20 12:37:50 +00004748<p>The '<tt>zext</tt>' instruction takes a value to cast, and a type to cast it to.
4749 Both types must be of <a href="#t_integer">integer</a> types, or vectors
4750 of the same number of integers.
4751 The bit size of the <tt>value</tt> must be smaller than
4752 the bit size of the destination type,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004753 <tt>ty2</tt>.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004754
4755<h5>Semantics:</h5>
4756<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004757 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004758
Reid Spencerb5929522007-01-12 15:46:11 +00004759<p>When zero extending from i1, the result will always be either 0 or 1.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004760
4761<h5>Example:</h5>
4762<pre>
Reid Spencerca86e162006-12-31 07:07:53 +00004763 %X = zext i32 257 to i64 <i>; yields i64:257</i>
Reid Spencerc78f3372007-01-12 03:35:51 +00004764 %Y = zext i1 true to i32 <i>; yields i32:1</i>
Nadav Rotemed9b9342011-02-20 12:37:50 +00004765 %Z = zext &lt;2 x i16&gt; &lt;i16 8, i16 7&gt; to &lt;2 x i32&gt; <i>; yields &lt;i32 8, i32 7&gt;</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004766</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004767
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004768</div>
4769
4770<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004771<h4>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004772 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004773</h4>
4774
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004775<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004776
4777<h5>Syntax:</h5>
4778<pre>
4779 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4780</pre>
4781
4782<h5>Overview:</h5>
4783<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
4784
4785<h5>Arguments:</h5>
Nadav Rotem8c20ec52011-02-24 21:01:34 +00004786<p>The '<tt>sext</tt>' instruction takes a value to cast, and a type to cast it to.
4787 Both types must be of <a href="#t_integer">integer</a> types, or vectors
4788 of the same number of integers.
4789 The bit size of the <tt>value</tt> must be smaller than
4790 the bit size of the destination type,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004791 <tt>ty2</tt>.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004792
4793<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004794<p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
4795 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
4796 of the type <tt>ty2</tt>.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004797
Reid Spencerc78f3372007-01-12 03:35:51 +00004798<p>When sign extending from i1, the extension always results in -1 or 0.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004799
4800<h5>Example:</h5>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004801<pre>
Reid Spencerca86e162006-12-31 07:07:53 +00004802 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
Reid Spencerc78f3372007-01-12 03:35:51 +00004803 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
Nadav Rotem8c20ec52011-02-24 21:01:34 +00004804 %Z = sext &lt;2 x i16&gt; &lt;i16 8, i16 7&gt; to &lt;2 x i32&gt; <i>; yields &lt;i32 8, i32 7&gt;</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004805</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004806
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004807</div>
4808
4809<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004810<h4>
Reid Spencer3fa91b02006-11-09 21:48:10 +00004811 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004812</h4>
Reid Spencer3fa91b02006-11-09 21:48:10 +00004813
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004814<div>
Reid Spencer3fa91b02006-11-09 21:48:10 +00004815
4816<h5>Syntax:</h5>
Reid Spencer3fa91b02006-11-09 21:48:10 +00004817<pre>
4818 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4819</pre>
4820
4821<h5>Overview:</h5>
4822<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004823 <tt>ty2</tt>.</p>
Reid Spencer3fa91b02006-11-09 21:48:10 +00004824
4825<h5>Arguments:</h5>
4826<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004827 point</a> value to cast and a <a href="#t_floating">floating point</a> type
4828 to cast it to. The size of <tt>value</tt> must be larger than the size of
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004829 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004830 <i>no-op cast</i>.</p>
Reid Spencer3fa91b02006-11-09 21:48:10 +00004831
4832<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004833<p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004834 <a href="#t_floating">floating point</a> type to a smaller
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004835 <a href="#t_floating">floating point</a> type. If the value cannot fit
4836 within the destination type, <tt>ty2</tt>, then the results are
4837 undefined.</p>
Reid Spencer3fa91b02006-11-09 21:48:10 +00004838
4839<h5>Example:</h5>
4840<pre>
4841 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4842 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4843</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004844
Reid Spencer3fa91b02006-11-09 21:48:10 +00004845</div>
4846
4847<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004848<h4>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004849 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004850</h4>
4851
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004852<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004853
4854<h5>Syntax:</h5>
4855<pre>
4856 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4857</pre>
4858
4859<h5>Overview:</h5>
4860<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004861 floating point value.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004862
4863<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004864<p>The '<tt>fpext</tt>' instruction takes a
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004865 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
4866 a <a href="#t_floating">floating point</a> type to cast it to. The source
4867 type must be smaller than the destination type.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004868
4869<h5>Semantics:</h5>
Reid Spencerd4448792006-11-09 23:03:26 +00004870<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004871 <a href="#t_floating">floating point</a> type to a larger
4872 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4873 used to make a <i>no-op cast</i> because it always changes bits. Use
4874 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004875
4876<h5>Example:</h5>
4877<pre>
Nick Lewycky5bb3ece2011-03-31 18:20:19 +00004878 %X = fpext float 3.125 to double <i>; yields double:3.125000e+00</i>
4879 %Y = fpext double %X to fp128 <i>; yields fp128:0xL00000000000000004000900000000000</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004880</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004881
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004882</div>
4883
4884<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004885<h4>
Reid Spencer24d6da52007-01-21 00:29:26 +00004886 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004887</h4>
4888
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004889<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004890
4891<h5>Syntax:</h5>
4892<pre>
Reid Spencer1539a1c2007-07-31 14:40:14 +00004893 &lt;result&gt; = fptoui &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004894</pre>
4895
4896<h5>Overview:</h5>
Reid Spencer1539a1c2007-07-31 14:40:14 +00004897<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004898 unsigned integer equivalent of type <tt>ty2</tt>.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004899
4900<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004901<p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
4902 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4903 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4904 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4905 vector integer type with the same number of elements as <tt>ty</tt></p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004906
4907<h5>Semantics:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004908<p>The '<tt>fptoui</tt>' instruction converts its
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004909 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4910 towards zero) unsigned integer value. If the value cannot fit
4911 in <tt>ty2</tt>, the results are undefined.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004912
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004913<h5>Example:</h5>
4914<pre>
Reid Spencer1539a1c2007-07-31 14:40:14 +00004915 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner88519042007-09-22 03:17:52 +00004916 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Gabor Greif6a292012009-10-28 09:21:30 +00004917 %Z = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004918</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004919
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004920</div>
4921
4922<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004923<h4>
Reid Spencerd4448792006-11-09 23:03:26 +00004924 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004925</h4>
4926
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004927<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004928
4929<h5>Syntax:</h5>
4930<pre>
Reid Spencerd4448792006-11-09 23:03:26 +00004931 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004932</pre>
4933
4934<h5>Overview:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004935<p>The '<tt>fptosi</tt>' instruction converts
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004936 <a href="#t_floating">floating point</a> <tt>value</tt> to
4937 type <tt>ty2</tt>.</p>
Chris Lattnercc37aae2004-03-12 05:50:16 +00004938
Chris Lattner6536cfe2002-05-06 22:08:29 +00004939<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004940<p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
4941 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4942 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4943 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4944 vector integer type with the same number of elements as <tt>ty</tt></p>
Chris Lattnercc37aae2004-03-12 05:50:16 +00004945
Chris Lattner6536cfe2002-05-06 22:08:29 +00004946<h5>Semantics:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004947<p>The '<tt>fptosi</tt>' instruction converts its
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004948 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4949 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4950 the results are undefined.</p>
Chris Lattnercc37aae2004-03-12 05:50:16 +00004951
Chris Lattner33ba0d92001-07-09 00:26:23 +00004952<h5>Example:</h5>
Chris Lattnercc37aae2004-03-12 05:50:16 +00004953<pre>
Reid Spencerc78f3372007-01-12 03:35:51 +00004954 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner88519042007-09-22 03:17:52 +00004955 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Gabor Greif6a292012009-10-28 09:21:30 +00004956 %Z = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004957</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004958
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004959</div>
4960
4961<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004962<h4>
Reid Spencerd4448792006-11-09 23:03:26 +00004963 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004964</h4>
4965
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004966<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004967
4968<h5>Syntax:</h5>
4969<pre>
Reid Spencerd4448792006-11-09 23:03:26 +00004970 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004971</pre>
4972
4973<h5>Overview:</h5>
Reid Spencerd4448792006-11-09 23:03:26 +00004974<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004975 integer and converts that value to the <tt>ty2</tt> type.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004976
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004977<h5>Arguments:</h5>
Nate Begemanb348d182007-11-17 03:58:34 +00004978<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004979 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4980 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4981 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4982 floating point type with the same number of elements as <tt>ty</tt></p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004983
4984<h5>Semantics:</h5>
Reid Spencerd4448792006-11-09 23:03:26 +00004985<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004986 integer quantity and converts it to the corresponding floating point
4987 value. If the value cannot fit in the floating point value, the results are
4988 undefined.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004989
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004990<h5>Example:</h5>
4991<pre>
Reid Spencerca86e162006-12-31 07:07:53 +00004992 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohman0e451ce2008-10-14 16:51:45 +00004993 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004994</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004995
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004996</div>
4997
4998<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004999<h4>
Reid Spencerd4448792006-11-09 23:03:26 +00005000 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005001</h4>
5002
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005003<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005004
5005<h5>Syntax:</h5>
5006<pre>
Reid Spencerd4448792006-11-09 23:03:26 +00005007 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005008</pre>
5009
5010<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005011<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
5012 and converts that value to the <tt>ty2</tt> type.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005013
5014<h5>Arguments:</h5>
Nate Begemanb348d182007-11-17 03:58:34 +00005015<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005016 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
5017 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
5018 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
5019 floating point type with the same number of elements as <tt>ty</tt></p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005020
5021<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005022<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
5023 quantity and converts it to the corresponding floating point value. If the
5024 value cannot fit in the floating point value, the results are undefined.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005025
5026<h5>Example:</h5>
5027<pre>
Reid Spencerca86e162006-12-31 07:07:53 +00005028 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohman0e451ce2008-10-14 16:51:45 +00005029 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005030</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005031
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005032</div>
5033
5034<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005035<h4>
Reid Spencer72679252006-11-11 21:00:47 +00005036 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005037</h4>
5038
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005039<div>
Reid Spencer72679252006-11-11 21:00:47 +00005040
5041<h5>Syntax:</h5>
5042<pre>
5043 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
5044</pre>
5045
5046<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005047<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
5048 the integer type <tt>ty2</tt>.</p>
Reid Spencer72679252006-11-11 21:00:47 +00005049
5050<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005051<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
5052 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
5053 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
Reid Spencer72679252006-11-11 21:00:47 +00005054
5055<h5>Semantics:</h5>
5056<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005057 <tt>ty2</tt> by interpreting the pointer value as an integer and either
5058 truncating or zero extending that value to the size of the integer type. If
5059 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
5060 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
5061 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
5062 change.</p>
Reid Spencer72679252006-11-11 21:00:47 +00005063
5064<h5>Example:</h5>
5065<pre>
Jeff Cohenb627eab2007-04-29 01:07:00 +00005066 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
5067 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
Reid Spencer72679252006-11-11 21:00:47 +00005068</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005069
Reid Spencer72679252006-11-11 21:00:47 +00005070</div>
5071
5072<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005073<h4>
Reid Spencer72679252006-11-11 21:00:47 +00005074 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005075</h4>
5076
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005077<div>
Reid Spencer72679252006-11-11 21:00:47 +00005078
5079<h5>Syntax:</h5>
5080<pre>
5081 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
5082</pre>
5083
5084<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005085<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
5086 pointer type, <tt>ty2</tt>.</p>
Reid Spencer72679252006-11-11 21:00:47 +00005087
5088<h5>Arguments:</h5>
Duncan Sands8036ca42007-03-30 12:22:09 +00005089<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005090 value to cast, and a type to cast it to, which must be a
5091 <a href="#t_pointer">pointer</a> type.</p>
Reid Spencer72679252006-11-11 21:00:47 +00005092
5093<h5>Semantics:</h5>
5094<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005095 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
5096 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
5097 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
5098 than the size of a pointer then a zero extension is done. If they are the
5099 same size, nothing is done (<i>no-op cast</i>).</p>
Reid Spencer72679252006-11-11 21:00:47 +00005100
5101<h5>Example:</h5>
5102<pre>
Jeff Cohenb627eab2007-04-29 01:07:00 +00005103 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
Gabor Greif6a292012009-10-28 09:21:30 +00005104 %Y = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
5105 %Z = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
Reid Spencer72679252006-11-11 21:00:47 +00005106</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005107
Reid Spencer72679252006-11-11 21:00:47 +00005108</div>
5109
5110<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005111<h4>
Reid Spencer5c0ef472006-11-11 23:08:07 +00005112 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005113</h4>
5114
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005115<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005116
5117<h5>Syntax:</h5>
5118<pre>
Reid Spencer5c0ef472006-11-11 23:08:07 +00005119 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005120</pre>
5121
5122<h5>Overview:</h5>
Reid Spencer5c0ef472006-11-11 23:08:07 +00005123<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005124 <tt>ty2</tt> without changing any bits.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005125
5126<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005127<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
5128 non-aggregate first class value, and a type to cast it to, which must also be
5129 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
5130 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
5131 identical. If the source type is a pointer, the destination type must also be
5132 a pointer. This instruction supports bitwise conversion of vectors to
5133 integers and to vectors of other types (as long as they have the same
5134 size).</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005135
5136<h5>Semantics:</h5>
Reid Spencer5c0ef472006-11-11 23:08:07 +00005137<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005138 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
5139 this conversion. The conversion is done as if the <tt>value</tt> had been
5140 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
5141 be converted to other pointer types with this instruction. To convert
5142 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
5143 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005144
5145<h5>Example:</h5>
5146<pre>
Jeff Cohenb627eab2007-04-29 01:07:00 +00005147 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
Reid Spencerca86e162006-12-31 07:07:53 +00005148 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005149 %Z = bitcast &lt;2 x int&gt; %V to i64; <i>; yields i64: %V</i>
Chris Lattner33ba0d92001-07-09 00:26:23 +00005150</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005151
Misha Brukman9d0919f2003-11-08 01:05:38 +00005152</div>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005153
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005154</div>
5155
Reid Spencer2fd21e62006-11-08 01:18:52 +00005156<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005157<h3>
5158 <a name="otherops">Other Operations</a>
5159</h3>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005160
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005161<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005162
5163<p>The instructions in this category are the "miscellaneous" instructions, which
5164 defy better classification.</p>
5165
Reid Spencerf3a70a62006-11-18 21:50:54 +00005166<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005167<h4>
5168 <a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
5169</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005170
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005171<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005172
Reid Spencerf3a70a62006-11-18 21:50:54 +00005173<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005174<pre>
5175 &lt;result&gt; = icmp &lt;cond&gt; &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {i1} or {&lt;N x i1&gt;}:result</i>
Reid Spencerf3a70a62006-11-18 21:50:54 +00005176</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005177
Reid Spencerf3a70a62006-11-18 21:50:54 +00005178<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005179<p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
5180 boolean values based on comparison of its two integer, integer vector, or
5181 pointer operands.</p>
5182
Reid Spencerf3a70a62006-11-18 21:50:54 +00005183<h5>Arguments:</h5>
5184<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005185 the condition code indicating the kind of comparison to perform. It is not a
5186 value, just a keyword. The possible condition code are:</p>
5187
Reid Spencerf3a70a62006-11-18 21:50:54 +00005188<ol>
5189 <li><tt>eq</tt>: equal</li>
5190 <li><tt>ne</tt>: not equal </li>
5191 <li><tt>ugt</tt>: unsigned greater than</li>
5192 <li><tt>uge</tt>: unsigned greater or equal</li>
5193 <li><tt>ult</tt>: unsigned less than</li>
5194 <li><tt>ule</tt>: unsigned less or equal</li>
5195 <li><tt>sgt</tt>: signed greater than</li>
5196 <li><tt>sge</tt>: signed greater or equal</li>
5197 <li><tt>slt</tt>: signed less than</li>
5198 <li><tt>sle</tt>: signed less or equal</li>
5199</ol>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005200
Chris Lattner3b19d652007-01-15 01:54:13 +00005201<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005202 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
5203 typed. They must also be identical types.</p>
5204
Reid Spencerf3a70a62006-11-18 21:50:54 +00005205<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005206<p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
5207 condition code given as <tt>cond</tt>. The comparison performed always yields
Nick Lewyckyec38da42009-09-27 00:45:11 +00005208 either an <a href="#t_integer"><tt>i1</tt></a> or vector of <tt>i1</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005209 result, as follows:</p>
5210
Reid Spencerf3a70a62006-11-18 21:50:54 +00005211<ol>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005212 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005213 <tt>false</tt> otherwise. No sign interpretation is necessary or
5214 performed.</li>
5215
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005216 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005217 <tt>false</tt> otherwise. No sign interpretation is necessary or
5218 performed.</li>
5219
Reid Spencerf3a70a62006-11-18 21:50:54 +00005220 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005221 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5222
Reid Spencerf3a70a62006-11-18 21:50:54 +00005223 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005224 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5225 to <tt>op2</tt>.</li>
5226
Reid Spencerf3a70a62006-11-18 21:50:54 +00005227 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005228 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5229
Reid Spencerf3a70a62006-11-18 21:50:54 +00005230 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005231 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5232
Reid Spencerf3a70a62006-11-18 21:50:54 +00005233 <li><tt>sgt</tt>: interprets the operands as signed values and yields
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005234 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5235
Reid Spencerf3a70a62006-11-18 21:50:54 +00005236 <li><tt>sge</tt>: interprets the operands as signed values and yields
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005237 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5238 to <tt>op2</tt>.</li>
5239
Reid Spencerf3a70a62006-11-18 21:50:54 +00005240 <li><tt>slt</tt>: interprets the operands as signed values and yields
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005241 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5242
Reid Spencerf3a70a62006-11-18 21:50:54 +00005243 <li><tt>sle</tt>: interprets the operands as signed values and yields
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005244 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Reid Spencerf3a70a62006-11-18 21:50:54 +00005245</ol>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005246
Reid Spencerf3a70a62006-11-18 21:50:54 +00005247<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005248 values are compared as if they were integers.</p>
5249
5250<p>If the operands are integer vectors, then they are compared element by
5251 element. The result is an <tt>i1</tt> vector with the same number of elements
5252 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
Reid Spencerf3a70a62006-11-18 21:50:54 +00005253
5254<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005255<pre>
5256 &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
Reid Spencerca86e162006-12-31 07:07:53 +00005257 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
5258 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
5259 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
5260 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
5261 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
Reid Spencerf3a70a62006-11-18 21:50:54 +00005262</pre>
Dan Gohmanc766f722009-01-22 01:39:38 +00005263
5264<p>Note that the code generator does not yet support vector types with
5265 the <tt>icmp</tt> instruction.</p>
5266
Reid Spencerf3a70a62006-11-18 21:50:54 +00005267</div>
5268
5269<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005270<h4>
5271 <a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
5272</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005273
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005274<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005275
Reid Spencerf3a70a62006-11-18 21:50:54 +00005276<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005277<pre>
5278 &lt;result&gt; = fcmp &lt;cond&gt; &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {i1} or {&lt;N x i1&gt;}:result</i>
Reid Spencerf3a70a62006-11-18 21:50:54 +00005279</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005280
Reid Spencerf3a70a62006-11-18 21:50:54 +00005281<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005282<p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
5283 values based on comparison of its operands.</p>
5284
5285<p>If the operands are floating point scalars, then the result type is a boolean
Nick Lewyckyec38da42009-09-27 00:45:11 +00005286(<a href="#t_integer"><tt>i1</tt></a>).</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005287
5288<p>If the operands are floating point vectors, then the result type is a vector
5289 of boolean with the same number of elements as the operands being
5290 compared.</p>
5291
Reid Spencerf3a70a62006-11-18 21:50:54 +00005292<h5>Arguments:</h5>
5293<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005294 the condition code indicating the kind of comparison to perform. It is not a
5295 value, just a keyword. The possible condition code are:</p>
5296
Reid Spencerf3a70a62006-11-18 21:50:54 +00005297<ol>
Reid Spencerb7f26282006-11-19 03:00:14 +00005298 <li><tt>false</tt>: no comparison, always returns false</li>
Reid Spencerf3a70a62006-11-18 21:50:54 +00005299 <li><tt>oeq</tt>: ordered and equal</li>
5300 <li><tt>ogt</tt>: ordered and greater than </li>
5301 <li><tt>oge</tt>: ordered and greater than or equal</li>
5302 <li><tt>olt</tt>: ordered and less than </li>
5303 <li><tt>ole</tt>: ordered and less than or equal</li>
5304 <li><tt>one</tt>: ordered and not equal</li>
5305 <li><tt>ord</tt>: ordered (no nans)</li>
5306 <li><tt>ueq</tt>: unordered or equal</li>
5307 <li><tt>ugt</tt>: unordered or greater than </li>
5308 <li><tt>uge</tt>: unordered or greater than or equal</li>
5309 <li><tt>ult</tt>: unordered or less than </li>
5310 <li><tt>ule</tt>: unordered or less than or equal</li>
5311 <li><tt>une</tt>: unordered or not equal</li>
5312 <li><tt>uno</tt>: unordered (either nans)</li>
Reid Spencerb7f26282006-11-19 03:00:14 +00005313 <li><tt>true</tt>: no comparison, always returns true</li>
Reid Spencerf3a70a62006-11-18 21:50:54 +00005314</ol>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005315
Jeff Cohenb627eab2007-04-29 01:07:00 +00005316<p><i>Ordered</i> means that neither operand is a QNAN while
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005317 <i>unordered</i> means that either operand may be a QNAN.</p>
5318
5319<p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
5320 a <a href="#t_floating">floating point</a> type or
5321 a <a href="#t_vector">vector</a> of floating point type. They must have
5322 identical types.</p>
5323
Reid Spencerf3a70a62006-11-18 21:50:54 +00005324<h5>Semantics:</h5>
Gabor Greiffb224a22008-08-07 21:46:00 +00005325<p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005326 according to the condition code given as <tt>cond</tt>. If the operands are
5327 vectors, then the vectors are compared element by element. Each comparison
Nick Lewyckyec38da42009-09-27 00:45:11 +00005328 performed always yields an <a href="#t_integer">i1</a> result, as
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005329 follows:</p>
5330
Reid Spencerf3a70a62006-11-18 21:50:54 +00005331<ol>
5332 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005333
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005334 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005335 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5336
Reid Spencerb7f26282006-11-19 03:00:14 +00005337 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00005338 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005339
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005340 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005341 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5342
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005343 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005344 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5345
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005346 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005347 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5348
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005349 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005350 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5351
Reid Spencerb7f26282006-11-19 03:00:14 +00005352 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005353
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005354 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005355 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5356
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005357 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005358 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5359
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005360 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005361 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5362
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005363 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005364 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5365
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005366 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005367 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5368
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005369 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005370 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5371
Reid Spencerb7f26282006-11-19 03:00:14 +00005372 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005373
Reid Spencerf3a70a62006-11-18 21:50:54 +00005374 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
5375</ol>
Reid Spencerf3a70a62006-11-18 21:50:54 +00005376
5377<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005378<pre>
5379 &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
Dan Gohmanf72fb672008-09-09 01:02:47 +00005380 &lt;result&gt; = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
5381 &lt;result&gt; = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
5382 &lt;result&gt; = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
Reid Spencerf3a70a62006-11-18 21:50:54 +00005383</pre>
Dan Gohmanc766f722009-01-22 01:39:38 +00005384
5385<p>Note that the code generator does not yet support vector types with
5386 the <tt>fcmp</tt> instruction.</p>
5387
Reid Spencerf3a70a62006-11-18 21:50:54 +00005388</div>
5389
Reid Spencer2fd21e62006-11-08 01:18:52 +00005390<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005391<h4>
Chris Lattner5568e942008-05-20 20:48:21 +00005392 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005393</h4>
Chris Lattner5568e942008-05-20 20:48:21 +00005394
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005395<div>
Chris Lattner5568e942008-05-20 20:48:21 +00005396
Reid Spencer2fd21e62006-11-08 01:18:52 +00005397<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005398<pre>
5399 &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...
5400</pre>
Chris Lattner5568e942008-05-20 20:48:21 +00005401
Reid Spencer2fd21e62006-11-08 01:18:52 +00005402<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005403<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in the
5404 SSA graph representing the function.</p>
5405
Reid Spencer2fd21e62006-11-08 01:18:52 +00005406<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005407<p>The type of the incoming values is specified with the first type field. After
5408 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
5409 one pair for each predecessor basic block of the current block. Only values
5410 of <a href="#t_firstclass">first class</a> type may be used as the value
5411 arguments to the PHI node. Only labels may be used as the label
5412 arguments.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00005413
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005414<p>There must be no non-phi instructions between the start of a basic block and
5415 the PHI instructions: i.e. PHI instructions must be first in a basic
5416 block.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00005417
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005418<p>For the purposes of the SSA form, the use of each incoming value is deemed to
5419 occur on the edge from the corresponding predecessor block to the current
5420 block (but after any definition of an '<tt>invoke</tt>' instruction's return
5421 value on the same edge).</p>
Jay Foadd2449092009-06-03 10:20:10 +00005422
Reid Spencer2fd21e62006-11-08 01:18:52 +00005423<h5>Semantics:</h5>
Jeff Cohenb627eab2007-04-29 01:07:00 +00005424<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005425 specified by the pair corresponding to the predecessor basic block that
5426 executed just prior to the current block.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00005427
Reid Spencer2fd21e62006-11-08 01:18:52 +00005428<h5>Example:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00005429<pre>
5430Loop: ; Infinite loop that counts from 0 on up...
5431 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
5432 %nextindvar = add i32 %indvar, 1
5433 br label %Loop
5434</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005435
Reid Spencer2fd21e62006-11-08 01:18:52 +00005436</div>
5437
Chris Lattnercc37aae2004-03-12 05:50:16 +00005438<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005439<h4>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005440 <a name="i_select">'<tt>select</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005441</h4>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005442
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005443<div>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005444
5445<h5>Syntax:</h5>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005446<pre>
Dan Gohmanf72fb672008-09-09 01:02:47 +00005447 &lt;result&gt; = select <i>selty</i> &lt;cond&gt;, &lt;ty&gt; &lt;val1&gt;, &lt;ty&gt; &lt;val2&gt; <i>; yields ty</i>
5448
Dan Gohman0e451ce2008-10-14 16:51:45 +00005449 <i>selty</i> is either i1 or {&lt;N x i1&gt;}
Chris Lattnercc37aae2004-03-12 05:50:16 +00005450</pre>
5451
5452<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005453<p>The '<tt>select</tt>' instruction is used to choose one value based on a
5454 condition, without branching.</p>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005455
5456
5457<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005458<p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
5459 values indicating the condition, and two values of the
5460 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
5461 vectors and the condition is a scalar, then entire vectors are selected, not
5462 individual elements.</p>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005463
5464<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005465<p>If the condition is an i1 and it evaluates to 1, the instruction returns the
5466 first value argument; otherwise, it returns the second value argument.</p>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005467
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005468<p>If the condition is a vector of i1, then the value arguments must be vectors
5469 of the same size, and the selection is done element by element.</p>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005470
5471<h5>Example:</h5>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005472<pre>
Reid Spencerc78f3372007-01-12 03:35:51 +00005473 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005474</pre>
Dan Gohmanc766f722009-01-22 01:39:38 +00005475
5476<p>Note that the code generator does not yet support conditions
5477 with vector type.</p>
5478
Chris Lattnercc37aae2004-03-12 05:50:16 +00005479</div>
5480
Robert Bocchino05ccd702006-01-15 20:48:27 +00005481<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005482<h4>
Chris Lattner2bff5242005-05-06 05:47:36 +00005483 <a name="i_call">'<tt>call</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005484</h4>
Chris Lattner2bff5242005-05-06 05:47:36 +00005485
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005486<div>
Chris Lattner2bff5242005-05-06 05:47:36 +00005487
Chris Lattner00950542001-06-06 20:29:01 +00005488<h5>Syntax:</h5>
Chris Lattner2bff5242005-05-06 05:47:36 +00005489<pre>
Devang Patel307e8ab2008-10-07 17:48:33 +00005490 &lt;result&gt; = [tail] call [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>] &lt;ty&gt; [&lt;fnty&gt;*] &lt;fnptrval&gt;(&lt;function args&gt;) [<a href="#fnattrs">fn attrs</a>]
Chris Lattner2bff5242005-05-06 05:47:36 +00005491</pre>
5492
Chris Lattner00950542001-06-06 20:29:01 +00005493<h5>Overview:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00005494<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
Chris Lattner2bff5242005-05-06 05:47:36 +00005495
Chris Lattner00950542001-06-06 20:29:01 +00005496<h5>Arguments:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00005497<p>This instruction requires several arguments:</p>
Chris Lattner2bff5242005-05-06 05:47:36 +00005498
Chris Lattner6536cfe2002-05-06 22:08:29 +00005499<ol>
Jeffrey Yasskin95fa80a2010-01-09 19:44:16 +00005500 <li>The optional "tail" marker indicates that the callee function does not
5501 access any allocas or varargs in the caller. Note that calls may be
5502 marked "tail" even if they do not occur before
5503 a <a href="#i_ret"><tt>ret</tt></a> instruction. If the "tail" marker is
5504 present, the function call is eligible for tail call optimization,
5505 but <a href="CodeGenerator.html#tailcallopt">might not in fact be
Evan Chengdc444e92010-03-08 21:05:02 +00005506 optimized into a jump</a>. The code generator may optimize calls marked
5507 "tail" with either 1) automatic <a href="CodeGenerator.html#sibcallopt">
5508 sibling call optimization</a> when the caller and callee have
5509 matching signatures, or 2) forced tail call optimization when the
5510 following extra requirements are met:
Jeffrey Yasskin95fa80a2010-01-09 19:44:16 +00005511 <ul>
5512 <li>Caller and callee both have the calling
5513 convention <tt>fastcc</tt>.</li>
5514 <li>The call is in tail position (ret immediately follows call and ret
5515 uses value of call or is void).</li>
5516 <li>Option <tt>-tailcallopt</tt> is enabled,
Dan Gohmanfbbee8d2010-03-02 01:08:11 +00005517 or <code>llvm::GuaranteedTailCallOpt</code> is <code>true</code>.</li>
Jeffrey Yasskin95fa80a2010-01-09 19:44:16 +00005518 <li><a href="CodeGenerator.html#tailcallopt">Platform specific
5519 constraints are met.</a></li>
5520 </ul>
5521 </li>
Devang Patelf642f472008-10-06 18:50:38 +00005522
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005523 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
5524 convention</a> the call should use. If none is specified, the call
Jeffrey Yasskin95fa80a2010-01-09 19:44:16 +00005525 defaults to using C calling conventions. The calling convention of the
5526 call must match the calling convention of the target function, or else the
5527 behavior is undefined.</li>
Devang Patelf642f472008-10-06 18:50:38 +00005528
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005529 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
5530 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
5531 '<tt>inreg</tt>' attributes are valid here.</li>
5532
5533 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
5534 type of the return value. Functions that return no value are marked
5535 <tt><a href="#t_void">void</a></tt>.</li>
5536
5537 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
5538 being invoked. The argument types must match the types implied by this
5539 signature. This type can be omitted if the function is not varargs and if
5540 the function type does not return a pointer to a function.</li>
5541
5542 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
5543 be invoked. In most cases, this is a direct function invocation, but
5544 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
5545 to function value.</li>
5546
5547 <li>'<tt>function args</tt>': argument list whose types match the function
Chris Lattner0724fbd2010-03-02 06:36:51 +00005548 signature argument types and parameter attributes. All arguments must be
5549 of <a href="#t_firstclass">first class</a> type. If the function
5550 signature indicates the function accepts a variable number of arguments,
5551 the extra arguments can be specified.</li>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005552
5553 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
5554 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
5555 '<tt>readnone</tt>' attributes are valid here.</li>
Chris Lattner6536cfe2002-05-06 22:08:29 +00005556</ol>
Chris Lattner2bff5242005-05-06 05:47:36 +00005557
Chris Lattner00950542001-06-06 20:29:01 +00005558<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005559<p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
5560 a specified function, with its incoming arguments bound to the specified
5561 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
5562 function, control flow continues with the instruction after the function
5563 call, and the return value of the function is bound to the result
5564 argument.</p>
Chris Lattner2bff5242005-05-06 05:47:36 +00005565
Chris Lattner00950542001-06-06 20:29:01 +00005566<h5>Example:</h5>
Chris Lattner2bff5242005-05-06 05:47:36 +00005567<pre>
Nick Lewyckydb7e3c92007-09-08 13:57:50 +00005568 %retval = call i32 @test(i32 %argc)
Dan Gohmanfe47aae2010-05-28 17:13:49 +00005569 call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42) <i>; yields i32</i>
Chris Lattner772fccf2008-03-21 17:24:17 +00005570 %X = tail call i32 @foo() <i>; yields i32</i>
5571 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
5572 call void %foo(i8 97 signext)
Devang Patelc3fc6df2008-03-10 20:49:15 +00005573
5574 %struct.A = type { i32, i8 }
Devang Patelf642f472008-10-06 18:50:38 +00005575 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
Dan Gohmanb1e6b962008-10-04 19:00:07 +00005576 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
5577 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
Chris Lattner85a350f2008-10-08 06:26:11 +00005578 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
Matthijs Kooijmancb73d192008-10-07 10:03:45 +00005579 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
Chris Lattner2bff5242005-05-06 05:47:36 +00005580</pre>
5581
Dale Johannesen07de8d12009-09-24 18:38:21 +00005582<p>llvm treats calls to some functions with names and arguments that match the
Dale Johannesen9f8380b2009-09-25 17:04:42 +00005583standard C99 library as being the C99 library functions, and may perform
5584optimizations or generate code for them under that assumption. This is
5585something we'd like to change in the future to provide better support for
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00005586freestanding environments and non-C-based languages.</p>
Dale Johannesen07de8d12009-09-24 18:38:21 +00005587
Misha Brukman9d0919f2003-11-08 01:05:38 +00005588</div>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005589
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005590<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005591<h4>
Chris Lattnerfb6977d2006-01-13 23:26:01 +00005592 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005593</h4>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005594
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005595<div>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005596
Chris Lattner8d1a81d2003-10-18 05:51:36 +00005597<h5>Syntax:</h5>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005598<pre>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00005599 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
Chris Lattnere19d7a72004-09-27 21:51:25 +00005600</pre>
5601
Chris Lattner8d1a81d2003-10-18 05:51:36 +00005602<h5>Overview:</h5>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00005603<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005604 the "variable argument" area of a function call. It is used to implement the
5605 <tt>va_arg</tt> macro in C.</p>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005606
Chris Lattner8d1a81d2003-10-18 05:51:36 +00005607<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005608<p>This instruction takes a <tt>va_list*</tt> value and the type of the
5609 argument. It returns a value of the specified argument type and increments
5610 the <tt>va_list</tt> to point to the next argument. The actual type
5611 of <tt>va_list</tt> is target specific.</p>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005612
Chris Lattner8d1a81d2003-10-18 05:51:36 +00005613<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005614<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
5615 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
5616 to the next argument. For more information, see the variable argument
5617 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005618
5619<p>It is legal for this instruction to be called in a function which does not
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005620 take a variable number of arguments, for example, the <tt>vfprintf</tt>
5621 function.</p>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005622
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005623<p><tt>va_arg</tt> is an LLVM instruction instead of
5624 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
5625 argument.</p>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005626
Chris Lattner8d1a81d2003-10-18 05:51:36 +00005627<h5>Example:</h5>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005628<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
5629
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005630<p>Note that the code generator does not yet fully support va_arg on many
5631 targets. Also, it does not currently support va_arg with aggregate types on
5632 any target.</p>
Dan Gohmanf3e60bd2009-01-12 23:12:39 +00005633
Misha Brukman9d0919f2003-11-08 01:05:38 +00005634</div>
Chris Lattner8ff75902004-01-06 05:31:32 +00005635
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005636</div>
5637
5638</div>
5639
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005640<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005641<h2><a name="intrinsics">Intrinsic Functions</a></h2>
Chris Lattner261efe92003-11-25 01:02:51 +00005642<!-- *********************************************************************** -->
Chris Lattner8ff75902004-01-06 05:31:32 +00005643
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005644<div>
Chris Lattner33aec9e2004-02-12 17:01:32 +00005645
5646<p>LLVM supports the notion of an "intrinsic function". These functions have
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005647 well known names and semantics and are required to follow certain
5648 restrictions. Overall, these intrinsics represent an extension mechanism for
5649 the LLVM language that does not require changing all of the transformations
5650 in LLVM when adding to the language (or the bitcode reader/writer, the
5651 parser, etc...).</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00005652
John Criswellfc6b8952005-05-16 16:17:45 +00005653<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005654 prefix is reserved in LLVM for intrinsic names; thus, function names may not
5655 begin with this prefix. Intrinsic functions must always be external
5656 functions: you cannot define the body of intrinsic functions. Intrinsic
5657 functions may only be used in call or invoke instructions: it is illegal to
5658 take the address of an intrinsic function. Additionally, because intrinsic
5659 functions are part of the LLVM language, it is required if any are added that
5660 they be documented here.</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00005661
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005662<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
5663 family of functions that perform the same operation but on different data
5664 types. Because LLVM can represent over 8 million different integer types,
5665 overloading is used commonly to allow an intrinsic function to operate on any
5666 integer type. One or more of the argument types or the result type can be
5667 overloaded to accept any integer type. Argument types may also be defined as
5668 exactly matching a previous argument's type or the result type. This allows
5669 an intrinsic function which accepts multiple arguments, but needs all of them
5670 to be of the same type, to only be overloaded with respect to a single
5671 argument or the result.</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00005672
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005673<p>Overloaded intrinsics will have the names of its overloaded argument types
5674 encoded into its function name, each preceded by a period. Only those types
5675 which are overloaded result in a name suffix. Arguments whose type is matched
5676 against another type do not. For example, the <tt>llvm.ctpop</tt> function
5677 can take an integer of any width and returns an integer of exactly the same
5678 integer width. This leads to a family of functions such as
5679 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
5680 %val)</tt>. Only one type, the return type, is overloaded, and only one type
5681 suffix is required. Because the argument's type is matched against the return
5682 type, it does not require its own name suffix.</p>
Reid Spencer409e28f2007-04-01 08:04:23 +00005683
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005684<p>To learn how to add an intrinsic function, please see the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005685 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00005686
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005687<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005688<h3>
Chris Lattner8ff75902004-01-06 05:31:32 +00005689 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005690</h3>
Chris Lattner8ff75902004-01-06 05:31:32 +00005691
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005692<div>
Chris Lattnerd7923912004-05-23 21:06:01 +00005693
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005694<p>Variable argument support is defined in LLVM with
5695 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
5696 intrinsic functions. These functions are related to the similarly named
5697 macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005698
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005699<p>All of these functions operate on arguments that use a target-specific value
5700 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
5701 not define what this type is, so all transformations should be prepared to
5702 handle these functions regardless of the type used.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005703
Chris Lattner374ab302006-05-15 17:26:46 +00005704<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005705 instruction and the variable argument handling intrinsic functions are
5706 used.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005707
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00005708<pre class="doc_code">
Anton Korobeynikov5d522f32007-03-21 23:58:04 +00005709define i32 @test(i32 %X, ...) {
Chris Lattner33aec9e2004-02-12 17:01:32 +00005710 ; Initialize variable argument processing
Jeff Cohenb627eab2007-04-29 01:07:00 +00005711 %ap = alloca i8*
Chris Lattnerb75137d2007-01-08 07:55:15 +00005712 %ap2 = bitcast i8** %ap to i8*
Anton Korobeynikov5d522f32007-03-21 23:58:04 +00005713 call void @llvm.va_start(i8* %ap2)
Chris Lattner33aec9e2004-02-12 17:01:32 +00005714
5715 ; Read a single integer argument
Jeff Cohenb627eab2007-04-29 01:07:00 +00005716 %tmp = va_arg i8** %ap, i32
Chris Lattner33aec9e2004-02-12 17:01:32 +00005717
5718 ; Demonstrate usage of llvm.va_copy and llvm.va_end
Jeff Cohenb627eab2007-04-29 01:07:00 +00005719 %aq = alloca i8*
Chris Lattnerb75137d2007-01-08 07:55:15 +00005720 %aq2 = bitcast i8** %aq to i8*
Jeff Cohenb627eab2007-04-29 01:07:00 +00005721 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
Anton Korobeynikov5d522f32007-03-21 23:58:04 +00005722 call void @llvm.va_end(i8* %aq2)
Chris Lattner33aec9e2004-02-12 17:01:32 +00005723
5724 ; Stop processing of arguments.
Anton Korobeynikov5d522f32007-03-21 23:58:04 +00005725 call void @llvm.va_end(i8* %ap2)
Reid Spencerca86e162006-12-31 07:07:53 +00005726 ret i32 %tmp
Chris Lattner33aec9e2004-02-12 17:01:32 +00005727}
Anton Korobeynikov5d522f32007-03-21 23:58:04 +00005728
5729declare void @llvm.va_start(i8*)
5730declare void @llvm.va_copy(i8*, i8*)
5731declare void @llvm.va_end(i8*)
Chris Lattner33aec9e2004-02-12 17:01:32 +00005732</pre>
Chris Lattner8ff75902004-01-06 05:31:32 +00005733
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005734<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005735<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00005736 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005737</h4>
Chris Lattner8ff75902004-01-06 05:31:32 +00005738
5739
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005740<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005741
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005742<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005743<pre>
5744 declare void %llvm.va_start(i8* &lt;arglist&gt;)
5745</pre>
5746
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005747<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005748<p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*&lt;arglist&gt;</tt>
5749 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00005750
5751<h5>Arguments:</h5>
Dan Gohman0e451ce2008-10-14 16:51:45 +00005752<p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00005753
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005754<h5>Semantics:</h5>
Dan Gohman0e451ce2008-10-14 16:51:45 +00005755<p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005756 macro available in C. In a target-dependent way, it initializes
5757 the <tt>va_list</tt> element to which the argument points, so that the next
5758 call to <tt>va_arg</tt> will produce the first variable argument passed to
5759 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
5760 need to know the last argument of the function as the compiler can figure
5761 that out.</p>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00005762
Misha Brukman9d0919f2003-11-08 01:05:38 +00005763</div>
Chris Lattner8ff75902004-01-06 05:31:32 +00005764
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005765<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005766<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00005767 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005768</h4>
Chris Lattner8ff75902004-01-06 05:31:32 +00005769
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005770<div>
Chris Lattnerb75137d2007-01-08 07:55:15 +00005771
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005772<h5>Syntax:</h5>
5773<pre>
5774 declare void @llvm.va_end(i8* &lt;arglist&gt;)
5775</pre>
5776
5777<h5>Overview:</h5>
Jeff Cohenb627eab2007-04-29 01:07:00 +00005778<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005779 which has been initialized previously
5780 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
5781 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
Chris Lattnerb75137d2007-01-08 07:55:15 +00005782
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005783<h5>Arguments:</h5>
Jeff Cohenb627eab2007-04-29 01:07:00 +00005784<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
Chris Lattnerb75137d2007-01-08 07:55:15 +00005785
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005786<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00005787<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005788 macro available in C. In a target-dependent way, it destroys
5789 the <tt>va_list</tt> element to which the argument points. Calls
5790 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
5791 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
5792 with calls to <tt>llvm.va_end</tt>.</p>
Chris Lattnerb75137d2007-01-08 07:55:15 +00005793
Misha Brukman9d0919f2003-11-08 01:05:38 +00005794</div>
Chris Lattner8ff75902004-01-06 05:31:32 +00005795
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005796<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005797<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00005798 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005799</h4>
Chris Lattner8ff75902004-01-06 05:31:32 +00005800
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005801<div>
Chris Lattnerd7923912004-05-23 21:06:01 +00005802
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005803<h5>Syntax:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00005804<pre>
Anton Korobeynikovec43a062007-03-22 00:02:17 +00005805 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
Chris Lattnerd7923912004-05-23 21:06:01 +00005806</pre>
5807
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005808<h5>Overview:</h5>
Jeff Cohenb627eab2007-04-29 01:07:00 +00005809<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005810 from the source argument list to the destination argument list.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005811
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005812<h5>Arguments:</h5>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00005813<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005814 The second argument is a pointer to a <tt>va_list</tt> element to copy
5815 from.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005816
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005817<h5>Semantics:</h5>
Jeff Cohenb627eab2007-04-29 01:07:00 +00005818<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005819 macro available in C. In a target-dependent way, it copies the
5820 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
5821 element. This intrinsic is necessary because
5822 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
5823 arbitrarily complex and require, for example, memory allocation.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005824
Misha Brukman9d0919f2003-11-08 01:05:38 +00005825</div>
Chris Lattner8ff75902004-01-06 05:31:32 +00005826
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005827</div>
5828
Chris Lattner33aec9e2004-02-12 17:01:32 +00005829<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005830<h3>
Chris Lattnerd7923912004-05-23 21:06:01 +00005831 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005832</h3>
Chris Lattnerd7923912004-05-23 21:06:01 +00005833
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005834<div>
Chris Lattnerd7923912004-05-23 21:06:01 +00005835
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005836<p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
Chris Lattnerd3eda892008-08-05 18:29:16 +00005837Collection</a> (GC) requires the implementation and generation of these
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005838intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
5839roots on the stack</a>, as well as garbage collector implementations that
5840require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
5841barriers. Front-ends for type-safe garbage collected languages should generate
5842these intrinsics to make use of the LLVM garbage collectors. For more details,
5843see <a href="GarbageCollection.html">Accurate Garbage Collection with
5844LLVM</a>.</p>
Christopher Lamb303dae92007-12-17 01:00:21 +00005845
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005846<p>The garbage collection intrinsics only operate on objects in the generic
5847 address space (address space zero).</p>
Christopher Lamb303dae92007-12-17 01:00:21 +00005848
Chris Lattnerd7923912004-05-23 21:06:01 +00005849<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005850<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00005851 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005852</h4>
Chris Lattnerd7923912004-05-23 21:06:01 +00005853
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005854<div>
Chris Lattnerd7923912004-05-23 21:06:01 +00005855
5856<h5>Syntax:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00005857<pre>
Chris Lattner1df4f752007-09-21 17:30:40 +00005858 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Chris Lattnerd7923912004-05-23 21:06:01 +00005859</pre>
5860
5861<h5>Overview:</h5>
John Criswell9e2485c2004-12-10 15:51:16 +00005862<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005863 the code generator, and allows some metadata to be associated with it.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005864
5865<h5>Arguments:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00005866<p>The first argument specifies the address of a stack object that contains the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005867 root pointer. The second pointer (which must be either a constant or a
5868 global value address) contains the meta-data to be associated with the
5869 root.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005870
5871<h5>Semantics:</h5>
Chris Lattner05d67092008-04-24 05:59:56 +00005872<p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005873 location. At compile-time, the code generator generates information to allow
5874 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5875 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5876 algorithm</a>.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005877
5878</div>
5879
Chris Lattnerd7923912004-05-23 21:06:01 +00005880<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005881<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00005882 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005883</h4>
Chris Lattnerd7923912004-05-23 21:06:01 +00005884
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005885<div>
Chris Lattnerd7923912004-05-23 21:06:01 +00005886
5887<h5>Syntax:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00005888<pre>
Chris Lattner1df4f752007-09-21 17:30:40 +00005889 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Chris Lattnerd7923912004-05-23 21:06:01 +00005890</pre>
5891
5892<h5>Overview:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00005893<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005894 locations, allowing garbage collector implementations that require read
5895 barriers.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005896
5897<h5>Arguments:</h5>
Chris Lattner80626e92006-03-14 20:02:51 +00005898<p>The second argument is the address to read from, which should be an address
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005899 allocated from the garbage collector. The first object is a pointer to the
5900 start of the referenced object, if needed by the language runtime (otherwise
5901 null).</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005902
5903<h5>Semantics:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00005904<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005905 instruction, but may be replaced with substantially more complex code by the
5906 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5907 may only be used in a function which <a href="#gc">specifies a GC
5908 algorithm</a>.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005909
5910</div>
5911
Chris Lattnerd7923912004-05-23 21:06:01 +00005912<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005913<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00005914 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005915</h4>
Chris Lattnerd7923912004-05-23 21:06:01 +00005916
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005917<div>
Chris Lattnerd7923912004-05-23 21:06:01 +00005918
5919<h5>Syntax:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00005920<pre>
Chris Lattner1df4f752007-09-21 17:30:40 +00005921 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Chris Lattnerd7923912004-05-23 21:06:01 +00005922</pre>
5923
5924<h5>Overview:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00005925<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005926 locations, allowing garbage collector implementations that require write
5927 barriers (such as generational or reference counting collectors).</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005928
5929<h5>Arguments:</h5>
Chris Lattner80626e92006-03-14 20:02:51 +00005930<p>The first argument is the reference to store, the second is the start of the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005931 object to store it to, and the third is the address of the field of Obj to
5932 store to. If the runtime does not require a pointer to the object, Obj may
5933 be null.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005934
5935<h5>Semantics:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00005936<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005937 instruction, but may be replaced with substantially more complex code by the
5938 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5939 may only be used in a function which <a href="#gc">specifies a GC
5940 algorithm</a>.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005941
5942</div>
5943
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005944</div>
5945
Chris Lattnerd7923912004-05-23 21:06:01 +00005946<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005947<h3>
Chris Lattner10610642004-02-14 04:08:35 +00005948 <a name="int_codegen">Code Generator Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005949</h3>
Chris Lattner10610642004-02-14 04:08:35 +00005950
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005951<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005952
5953<p>These intrinsics are provided by LLVM to expose special features that may
5954 only be implemented with code generator support.</p>
Chris Lattner10610642004-02-14 04:08:35 +00005955
Chris Lattner10610642004-02-14 04:08:35 +00005956<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005957<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00005958 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005959</h4>
Chris Lattner10610642004-02-14 04:08:35 +00005960
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005961<div>
Chris Lattner10610642004-02-14 04:08:35 +00005962
5963<h5>Syntax:</h5>
5964<pre>
Anton Korobeynikovec43a062007-03-22 00:02:17 +00005965 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
Chris Lattner10610642004-02-14 04:08:35 +00005966</pre>
5967
5968<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005969<p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5970 target-specific value indicating the return address of the current function
5971 or one of its callers.</p>
Chris Lattner10610642004-02-14 04:08:35 +00005972
5973<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005974<p>The argument to this intrinsic indicates which function to return the address
5975 for. Zero indicates the calling function, one indicates its caller, etc.
5976 The argument is <b>required</b> to be a constant integer value.</p>
Chris Lattner10610642004-02-14 04:08:35 +00005977
5978<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005979<p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
5980 indicating the return address of the specified call frame, or zero if it
5981 cannot be identified. The value returned by this intrinsic is likely to be
5982 incorrect or 0 for arguments other than zero, so it should only be used for
5983 debugging purposes.</p>
Chris Lattner10610642004-02-14 04:08:35 +00005984
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005985<p>Note that calling this intrinsic does not prevent function inlining or other
5986 aggressive transformations, so the value returned may not be that of the
5987 obvious source-language caller.</p>
Chris Lattner10610642004-02-14 04:08:35 +00005988
Chris Lattner10610642004-02-14 04:08:35 +00005989</div>
5990
Chris Lattner10610642004-02-14 04:08:35 +00005991<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005992<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00005993 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005994</h4>
Chris Lattner10610642004-02-14 04:08:35 +00005995
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005996<div>
Chris Lattner10610642004-02-14 04:08:35 +00005997
5998<h5>Syntax:</h5>
5999<pre>
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006000 declare i8* @llvm.frameaddress(i32 &lt;level&gt;)
Chris Lattner10610642004-02-14 04:08:35 +00006001</pre>
6002
6003<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006004<p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
6005 target-specific frame pointer value for the specified stack frame.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006006
6007<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006008<p>The argument to this intrinsic indicates which function to return the frame
6009 pointer for. Zero indicates the calling function, one indicates its caller,
6010 etc. The argument is <b>required</b> to be a constant integer value.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006011
6012<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006013<p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
6014 indicating the frame address of the specified call frame, or zero if it
6015 cannot be identified. The value returned by this intrinsic is likely to be
6016 incorrect or 0 for arguments other than zero, so it should only be used for
6017 debugging purposes.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006018
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006019<p>Note that calling this intrinsic does not prevent function inlining or other
6020 aggressive transformations, so the value returned may not be that of the
6021 obvious source-language caller.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006022
Chris Lattner10610642004-02-14 04:08:35 +00006023</div>
6024
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00006025<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006026<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006027 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006028</h4>
Chris Lattner57e1f392006-01-13 02:03:13 +00006029
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006030<div>
Chris Lattner57e1f392006-01-13 02:03:13 +00006031
6032<h5>Syntax:</h5>
6033<pre>
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006034 declare i8* @llvm.stacksave()
Chris Lattner57e1f392006-01-13 02:03:13 +00006035</pre>
6036
6037<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006038<p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
6039 of the function stack, for use
6040 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
6041 useful for implementing language features like scoped automatic variable
6042 sized arrays in C99.</p>
Chris Lattner57e1f392006-01-13 02:03:13 +00006043
6044<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006045<p>This intrinsic returns a opaque pointer value that can be passed
6046 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
6047 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
6048 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
6049 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
6050 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
6051 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
Chris Lattner57e1f392006-01-13 02:03:13 +00006052
6053</div>
6054
6055<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006056<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006057 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006058</h4>
Chris Lattner57e1f392006-01-13 02:03:13 +00006059
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006060<div>
Chris Lattner57e1f392006-01-13 02:03:13 +00006061
6062<h5>Syntax:</h5>
6063<pre>
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006064 declare void @llvm.stackrestore(i8* %ptr)
Chris Lattner57e1f392006-01-13 02:03:13 +00006065</pre>
6066
6067<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006068<p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
6069 the function stack to the state it was in when the
6070 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
6071 executed. This is useful for implementing language features like scoped
6072 automatic variable sized arrays in C99.</p>
Chris Lattner57e1f392006-01-13 02:03:13 +00006073
6074<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006075<p>See the description
6076 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
Chris Lattner57e1f392006-01-13 02:03:13 +00006077
6078</div>
6079
Chris Lattner57e1f392006-01-13 02:03:13 +00006080<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006081<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006082 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006083</h4>
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00006084
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006085<div>
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00006086
6087<h5>Syntax:</h5>
6088<pre>
Bruno Cardoso Lopes9a767332011-06-14 04:58:37 +00006089 declare void @llvm.prefetch(i8* &lt;address&gt;, i32 &lt;rw&gt;, i32 &lt;locality&gt;, i32 &lt;cache type&gt;)
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00006090</pre>
6091
6092<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006093<p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
6094 insert a prefetch instruction if supported; otherwise, it is a noop.
6095 Prefetches have no effect on the behavior of the program but can change its
6096 performance characteristics.</p>
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00006097
6098<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006099<p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
6100 specifier determining if the fetch should be for a read (0) or write (1),
6101 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
Bruno Cardoso Lopes9a767332011-06-14 04:58:37 +00006102 locality, to (3) - extremely local keep in cache. The <tt>cache type</tt>
6103 specifies whether the prefetch is performed on the data (1) or instruction (0)
6104 cache. The <tt>rw</tt>, <tt>locality</tt> and <tt>cache type</tt> arguments
6105 must be constant integers.</p>
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00006106
6107<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006108<p>This intrinsic does not modify the behavior of the program. In particular,
6109 prefetches cannot trap and do not produce a value. On targets that support
6110 this intrinsic, the prefetch can provide hints to the processor cache for
6111 better performance.</p>
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00006112
6113</div>
6114
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00006115<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006116<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006117 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006118</h4>
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00006119
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006120<div>
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00006121
6122<h5>Syntax:</h5>
6123<pre>
Chris Lattner1df4f752007-09-21 17:30:40 +00006124 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00006125</pre>
6126
6127<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006128<p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
6129 Counter (PC) in a region of code to simulators and other tools. The method
6130 is target specific, but it is expected that the marker will use exported
6131 symbols to transmit the PC of the marker. The marker makes no guarantees
6132 that it will remain with any specific instruction after optimizations. It is
6133 possible that the presence of a marker will inhibit optimizations. The
6134 intended use is to be inserted after optimizations to allow correlations of
6135 simulation runs.</p>
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00006136
6137<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006138<p><tt>id</tt> is a numerical id identifying the marker.</p>
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00006139
6140<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006141<p>This intrinsic does not modify the behavior of the program. Backends that do
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00006142 not support this intrinsic may ignore it.</p>
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00006143
6144</div>
6145
Andrew Lenharth51b8d542005-11-11 16:47:30 +00006146<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006147<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006148 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006149</h4>
Andrew Lenharth51b8d542005-11-11 16:47:30 +00006150
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006151<div>
Andrew Lenharth51b8d542005-11-11 16:47:30 +00006152
6153<h5>Syntax:</h5>
6154<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00006155 declare i64 @llvm.readcyclecounter()
Andrew Lenharth51b8d542005-11-11 16:47:30 +00006156</pre>
6157
6158<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006159<p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
6160 counter register (or similar low latency, high accuracy clocks) on those
6161 targets that support it. On X86, it should map to RDTSC. On Alpha, it
6162 should map to RPCC. As the backing counters overflow quickly (on the order
6163 of 9 seconds on alpha), this should only be used for small timings.</p>
Andrew Lenharth51b8d542005-11-11 16:47:30 +00006164
6165<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006166<p>When directly supported, reading the cycle counter should not modify any
6167 memory. Implementations are allowed to either return a application specific
6168 value or a system wide value. On backends without support, this is lowered
6169 to a constant 0.</p>
Andrew Lenharth51b8d542005-11-11 16:47:30 +00006170
6171</div>
6172
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006173</div>
6174
Chris Lattner10610642004-02-14 04:08:35 +00006175<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006176<h3>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006177 <a name="int_libc">Standard C Library Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006178</h3>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006179
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006180<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006181
6182<p>LLVM provides intrinsics for a few important standard C library functions.
6183 These intrinsics allow source-language front-ends to pass information about
6184 the alignment of the pointer arguments to the code generator, providing
6185 opportunity for more efficient code generation.</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006186
Chris Lattner33aec9e2004-02-12 17:01:32 +00006187<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006188<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006189 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006190</h4>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006191
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006192<div>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006193
6194<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006195<p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
Mon P Wange88909b2010-04-07 06:35:53 +00006196 integer bit width and for different address spaces. Not all targets support
6197 all bit widths however.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006198
Chris Lattner33aec9e2004-02-12 17:01:32 +00006199<pre>
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006200 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* &lt;dest&gt;, i8* &lt;src&gt;,
Chris Lattner9f636de2010-04-08 00:53:57 +00006201 i32 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006202 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* &lt;dest&gt;, i8* &lt;src&gt;,
Chris Lattner9f636de2010-04-08 00:53:57 +00006203 i64 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Chris Lattner33aec9e2004-02-12 17:01:32 +00006204</pre>
6205
6206<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006207<p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6208 source location to the destination location.</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006209
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006210<p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
Chris Lattner9f636de2010-04-08 00:53:57 +00006211 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6212 and the pointers can be in specified address spaces.</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006213
6214<h5>Arguments:</h5>
Chris Lattner9f636de2010-04-08 00:53:57 +00006215
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006216<p>The first argument is a pointer to the destination, the second is a pointer
6217 to the source. The third argument is an integer argument specifying the
Chris Lattner9f636de2010-04-08 00:53:57 +00006218 number of bytes to copy, the fourth argument is the alignment of the
6219 source and destination locations, and the fifth is a boolean indicating a
6220 volatile access.</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006221
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00006222<p>If the call to this intrinsic has an alignment value that is not 0 or 1,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006223 then the caller guarantees that both the source and destination pointers are
6224 aligned to that boundary.</p>
Chris Lattner3301ced2004-02-12 21:18:15 +00006225
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00006226<p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6227 <tt>llvm.memcpy</tt> call is a <a href="#volatile">volatile operation</a>.
6228 The detailed access behavior is not very cleanly specified and it is unwise
6229 to depend on it.</p>
Chris Lattner9f636de2010-04-08 00:53:57 +00006230
Chris Lattner33aec9e2004-02-12 17:01:32 +00006231<h5>Semantics:</h5>
Chris Lattner9f636de2010-04-08 00:53:57 +00006232
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006233<p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6234 source location to the destination location, which are not allowed to
6235 overlap. It copies "len" bytes of memory over. If the argument is known to
6236 be aligned to some boundary, this can be specified as the fourth argument,
6237 otherwise it should be set to 0 or 1.</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006238
Chris Lattner33aec9e2004-02-12 17:01:32 +00006239</div>
6240
Chris Lattner0eb51b42004-02-12 18:10:10 +00006241<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006242<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006243 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006244</h4>
Chris Lattner0eb51b42004-02-12 18:10:10 +00006245
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006246<div>
Chris Lattner0eb51b42004-02-12 18:10:10 +00006247
6248<h5>Syntax:</h5>
Chris Lattner824b9582008-11-21 16:42:48 +00006249<p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
Mon P Wange88909b2010-04-07 06:35:53 +00006250 width and for different address space. Not all targets support all bit
6251 widths however.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006252
Chris Lattner0eb51b42004-02-12 18:10:10 +00006253<pre>
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006254 declare void @llvm.memmove.p0i8.p0i8.i32(i8* &lt;dest&gt;, i8* &lt;src&gt;,
Chris Lattner9f636de2010-04-08 00:53:57 +00006255 i32 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006256 declare void @llvm.memmove.p0i8.p0i8.i64(i8* &lt;dest&gt;, i8* &lt;src&gt;,
Chris Lattner9f636de2010-04-08 00:53:57 +00006257 i64 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Chris Lattner0eb51b42004-02-12 18:10:10 +00006258</pre>
6259
6260<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006261<p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
6262 source location to the destination location. It is similar to the
6263 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
6264 overlap.</p>
Chris Lattner0eb51b42004-02-12 18:10:10 +00006265
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006266<p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
Chris Lattner9f636de2010-04-08 00:53:57 +00006267 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6268 and the pointers can be in specified address spaces.</p>
Chris Lattner0eb51b42004-02-12 18:10:10 +00006269
6270<h5>Arguments:</h5>
Chris Lattner9f636de2010-04-08 00:53:57 +00006271
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006272<p>The first argument is a pointer to the destination, the second is a pointer
6273 to the source. The third argument is an integer argument specifying the
Chris Lattner9f636de2010-04-08 00:53:57 +00006274 number of bytes to copy, the fourth argument is the alignment of the
6275 source and destination locations, and the fifth is a boolean indicating a
6276 volatile access.</p>
Chris Lattner0eb51b42004-02-12 18:10:10 +00006277
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00006278<p>If the call to this intrinsic has an alignment value that is not 0 or 1,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006279 then the caller guarantees that the source and destination pointers are
6280 aligned to that boundary.</p>
Chris Lattner3301ced2004-02-12 21:18:15 +00006281
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00006282<p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6283 <tt>llvm.memmove</tt> call is a <a href="#volatile">volatile operation</a>.
6284 The detailed access behavior is not very cleanly specified and it is unwise
6285 to depend on it.</p>
Chris Lattner9f636de2010-04-08 00:53:57 +00006286
Chris Lattner0eb51b42004-02-12 18:10:10 +00006287<h5>Semantics:</h5>
Chris Lattner9f636de2010-04-08 00:53:57 +00006288
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006289<p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
6290 source location to the destination location, which may overlap. It copies
6291 "len" bytes of memory over. If the argument is known to be aligned to some
6292 boundary, this can be specified as the fourth argument, otherwise it should
6293 be set to 0 or 1.</p>
Chris Lattner0eb51b42004-02-12 18:10:10 +00006294
Chris Lattner0eb51b42004-02-12 18:10:10 +00006295</div>
6296
Chris Lattner10610642004-02-14 04:08:35 +00006297<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006298<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006299 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006300</h4>
Chris Lattner10610642004-02-14 04:08:35 +00006301
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006302<div>
Chris Lattner10610642004-02-14 04:08:35 +00006303
6304<h5>Syntax:</h5>
Chris Lattner824b9582008-11-21 16:42:48 +00006305<p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
John Criswellcdcbbfc2010-07-30 16:30:28 +00006306 width and for different address spaces. However, not all targets support all
6307 bit widths.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006308
Chris Lattner10610642004-02-14 04:08:35 +00006309<pre>
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006310 declare void @llvm.memset.p0i8.i32(i8* &lt;dest&gt;, i8 &lt;val&gt;,
Chris Lattnerff35c3f2010-04-08 00:54:34 +00006311 i32 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006312 declare void @llvm.memset.p0i8.i64(i8* &lt;dest&gt;, i8 &lt;val&gt;,
Chris Lattnerff35c3f2010-04-08 00:54:34 +00006313 i64 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Chris Lattner10610642004-02-14 04:08:35 +00006314</pre>
6315
6316<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006317<p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
6318 particular byte value.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006319
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006320<p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
John Criswellcdcbbfc2010-07-30 16:30:28 +00006321 intrinsic does not return a value and takes extra alignment/volatile
6322 arguments. Also, the destination can be in an arbitrary address space.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006323
6324<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006325<p>The first argument is a pointer to the destination to fill, the second is the
John Criswellcdcbbfc2010-07-30 16:30:28 +00006326 byte value with which to fill it, the third argument is an integer argument
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006327 specifying the number of bytes to fill, and the fourth argument is the known
John Criswellcdcbbfc2010-07-30 16:30:28 +00006328 alignment of the destination location.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006329
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00006330<p>If the call to this intrinsic has an alignment value that is not 0 or 1,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006331 then the caller guarantees that the destination pointer is aligned to that
6332 boundary.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006333
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00006334<p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6335 <tt>llvm.memset</tt> call is a <a href="#volatile">volatile operation</a>.
6336 The detailed access behavior is not very cleanly specified and it is unwise
6337 to depend on it.</p>
Chris Lattner9f636de2010-04-08 00:53:57 +00006338
Chris Lattner10610642004-02-14 04:08:35 +00006339<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006340<p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
6341 at the destination location. If the argument is known to be aligned to some
6342 boundary, this can be specified as the fourth argument, otherwise it should
6343 be set to 0 or 1.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006344
Chris Lattner10610642004-02-14 04:08:35 +00006345</div>
6346
Chris Lattner32006282004-06-11 02:28:03 +00006347<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006348<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006349 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006350</h4>
Chris Lattnera4d74142005-07-21 01:29:16 +00006351
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006352<div>
Chris Lattnera4d74142005-07-21 01:29:16 +00006353
6354<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006355<p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
6356 floating point or vector of floating point type. Not all targets support all
6357 types however.</p>
6358
Chris Lattnera4d74142005-07-21 01:29:16 +00006359<pre>
Dale Johannesen408f9c12007-10-02 17:47:38 +00006360 declare float @llvm.sqrt.f32(float %Val)
6361 declare double @llvm.sqrt.f64(double %Val)
6362 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
6363 declare fp128 @llvm.sqrt.f128(fp128 %Val)
6364 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Chris Lattnera4d74142005-07-21 01:29:16 +00006365</pre>
6366
6367<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006368<p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
6369 returning the same value as the libm '<tt>sqrt</tt>' functions would.
6370 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
6371 behavior for negative numbers other than -0.0 (which allows for better
6372 optimization, because there is no need to worry about errno being
6373 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
Chris Lattnera4d74142005-07-21 01:29:16 +00006374
6375<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006376<p>The argument and return value are floating point numbers of the same
6377 type.</p>
Chris Lattnera4d74142005-07-21 01:29:16 +00006378
6379<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006380<p>This function returns the sqrt of the specified operand if it is a
6381 nonnegative floating point number.</p>
Chris Lattnera4d74142005-07-21 01:29:16 +00006382
Chris Lattnera4d74142005-07-21 01:29:16 +00006383</div>
6384
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006385<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006386<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006387 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006388</h4>
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006389
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006390<div>
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006391
6392<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006393<p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
6394 floating point or vector of floating point type. Not all targets support all
6395 types however.</p>
6396
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006397<pre>
Dale Johannesen408f9c12007-10-02 17:47:38 +00006398 declare float @llvm.powi.f32(float %Val, i32 %power)
6399 declare double @llvm.powi.f64(double %Val, i32 %power)
6400 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
6401 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
6402 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006403</pre>
6404
6405<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006406<p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
6407 specified (positive or negative) power. The order of evaluation of
6408 multiplications is not defined. When a vector of floating point type is
6409 used, the second argument remains a scalar integer value.</p>
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006410
6411<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006412<p>The second argument is an integer power, and the first is a value to raise to
6413 that power.</p>
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006414
6415<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006416<p>This function returns the first value raised to the second power with an
6417 unspecified sequence of rounding operations.</p>
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006418
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006419</div>
6420
Dan Gohman91c284c2007-10-15 20:30:11 +00006421<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006422<h4>
Dan Gohman91c284c2007-10-15 20:30:11 +00006423 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006424</h4>
Dan Gohman91c284c2007-10-15 20:30:11 +00006425
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006426<div>
Dan Gohman91c284c2007-10-15 20:30:11 +00006427
6428<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006429<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
6430 floating point or vector of floating point type. Not all targets support all
6431 types however.</p>
6432
Dan Gohman91c284c2007-10-15 20:30:11 +00006433<pre>
6434 declare float @llvm.sin.f32(float %Val)
6435 declare double @llvm.sin.f64(double %Val)
6436 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
6437 declare fp128 @llvm.sin.f128(fp128 %Val)
6438 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
6439</pre>
6440
6441<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006442<p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006443
6444<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006445<p>The argument and return value are floating point numbers of the same
6446 type.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006447
6448<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006449<p>This function returns the sine of the specified operand, returning the same
6450 values as the libm <tt>sin</tt> functions would, and handles error conditions
6451 in the same way.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006452
Dan Gohman91c284c2007-10-15 20:30:11 +00006453</div>
6454
6455<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006456<h4>
Dan Gohman91c284c2007-10-15 20:30:11 +00006457 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006458</h4>
Dan Gohman91c284c2007-10-15 20:30:11 +00006459
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006460<div>
Dan Gohman91c284c2007-10-15 20:30:11 +00006461
6462<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006463<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
6464 floating point or vector of floating point type. Not all targets support all
6465 types however.</p>
6466
Dan Gohman91c284c2007-10-15 20:30:11 +00006467<pre>
6468 declare float @llvm.cos.f32(float %Val)
6469 declare double @llvm.cos.f64(double %Val)
6470 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
6471 declare fp128 @llvm.cos.f128(fp128 %Val)
6472 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
6473</pre>
6474
6475<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006476<p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006477
6478<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006479<p>The argument and return value are floating point numbers of the same
6480 type.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006481
6482<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006483<p>This function returns the cosine of the specified operand, returning the same
6484 values as the libm <tt>cos</tt> functions would, and handles error conditions
6485 in the same way.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006486
Dan Gohman91c284c2007-10-15 20:30:11 +00006487</div>
6488
6489<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006490<h4>
Dan Gohman91c284c2007-10-15 20:30:11 +00006491 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006492</h4>
Dan Gohman91c284c2007-10-15 20:30:11 +00006493
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006494<div>
Dan Gohman91c284c2007-10-15 20:30:11 +00006495
6496<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006497<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
6498 floating point or vector of floating point type. Not all targets support all
6499 types however.</p>
6500
Dan Gohman91c284c2007-10-15 20:30:11 +00006501<pre>
6502 declare float @llvm.pow.f32(float %Val, float %Power)
6503 declare double @llvm.pow.f64(double %Val, double %Power)
6504 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
6505 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
6506 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
6507</pre>
6508
6509<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006510<p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
6511 specified (positive or negative) power.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006512
6513<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006514<p>The second argument is a floating point power, and the first is a value to
6515 raise to that power.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006516
6517<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006518<p>This function returns the first value raised to the second power, returning
6519 the same values as the libm <tt>pow</tt> functions would, and handles error
6520 conditions in the same way.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006521
Dan Gohman91c284c2007-10-15 20:30:11 +00006522</div>
6523
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006524</div>
6525
Dan Gohman4e9011c2011-05-23 21:13:03 +00006526<!-- _______________________________________________________________________ -->
6527<h4>
6528 <a name="int_exp">'<tt>llvm.exp.*</tt>' Intrinsic</a>
6529</h4>
6530
6531<div>
6532
6533<h5>Syntax:</h5>
6534<p>This is an overloaded intrinsic. You can use <tt>llvm.exp</tt> on any
6535 floating point or vector of floating point type. Not all targets support all
6536 types however.</p>
6537
6538<pre>
6539 declare float @llvm.exp.f32(float %Val)
6540 declare double @llvm.exp.f64(double %Val)
6541 declare x86_fp80 @llvm.exp.f80(x86_fp80 %Val)
6542 declare fp128 @llvm.exp.f128(fp128 %Val)
6543 declare ppc_fp128 @llvm.exp.ppcf128(ppc_fp128 %Val)
6544</pre>
6545
6546<h5>Overview:</h5>
6547<p>The '<tt>llvm.exp.*</tt>' intrinsics perform the exp function.</p>
6548
6549<h5>Arguments:</h5>
6550<p>The argument and return value are floating point numbers of the same
6551 type.</p>
6552
6553<h5>Semantics:</h5>
6554<p>This function returns the same values as the libm <tt>exp</tt> functions
6555 would, and handles error conditions in the same way.</p>
6556
6557</div>
6558
6559<!-- _______________________________________________________________________ -->
6560<h4>
6561 <a name="int_log">'<tt>llvm.log.*</tt>' Intrinsic</a>
6562</h4>
6563
6564<div>
6565
6566<h5>Syntax:</h5>
6567<p>This is an overloaded intrinsic. You can use <tt>llvm.log</tt> on any
6568 floating point or vector of floating point type. Not all targets support all
6569 types however.</p>
6570
6571<pre>
6572 declare float @llvm.log.f32(float %Val)
6573 declare double @llvm.log.f64(double %Val)
6574 declare x86_fp80 @llvm.log.f80(x86_fp80 %Val)
6575 declare fp128 @llvm.log.f128(fp128 %Val)
6576 declare ppc_fp128 @llvm.log.ppcf128(ppc_fp128 %Val)
6577</pre>
6578
6579<h5>Overview:</h5>
6580<p>The '<tt>llvm.log.*</tt>' intrinsics perform the log function.</p>
6581
6582<h5>Arguments:</h5>
6583<p>The argument and return value are floating point numbers of the same
6584 type.</p>
6585
6586<h5>Semantics:</h5>
6587<p>This function returns the same values as the libm <tt>log</tt> functions
6588 would, and handles error conditions in the same way.</p>
6589
Cameron Zwarich33390842011-07-08 21:39:21 +00006590<h4>
6591 <a name="int_fma">'<tt>llvm.fma.*</tt>' Intrinsic</a>
6592</h4>
6593
6594<div>
6595
6596<h5>Syntax:</h5>
6597<p>This is an overloaded intrinsic. You can use <tt>llvm.fma</tt> on any
6598 floating point or vector of floating point type. Not all targets support all
6599 types however.</p>
6600
6601<pre>
6602 declare float @llvm.fma.f32(float %a, float %b, float %c)
6603 declare double @llvm.fma.f64(double %a, double %b, double %c)
6604 declare x86_fp80 @llvm.fma.f80(x86_fp80 %a, x86_fp80 %b, x86_fp80 %c)
6605 declare fp128 @llvm.fma.f128(fp128 %a, fp128 %b, fp128 %c)
6606 declare ppc_fp128 @llvm.fma.ppcf128(ppc_fp128 %a, ppc_fp128 %b, ppc_fp128 %c)
6607</pre>
6608
6609<h5>Overview:</h5>
Cameron Zwarichabc43e62011-07-08 22:13:55 +00006610<p>The '<tt>llvm.fma.*</tt>' intrinsics perform the fused multiply-add
Cameron Zwarich33390842011-07-08 21:39:21 +00006611 operation.</p>
6612
6613<h5>Arguments:</h5>
6614<p>The argument and return value are floating point numbers of the same
6615 type.</p>
6616
6617<h5>Semantics:</h5>
6618<p>This function returns the same values as the libm <tt>fma</tt> functions
6619 would.</p>
6620
Dan Gohman4e9011c2011-05-23 21:13:03 +00006621</div>
6622
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006623<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006624<h3>
Nate Begeman7e36c472006-01-13 23:26:38 +00006625 <a name="int_manip">Bit Manipulation Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006626</h3>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006627
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006628<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006629
6630<p>LLVM provides intrinsics for a few important bit manipulation operations.
6631 These allow efficient code generation for some algorithms.</p>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006632
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006633<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006634<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006635 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006636</h4>
Nate Begeman7e36c472006-01-13 23:26:38 +00006637
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006638<div>
Nate Begeman7e36c472006-01-13 23:26:38 +00006639
6640<h5>Syntax:</h5>
Reid Spencer409e28f2007-04-01 08:04:23 +00006641<p>This is an overloaded intrinsic function. You can use bswap on any integer
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006642 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
6643
Nate Begeman7e36c472006-01-13 23:26:38 +00006644<pre>
Chandler Carruth69940402007-08-04 01:51:18 +00006645 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
6646 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
6647 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Nate Begeman7e36c472006-01-13 23:26:38 +00006648</pre>
6649
6650<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006651<p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
6652 values with an even number of bytes (positive multiple of 16 bits). These
6653 are useful for performing operations on data that is not in the target's
6654 native byte order.</p>
Nate Begeman7e36c472006-01-13 23:26:38 +00006655
6656<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006657<p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
6658 and low byte of the input i16 swapped. Similarly,
6659 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
6660 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
6661 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
6662 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
6663 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
6664 more, respectively).</p>
Nate Begeman7e36c472006-01-13 23:26:38 +00006665
6666</div>
6667
6668<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006669<h4>
Reid Spencer0b118202006-01-16 21:12:35 +00006670 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006671</h4>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006672
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006673<div>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006674
6675<h5>Syntax:</h5>
Reid Spencer409e28f2007-04-01 08:04:23 +00006676<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
Owen Andersonf1ac4652011-07-01 21:52:38 +00006677 width, or on any vector with integer elements. Not all targets support all
6678 bit widths or vector types, however.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006679
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006680<pre>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006681 declare i8 @llvm.ctpop.i8(i8 &lt;src&gt;)
Chandler Carruth69940402007-08-04 01:51:18 +00006682 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Anton Korobeynikovec43a062007-03-22 00:02:17 +00006683 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carruth69940402007-08-04 01:51:18 +00006684 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
6685 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Owen Andersonf1ac4652011-07-01 21:52:38 +00006686 declare &lt;2 x i32&gt; @llvm.ctpop.v2i32(&lt;2 x i32&gt; &lt;src&gt;)
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006687</pre>
6688
6689<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006690<p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
6691 in a value.</p>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006692
6693<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006694<p>The only argument is the value to be counted. The argument may be of any
Owen Andersonf1ac4652011-07-01 21:52:38 +00006695 integer type, or a vector with integer elements.
6696 The return type must match the argument type.</p>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006697
6698<h5>Semantics:</h5>
Owen Andersonf1ac4652011-07-01 21:52:38 +00006699<p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable, or within each
6700 element of a vector.</p>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006701
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006702</div>
6703
6704<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006705<h4>
Chris Lattner8a886be2006-01-16 22:34:14 +00006706 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006707</h4>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006708
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006709<div>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006710
6711<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006712<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
Owen Andersonf1ac4652011-07-01 21:52:38 +00006713 integer bit width, or any vector whose elements are integers. Not all
6714 targets support all bit widths or vector types, however.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006715
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006716<pre>
Chandler Carruth69940402007-08-04 01:51:18 +00006717 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
6718 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Anton Korobeynikovec43a062007-03-22 00:02:17 +00006719 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carruth69940402007-08-04 01:51:18 +00006720 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
6721 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Owen Andersonf1ac4652011-07-01 21:52:38 +00006722 declare &lt;2 x i32&gt; @llvm.ctlz.v2i32(&lt;2 x i32&gt; &lt;src;gt)
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006723</pre>
6724
6725<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006726<p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
6727 leading zeros in a variable.</p>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006728
6729<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006730<p>The only argument is the value to be counted. The argument may be of any
Owen Andersonf1ac4652011-07-01 21:52:38 +00006731 integer type, or any vector type with integer element type.
6732 The return type must match the argument type.</p>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006733
6734<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006735<p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
Owen Andersonf1ac4652011-07-01 21:52:38 +00006736 zeros in a variable, or within each element of the vector if the operation
6737 is of vector type. If the src == 0 then the result is the size in bits of
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006738 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006739
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006740</div>
Chris Lattner32006282004-06-11 02:28:03 +00006741
Chris Lattnereff29ab2005-05-15 19:39:26 +00006742<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006743<h4>
Chris Lattner8a886be2006-01-16 22:34:14 +00006744 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006745</h4>
Chris Lattnereff29ab2005-05-15 19:39:26 +00006746
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006747<div>
Chris Lattnereff29ab2005-05-15 19:39:26 +00006748
6749<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006750<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
Owen Andersonf1ac4652011-07-01 21:52:38 +00006751 integer bit width, or any vector of integer elements. Not all targets
6752 support all bit widths or vector types, however.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006753
Chris Lattnereff29ab2005-05-15 19:39:26 +00006754<pre>
Chandler Carruth69940402007-08-04 01:51:18 +00006755 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
6756 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Anton Korobeynikovec43a062007-03-22 00:02:17 +00006757 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carruth69940402007-08-04 01:51:18 +00006758 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
6759 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Owen Andersonf1ac4652011-07-01 21:52:38 +00006760 declase &lt;2 x i32&gt; @llvm.cttz.v2i32(&lt;2 x i32&gt; &lt;src&gt;)
Chris Lattnereff29ab2005-05-15 19:39:26 +00006761</pre>
6762
6763<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006764<p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
6765 trailing zeros.</p>
Chris Lattnereff29ab2005-05-15 19:39:26 +00006766
6767<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006768<p>The only argument is the value to be counted. The argument may be of any
Owen Andersonf1ac4652011-07-01 21:52:38 +00006769 integer type, or a vectory with integer element type.. The return type
6770 must match the argument type.</p>
Chris Lattnereff29ab2005-05-15 19:39:26 +00006771
6772<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006773<p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
Owen Andersonf1ac4652011-07-01 21:52:38 +00006774 zeros in a variable, or within each element of a vector.
6775 If the src == 0 then the result is the size in bits of
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006776 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
Chris Lattnereff29ab2005-05-15 19:39:26 +00006777
Chris Lattnereff29ab2005-05-15 19:39:26 +00006778</div>
6779
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006780</div>
6781
Bill Wendlingda01af72009-02-08 04:04:40 +00006782<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006783<h3>
Bill Wendlingda01af72009-02-08 04:04:40 +00006784 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006785</h3>
Bill Wendlingda01af72009-02-08 04:04:40 +00006786
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006787<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006788
6789<p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
Bill Wendlingda01af72009-02-08 04:04:40 +00006790
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006791<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006792<h4>
6793 <a name="int_sadd_overflow">
6794 '<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics
6795 </a>
6796</h4>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006797
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006798<div>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006799
6800<h5>Syntax:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006801<p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006802 on any integer bit width.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006803
6804<pre>
6805 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
6806 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6807 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6808</pre>
6809
6810<h5>Overview:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006811<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006812 a signed addition of the two arguments, and indicate whether an overflow
6813 occurred during the signed summation.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006814
6815<h5>Arguments:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006816<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006817 be of integer types of any bit width, but they must have the same bit
6818 width. The second element of the result structure must be of
6819 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6820 undergo signed addition.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006821
6822<h5>Semantics:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006823<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006824 a signed addition of the two variables. They return a structure &mdash; the
6825 first element of which is the signed summation, and the second element of
6826 which is a bit specifying if the signed summation resulted in an
6827 overflow.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006828
6829<h5>Examples:</h5>
6830<pre>
6831 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6832 %sum = extractvalue {i32, i1} %res, 0
6833 %obit = extractvalue {i32, i1} %res, 1
6834 br i1 %obit, label %overflow, label %normal
6835</pre>
6836
6837</div>
6838
6839<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006840<h4>
6841 <a name="int_uadd_overflow">
6842 '<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics
6843 </a>
6844</h4>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006845
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006846<div>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006847
6848<h5>Syntax:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006849<p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006850 on any integer bit width.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006851
6852<pre>
6853 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6854 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6855 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6856</pre>
6857
6858<h5>Overview:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006859<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006860 an unsigned addition of the two arguments, and indicate whether a carry
6861 occurred during the unsigned summation.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006862
6863<h5>Arguments:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006864<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006865 be of integer types of any bit width, but they must have the same bit
6866 width. The second element of the result structure must be of
6867 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6868 undergo unsigned addition.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006869
6870<h5>Semantics:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006871<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006872 an unsigned addition of the two arguments. They return a structure &mdash;
6873 the first element of which is the sum, and the second element of which is a
6874 bit specifying if the unsigned summation resulted in a carry.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006875
6876<h5>Examples:</h5>
6877<pre>
6878 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6879 %sum = extractvalue {i32, i1} %res, 0
6880 %obit = extractvalue {i32, i1} %res, 1
6881 br i1 %obit, label %carry, label %normal
6882</pre>
6883
6884</div>
6885
6886<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006887<h4>
6888 <a name="int_ssub_overflow">
6889 '<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics
6890 </a>
6891</h4>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006892
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006893<div>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006894
6895<h5>Syntax:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006896<p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006897 on any integer bit width.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006898
6899<pre>
6900 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6901 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6902 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6903</pre>
6904
6905<h5>Overview:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006906<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006907 a signed subtraction of the two arguments, and indicate whether an overflow
6908 occurred during the signed subtraction.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006909
6910<h5>Arguments:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006911<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006912 be of integer types of any bit width, but they must have the same bit
6913 width. The second element of the result structure must be of
6914 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6915 undergo signed subtraction.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006916
6917<h5>Semantics:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006918<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006919 a signed subtraction of the two arguments. They return a structure &mdash;
6920 the first element of which is the subtraction, and the second element of
6921 which is a bit specifying if the signed subtraction resulted in an
6922 overflow.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006923
6924<h5>Examples:</h5>
6925<pre>
6926 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6927 %sum = extractvalue {i32, i1} %res, 0
6928 %obit = extractvalue {i32, i1} %res, 1
6929 br i1 %obit, label %overflow, label %normal
6930</pre>
6931
6932</div>
6933
6934<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006935<h4>
6936 <a name="int_usub_overflow">
6937 '<tt>llvm.usub.with.overflow.*</tt>' Intrinsics
6938 </a>
6939</h4>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006940
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006941<div>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006942
6943<h5>Syntax:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006944<p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006945 on any integer bit width.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006946
6947<pre>
6948 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6949 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6950 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6951</pre>
6952
6953<h5>Overview:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006954<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006955 an unsigned subtraction of the two arguments, and indicate whether an
6956 overflow occurred during the unsigned subtraction.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006957
6958<h5>Arguments:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006959<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006960 be of integer types of any bit width, but they must have the same bit
6961 width. The second element of the result structure must be of
6962 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6963 undergo unsigned subtraction.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006964
6965<h5>Semantics:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006966<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006967 an unsigned subtraction of the two arguments. They return a structure &mdash;
6968 the first element of which is the subtraction, and the second element of
6969 which is a bit specifying if the unsigned subtraction resulted in an
6970 overflow.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006971
6972<h5>Examples:</h5>
6973<pre>
6974 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6975 %sum = extractvalue {i32, i1} %res, 0
6976 %obit = extractvalue {i32, i1} %res, 1
6977 br i1 %obit, label %overflow, label %normal
6978</pre>
6979
6980</div>
6981
6982<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006983<h4>
6984 <a name="int_smul_overflow">
6985 '<tt>llvm.smul.with.overflow.*</tt>' Intrinsics
6986 </a>
6987</h4>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006988
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006989<div>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006990
6991<h5>Syntax:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006992<p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006993 on any integer bit width.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006994
6995<pre>
6996 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6997 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6998 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6999</pre>
7000
7001<h5>Overview:</h5>
7002
7003<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007004 a signed multiplication of the two arguments, and indicate whether an
7005 overflow occurred during the signed multiplication.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007006
7007<h5>Arguments:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007008<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007009 be of integer types of any bit width, but they must have the same bit
7010 width. The second element of the result structure must be of
7011 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7012 undergo signed multiplication.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007013
7014<h5>Semantics:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007015<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007016 a signed multiplication of the two arguments. They return a structure &mdash;
7017 the first element of which is the multiplication, and the second element of
7018 which is a bit specifying if the signed multiplication resulted in an
7019 overflow.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007020
7021<h5>Examples:</h5>
7022<pre>
7023 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
7024 %sum = extractvalue {i32, i1} %res, 0
7025 %obit = extractvalue {i32, i1} %res, 1
7026 br i1 %obit, label %overflow, label %normal
7027</pre>
7028
Reid Spencerf86037f2007-04-11 23:23:49 +00007029</div>
7030
Bill Wendling41b485c2009-02-08 23:00:09 +00007031<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007032<h4>
7033 <a name="int_umul_overflow">
7034 '<tt>llvm.umul.with.overflow.*</tt>' Intrinsics
7035 </a>
7036</h4>
Bill Wendling41b485c2009-02-08 23:00:09 +00007037
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007038<div>
Bill Wendling41b485c2009-02-08 23:00:09 +00007039
7040<h5>Syntax:</h5>
Bill Wendling41b485c2009-02-08 23:00:09 +00007041<p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007042 on any integer bit width.</p>
Bill Wendling41b485c2009-02-08 23:00:09 +00007043
7044<pre>
7045 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
7046 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
7047 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
7048</pre>
7049
7050<h5>Overview:</h5>
Bill Wendling41b485c2009-02-08 23:00:09 +00007051<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007052 a unsigned multiplication of the two arguments, and indicate whether an
7053 overflow occurred during the unsigned multiplication.</p>
Bill Wendling41b485c2009-02-08 23:00:09 +00007054
7055<h5>Arguments:</h5>
Bill Wendling41b485c2009-02-08 23:00:09 +00007056<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007057 be of integer types of any bit width, but they must have the same bit
7058 width. The second element of the result structure must be of
7059 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7060 undergo unsigned multiplication.</p>
Bill Wendling41b485c2009-02-08 23:00:09 +00007061
7062<h5>Semantics:</h5>
Bill Wendling41b485c2009-02-08 23:00:09 +00007063<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007064 an unsigned multiplication of the two arguments. They return a structure
7065 &mdash; the first element of which is the multiplication, and the second
7066 element of which is a bit specifying if the unsigned multiplication resulted
7067 in an overflow.</p>
Bill Wendling41b485c2009-02-08 23:00:09 +00007068
7069<h5>Examples:</h5>
7070<pre>
7071 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
7072 %sum = extractvalue {i32, i1} %res, 0
7073 %obit = extractvalue {i32, i1} %res, 1
7074 br i1 %obit, label %overflow, label %normal
7075</pre>
7076
7077</div>
7078
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007079</div>
7080
Chris Lattner8ff75902004-01-06 05:31:32 +00007081<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007082<h3>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007083 <a name="int_fp16">Half Precision Floating Point Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007084</h3>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007085
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007086<div>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007087
Chris Lattner0cec9c82010-03-15 04:12:21 +00007088<p>Half precision floating point is a storage-only format. This means that it is
7089 a dense encoding (in memory) but does not support computation in the
7090 format.</p>
Chris Lattner82c3dc62010-03-14 23:03:31 +00007091
Chris Lattner0cec9c82010-03-15 04:12:21 +00007092<p>This means that code must first load the half-precision floating point
Chris Lattner82c3dc62010-03-14 23:03:31 +00007093 value as an i16, then convert it to float with <a
7094 href="#int_convert_from_fp16"><tt>llvm.convert.from.fp16</tt></a>.
7095 Computation can then be performed on the float value (including extending to
Chris Lattner0cec9c82010-03-15 04:12:21 +00007096 double etc). To store the value back to memory, it is first converted to
7097 float if needed, then converted to i16 with
Chris Lattner82c3dc62010-03-14 23:03:31 +00007098 <a href="#int_convert_to_fp16"><tt>llvm.convert.to.fp16</tt></a>, then
7099 storing as an i16 value.</p>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007100
7101<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007102<h4>
7103 <a name="int_convert_to_fp16">
7104 '<tt>llvm.convert.to.fp16</tt>' Intrinsic
7105 </a>
7106</h4>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007107
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007108<div>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007109
7110<h5>Syntax:</h5>
7111<pre>
7112 declare i16 @llvm.convert.to.fp16(f32 %a)
7113</pre>
7114
7115<h5>Overview:</h5>
7116<p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
7117 a conversion from single precision floating point format to half precision
7118 floating point format.</p>
7119
7120<h5>Arguments:</h5>
7121<p>The intrinsic function contains single argument - the value to be
7122 converted.</p>
7123
7124<h5>Semantics:</h5>
7125<p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
7126 a conversion from single precision floating point format to half precision
Chris Lattner0cec9c82010-03-15 04:12:21 +00007127 floating point format. The return value is an <tt>i16</tt> which
Chris Lattner82c3dc62010-03-14 23:03:31 +00007128 contains the converted number.</p>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007129
7130<h5>Examples:</h5>
7131<pre>
7132 %res = call i16 @llvm.convert.to.fp16(f32 %a)
7133 store i16 %res, i16* @x, align 2
7134</pre>
7135
7136</div>
7137
7138<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007139<h4>
7140 <a name="int_convert_from_fp16">
7141 '<tt>llvm.convert.from.fp16</tt>' Intrinsic
7142 </a>
7143</h4>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007144
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007145<div>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007146
7147<h5>Syntax:</h5>
7148<pre>
7149 declare f32 @llvm.convert.from.fp16(i16 %a)
7150</pre>
7151
7152<h5>Overview:</h5>
7153<p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs
7154 a conversion from half precision floating point format to single precision
7155 floating point format.</p>
7156
7157<h5>Arguments:</h5>
7158<p>The intrinsic function contains single argument - the value to be
7159 converted.</p>
7160
7161<h5>Semantics:</h5>
7162<p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs a
Chris Lattner0cec9c82010-03-15 04:12:21 +00007163 conversion from half single precision floating point format to single
Chris Lattner82c3dc62010-03-14 23:03:31 +00007164 precision floating point format. The input half-float value is represented by
7165 an <tt>i16</tt> value.</p>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007166
7167<h5>Examples:</h5>
7168<pre>
7169 %a = load i16* @x, align 2
7170 %res = call f32 @llvm.convert.from.fp16(i16 %a)
7171</pre>
7172
7173</div>
7174
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007175</div>
7176
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007177<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007178<h3>
Chris Lattner8ff75902004-01-06 05:31:32 +00007179 <a name="int_debugger">Debugger Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007180</h3>
Chris Lattner8ff75902004-01-06 05:31:32 +00007181
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007182<div>
Chris Lattner8ff75902004-01-06 05:31:32 +00007183
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007184<p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
7185 prefix), are described in
7186 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
7187 Level Debugging</a> document.</p>
7188
7189</div>
Chris Lattner8ff75902004-01-06 05:31:32 +00007190
Jim Laskeydd4ef1b2007-03-14 19:31:19 +00007191<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007192<h3>
Jim Laskeydd4ef1b2007-03-14 19:31:19 +00007193 <a name="int_eh">Exception Handling Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007194</h3>
Jim Laskeydd4ef1b2007-03-14 19:31:19 +00007195
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007196<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007197
7198<p>The LLVM exception handling intrinsics (which all start with
7199 <tt>llvm.eh.</tt> prefix), are described in
7200 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
7201 Handling</a> document.</p>
7202
Jim Laskeydd4ef1b2007-03-14 19:31:19 +00007203</div>
7204
Tanya Lattner6d806e92007-06-15 20:50:54 +00007205<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007206<h3>
Duncan Sandsf7331b32007-09-11 14:10:23 +00007207 <a name="int_trampoline">Trampoline Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007208</h3>
Duncan Sands36397f52007-07-27 12:58:54 +00007209
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007210<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007211
7212<p>This intrinsic makes it possible to excise one parameter, marked with
Dan Gohmanff235352010-07-02 23:18:08 +00007213 the <a href="#nest"><tt>nest</tt></a> attribute, from a function.
7214 The result is a callable
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007215 function pointer lacking the nest parameter - the caller does not need to
7216 provide a value for it. Instead, the value to use is stored in advance in a
7217 "trampoline", a block of memory usually allocated on the stack, which also
7218 contains code to splice the nest value into the argument list. This is used
7219 to implement the GCC nested function address extension.</p>
7220
7221<p>For example, if the function is
7222 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
7223 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
7224 follows:</p>
7225
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00007226<pre class="doc_code">
Duncan Sandsf7331b32007-09-11 14:10:23 +00007227 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
7228 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007229 %p = call i8* @llvm.init.trampoline(i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval)
Duncan Sandsf7331b32007-09-11 14:10:23 +00007230 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands36397f52007-07-27 12:58:54 +00007231</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007232
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007233<p>The call <tt>%val = call i32 %fp(i32 %x, i32 %y)</tt> is then equivalent
7234 to <tt>%val = call i32 %f(i8* %nval, i32 %x, i32 %y)</tt>.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007235
Duncan Sands36397f52007-07-27 12:58:54 +00007236<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007237<h4>
7238 <a name="int_it">
7239 '<tt>llvm.init.trampoline</tt>' Intrinsic
7240 </a>
7241</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007242
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007243<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007244
Duncan Sands36397f52007-07-27 12:58:54 +00007245<h5>Syntax:</h5>
7246<pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007247 declare i8* @llvm.init.trampoline(i8* &lt;tramp&gt;, i8* &lt;func&gt;, i8* &lt;nval&gt;)
Duncan Sands36397f52007-07-27 12:58:54 +00007248</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007249
Duncan Sands36397f52007-07-27 12:58:54 +00007250<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007251<p>This fills the memory pointed to by <tt>tramp</tt> with code and returns a
7252 function pointer suitable for executing it.</p>
7253
Duncan Sands36397f52007-07-27 12:58:54 +00007254<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007255<p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
7256 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
7257 sufficiently aligned block of memory; this memory is written to by the
7258 intrinsic. Note that the size and the alignment are target-specific - LLVM
7259 currently provides no portable way of determining them, so a front-end that
7260 generates this intrinsic needs to have some target-specific knowledge.
7261 The <tt>func</tt> argument must hold a function bitcast to
7262 an <tt>i8*</tt>.</p>
7263
Duncan Sands36397f52007-07-27 12:58:54 +00007264<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007265<p>The block of memory pointed to by <tt>tramp</tt> is filled with target
7266 dependent code, turning it into a function. A pointer to this function is
7267 returned, but needs to be bitcast to an <a href="#int_trampoline">appropriate
7268 function pointer type</a> before being called. The new function's signature
7269 is the same as that of <tt>func</tt> with any arguments marked with
7270 the <tt>nest</tt> attribute removed. At most one such <tt>nest</tt> argument
7271 is allowed, and it must be of pointer type. Calling the new function is
7272 equivalent to calling <tt>func</tt> with the same argument list, but
7273 with <tt>nval</tt> used for the missing <tt>nest</tt> argument. If, after
7274 calling <tt>llvm.init.trampoline</tt>, the memory pointed to
7275 by <tt>tramp</tt> is modified, then the effect of any later call to the
7276 returned function pointer is undefined.</p>
7277
Duncan Sands36397f52007-07-27 12:58:54 +00007278</div>
7279
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007280</div>
7281
Duncan Sands36397f52007-07-27 12:58:54 +00007282<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007283<h3>
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007284 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007285</h3>
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007286
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007287<div>
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007288
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007289<p>These intrinsic functions expand the "universal IR" of LLVM to represent
7290 hardware constructs for atomic operations and memory synchronization. This
7291 provides an interface to the hardware, not an interface to the programmer. It
7292 is aimed at a low enough level to allow any programming models or APIs
7293 (Application Programming Interfaces) which need atomic behaviors to map
7294 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
7295 hardware provides a "universal IR" for source languages, it also provides a
7296 starting point for developing a "universal" atomic operation and
7297 synchronization IR.</p>
7298
7299<p>These do <em>not</em> form an API such as high-level threading libraries,
7300 software transaction memory systems, atomic primitives, and intrinsic
7301 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
7302 application libraries. The hardware interface provided by LLVM should allow
7303 a clean implementation of all of these APIs and parallel programming models.
7304 No one model or paradigm should be selected above others unless the hardware
7305 itself ubiquitously does so.</p>
7306
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007307<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007308<h4>
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007309 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007310</h4>
7311
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007312<div>
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007313<h5>Syntax:</h5>
7314<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007315 declare void @llvm.memory.barrier(i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt;, i1 &lt;device&gt;)
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007316</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007317
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007318<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007319<p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
7320 specific pairs of memory access types.</p>
7321
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007322<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007323<p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
7324 The first four arguments enables a specific barrier as listed below. The
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00007325 fifth argument specifies that the barrier applies to io or device or uncached
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007326 memory.</p>
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007327
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007328<ul>
7329 <li><tt>ll</tt>: load-load barrier</li>
7330 <li><tt>ls</tt>: load-store barrier</li>
7331 <li><tt>sl</tt>: store-load barrier</li>
7332 <li><tt>ss</tt>: store-store barrier</li>
7333 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
7334</ul>
7335
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007336<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007337<p>This intrinsic causes the system to enforce some ordering constraints upon
7338 the loads and stores of the program. This barrier does not
7339 indicate <em>when</em> any events will occur, it only enforces
7340 an <em>order</em> in which they occur. For any of the specified pairs of load
7341 and store operations (f.ex. load-load, or store-load), all of the first
7342 operations preceding the barrier will complete before any of the second
7343 operations succeeding the barrier begin. Specifically the semantics for each
7344 pairing is as follows:</p>
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007345
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007346<ul>
7347 <li><tt>ll</tt>: All loads before the barrier must complete before any load
7348 after the barrier begins.</li>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00007349 <li><tt>ls</tt>: All loads before the barrier must complete before any
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007350 store after the barrier begins.</li>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00007351 <li><tt>ss</tt>: All stores before the barrier must complete before any
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007352 store after the barrier begins.</li>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00007353 <li><tt>sl</tt>: All stores before the barrier must complete before any
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007354 load after the barrier begins.</li>
7355</ul>
7356
7357<p>These semantics are applied with a logical "and" behavior when more than one
7358 is enabled in a single memory barrier intrinsic.</p>
7359
7360<p>Backends may implement stronger barriers than those requested when they do
7361 not support as fine grained a barrier as requested. Some architectures do
7362 not need all types of barriers and on such architectures, these become
7363 noops.</p>
7364
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007365<h5>Example:</h5>
7366<pre>
Victor Hernandez2fee2942009-10-26 23:44:29 +00007367%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7368%ptr = bitcast i8* %mallocP to i32*
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007369 store i32 4, %ptr
7370
7371%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
Evan Cheng0b0669a2011-06-29 17:14:00 +00007372 call void @llvm.memory.barrier(i1 false, i1 true, i1 false, i1 false, i1 true)
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007373 <i>; guarantee the above finishes</i>
7374 store i32 8, %ptr <i>; before this begins</i>
7375</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007376
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007377</div>
7378
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007379<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007380<h4>
Mon P Wang28873102008-06-25 08:15:39 +00007381 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007382</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007383
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007384<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007385
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007386<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007387<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
7388 any integer bit width and for different address spaces. Not all targets
7389 support all bit widths however.</p>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007390
7391<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007392 declare i8 @llvm.atomic.cmp.swap.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt;)
7393 declare i16 @llvm.atomic.cmp.swap.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt;)
7394 declare i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt;)
7395 declare i64 @llvm.atomic.cmp.swap.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;cmp&gt;, i64 &lt;val&gt;)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007396</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007397
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007398<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007399<p>This loads a value in memory and compares it to a given value. If they are
7400 equal, it stores a new value into the memory.</p>
7401
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007402<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007403<p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
7404 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
7405 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
7406 this integer type. While any bit width integer may be used, targets may only
7407 lower representations they support in hardware.</p>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007408
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007409<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007410<p>This entire intrinsic must be executed atomically. It first loads the value
7411 in memory pointed to by <tt>ptr</tt> and compares it with the
7412 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
7413 memory. The loaded value is yielded in all cases. This provides the
7414 equivalent of an atomic compare-and-swap operation within the SSA
7415 framework.</p>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007416
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007417<h5>Examples:</h5>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007418<pre>
Victor Hernandez2fee2942009-10-26 23:44:29 +00007419%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7420%ptr = bitcast i8* %mallocP to i32*
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007421 store i32 4, %ptr
7422
7423%val1 = add i32 4, 4
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007424%result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 4, %val1)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007425 <i>; yields {i32}:result1 = 4</i>
7426%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7427%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7428
7429%val2 = add i32 1, 1
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007430%result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 5, %val2)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007431 <i>; yields {i32}:result2 = 8</i>
7432%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
7433
7434%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
7435</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007436
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007437</div>
7438
7439<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007440<h4>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007441 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007442</h4>
7443
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007444<div>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007445<h5>Syntax:</h5>
7446
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007447<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
7448 integer bit width. Not all targets support all bit widths however.</p>
7449
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007450<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007451 declare i8 @llvm.atomic.swap.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;val&gt;)
7452 declare i16 @llvm.atomic.swap.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;val&gt;)
7453 declare i32 @llvm.atomic.swap.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;val&gt;)
7454 declare i64 @llvm.atomic.swap.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;val&gt;)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007455</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007456
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007457<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007458<p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
7459 the value from memory. It then stores the value in <tt>val</tt> in the memory
7460 at <tt>ptr</tt>.</p>
7461
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007462<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007463<p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
7464 the <tt>val</tt> argument and the result must be integers of the same bit
7465 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
7466 integer type. The targets may only lower integer representations they
7467 support.</p>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007468
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007469<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007470<p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
7471 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
7472 equivalent of an atomic swap operation within the SSA framework.</p>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007473
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007474<h5>Examples:</h5>
7475<pre>
Victor Hernandez2fee2942009-10-26 23:44:29 +00007476%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7477%ptr = bitcast i8* %mallocP to i32*
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007478 store i32 4, %ptr
7479
7480%val1 = add i32 4, 4
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007481%result1 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val1)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007482 <i>; yields {i32}:result1 = 4</i>
7483%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7484%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7485
7486%val2 = add i32 1, 1
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007487%result2 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val2)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007488 <i>; yields {i32}:result2 = 8</i>
7489
7490%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
7491%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
7492</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007493
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007494</div>
7495
7496<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007497<h4>
Mon P Wang28873102008-06-25 08:15:39 +00007498 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007499</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007500
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007501<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007502
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007503<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007504<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
7505 any integer bit width. Not all targets support all bit widths however.</p>
7506
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007507<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007508 declare i8 @llvm.atomic.load.add.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7509 declare i16 @llvm.atomic.load.add.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7510 declare i32 @llvm.atomic.load.add.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7511 declare i64 @llvm.atomic.load.add.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007512</pre>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007513
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007514<h5>Overview:</h5>
7515<p>This intrinsic adds <tt>delta</tt> to the value stored in memory
7516 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7517
7518<h5>Arguments:</h5>
7519<p>The intrinsic takes two arguments, the first a pointer to an integer value
7520 and the second an integer value. The result is also an integer value. These
7521 integer types can have any bit width, but they must all have the same bit
7522 width. The targets may only lower integer representations they support.</p>
7523
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007524<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007525<p>This intrinsic does a series of operations atomically. It first loads the
7526 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
7527 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007528
7529<h5>Examples:</h5>
7530<pre>
Victor Hernandez2fee2942009-10-26 23:44:29 +00007531%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7532%ptr = bitcast i8* %mallocP to i32*
7533 store i32 4, %ptr
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007534%result1 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 4)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007535 <i>; yields {i32}:result1 = 4</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007536%result2 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 2)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007537 <i>; yields {i32}:result2 = 8</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007538%result3 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 5)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007539 <i>; yields {i32}:result3 = 10</i>
Mon P Wang28873102008-06-25 08:15:39 +00007540%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007541</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007542
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007543</div>
7544
Mon P Wang28873102008-06-25 08:15:39 +00007545<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007546<h4>
Mon P Wang28873102008-06-25 08:15:39 +00007547 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007548</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007549
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007550<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007551
Mon P Wang28873102008-06-25 08:15:39 +00007552<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007553<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
7554 any integer bit width and for different address spaces. Not all targets
7555 support all bit widths however.</p>
7556
Mon P Wang28873102008-06-25 08:15:39 +00007557<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007558 declare i8 @llvm.atomic.load.sub.i8.p0i32(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7559 declare i16 @llvm.atomic.load.sub.i16.p0i32(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7560 declare i32 @llvm.atomic.load.sub.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7561 declare i64 @llvm.atomic.load.sub.i64.p0i32(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007562</pre>
Mon P Wang28873102008-06-25 08:15:39 +00007563
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007564<h5>Overview:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00007565<p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007566 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7567
7568<h5>Arguments:</h5>
7569<p>The intrinsic takes two arguments, the first a pointer to an integer value
7570 and the second an integer value. The result is also an integer value. These
7571 integer types can have any bit width, but they must all have the same bit
7572 width. The targets may only lower integer representations they support.</p>
7573
Mon P Wang28873102008-06-25 08:15:39 +00007574<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007575<p>This intrinsic does a series of operations atomically. It first loads the
7576 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
7577 result to <tt>ptr</tt>. It yields the original value stored
7578 at <tt>ptr</tt>.</p>
Mon P Wang28873102008-06-25 08:15:39 +00007579
7580<h5>Examples:</h5>
7581<pre>
Victor Hernandez2fee2942009-10-26 23:44:29 +00007582%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7583%ptr = bitcast i8* %mallocP to i32*
7584 store i32 8, %ptr
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007585%result1 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 4)
Mon P Wang28873102008-06-25 08:15:39 +00007586 <i>; yields {i32}:result1 = 8</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007587%result2 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 2)
Mon P Wang28873102008-06-25 08:15:39 +00007588 <i>; yields {i32}:result2 = 4</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007589%result3 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 5)
Mon P Wang28873102008-06-25 08:15:39 +00007590 <i>; yields {i32}:result3 = 2</i>
7591%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
7592</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007593
Mon P Wang28873102008-06-25 08:15:39 +00007594</div>
7595
7596<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007597<h4>
7598 <a name="int_atomic_load_and">
7599 '<tt>llvm.atomic.load.and.*</tt>' Intrinsic
7600 </a>
7601 <br>
7602 <a name="int_atomic_load_nand">
7603 '<tt>llvm.atomic.load.nand.*</tt>' Intrinsic
7604 </a>
7605 <br>
7606 <a name="int_atomic_load_or">
7607 '<tt>llvm.atomic.load.or.*</tt>' Intrinsic
7608 </a>
7609 <br>
7610 <a name="int_atomic_load_xor">
7611 '<tt>llvm.atomic.load.xor.*</tt>' Intrinsic
7612 </a>
7613</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007614
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007615<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007616
Mon P Wang28873102008-06-25 08:15:39 +00007617<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007618<p>These are overloaded intrinsics. You can
7619 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
7620 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
7621 bit width and for different address spaces. Not all targets support all bit
7622 widths however.</p>
Mon P Wang28873102008-06-25 08:15:39 +00007623
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007624<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007625 declare i8 @llvm.atomic.load.and.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7626 declare i16 @llvm.atomic.load.and.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7627 declare i32 @llvm.atomic.load.and.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7628 declare i64 @llvm.atomic.load.and.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007629</pre>
7630
7631<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007632 declare i8 @llvm.atomic.load.or.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7633 declare i16 @llvm.atomic.load.or.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7634 declare i32 @llvm.atomic.load.or.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7635 declare i64 @llvm.atomic.load.or.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007636</pre>
7637
7638<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007639 declare i8 @llvm.atomic.load.nand.i8.p0i32(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7640 declare i16 @llvm.atomic.load.nand.i16.p0i32(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7641 declare i32 @llvm.atomic.load.nand.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7642 declare i64 @llvm.atomic.load.nand.i64.p0i32(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007643</pre>
7644
7645<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007646 declare i8 @llvm.atomic.load.xor.i8.p0i32(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7647 declare i16 @llvm.atomic.load.xor.i16.p0i32(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7648 declare i32 @llvm.atomic.load.xor.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7649 declare i64 @llvm.atomic.load.xor.i64.p0i32(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007650</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007651
Mon P Wang28873102008-06-25 08:15:39 +00007652<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007653<p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
7654 the value stored in memory at <tt>ptr</tt>. It yields the original value
7655 at <tt>ptr</tt>.</p>
Mon P Wang28873102008-06-25 08:15:39 +00007656
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007657<h5>Arguments:</h5>
7658<p>These intrinsics take two arguments, the first a pointer to an integer value
7659 and the second an integer value. The result is also an integer value. These
7660 integer types can have any bit width, but they must all have the same bit
7661 width. The targets may only lower integer representations they support.</p>
7662
Mon P Wang28873102008-06-25 08:15:39 +00007663<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007664<p>These intrinsics does a series of operations atomically. They first load the
7665 value stored at <tt>ptr</tt>. They then do the bitwise
7666 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
7667 original value stored at <tt>ptr</tt>.</p>
Mon P Wang28873102008-06-25 08:15:39 +00007668
7669<h5>Examples:</h5>
7670<pre>
Victor Hernandez2fee2942009-10-26 23:44:29 +00007671%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7672%ptr = bitcast i8* %mallocP to i32*
7673 store i32 0x0F0F, %ptr
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007674%result0 = call i32 @llvm.atomic.load.nand.i32.p0i32(i32* %ptr, i32 0xFF)
Mon P Wang28873102008-06-25 08:15:39 +00007675 <i>; yields {i32}:result0 = 0x0F0F</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007676%result1 = call i32 @llvm.atomic.load.and.i32.p0i32(i32* %ptr, i32 0xFF)
Mon P Wang28873102008-06-25 08:15:39 +00007677 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007678%result2 = call i32 @llvm.atomic.load.or.i32.p0i32(i32* %ptr, i32 0F)
Mon P Wang28873102008-06-25 08:15:39 +00007679 <i>; yields {i32}:result2 = 0xF0</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007680%result3 = call i32 @llvm.atomic.load.xor.i32.p0i32(i32* %ptr, i32 0F)
Mon P Wang28873102008-06-25 08:15:39 +00007681 <i>; yields {i32}:result3 = FF</i>
7682%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
7683</pre>
Mon P Wang28873102008-06-25 08:15:39 +00007684
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007685</div>
Mon P Wang28873102008-06-25 08:15:39 +00007686
7687<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007688<h4>
7689 <a name="int_atomic_load_max">
7690 '<tt>llvm.atomic.load.max.*</tt>' Intrinsic
7691 </a>
7692 <br>
7693 <a name="int_atomic_load_min">
7694 '<tt>llvm.atomic.load.min.*</tt>' Intrinsic
7695 </a>
7696 <br>
7697 <a name="int_atomic_load_umax">
7698 '<tt>llvm.atomic.load.umax.*</tt>' Intrinsic
7699 </a>
7700 <br>
7701 <a name="int_atomic_load_umin">
7702 '<tt>llvm.atomic.load.umin.*</tt>' Intrinsic
7703 </a>
7704</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007705
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007706<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007707
Mon P Wang28873102008-06-25 08:15:39 +00007708<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007709<p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
7710 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
7711 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
7712 address spaces. Not all targets support all bit widths however.</p>
Mon P Wang28873102008-06-25 08:15:39 +00007713
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007714<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007715 declare i8 @llvm.atomic.load.max.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7716 declare i16 @llvm.atomic.load.max.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7717 declare i32 @llvm.atomic.load.max.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7718 declare i64 @llvm.atomic.load.max.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007719</pre>
7720
7721<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007722 declare i8 @llvm.atomic.load.min.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7723 declare i16 @llvm.atomic.load.min.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7724 declare i32 @llvm.atomic.load.min.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7725 declare i64 @llvm.atomic.load.min.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007726</pre>
7727
7728<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007729 declare i8 @llvm.atomic.load.umax.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7730 declare i16 @llvm.atomic.load.umax.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7731 declare i32 @llvm.atomic.load.umax.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7732 declare i64 @llvm.atomic.load.umax.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007733</pre>
7734
7735<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007736 declare i8 @llvm.atomic.load.umin.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7737 declare i16 @llvm.atomic.load.umin.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7738 declare i32 @llvm.atomic.load.umin.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7739 declare i64 @llvm.atomic.load.umin.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007740</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007741
Mon P Wang28873102008-06-25 08:15:39 +00007742<h5>Overview:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00007743<p>These intrinsics takes the signed or unsigned minimum or maximum of
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007744 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
7745 original value at <tt>ptr</tt>.</p>
Mon P Wang28873102008-06-25 08:15:39 +00007746
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007747<h5>Arguments:</h5>
7748<p>These intrinsics take two arguments, the first a pointer to an integer value
7749 and the second an integer value. The result is also an integer value. These
7750 integer types can have any bit width, but they must all have the same bit
7751 width. The targets may only lower integer representations they support.</p>
7752
Mon P Wang28873102008-06-25 08:15:39 +00007753<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007754<p>These intrinsics does a series of operations atomically. They first load the
7755 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
7756 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
7757 yield the original value stored at <tt>ptr</tt>.</p>
Mon P Wang28873102008-06-25 08:15:39 +00007758
7759<h5>Examples:</h5>
7760<pre>
Victor Hernandez2fee2942009-10-26 23:44:29 +00007761%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7762%ptr = bitcast i8* %mallocP to i32*
7763 store i32 7, %ptr
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007764%result0 = call i32 @llvm.atomic.load.min.i32.p0i32(i32* %ptr, i32 -2)
Mon P Wang28873102008-06-25 08:15:39 +00007765 <i>; yields {i32}:result0 = 7</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007766%result1 = call i32 @llvm.atomic.load.max.i32.p0i32(i32* %ptr, i32 8)
Mon P Wang28873102008-06-25 08:15:39 +00007767 <i>; yields {i32}:result1 = -2</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007768%result2 = call i32 @llvm.atomic.load.umin.i32.p0i32(i32* %ptr, i32 10)
Mon P Wang28873102008-06-25 08:15:39 +00007769 <i>; yields {i32}:result2 = 8</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007770%result3 = call i32 @llvm.atomic.load.umax.i32.p0i32(i32* %ptr, i32 30)
Mon P Wang28873102008-06-25 08:15:39 +00007771 <i>; yields {i32}:result3 = 8</i>
7772%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
7773</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007774
Mon P Wang28873102008-06-25 08:15:39 +00007775</div>
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007776
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007777</div>
Nick Lewyckycc271862009-10-13 07:03:23 +00007778
7779<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007780<h3>
Nick Lewyckycc271862009-10-13 07:03:23 +00007781 <a name="int_memorymarkers">Memory Use Markers</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007782</h3>
Nick Lewyckycc271862009-10-13 07:03:23 +00007783
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007784<div>
Nick Lewyckycc271862009-10-13 07:03:23 +00007785
7786<p>This class of intrinsics exists to information about the lifetime of memory
7787 objects and ranges where variables are immutable.</p>
7788
Nick Lewyckycc271862009-10-13 07:03:23 +00007789<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007790<h4>
Nick Lewyckycc271862009-10-13 07:03:23 +00007791 <a name="int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007792</h4>
Nick Lewyckycc271862009-10-13 07:03:23 +00007793
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007794<div>
Nick Lewyckycc271862009-10-13 07:03:23 +00007795
7796<h5>Syntax:</h5>
7797<pre>
7798 declare void @llvm.lifetime.start(i64 &lt;size&gt;, i8* nocapture &lt;ptr&gt;)
7799</pre>
7800
7801<h5>Overview:</h5>
7802<p>The '<tt>llvm.lifetime.start</tt>' intrinsic specifies the start of a memory
7803 object's lifetime.</p>
7804
7805<h5>Arguments:</h5>
Nick Lewycky321333e2009-10-13 07:57:33 +00007806<p>The first argument is a constant integer representing the size of the
7807 object, or -1 if it is variable sized. The second argument is a pointer to
7808 the object.</p>
Nick Lewyckycc271862009-10-13 07:03:23 +00007809
7810<h5>Semantics:</h5>
7811<p>This intrinsic indicates that before this point in the code, the value of the
7812 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
Nick Lewycky8d336592009-10-27 16:56:58 +00007813 never be used and has an undefined value. A load from the pointer that
7814 precedes this intrinsic can be replaced with
Nick Lewyckycc271862009-10-13 07:03:23 +00007815 <tt>'<a href="#undefvalues">undef</a>'</tt>.</p>
7816
7817</div>
7818
7819<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007820<h4>
Nick Lewyckycc271862009-10-13 07:03:23 +00007821 <a name="int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007822</h4>
Nick Lewyckycc271862009-10-13 07:03:23 +00007823
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007824<div>
Nick Lewyckycc271862009-10-13 07:03:23 +00007825
7826<h5>Syntax:</h5>
7827<pre>
7828 declare void @llvm.lifetime.end(i64 &lt;size&gt;, i8* nocapture &lt;ptr&gt;)
7829</pre>
7830
7831<h5>Overview:</h5>
7832<p>The '<tt>llvm.lifetime.end</tt>' intrinsic specifies the end of a memory
7833 object's lifetime.</p>
7834
7835<h5>Arguments:</h5>
Nick Lewycky321333e2009-10-13 07:57:33 +00007836<p>The first argument is a constant integer representing the size of the
7837 object, or -1 if it is variable sized. The second argument is a pointer to
7838 the object.</p>
Nick Lewyckycc271862009-10-13 07:03:23 +00007839
7840<h5>Semantics:</h5>
7841<p>This intrinsic indicates that after this point in the code, the value of the
7842 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7843 never be used and has an undefined value. Any stores into the memory object
7844 following this intrinsic may be removed as dead.
7845
7846</div>
7847
7848<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007849<h4>
Nick Lewyckycc271862009-10-13 07:03:23 +00007850 <a name="int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007851</h4>
Nick Lewyckycc271862009-10-13 07:03:23 +00007852
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007853<div>
Nick Lewyckycc271862009-10-13 07:03:23 +00007854
7855<h5>Syntax:</h5>
7856<pre>
Nick Lewycky29b6cb42010-11-30 04:13:41 +00007857 declare {}* @llvm.invariant.start(i64 &lt;size&gt;, i8* nocapture &lt;ptr&gt;)
Nick Lewyckycc271862009-10-13 07:03:23 +00007858</pre>
7859
7860<h5>Overview:</h5>
7861<p>The '<tt>llvm.invariant.start</tt>' intrinsic specifies that the contents of
7862 a memory object will not change.</p>
7863
7864<h5>Arguments:</h5>
Nick Lewycky321333e2009-10-13 07:57:33 +00007865<p>The first argument is a constant integer representing the size of the
7866 object, or -1 if it is variable sized. The second argument is a pointer to
7867 the object.</p>
Nick Lewyckycc271862009-10-13 07:03:23 +00007868
7869<h5>Semantics:</h5>
7870<p>This intrinsic indicates that until an <tt>llvm.invariant.end</tt> that uses
7871 the return value, the referenced memory location is constant and
7872 unchanging.</p>
7873
7874</div>
7875
7876<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007877<h4>
Nick Lewyckycc271862009-10-13 07:03:23 +00007878 <a name="int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007879</h4>
Nick Lewyckycc271862009-10-13 07:03:23 +00007880
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007881<div>
Nick Lewyckycc271862009-10-13 07:03:23 +00007882
7883<h5>Syntax:</h5>
7884<pre>
7885 declare void @llvm.invariant.end({}* &lt;start&gt;, i64 &lt;size&gt;, i8* nocapture &lt;ptr&gt;)
7886</pre>
7887
7888<h5>Overview:</h5>
7889<p>The '<tt>llvm.invariant.end</tt>' intrinsic specifies that the contents of
7890 a memory object are mutable.</p>
7891
7892<h5>Arguments:</h5>
7893<p>The first argument is the matching <tt>llvm.invariant.start</tt> intrinsic.
Nick Lewycky321333e2009-10-13 07:57:33 +00007894 The second argument is a constant integer representing the size of the
7895 object, or -1 if it is variable sized and the third argument is a pointer
7896 to the object.</p>
Nick Lewyckycc271862009-10-13 07:03:23 +00007897
7898<h5>Semantics:</h5>
7899<p>This intrinsic indicates that the memory is mutable again.</p>
7900
7901</div>
7902
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007903</div>
7904
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007905<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007906<h3>
Tanya Lattner6d806e92007-06-15 20:50:54 +00007907 <a name="int_general">General Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007908</h3>
Tanya Lattner6d806e92007-06-15 20:50:54 +00007909
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007910<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007911
7912<p>This class of intrinsics is designed to be generic and has no specific
7913 purpose.</p>
7914
Tanya Lattner6d806e92007-06-15 20:50:54 +00007915<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007916<h4>
Tanya Lattner6d806e92007-06-15 20:50:54 +00007917 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007918</h4>
Tanya Lattner6d806e92007-06-15 20:50:54 +00007919
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007920<div>
Tanya Lattner6d806e92007-06-15 20:50:54 +00007921
7922<h5>Syntax:</h5>
7923<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007924 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
Tanya Lattner6d806e92007-06-15 20:50:54 +00007925</pre>
7926
7927<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007928<p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
Tanya Lattner6d806e92007-06-15 20:50:54 +00007929
7930<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007931<p>The first argument is a pointer to a value, the second is a pointer to a
7932 global string, the third is a pointer to a global string which is the source
7933 file name, and the last argument is the line number.</p>
Tanya Lattner6d806e92007-06-15 20:50:54 +00007934
7935<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007936<p>This intrinsic allows annotation of local variables with arbitrary strings.
7937 This can be useful for special purpose optimizations that want to look for
7938 these annotations. These have no other defined use, they are ignored by code
7939 generation and optimization.</p>
Tanya Lattner6d806e92007-06-15 20:50:54 +00007940
Tanya Lattner6d806e92007-06-15 20:50:54 +00007941</div>
7942
Tanya Lattnerb6367882007-09-21 22:59:12 +00007943<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007944<h4>
Tanya Lattnere1a8da02007-09-21 23:57:59 +00007945 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007946</h4>
Tanya Lattnerb6367882007-09-21 22:59:12 +00007947
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007948<div>
Tanya Lattnerb6367882007-09-21 22:59:12 +00007949
7950<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007951<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
7952 any integer bit width.</p>
7953
Tanya Lattnerb6367882007-09-21 22:59:12 +00007954<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007955 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
7956 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
7957 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
7958 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
7959 declare i256 @llvm.annotation.i256(i256 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
Tanya Lattnerb6367882007-09-21 22:59:12 +00007960</pre>
7961
7962<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007963<p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
Tanya Lattnerb6367882007-09-21 22:59:12 +00007964
7965<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007966<p>The first argument is an integer value (result of some expression), the
7967 second is a pointer to a global string, the third is a pointer to a global
7968 string which is the source file name, and the last argument is the line
7969 number. It returns the value of the first argument.</p>
Tanya Lattnerb6367882007-09-21 22:59:12 +00007970
7971<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007972<p>This intrinsic allows annotations to be put on arbitrary expressions with
7973 arbitrary strings. This can be useful for special purpose optimizations that
7974 want to look for these annotations. These have no other defined use, they
7975 are ignored by code generation and optimization.</p>
Tanya Lattnerb6367882007-09-21 22:59:12 +00007976
Tanya Lattnerb6367882007-09-21 22:59:12 +00007977</div>
Jim Laskeydd4ef1b2007-03-14 19:31:19 +00007978
Anton Korobeynikov4cb86182008-01-15 22:31:34 +00007979<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007980<h4>
Anton Korobeynikov4cb86182008-01-15 22:31:34 +00007981 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007982</h4>
Anton Korobeynikov4cb86182008-01-15 22:31:34 +00007983
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007984<div>
Anton Korobeynikov4cb86182008-01-15 22:31:34 +00007985
7986<h5>Syntax:</h5>
7987<pre>
7988 declare void @llvm.trap()
7989</pre>
7990
7991<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007992<p>The '<tt>llvm.trap</tt>' intrinsic.</p>
Anton Korobeynikov4cb86182008-01-15 22:31:34 +00007993
7994<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007995<p>None.</p>
Anton Korobeynikov4cb86182008-01-15 22:31:34 +00007996
7997<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007998<p>This intrinsics is lowered to the target dependent trap instruction. If the
7999 target does not have a trap instruction, this intrinsic will be lowered to
8000 the call of the <tt>abort()</tt> function.</p>
Anton Korobeynikov4cb86182008-01-15 22:31:34 +00008001
Anton Korobeynikov4cb86182008-01-15 22:31:34 +00008002</div>
8003
Bill Wendling69e4adb2008-11-19 05:56:17 +00008004<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00008005<h4>
Misha Brukmandccb0252008-11-22 23:55:29 +00008006 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00008007</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008008
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00008009<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008010
Bill Wendling69e4adb2008-11-19 05:56:17 +00008011<h5>Syntax:</h5>
8012<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00008013 declare void @llvm.stackprotector(i8* &lt;guard&gt;, i8** &lt;slot&gt;)
Bill Wendling69e4adb2008-11-19 05:56:17 +00008014</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008015
Bill Wendling69e4adb2008-11-19 05:56:17 +00008016<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008017<p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
8018 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
8019 ensure that it is placed on the stack before local variables.</p>
8020
Bill Wendling69e4adb2008-11-19 05:56:17 +00008021<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008022<p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
8023 arguments. The first argument is the value loaded from the stack
8024 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
8025 that has enough space to hold the value of the guard.</p>
8026
Bill Wendling69e4adb2008-11-19 05:56:17 +00008027<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008028<p>This intrinsic causes the prologue/epilogue inserter to force the position of
8029 the <tt>AllocaInst</tt> stack slot to be before local variables on the
8030 stack. This is to ensure that if a local variable on the stack is
8031 overwritten, it will destroy the value of the guard. When the function exits,
Bill Wendling1b383ba2010-10-27 01:07:41 +00008032 the guard on the stack is checked against the original guard. If they are
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008033 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
8034 function.</p>
8035
Bill Wendling69e4adb2008-11-19 05:56:17 +00008036</div>
8037
Eric Christopher0e671492009-11-30 08:03:53 +00008038<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00008039<h4>
Eric Christopher0e671492009-11-30 08:03:53 +00008040 <a name="int_objectsize">'<tt>llvm.objectsize</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00008041</h4>
Eric Christopher0e671492009-11-30 08:03:53 +00008042
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00008043<div>
Eric Christopher0e671492009-11-30 08:03:53 +00008044
8045<h5>Syntax:</h5>
8046<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00008047 declare i32 @llvm.objectsize.i32(i8* &lt;object&gt;, i1 &lt;type&gt;)
8048 declare i64 @llvm.objectsize.i64(i8* &lt;object&gt;, i1 &lt;type&gt;)
Eric Christopher0e671492009-11-30 08:03:53 +00008049</pre>
8050
8051<h5>Overview:</h5>
Bill Wendling1b383ba2010-10-27 01:07:41 +00008052<p>The <tt>llvm.objectsize</tt> intrinsic is designed to provide information to
8053 the optimizers to determine at compile time whether a) an operation (like
8054 memcpy) will overflow a buffer that corresponds to an object, or b) that a
8055 runtime check for overflow isn't necessary. An object in this context means
8056 an allocation of a specific class, structure, array, or other object.</p>
Eric Christopher0e671492009-11-30 08:03:53 +00008057
8058<h5>Arguments:</h5>
Bill Wendling1b383ba2010-10-27 01:07:41 +00008059<p>The <tt>llvm.objectsize</tt> intrinsic takes two arguments. The first
Eric Christopher8295a0a2009-12-23 00:29:49 +00008060 argument is a pointer to or into the <tt>object</tt>. The second argument
Bill Wendling1b383ba2010-10-27 01:07:41 +00008061 is a boolean 0 or 1. This argument determines whether you want the
8062 maximum (0) or minimum (1) bytes remaining. This needs to be a literal 0 or
Eric Christopher8295a0a2009-12-23 00:29:49 +00008063 1, variables are not allowed.</p>
8064
Eric Christopher0e671492009-11-30 08:03:53 +00008065<h5>Semantics:</h5>
8066<p>The <tt>llvm.objectsize</tt> intrinsic is lowered to either a constant
Bill Wendling1b383ba2010-10-27 01:07:41 +00008067 representing the size of the object concerned, or <tt>i32/i64 -1 or 0</tt>,
8068 depending on the <tt>type</tt> argument, if the size cannot be determined at
8069 compile time.</p>
Eric Christopher0e671492009-11-30 08:03:53 +00008070
8071</div>
8072
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00008073</div>
8074
8075</div>
8076
Chris Lattner00950542001-06-06 20:29:01 +00008077<!-- *********************************************************************** -->
Chris Lattner00950542001-06-06 20:29:01 +00008078<hr>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00008079<address>
8080 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman44408702008-12-11 17:34:48 +00008081 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00008082 <a href="http://validator.w3.org/check/referer"><img
Misha Brukman44408702008-12-11 17:34:48 +00008083 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00008084
8085 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
NAKAMURA Takumib9a33632011-04-09 02:13:37 +00008086 <a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00008087 Last modified: $Date$
8088</address>
Chris Lattnerc7d3ab32008-01-04 04:33:49 +00008089
Misha Brukman9d0919f2003-11-08 01:05:38 +00008090</body>
8091</html>