blob: 40affb7e917068c6ba527b72c17aabd4f6fabdfc [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 Lattner628ed392011-07-23 19:59:08 +000078 <li><a href="#t_opaque">Opaque Structure Types</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>
Bill Wendling772fe172011-07-27 20:18:04 +0000126 <li><a href="#i_resume">'<tt>resume</tt>' Instruction</a></li>
Chris Lattner35eca582004-10-16 18:04:13 +0000127 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000128 </ol>
129 </li>
Chris Lattner00950542001-06-06 20:29:01 +0000130 <li><a href="#binaryops">Binary Operations</a>
131 <ol>
Chris Lattner261efe92003-11-25 01:02:51 +0000132 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
Dan Gohmanae3a0be2009-06-04 22:49:04 +0000133 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000134 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
Dan Gohmanae3a0be2009-06-04 22:49:04 +0000135 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000136 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
Dan Gohmanae3a0be2009-06-04 22:49:04 +0000137 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
Reid Spencer1628cec2006-10-26 06:15:43 +0000138 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
139 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
140 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
Reid Spencer0a783f72006-11-02 01:53:59 +0000141 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
142 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
143 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000144 </ol>
145 </li>
Chris Lattner00950542001-06-06 20:29:01 +0000146 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
147 <ol>
Reid Spencer8e11bf82007-02-02 13:57:07 +0000148 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
149 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
150 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000151 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000152 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000153 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000154 </ol>
155 </li>
Chris Lattner3df241e2006-04-08 23:07:04 +0000156 <li><a href="#vectorops">Vector Operations</a>
157 <ol>
158 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
159 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
160 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
Chris Lattner3df241e2006-04-08 23:07:04 +0000161 </ol>
162 </li>
Dan Gohmana334d5f2008-05-12 23:51:09 +0000163 <li><a href="#aggregateops">Aggregate Operations</a>
164 <ol>
165 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
166 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
167 </ol>
168 </li>
Chris Lattner884a9702006-08-15 00:45:58 +0000169 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
Chris Lattner00950542001-06-06 20:29:01 +0000170 <ol>
Chris Lattner261efe92003-11-25 01:02:51 +0000171 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
Robert Bocchino7b81c752006-02-17 21:18:08 +0000172 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
173 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
Eli Friedman47f35132011-07-25 23:16:38 +0000174 <li><a href="#i_fence">'<tt>fence</tt>' Instruction</a></li>
Robert Bocchino7b81c752006-02-17 21:18:08 +0000175 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000176 </ol>
177 </li>
Reid Spencer2fd21e62006-11-08 01:18:52 +0000178 <li><a href="#convertops">Conversion Operations</a>
Reid Spencer9dee3ac2006-11-08 01:11:31 +0000179 <ol>
180 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
181 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
182 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
183 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
184 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
Reid Spencerd4448792006-11-09 23:03:26 +0000185 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
186 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
187 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
188 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
Reid Spencer72679252006-11-11 21:00:47 +0000189 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
190 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
Reid Spencer5c0ef472006-11-11 23:08:07 +0000191 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
Reid Spencer9dee3ac2006-11-08 01:11:31 +0000192 </ol>
Dan Gohman0e451ce2008-10-14 16:51:45 +0000193 </li>
Chris Lattner00950542001-06-06 20:29:01 +0000194 <li><a href="#otherops">Other Operations</a>
195 <ol>
Reid Spencerf3a70a62006-11-18 21:50:54 +0000196 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
197 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000198 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
Chris Lattnercc37aae2004-03-12 05:50:16 +0000199 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000200 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
Chris Lattnerfb6977d2006-01-13 23:26:01 +0000201 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
Chris Lattner00950542001-06-06 20:29:01 +0000202 </ol>
Chris Lattner261efe92003-11-25 01:02:51 +0000203 </li>
Chris Lattner00950542001-06-06 20:29:01 +0000204 </ol>
Chris Lattner261efe92003-11-25 01:02:51 +0000205 </li>
Chris Lattnerd9ad5b32003-05-08 04:57:36 +0000206 <li><a href="#intrinsics">Intrinsic Functions</a>
Chris Lattnerd9ad5b32003-05-08 04:57:36 +0000207 <ol>
Chris Lattner261efe92003-11-25 01:02:51 +0000208 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
209 <ol>
Reid Spencera3e435f2007-04-04 02:42:35 +0000210 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
211 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
212 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000213 </ol>
214 </li>
Chris Lattnerd7923912004-05-23 21:06:01 +0000215 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
216 <ol>
Reid Spencera3e435f2007-04-04 02:42:35 +0000217 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
218 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
219 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
Chris Lattnerd7923912004-05-23 21:06:01 +0000220 </ol>
221 </li>
Chris Lattner10610642004-02-14 04:08:35 +0000222 <li><a href="#int_codegen">Code Generator Intrinsics</a>
223 <ol>
Reid Spencera3e435f2007-04-04 02:42:35 +0000224 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
225 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
226 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
227 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
228 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
229 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
Dan Gohman31f1af12010-05-26 21:56:15 +0000230 <li><a href="#int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
John Criswell7123e272004-04-09 16:43:20 +0000231 </ol>
232 </li>
Chris Lattner33aec9e2004-02-12 17:01:32 +0000233 <li><a href="#int_libc">Standard C Library Intrinsics</a>
234 <ol>
Reid Spencera3e435f2007-04-04 02:42:35 +0000235 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
236 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
237 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
238 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
239 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohman91c284c2007-10-15 20:30:11 +0000240 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
241 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
242 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Dan Gohman08b280b2011-05-27 00:36:31 +0000243 <li><a href="#int_exp">'<tt>llvm.exp.*</tt>' Intrinsic</a></li>
244 <li><a href="#int_log">'<tt>llvm.log.*</tt>' Intrinsic</a></li>
Cameron Zwarich33390842011-07-08 21:39:21 +0000245 <li><a href="#int_fma">'<tt>llvm.fma.*</tt>' Intrinsic</a></li>
Chris Lattner33aec9e2004-02-12 17:01:32 +0000246 </ol>
247 </li>
Nate Begeman7e36c472006-01-13 23:26:38 +0000248 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
Andrew Lenharthec370fd2005-05-03 18:01:48 +0000249 <ol>
Reid Spencera3e435f2007-04-04 02:42:35 +0000250 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
Chris Lattner8a886be2006-01-16 22:34:14 +0000251 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
252 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
253 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
Andrew Lenharthec370fd2005-05-03 18:01:48 +0000254 </ol>
255 </li>
Bill Wendlingac1df8e2009-02-08 01:40:31 +0000256 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
257 <ol>
Bill Wendlingda01af72009-02-08 04:04:40 +0000258 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
259 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
260 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
261 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
262 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
Bill Wendling41b485c2009-02-08 23:00:09 +0000263 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
Bill Wendlingac1df8e2009-02-08 01:40:31 +0000264 </ol>
265 </li>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +0000266 <li><a href="#int_fp16">Half Precision Floating Point Intrinsics</a>
267 <ol>
Chris Lattner82c3dc62010-03-14 23:03:31 +0000268 <li><a href="#int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a></li>
269 <li><a href="#int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a></li>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +0000270 </ol>
271 </li>
Chris Lattnerd7923912004-05-23 21:06:01 +0000272 <li><a href="#int_debugger">Debugger intrinsics</a></li>
Jim Laskeydd4ef1b2007-03-14 19:31:19 +0000273 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sandsf7331b32007-09-11 14:10:23 +0000274 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands36397f52007-07-27 12:58:54 +0000275 <ol>
276 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands36397f52007-07-27 12:58:54 +0000277 </ol>
278 </li>
Bill Wendling3c44f5b2008-11-18 22:10:53 +0000279 <li><a href="#int_atomics">Atomic intrinsics</a>
280 <ol>
281 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
282 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
283 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
284 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
285 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
286 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
287 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
288 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
289 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
290 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
291 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
292 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
293 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
294 </ol>
295 </li>
Nick Lewyckycc271862009-10-13 07:03:23 +0000296 <li><a href="#int_memorymarkers">Memory Use Markers</a>
297 <ol>
298 <li><a href="#int_lifetime_start"><tt>llvm.lifetime.start</tt></a></li>
299 <li><a href="#int_lifetime_end"><tt>llvm.lifetime.end</tt></a></li>
300 <li><a href="#int_invariant_start"><tt>llvm.invariant.start</tt></a></li>
301 <li><a href="#int_invariant_end"><tt>llvm.invariant.end</tt></a></li>
302 </ol>
303 </li>
Reid Spencer20677642007-07-20 19:59:11 +0000304 <li><a href="#int_general">General intrinsics</a>
Tanya Lattner6d806e92007-06-15 20:50:54 +0000305 <ol>
Reid Spencer20677642007-07-20 19:59:11 +0000306 <li><a href="#int_var_annotation">
Bill Wendling69e4adb2008-11-19 05:56:17 +0000307 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Tanya Lattnerb6367882007-09-21 22:59:12 +0000308 <li><a href="#int_annotation">
Bill Wendling69e4adb2008-11-19 05:56:17 +0000309 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Anton Korobeynikov4cb86182008-01-15 22:31:34 +0000310 <li><a href="#int_trap">
Bill Wendling69e4adb2008-11-19 05:56:17 +0000311 '<tt>llvm.trap</tt>' Intrinsic</a></li>
312 <li><a href="#int_stackprotector">
313 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
Eric Christopher0e671492009-11-30 08:03:53 +0000314 <li><a href="#int_objectsize">
315 '<tt>llvm.objectsize</tt>' Intrinsic</a></li>
Tanya Lattnerb6367882007-09-21 22:59:12 +0000316 </ol>
Tanya Lattner6d806e92007-06-15 20:50:54 +0000317 </li>
Chris Lattner261efe92003-11-25 01:02:51 +0000318 </ol>
319 </li>
Chris Lattner00950542001-06-06 20:29:01 +0000320</ol>
Chris Lattnerd7923912004-05-23 21:06:01 +0000321
322<div class="doc_author">
323 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
324 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000325</div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000326
Chris Lattner00950542001-06-06 20:29:01 +0000327<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000328<h2><a name="abstract">Abstract</a></h2>
Chris Lattner261efe92003-11-25 01:02:51 +0000329<!-- *********************************************************************** -->
Chris Lattnerd7923912004-05-23 21:06:01 +0000330
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000331<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000332
333<p>This document is a reference manual for the LLVM assembly language. LLVM is
334 a Static Single Assignment (SSA) based representation that provides type
335 safety, low-level operations, flexibility, and the capability of representing
336 'all' high-level languages cleanly. It is the common code representation
337 used throughout all phases of the LLVM compilation strategy.</p>
338
Misha Brukman9d0919f2003-11-08 01:05:38 +0000339</div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000340
Chris Lattner00950542001-06-06 20:29:01 +0000341<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000342<h2><a name="introduction">Introduction</a></h2>
Chris Lattner261efe92003-11-25 01:02:51 +0000343<!-- *********************************************************************** -->
Chris Lattnerd7923912004-05-23 21:06:01 +0000344
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000345<div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000346
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000347<p>The LLVM code representation is designed to be used in three different forms:
348 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
349 for fast loading by a Just-In-Time compiler), and as a human readable
350 assembly language representation. This allows LLVM to provide a powerful
351 intermediate representation for efficient compiler transformations and
352 analysis, while providing a natural means to debug and visualize the
353 transformations. The three different forms of LLVM are all equivalent. This
354 document describes the human readable representation and notation.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +0000355
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000356<p>The LLVM representation aims to be light-weight and low-level while being
357 expressive, typed, and extensible at the same time. It aims to be a
358 "universal IR" of sorts, by being at a low enough level that high-level ideas
359 may be cleanly mapped to it (similar to how microprocessors are "universal
360 IR's", allowing many source languages to be mapped to them). By providing
361 type information, LLVM can be used as the target of optimizations: for
362 example, through pointer analysis, it can be proven that a C automatic
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000363 variable is never accessed outside of the current function, allowing it to
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000364 be promoted to a simple SSA value instead of a memory location.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +0000365
Chris Lattner00950542001-06-06 20:29:01 +0000366<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000367<h4>
368 <a name="wellformed">Well-Formedness</a>
369</h4>
Chris Lattnerd7923912004-05-23 21:06:01 +0000370
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000371<div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000372
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000373<p>It is important to note that this document describes 'well formed' LLVM
374 assembly language. There is a difference between what the parser accepts and
375 what is considered 'well formed'. For example, the following instruction is
376 syntactically okay, but not well formed:</p>
Chris Lattnerd7923912004-05-23 21:06:01 +0000377
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000378<pre class="doc_code">
Bill Wendling2f7a8b02007-05-29 09:04:49 +0000379%x = <a href="#i_add">add</a> i32 1, %x
Chris Lattnerd7923912004-05-23 21:06:01 +0000380</pre>
381
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000382<p>because the definition of <tt>%x</tt> does not dominate all of its uses. The
383 LLVM infrastructure provides a verification pass that may be used to verify
384 that an LLVM module is well formed. This pass is automatically run by the
385 parser after parsing input assembly and by the optimizer before it outputs
386 bitcode. The violations pointed out by the verifier pass indicate bugs in
387 transformation passes or input to the parser.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000388
Bill Wendling2f7a8b02007-05-29 09:04:49 +0000389</div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000390
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000391</div>
392
Chris Lattnercc689392007-10-03 17:34:29 +0000393<!-- Describe the typesetting conventions here. -->
Chris Lattnerd7923912004-05-23 21:06:01 +0000394
Chris Lattner00950542001-06-06 20:29:01 +0000395<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000396<h2><a name="identifiers">Identifiers</a></h2>
Chris Lattner00950542001-06-06 20:29:01 +0000397<!-- *********************************************************************** -->
Chris Lattnerd7923912004-05-23 21:06:01 +0000398
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000399<div>
Chris Lattnerd7923912004-05-23 21:06:01 +0000400
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000401<p>LLVM identifiers come in two basic types: global and local. Global
402 identifiers (functions, global variables) begin with the <tt>'@'</tt>
403 character. Local identifiers (register names, types) begin with
404 the <tt>'%'</tt> character. Additionally, there are three different formats
405 for identifiers, for different purposes:</p>
Chris Lattnerd7923912004-05-23 21:06:01 +0000406
Chris Lattner00950542001-06-06 20:29:01 +0000407<ol>
Reid Spencer2c452282007-08-07 14:34:28 +0000408 <li>Named values are represented as a string of characters with their prefix.
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000409 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
410 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
411 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
412 other characters in their names can be surrounded with quotes. Special
413 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
414 ASCII code for the character in hexadecimal. In this way, any character
415 can be used in a name value, even quotes themselves.</li>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000416
Reid Spencer2c452282007-08-07 14:34:28 +0000417 <li>Unnamed values are represented as an unsigned numeric value with their
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000418 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000419
Reid Spencercc16dc32004-12-09 18:02:53 +0000420 <li>Constants, which are described in a <a href="#constants">section about
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000421 constants</a>, below.</li>
Misha Brukman9d0919f2003-11-08 01:05:38 +0000422</ol>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000423
Reid Spencer2c452282007-08-07 14:34:28 +0000424<p>LLVM requires that values start with a prefix for two reasons: Compilers
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000425 don't need to worry about name clashes with reserved words, and the set of
426 reserved words may be expanded in the future without penalty. Additionally,
427 unnamed identifiers allow a compiler to quickly come up with a temporary
428 variable without having to avoid symbol table conflicts.</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000429
Chris Lattner261efe92003-11-25 01:02:51 +0000430<p>Reserved words in LLVM are very similar to reserved words in other
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000431 languages. There are keywords for different opcodes
432 ('<tt><a href="#i_add">add</a></tt>',
433 '<tt><a href="#i_bitcast">bitcast</a></tt>',
434 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
435 ('<tt><a href="#t_void">void</a></tt>',
436 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
437 reserved words cannot conflict with variable names, because none of them
438 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000439
440<p>Here is an example of LLVM code to multiply the integer variable
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000441 '<tt>%X</tt>' by 8:</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000442
Misha Brukman9d0919f2003-11-08 01:05:38 +0000443<p>The easy way:</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000444
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000445<pre class="doc_code">
Bill Wendling2f7a8b02007-05-29 09:04:49 +0000446%result = <a href="#i_mul">mul</a> i32 %X, 8
Chris Lattnere5d947b2004-12-09 16:36:40 +0000447</pre>
448
Misha Brukman9d0919f2003-11-08 01:05:38 +0000449<p>After strength reduction:</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000450
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000451<pre class="doc_code">
Bill Wendling2f7a8b02007-05-29 09:04:49 +0000452%result = <a href="#i_shl">shl</a> i32 %X, i8 3
Chris Lattnere5d947b2004-12-09 16:36:40 +0000453</pre>
454
Misha Brukman9d0919f2003-11-08 01:05:38 +0000455<p>And the hard way:</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000456
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000457<pre class="doc_code">
Gabor Greifec58f752009-10-28 13:05:07 +0000458%0 = <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
459%1 = <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
Bill Wendling2f7a8b02007-05-29 09:04:49 +0000460%result = <a href="#i_add">add</a> i32 %1, %1
Chris Lattnere5d947b2004-12-09 16:36:40 +0000461</pre>
462
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000463<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
464 lexical features of LLVM:</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000465
Chris Lattner00950542001-06-06 20:29:01 +0000466<ol>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000467 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000468 line.</li>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000469
470 <li>Unnamed temporaries are created when the result of a computation is not
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000471 assigned to a named value.</li>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000472
Misha Brukman9d0919f2003-11-08 01:05:38 +0000473 <li>Unnamed temporaries are numbered sequentially</li>
474</ol>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000475
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000476<p>It also shows a convention that we follow in this document. When
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000477 demonstrating instructions, we will follow an instruction with a comment that
478 defines the type and name of value produced. Comments are shown in italic
479 text.</p>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000480
Misha Brukman9d0919f2003-11-08 01:05:38 +0000481</div>
Chris Lattnerfa730212004-12-09 16:11:40 +0000482
483<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000484<h2><a name="highlevel">High Level Structure</a></h2>
Chris Lattnerfa730212004-12-09 16:11:40 +0000485<!-- *********************************************************************** -->
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000486<div>
Chris Lattnerfa730212004-12-09 16:11:40 +0000487<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000488<h3>
489 <a name="modulestructure">Module Structure</a>
490</h3>
Chris Lattnerfa730212004-12-09 16:11:40 +0000491
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000492<div>
Chris Lattnerfa730212004-12-09 16:11:40 +0000493
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000494<p>LLVM programs are composed of "Module"s, each of which is a translation unit
495 of the input programs. Each module consists of functions, global variables,
496 and symbol table entries. Modules may be combined together with the LLVM
497 linker, which merges function (and global variable) definitions, resolves
498 forward declarations, and merges symbol table entries. Here is an example of
499 the "hello world" module:</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000500
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000501<pre class="doc_code">
Chris Lattner63e4ccb2010-08-17 17:13:42 +0000502<i>; Declare the string constant as a global constant.</i>&nbsp;
Nick Lewyckydb9cd762011-01-29 01:09:53 +0000503<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 +0000504
Chris Lattner63e4ccb2010-08-17 17:13:42 +0000505<i>; External declaration of the puts function</i>&nbsp;
506<a href="#functionstructure">declare</a> i32 @puts(i8*) <i>; i32 (i8*)* </i>&nbsp;
Chris Lattnerfa730212004-12-09 16:11:40 +0000507
508<i>; Definition of main function</i>
Chris Lattner63e4ccb2010-08-17 17:13:42 +0000509define i32 @main() { <i>; i32()* </i>&nbsp;
510 <i>; Convert [13 x i8]* to i8 *...</i>&nbsp;
511 %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 +0000512
Chris Lattner63e4ccb2010-08-17 17:13:42 +0000513 <i>; Call puts function to write out the string to stdout.</i>&nbsp;
514 <a href="#i_call">call</a> i32 @puts(i8* %cast210) <i>; i32</i>&nbsp;
515 <a href="#i_ret">ret</a> i32 0&nbsp;
516}
Devang Patelcd1fd252010-01-11 19:35:55 +0000517
518<i>; Named metadata</i>
519!1 = metadata !{i32 41}
520!foo = !{!1, null}
Bill Wendling2f7a8b02007-05-29 09:04:49 +0000521</pre>
Chris Lattnerfa730212004-12-09 16:11:40 +0000522
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000523<p>This example is made up of a <a href="#globalvars">global variable</a> named
Devang Patelcd1fd252010-01-11 19:35:55 +0000524 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function,
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000525 a <a href="#functionstructure">function definition</a> for
Devang Patelcd1fd252010-01-11 19:35:55 +0000526 "<tt>main</tt>" and <a href="#namedmetadatastructure">named metadata</a>
527 "<tt>foo"</tt>.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000528
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000529<p>In general, a module is made up of a list of global values, where both
530 functions and global variables are global values. Global values are
531 represented by a pointer to a memory location (in this case, a pointer to an
532 array of char, and a pointer to a function), and have one of the
533 following <a href="#linkage">linkage types</a>.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000534
Chris Lattnere5d947b2004-12-09 16:36:40 +0000535</div>
536
537<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000538<h3>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000539 <a name="linkage">Linkage Types</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000540</h3>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000541
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000542<div>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000543
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000544<p>All Global Variables and Functions have one of the following types of
545 linkage:</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000546
547<dl>
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000548 <dt><tt><b><a name="linkage_private">private</a></b></tt></dt>
Bill Wendling5e721d72010-07-01 21:55:59 +0000549 <dd>Global values with "<tt>private</tt>" linkage are only directly accessible
550 by objects in the current module. In particular, linking code into a
551 module with an private global value may cause the private to be renamed as
552 necessary to avoid collisions. Because the symbol is private to the
553 module, all references can be updated. This doesn't show up in any symbol
554 table in the object file.</dd>
Rafael Espindolabb46f522009-01-15 20:18:42 +0000555
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000556 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt></dt>
Bill Wendling5e721d72010-07-01 21:55:59 +0000557 <dd>Similar to <tt>private</tt>, but the symbol is passed through the
558 assembler and evaluated by the linker. Unlike normal strong symbols, they
559 are removed by the linker from the final linked image (executable or
560 dynamic library).</dd>
561
562 <dt><tt><b><a name="linkage_linker_private_weak">linker_private_weak</a></b></tt></dt>
563 <dd>Similar to "<tt>linker_private</tt>", but the symbol is weak. Note that
564 <tt>linker_private_weak</tt> symbols are subject to coalescing by the
565 linker. The symbols are removed by the linker from the final linked image
566 (executable or dynamic library).</dd>
Bill Wendling3d10a5a2009-07-20 01:03:30 +0000567
Bill Wendling55ae5152010-08-20 22:05:50 +0000568 <dt><tt><b><a name="linkage_linker_private_weak_def_auto">linker_private_weak_def_auto</a></b></tt></dt>
569 <dd>Similar to "<tt>linker_private_weak</tt>", but it's known that the address
570 of the object is not taken. For instance, functions that had an inline
571 definition, but the compiler decided not to inline it. Note,
572 unlike <tt>linker_private</tt> and <tt>linker_private_weak</tt>,
573 <tt>linker_private_weak_def_auto</tt> may have only <tt>default</tt>
574 visibility. The symbols are removed by the linker from the final linked
575 image (executable or dynamic library).</dd>
576
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000577 <dt><tt><b><a name="linkage_internal">internal</a></b></tt></dt>
Bill Wendling07d31772010-06-29 22:34:52 +0000578 <dd>Similar to private, but the value shows as a local symbol
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000579 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
580 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000581
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000582 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt></dt>
Chris Lattner266c7bb2009-04-13 05:44:34 +0000583 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000584 into the object file corresponding to the LLVM module. They exist to
585 allow inlining and other optimizations to take place given knowledge of
586 the definition of the global, which is known to be somewhere outside the
587 module. Globals with <tt>available_externally</tt> linkage are allowed to
588 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
589 This linkage type is only allowed on definitions, not declarations.</dd>
Chris Lattner266c7bb2009-04-13 05:44:34 +0000590
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000591 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt></dt>
Chris Lattner4887bd82007-01-14 06:51:48 +0000592 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
Chris Lattner873187c2010-01-09 19:15:14 +0000593 the same name when linkage occurs. This can be used to implement
594 some forms of inline functions, templates, or other code which must be
595 generated in each translation unit that uses it, but where the body may
596 be overridden with a more definitive definition later. Unreferenced
597 <tt>linkonce</tt> globals are allowed to be discarded. Note that
598 <tt>linkonce</tt> linkage does not actually allow the optimizer to
599 inline the body of this function into callers because it doesn't know if
600 this definition of the function is the definitive definition within the
601 program or whether it will be overridden by a stronger definition.
602 To enable inlining and other optimizations, use "<tt>linkonce_odr</tt>"
603 linkage.</dd>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000604
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000605 <dt><tt><b><a name="linkage_weak">weak</a></b></tt></dt>
Chris Lattner26d054d2009-08-05 05:21:07 +0000606 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
607 <tt>linkonce</tt> linkage, except that unreferenced globals with
608 <tt>weak</tt> linkage may not be discarded. This is used for globals that
609 are declared "weak" in C source code.</dd>
610
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000611 <dt><tt><b><a name="linkage_common">common</a></b></tt></dt>
Chris Lattner26d054d2009-08-05 05:21:07 +0000612 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
613 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
614 global scope.
615 Symbols with "<tt>common</tt>" linkage are merged in the same way as
616 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
Chris Lattnercd81f5d2009-08-05 05:41:44 +0000617 <tt>common</tt> symbols may not have an explicit section,
Eric Christopher6c7e8a02009-12-05 02:46:03 +0000618 must have a zero initializer, and may not be marked '<a
Chris Lattnercd81f5d2009-08-05 05:41:44 +0000619 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
620 have common linkage.</dd>
Chris Lattner26d054d2009-08-05 05:21:07 +0000621
Chris Lattnere5d947b2004-12-09 16:36:40 +0000622
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000623 <dt><tt><b><a name="linkage_appending">appending</a></b></tt></dt>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000624 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000625 pointer to array type. When two global variables with appending linkage
626 are linked together, the two global arrays are appended together. This is
627 the LLVM, typesafe, equivalent of having the system linker append together
628 "sections" with identical names when .o files are linked.</dd>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000629
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000630 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000631 <dd>The semantics of this linkage follow the ELF object file model: the symbol
632 is weak until linked, if not linked, the symbol becomes null instead of
633 being an undefined reference.</dd>
Anton Korobeynikov7f705592007-01-12 19:20:47 +0000634
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000635 <dt><tt><b><a name="linkage_linkonce_odr">linkonce_odr</a></b></tt></dt>
636 <dt><tt><b><a name="linkage_weak_odr">weak_odr</a></b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000637 <dd>Some languages allow differing globals to be merged, such as two functions
638 with different semantics. Other languages, such as <tt>C++</tt>, ensure
Bill Wendling5e721d72010-07-01 21:55:59 +0000639 that only equivalent globals are ever merged (the "one definition rule"
640 &mdash; "ODR"). Such languages can use the <tt>linkonce_odr</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000641 and <tt>weak_odr</tt> linkage types to indicate that the global will only
642 be merged with equivalent globals. These linkage types are otherwise the
643 same as their non-<tt>odr</tt> versions.</dd>
Duncan Sands667d4b82009-03-07 15:45:40 +0000644
Chris Lattnerfa730212004-12-09 16:11:40 +0000645 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
Chris Lattnere5d947b2004-12-09 16:36:40 +0000646 <dd>If none of the above identifiers are used, the global is externally
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000647 visible, meaning that it participates in linkage and can be used to
648 resolve external symbol references.</dd>
Reid Spencerc8910842007-04-11 23:49:50 +0000649</dl>
Anton Korobeynikovb74ed072006-09-14 18:23:27 +0000650
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000651<p>The next two types of linkage are targeted for Microsoft Windows platform
652 only. They are designed to support importing (exporting) symbols from (to)
653 DLLs (Dynamic Link Libraries).</p>
Anton Korobeynikovb74ed072006-09-14 18:23:27 +0000654
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000655<dl>
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000656 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt></dt>
Anton Korobeynikovb74ed072006-09-14 18:23:27 +0000657 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000658 or variable via a global pointer to a pointer that is set up by the DLL
659 exporting the symbol. On Microsoft Windows targets, the pointer name is
660 formed by combining <code>__imp_</code> and the function or variable
661 name.</dd>
Anton Korobeynikovb74ed072006-09-14 18:23:27 +0000662
Bill Wendlingf82d40a2009-11-02 00:24:16 +0000663 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt></dt>
Anton Korobeynikovb74ed072006-09-14 18:23:27 +0000664 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000665 pointer to a pointer in a DLL, so that it can be referenced with the
666 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
667 name is formed by combining <code>__imp_</code> and the function or
668 variable name.</dd>
Chris Lattnerfa730212004-12-09 16:11:40 +0000669</dl>
670
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000671<p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
672 another module defined a "<tt>.LC0</tt>" variable and was linked with this
673 one, one of the two would be renamed, preventing a collision. Since
674 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
675 declarations), they are accessible outside of the current module.</p>
676
677<p>It is illegal for a function <i>declaration</i> to have any linkage type
678 other than "externally visible", <tt>dllimport</tt>
679 or <tt>extern_weak</tt>.</p>
680
Duncan Sands667d4b82009-03-07 15:45:40 +0000681<p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000682 or <tt>weak_odr</tt> linkages.</p>
683
Chris Lattnerfa730212004-12-09 16:11:40 +0000684</div>
685
686<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000687<h3>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000688 <a name="callingconv">Calling Conventions</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000689</h3>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000690
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000691<div>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000692
693<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000694 and <a href="#i_invoke">invokes</a> can all have an optional calling
695 convention specified for the call. The calling convention of any pair of
696 dynamic caller/callee must match, or the behavior of the program is
697 undefined. The following calling conventions are supported by LLVM, and more
698 may be added in the future:</p>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000699
700<dl>
701 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000702 <dd>This calling convention (the default if no other calling convention is
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000703 specified) matches the target C calling conventions. This calling
704 convention supports varargs function calls and tolerates some mismatch in
705 the declared prototype and implemented declaration of the function (as
706 does normal C).</dd>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000707
708 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000709 <dd>This calling convention attempts to make calls as fast as possible
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000710 (e.g. by passing things in registers). This calling convention allows the
711 target to use whatever tricks it wants to produce fast code for the
712 target, without having to conform to an externally specified ABI
Jeffrey Yasskin95fa80a2010-01-09 19:44:16 +0000713 (Application Binary Interface).
714 <a href="CodeGenerator.html#tailcallopt">Tail calls can only be optimized
Chris Lattner29689432010-03-11 00:22:57 +0000715 when this or the GHC convention is used.</a> This calling convention
716 does not support varargs and requires the prototype of all callees to
717 exactly match the prototype of the function definition.</dd>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000718
719 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000720 <dd>This calling convention attempts to make code in the caller as efficient
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000721 as possible under the assumption that the call is not commonly executed.
722 As such, these calls often preserve all registers so that the call does
723 not break any live ranges in the caller side. This calling convention
724 does not support varargs and requires the prototype of all callees to
725 exactly match the prototype of the function definition.</dd>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000726
Chris Lattner29689432010-03-11 00:22:57 +0000727 <dt><b>"<tt>cc <em>10</em></tt>" - GHC convention</b>:</dt>
728 <dd>This calling convention has been implemented specifically for use by the
729 <a href="http://www.haskell.org/ghc">Glasgow Haskell Compiler (GHC)</a>.
730 It passes everything in registers, going to extremes to achieve this by
731 disabling callee save registers. This calling convention should not be
732 used lightly but only for specific situations such as an alternative to
733 the <em>register pinning</em> performance technique often used when
734 implementing functional programming languages.At the moment only X86
735 supports this convention and it has the following limitations:
736 <ul>
737 <li>On <em>X86-32</em> only supports up to 4 bit type parameters. No
738 floating point types are supported.</li>
739 <li>On <em>X86-64</em> only supports up to 10 bit type parameters and
740 6 floating point parameters.</li>
741 </ul>
742 This calling convention supports
743 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> but
744 requires both the caller and callee are using it.
745 </dd>
746
Chris Lattnercfe6b372005-05-07 01:46:40 +0000747 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000748 <dd>Any calling convention may be specified by number, allowing
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000749 target-specific calling conventions to be used. Target specific calling
750 conventions start at 64.</dd>
Chris Lattnercfe6b372005-05-07 01:46:40 +0000751</dl>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000752
753<p>More calling conventions can be added/defined on an as-needed basis, to
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000754 support Pascal conventions or any other well-known target-independent
755 convention.</p>
Chris Lattnerbad10ee2005-05-06 22:57:40 +0000756
757</div>
758
759<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000760<h3>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000761 <a name="visibility">Visibility Styles</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000762</h3>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000763
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000764<div>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000765
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000766<p>All Global Variables and Functions have one of the following visibility
767 styles:</p>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000768
769<dl>
770 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
Chris Lattnerd3eda892008-08-05 18:29:16 +0000771 <dd>On targets that use the ELF object file format, default visibility means
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000772 that the declaration is visible to other modules and, in shared libraries,
773 means that the declared entity may be overridden. On Darwin, default
774 visibility means that the declaration is visible to other modules. Default
775 visibility corresponds to "external linkage" in the language.</dd>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000776
777 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000778 <dd>Two declarations of an object with hidden visibility refer to the same
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000779 object if they are in the same shared object. Usually, hidden visibility
780 indicates that the symbol will not be placed into the dynamic symbol
781 table, so no other module (executable or shared library) can reference it
782 directly.</dd>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000783
Anton Korobeynikov6f9896f2007-04-29 18:35:00 +0000784 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
Anton Korobeynikov6f9896f2007-04-29 18:35:00 +0000785 <dd>On ELF, protected visibility indicates that the symbol will be placed in
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000786 the dynamic symbol table, but that references within the defining module
787 will bind to the local symbol. That is, the symbol cannot be overridden by
788 another module.</dd>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000789</dl>
790
791</div>
792
793<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000794<h3>
Chris Lattnere7886e42009-01-11 20:53:49 +0000795 <a name="namedtypes">Named Types</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000796</h3>
Chris Lattnere7886e42009-01-11 20:53:49 +0000797
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000798<div>
Chris Lattnere7886e42009-01-11 20:53:49 +0000799
800<p>LLVM IR allows you to specify name aliases for certain types. This can make
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000801 it easier to read the IR and make the IR more condensed (particularly when
802 recursive types are involved). An example of a name specification is:</p>
Chris Lattnere7886e42009-01-11 20:53:49 +0000803
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000804<pre class="doc_code">
Chris Lattnere7886e42009-01-11 20:53:49 +0000805%mytype = type { %mytype*, i32 }
806</pre>
Chris Lattnere7886e42009-01-11 20:53:49 +0000807
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000808<p>You may give a name to any <a href="#typesystem">type</a> except
Chris Lattnerdc65f222010-08-17 23:26:04 +0000809 "<a href="#t_void">void</a>". Type name aliases may be used anywhere a type
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000810 is expected with the syntax "%mytype".</p>
Chris Lattnere7886e42009-01-11 20:53:49 +0000811
812<p>Note that type names are aliases for the structural type that they indicate,
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000813 and that you can therefore specify multiple names for the same type. This
814 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
815 uses structural typing, the name is not part of the type. When printing out
816 LLVM IR, the printer will pick <em>one name</em> to render all types of a
817 particular shape. This means that if you have code where two different
818 source types end up having the same LLVM type, that the dumper will sometimes
819 print the "wrong" or unexpected type. This is an important design point and
820 isn't going to change.</p>
Chris Lattnere7886e42009-01-11 20:53:49 +0000821
822</div>
823
Chris Lattnere7886e42009-01-11 20:53:49 +0000824<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000825<h3>
Chris Lattnerfa730212004-12-09 16:11:40 +0000826 <a name="globalvars">Global Variables</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000827</h3>
Chris Lattnerfa730212004-12-09 16:11:40 +0000828
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000829<div>
Chris Lattnerfa730212004-12-09 16:11:40 +0000830
Chris Lattner3689a342005-02-12 19:30:21 +0000831<p>Global variables define regions of memory allocated at compilation time
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000832 instead of run-time. Global variables may optionally be initialized, may
833 have an explicit section to be placed in, and may have an optional explicit
834 alignment specified. A variable may be defined as "thread_local", which
835 means that it will not be shared by threads (each thread will have a
836 separated copy of the variable). A variable may be defined as a global
837 "constant," which indicates that the contents of the variable
838 will <b>never</b> be modified (enabling better optimization, allowing the
839 global data to be placed in the read-only section of an executable, etc).
840 Note that variables that need runtime initialization cannot be marked
841 "constant" as there is a store to the variable.</p>
Chris Lattner3689a342005-02-12 19:30:21 +0000842
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000843<p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
844 constant, even if the final definition of the global is not. This capability
845 can be used to enable slightly better optimization of the program, but
846 requires the language definition to guarantee that optimizations based on the
847 'constantness' are valid for the translation units that do not include the
848 definition.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000849
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000850<p>As SSA values, global variables define pointer values that are in scope
851 (i.e. they dominate) all basic blocks in the program. Global variables
852 always define a pointer to their "content" type because they describe a
853 region of memory, and all memory objects in LLVM are accessed through
854 pointers.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000855
Rafael Espindolabea46262011-01-08 16:42:36 +0000856<p>Global variables can be marked with <tt>unnamed_addr</tt> which indicates
857 that the address is not significant, only the content. Constants marked
Rafael Espindolaa5eaa862011-01-15 08:20:57 +0000858 like this can be merged with other constants if they have the same
859 initializer. Note that a constant with significant address <em>can</em>
860 be merged with a <tt>unnamed_addr</tt> constant, the result being a
861 constant whose address is significant.</p>
Rafael Espindolabea46262011-01-08 16:42:36 +0000862
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000863<p>A global variable may be declared to reside in a target-specific numbered
864 address space. For targets that support them, address spaces may affect how
865 optimizations are performed and/or what target instructions are used to
866 access the variable. The default address space is zero. The address space
867 qualifier must precede any other attributes.</p>
Christopher Lamb284d9922007-12-11 09:31:00 +0000868
Chris Lattner88f6c462005-11-12 00:45:07 +0000869<p>LLVM allows an explicit section to be specified for globals. If the target
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000870 supports it, it will emit globals to the section specified.</p>
Chris Lattner88f6c462005-11-12 00:45:07 +0000871
Chris Lattnerce99fa92010-04-28 00:13:42 +0000872<p>An explicit alignment may be specified for a global, which must be a power
873 of 2. If not present, or if the alignment is set to zero, the alignment of
874 the global is set by the target to whatever it feels convenient. If an
875 explicit alignment is specified, the global is forced to have exactly that
Chris Lattner2d4b8ee2010-04-28 00:31:12 +0000876 alignment. Targets and optimizers are not allowed to over-align the global
877 if the global has an assigned section. In this case, the extra alignment
878 could be observable: for example, code could assume that the globals are
879 densely packed in their section and try to iterate over them as an array,
880 alignment padding would break this iteration.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +0000881
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000882<p>For example, the following defines a global in a numbered address space with
883 an initializer, section, and alignment:</p>
Chris Lattner68027ea2007-01-14 00:27:09 +0000884
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000885<pre class="doc_code">
Dan Gohman398873c2009-01-11 00:40:00 +0000886@G = addrspace(5) constant float 1.0, section "foo", align 4
Chris Lattner68027ea2007-01-14 00:27:09 +0000887</pre>
888
Chris Lattnerfa730212004-12-09 16:11:40 +0000889</div>
890
891
892<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000893<h3>
Chris Lattnerfa730212004-12-09 16:11:40 +0000894 <a name="functionstructure">Functions</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000895</h3>
Chris Lattnerfa730212004-12-09 16:11:40 +0000896
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000897<div>
Chris Lattnerfa730212004-12-09 16:11:40 +0000898
Dan Gohmanb55a1ee2010-03-01 17:41:39 +0000899<p>LLVM function definitions consist of the "<tt>define</tt>" keyword, an
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000900 optional <a href="#linkage">linkage type</a>, an optional
901 <a href="#visibility">visibility style</a>, an optional
Rafael Espindolabea46262011-01-08 16:42:36 +0000902 <a href="#callingconv">calling convention</a>,
903 an optional <tt>unnamed_addr</tt> attribute, a return type, an optional
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000904 <a href="#paramattrs">parameter attribute</a> for the return type, a function
905 name, a (possibly empty) argument list (each with optional
906 <a href="#paramattrs">parameter attributes</a>), optional
907 <a href="#fnattrs">function attributes</a>, an optional section, an optional
908 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
909 curly brace, a list of basic blocks, and a closing curly brace.</p>
Anton Korobeynikov8cea37b2007-01-23 12:35:46 +0000910
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000911<p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
912 optional <a href="#linkage">linkage type</a>, an optional
Eric Christopher6c7e8a02009-12-05 02:46:03 +0000913 <a href="#visibility">visibility style</a>, an optional
Rafael Espindolabea46262011-01-08 16:42:36 +0000914 <a href="#callingconv">calling convention</a>,
915 an optional <tt>unnamed_addr</tt> attribute, a return type, an optional
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000916 <a href="#paramattrs">parameter attribute</a> for the return type, a function
917 name, a possibly empty list of arguments, an optional alignment, and an
918 optional <a href="#gc">garbage collector name</a>.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000919
Chris Lattnerd3eda892008-08-05 18:29:16 +0000920<p>A function definition contains a list of basic blocks, forming the CFG
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000921 (Control Flow Graph) for the function. Each basic block may optionally start
922 with a label (giving the basic block a symbol table entry), contains a list
923 of instructions, and ends with a <a href="#terminators">terminator</a>
924 instruction (such as a branch or function return).</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000925
Chris Lattner4a3c9012007-06-08 16:52:14 +0000926<p>The first basic block in a function is special in two ways: it is immediately
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000927 executed on entrance to the function, and it is not allowed to have
928 predecessor basic blocks (i.e. there can not be any branches to the entry
929 block of a function). Because the block can have no predecessors, it also
930 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +0000931
Chris Lattner88f6c462005-11-12 00:45:07 +0000932<p>LLVM allows an explicit section to be specified for functions. If the target
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000933 supports it, it will emit functions to the section specified.</p>
Chris Lattner88f6c462005-11-12 00:45:07 +0000934
Chris Lattner2cbdc452005-11-06 08:02:57 +0000935<p>An explicit alignment may be specified for a function. If not present, or if
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000936 the alignment is set to zero, the alignment of the function is set by the
937 target to whatever it feels convenient. If an explicit alignment is
938 specified, the function is forced to have at least that much alignment. All
939 alignments must be a power of 2.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +0000940
Rafael Espindolabea46262011-01-08 16:42:36 +0000941<p>If the <tt>unnamed_addr</tt> attribute is given, the address is know to not
942 be significant and two identical functions can be merged</p>.
943
Bill Wendlingc39e3e02009-07-20 02:39:26 +0000944<h5>Syntax:</h5>
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000945<pre class="doc_code">
Chris Lattner50ad45c2008-10-13 16:55:18 +0000946define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000947 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
948 &lt;ResultType&gt; @&lt;FunctionName&gt; ([argument list])
949 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
950 [<a href="#gc">gc</a>] { ... }
951</pre>
Devang Patel307e8ab2008-10-07 17:48:33 +0000952
Chris Lattnerfa730212004-12-09 16:11:40 +0000953</div>
954
Anton Korobeynikov8b0a8c82007-04-25 14:27:10 +0000955<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000956<h3>
Anton Korobeynikov8b0a8c82007-04-25 14:27:10 +0000957 <a name="aliasstructure">Aliases</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000958</h3>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000959
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000960<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +0000961
962<p>Aliases act as "second name" for the aliasee value (which can be either
963 function, global variable, another alias or bitcast of global value). Aliases
964 may have an optional <a href="#linkage">linkage type</a>, and an
965 optional <a href="#visibility">visibility style</a>.</p>
Anton Korobeynikov8b0a8c82007-04-25 14:27:10 +0000966
Bill Wendlingc39e3e02009-07-20 02:39:26 +0000967<h5>Syntax:</h5>
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000968<pre class="doc_code">
Duncan Sands0b23ac12008-09-12 20:48:21 +0000969@&lt;Name&gt; = alias [Linkage] [Visibility] &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
Bill Wendlingaac388b2007-05-29 09:42:13 +0000970</pre>
Anton Korobeynikov8b0a8c82007-04-25 14:27:10 +0000971
972</div>
973
Chris Lattner4e9aba72006-01-23 23:23:47 +0000974<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000975<h3>
Devang Patelcd1fd252010-01-11 19:35:55 +0000976 <a name="namedmetadatastructure">Named Metadata</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000977</h3>
Devang Patelcd1fd252010-01-11 19:35:55 +0000978
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +0000979<div>
Devang Patelcd1fd252010-01-11 19:35:55 +0000980
Chris Lattnere6a5ddd2010-01-15 21:50:19 +0000981<p>Named metadata is a collection of metadata. <a href="#metadata">Metadata
Dan Gohman872814a2010-07-21 18:54:18 +0000982 nodes</a> (but not metadata strings) are the only valid operands for
Chris Lattnere6a5ddd2010-01-15 21:50:19 +0000983 a named metadata.</p>
Devang Patelcd1fd252010-01-11 19:35:55 +0000984
985<h5>Syntax:</h5>
Benjamin Kramer26fe25f2010-07-13 12:26:09 +0000986<pre class="doc_code">
Dan Gohman872814a2010-07-21 18:54:18 +0000987; Some unnamed metadata nodes, which are referenced by the named metadata.
988!0 = metadata !{metadata !"zero"}
Devang Patelcd1fd252010-01-11 19:35:55 +0000989!1 = metadata !{metadata !"one"}
Dan Gohman872814a2010-07-21 18:54:18 +0000990!2 = metadata !{metadata !"two"}
Dan Gohman1005bc52010-07-13 19:48:13 +0000991; A named metadata.
Dan Gohman872814a2010-07-21 18:54:18 +0000992!name = !{!0, !1, !2}
Devang Patelcd1fd252010-01-11 19:35:55 +0000993</pre>
Devang Patelcd1fd252010-01-11 19:35:55 +0000994
995</div>
996
997<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +0000998<h3>
999 <a name="paramattrs">Parameter Attributes</a>
1000</h3>
Reid Spencerca86e162006-12-31 07:07:53 +00001001
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001002<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001003
1004<p>The return type and each parameter of a function type may have a set of
1005 <i>parameter attributes</i> associated with them. Parameter attributes are
1006 used to communicate additional information about the result or parameters of
1007 a function. Parameter attributes are considered to be part of the function,
1008 not of the function type, so functions with different parameter attributes
1009 can have the same function type.</p>
1010
1011<p>Parameter attributes are simple keywords that follow the type specified. If
1012 multiple parameter attributes are needed, they are space separated. For
1013 example:</p>
Bill Wendling2f7a8b02007-05-29 09:04:49 +00001014
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00001015<pre class="doc_code">
Nick Lewyckyb6a7d252009-02-15 23:06:14 +00001016declare i32 @printf(i8* noalias nocapture, ...)
Chris Lattner66d922c2008-10-04 18:33:34 +00001017declare i32 @atoi(i8 zeroext)
1018declare signext i8 @returns_signed_char()
Bill Wendling2f7a8b02007-05-29 09:04:49 +00001019</pre>
Bill Wendling2f7a8b02007-05-29 09:04:49 +00001020
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001021<p>Note that any attributes for the function result (<tt>nounwind</tt>,
1022 <tt>readonly</tt>) come immediately after the argument list.</p>
Reid Spencerca86e162006-12-31 07:07:53 +00001023
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001024<p>Currently, only the following parameter attributes are defined:</p>
Chris Lattner47507de2008-01-11 06:20:47 +00001025
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001026<dl>
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001027 <dt><tt><b>zeroext</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001028 <dd>This indicates to the code generator that the parameter or return value
Cameron Zwarichebe81732011-03-16 22:20:18 +00001029 should be zero-extended to the extent required by the target's ABI (which
1030 is usually 32-bits, but is 8-bits for a i1 on x86-64) by the caller (for a
1031 parameter) or the callee (for a return value).</dd>
Chris Lattner47507de2008-01-11 06:20:47 +00001032
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001033 <dt><tt><b>signext</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001034 <dd>This indicates to the code generator that the parameter or return value
Cameron Zwarich9e69ff92011-03-17 14:21:58 +00001035 should be sign-extended to the extent required by the target's ABI (which
1036 is usually 32-bits) by the caller (for a parameter) or the callee (for a
1037 return value).</dd>
Chris Lattner47507de2008-01-11 06:20:47 +00001038
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001039 <dt><tt><b>inreg</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001040 <dd>This indicates that this parameter or return value should be treated in a
1041 special target-dependent fashion during while emitting code for a function
1042 call or return (usually, by putting it in a register as opposed to memory,
1043 though some targets use it to distinguish between two different kinds of
1044 registers). Use of this attribute is target-specific.</dd>
Chris Lattner47507de2008-01-11 06:20:47 +00001045
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001046 <dt><tt><b><a name="byval">byval</a></b></tt></dt>
Chris Lattnera6fd81d2010-11-20 23:49:06 +00001047 <dd><p>This indicates that the pointer parameter should really be passed by
1048 value to the function. The attribute implies that a hidden copy of the
1049 pointee
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001050 is made between the caller and the callee, so the callee is unable to
1051 modify the value in the callee. This attribute is only valid on LLVM
1052 pointer arguments. It is generally used to pass structs and arrays by
1053 value, but is also valid on pointers to scalars. The copy is considered
1054 to belong to the caller not the callee (for example,
1055 <tt><a href="#readonly">readonly</a></tt> functions should not write to
1056 <tt>byval</tt> parameters). This is not a valid attribute for return
Chris Lattnera6fd81d2010-11-20 23:49:06 +00001057 values.</p>
1058
1059 <p>The byval attribute also supports specifying an alignment with
1060 the align attribute. It indicates the alignment of the stack slot to
1061 form and the known alignment of the pointer specified to the call site. If
1062 the alignment is not specified, then the code generator makes a
1063 target-specific assumption.</p></dd>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001064
Dan Gohmanff235352010-07-02 23:18:08 +00001065 <dt><tt><b><a name="sret">sret</a></b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001066 <dd>This indicates that the pointer parameter specifies the address of a
1067 structure that is the return value of the function in the source program.
1068 This pointer must be guaranteed by the caller to be valid: loads and
1069 stores to the structure may be assumed by the callee to not to trap. This
1070 may only be applied to the first parameter. This is not a valid attribute
1071 for return values. </dd>
1072
Dan Gohmanff235352010-07-02 23:18:08 +00001073 <dt><tt><b><a name="noalias">noalias</a></b></tt></dt>
Dan Gohman1e109622010-07-02 18:41:32 +00001074 <dd>This indicates that pointer values
1075 <a href="#pointeraliasing"><i>based</i></a> on the argument or return
Dan Gohmanefca7f92010-07-02 23:46:54 +00001076 value do not alias pointer values which are not <i>based</i> on it,
1077 ignoring certain "irrelevant" dependencies.
1078 For a call to the parent function, dependencies between memory
1079 references from before or after the call and from those during the call
1080 are "irrelevant" to the <tt>noalias</tt> keyword for the arguments and
1081 return value used in that call.
Dan Gohman1e109622010-07-02 18:41:32 +00001082 The caller shares the responsibility with the callee for ensuring that
1083 these requirements are met.
1084 For further details, please see the discussion of the NoAlias response in
Dan Gohmanff70fe42010-07-06 15:26:33 +00001085 <a href="AliasAnalysis.html#MustMayNo">alias analysis</a>.<br>
1086<br>
John McCall191d4ee2010-07-06 21:07:14 +00001087 Note that this definition of <tt>noalias</tt> is intentionally
1088 similar to the definition of <tt>restrict</tt> in C99 for function
Chris Lattner211244a2010-07-06 20:51:35 +00001089 arguments, though it is slightly weaker.
Dan Gohmanff70fe42010-07-06 15:26:33 +00001090<br>
1091 For function return values, C99's <tt>restrict</tt> is not meaningful,
1092 while LLVM's <tt>noalias</tt> is.
1093 </dd>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001094
Dan Gohmanff235352010-07-02 23:18:08 +00001095 <dt><tt><b><a name="nocapture">nocapture</a></b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001096 <dd>This indicates that the callee does not make any copies of the pointer
1097 that outlive the callee itself. This is not a valid attribute for return
1098 values.</dd>
1099
Dan Gohmanff235352010-07-02 23:18:08 +00001100 <dt><tt><b><a name="nest">nest</a></b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001101 <dd>This indicates that the pointer parameter can be excised using the
1102 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
1103 attribute for return values.</dd>
1104</dl>
Reid Spencerca86e162006-12-31 07:07:53 +00001105
Reid Spencerca86e162006-12-31 07:07:53 +00001106</div>
1107
1108<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001109<h3>
Gordon Henriksen80a75bf2007-12-10 03:18:06 +00001110 <a name="gc">Garbage Collector Names</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001111</h3>
Gordon Henriksen80a75bf2007-12-10 03:18:06 +00001112
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001113<div>
Gordon Henriksen80a75bf2007-12-10 03:18:06 +00001114
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001115<p>Each function may specify a garbage collector name, which is simply a
1116 string:</p>
1117
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00001118<pre class="doc_code">
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001119define void @f() gc "name" { ... }
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001120</pre>
Gordon Henriksen80a75bf2007-12-10 03:18:06 +00001121
1122<p>The compiler declares the supported values of <i>name</i>. Specifying a
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001123 collector which will cause the compiler to alter its output in order to
1124 support the named garbage collection algorithm.</p>
1125
Gordon Henriksen80a75bf2007-12-10 03:18:06 +00001126</div>
1127
1128<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001129<h3>
Devang Patel2c9c3e72008-09-26 23:51:19 +00001130 <a name="fnattrs">Function Attributes</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001131</h3>
Devang Patelf8b94812008-09-04 23:05:13 +00001132
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001133<div>
Devang Patel2c9c3e72008-09-26 23:51:19 +00001134
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001135<p>Function attributes are set to communicate additional information about a
1136 function. Function attributes are considered to be part of the function, not
1137 of the function type, so functions with different parameter attributes can
1138 have the same function type.</p>
Devang Patel2c9c3e72008-09-26 23:51:19 +00001139
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001140<p>Function attributes are simple keywords that follow the type specified. If
1141 multiple attributes are needed, they are space separated. For example:</p>
Devang Patelf8b94812008-09-04 23:05:13 +00001142
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00001143<pre class="doc_code">
Devang Patel2c9c3e72008-09-26 23:51:19 +00001144define void @f() noinline { ... }
1145define void @f() alwaysinline { ... }
1146define void @f() alwaysinline optsize { ... }
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001147define void @f() optsize { ... }
Bill Wendlinge36dccc2008-09-07 10:26:33 +00001148</pre>
Devang Patelf8b94812008-09-04 23:05:13 +00001149
Bill Wendlinge36dccc2008-09-07 10:26:33 +00001150<dl>
Charles Davis1e063d12010-02-12 00:31:15 +00001151 <dt><tt><b>alignstack(&lt;<em>n</em>&gt;)</b></tt></dt>
1152 <dd>This attribute indicates that, when emitting the prologue and epilogue,
1153 the backend should forcibly align the stack pointer. Specify the
1154 desired alignment, which must be a power of two, in parentheses.
1155
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001156 <dt><tt><b>alwaysinline</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001157 <dd>This attribute indicates that the inliner should attempt to inline this
1158 function into callers whenever possible, ignoring any active inlining size
1159 threshold for this caller.</dd>
Bill Wendlinge36dccc2008-09-07 10:26:33 +00001160
Charles Davis970bfcc2010-10-25 15:37:09 +00001161 <dt><tt><b>hotpatch</b></tt></dt>
Charles Davis6f12e292010-10-25 16:29:03 +00001162 <dd>This attribute indicates that the function should be 'hotpatchable',
Charles Davis0076d202010-10-25 19:07:39 +00001163 meaning the function can be patched and/or hooked even while it is
1164 loaded into memory. On x86, the function prologue will be preceded
1165 by six bytes of padding and will begin with a two-byte instruction.
1166 Most of the functions in the Windows system DLLs in Windows XP SP2 or
1167 higher were compiled in this fashion.</dd>
Charles Davis970bfcc2010-10-25 15:37:09 +00001168
Dan Gohman129bd562011-06-16 16:03:13 +00001169 <dt><tt><b>nonlazybind</b></tt></dt>
1170 <dd>This attribute suppresses lazy symbol binding for the function. This
1171 may make calls to the function faster, at the cost of extra program
1172 startup time if the function is not called during program startup.</dd>
1173
Jakob Stoklund Olesen570a4a52010-02-06 01:16:28 +00001174 <dt><tt><b>inlinehint</b></tt></dt>
1175 <dd>This attribute indicates that the source code contained a hint that inlining
1176 this function is desirable (such as the "inline" keyword in C/C++). It
1177 is just a hint; it imposes no requirements on the inliner.</dd>
1178
Nick Lewycky76ec37a2010-07-06 18:24:09 +00001179 <dt><tt><b>naked</b></tt></dt>
1180 <dd>This attribute disables prologue / epilogue emission for the function.
1181 This can have very system-specific consequences.</dd>
1182
1183 <dt><tt><b>noimplicitfloat</b></tt></dt>
1184 <dd>This attributes disables implicit floating point instructions.</dd>
1185
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001186 <dt><tt><b>noinline</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001187 <dd>This attribute indicates that the inliner should never inline this
1188 function in any situation. This attribute may not be used together with
1189 the <tt>alwaysinline</tt> attribute.</dd>
Devang Patel2c9c3e72008-09-26 23:51:19 +00001190
Nick Lewycky76ec37a2010-07-06 18:24:09 +00001191 <dt><tt><b>noredzone</b></tt></dt>
1192 <dd>This attribute indicates that the code generator should not use a red
1193 zone, even if the target-specific ABI normally permits it.</dd>
Devang Patel2c9c3e72008-09-26 23:51:19 +00001194
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001195 <dt><tt><b>noreturn</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001196 <dd>This function attribute indicates that the function never returns
1197 normally. This produces undefined behavior at runtime if the function
1198 ever does dynamically return.</dd>
Bill Wendling31359ba2008-11-13 01:02:51 +00001199
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001200 <dt><tt><b>nounwind</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001201 <dd>This function attribute indicates that the function never returns with an
1202 unwind or exceptional control flow. If the function does unwind, its
1203 runtime behavior is undefined.</dd>
Bill Wendlingfbaa7ed2008-11-26 19:07:40 +00001204
Nick Lewycky76ec37a2010-07-06 18:24:09 +00001205 <dt><tt><b>optsize</b></tt></dt>
1206 <dd>This attribute suggests that optimization passes and code generator passes
1207 make choices that keep the code size of this function low, and otherwise
1208 do optimizations specifically to reduce code size.</dd>
1209
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001210 <dt><tt><b>readnone</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001211 <dd>This attribute indicates that the function computes its result (or decides
1212 to unwind an exception) based strictly on its arguments, without
1213 dereferencing any pointer arguments or otherwise accessing any mutable
1214 state (e.g. memory, control registers, etc) visible to caller functions.
1215 It does not write through any pointer arguments
1216 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1217 changes any state visible to callers. This means that it cannot unwind
1218 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1219 could use the <tt>unwind</tt> instruction.</dd>
Devang Patel5d96fda2009-06-12 19:45:19 +00001220
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001221 <dt><tt><b><a name="readonly">readonly</a></b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001222 <dd>This attribute indicates that the function does not write through any
1223 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1224 arguments) or otherwise modify any state (e.g. memory, control registers,
1225 etc) visible to caller functions. It may dereference pointer arguments
1226 and read state that may be set in the caller. A readonly function always
1227 returns the same value (or unwinds an exception identically) when called
1228 with the same set of arguments and global state. It cannot unwind an
1229 exception by calling the <tt>C++</tt> exception throwing methods, but may
1230 use the <tt>unwind</tt> instruction.</dd>
Anton Korobeynikovc5ec8a72009-07-17 18:07:26 +00001231
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001232 <dt><tt><b><a name="ssp">ssp</a></b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001233 <dd>This attribute indicates that the function should emit a stack smashing
1234 protector. It is in the form of a "canary"&mdash;a random value placed on
1235 the stack before the local variables that's checked upon return from the
1236 function to see if it has been overwritten. A heuristic is used to
1237 determine if a function needs stack protectors or not.<br>
1238<br>
1239 If a function that has an <tt>ssp</tt> attribute is inlined into a
1240 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1241 function will have an <tt>ssp</tt> attribute.</dd>
1242
Bill Wendlingf82d40a2009-11-02 00:24:16 +00001243 <dt><tt><b>sspreq</b></tt></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001244 <dd>This attribute indicates that the function should <em>always</em> emit a
1245 stack smashing protector. This overrides
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001246 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1247<br>
1248 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1249 function that doesn't have an <tt>sspreq</tt> attribute or which has
1250 an <tt>ssp</tt> attribute, then the resulting function will have
1251 an <tt>sspreq</tt> attribute.</dd>
Rafael Espindolafbff0ec2011-07-25 15:27:59 +00001252
1253 <dt><tt><b><a name="uwtable">uwtable</a></b></tt></dt>
1254 <dd>This attribute indicates that the ABI being targeted requires that
1255 an unwind table entry be produce for this function even if we can
1256 show that no exceptions passes by it. This is normally the case for
1257 the ELF x86-64 abi, but it can be disabled for some compilation
1258 units.</dd>
1259
Bill Wendlinge36dccc2008-09-07 10:26:33 +00001260</dl>
1261
Devang Patelf8b94812008-09-04 23:05:13 +00001262</div>
1263
1264<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001265<h3>
Chris Lattner1eeeb0c2006-04-08 04:40:53 +00001266 <a name="moduleasm">Module-Level Inline Assembly</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001267</h3>
Chris Lattner4e9aba72006-01-23 23:23:47 +00001268
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001269<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001270
1271<p>Modules may contain "module-level inline asm" blocks, which corresponds to
1272 the GCC "file scope inline asm" blocks. These blocks are internally
1273 concatenated by LLVM and treated as a single unit, but may be separated in
1274 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
Chris Lattner4e9aba72006-01-23 23:23:47 +00001275
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00001276<pre class="doc_code">
Bill Wendling2f7a8b02007-05-29 09:04:49 +00001277module asm "inline asm code goes here"
1278module asm "more can go here"
1279</pre>
Chris Lattner4e9aba72006-01-23 23:23:47 +00001280
1281<p>The strings can contain any character by escaping non-printable characters.
1282 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001283 for the number.</p>
Chris Lattner4e9aba72006-01-23 23:23:47 +00001284
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001285<p>The inline asm code is simply printed to the machine code .s file when
1286 assembly code is generated.</p>
1287
Chris Lattner4e9aba72006-01-23 23:23:47 +00001288</div>
Chris Lattnerfa730212004-12-09 16:11:40 +00001289
Reid Spencerde151942007-02-19 23:54:10 +00001290<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001291<h3>
Reid Spencerde151942007-02-19 23:54:10 +00001292 <a name="datalayout">Data Layout</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001293</h3>
Reid Spencerde151942007-02-19 23:54:10 +00001294
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001295<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001296
Reid Spencerde151942007-02-19 23:54:10 +00001297<p>A module may specify a target specific data layout string that specifies how
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001298 data is to be laid out in memory. The syntax for the data layout is
1299 simply:</p>
1300
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00001301<pre class="doc_code">
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001302target datalayout = "<i>layout specification</i>"
1303</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001304
1305<p>The <i>layout specification</i> consists of a list of specifications
1306 separated by the minus sign character ('-'). Each specification starts with
1307 a letter and may include other information after the letter to define some
1308 aspect of the data layout. The specifications accepted are as follows:</p>
1309
Reid Spencerde151942007-02-19 23:54:10 +00001310<dl>
1311 <dt><tt>E</tt></dt>
1312 <dd>Specifies that the target lays out data in big-endian form. That is, the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001313 bits with the most significance have the lowest address location.</dd>
1314
Reid Spencerde151942007-02-19 23:54:10 +00001315 <dt><tt>e</tt></dt>
Chris Lattnerd3eda892008-08-05 18:29:16 +00001316 <dd>Specifies that the target lays out data in little-endian form. That is,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001317 the bits with the least significance have the lowest address
1318 location.</dd>
1319
Reid Spencerde151942007-02-19 23:54:10 +00001320 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00001321 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001322 <i>preferred</i> alignments. All sizes are in bits. Specifying
1323 the <i>pref</i> alignment is optional. If omitted, the
1324 preceding <tt>:</tt> should be omitted too.</dd>
1325
Reid Spencerde151942007-02-19 23:54:10 +00001326 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1327 <dd>This specifies the alignment for an integer type of a given bit
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001328 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1329
Reid Spencerde151942007-02-19 23:54:10 +00001330 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00001331 <dd>This specifies the alignment for a vector type of a given bit
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001332 <i>size</i>.</dd>
1333
Reid Spencerde151942007-02-19 23:54:10 +00001334 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00001335 <dd>This specifies the alignment for a floating point type of a given bit
Dale Johannesen9d8d2212010-05-28 18:54:47 +00001336 <i>size</i>. Only values of <i>size</i> that are supported by the target
1337 will work. 32 (float) and 64 (double) are supported on all targets;
1338 80 or 128 (different flavors of long double) are also supported on some
1339 targets.
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001340
Reid Spencerde151942007-02-19 23:54:10 +00001341 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1342 <dd>This specifies the alignment for an aggregate type of a given bit
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001343 <i>size</i>.</dd>
1344
Daniel Dunbar87bde0b2009-06-08 22:17:53 +00001345 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1346 <dd>This specifies the alignment for a stack object of a given bit
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001347 <i>size</i>.</dd>
Chris Lattnere82bdc42009-11-07 09:35:34 +00001348
1349 <dt><tt>n<i>size1</i>:<i>size2</i>:<i>size3</i>...</tt></dt>
1350 <dd>This specifies a set of native integer widths for the target CPU
1351 in bits. For example, it might contain "n32" for 32-bit PowerPC,
1352 "n32:64" for PowerPC 64, or "n8:16:32:64" for X86-64. Elements of
Eric Christopher6c7e8a02009-12-05 02:46:03 +00001353 this set are considered to support most general arithmetic
Chris Lattnere82bdc42009-11-07 09:35:34 +00001354 operations efficiently.</dd>
Reid Spencerde151942007-02-19 23:54:10 +00001355</dl>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001356
Reid Spencerde151942007-02-19 23:54:10 +00001357<p>When constructing the data layout for a given target, LLVM starts with a
Dan Gohman1c70c002010-04-28 00:36:01 +00001358 default set of specifications which are then (possibly) overridden by the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001359 specifications in the <tt>datalayout</tt> keyword. The default specifications
1360 are given in this list:</p>
1361
Reid Spencerde151942007-02-19 23:54:10 +00001362<ul>
1363 <li><tt>E</tt> - big endian</li>
Dan Gohmanfdf2e8c2010-02-23 02:44:03 +00001364 <li><tt>p:64:64:64</tt> - 64-bit pointers with 64-bit alignment</li>
Reid Spencerde151942007-02-19 23:54:10 +00001365 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1366 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1367 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1368 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
Chris Lattnerd3eda892008-08-05 18:29:16 +00001369 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
Reid Spencerde151942007-02-19 23:54:10 +00001370 alignment of 64-bits</li>
1371 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1372 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1373 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1374 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1375 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
Daniel Dunbar87bde0b2009-06-08 22:17:53 +00001376 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
Reid Spencerde151942007-02-19 23:54:10 +00001377</ul>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001378
1379<p>When LLVM is determining the alignment for a given type, it uses the
1380 following rules:</p>
1381
Reid Spencerde151942007-02-19 23:54:10 +00001382<ol>
1383 <li>If the type sought is an exact match for one of the specifications, that
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001384 specification is used.</li>
1385
Reid Spencerde151942007-02-19 23:54:10 +00001386 <li>If no match is found, and the type sought is an integer type, then the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001387 smallest integer type that is larger than the bitwidth of the sought type
1388 is used. If none of the specifications are larger than the bitwidth then
1389 the the largest integer type is used. For example, given the default
1390 specifications above, the i7 type will use the alignment of i8 (next
1391 largest) while both i65 and i256 will use the alignment of i64 (largest
1392 specified).</li>
1393
Reid Spencerde151942007-02-19 23:54:10 +00001394 <li>If no match is found, and the type sought is a vector type, then the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001395 largest vector type that is smaller than the sought vector type will be
1396 used as a fall back. This happens because &lt;128 x double&gt; can be
1397 implemented in terms of 64 &lt;2 x double&gt;, for example.</li>
Reid Spencerde151942007-02-19 23:54:10 +00001398</ol>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001399
Reid Spencerde151942007-02-19 23:54:10 +00001400</div>
Chris Lattnerfa730212004-12-09 16:11:40 +00001401
Dan Gohman556ca272009-07-27 18:07:55 +00001402<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001403<h3>
Dan Gohman556ca272009-07-27 18:07:55 +00001404 <a name="pointeraliasing">Pointer Aliasing Rules</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001405</h3>
Dan Gohman556ca272009-07-27 18:07:55 +00001406
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001407<div>
Dan Gohman556ca272009-07-27 18:07:55 +00001408
Andreas Bolka55e459a2009-07-29 00:02:05 +00001409<p>Any memory access must be done through a pointer value associated
Andreas Bolka99a82052009-07-27 20:37:10 +00001410with an address range of the memory access, otherwise the behavior
Dan Gohman556ca272009-07-27 18:07:55 +00001411is undefined. Pointer values are associated with address ranges
1412according to the following rules:</p>
1413
1414<ul>
Dan Gohman1e109622010-07-02 18:41:32 +00001415 <li>A pointer value is associated with the addresses associated with
1416 any value it is <i>based</i> on.
Andreas Bolka55e459a2009-07-29 00:02:05 +00001417 <li>An address of a global variable is associated with the address
Dan Gohman556ca272009-07-27 18:07:55 +00001418 range of the variable's storage.</li>
1419 <li>The result value of an allocation instruction is associated with
1420 the address range of the allocated storage.</li>
1421 <li>A null pointer in the default address-space is associated with
Andreas Bolka55e459a2009-07-29 00:02:05 +00001422 no address.</li>
Dan Gohman556ca272009-07-27 18:07:55 +00001423 <li>An integer constant other than zero or a pointer value returned
1424 from a function not defined within LLVM may be associated with address
1425 ranges allocated through mechanisms other than those provided by
Andreas Bolka55e459a2009-07-29 00:02:05 +00001426 LLVM. Such ranges shall not overlap with any ranges of addresses
Dan Gohman556ca272009-07-27 18:07:55 +00001427 allocated by mechanisms provided by LLVM.</li>
Dan Gohman1e109622010-07-02 18:41:32 +00001428</ul>
1429
1430<p>A pointer value is <i>based</i> on another pointer value according
1431 to the following rules:</p>
1432
1433<ul>
1434 <li>A pointer value formed from a
1435 <tt><a href="#i_getelementptr">getelementptr</a></tt> operation
1436 is <i>based</i> on the first operand of the <tt>getelementptr</tt>.</li>
1437 <li>The result value of a
1438 <tt><a href="#i_bitcast">bitcast</a></tt> is <i>based</i> on the operand
1439 of the <tt>bitcast</tt>.</li>
1440 <li>A pointer value formed by an
1441 <tt><a href="#i_inttoptr">inttoptr</a></tt> is <i>based</i> on all
1442 pointer values that contribute (directly or indirectly) to the
1443 computation of the pointer's value.</li>
1444 <li>The "<i>based</i> on" relationship is transitive.</li>
1445</ul>
1446
1447<p>Note that this definition of <i>"based"</i> is intentionally
1448 similar to the definition of <i>"based"</i> in C99, though it is
1449 slightly weaker.</p>
Dan Gohman556ca272009-07-27 18:07:55 +00001450
1451<p>LLVM IR does not associate types with memory. The result type of a
Andreas Bolka55e459a2009-07-29 00:02:05 +00001452<tt><a href="#i_load">load</a></tt> merely indicates the size and
1453alignment of the memory from which to load, as well as the
Dan Gohmanc22c0f32010-06-17 19:23:50 +00001454interpretation of the value. The first operand type of a
Andreas Bolka55e459a2009-07-29 00:02:05 +00001455<tt><a href="#i_store">store</a></tt> similarly only indicates the size
1456and alignment of the store.</p>
Dan Gohman556ca272009-07-27 18:07:55 +00001457
1458<p>Consequently, type-based alias analysis, aka TBAA, aka
1459<tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1460LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1461additional information which specialized optimization passes may use
1462to implement type-based alias analysis.</p>
1463
1464</div>
1465
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00001466<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001467<h3>
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00001468 <a name="volatile">Volatile Memory Accesses</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001469</h3>
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00001470
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001471<div>
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00001472
1473<p>Certain memory accesses, such as <a href="#i_load"><tt>load</tt></a>s, <a
1474href="#i_store"><tt>store</tt></a>s, and <a
1475href="#int_memcpy"><tt>llvm.memcpy</tt></a>s may be marked <tt>volatile</tt>.
1476The optimizers must not change the number of volatile operations or change their
1477order of execution relative to other volatile operations. The optimizers
1478<i>may</i> change the order of volatile operations relative to non-volatile
1479operations. This is not Java's "volatile" and has no cross-thread
1480synchronization behavior.</p>
1481
1482</div>
1483
Eli Friedman5b60e1b2011-07-20 21:35:53 +00001484<!-- ======================================================================= -->
1485<h3>
1486 <a name="memmodel">Memory Model for Concurrent Operations</a>
1487</h3>
1488
1489<div>
1490
1491<p>The LLVM IR does not define any way to start parallel threads of execution
1492or to register signal handlers. Nonetheless, there are platform-specific
1493ways to create them, and we define LLVM IR's behavior in their presence. This
1494model is inspired by the C++0x memory model.</p>
1495
1496<p>We define a <i>happens-before</i> partial order as the least partial order
1497that</p>
1498<ul>
1499 <li>Is a superset of single-thread program order, and</li>
1500 <li>When a <i>synchronizes-with</i> <tt>b</tt>, includes an edge from
1501 <tt>a</tt> to <tt>b</tt>. <i>Synchronizes-with</i> pairs are introduced
1502 by platform-specific techniques, like pthread locks, thread
1503 creation, thread joining, etc., and by the atomic operations described
1504 in the <a href="#int_atomics">Atomic intrinsics</a> section.</li>
1505</ul>
1506
1507<p>Note that program order does not introduce <i>happens-before</i> edges
1508between a thread and signals executing inside that thread.</p>
1509
1510<p>Every (defined) read operation (load instructions, memcpy, atomic
1511loads/read-modify-writes, etc.) <var>R</var> reads a series of bytes written by
1512(defined) write operations (store instructions, atomic
Eli Friedman118973a2011-07-22 03:04:45 +00001513stores/read-modify-writes, memcpy, etc.). For the purposes of this section,
1514initialized globals are considered to have a write of the initializer which is
1515atomic and happens before any other read or write of the memory in question.
1516For each byte of a read <var>R</var>, <var>R<sub>byte</sub></var> may see
1517any write to the same byte, except:</p>
Eli Friedman5b60e1b2011-07-20 21:35:53 +00001518
1519<ul>
1520 <li>If <var>write<sub>1</sub></var> happens before
1521 <var>write<sub>2</sub></var>, and <var>write<sub>2</sub></var> happens
1522 before <var>R<sub>byte</sub></var>, then <var>R<sub>byte</sub></var>
Eli Friedman118973a2011-07-22 03:04:45 +00001523 does not see <var>write<sub>1</sub></var>.
Eli Friedman5b60e1b2011-07-20 21:35:53 +00001524 <li>If <var>R<sub>byte</sub></var> happens before <var>write<sub>3</var>,
Eli Friedman118973a2011-07-22 03:04:45 +00001525 then <var>R<sub>byte</sub></var> does not see
Eli Friedman5b60e1b2011-07-20 21:35:53 +00001526 <var>write<sub>3</sub></var>.
1527</ul>
1528
1529<p>Given that definition, <var>R<sub>byte</sub></var> is defined as follows:
1530<ul>
1531 <li>If there is no write to the same byte that happens before
1532 <var>R<sub>byte</sub></var>, <var>R<sub>byte</sub></var> returns
1533 <tt>undef</tt> for that byte.
Eli Friedman118973a2011-07-22 03:04:45 +00001534 <li>Otherwise, if <var>R<sub>byte</sub></var> may see exactly one write,
Eli Friedman5b60e1b2011-07-20 21:35:53 +00001535 <var>R<sub>byte</sub></var> returns the value written by that
1536 write.</li>
Eli Friedman118973a2011-07-22 03:04:45 +00001537 <li>Otherwise, if <var>R</var> is atomic, and all the writes
1538 <var>R<sub>byte</sub></var> may see are atomic, it chooses one of the
1539 values written. See the <a href="#int_atomics">Atomic intrinsics</a>
1540 section for additional guarantees on how the choice is made.
Eli Friedman5b60e1b2011-07-20 21:35:53 +00001541 <li>Otherwise <var>R<sub>byte</sub></var> returns <tt>undef</tt>.</li>
1542</ul>
1543
1544<p><var>R</var> returns the value composed of the series of bytes it read.
1545This implies that some bytes within the value may be <tt>undef</tt>
1546<b>without</b> the entire value being <tt>undef</tt>. Note that this only
1547defines the semantics of the operation; it doesn't mean that targets will
1548emit more than one instruction to read the series of bytes.</p>
1549
1550<p>Note that in cases where none of the atomic intrinsics are used, this model
1551places only one restriction on IR transformations on top of what is required
1552for single-threaded execution: introducing a store to a byte which might not
Eli Friedman118973a2011-07-22 03:04:45 +00001553otherwise be stored to can introduce undefined behavior. (Specifically, in
1554the case where another thread might write to and read from an address,
1555introducing a store can change a load that may see exactly one write into
1556a load that may see multiple writes.)</p>
Eli Friedman5b60e1b2011-07-20 21:35:53 +00001557
1558<!-- FIXME: This model assumes all targets where concurrency is relevant have
1559a byte-size store which doesn't affect adjacent bytes. As far as I can tell,
1560none of the backends currently in the tree fall into this category; however,
1561there might be targets which care. If there are, we want a paragraph
1562like the following:
1563
1564Targets may specify that stores narrower than a certain width are not
1565available; on such a target, for the purposes of this model, treat any
1566non-atomic write with an alignment or width less than the minimum width
1567as if it writes to the relevant surrounding bytes.
1568-->
1569
1570</div>
1571
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001572</div>
1573
Chris Lattner00950542001-06-06 20:29:01 +00001574<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001575<h2><a name="typesystem">Type System</a></h2>
Chris Lattner261efe92003-11-25 01:02:51 +00001576<!-- *********************************************************************** -->
Chris Lattnerfa730212004-12-09 16:11:40 +00001577
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001578<div>
Chris Lattnerfa730212004-12-09 16:11:40 +00001579
Misha Brukman9d0919f2003-11-08 01:05:38 +00001580<p>The LLVM type system is one of the most important features of the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001581 intermediate representation. Being typed enables a number of optimizations
1582 to be performed on the intermediate representation directly, without having
1583 to do extra analyses on the side before the transformation. A strong type
1584 system makes it easier to read the generated code and enables novel analyses
1585 and transformations that are not feasible to perform on normal three address
1586 code representations.</p>
Chris Lattnerfa730212004-12-09 16:11:40 +00001587
Chris Lattner00950542001-06-06 20:29:01 +00001588<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001589<h3>
1590 <a name="t_classifications">Type Classifications</a>
1591</h3>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001592
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001593<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001594
1595<p>The types fall into a few useful classifications:</p>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00001596
1597<table border="1" cellspacing="0" cellpadding="4">
Chris Lattner261efe92003-11-25 01:02:51 +00001598 <tbody>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001599 <tr><th>Classification</th><th>Types</th></tr>
Chris Lattner261efe92003-11-25 01:02:51 +00001600 <tr>
Chris Lattner4f69f462008-01-04 04:32:38 +00001601 <td><a href="#t_integer">integer</a></td>
Reid Spencer2b916312007-05-16 18:44:01 +00001602 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
Chris Lattner261efe92003-11-25 01:02:51 +00001603 </tr>
1604 <tr>
Chris Lattner4f69f462008-01-04 04:32:38 +00001605 <td><a href="#t_floating">floating point</a></td>
1606 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Chris Lattner261efe92003-11-25 01:02:51 +00001607 </tr>
1608 <tr>
1609 <td><a name="t_firstclass">first class</a></td>
Chris Lattner4f69f462008-01-04 04:32:38 +00001610 <td><a href="#t_integer">integer</a>,
1611 <a href="#t_floating">floating point</a>,
1612 <a href="#t_pointer">pointer</a>,
Dan Gohman0066db62008-06-18 18:42:13 +00001613 <a href="#t_vector">vector</a>,
Dan Gohmana334d5f2008-05-12 23:51:09 +00001614 <a href="#t_struct">structure</a>,
1615 <a href="#t_array">array</a>,
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001616 <a href="#t_label">label</a>,
1617 <a href="#t_metadata">metadata</a>.
Reid Spencerca86e162006-12-31 07:07:53 +00001618 </td>
Chris Lattner261efe92003-11-25 01:02:51 +00001619 </tr>
Chris Lattner4f69f462008-01-04 04:32:38 +00001620 <tr>
1621 <td><a href="#t_primitive">primitive</a></td>
1622 <td><a href="#t_label">label</a>,
1623 <a href="#t_void">void</a>,
Tobias Grosser05387292010-12-28 20:29:31 +00001624 <a href="#t_integer">integer</a>,
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001625 <a href="#t_floating">floating point</a>,
Dale Johannesen21fe99b2010-10-01 00:48:59 +00001626 <a href="#t_x86mmx">x86mmx</a>,
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001627 <a href="#t_metadata">metadata</a>.</td>
Chris Lattner4f69f462008-01-04 04:32:38 +00001628 </tr>
1629 <tr>
1630 <td><a href="#t_derived">derived</a></td>
Chris Lattnerfdfeb692010-02-12 20:49:41 +00001631 <td><a href="#t_array">array</a>,
Chris Lattner4f69f462008-01-04 04:32:38 +00001632 <a href="#t_function">function</a>,
1633 <a href="#t_pointer">pointer</a>,
1634 <a href="#t_struct">structure</a>,
Chris Lattner4f69f462008-01-04 04:32:38 +00001635 <a href="#t_vector">vector</a>,
1636 <a href="#t_opaque">opaque</a>.
Dan Gohman01ac1012008-10-14 16:32:04 +00001637 </td>
Chris Lattner4f69f462008-01-04 04:32:38 +00001638 </tr>
Chris Lattner261efe92003-11-25 01:02:51 +00001639 </tbody>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001640</table>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00001641
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001642<p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1643 important. Values of these types are the only ones which can be produced by
Nick Lewyckyec38da42009-09-27 00:45:11 +00001644 instructions.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001645
Misha Brukman9d0919f2003-11-08 01:05:38 +00001646</div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001647
Chris Lattner00950542001-06-06 20:29:01 +00001648<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001649<h3>
1650 <a name="t_primitive">Primitive Types</a>
1651</h3>
Chris Lattner8f8c7b72008-01-04 04:34:14 +00001652
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001653<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001654
Chris Lattner4f69f462008-01-04 04:32:38 +00001655<p>The primitive types are the fundamental building blocks of the LLVM
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001656 system.</p>
Chris Lattner4f69f462008-01-04 04:32:38 +00001657
1658<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001659<h4>
1660 <a name="t_integer">Integer Type</a>
1661</h4>
Nick Lewyckyec38da42009-09-27 00:45:11 +00001662
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001663<div>
Nick Lewyckyec38da42009-09-27 00:45:11 +00001664
1665<h5>Overview:</h5>
1666<p>The integer type is a very simple type that simply specifies an arbitrary
1667 bit width for the integer type desired. Any bit width from 1 bit to
1668 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
1669
1670<h5>Syntax:</h5>
1671<pre>
1672 iN
1673</pre>
1674
1675<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1676 value.</p>
1677
1678<h5>Examples:</h5>
1679<table class="layout">
1680 <tr class="layout">
1681 <td class="left"><tt>i1</tt></td>
1682 <td class="left">a single-bit integer.</td>
1683 </tr>
1684 <tr class="layout">
1685 <td class="left"><tt>i32</tt></td>
1686 <td class="left">a 32-bit integer.</td>
1687 </tr>
1688 <tr class="layout">
1689 <td class="left"><tt>i1942652</tt></td>
1690 <td class="left">a really big integer of over 1 million bits.</td>
1691 </tr>
1692</table>
1693
Nick Lewyckyec38da42009-09-27 00:45:11 +00001694</div>
1695
1696<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001697<h4>
1698 <a name="t_floating">Floating Point Types</a>
1699</h4>
Chris Lattner4f69f462008-01-04 04:32:38 +00001700
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001701<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001702
1703<table>
1704 <tbody>
1705 <tr><th>Type</th><th>Description</th></tr>
1706 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1707 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1708 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1709 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1710 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1711 </tbody>
1712</table>
1713
Chris Lattner4f69f462008-01-04 04:32:38 +00001714</div>
1715
1716<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001717<h4>
1718 <a name="t_x86mmx">X86mmx Type</a>
1719</h4>
Dale Johannesen21fe99b2010-10-01 00:48:59 +00001720
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001721<div>
Dale Johannesen21fe99b2010-10-01 00:48:59 +00001722
1723<h5>Overview:</h5>
1724<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>
1725
1726<h5>Syntax:</h5>
1727<pre>
Dale Johannesen473a8c82010-10-01 01:07:02 +00001728 x86mmx
Dale Johannesen21fe99b2010-10-01 00:48:59 +00001729</pre>
1730
1731</div>
1732
1733<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001734<h4>
1735 <a name="t_void">Void Type</a>
1736</h4>
Chris Lattner4f69f462008-01-04 04:32:38 +00001737
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001738<div>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001739
Chris Lattner4f69f462008-01-04 04:32:38 +00001740<h5>Overview:</h5>
1741<p>The void type does not represent any value and has no size.</p>
1742
1743<h5>Syntax:</h5>
Chris Lattner4f69f462008-01-04 04:32:38 +00001744<pre>
1745 void
1746</pre>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001747
Chris Lattner4f69f462008-01-04 04:32:38 +00001748</div>
1749
1750<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001751<h4>
1752 <a name="t_label">Label Type</a>
1753</h4>
Chris Lattner4f69f462008-01-04 04:32:38 +00001754
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001755<div>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001756
Chris Lattner4f69f462008-01-04 04:32:38 +00001757<h5>Overview:</h5>
1758<p>The label type represents code labels.</p>
1759
1760<h5>Syntax:</h5>
Chris Lattner4f69f462008-01-04 04:32:38 +00001761<pre>
1762 label
1763</pre>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001764
Chris Lattner4f69f462008-01-04 04:32:38 +00001765</div>
1766
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001767<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001768<h4>
1769 <a name="t_metadata">Metadata Type</a>
1770</h4>
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001771
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001772<div>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001773
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001774<h5>Overview:</h5>
Nick Lewyckyc261df92009-09-27 23:27:42 +00001775<p>The metadata type represents embedded metadata. No derived types may be
1776 created from metadata except for <a href="#t_function">function</a>
1777 arguments.
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001778
1779<h5>Syntax:</h5>
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001780<pre>
1781 metadata
1782</pre>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001783
Nick Lewycky7a0370f2009-05-30 05:06:04 +00001784</div>
1785
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001786</div>
Chris Lattner4f69f462008-01-04 04:32:38 +00001787
1788<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001789<h3>
1790 <a name="t_derived">Derived Types</a>
1791</h3>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001792
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001793<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001794
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001795<p>The real power in LLVM comes from the derived types in the system. This is
1796 what allows a programmer to represent arrays, functions, pointers, and other
Nick Lewyckyec38da42009-09-27 00:45:11 +00001797 useful types. Each of these types contain one or more element types which
1798 may be a primitive type, or another derived type. For example, it is
1799 possible to have a two dimensional array, using an array as the element type
1800 of another array.</p>
Dan Gohmand8791e52009-01-24 15:58:40 +00001801
Chris Lattner1afcace2011-07-09 17:41:24 +00001802</div>
1803
1804
Chris Lattnerfdfeb692010-02-12 20:49:41 +00001805<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001806<h4>
1807 <a name="t_aggregate">Aggregate Types</a>
1808</h4>
Chris Lattnerfdfeb692010-02-12 20:49:41 +00001809
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001810<div>
Chris Lattnerfdfeb692010-02-12 20:49:41 +00001811
1812<p>Aggregate Types are a subset of derived types that can contain multiple
1813 member types. <a href="#t_array">Arrays</a>,
Chris Lattner61c70e92010-08-28 04:09:24 +00001814 <a href="#t_struct">structs</a>, and <a href="#t_vector">vectors</a> are
1815 aggregate types.</p>
Chris Lattnerfdfeb692010-02-12 20:49:41 +00001816
1817</div>
1818
Reid Spencer2b916312007-05-16 18:44:01 +00001819<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001820<h4>
1821 <a name="t_array">Array Type</a>
1822</h4>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001823
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001824<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001825
Chris Lattner00950542001-06-06 20:29:01 +00001826<h5>Overview:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00001827<p>The array type is a very simple derived type that arranges elements
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001828 sequentially in memory. The array type requires a size (number of elements)
1829 and an underlying data type.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001830
Chris Lattner7faa8832002-04-14 06:13:44 +00001831<h5>Syntax:</h5>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001832<pre>
1833 [&lt;# elements&gt; x &lt;elementtype&gt;]
1834</pre>
1835
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001836<p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1837 be any type with a size.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00001838
Chris Lattner7faa8832002-04-14 06:13:44 +00001839<h5>Examples:</h5>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001840<table class="layout">
1841 <tr class="layout">
Chris Lattner23ff1f92007-12-19 05:04:11 +00001842 <td class="left"><tt>[40 x i32]</tt></td>
1843 <td class="left">Array of 40 32-bit integer values.</td>
1844 </tr>
1845 <tr class="layout">
1846 <td class="left"><tt>[41 x i32]</tt></td>
1847 <td class="left">Array of 41 32-bit integer values.</td>
1848 </tr>
1849 <tr class="layout">
1850 <td class="left"><tt>[4 x i8]</tt></td>
1851 <td class="left">Array of 4 8-bit integer values.</td>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001852 </tr>
Chris Lattner00950542001-06-06 20:29:01 +00001853</table>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001854<p>Here are some examples of multidimensional arrays:</p>
1855<table class="layout">
1856 <tr class="layout">
Chris Lattner23ff1f92007-12-19 05:04:11 +00001857 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1858 <td class="left">3x4 array of 32-bit integer values.</td>
1859 </tr>
1860 <tr class="layout">
1861 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1862 <td class="left">12x10 array of single precision floating point values.</td>
1863 </tr>
1864 <tr class="layout">
1865 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1866 <td class="left">2x3x4 array of 16-bit integer values.</td>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001867 </tr>
1868</table>
Chris Lattnere67a9512005-06-24 17:22:57 +00001869
Dan Gohman7657f6b2009-11-09 19:01:53 +00001870<p>There is no restriction on indexing beyond the end of the array implied by
1871 a static type (though there are restrictions on indexing beyond the bounds
1872 of an allocated object in some cases). This means that single-dimension
1873 'variable sized array' addressing can be implemented in LLVM with a zero
1874 length array type. An implementation of 'pascal style arrays' in LLVM could
1875 use the type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
Chris Lattnere67a9512005-06-24 17:22:57 +00001876
Misha Brukman9d0919f2003-11-08 01:05:38 +00001877</div>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001878
Chris Lattner00950542001-06-06 20:29:01 +00001879<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001880<h4>
1881 <a name="t_function">Function Type</a>
1882</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001883
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001884<div>
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00001885
Chris Lattner00950542001-06-06 20:29:01 +00001886<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001887<p>The function type can be thought of as a function signature. It consists of
1888 a return type and a list of formal parameter types. The return type of a
Chris Lattner61c70e92010-08-28 04:09:24 +00001889 function type is a first class type or a void type.</p>
Devang Patelc3fc6df2008-03-10 20:49:15 +00001890
Chris Lattner00950542001-06-06 20:29:01 +00001891<h5>Syntax:</h5>
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00001892<pre>
Nick Lewycky51386942009-09-27 07:55:32 +00001893 &lt;returntype&gt; (&lt;parameter list&gt;)
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00001894</pre>
1895
John Criswell0ec250c2005-10-24 16:17:18 +00001896<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001897 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1898 which indicates that the function takes a variable number of arguments.
1899 Variable argument functions can access their arguments with
1900 the <a href="#int_varargs">variable argument handling intrinsic</a>
Chris Lattner0724fbd2010-03-02 06:36:51 +00001901 functions. '<tt>&lt;returntype&gt;</tt>' is any type except
Nick Lewyckyc261df92009-09-27 23:27:42 +00001902 <a href="#t_label">label</a>.</p>
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00001903
Chris Lattner00950542001-06-06 20:29:01 +00001904<h5>Examples:</h5>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001905<table class="layout">
1906 <tr class="layout">
Reid Spencer92f82302006-12-31 07:18:34 +00001907 <td class="left"><tt>i32 (i32)</tt></td>
1908 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001909 </td>
Reid Spencer92f82302006-12-31 07:18:34 +00001910 </tr><tr class="layout">
Chris Lattner0724fbd2010-03-02 06:36:51 +00001911 <td class="left"><tt>float&nbsp;(i16,&nbsp;i32&nbsp;*)&nbsp;*
Reid Spencerf17a0b72006-12-31 07:20:23 +00001912 </tt></td>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00001913 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
Chris Lattner0724fbd2010-03-02 06:36:51 +00001914 an <tt>i16</tt> and a <a href="#t_pointer">pointer</a> to <tt>i32</tt>,
1915 returning <tt>float</tt>.
Reid Spencer92f82302006-12-31 07:18:34 +00001916 </td>
1917 </tr><tr class="layout">
1918 <td class="left"><tt>i32 (i8*, ...)</tt></td>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00001919 <td class="left">A vararg function that takes at least one
1920 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1921 which returns an integer. This is the signature for <tt>printf</tt> in
Reid Spencer92f82302006-12-31 07:18:34 +00001922 LLVM.
Reid Spencerd3f876c2004-11-01 08:19:36 +00001923 </td>
Devang Patela582f402008-03-24 05:35:41 +00001924 </tr><tr class="layout">
1925 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Nick Lewycky51386942009-09-27 07:55:32 +00001926 <td class="left">A function taking an <tt>i32</tt>, returning a
1927 <a href="#t_struct">structure</a> containing two <tt>i32</tt> values
Devang Patela582f402008-03-24 05:35:41 +00001928 </td>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001929 </tr>
Chris Lattner00950542001-06-06 20:29:01 +00001930</table>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00001931
Misha Brukman9d0919f2003-11-08 01:05:38 +00001932</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001933
Chris Lattner00950542001-06-06 20:29:01 +00001934<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001935<h4>
1936 <a name="t_struct">Structure Type</a>
1937</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001938
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001939<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001940
Chris Lattner00950542001-06-06 20:29:01 +00001941<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001942<p>The structure type is used to represent a collection of data members together
Chris Lattner1afcace2011-07-09 17:41:24 +00001943 in memory. The elements of a structure may be any type that has a size.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001944
Jeffrey Yasskin7a088cf2010-01-11 19:19:26 +00001945<p>Structures in memory are accessed using '<tt><a href="#i_load">load</a></tt>'
1946 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field
1947 with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
1948 Structures in registers are accessed using the
1949 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' and
1950 '<tt><a href="#i_insertvalue">insertvalue</a></tt>' instructions.</p>
Chris Lattner1afcace2011-07-09 17:41:24 +00001951
1952<p>Structures may optionally be "packed" structures, which indicate that the
1953 alignment of the struct is one byte, and that there is no padding between
1954 the elements. In non-packed structs, padding between field types is defined
1955 by the target data string to match the underlying processor.</p>
1956
1957<p>Structures can either be "anonymous" or "named". An anonymous structure is
1958 defined inline with other types (e.g. <tt>{i32, i32}*</tt>) and a named types
1959 are always defined at the top level with a name. Anonmyous types are uniqued
1960 by their contents and can never be recursive since there is no way to write
1961 one. Named types can be recursive.
1962</p>
1963
Chris Lattner00950542001-06-06 20:29:01 +00001964<h5>Syntax:</h5>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001965<pre>
Chris Lattner1afcace2011-07-09 17:41:24 +00001966 %T1 = type { &lt;type list&gt; } <i>; Named normal struct type</i>
1967 %T2 = type &lt;{ &lt;type list&gt; }&gt; <i>; Named packed struct type</i>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00001968</pre>
Chris Lattner1afcace2011-07-09 17:41:24 +00001969
Chris Lattner00950542001-06-06 20:29:01 +00001970<h5>Examples:</h5>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001971<table class="layout">
1972 <tr class="layout">
Jeff Cohen6f1cc772007-04-22 01:17:39 +00001973 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1974 <td class="left">A triple of three <tt>i32</tt> values</td>
Chris Lattner1afcace2011-07-09 17:41:24 +00001975 </tr>
1976 <tr class="layout">
Jeff Cohen6f1cc772007-04-22 01:17:39 +00001977 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1978 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1979 second element is a <a href="#t_pointer">pointer</a> to a
1980 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1981 an <tt>i32</tt>.</td>
Reid Spencerd3f876c2004-11-01 08:19:36 +00001982 </tr>
Chris Lattner1afcace2011-07-09 17:41:24 +00001983 <tr class="layout">
1984 <td class="left"><tt>&lt;{ i8, i32 }&gt;</tt></td>
1985 <td class="left">A packed struct known to be 5 bytes in size.</td>
1986 </tr>
Chris Lattner00950542001-06-06 20:29:01 +00001987</table>
Dan Gohmand8791e52009-01-24 15:58:40 +00001988
Misha Brukman9d0919f2003-11-08 01:05:38 +00001989</div>
Chris Lattner1afcace2011-07-09 17:41:24 +00001990
Chris Lattner00950542001-06-06 20:29:01 +00001991<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001992<h4>
Chris Lattner628ed392011-07-23 19:59:08 +00001993 <a name="t_opaque">Opaque Structure Types</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00001994</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001995
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00001996<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00001997
Andrew Lenharth75e10682006-12-08 17:13:00 +00001998<h5>Overview:</h5>
Chris Lattner628ed392011-07-23 19:59:08 +00001999<p>Opaque structure types are used to represent named structure types that do
2000 not have a body specified. This corresponds (for example) to the C notion of
2001 a forward declared structure.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002002
Andrew Lenharth75e10682006-12-08 17:13:00 +00002003<h5>Syntax:</h5>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00002004<pre>
Chris Lattner1afcace2011-07-09 17:41:24 +00002005 %X = type opaque
2006 %52 = type opaque
Bill Wendlingc39e3e02009-07-20 02:39:26 +00002007</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002008
Andrew Lenharth75e10682006-12-08 17:13:00 +00002009<h5>Examples:</h5>
2010<table class="layout">
2011 <tr class="layout">
Chris Lattner1afcace2011-07-09 17:41:24 +00002012 <td class="left"><tt>opaque</tt></td>
2013 <td class="left">An opaque type.</td>
Andrew Lenharth75e10682006-12-08 17:13:00 +00002014 </tr>
2015</table>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002016
Andrew Lenharth75e10682006-12-08 17:13:00 +00002017</div>
2018
Chris Lattner1afcace2011-07-09 17:41:24 +00002019
2020
Andrew Lenharth75e10682006-12-08 17:13:00 +00002021<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002022<h4>
2023 <a name="t_pointer">Pointer Type</a>
2024</h4>
Chris Lattner0fd4a272009-02-08 19:53:29 +00002025
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002026<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002027
2028<h5>Overview:</h5>
Dan Gohmanff3ef322010-02-25 16:50:07 +00002029<p>The pointer type is used to specify memory locations.
2030 Pointers are commonly used to reference objects in memory.</p>
2031
2032<p>Pointer types may have an optional address space attribute defining the
2033 numbered address space where the pointed-to object resides. The default
2034 address space is number zero. The semantics of non-zero address
2035 spaces are target-specific.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002036
2037<p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
2038 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
Chris Lattner0fd4a272009-02-08 19:53:29 +00002039
Chris Lattner7faa8832002-04-14 06:13:44 +00002040<h5>Syntax:</h5>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00002041<pre>
2042 &lt;type&gt; *
2043</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002044
Chris Lattner7faa8832002-04-14 06:13:44 +00002045<h5>Examples:</h5>
Reid Spencerd3f876c2004-11-01 08:19:36 +00002046<table class="layout">
2047 <tr class="layout">
Dan Gohman2a08c532009-01-04 23:44:43 +00002048 <td class="left"><tt>[4 x i32]*</tt></td>
Chris Lattner23ff1f92007-12-19 05:04:11 +00002049 <td class="left">A <a href="#t_pointer">pointer</a> to <a
2050 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
2051 </tr>
2052 <tr class="layout">
Dan Gohmanfe47aae2010-05-28 17:13:49 +00002053 <td class="left"><tt>i32 (i32*) *</tt></td>
Chris Lattner23ff1f92007-12-19 05:04:11 +00002054 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Reid Spencerca86e162006-12-31 07:07:53 +00002055 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner23ff1f92007-12-19 05:04:11 +00002056 <tt>i32</tt>.</td>
2057 </tr>
2058 <tr class="layout">
2059 <td class="left"><tt>i32 addrspace(5)*</tt></td>
2060 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
2061 that resides in address space #5.</td>
Reid Spencerd3f876c2004-11-01 08:19:36 +00002062 </tr>
Misha Brukman9d0919f2003-11-08 01:05:38 +00002063</table>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002064
Misha Brukman9d0919f2003-11-08 01:05:38 +00002065</div>
Reid Spencerd3f876c2004-11-01 08:19:36 +00002066
Chris Lattnera58561b2004-08-12 19:12:28 +00002067<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002068<h4>
2069 <a name="t_vector">Vector Type</a>
2070</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002071
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002072<div>
Chris Lattner69c11bb2005-04-25 17:34:15 +00002073
Chris Lattnera58561b2004-08-12 19:12:28 +00002074<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002075<p>A vector type is a simple derived type that represents a vector of elements.
2076 Vector types are used when multiple primitive data are operated in parallel
2077 using a single instruction (SIMD). A vector type requires a size (number of
Duncan Sandsd40d14e2009-11-27 13:38:03 +00002078 elements) and an underlying primitive data type. Vector types are considered
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002079 <a href="#t_firstclass">first class</a>.</p>
Chris Lattner69c11bb2005-04-25 17:34:15 +00002080
Chris Lattnera58561b2004-08-12 19:12:28 +00002081<h5>Syntax:</h5>
Chris Lattner69c11bb2005-04-25 17:34:15 +00002082<pre>
2083 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
2084</pre>
2085
Chris Lattner7d2e7be2010-10-10 18:20:35 +00002086<p>The number of elements is a constant integer value larger than 0; elementtype
2087 may be any integer or floating point type. Vectors of size zero are not
2088 allowed, and pointers are not allowed as the element type.</p>
Chris Lattner69c11bb2005-04-25 17:34:15 +00002089
Chris Lattnera58561b2004-08-12 19:12:28 +00002090<h5>Examples:</h5>
Reid Spencerd3f876c2004-11-01 08:19:36 +00002091<table class="layout">
2092 <tr class="layout">
Chris Lattner23ff1f92007-12-19 05:04:11 +00002093 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
2094 <td class="left">Vector of 4 32-bit integer values.</td>
2095 </tr>
2096 <tr class="layout">
2097 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
2098 <td class="left">Vector of 8 32-bit floating-point values.</td>
2099 </tr>
2100 <tr class="layout">
2101 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
2102 <td class="left">Vector of 2 64-bit integer values.</td>
Reid Spencerd3f876c2004-11-01 08:19:36 +00002103 </tr>
2104</table>
Dan Gohmand8791e52009-01-24 15:58:40 +00002105
Misha Brukman9d0919f2003-11-08 01:05:38 +00002106</div>
2107
Chris Lattnerc3f59762004-12-09 17:30:23 +00002108<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002109<h2><a name="constants">Constants</a></h2>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002110<!-- *********************************************************************** -->
2111
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002112<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002113
2114<p>LLVM has several different basic types of constants. This section describes
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002115 them all and their syntax.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002116
Chris Lattnerc3f59762004-12-09 17:30:23 +00002117<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002118<h3>
2119 <a name="simpleconstants">Simple Constants</a>
2120</h3>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002121
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002122<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002123
2124<dl>
2125 <dt><b>Boolean constants</b></dt>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002126 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
Nick Lewyckyec38da42009-09-27 00:45:11 +00002127 constants of the <tt><a href="#t_integer">i1</a></tt> type.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002128
2129 <dt><b>Integer constants</b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002130 <dd>Standard integers (such as '4') are constants of
2131 the <a href="#t_integer">integer</a> type. Negative numbers may be used
2132 with integer types.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002133
2134 <dt><b>Floating point constants</b></dt>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002135 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002136 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
2137 notation (see below). The assembler requires the exact decimal value of a
2138 floating-point constant. For example, the assembler accepts 1.25 but
2139 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
2140 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002141
2142 <dt><b>Null pointer constants</b></dt>
John Criswell9e2485c2004-12-10 15:51:16 +00002143 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002144 and must be of <a href="#t_pointer">pointer type</a>.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002145</dl>
2146
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002147<p>The one non-intuitive notation for constants is the hexadecimal form of
2148 floating point constants. For example, the form '<tt>double
2149 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
2150 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
2151 constants are required (and the only time that they are generated by the
2152 disassembler) is when a floating point constant must be emitted but it cannot
2153 be represented as a decimal floating point number in a reasonable number of
2154 digits. For example, NaN's, infinities, and other special values are
2155 represented in their IEEE hexadecimal format so that assembly and disassembly
2156 do not cause any bits to change in the constants.</p>
2157
Dale Johannesenbd5e5a82009-02-11 22:14:51 +00002158<p>When using the hexadecimal form, constants of types float and double are
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002159 represented using the 16-digit form shown above (which matches the IEEE754
2160 representation for double); float values must, however, be exactly
2161 representable as IEE754 single precision. Hexadecimal format is always used
2162 for long double, and there are three forms of long double. The 80-bit format
2163 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
2164 The 128-bit format used by PowerPC (two adjacent doubles) is represented
2165 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
2166 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
2167 currently supported target uses this format. Long doubles will only work if
2168 they match the long double format on your target. All hexadecimal formats
2169 are big-endian (sign bit at the left).</p>
2170
Dale Johannesen21fe99b2010-10-01 00:48:59 +00002171<p>There are no constants of type x86mmx.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002172</div>
2173
2174<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002175<h3>
Bill Wendlingd9fe2982009-07-20 02:32:41 +00002176<a name="aggregateconstants"></a> <!-- old anchor -->
2177<a name="complexconstants">Complex Constants</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002178</h3>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002179
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002180<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002181
Chris Lattner70882792009-02-28 18:32:25 +00002182<p>Complex constants are a (potentially recursive) combination of simple
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002183 constants and smaller complex constants.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002184
2185<dl>
2186 <dt><b>Structure constants</b></dt>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002187 <dd>Structure constants are represented with notation similar to structure
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002188 type definitions (a comma separated list of elements, surrounded by braces
2189 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
2190 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
2191 Structure constants must have <a href="#t_struct">structure type</a>, and
2192 the number and types of elements must match those specified by the
2193 type.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002194
2195 <dt><b>Array constants</b></dt>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002196 <dd>Array constants are represented with notation similar to array type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002197 definitions (a comma separated list of elements, surrounded by square
2198 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
2199 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
2200 the number and types of elements must match those specified by the
2201 type.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002202
Reid Spencer485bad12007-02-15 03:07:05 +00002203 <dt><b>Vector constants</b></dt>
Reid Spencer485bad12007-02-15 03:07:05 +00002204 <dd>Vector constants are represented with notation similar to vector type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002205 definitions (a comma separated list of elements, surrounded by
2206 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32
2207 42, i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must
2208 have <a href="#t_vector">vector type</a>, and the number and types of
2209 elements must match those specified by the type.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002210
2211 <dt><b>Zero initialization</b></dt>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002212 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
Chris Lattnerfdfeb692010-02-12 20:49:41 +00002213 value to zero of <em>any</em> type, including scalar and
2214 <a href="#t_aggregate">aggregate</a> types.
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002215 This is often used to avoid having to print large zero initializers
2216 (e.g. for large arrays) and is always exactly equivalent to using explicit
2217 zero initializers.</dd>
Nick Lewycky21cc4462009-04-04 07:22:01 +00002218
2219 <dt><b>Metadata node</b></dt>
Nick Lewycky1e8c7a62009-05-30 16:08:30 +00002220 <dd>A metadata node is a structure-like constant with
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002221 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
2222 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
2223 be interpreted as part of the instruction stream, metadata is a place to
2224 attach additional information such as debug info.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002225</dl>
2226
2227</div>
2228
2229<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002230<h3>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002231 <a name="globalconstants">Global Variable and Function Addresses</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002232</h3>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002233
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002234<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002235
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002236<p>The addresses of <a href="#globalvars">global variables</a>
2237 and <a href="#functionstructure">functions</a> are always implicitly valid
2238 (link-time) constants. These constants are explicitly referenced when
2239 the <a href="#identifiers">identifier for the global</a> is used and always
2240 have <a href="#t_pointer">pointer</a> type. For example, the following is a
2241 legal LLVM file:</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002242
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002243<pre class="doc_code">
Chris Lattnera18a4242007-06-06 18:28:13 +00002244@X = global i32 17
2245@Y = global i32 42
2246@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
Chris Lattnerc3f59762004-12-09 17:30:23 +00002247</pre>
2248
2249</div>
2250
2251<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002252<h3>
2253 <a name="undefvalues">Undefined Values</a>
2254</h3>
2255
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002256<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002257
Chris Lattner48a109c2009-09-07 22:52:39 +00002258<p>The string '<tt>undef</tt>' can be used anywhere a constant is expected, and
Benjamin Kramer8040cd32009-10-12 14:46:08 +00002259 indicates that the user of the value may receive an unspecified bit-pattern.
Bill Wendling1b383ba2010-10-27 01:07:41 +00002260 Undefined values may be of any type (other than '<tt>label</tt>'
2261 or '<tt>void</tt>') and be used anywhere a constant is permitted.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002262
Chris Lattnerc608cb12009-09-11 01:49:31 +00002263<p>Undefined values are useful because they indicate to the compiler that the
Chris Lattner48a109c2009-09-07 22:52:39 +00002264 program is well defined no matter what value is used. This gives the
2265 compiler more freedom to optimize. Here are some examples of (potentially
2266 surprising) transformations that are valid (in pseudo IR):</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002267
Chris Lattner48a109c2009-09-07 22:52:39 +00002268
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002269<pre class="doc_code">
Chris Lattner48a109c2009-09-07 22:52:39 +00002270 %A = add %X, undef
2271 %B = sub %X, undef
2272 %C = xor %X, undef
2273Safe:
2274 %A = undef
2275 %B = undef
2276 %C = undef
2277</pre>
Chris Lattner48a109c2009-09-07 22:52:39 +00002278
2279<p>This is safe because all of the output bits are affected by the undef bits.
Bill Wendling1b383ba2010-10-27 01:07:41 +00002280 Any output bit can have a zero or one depending on the input bits.</p>
Chris Lattner48a109c2009-09-07 22:52:39 +00002281
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002282<pre class="doc_code">
Chris Lattner48a109c2009-09-07 22:52:39 +00002283 %A = or %X, undef
2284 %B = and %X, undef
2285Safe:
2286 %A = -1
2287 %B = 0
2288Unsafe:
2289 %A = undef
2290 %B = undef
2291</pre>
Chris Lattner48a109c2009-09-07 22:52:39 +00002292
2293<p>These logical operations have bits that are not always affected by the input.
Bill Wendling1b383ba2010-10-27 01:07:41 +00002294 For example, if <tt>%X</tt> has a zero bit, then the output of the
2295 '<tt>and</tt>' operation will always be a zero for that bit, no matter what
2296 the corresponding bit from the '<tt>undef</tt>' is. As such, it is unsafe to
2297 optimize or assume that the result of the '<tt>and</tt>' is '<tt>undef</tt>'.
2298 However, it is safe to assume that all bits of the '<tt>undef</tt>' could be
2299 0, and optimize the '<tt>and</tt>' to 0. Likewise, it is safe to assume that
2300 all the bits of the '<tt>undef</tt>' operand to the '<tt>or</tt>' could be
2301 set, allowing the '<tt>or</tt>' to be folded to -1.</p>
Chris Lattner48a109c2009-09-07 22:52:39 +00002302
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002303<pre class="doc_code">
Chris Lattner48a109c2009-09-07 22:52:39 +00002304 %A = select undef, %X, %Y
2305 %B = select undef, 42, %Y
2306 %C = select %X, %Y, undef
2307Safe:
2308 %A = %X (or %Y)
2309 %B = 42 (or %Y)
2310 %C = %Y
2311Unsafe:
2312 %A = undef
2313 %B = undef
2314 %C = undef
2315</pre>
Chris Lattner48a109c2009-09-07 22:52:39 +00002316
Bill Wendling1b383ba2010-10-27 01:07:41 +00002317<p>This set of examples shows that undefined '<tt>select</tt>' (and conditional
2318 branch) conditions can go <em>either way</em>, but they have to come from one
2319 of the two operands. In the <tt>%A</tt> example, if <tt>%X</tt> and
2320 <tt>%Y</tt> were both known to have a clear low bit, then <tt>%A</tt> would
2321 have to have a cleared low bit. However, in the <tt>%C</tt> example, the
2322 optimizer is allowed to assume that the '<tt>undef</tt>' operand could be the
2323 same as <tt>%Y</tt>, allowing the whole '<tt>select</tt>' to be
2324 eliminated.</p>
Chris Lattner48a109c2009-09-07 22:52:39 +00002325
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002326<pre class="doc_code">
Chris Lattner48a109c2009-09-07 22:52:39 +00002327 %A = xor undef, undef
Eric Christopher6c7e8a02009-12-05 02:46:03 +00002328
Chris Lattner48a109c2009-09-07 22:52:39 +00002329 %B = undef
2330 %C = xor %B, %B
2331
2332 %D = undef
2333 %E = icmp lt %D, 4
2334 %F = icmp gte %D, 4
2335
2336Safe:
2337 %A = undef
2338 %B = undef
2339 %C = undef
2340 %D = undef
2341 %E = undef
2342 %F = undef
2343</pre>
Chris Lattner48a109c2009-09-07 22:52:39 +00002344
Bill Wendling1b383ba2010-10-27 01:07:41 +00002345<p>This example points out that two '<tt>undef</tt>' operands are not
2346 necessarily the same. This can be surprising to people (and also matches C
2347 semantics) where they assume that "<tt>X^X</tt>" is always zero, even
2348 if <tt>X</tt> is undefined. This isn't true for a number of reasons, but the
2349 short answer is that an '<tt>undef</tt>' "variable" can arbitrarily change
2350 its value over its "live range". This is true because the variable doesn't
2351 actually <em>have a live range</em>. Instead, the value is logically read
2352 from arbitrary registers that happen to be around when needed, so the value
2353 is not necessarily consistent over time. In fact, <tt>%A</tt> and <tt>%C</tt>
2354 need to have the same semantics or the core LLVM "replace all uses with"
2355 concept would not hold.</p>
Chris Lattner6e9057b2009-09-07 23:33:52 +00002356
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002357<pre class="doc_code">
Chris Lattner6e9057b2009-09-07 23:33:52 +00002358 %A = fdiv undef, %X
2359 %B = fdiv %X, undef
2360Safe:
2361 %A = undef
2362b: unreachable
2363</pre>
Chris Lattner6e9057b2009-09-07 23:33:52 +00002364
2365<p>These examples show the crucial difference between an <em>undefined
Bill Wendling1b383ba2010-10-27 01:07:41 +00002366 value</em> and <em>undefined behavior</em>. An undefined value (like
2367 '<tt>undef</tt>') is allowed to have an arbitrary bit-pattern. This means that
2368 the <tt>%A</tt> operation can be constant folded to '<tt>undef</tt>', because
2369 the '<tt>undef</tt>' could be an SNaN, and <tt>fdiv</tt> is not (currently)
2370 defined on SNaN's. However, in the second example, we can make a more
2371 aggressive assumption: because the <tt>undef</tt> is allowed to be an
2372 arbitrary value, we are allowed to assume that it could be zero. Since a
2373 divide by zero has <em>undefined behavior</em>, we are allowed to assume that
2374 the operation does not execute at all. This allows us to delete the divide and
2375 all code after it. Because the undefined operation "can't happen", the
2376 optimizer can assume that it occurs in dead code.</p>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00002377
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002378<pre class="doc_code">
Chris Lattner6e9057b2009-09-07 23:33:52 +00002379a: store undef -> %X
2380b: store %X -> undef
2381Safe:
2382a: &lt;deleted&gt;
2383b: unreachable
2384</pre>
Chris Lattner6e9057b2009-09-07 23:33:52 +00002385
Bill Wendling1b383ba2010-10-27 01:07:41 +00002386<p>These examples reiterate the <tt>fdiv</tt> example: a store <em>of</em> an
2387 undefined value can be assumed to not have any effect; we can assume that the
2388 value is overwritten with bits that happen to match what was already there.
2389 However, a store <em>to</em> an undefined location could clobber arbitrary
2390 memory, therefore, it has undefined behavior.</p>
Chris Lattner6e9057b2009-09-07 23:33:52 +00002391
Chris Lattnerc3f59762004-12-09 17:30:23 +00002392</div>
2393
2394<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002395<h3>
2396 <a name="trapvalues">Trap Values</a>
2397</h3>
2398
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002399<div>
Dan Gohmanfff6c532010-04-22 23:14:21 +00002400
Dan Gohmanc68ce062010-04-26 20:21:21 +00002401<p>Trap values are similar to <a href="#undefvalues">undef values</a>, however
Dan Gohmanfff6c532010-04-22 23:14:21 +00002402 instead of representing an unspecified bit pattern, they represent the
2403 fact that an instruction or constant expression which cannot evoke side
2404 effects has nevertheless detected a condition which results in undefined
Dan Gohmanc68ce062010-04-26 20:21:21 +00002405 behavior.</p>
Dan Gohmanfff6c532010-04-22 23:14:21 +00002406
Dan Gohman34b3d992010-04-28 00:49:41 +00002407<p>There is currently no way of representing a trap value in the IR; they
Dan Gohman855abed2010-05-03 14:51:43 +00002408 only exist when produced by operations such as
Dan Gohman34b3d992010-04-28 00:49:41 +00002409 <a href="#i_add"><tt>add</tt></a> with the <tt>nsw</tt> flag.</p>
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002410
Dan Gohman34b3d992010-04-28 00:49:41 +00002411<p>Trap value behavior is defined in terms of value <i>dependence</i>:</p>
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002412
Dan Gohman34b3d992010-04-28 00:49:41 +00002413<ul>
2414<li>Values other than <a href="#i_phi"><tt>phi</tt></a> nodes depend on
2415 their operands.</li>
2416
2417<li><a href="#i_phi"><tt>Phi</tt></a> nodes depend on the operand corresponding
2418 to their dynamic predecessor basic block.</li>
2419
2420<li>Function arguments depend on the corresponding actual argument values in
2421 the dynamic callers of their functions.</li>
2422
2423<li><a href="#i_call"><tt>Call</tt></a> instructions depend on the
2424 <a href="#i_ret"><tt>ret</tt></a> instructions that dynamically transfer
2425 control back to them.</li>
2426
Dan Gohmanb5328162010-05-03 14:55:22 +00002427<li><a href="#i_invoke"><tt>Invoke</tt></a> instructions depend on the
2428 <a href="#i_ret"><tt>ret</tt></a>, <a href="#i_unwind"><tt>unwind</tt></a>,
2429 or exception-throwing call instructions that dynamically transfer control
2430 back to them.</li>
2431
Dan Gohman34b3d992010-04-28 00:49:41 +00002432<li>Non-volatile loads and stores depend on the most recent stores to all of the
2433 referenced memory addresses, following the order in the IR
2434 (including loads and stores implied by intrinsics such as
2435 <a href="#int_memcpy"><tt>@llvm.memcpy</tt></a>.)</li>
2436
Dan Gohman7c24ff12010-05-03 14:59:34 +00002437<!-- TODO: In the case of multiple threads, this only applies if the store
2438 "happens-before" the load or store. -->
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002439
Dan Gohman34b3d992010-04-28 00:49:41 +00002440<!-- TODO: floating-point exception state -->
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002441
Dan Gohman34b3d992010-04-28 00:49:41 +00002442<li>An instruction with externally visible side effects depends on the most
2443 recent preceding instruction with externally visible side effects, following
Dan Gohmanff70fe42010-07-06 15:26:33 +00002444 the order in the IR. (This includes
2445 <a href="#volatile">volatile operations</a>.)</li>
Dan Gohmanfff6c532010-04-22 23:14:21 +00002446
Dan Gohmanb5328162010-05-03 14:55:22 +00002447<li>An instruction <i>control-depends</i> on a
2448 <a href="#terminators">terminator instruction</a>
2449 if the terminator instruction has multiple successors and the instruction
2450 is always executed when control transfers to one of the successors, and
Chris Lattner7a2bdde2011-04-15 05:18:47 +00002451 may not be executed when control is transferred to another.</li>
Dan Gohman34b3d992010-04-28 00:49:41 +00002452
Dan Gohmanca4cac42011-04-12 23:05:59 +00002453<li>Additionally, an instruction also <i>control-depends</i> on a terminator
2454 instruction if the set of instructions it otherwise depends on would be
Chris Lattner7a2bdde2011-04-15 05:18:47 +00002455 different if the terminator had transferred control to a different
Dan Gohmanca4cac42011-04-12 23:05:59 +00002456 successor.</li>
2457
Dan Gohman34b3d992010-04-28 00:49:41 +00002458<li>Dependence is transitive.</li>
2459
2460</ul>
Dan Gohman34b3d992010-04-28 00:49:41 +00002461
2462<p>Whenever a trap value is generated, all values which depend on it evaluate
2463 to trap. If they have side effects, the evoke their side effects as if each
2464 operand with a trap value were undef. If they have externally-visible side
2465 effects, the behavior is undefined.</p>
2466
2467<p>Here are some examples:</p>
Dan Gohmanc30f6e12010-04-26 20:54:53 +00002468
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002469<pre class="doc_code">
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002470entry:
2471 %trap = sub nuw i32 0, 1 ; Results in a trap value.
Dan Gohman34b3d992010-04-28 00:49:41 +00002472 %still_trap = and i32 %trap, 0 ; Whereas (and i32 undef, 0) would return 0.
2473 %trap_yet_again = getelementptr i32* @h, i32 %still_trap
2474 store i32 0, i32* %trap_yet_again ; undefined behavior
2475
2476 store i32 %trap, i32* @g ; Trap value conceptually stored to memory.
2477 %trap2 = load i32* @g ; Returns a trap value, not just undef.
2478
2479 volatile store i32 %trap, i32* @g ; External observation; undefined behavior.
2480
2481 %narrowaddr = bitcast i32* @g to i16*
2482 %wideaddr = bitcast i32* @g to i64*
Nick Lewycky64f9fb12011-05-16 19:29:30 +00002483 %trap3 = load i16* %narrowaddr ; Returns a trap value.
2484 %trap4 = load i64* %wideaddr ; Returns a trap value.
Dan Gohman34b3d992010-04-28 00:49:41 +00002485
Nick Lewycky64f9fb12011-05-16 19:29:30 +00002486 %cmp = icmp slt i32 %trap, 0 ; Returns a trap value.
2487 br i1 %cmp, label %true, label %end ; Branch to either destination.
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002488
2489true:
Dan Gohman34b3d992010-04-28 00:49:41 +00002490 volatile store i32 0, i32* @g ; This is control-dependent on %cmp, so
2491 ; it has undefined behavior.
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002492 br label %end
2493
2494end:
2495 %p = phi i32 [ 0, %entry ], [ 1, %true ]
2496 ; Both edges into this PHI are
2497 ; control-dependent on %cmp, so this
Dan Gohman34b3d992010-04-28 00:49:41 +00002498 ; always results in a trap value.
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002499
Dan Gohmanca4cac42011-04-12 23:05:59 +00002500 volatile store i32 0, i32* @g ; This would depend on the store in %true
2501 ; if %cmp is true, or the store in %entry
2502 ; otherwise, so this is undefined behavior.
2503
Nick Lewycky64f9fb12011-05-16 19:29:30 +00002504 br i1 %cmp, label %second_true, label %second_end
Dan Gohmanca4cac42011-04-12 23:05:59 +00002505 ; The same branch again, but this time the
2506 ; true block doesn't have side effects.
2507
2508second_true:
2509 ; No side effects!
Nick Lewycky64f9fb12011-05-16 19:29:30 +00002510 ret void
Dan Gohmanca4cac42011-04-12 23:05:59 +00002511
2512second_end:
2513 volatile store i32 0, i32* @g ; This time, the instruction always depends
2514 ; on the store in %end. Also, it is
2515 ; control-equivalent to %end, so this is
Nick Lewycky64f9fb12011-05-16 19:29:30 +00002516 ; well-defined (again, ignoring earlier
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002517 ; undefined behavior in this example).
Dan Gohmanae11c3f2010-04-26 23:36:52 +00002518</pre>
Dan Gohmanfff6c532010-04-22 23:14:21 +00002519
Dan Gohmanfff6c532010-04-22 23:14:21 +00002520</div>
2521
2522<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002523<h3>
2524 <a name="blockaddress">Addresses of Basic Blocks</a>
2525</h3>
2526
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002527<div>
Chris Lattnerc6f44362009-10-27 21:01:34 +00002528
Chris Lattnercdfc9402009-11-01 01:27:45 +00002529<p><b><tt>blockaddress(@function, %block)</tt></b></p>
Chris Lattnerc6f44362009-10-27 21:01:34 +00002530
2531<p>The '<tt>blockaddress</tt>' constant computes the address of the specified
Chris Lattner2dfdf2a2009-10-27 21:49:40 +00002532 basic block in the specified function, and always has an i8* type. Taking
Chris Lattnercdfc9402009-11-01 01:27:45 +00002533 the address of the entry block is illegal.</p>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00002534
Chris Lattnerc6f44362009-10-27 21:01:34 +00002535<p>This value only has defined behavior when used as an operand to the
Bill Wendling1b383ba2010-10-27 01:07:41 +00002536 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>' instruction, or for
2537 comparisons against null. Pointer equality tests between labels addresses
2538 results in undefined behavior &mdash; though, again, comparison against null
2539 is ok, and no label is equal to the null pointer. This may be passed around
2540 as an opaque pointer sized value as long as the bits are not inspected. This
2541 allows <tt>ptrtoint</tt> and arithmetic to be performed on these values so
2542 long as the original value is reconstituted before the <tt>indirectbr</tt>
2543 instruction.</p>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00002544
Bill Wendling1b383ba2010-10-27 01:07:41 +00002545<p>Finally, some targets may provide defined semantics when using the value as
2546 the operand to an inline assembly, but that is target specific.</p>
Chris Lattnerc6f44362009-10-27 21:01:34 +00002547
2548</div>
2549
2550
2551<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002552<h3>
2553 <a name="constantexprs">Constant Expressions</a>
2554</h3>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002555
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002556<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002557
2558<p>Constant expressions are used to allow expressions involving other constants
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002559 to be used as constants. Constant expressions may be of
2560 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2561 operation that does not have side effects (e.g. load and call are not
Bill Wendling1b383ba2010-10-27 01:07:41 +00002562 supported). The following is the syntax for constant expressions:</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002563
2564<dl>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002565 <dt><b><tt>trunc (CST to TYPE)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002566 <dd>Truncate a constant to another type. The bit size of CST must be larger
2567 than the bit size of TYPE. Both types must be integers.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002568
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002569 <dt><b><tt>zext (CST to TYPE)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002570 <dd>Zero extend a constant to another type. The bit size of CST must be
Duncan Sands28afd432010-07-13 12:06:14 +00002571 smaller than the bit size of TYPE. Both types must be integers.</dd>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002572
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002573 <dt><b><tt>sext (CST to TYPE)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002574 <dd>Sign extend a constant to another type. The bit size of CST must be
Duncan Sands28afd432010-07-13 12:06:14 +00002575 smaller than the bit size of TYPE. Both types must be integers.</dd>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002576
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002577 <dt><b><tt>fptrunc (CST to TYPE)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002578 <dd>Truncate a floating point constant to another floating point type. The
2579 size of CST must be larger than the size of TYPE. Both types must be
2580 floating point.</dd>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002581
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002582 <dt><b><tt>fpext (CST to TYPE)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002583 <dd>Floating point extend a constant to another type. The size of CST must be
2584 smaller or equal to the size of TYPE. Both types must be floating
2585 point.</dd>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002586
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002587 <dt><b><tt>fptoui (CST to TYPE)</tt></b></dt>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002588 <dd>Convert a floating point constant to the corresponding unsigned integer
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002589 constant. TYPE must be a scalar or vector integer type. CST must be of
2590 scalar or vector floating point type. Both CST and TYPE must be scalars,
2591 or vectors of the same number of elements. If the value won't fit in the
2592 integer type, the results are undefined.</dd>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002593
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002594 <dt><b><tt>fptosi (CST to TYPE)</tt></b></dt>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002595 <dd>Convert a floating point constant to the corresponding signed integer
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002596 constant. TYPE must be a scalar or vector integer type. CST must be of
2597 scalar or vector floating point type. Both CST and TYPE must be scalars,
2598 or vectors of the same number of elements. If the value won't fit in the
2599 integer type, the results are undefined.</dd>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002600
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002601 <dt><b><tt>uitofp (CST to TYPE)</tt></b></dt>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002602 <dd>Convert an unsigned integer constant to the corresponding floating point
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002603 constant. TYPE must be a scalar or vector floating point type. CST must be
2604 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2605 vectors of the same number of elements. If the value won't fit in the
2606 floating point type, the results are undefined.</dd>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002607
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002608 <dt><b><tt>sitofp (CST to TYPE)</tt></b></dt>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002609 <dd>Convert a signed integer constant to the corresponding floating point
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002610 constant. TYPE must be a scalar or vector floating point type. CST must be
2611 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2612 vectors of the same number of elements. If the value won't fit in the
2613 floating point type, the results are undefined.</dd>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00002614
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002615 <dt><b><tt>ptrtoint (CST to TYPE)</tt></b></dt>
Reid Spencer5c0ef472006-11-11 23:08:07 +00002616 <dd>Convert a pointer typed constant to the corresponding integer constant
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002617 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2618 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2619 make it fit in <tt>TYPE</tt>.</dd>
Reid Spencer5c0ef472006-11-11 23:08:07 +00002620
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002621 <dt><b><tt>inttoptr (CST to TYPE)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002622 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2623 type. CST must be of integer type. The CST value is zero extended,
2624 truncated, or unchanged to make it fit in a pointer size. This one is
2625 <i>really</i> dangerous!</dd>
Reid Spencer5c0ef472006-11-11 23:08:07 +00002626
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002627 <dt><b><tt>bitcast (CST to TYPE)</tt></b></dt>
Chris Lattner03bbad62009-02-28 18:27:03 +00002628 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2629 are the same as those for the <a href="#i_bitcast">bitcast
2630 instruction</a>.</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002631
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002632 <dt><b><tt>getelementptr (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2633 <dt><b><tt>getelementptr inbounds (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002634 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002635 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2636 instruction, the index list may have zero or more indexes, which are
2637 required to make sense for the type of "CSTPTR".</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002638
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002639 <dt><b><tt>select (COND, VAL1, VAL2)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002640 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
Reid Spencer01c42592006-12-04 19:23:19 +00002641
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002642 <dt><b><tt>icmp COND (VAL1, VAL2)</tt></b></dt>
Reid Spencer01c42592006-12-04 19:23:19 +00002643 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2644
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002645 <dt><b><tt>fcmp COND (VAL1, VAL2)</tt></b></dt>
Reid Spencer01c42592006-12-04 19:23:19 +00002646 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
Robert Bocchino9fbe1452006-01-10 19:31:34 +00002647
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002648 <dt><b><tt>extractelement (VAL, IDX)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002649 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2650 constants.</dd>
Robert Bocchino9fbe1452006-01-10 19:31:34 +00002651
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002652 <dt><b><tt>insertelement (VAL, ELT, IDX)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002653 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2654 constants.</dd>
Chris Lattnerc1989542006-04-08 00:13:41 +00002655
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002656 <dt><b><tt>shufflevector (VEC1, VEC2, IDXMASK)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002657 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2658 constants.</dd>
Chris Lattnerc1989542006-04-08 00:13:41 +00002659
Nick Lewycky9e130ce2010-05-29 06:44:15 +00002660 <dt><b><tt>extractvalue (VAL, IDX0, IDX1, ...)</tt></b></dt>
2661 <dd>Perform the <a href="#i_extractvalue">extractvalue operation</a> on
2662 constants. The index list is interpreted in a similar manner as indices in
2663 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2664 index value must be specified.</dd>
2665
2666 <dt><b><tt>insertvalue (VAL, ELT, IDX0, IDX1, ...)</tt></b></dt>
2667 <dd>Perform the <a href="#i_insertvalue">insertvalue operation</a> on
2668 constants. The index list is interpreted in a similar manner as indices in
2669 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2670 index value must be specified.</dd>
2671
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00002672 <dt><b><tt>OPCODE (LHS, RHS)</tt></b></dt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002673 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2674 be any of the <a href="#binaryops">binary</a>
2675 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2676 on operands are the same as those for the corresponding instruction
2677 (e.g. no bitwise operations on floating point values are allowed).</dd>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002678</dl>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002679
Chris Lattnerc3f59762004-12-09 17:30:23 +00002680</div>
Chris Lattner9ee5d222004-03-08 16:49:10 +00002681
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002682</div>
2683
Chris Lattner00950542001-06-06 20:29:01 +00002684<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002685<h2><a name="othervalues">Other Values</a></h2>
Chris Lattnere87d6532006-01-25 23:47:57 +00002686<!-- *********************************************************************** -->
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002687<div>
Chris Lattnere87d6532006-01-25 23:47:57 +00002688<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002689<h3>
Chris Lattnere87d6532006-01-25 23:47:57 +00002690<a name="inlineasm">Inline Assembler Expressions</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002691</h3>
Chris Lattnere87d6532006-01-25 23:47:57 +00002692
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002693<div>
Chris Lattnere87d6532006-01-25 23:47:57 +00002694
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002695<p>LLVM supports inline assembler expressions (as opposed
2696 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2697 a special value. This value represents the inline assembler as a string
2698 (containing the instructions to emit), a list of operand constraints (stored
Dale Johannesen09fed252009-10-13 21:56:55 +00002699 as a string), a flag that indicates whether or not the inline asm
Dale Johannesen8ba2d5b2009-10-21 23:28:00 +00002700 expression has side effects, and a flag indicating whether the function
2701 containing the asm needs to align its stack conservatively. An example
2702 inline assembler expression is:</p>
Chris Lattnere87d6532006-01-25 23:47:57 +00002703
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002704<pre class="doc_code">
Bill Wendling2f7a8b02007-05-29 09:04:49 +00002705i32 (i32) asm "bswap $0", "=r,r"
Chris Lattnere87d6532006-01-25 23:47:57 +00002706</pre>
2707
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002708<p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2709 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2710 have:</p>
Chris Lattnere87d6532006-01-25 23:47:57 +00002711
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002712<pre class="doc_code">
Bill Wendling2f7a8b02007-05-29 09:04:49 +00002713%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
Chris Lattnere87d6532006-01-25 23:47:57 +00002714</pre>
2715
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002716<p>Inline asms with side effects not visible in the constraint list must be
2717 marked as having side effects. This is done through the use of the
2718 '<tt>sideeffect</tt>' keyword, like so:</p>
Chris Lattnere87d6532006-01-25 23:47:57 +00002719
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002720<pre class="doc_code">
Bill Wendling2f7a8b02007-05-29 09:04:49 +00002721call void asm sideeffect "eieio", ""()
Chris Lattnere87d6532006-01-25 23:47:57 +00002722</pre>
2723
Dale Johannesen8ba2d5b2009-10-21 23:28:00 +00002724<p>In some cases inline asms will contain code that will not work unless the
2725 stack is aligned in some way, such as calls or SSE instructions on x86,
2726 yet will not contain code that does that alignment within the asm.
2727 The compiler should make conservative assumptions about what the asm might
2728 contain and should generate its usual stack alignment code in the prologue
2729 if the '<tt>alignstack</tt>' keyword is present:</p>
Dale Johannesen09fed252009-10-13 21:56:55 +00002730
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002731<pre class="doc_code">
Dale Johannesen8ba2d5b2009-10-21 23:28:00 +00002732call void asm alignstack "eieio", ""()
Dale Johannesen09fed252009-10-13 21:56:55 +00002733</pre>
Dale Johannesen09fed252009-10-13 21:56:55 +00002734
2735<p>If both keywords appear the '<tt>sideeffect</tt>' keyword must come
2736 first.</p>
2737
Chris Lattnere87d6532006-01-25 23:47:57 +00002738<p>TODO: The format of the asm and constraints string still need to be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002739 documented here. Constraints on what can be done (e.g. duplication, moving,
2740 etc need to be documented). This is probably best done by reference to
2741 another document that covers inline asm from a holistic perspective.</p>
Chris Lattnercf9a4152010-04-07 05:38:05 +00002742
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002743<h4>
Chris Lattnercf9a4152010-04-07 05:38:05 +00002744<a name="inlineasm_md">Inline Asm Metadata</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002745</h4>
Chris Lattnercf9a4152010-04-07 05:38:05 +00002746
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002747<div>
Chris Lattnercf9a4152010-04-07 05:38:05 +00002748
2749<p>The call instructions that wrap inline asm nodes may have a "!srcloc" MDNode
Chris Lattnerce1b9ad2010-11-17 08:20:42 +00002750 attached to it that contains a list of constant integers. If present, the
2751 code generator will use the integer as the location cookie value when report
Chris Lattnercf9a4152010-04-07 05:38:05 +00002752 errors through the LLVMContext error reporting mechanisms. This allows a
Dan Gohman1c70c002010-04-28 00:36:01 +00002753 front-end to correlate backend errors that occur with inline asm back to the
Chris Lattnercf9a4152010-04-07 05:38:05 +00002754 source code that produced it. For example:</p>
2755
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002756<pre class="doc_code">
Chris Lattnercf9a4152010-04-07 05:38:05 +00002757call void asm sideeffect "something bad", ""()<b>, !srcloc !42</b>
2758...
2759!42 = !{ i32 1234567 }
2760</pre>
Chris Lattnercf9a4152010-04-07 05:38:05 +00002761
2762<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 +00002763 IR. If the MDNode contains multiple constants, the code generator will use
2764 the one that corresponds to the line of the asm that the error occurs on.</p>
Chris Lattnere87d6532006-01-25 23:47:57 +00002765
2766</div>
2767
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002768</div>
2769
Chris Lattnere6a5ddd2010-01-15 21:50:19 +00002770<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002771<h3>
2772 <a name="metadata">Metadata Nodes and Metadata Strings</a>
2773</h3>
Chris Lattnere6a5ddd2010-01-15 21:50:19 +00002774
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002775<div>
Chris Lattnere6a5ddd2010-01-15 21:50:19 +00002776
2777<p>LLVM IR allows metadata to be attached to instructions in the program that
2778 can convey extra information about the code to the optimizers and code
2779 generator. One example application of metadata is source-level debug
2780 information. There are two metadata primitives: strings and nodes. All
2781 metadata has the <tt>metadata</tt> type and is identified in syntax by a
2782 preceding exclamation point ('<tt>!</tt>').</p>
2783
2784<p>A metadata string is a string surrounded by double quotes. It can contain
2785 any character by escaping non-printable characters with "\xx" where "xx" is
2786 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
2787
2788<p>Metadata nodes are represented with notation similar to structure constants
2789 (a comma separated list of elements, surrounded by braces and preceded by an
2790 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2791 10}</tt>". Metadata nodes can have any values as their operand.</p>
2792
2793<p>A <a href="#namedmetadatastructure">named metadata</a> is a collection of
2794 metadata nodes, which can be looked up in the module symbol table. For
2795 example: "<tt>!foo = metadata !{!4, !3}</tt>".
2796
Devang Patele1d50cd2010-03-04 23:44:48 +00002797<p>Metadata can be used as function arguments. Here <tt>llvm.dbg.value</tt>
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002798 function is using two metadata arguments.</p>
Devang Patele1d50cd2010-03-04 23:44:48 +00002799
Bill Wendling9ff5de92011-03-02 02:17:11 +00002800<div class="doc_code">
2801<pre>
2802call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
2803</pre>
2804</div>
Devang Patele1d50cd2010-03-04 23:44:48 +00002805
2806<p>Metadata can be attached with an instruction. Here metadata <tt>!21</tt> is
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00002807 attached with <tt>add</tt> instruction using <tt>!dbg</tt> identifier.</p>
Devang Patele1d50cd2010-03-04 23:44:48 +00002808
Bill Wendling9ff5de92011-03-02 02:17:11 +00002809<div class="doc_code">
2810<pre>
2811%indvar.next = add i64 %indvar, 1, !dbg !21
2812</pre>
2813</div>
2814
Chris Lattnere6a5ddd2010-01-15 21:50:19 +00002815</div>
2816
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002817</div>
Chris Lattner857755c2009-07-20 05:55:19 +00002818
2819<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002820<h2>
Chris Lattner857755c2009-07-20 05:55:19 +00002821 <a name="intrinsic_globals">Intrinsic Global Variables</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002822</h2>
Chris Lattner857755c2009-07-20 05:55:19 +00002823<!-- *********************************************************************** -->
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002824<div>
Chris Lattner857755c2009-07-20 05:55:19 +00002825<p>LLVM has a number of "magic" global variables that contain data that affect
2826code generation or other IR semantics. These are documented here. All globals
Chris Lattner401e10c2009-07-20 06:14:25 +00002827of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2828section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2829by LLVM.</p>
Chris Lattner857755c2009-07-20 05:55:19 +00002830
2831<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002832<h3>
Chris Lattner857755c2009-07-20 05:55:19 +00002833<a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002834</h3>
Chris Lattner857755c2009-07-20 05:55:19 +00002835
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002836<div>
Chris Lattner857755c2009-07-20 05:55:19 +00002837
2838<p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2839href="#linkage_appending">appending linkage</a>. This array contains a list of
2840pointers to global variables and functions which may optionally have a pointer
2841cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2842
2843<pre>
2844 @X = global i8 4
2845 @Y = global i32 123
2846
2847 @llvm.used = appending global [2 x i8*] [
2848 i8* @X,
2849 i8* bitcast (i32* @Y to i8*)
2850 ], section "llvm.metadata"
2851</pre>
2852
2853<p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2854compiler, assembler, and linker are required to treat the symbol as if there is
2855a reference to the global that it cannot see. For example, if a variable has
2856internal linkage and no references other than that from the <tt>@llvm.used</tt>
2857list, it cannot be deleted. This is commonly used to represent references from
2858inline asms and other things the compiler cannot "see", and corresponds to
2859"attribute((used))" in GNU C.</p>
2860
2861<p>On some targets, the code generator must emit a directive to the assembler or
2862object file to prevent the assembler and linker from molesting the symbol.</p>
2863
2864</div>
2865
2866<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002867<h3>
2868 <a name="intg_compiler_used">
2869 The '<tt>llvm.compiler.used</tt>' Global Variable
2870 </a>
2871</h3>
Chris Lattner401e10c2009-07-20 06:14:25 +00002872
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002873<div>
Chris Lattner401e10c2009-07-20 06:14:25 +00002874
2875<p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2876<tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2877touching the symbol. On targets that support it, this allows an intelligent
2878linker to optimize references to the symbol without being impeded as it would be
2879by <tt>@llvm.used</tt>.</p>
2880
2881<p>This is a rare construct that should only be used in rare circumstances, and
2882should not be exposed to source languages.</p>
2883
2884</div>
2885
2886<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002887<h3>
Chris Lattner857755c2009-07-20 05:55:19 +00002888<a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002889</h3>
Chris Lattner857755c2009-07-20 05:55:19 +00002890
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002891<div>
David Chisnalle31e9962010-04-30 19:23:49 +00002892<pre>
2893%0 = type { i32, void ()* }
David Chisnall27195a52010-04-30 19:27:35 +00002894@llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor }]
David Chisnalle31e9962010-04-30 19:23:49 +00002895</pre>
2896<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.
2897</p>
Chris Lattner857755c2009-07-20 05:55:19 +00002898
2899</div>
2900
2901<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002902<h3>
Chris Lattner857755c2009-07-20 05:55:19 +00002903<a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002904</h3>
Chris Lattner857755c2009-07-20 05:55:19 +00002905
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002906<div>
David Chisnalle31e9962010-04-30 19:23:49 +00002907<pre>
2908%0 = type { i32, void ()* }
David Chisnall27195a52010-04-30 19:27:35 +00002909@llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor }]
David Chisnalle31e9962010-04-30 19:23:49 +00002910</pre>
Chris Lattner857755c2009-07-20 05:55:19 +00002911
David Chisnalle31e9962010-04-30 19:23:49 +00002912<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.
2913</p>
Chris Lattner857755c2009-07-20 05:55:19 +00002914
2915</div>
2916
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002917</div>
Chris Lattner857755c2009-07-20 05:55:19 +00002918
Chris Lattnere87d6532006-01-25 23:47:57 +00002919<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002920<h2><a name="instref">Instruction Reference</a></h2>
Chris Lattner261efe92003-11-25 01:02:51 +00002921<!-- *********************************************************************** -->
Chris Lattnerc3f59762004-12-09 17:30:23 +00002922
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002923<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002924
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002925<p>The LLVM instruction set consists of several different classifications of
2926 instructions: <a href="#terminators">terminator
2927 instructions</a>, <a href="#binaryops">binary instructions</a>,
2928 <a href="#bitwiseops">bitwise binary instructions</a>,
2929 <a href="#memoryops">memory instructions</a>, and
2930 <a href="#otherops">other instructions</a>.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002931
Chris Lattner00950542001-06-06 20:29:01 +00002932<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002933<h3>
2934 <a name="terminators">Terminator Instructions</a>
2935</h3>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002936
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002937<div>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002938
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002939<p>As mentioned <a href="#functionstructure">previously</a>, every basic block
2940 in a program ends with a "Terminator" instruction, which indicates which
2941 block should be executed after the current block is finished. These
2942 terminator instructions typically yield a '<tt>void</tt>' value: they produce
2943 control flow, not values (the one exception being the
2944 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2945
Bill Wendling772fe172011-07-27 20:18:04 +00002946<p>There are eight different terminator instructions: the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002947 '<a href="#i_ret"><tt>ret</tt></a>' instruction, the
2948 '<a href="#i_br"><tt>br</tt></a>' instruction, the
2949 '<a href="#i_switch"><tt>switch</tt></a>' instruction, the
Bill Wendling21c346e2009-11-02 00:25:26 +00002950 '<a href="#i_indirectbr">'<tt>indirectbr</tt></a>' Instruction, the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002951 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the
Bill Wendling772fe172011-07-27 20:18:04 +00002952 '<a href="#i_unwind"><tt>unwind</tt></a>' instruction, the
2953 '<a href="#i_resume"><tt>resume</tt></a>' instruction, and the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002954 '<a href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
Chris Lattnerc3f59762004-12-09 17:30:23 +00002955
Chris Lattner00950542001-06-06 20:29:01 +00002956<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00002957<h4>
2958 <a name="i_ret">'<tt>ret</tt>' Instruction</a>
2959</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002960
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00002961<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002962
Chris Lattner00950542001-06-06 20:29:01 +00002963<h5>Syntax:</h5>
Dan Gohmanb1e6b962008-10-04 19:00:07 +00002964<pre>
2965 ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
Chris Lattner7faa8832002-04-14 06:13:44 +00002966 ret void <i>; Return from void function</i>
Chris Lattner00950542001-06-06 20:29:01 +00002967</pre>
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00002968
Chris Lattner00950542001-06-06 20:29:01 +00002969<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002970<p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
2971 a value) from a function back to the caller.</p>
2972
2973<p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
2974 value and then causes control flow, and one that just causes control flow to
2975 occur.</p>
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00002976
Chris Lattner00950542001-06-06 20:29:01 +00002977<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002978<p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
2979 return value. The type of the return value must be a
2980 '<a href="#t_firstclass">first class</a>' type.</p>
Dan Gohmanb1e6b962008-10-04 19:00:07 +00002981
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002982<p>A function is not <a href="#wellformed">well formed</a> if it it has a
2983 non-void return type and contains a '<tt>ret</tt>' instruction with no return
2984 value or a return value with a type that does not match its type, or if it
2985 has a void return type and contains a '<tt>ret</tt>' instruction with a
2986 return value.</p>
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00002987
Chris Lattner00950542001-06-06 20:29:01 +00002988<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00002989<p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
2990 the calling function's context. If the caller is a
2991 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
2992 instruction after the call. If the caller was an
2993 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
2994 the beginning of the "normal" destination block. If the instruction returns
2995 a value, that value shall set the call or invoke instruction's return
2996 value.</p>
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00002997
Chris Lattner00950542001-06-06 20:29:01 +00002998<h5>Example:</h5>
Chris Lattnerf4cde4e2008-04-23 04:59:35 +00002999<pre>
3000 ret i32 5 <i>; Return an integer value of 5</i>
Chris Lattner7faa8832002-04-14 06:13:44 +00003001 ret void <i>; Return from a void function</i>
Bill Wendling0a4bbbf2009-02-28 22:12:54 +00003002 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
Chris Lattner00950542001-06-06 20:29:01 +00003003</pre>
Dan Gohmanf3e60bd2009-01-12 23:12:39 +00003004
Misha Brukman9d0919f2003-11-08 01:05:38 +00003005</div>
Chris Lattner00950542001-06-06 20:29:01 +00003006<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003007<h4>
3008 <a name="i_br">'<tt>br</tt>' Instruction</a>
3009</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003010
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003011<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003012
Chris Lattner00950542001-06-06 20:29:01 +00003013<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003014<pre>
Bill Wendlingb3aa4712011-07-26 10:41:15 +00003015 br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;
3016 br label &lt;dest&gt; <i>; Unconditional branch</i>
Chris Lattner00950542001-06-06 20:29:01 +00003017</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003018
Chris Lattner00950542001-06-06 20:29:01 +00003019<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003020<p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
3021 different basic block in the current function. There are two forms of this
3022 instruction, corresponding to a conditional branch and an unconditional
3023 branch.</p>
3024
Chris Lattner00950542001-06-06 20:29:01 +00003025<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003026<p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
3027 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
3028 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
3029 target.</p>
3030
Chris Lattner00950542001-06-06 20:29:01 +00003031<h5>Semantics:</h5>
Reid Spencerc78f3372007-01-12 03:35:51 +00003032<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003033 argument is evaluated. If the value is <tt>true</tt>, control flows to the
3034 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
3035 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
3036
Chris Lattner00950542001-06-06 20:29:01 +00003037<h5>Example:</h5>
Bill Wendlingc39e3e02009-07-20 02:39:26 +00003038<pre>
3039Test:
3040 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
3041 br i1 %cond, label %IfEqual, label %IfUnequal
3042IfEqual:
3043 <a href="#i_ret">ret</a> i32 1
3044IfUnequal:
3045 <a href="#i_ret">ret</a> i32 0
3046</pre>
3047
Misha Brukman9d0919f2003-11-08 01:05:38 +00003048</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003049
Chris Lattner00950542001-06-06 20:29:01 +00003050<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003051<h4>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003052 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003053</h4>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003054
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003055<div>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003056
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003057<h5>Syntax:</h5>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003058<pre>
3059 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
3060</pre>
3061
Chris Lattner00950542001-06-06 20:29:01 +00003062<h5>Overview:</h5>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003063<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003064 several different places. It is a generalization of the '<tt>br</tt>'
3065 instruction, allowing a branch to occur to one of many possible
3066 destinations.</p>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003067
Chris Lattner00950542001-06-06 20:29:01 +00003068<h5>Arguments:</h5>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003069<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003070 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
3071 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
3072 The table is not allowed to contain duplicate constant entries.</p>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003073
Chris Lattner00950542001-06-06 20:29:01 +00003074<h5>Semantics:</h5>
Chris Lattner261efe92003-11-25 01:02:51 +00003075<p>The <tt>switch</tt> instruction specifies a table of values and
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003076 destinations. When the '<tt>switch</tt>' instruction is executed, this table
3077 is searched for the given value. If the value is found, control flow is
Benjamin Kramer8040cd32009-10-12 14:46:08 +00003078 transferred to the corresponding destination; otherwise, control flow is
3079 transferred to the default destination.</p>
Chris Lattner00950542001-06-06 20:29:01 +00003080
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003081<h5>Implementation:</h5>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003082<p>Depending on properties of the target machine and the particular
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003083 <tt>switch</tt> instruction, this instruction may be code generated in
3084 different ways. For example, it could be generated as a series of chained
3085 conditional branches or with a lookup table.</p>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003086
3087<h5>Example:</h5>
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003088<pre>
3089 <i>; Emulate a conditional br instruction</i>
Reid Spencerc78f3372007-01-12 03:35:51 +00003090 %Val = <a href="#i_zext">zext</a> i1 %value to i32
Dan Gohman2a08c532009-01-04 23:44:43 +00003091 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003092
3093 <i>; Emulate an unconditional br instruction</i>
Reid Spencerca86e162006-12-31 07:07:53 +00003094 switch i32 0, label %dest [ ]
Chris Lattnerc88c17b2004-02-24 04:54:45 +00003095
3096 <i>; Implement a jump table:</i>
Dan Gohman2a08c532009-01-04 23:44:43 +00003097 switch i32 %val, label %otherwise [ i32 0, label %onzero
3098 i32 1, label %onone
3099 i32 2, label %ontwo ]
Chris Lattner00950542001-06-06 20:29:01 +00003100</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003101
Misha Brukman9d0919f2003-11-08 01:05:38 +00003102</div>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003103
Chris Lattnerf9be95f2009-10-27 19:13:16 +00003104
3105<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003106<h4>
Chris Lattnerab21db72009-10-28 00:19:10 +00003107 <a name="i_indirectbr">'<tt>indirectbr</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003108</h4>
Chris Lattnerf9be95f2009-10-27 19:13:16 +00003109
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003110<div>
Chris Lattnerf9be95f2009-10-27 19:13:16 +00003111
3112<h5>Syntax:</h5>
3113<pre>
Chris Lattnerab21db72009-10-28 00:19:10 +00003114 indirectbr &lt;somety&gt;* &lt;address&gt;, [ label &lt;dest1&gt;, label &lt;dest2&gt;, ... ]
Chris Lattnerf9be95f2009-10-27 19:13:16 +00003115</pre>
3116
3117<h5>Overview:</h5>
3118
Chris Lattnerab21db72009-10-28 00:19:10 +00003119<p>The '<tt>indirectbr</tt>' instruction implements an indirect branch to a label
Chris Lattnerf9be95f2009-10-27 19:13:16 +00003120 within the current function, whose address is specified by
Chris Lattnerc6f44362009-10-27 21:01:34 +00003121 "<tt>address</tt>". Address must be derived from a <a
3122 href="#blockaddress">blockaddress</a> constant.</p>
Chris Lattnerf9be95f2009-10-27 19:13:16 +00003123
3124<h5>Arguments:</h5>
3125
3126<p>The '<tt>address</tt>' argument is the address of the label to jump to. The
3127 rest of the arguments indicate the full set of possible destinations that the
3128 address may point to. Blocks are allowed to occur multiple times in the
3129 destination list, though this isn't particularly useful.</p>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003130
Chris Lattnerf9be95f2009-10-27 19:13:16 +00003131<p>This destination list is required so that dataflow analysis has an accurate
3132 understanding of the CFG.</p>
3133
3134<h5>Semantics:</h5>
3135
3136<p>Control transfers to the block specified in the address argument. All
3137 possible destination blocks must be listed in the label list, otherwise this
3138 instruction has undefined behavior. This implies that jumps to labels
3139 defined in other functions have undefined behavior as well.</p>
3140
3141<h5>Implementation:</h5>
3142
3143<p>This is typically implemented with a jump through a register.</p>
3144
3145<h5>Example:</h5>
3146<pre>
Chris Lattnerab21db72009-10-28 00:19:10 +00003147 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
Chris Lattnerf9be95f2009-10-27 19:13:16 +00003148</pre>
3149
3150</div>
3151
3152
Chris Lattner00950542001-06-06 20:29:01 +00003153<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003154<h4>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003155 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003156</h4>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003157
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003158<div>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003159
Chris Lattner00950542001-06-06 20:29:01 +00003160<h5>Syntax:</h5>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003161<pre>
Devang Patel307e8ab2008-10-07 17:48:33 +00003162 &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 +00003163 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003164</pre>
3165
Chris Lattner6536cfe2002-05-06 22:08:29 +00003166<h5>Overview:</h5>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003167<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003168 function, with the possibility of control flow transfer to either the
3169 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
3170 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
3171 control flow will return to the "normal" label. If the callee (or any
3172 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
3173 instruction, control is interrupted and continued at the dynamically nearest
3174 "exception" label.</p>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003175
Chris Lattner00950542001-06-06 20:29:01 +00003176<h5>Arguments:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00003177<p>This instruction requires several arguments:</p>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003178
Chris Lattner00950542001-06-06 20:29:01 +00003179<ol>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003180 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
3181 convention</a> the call should use. If none is specified, the call
3182 defaults to using C calling conventions.</li>
Devang Patelf642f472008-10-06 18:50:38 +00003183
3184 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003185 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
3186 '<tt>inreg</tt>' attributes are valid here.</li>
Devang Patelf642f472008-10-06 18:50:38 +00003187
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003188 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003189 function value being invoked. In most cases, this is a direct function
3190 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
3191 off an arbitrary pointer to function value.</li>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003192
3193 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003194 function to be invoked. </li>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003195
3196 <li>'<tt>function args</tt>': argument list whose types match the function
Chris Lattner0724fbd2010-03-02 06:36:51 +00003197 signature argument types and parameter attributes. All arguments must be
3198 of <a href="#t_firstclass">first class</a> type. If the function
3199 signature indicates the function accepts a variable number of arguments,
3200 the extra arguments can be specified.</li>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003201
3202 <li>'<tt>normal label</tt>': the label reached when the called function
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003203 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003204
3205 <li>'<tt>exception label</tt>': the label reached when a callee returns with
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003206 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003207
Devang Patel307e8ab2008-10-07 17:48:33 +00003208 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003209 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
3210 '<tt>readnone</tt>' attributes are valid here.</li>
Chris Lattner00950542001-06-06 20:29:01 +00003211</ol>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003212
Chris Lattner00950542001-06-06 20:29:01 +00003213<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003214<p>This instruction is designed to operate as a standard
3215 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
3216 primary difference is that it establishes an association with a label, which
3217 is used by the runtime library to unwind the stack.</p>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003218
3219<p>This instruction is used in languages with destructors to ensure that proper
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003220 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
3221 exception. Additionally, this is important for implementation of
3222 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003223
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003224<p>For the purposes of the SSA form, the definition of the value returned by the
3225 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
3226 block to the "normal" label. If the callee unwinds then no return value is
3227 available.</p>
Dan Gohmanf96a4992009-05-22 21:47:08 +00003228
Chris Lattnerdf7a6802010-01-15 18:08:37 +00003229<p>Note that the code generator does not yet completely support unwind, and
3230that the invoke/unwind semantics are likely to change in future versions.</p>
3231
Chris Lattner00950542001-06-06 20:29:01 +00003232<h5>Example:</h5>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00003233<pre>
Nick Lewyckyd703f652008-03-16 07:18:12 +00003234 %retval = invoke i32 @Test(i32 15) to label %Continue
Jeff Cohen6f1cc772007-04-22 01:17:39 +00003235 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewyckyd703f652008-03-16 07:18:12 +00003236 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Jeff Cohen6f1cc772007-04-22 01:17:39 +00003237 unwind label %TestCleanup <i>; {i32}:retval set</i>
Chris Lattner00950542001-06-06 20:29:01 +00003238</pre>
Chris Lattner35eca582004-10-16 18:04:13 +00003239
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003240</div>
Chris Lattner35eca582004-10-16 18:04:13 +00003241
Chris Lattner27f71f22003-09-03 00:41:47 +00003242<!-- _______________________________________________________________________ -->
Chris Lattner35eca582004-10-16 18:04:13 +00003243
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003244<h4>
3245 <a name="i_unwind">'<tt>unwind</tt>' Instruction</a>
3246</h4>
Chris Lattner35eca582004-10-16 18:04:13 +00003247
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003248<div>
Chris Lattner35eca582004-10-16 18:04:13 +00003249
Chris Lattner27f71f22003-09-03 00:41:47 +00003250<h5>Syntax:</h5>
Chris Lattner35eca582004-10-16 18:04:13 +00003251<pre>
3252 unwind
3253</pre>
3254
Chris Lattner27f71f22003-09-03 00:41:47 +00003255<h5>Overview:</h5>
Chris Lattner35eca582004-10-16 18:04:13 +00003256<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003257 at the first callee in the dynamic call stack which used
3258 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
3259 This is primarily used to implement exception handling.</p>
Chris Lattner35eca582004-10-16 18:04:13 +00003260
Chris Lattner27f71f22003-09-03 00:41:47 +00003261<h5>Semantics:</h5>
Chris Lattner72ed2002008-04-19 21:01:16 +00003262<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003263 immediately halt. The dynamic call stack is then searched for the
3264 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
3265 Once found, execution continues at the "exceptional" destination block
3266 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
3267 instruction in the dynamic call chain, undefined behavior results.</p>
3268
Chris Lattnerdf7a6802010-01-15 18:08:37 +00003269<p>Note that the code generator does not yet completely support unwind, and
3270that the invoke/unwind semantics are likely to change in future versions.</p>
3271
Misha Brukman9d0919f2003-11-08 01:05:38 +00003272</div>
Chris Lattner35eca582004-10-16 18:04:13 +00003273
3274<!-- _______________________________________________________________________ -->
3275
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003276<h4>
Bill Wendling772fe172011-07-27 20:18:04 +00003277 <a name="i_resume">'<tt>resume</tt>' Instruction</a>
3278</h4>
3279
3280<div>
3281
3282<h5>Syntax:</h5>
3283<pre>
3284 resume &lt;type&gt; &lt;value&gt;
3285</pre>
3286
3287<h5>Overview:</h5>
3288<p>The '<tt>resume</tt>' instruction is a terminator instruction that has no
3289 successors. Its operand must have the same type as the result of any
3290 '<tt>landingpad</tt>' instruction in the same function.</p>
3291
3292<h5>Semantics:</h5>
3293<p>The '<tt>resume</tt>' instruction resumes propagation of an existing
3294 (in-flight) exception.</p>
3295
3296<h5>Example:</h5>
3297<pre>
3298 resume { i8*, i32 } %exn
3299</pre>
3300
3301</div>
3302
3303<!-- _______________________________________________________________________ -->
3304
3305<h4>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003306 <a name="i_unreachable">'<tt>unreachable</tt>' Instruction</a>
3307</h4>
Chris Lattner35eca582004-10-16 18:04:13 +00003308
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003309<div>
Chris Lattner35eca582004-10-16 18:04:13 +00003310
3311<h5>Syntax:</h5>
3312<pre>
3313 unreachable
3314</pre>
3315
3316<h5>Overview:</h5>
Chris Lattner35eca582004-10-16 18:04:13 +00003317<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003318 instruction is used to inform the optimizer that a particular portion of the
3319 code is not reachable. This can be used to indicate that the code after a
3320 no-return function cannot be reached, and other facts.</p>
Chris Lattner35eca582004-10-16 18:04:13 +00003321
3322<h5>Semantics:</h5>
Chris Lattner35eca582004-10-16 18:04:13 +00003323<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003324
Chris Lattner35eca582004-10-16 18:04:13 +00003325</div>
3326
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003327</div>
3328
Chris Lattner00950542001-06-06 20:29:01 +00003329<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003330<h3>
3331 <a name="binaryops">Binary Operations</a>
3332</h3>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003333
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003334<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003335
3336<p>Binary operators are used to do most of the computation in a program. They
3337 require two operands of the same type, execute an operation on them, and
3338 produce a single value. The operands might represent multiple data, as is
3339 the case with the <a href="#t_vector">vector</a> data type. The result value
3340 has the same type as its operands.</p>
3341
Misha Brukman9d0919f2003-11-08 01:05:38 +00003342<p>There are several different binary operators:</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003343
Chris Lattner00950542001-06-06 20:29:01 +00003344<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003345<h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003346 <a name="i_add">'<tt>add</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003347</h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003348
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003349<div>
Chris Lattner5568e942008-05-20 20:48:21 +00003350
Chris Lattner00950542001-06-06 20:29:01 +00003351<h5>Syntax:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003352<pre>
Dan Gohmancbb38f22009-07-20 22:41:19 +00003353 &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 +00003354 &lt;result&gt; = add nuw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3355 &lt;result&gt; = add nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3356 &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 +00003357</pre>
Chris Lattner5568e942008-05-20 20:48:21 +00003358
Chris Lattner00950542001-06-06 20:29:01 +00003359<h5>Overview:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00003360<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003361
Chris Lattner00950542001-06-06 20:29:01 +00003362<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003363<p>The two arguments to the '<tt>add</tt>' instruction must
3364 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3365 integer values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003366
Chris Lattner00950542001-06-06 20:29:01 +00003367<h5>Semantics:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003368<p>The value produced is the integer sum of the two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003369
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003370<p>If the sum has unsigned overflow, the result returned is the mathematical
3371 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003372
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003373<p>Because LLVM integers use a two's complement representation, this instruction
3374 is appropriate for both signed and unsigned integers.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003375
Dan Gohman08d012e2009-07-22 22:44:56 +00003376<p><tt>nuw</tt> and <tt>nsw</tt> stand for &quot;No Unsigned Wrap&quot;
3377 and &quot;No Signed Wrap&quot;, respectively. If the <tt>nuw</tt> and/or
3378 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
Dan Gohmanfff6c532010-04-22 23:14:21 +00003379 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3380 respectively, occurs.</p>
Dan Gohmancbb38f22009-07-20 22:41:19 +00003381
Chris Lattner00950542001-06-06 20:29:01 +00003382<h5>Example:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003383<pre>
3384 &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Chris Lattner00950542001-06-06 20:29:01 +00003385</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003386
Misha Brukman9d0919f2003-11-08 01:05:38 +00003387</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003388
Chris Lattner00950542001-06-06 20:29:01 +00003389<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003390<h4>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003391 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003392</h4>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003393
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003394<div>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003395
3396<h5>Syntax:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003397<pre>
3398 &lt;result&gt; = fadd &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3399</pre>
3400
3401<h5>Overview:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003402<p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
3403
3404<h5>Arguments:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003405<p>The two arguments to the '<tt>fadd</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003406 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3407 floating point values. Both arguments must have identical types.</p>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003408
3409<h5>Semantics:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003410<p>The value produced is the floating point sum of the two operands.</p>
3411
3412<h5>Example:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003413<pre>
3414 &lt;result&gt; = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
3415</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003416
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003417</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003418
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003419<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003420<h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003421 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003422</h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003423
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003424<div>
Chris Lattner5568e942008-05-20 20:48:21 +00003425
Chris Lattner00950542001-06-06 20:29:01 +00003426<h5>Syntax:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003427<pre>
Dan Gohman08d012e2009-07-22 22:44:56 +00003428 &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 +00003429 &lt;result&gt; = sub nuw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3430 &lt;result&gt; = sub nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3431 &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 +00003432</pre>
Chris Lattner5568e942008-05-20 20:48:21 +00003433
Chris Lattner00950542001-06-06 20:29:01 +00003434<h5>Overview:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00003435<p>The '<tt>sub</tt>' instruction returns the difference of its two
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003436 operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003437
3438<p>Note that the '<tt>sub</tt>' instruction is used to represent the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003439 '<tt>neg</tt>' instruction present in most other intermediate
3440 representations.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003441
Chris Lattner00950542001-06-06 20:29:01 +00003442<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003443<p>The two arguments to the '<tt>sub</tt>' instruction must
3444 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3445 integer values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003446
Chris Lattner00950542001-06-06 20:29:01 +00003447<h5>Semantics:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003448<p>The value produced is the integer difference of the two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003449
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003450<p>If the difference has unsigned overflow, the result returned is the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003451 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
3452 result.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003453
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003454<p>Because LLVM integers use a two's complement representation, this instruction
3455 is appropriate for both signed and unsigned integers.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003456
Dan Gohman08d012e2009-07-22 22:44:56 +00003457<p><tt>nuw</tt> and <tt>nsw</tt> stand for &quot;No Unsigned Wrap&quot;
3458 and &quot;No Signed Wrap&quot;, respectively. If the <tt>nuw</tt> and/or
3459 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
Dan Gohmanfff6c532010-04-22 23:14:21 +00003460 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3461 respectively, occurs.</p>
Dan Gohmancbb38f22009-07-20 22:41:19 +00003462
Chris Lattner00950542001-06-06 20:29:01 +00003463<h5>Example:</h5>
Bill Wendlingaac388b2007-05-29 09:42:13 +00003464<pre>
3465 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
Reid Spencerca86e162006-12-31 07:07:53 +00003466 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
Chris Lattner00950542001-06-06 20:29:01 +00003467</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003468
Misha Brukman9d0919f2003-11-08 01:05:38 +00003469</div>
Chris Lattner5568e942008-05-20 20:48:21 +00003470
Chris Lattner00950542001-06-06 20:29:01 +00003471<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003472<h4>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003473 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003474</h4>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003475
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003476<div>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003477
3478<h5>Syntax:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003479<pre>
3480 &lt;result&gt; = fsub &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3481</pre>
3482
3483<h5>Overview:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003484<p>The '<tt>fsub</tt>' instruction returns the difference of its two
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003485 operands.</p>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003486
3487<p>Note that the '<tt>fsub</tt>' instruction is used to represent the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003488 '<tt>fneg</tt>' instruction present in most other intermediate
3489 representations.</p>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003490
3491<h5>Arguments:</h5>
Bill Wendlingd9fe2982009-07-20 02:32:41 +00003492<p>The two arguments to the '<tt>fsub</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003493 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3494 floating point values. Both arguments must have identical types.</p>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003495
3496<h5>Semantics:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003497<p>The value produced is the floating point difference of the two operands.</p>
3498
3499<h5>Example:</h5>
3500<pre>
3501 &lt;result&gt; = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
3502 &lt;result&gt; = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
3503</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003504
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003505</div>
3506
3507<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003508<h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003509 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003510</h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003511
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003512<div>
Chris Lattner5568e942008-05-20 20:48:21 +00003513
Chris Lattner00950542001-06-06 20:29:01 +00003514<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003515<pre>
Dan Gohman08d012e2009-07-22 22:44:56 +00003516 &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 +00003517 &lt;result&gt; = mul nuw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3518 &lt;result&gt; = mul nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3519 &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 +00003520</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003521
Chris Lattner00950542001-06-06 20:29:01 +00003522<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003523<p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003524
Chris Lattner00950542001-06-06 20:29:01 +00003525<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003526<p>The two arguments to the '<tt>mul</tt>' instruction must
3527 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3528 integer values. Both arguments must have identical types.</p>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003529
Chris Lattner00950542001-06-06 20:29:01 +00003530<h5>Semantics:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003531<p>The value produced is the integer product of the two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003532
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003533<p>If the result of the multiplication has unsigned overflow, the result
3534 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
3535 width of the result.</p>
3536
3537<p>Because LLVM integers use a two's complement representation, and the result
3538 is the same width as the operands, this instruction returns the correct
3539 result for both signed and unsigned integers. If a full product
3540 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
3541 be sign-extended or zero-extended as appropriate to the width of the full
3542 product.</p>
3543
Dan Gohman08d012e2009-07-22 22:44:56 +00003544<p><tt>nuw</tt> and <tt>nsw</tt> stand for &quot;No Unsigned Wrap&quot;
3545 and &quot;No Signed Wrap&quot;, respectively. If the <tt>nuw</tt> and/or
3546 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
Dan Gohmanfff6c532010-04-22 23:14:21 +00003547 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3548 respectively, occurs.</p>
Dan Gohmancbb38f22009-07-20 22:41:19 +00003549
Chris Lattner00950542001-06-06 20:29:01 +00003550<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003551<pre>
3552 &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
Chris Lattner00950542001-06-06 20:29:01 +00003553</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003554
Misha Brukman9d0919f2003-11-08 01:05:38 +00003555</div>
Chris Lattner5568e942008-05-20 20:48:21 +00003556
Chris Lattner00950542001-06-06 20:29:01 +00003557<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003558<h4>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003559 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003560</h4>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003561
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003562<div>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003563
3564<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003565<pre>
3566 &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 +00003567</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003568
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003569<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003570<p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003571
3572<h5>Arguments:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003573<p>The two arguments to the '<tt>fmul</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003574 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3575 floating point values. Both arguments must have identical types.</p>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003576
3577<h5>Semantics:</h5>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003578<p>The value produced is the floating point product of the two operands.</p>
3579
3580<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003581<pre>
3582 &lt;result&gt; = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003583</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003584
Dan Gohmanae3a0be2009-06-04 22:49:04 +00003585</div>
3586
3587<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003588<h4>
3589 <a name="i_udiv">'<tt>udiv</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>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003595<pre>
Chris Lattner35bda892011-02-06 21:44:57 +00003596 &lt;result&gt; = udiv &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3597 &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 +00003598</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003599
Reid Spencer1628cec2006-10-26 06:15:43 +00003600<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003601<p>The '<tt>udiv</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>udiv</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>
Chris Lattner5ec89832008-01-28 00:36:27 +00003609<p>The value produced is the unsigned integer quotient of the two operands.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003610
Chris Lattner5ec89832008-01-28 00:36:27 +00003611<p>Note that unsigned integer division and signed integer division are distinct
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003612 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
3613
Chris Lattner5ec89832008-01-28 00:36:27 +00003614<p>Division by zero leads to undefined behavior.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003615
Chris Lattner35bda892011-02-06 21:44:57 +00003616<p>If the <tt>exact</tt> keyword is present, the result value of the
3617 <tt>udiv</tt> is a <a href="#trapvalues">trap value</a> if %op1 is not a
3618 multiple of %op2 (as such, "((a udiv exact b) mul b) == a").</p>
3619
3620
Reid Spencer1628cec2006-10-26 06:15:43 +00003621<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003622<pre>
3623 &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
Reid Spencer1628cec2006-10-26 06:15:43 +00003624</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003625
Reid Spencer1628cec2006-10-26 06:15:43 +00003626</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003627
Reid Spencer1628cec2006-10-26 06:15:43 +00003628<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003629<h4>
3630 <a name="i_sdiv">'<tt>sdiv</tt>' Instruction</a>
3631</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003632
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003633<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003634
Reid Spencer1628cec2006-10-26 06:15:43 +00003635<h5>Syntax:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003636<pre>
Dan Gohmancbb38f22009-07-20 22:41:19 +00003637 &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 +00003638 &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 +00003639</pre>
Chris Lattner5568e942008-05-20 20:48:21 +00003640
Reid Spencer1628cec2006-10-26 06:15:43 +00003641<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003642<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003643
Reid Spencer1628cec2006-10-26 06:15:43 +00003644<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003645<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003646 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3647 values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003648
Reid Spencer1628cec2006-10-26 06:15:43 +00003649<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003650<p>The value produced is the signed integer quotient of the two operands rounded
3651 towards zero.</p>
3652
Chris Lattner5ec89832008-01-28 00:36:27 +00003653<p>Note that signed integer division and unsigned integer division are distinct
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003654 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
3655
Chris Lattner5ec89832008-01-28 00:36:27 +00003656<p>Division by zero leads to undefined behavior. Overflow also leads to
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003657 undefined behavior; this is a rare case, but can occur, for example, by doing
3658 a 32-bit division of -2147483648 by -1.</p>
3659
Dan Gohman9c5beed2009-07-22 00:04:19 +00003660<p>If the <tt>exact</tt> keyword is present, the result value of the
Dan Gohman27ef9972010-04-23 15:23:32 +00003661 <tt>sdiv</tt> is a <a href="#trapvalues">trap value</a> if the result would
Dan Gohman38da9272010-07-11 00:08:34 +00003662 be rounded.</p>
Dan Gohmancbb38f22009-07-20 22:41:19 +00003663
Reid Spencer1628cec2006-10-26 06:15:43 +00003664<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003665<pre>
3666 &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
Reid Spencer1628cec2006-10-26 06:15:43 +00003667</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003668
Reid Spencer1628cec2006-10-26 06:15:43 +00003669</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003670
Reid Spencer1628cec2006-10-26 06:15:43 +00003671<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003672<h4>
3673 <a name="i_fdiv">'<tt>fdiv</tt>' Instruction</a>
3674</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003675
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003676<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003677
Chris Lattner00950542001-06-06 20:29:01 +00003678<h5>Syntax:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003679<pre>
Gabor Greiffb224a22008-08-07 21:46:00 +00003680 &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 +00003681</pre>
Chris Lattner5568e942008-05-20 20:48:21 +00003682
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003683<h5>Overview:</h5>
3684<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003685
Chris Lattner261efe92003-11-25 01:02:51 +00003686<h5>Arguments:</h5>
Jeff Cohen6f1cc772007-04-22 01:17:39 +00003687<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003688 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3689 floating point values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003690
Chris Lattner261efe92003-11-25 01:02:51 +00003691<h5>Semantics:</h5>
Reid Spencer1628cec2006-10-26 06:15:43 +00003692<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003693
Chris Lattner261efe92003-11-25 01:02:51 +00003694<h5>Example:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003695<pre>
3696 &lt;result&gt; = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
Chris Lattner261efe92003-11-25 01:02:51 +00003697</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003698
Chris Lattner261efe92003-11-25 01:02:51 +00003699</div>
Chris Lattner5568e942008-05-20 20:48:21 +00003700
Chris Lattner261efe92003-11-25 01:02:51 +00003701<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003702<h4>
3703 <a name="i_urem">'<tt>urem</tt>' Instruction</a>
3704</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003705
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003706<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003707
Reid Spencer0a783f72006-11-02 01:53:59 +00003708<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003709<pre>
3710 &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 +00003711</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003712
Reid Spencer0a783f72006-11-02 01:53:59 +00003713<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003714<p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3715 division of its two arguments.</p>
3716
Reid Spencer0a783f72006-11-02 01:53:59 +00003717<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003718<p>The two arguments to the '<tt>urem</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003719 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3720 values. Both arguments must have identical types.</p>
3721
Reid Spencer0a783f72006-11-02 01:53:59 +00003722<h5>Semantics:</h5>
3723<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003724 This instruction always performs an unsigned division to get the
3725 remainder.</p>
3726
Chris Lattner5ec89832008-01-28 00:36:27 +00003727<p>Note that unsigned integer remainder and signed integer remainder are
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003728 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3729
Chris Lattner5ec89832008-01-28 00:36:27 +00003730<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003731
Reid Spencer0a783f72006-11-02 01:53:59 +00003732<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003733<pre>
3734 &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
Reid Spencer0a783f72006-11-02 01:53:59 +00003735</pre>
3736
3737</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003738
Reid Spencer0a783f72006-11-02 01:53:59 +00003739<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003740<h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003741 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003742</h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003743
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003744<div>
Chris Lattner5568e942008-05-20 20:48:21 +00003745
Chris Lattner261efe92003-11-25 01:02:51 +00003746<h5>Syntax:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003747<pre>
Gabor Greiffb224a22008-08-07 21:46:00 +00003748 &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 +00003749</pre>
Chris Lattner5568e942008-05-20 20:48:21 +00003750
Chris Lattner261efe92003-11-25 01:02:51 +00003751<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003752<p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3753 division of its two operands. This instruction can also take
3754 <a href="#t_vector">vector</a> versions of the values in which case the
3755 elements must be integers.</p>
Chris Lattnerc7d3ab32008-01-04 04:33:49 +00003756
Chris Lattner261efe92003-11-25 01:02:51 +00003757<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003758<p>The two arguments to the '<tt>srem</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003759 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3760 values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003761
Chris Lattner261efe92003-11-25 01:02:51 +00003762<h5>Semantics:</h5>
Reid Spencer0a783f72006-11-02 01:53:59 +00003763<p>This instruction returns the <i>remainder</i> of a division (where the result
Duncan Sandsdea3a5e2011-03-07 09:12:24 +00003764 is either zero or has the same sign as the dividend, <tt>op1</tt>), not the
3765 <i>modulo</i> operator (where the result is either zero or has the same sign
3766 as the divisor, <tt>op2</tt>) of a value.
3767 For more information about the difference,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003768 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3769 Math Forum</a>. For a table of how this is implemented in various languages,
3770 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3771 Wikipedia: modulo operation</a>.</p>
3772
Chris Lattner5ec89832008-01-28 00:36:27 +00003773<p>Note that signed integer remainder and unsigned integer remainder are
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003774 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3775
Chris Lattner5ec89832008-01-28 00:36:27 +00003776<p>Taking the remainder of a division by zero leads to undefined behavior.
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003777 Overflow also leads to undefined behavior; this is a rare case, but can
3778 occur, for example, by taking the remainder of a 32-bit division of
3779 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3780 lets srem be implemented using instructions that return both the result of
3781 the division and the remainder.)</p>
3782
Chris Lattner261efe92003-11-25 01:02:51 +00003783<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003784<pre>
3785 &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
Reid Spencer0a783f72006-11-02 01:53:59 +00003786</pre>
3787
3788</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003789
Reid Spencer0a783f72006-11-02 01:53:59 +00003790<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003791<h4>
3792 <a name="i_frem">'<tt>frem</tt>' Instruction</a>
3793</h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003794
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003795<div>
Chris Lattner5568e942008-05-20 20:48:21 +00003796
Reid Spencer0a783f72006-11-02 01:53:59 +00003797<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003798<pre>
3799 &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 +00003800</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003801
Reid Spencer0a783f72006-11-02 01:53:59 +00003802<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003803<p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3804 its two operands.</p>
3805
Reid Spencer0a783f72006-11-02 01:53:59 +00003806<h5>Arguments:</h5>
3807<p>The two arguments to the '<tt>frem</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003808 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3809 floating point values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003810
Reid Spencer0a783f72006-11-02 01:53:59 +00003811<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003812<p>This instruction returns the <i>remainder</i> of a division. The remainder
3813 has the same sign as the dividend.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003814
Reid Spencer0a783f72006-11-02 01:53:59 +00003815<h5>Example:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003816<pre>
3817 &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
Chris Lattner261efe92003-11-25 01:02:51 +00003818</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003819
Misha Brukman9d0919f2003-11-08 01:05:38 +00003820</div>
Robert Bocchino7b81c752006-02-17 21:18:08 +00003821
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003822</div>
3823
Reid Spencer8e11bf82007-02-02 13:57:07 +00003824<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003825<h3>
3826 <a name="bitwiseops">Bitwise Binary Operations</a>
3827</h3>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003828
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003829<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003830
3831<p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3832 program. They are generally very efficient instructions and can commonly be
3833 strength reduced from other instructions. They require two operands of the
3834 same type, execute an operation on them, and produce a single value. The
3835 resulting value is the same type as its operands.</p>
3836
Reid Spencer569f2fa2007-01-31 21:39:12 +00003837<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003838<h4>
3839 <a name="i_shl">'<tt>shl</tt>' Instruction</a>
3840</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003841
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003842<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003843
Reid Spencer569f2fa2007-01-31 21:39:12 +00003844<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003845<pre>
Chris Lattnerf067d582011-02-07 16:40:21 +00003846 &lt;result&gt; = shl &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3847 &lt;result&gt; = shl nuw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3848 &lt;result&gt; = shl nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3849 &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 +00003850</pre>
Chris Lattner6ccc2d52007-10-03 21:01:14 +00003851
Reid Spencer569f2fa2007-01-31 21:39:12 +00003852<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003853<p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3854 a specified number of bits.</p>
Chris Lattner6ccc2d52007-10-03 21:01:14 +00003855
Reid Spencer569f2fa2007-01-31 21:39:12 +00003856<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003857<p>Both arguments to the '<tt>shl</tt>' instruction must be the
3858 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3859 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003860
Reid Spencer569f2fa2007-01-31 21:39:12 +00003861<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003862<p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3863 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3864 is (statically or dynamically) negative or equal to or larger than the number
3865 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3866 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3867 shift amount in <tt>op2</tt>.</p>
Chris Lattner6ccc2d52007-10-03 21:01:14 +00003868
Chris Lattnerf067d582011-02-07 16:40:21 +00003869<p>If the <tt>nuw</tt> keyword is present, then the shift produces a
3870 <a href="#trapvalues">trap value</a> if it shifts out any non-zero bits. If
Chris Lattner66298c12011-02-09 16:44:44 +00003871 the <tt>nsw</tt> keyword is present, then the shift produces a
Chris Lattnerf067d582011-02-07 16:40:21 +00003872 <a href="#trapvalues">trap value</a> if it shifts out any bits that disagree
3873 with the resultant sign bit. As such, NUW/NSW have the same semantics as
3874 they would if the shift were expressed as a mul instruction with the same
3875 nsw/nuw bits in (mul %op1, (shl 1, %op2)).</p>
3876
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003877<h5>Example:</h5>
3878<pre>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003879 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
3880 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
3881 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattner6ccc2d52007-10-03 21:01:14 +00003882 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Mon P Wange9f10152008-12-09 05:46:39 +00003883 &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 +00003884</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003885
Reid Spencer569f2fa2007-01-31 21:39:12 +00003886</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003887
Reid Spencer569f2fa2007-01-31 21:39:12 +00003888<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003889<h4>
3890 <a name="i_lshr">'<tt>lshr</tt>' Instruction</a>
3891</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003892
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003893<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003894
Reid Spencer569f2fa2007-01-31 21:39:12 +00003895<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003896<pre>
Chris Lattnerf067d582011-02-07 16:40:21 +00003897 &lt;result&gt; = lshr &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3898 &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 +00003899</pre>
3900
3901<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003902<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
3903 operand shifted to the right a specified number of bits with zero fill.</p>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003904
3905<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003906<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003907 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3908 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003909
3910<h5>Semantics:</h5>
3911<p>This instruction always performs a logical shift right operation. The most
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003912 significant bits of the result will be filled with zero bits after the shift.
3913 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
3914 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3915 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3916 shift amount in <tt>op2</tt>.</p>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003917
Chris Lattnerf067d582011-02-07 16:40:21 +00003918<p>If the <tt>exact</tt> keyword is present, the result value of the
3919 <tt>lshr</tt> is a <a href="#trapvalues">trap value</a> if any of the bits
3920 shifted out are non-zero.</p>
3921
3922
Reid Spencer569f2fa2007-01-31 21:39:12 +00003923<h5>Example:</h5>
3924<pre>
3925 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
3926 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
3927 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
3928 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattner6ccc2d52007-10-03 21:01:14 +00003929 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Mon P Wange9f10152008-12-09 05:46:39 +00003930 &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 +00003931</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003932
Reid Spencer569f2fa2007-01-31 21:39:12 +00003933</div>
3934
Reid Spencer8e11bf82007-02-02 13:57:07 +00003935<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003936<h4>
3937 <a name="i_ashr">'<tt>ashr</tt>' Instruction</a>
3938</h4>
3939
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003940<div>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003941
3942<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003943<pre>
Chris Lattnerf067d582011-02-07 16:40:21 +00003944 &lt;result&gt; = ashr &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3945 &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 +00003946</pre>
3947
3948<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003949<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3950 operand shifted to the right a specified number of bits with sign
3951 extension.</p>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003952
3953<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003954<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003955 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3956 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003957
3958<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003959<p>This instruction always performs an arithmetic shift right operation, The
3960 most significant bits of the result will be filled with the sign bit
3961 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3962 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
3963 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
3964 the corresponding shift amount in <tt>op2</tt>.</p>
Reid Spencer569f2fa2007-01-31 21:39:12 +00003965
Chris Lattnerf067d582011-02-07 16:40:21 +00003966<p>If the <tt>exact</tt> keyword is present, the result value of the
3967 <tt>ashr</tt> is a <a href="#trapvalues">trap value</a> if any of the bits
3968 shifted out are non-zero.</p>
3969
Reid Spencer569f2fa2007-01-31 21:39:12 +00003970<h5>Example:</h5>
3971<pre>
3972 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3973 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3974 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3975 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattner6ccc2d52007-10-03 21:01:14 +00003976 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Mon P Wange9f10152008-12-09 05:46:39 +00003977 &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 +00003978</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003979
Reid Spencer569f2fa2007-01-31 21:39:12 +00003980</div>
3981
Chris Lattner00950542001-06-06 20:29:01 +00003982<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00003983<h4>
3984 <a name="i_and">'<tt>and</tt>' Instruction</a>
3985</h4>
Chris Lattner5568e942008-05-20 20:48:21 +00003986
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00003987<div>
Chris Lattner5568e942008-05-20 20:48:21 +00003988
Chris Lattner00950542001-06-06 20:29:01 +00003989<h5>Syntax:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00003990<pre>
Gabor Greiffb224a22008-08-07 21:46:00 +00003991 &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 +00003992</pre>
Chris Lattner5568e942008-05-20 20:48:21 +00003993
Chris Lattner00950542001-06-06 20:29:01 +00003994<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00003995<p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
3996 operands.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00003997
Chris Lattner00950542001-06-06 20:29:01 +00003998<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00003999<p>The two arguments to the '<tt>and</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004000 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4001 values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00004002
Chris Lattner00950542001-06-06 20:29:01 +00004003<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00004004<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004005
Misha Brukman9d0919f2003-11-08 01:05:38 +00004006<table border="1" cellspacing="0" cellpadding="4">
Chris Lattner261efe92003-11-25 01:02:51 +00004007 <tbody>
4008 <tr>
4009 <td>In0</td>
4010 <td>In1</td>
4011 <td>Out</td>
4012 </tr>
4013 <tr>
4014 <td>0</td>
4015 <td>0</td>
4016 <td>0</td>
4017 </tr>
4018 <tr>
4019 <td>0</td>
4020 <td>1</td>
4021 <td>0</td>
4022 </tr>
4023 <tr>
4024 <td>1</td>
4025 <td>0</td>
4026 <td>0</td>
4027 </tr>
4028 <tr>
4029 <td>1</td>
4030 <td>1</td>
4031 <td>1</td>
4032 </tr>
4033 </tbody>
4034</table>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004035
Chris Lattner00950542001-06-06 20:29:01 +00004036<h5>Example:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00004037<pre>
4038 &lt;result&gt; = and i32 4, %var <i>; yields {i32}:result = 4 &amp; %var</i>
Reid Spencerca86e162006-12-31 07:07:53 +00004039 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
4040 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
Chris Lattner00950542001-06-06 20:29:01 +00004041</pre>
Misha Brukman9d0919f2003-11-08 01:05:38 +00004042</div>
Chris Lattner00950542001-06-06 20:29:01 +00004043<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004044<h4>
4045 <a name="i_or">'<tt>or</tt>' Instruction</a>
4046</h4>
Chris Lattner5568e942008-05-20 20:48:21 +00004047
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004048<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004049
4050<h5>Syntax:</h5>
4051<pre>
4052 &lt;result&gt; = or &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
4053</pre>
4054
4055<h5>Overview:</h5>
4056<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
4057 two operands.</p>
4058
4059<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004060<p>The two arguments to the '<tt>or</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004061 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4062 values. Both arguments must have identical types.</p>
4063
Chris Lattner00950542001-06-06 20:29:01 +00004064<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00004065<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004066
Chris Lattner261efe92003-11-25 01:02:51 +00004067<table border="1" cellspacing="0" cellpadding="4">
4068 <tbody>
4069 <tr>
4070 <td>In0</td>
4071 <td>In1</td>
4072 <td>Out</td>
4073 </tr>
4074 <tr>
4075 <td>0</td>
4076 <td>0</td>
4077 <td>0</td>
4078 </tr>
4079 <tr>
4080 <td>0</td>
4081 <td>1</td>
4082 <td>1</td>
4083 </tr>
4084 <tr>
4085 <td>1</td>
4086 <td>0</td>
4087 <td>1</td>
4088 </tr>
4089 <tr>
4090 <td>1</td>
4091 <td>1</td>
4092 <td>1</td>
4093 </tr>
4094 </tbody>
4095</table>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004096
Chris Lattner00950542001-06-06 20:29:01 +00004097<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004098<pre>
4099 &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
Reid Spencerca86e162006-12-31 07:07:53 +00004100 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
4101 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
Chris Lattner00950542001-06-06 20:29:01 +00004102</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004103
Misha Brukman9d0919f2003-11-08 01:05:38 +00004104</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004105
Chris Lattner00950542001-06-06 20:29:01 +00004106<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004107<h4>
4108 <a name="i_xor">'<tt>xor</tt>' Instruction</a>
4109</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004110
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004111<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004112
Chris Lattner00950542001-06-06 20:29:01 +00004113<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004114<pre>
4115 &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 +00004116</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004117
Chris Lattner00950542001-06-06 20:29:01 +00004118<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004119<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
4120 its two operands. The <tt>xor</tt> is used to implement the "one's
4121 complement" operation, which is the "~" operator in C.</p>
4122
Chris Lattner00950542001-06-06 20:29:01 +00004123<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004124<p>The two arguments to the '<tt>xor</tt>' instruction must be
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004125 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4126 values. Both arguments must have identical types.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00004127
Chris Lattner00950542001-06-06 20:29:01 +00004128<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00004129<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004130
Chris Lattner261efe92003-11-25 01:02:51 +00004131<table border="1" cellspacing="0" cellpadding="4">
4132 <tbody>
4133 <tr>
4134 <td>In0</td>
4135 <td>In1</td>
4136 <td>Out</td>
4137 </tr>
4138 <tr>
4139 <td>0</td>
4140 <td>0</td>
4141 <td>0</td>
4142 </tr>
4143 <tr>
4144 <td>0</td>
4145 <td>1</td>
4146 <td>1</td>
4147 </tr>
4148 <tr>
4149 <td>1</td>
4150 <td>0</td>
4151 <td>1</td>
4152 </tr>
4153 <tr>
4154 <td>1</td>
4155 <td>1</td>
4156 <td>0</td>
4157 </tr>
4158 </tbody>
4159</table>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004160
Chris Lattner00950542001-06-06 20:29:01 +00004161<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004162<pre>
4163 &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
Reid Spencerca86e162006-12-31 07:07:53 +00004164 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
4165 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
4166 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
Chris Lattner00950542001-06-06 20:29:01 +00004167</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004168
Misha Brukman9d0919f2003-11-08 01:05:38 +00004169</div>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004170
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004171</div>
4172
Chris Lattner00950542001-06-06 20:29:01 +00004173<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004174<h3>
Chris Lattner3df241e2006-04-08 23:07:04 +00004175 <a name="vectorops">Vector Operations</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004176</h3>
Chris Lattner3df241e2006-04-08 23:07:04 +00004177
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004178<div>
Chris Lattner3df241e2006-04-08 23:07:04 +00004179
4180<p>LLVM supports several instructions to represent vector operations in a
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004181 target-independent manner. These instructions cover the element-access and
4182 vector-specific operations needed to process vectors effectively. While LLVM
4183 does directly support these vector operations, many sophisticated algorithms
4184 will want to use target-specific intrinsics to take full advantage of a
4185 specific target.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004186
Chris Lattner3df241e2006-04-08 23:07:04 +00004187<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004188<h4>
Chris Lattner3df241e2006-04-08 23:07:04 +00004189 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004190</h4>
Chris Lattner3df241e2006-04-08 23:07:04 +00004191
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004192<div>
Chris Lattner3df241e2006-04-08 23:07:04 +00004193
4194<h5>Syntax:</h5>
Chris Lattner3df241e2006-04-08 23:07:04 +00004195<pre>
Reid Spencerca86e162006-12-31 07:07:53 +00004196 &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 +00004197</pre>
4198
4199<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004200<p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
4201 from a vector at a specified index.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004202
4203
4204<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004205<p>The first operand of an '<tt>extractelement</tt>' instruction is a value
4206 of <a href="#t_vector">vector</a> type. The second operand is an index
4207 indicating the position from which to extract the element. The index may be
4208 a variable.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004209
4210<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004211<p>The result is a scalar of the same type as the element type of
4212 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
4213 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4214 results are undefined.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004215
4216<h5>Example:</h5>
Chris Lattner3df241e2006-04-08 23:07:04 +00004217<pre>
Gabor Greifa5b6f452009-10-28 13:14:50 +00004218 &lt;result&gt; = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
Chris Lattner3df241e2006-04-08 23:07:04 +00004219</pre>
Chris Lattner3df241e2006-04-08 23:07:04 +00004220
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004221</div>
Chris Lattner3df241e2006-04-08 23:07:04 +00004222
4223<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004224<h4>
Chris Lattner3df241e2006-04-08 23:07:04 +00004225 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004226</h4>
Chris Lattner3df241e2006-04-08 23:07:04 +00004227
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004228<div>
Chris Lattner3df241e2006-04-08 23:07:04 +00004229
4230<h5>Syntax:</h5>
Chris Lattner3df241e2006-04-08 23:07:04 +00004231<pre>
Dan Gohmanf3480b92008-05-12 23:38:42 +00004232 &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 +00004233</pre>
4234
4235<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004236<p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
4237 vector at a specified index.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004238
4239<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004240<p>The first operand of an '<tt>insertelement</tt>' instruction is a value
4241 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
4242 whose type must equal the element type of the first operand. The third
4243 operand is an index indicating the position at which to insert the value.
4244 The index may be a variable.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004245
4246<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004247<p>The result is a vector of the same type as <tt>val</tt>. Its element values
4248 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
4249 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4250 results are undefined.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004251
4252<h5>Example:</h5>
Chris Lattner3df241e2006-04-08 23:07:04 +00004253<pre>
Gabor Greifa5b6f452009-10-28 13:14:50 +00004254 &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 +00004255</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004256
Chris Lattner3df241e2006-04-08 23:07:04 +00004257</div>
4258
4259<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004260<h4>
Chris Lattner3df241e2006-04-08 23:07:04 +00004261 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004262</h4>
Chris Lattner3df241e2006-04-08 23:07:04 +00004263
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004264<div>
Chris Lattner3df241e2006-04-08 23:07:04 +00004265
4266<h5>Syntax:</h5>
Chris Lattner3df241e2006-04-08 23:07:04 +00004267<pre>
Mon P Wangaeb06d22008-11-10 04:46:22 +00004268 &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 +00004269</pre>
4270
4271<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004272<p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
4273 from two input vectors, returning a vector with the same element type as the
4274 input and length that is the same as the shuffle mask.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004275
4276<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004277<p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
4278 with types that match each other. The third argument is a shuffle mask whose
4279 element type is always 'i32'. The result of the instruction is a vector
4280 whose length is the same as the shuffle mask and whose element type is the
4281 same as the element type of the first two operands.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004282
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004283<p>The shuffle mask operand is required to be a constant vector with either
4284 constant integer or undef values.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004285
4286<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004287<p>The elements of the two input vectors are numbered from left to right across
4288 both of the vectors. The shuffle mask operand specifies, for each element of
4289 the result vector, which element of the two input vectors the result element
4290 gets. The element selector may be undef (meaning "don't care") and the
4291 second operand may be undef if performing a shuffle from only one vector.</p>
Chris Lattner3df241e2006-04-08 23:07:04 +00004292
4293<h5>Example:</h5>
Chris Lattner3df241e2006-04-08 23:07:04 +00004294<pre>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004295 &lt;result&gt; = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
Jeff Cohen6f1cc772007-04-22 01:17:39 +00004296 &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 +00004297 &lt;result&gt; = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
Reid Spencerca86e162006-12-31 07:07:53 +00004298 &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 +00004299 &lt;result&gt; = shufflevector &lt;8 x i32&gt; %v1, &lt;8 x i32&gt; undef,
Mon P Wangaeb06d22008-11-10 04:46:22 +00004300 &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 +00004301 &lt;result&gt; = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
Mon P Wangaeb06d22008-11-10 04:46:22 +00004302 &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 +00004303</pre>
Chris Lattner3df241e2006-04-08 23:07:04 +00004304
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004305</div>
Tanya Lattner09474292006-04-14 19:24:33 +00004306
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004307</div>
4308
Chris Lattner3df241e2006-04-08 23:07:04 +00004309<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004310<h3>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004311 <a name="aggregateops">Aggregate Operations</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004312</h3>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004313
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004314<div>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004315
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004316<p>LLVM supports several instructions for working with
4317 <a href="#t_aggregate">aggregate</a> values.</p>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004318
Dan Gohmana334d5f2008-05-12 23:51:09 +00004319<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004320<h4>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004321 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004322</h4>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004323
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004324<div>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004325
4326<h5>Syntax:</h5>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004327<pre>
4328 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
4329</pre>
4330
4331<h5>Overview:</h5>
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004332<p>The '<tt>extractvalue</tt>' instruction extracts the value of a member field
4333 from an <a href="#t_aggregate">aggregate</a> value.</p>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004334
4335<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004336<p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
Chris Lattner61c70e92010-08-28 04:09:24 +00004337 of <a href="#t_struct">struct</a> or
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004338 <a href="#t_array">array</a> type. The operands are constant indices to
4339 specify which value to extract in a similar manner as indices in a
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004340 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
Frits van Bommel13242892010-12-05 20:54:38 +00004341 <p>The major differences to <tt>getelementptr</tt> indexing are:</p>
4342 <ul>
4343 <li>Since the value being indexed is not a pointer, the first index is
4344 omitted and assumed to be zero.</li>
4345 <li>At least one index must be specified.</li>
4346 <li>Not only struct indices but also array indices must be in
4347 bounds.</li>
4348 </ul>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004349
4350<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004351<p>The result is the value at the position in the aggregate specified by the
4352 index operands.</p>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004353
4354<h5>Example:</h5>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004355<pre>
Gabor Greifa5b6f452009-10-28 13:14:50 +00004356 &lt;result&gt; = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004357</pre>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004358
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004359</div>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004360
4361<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004362<h4>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004363 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004364</h4>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004365
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004366<div>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004367
4368<h5>Syntax:</h5>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004369<pre>
Bill Wendling194229e2011-07-26 20:42:28 +00004370 &lt;result&gt; = insertvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;ty&gt; &lt;elt&gt;, &lt;idx&gt;{, &lt;idx&gt;}* <i>; yields &lt;aggregate type&gt;</i>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004371</pre>
4372
4373<h5>Overview:</h5>
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004374<p>The '<tt>insertvalue</tt>' instruction inserts a value into a member field
4375 in an <a href="#t_aggregate">aggregate</a> value.</p>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004376
4377<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004378<p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
Chris Lattner61c70e92010-08-28 04:09:24 +00004379 of <a href="#t_struct">struct</a> or
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004380 <a href="#t_array">array</a> type. The second operand is a first-class
4381 value to insert. The following operands are constant indices indicating
4382 the position at which to insert the value in a similar manner as indices in a
Frits van Bommel13242892010-12-05 20:54:38 +00004383 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' instruction. The
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004384 value to insert must have the same type as the value identified by the
4385 indices.</p>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004386
4387<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004388<p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
4389 that of <tt>val</tt> except that the value at the position specified by the
4390 indices is that of <tt>elt</tt>.</p>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004391
4392<h5>Example:</h5>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004393<pre>
Chris Lattner8645d1a2011-05-22 07:18:08 +00004394 %agg1 = insertvalue {i32, float} undef, i32 1, 0 <i>; yields {i32 1, float undef}</i>
4395 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 <i>; yields {i32 1, float %val}</i>
4396 %agg3 = insertvalue {i32, {float}} %agg1, float %val, 1, 0 <i>; yields {i32 1, float %val}</i>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004397</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004398
Dan Gohmana334d5f2008-05-12 23:51:09 +00004399</div>
4400
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004401</div>
Dan Gohmana334d5f2008-05-12 23:51:09 +00004402
4403<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004404<h3>
Chris Lattner884a9702006-08-15 00:45:58 +00004405 <a name="memoryops">Memory Access and Addressing Operations</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004406</h3>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004407
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004408<div>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004409
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004410<p>A key design point of an SSA-based representation is how it represents
4411 memory. In LLVM, no memory locations are in SSA form, which makes things
Victor Hernandez2fee2942009-10-26 23:44:29 +00004412 very simple. This section describes how to read, write, and allocate
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004413 memory in LLVM.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004414
Chris Lattner00950542001-06-06 20:29:01 +00004415<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004416<h4>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004417 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004418</h4>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004419
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004420<div>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004421
Chris Lattner00950542001-06-06 20:29:01 +00004422<h5>Syntax:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004423<pre>
Dan Gohmanf75a7d32010-05-28 01:14:11 +00004424 &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 +00004425</pre>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004426
Chris Lattner00950542001-06-06 20:29:01 +00004427<h5>Overview:</h5>
Jeff Cohen6f1cc772007-04-22 01:17:39 +00004428<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004429 currently executing function, to be automatically released when this function
4430 returns to its caller. The object is always allocated in the generic address
4431 space (address space zero).</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004432
Chris Lattner00950542001-06-06 20:29:01 +00004433<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004434<p>The '<tt>alloca</tt>' instruction
4435 allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt> bytes of memory on the
4436 runtime stack, returning a pointer of the appropriate type to the program.
4437 If "NumElements" is specified, it is the number of elements allocated,
4438 otherwise "NumElements" is defaulted to be one. If a constant alignment is
4439 specified, the value result of the allocation is guaranteed to be aligned to
4440 at least that boundary. If not specified, or if zero, the target can choose
4441 to align the allocation on any convenient boundary compatible with the
4442 type.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004443
Misha Brukman9d0919f2003-11-08 01:05:38 +00004444<p>'<tt>type</tt>' may be any sized type.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004445
Chris Lattner00950542001-06-06 20:29:01 +00004446<h5>Semantics:</h5>
Bill Wendling871eb0a2009-05-08 20:49:29 +00004447<p>Memory is allocated; a pointer is returned. The operation is undefined if
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004448 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
4449 memory is automatically released when the function returns. The
4450 '<tt>alloca</tt>' instruction is commonly used to represent automatic
4451 variables that must have an address available. When the function returns
4452 (either with the <tt><a href="#i_ret">ret</a></tt>
4453 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
4454 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004455
Chris Lattner00950542001-06-06 20:29:01 +00004456<h5>Example:</h5>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004457<pre>
Dan Gohman81e21672009-01-04 23:49:44 +00004458 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
4459 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
4460 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
4461 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
Chris Lattner00950542001-06-06 20:29:01 +00004462</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004463
Misha Brukman9d0919f2003-11-08 01:05:38 +00004464</div>
Chris Lattner2cbdc452005-11-06 08:02:57 +00004465
Chris Lattner00950542001-06-06 20:29:01 +00004466<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004467<h4>
4468 <a name="i_load">'<tt>load</tt>' Instruction</a>
4469</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004470
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004471<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004472
Chris Lattner2b7d3202002-05-06 03:03:22 +00004473<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004474<pre>
Bill Wendling7c78dbb2010-02-25 21:23:24 +00004475 &lt;result&gt; = load &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;][, !nontemporal !&lt;index&gt;]
4476 &lt;result&gt; = volatile load &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;][, !nontemporal !&lt;index&gt;]
4477 !&lt;index&gt; = !{ i32 1 }
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004478</pre>
4479
Chris Lattner2b7d3202002-05-06 03:03:22 +00004480<h5>Overview:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00004481<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004482
Chris Lattner2b7d3202002-05-06 03:03:22 +00004483<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004484<p>The argument to the '<tt>load</tt>' instruction specifies the memory address
4485 from which to load. The pointer must point to
4486 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
4487 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00004488 number or order of execution of this <tt>load</tt> with other <a
4489 href="#volatile">volatile operations</a>.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004490
Bill Wendling7c78dbb2010-02-25 21:23:24 +00004491<p>The optional constant <tt>align</tt> argument specifies the alignment of the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004492 operation (that is, the alignment of the memory address). A value of 0 or an
Bill Wendling7c78dbb2010-02-25 21:23:24 +00004493 omitted <tt>align</tt> argument means that the operation has the preferential
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004494 alignment for the target. It is the responsibility of the code emitter to
4495 ensure that the alignment information is correct. Overestimating the
Bill Wendling7c78dbb2010-02-25 21:23:24 +00004496 alignment results in undefined behavior. Underestimating the alignment may
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004497 produce less efficient code. An alignment of 1 is always safe.</p>
4498
Bill Wendling7c78dbb2010-02-25 21:23:24 +00004499<p>The optional <tt>!nontemporal</tt> metadata must reference a single
4500 metatadata name &lt;index&gt; corresponding to a metadata node with
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00004501 one <tt>i32</tt> entry of value 1. The existence of
Bill Wendling7c78dbb2010-02-25 21:23:24 +00004502 the <tt>!nontemporal</tt> metatadata on the instruction tells the optimizer
4503 and code generator that this load is not expected to be reused in the cache.
4504 The code generator may select special instructions to save cache bandwidth,
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00004505 such as the <tt>MOVNT</tt> instruction on x86.</p>
David Greene8939b0d2010-02-16 20:50:18 +00004506
Chris Lattner2b7d3202002-05-06 03:03:22 +00004507<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004508<p>The location of memory pointed to is loaded. If the value being loaded is of
4509 scalar type then the number of bytes read does not exceed the minimum number
4510 of bytes needed to hold all bits of the type. For example, loading an
4511 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
4512 <tt>i20</tt> with a size that is not an integral number of bytes, the result
4513 is undefined if the value was not originally written using a store of the
4514 same type.</p>
4515
Chris Lattner2b7d3202002-05-06 03:03:22 +00004516<h5>Examples:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004517<pre>
4518 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4519 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
Reid Spencerca86e162006-12-31 07:07:53 +00004520 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
Chris Lattner2b7d3202002-05-06 03:03:22 +00004521</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004522
Misha Brukman9d0919f2003-11-08 01:05:38 +00004523</div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004524
Chris Lattner2b7d3202002-05-06 03:03:22 +00004525<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004526<h4>
4527 <a name="i_store">'<tt>store</tt>' Instruction</a>
4528</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004529
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004530<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004531
Chris Lattner2b7d3202002-05-06 03:03:22 +00004532<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004533<pre>
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00004534 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>
4535 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 +00004536</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004537
Chris Lattner2b7d3202002-05-06 03:03:22 +00004538<h5>Overview:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00004539<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004540
Chris Lattner2b7d3202002-05-06 03:03:22 +00004541<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004542<p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
4543 and an address at which to store it. The type of the
4544 '<tt>&lt;pointer&gt;</tt>' operand must be a pointer to
4545 the <a href="#t_firstclass">first class</a> type of the
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00004546 '<tt>&lt;value&gt;</tt>' operand. If the <tt>store</tt> is marked as
4547 <tt>volatile</tt>, then the optimizer is not allowed to modify the number or
4548 order of execution of this <tt>store</tt> with other <a
4549 href="#volatile">volatile operations</a>.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004550
4551<p>The optional constant "align" argument specifies the alignment of the
4552 operation (that is, the alignment of the memory address). A value of 0 or an
4553 omitted "align" argument means that the operation has the preferential
4554 alignment for the target. It is the responsibility of the code emitter to
4555 ensure that the alignment information is correct. Overestimating the
4556 alignment results in an undefined behavior. Underestimating the alignment may
4557 produce less efficient code. An alignment of 1 is always safe.</p>
4558
David Greene8939b0d2010-02-16 20:50:18 +00004559<p>The optional !nontemporal metadata must reference a single metatadata
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00004560 name &lt;index&gt; corresponding to a metadata node with one i32 entry of
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00004561 value 1. The existence of the !nontemporal metatadata on the
David Greene8939b0d2010-02-16 20:50:18 +00004562 instruction tells the optimizer and code generator that this load is
4563 not expected to be reused in the cache. The code generator may
4564 select special instructions to save cache bandwidth, such as the
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00004565 MOVNT instruction on x86.</p>
David Greene8939b0d2010-02-16 20:50:18 +00004566
4567
Chris Lattner261efe92003-11-25 01:02:51 +00004568<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004569<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>' at the
4570 location specified by the '<tt>&lt;pointer&gt;</tt>' operand. If
4571 '<tt>&lt;value&gt;</tt>' is of scalar type then the number of bytes written
4572 does not exceed the minimum number of bytes needed to hold all bits of the
4573 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
4574 writing a value of a type like <tt>i20</tt> with a size that is not an
4575 integral number of bytes, it is unspecified what happens to the extra bits
4576 that do not belong to the type, but they will typically be overwritten.</p>
4577
Chris Lattner2b7d3202002-05-06 03:03:22 +00004578<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004579<pre>
4580 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling8c6c72d2007-10-22 05:10:05 +00004581 store i32 3, i32* %ptr <i>; yields {void}</i>
4582 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
Chris Lattner2b7d3202002-05-06 03:03:22 +00004583</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004584
Reid Spencer47ce1792006-11-09 21:15:49 +00004585</div>
4586
Chris Lattner2b7d3202002-05-06 03:03:22 +00004587<!-- _______________________________________________________________________ -->
Eli Friedman47f35132011-07-25 23:16:38 +00004588<div class="doc_subsubsection"> <a name="i_fence">'<tt>fence</tt>'
4589Instruction</a> </div>
4590
4591<div class="doc_text">
4592
4593<h5>Syntax:</h5>
4594<pre>
4595 fence [singlethread] &lt;ordering&gt; <i>; yields {void}</i>
4596</pre>
4597
4598<h5>Overview:</h5>
4599<p>The '<tt>fence</tt>' instruction is used to introduce happens-before edges
4600between operations.</p>
4601
4602<h5>Arguments:</h5> <p>'<code>fence</code>' instructions take an <a
4603href="#ordering">ordering</a> argument which defines what
4604<i>synchronizes-with</i> edges they add. They can only be given
4605<code>acquire</code>, <code>release</code>, <code>acq_rel</code>, and
4606<code>seq_cst</code> orderings.</p>
4607
4608<h5>Semantics:</h5>
4609<p>A fence <var>A</var> which has (at least) <code>release</code> ordering
4610semantics <i>synchronizes with</i> a fence <var>B</var> with (at least)
4611<code>acquire</code> ordering semantics if and only if there exist atomic
4612operations <var>X</var> and <var>Y</var>, both operating on some atomic object
4613<var>M</var>, such that <var>A</var> is sequenced before <var>X</var>,
4614<var>X</var> modifies <var>M</var> (either directly or through some side effect
4615of a sequence headed by <var>X</var>), <var>Y</var> is sequenced before
4616<var>B</var>, and <var>Y</var> observes <var>M</var>. This provides a
4617<i>happens-before</i> dependency between <var>A</var> and <var>B</var>. Rather
4618than an explicit <code>fence</code>, one (but not both) of the atomic operations
4619<var>X</var> or <var>Y</var> might provide a <code>release</code> or
4620<code>acquire</code> (resp.) ordering constraint and still
4621<i>synchronize-with</i> the explicit <code>fence</code> and establish the
4622<i>happens-before</i> edge.</p>
4623
4624<p>A <code>fence</code> which has <code>seq_cst</code> ordering, in addition to
4625having both <code>acquire</code> and <code>release</code> semantics specified
4626above, participates in the global program order of other <code>seq_cst</code>
4627operations and/or fences.</p>
4628
4629<p>The optional "<a href="#singlethread"><code>singlethread</code></a>" argument
4630specifies that the fence only synchronizes with other fences in the same
4631thread. (This is useful for interacting with signal handlers.)</p>
4632
4633<p>FIXME: This instruction is a work in progress; until it is finished, use
4634 llvm.memory.barrier.
4635
4636<h5>Example:</h5>
4637<pre>
4638 fence acquire <i>; yields {void}</i>
4639 fence singlethread seq_cst <i>; yields {void}</i>
4640</pre>
4641
4642</div>
4643
4644<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004645<h4>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004646 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004647</h4>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004648
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004649<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004650
Chris Lattner7faa8832002-04-14 06:13:44 +00004651<h5>Syntax:</h5>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004652<pre>
Matthijs Kooijmane49d0bc2008-10-13 13:44:15 +00004653 &lt;result&gt; = getelementptr &lt;pty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
Dan Gohmandd8004d2009-07-27 21:53:46 +00004654 &lt;result&gt; = getelementptr inbounds &lt;pty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004655</pre>
4656
Chris Lattner7faa8832002-04-14 06:13:44 +00004657<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004658<p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004659 subelement of an <a href="#t_aggregate">aggregate</a> data structure.
4660 It performs address calculation only and does not access memory.</p>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004661
Chris Lattner7faa8832002-04-14 06:13:44 +00004662<h5>Arguments:</h5>
Matthijs Kooijmane49d0bc2008-10-13 13:44:15 +00004663<p>The first argument is always a pointer, and forms the basis of the
Chris Lattnerc8eef442009-07-29 06:44:13 +00004664 calculation. The remaining arguments are indices that indicate which of the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004665 elements of the aggregate object are indexed. The interpretation of each
4666 index is dependent on the type being indexed into. The first index always
4667 indexes the pointer value given as the first argument, the second index
4668 indexes a value of the type pointed to (not necessarily the value directly
4669 pointed to, since the first index can be non-zero), etc. The first type
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004670 indexed into must be a pointer value, subsequent types can be arrays,
Chris Lattner61c70e92010-08-28 04:09:24 +00004671 vectors, and structs. Note that subsequent types being indexed into
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004672 can never be pointers, since that would require loading the pointer before
4673 continuing calculation.</p>
Matthijs Kooijmane49d0bc2008-10-13 13:44:15 +00004674
4675<p>The type of each index argument depends on the type it is indexing into.
Chris Lattner61c70e92010-08-28 04:09:24 +00004676 When indexing into a (optionally packed) structure, only <tt>i32</tt>
Chris Lattnerfdfeb692010-02-12 20:49:41 +00004677 integer <b>constants</b> are allowed. When indexing into an array, pointer
4678 or vector, integers of any width are allowed, and they are not required to be
Chris Lattnerc8eef442009-07-29 06:44:13 +00004679 constant.</p>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004680
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004681<p>For example, let's consider a C code fragment and how it gets compiled to
4682 LLVM:</p>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004683
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00004684<pre class="doc_code">
Bill Wendling2f7a8b02007-05-29 09:04:49 +00004685struct RT {
4686 char A;
Chris Lattnercabc8462007-05-29 15:43:56 +00004687 int B[10][20];
Bill Wendling2f7a8b02007-05-29 09:04:49 +00004688 char C;
4689};
4690struct ST {
Chris Lattnercabc8462007-05-29 15:43:56 +00004691 int X;
Bill Wendling2f7a8b02007-05-29 09:04:49 +00004692 double Y;
4693 struct RT Z;
4694};
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004695
Chris Lattnercabc8462007-05-29 15:43:56 +00004696int *foo(struct ST *s) {
Bill Wendling2f7a8b02007-05-29 09:04:49 +00004697 return &amp;s[1].Z.B[5][13];
4698}
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004699</pre>
4700
Misha Brukman9d0919f2003-11-08 01:05:38 +00004701<p>The LLVM code generated by the GCC frontend is:</p>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004702
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00004703<pre class="doc_code">
Chris Lattnere7886e42009-01-11 20:53:49 +00004704%RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
4705%ST = <a href="#namedtypes">type</a> { i32, double, %RT }
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004706
Dan Gohman4df605b2009-07-25 02:23:48 +00004707define i32* @foo(%ST* %s) {
Bill Wendling2f7a8b02007-05-29 09:04:49 +00004708entry:
4709 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
4710 ret i32* %reg
4711}
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004712</pre>
4713
Chris Lattner7faa8832002-04-14 06:13:44 +00004714<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00004715<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004716 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
4717 }</tt>' type, a structure. The second index indexes into the third element
4718 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
4719 i8 }</tt>' type, another structure. The third index indexes into the second
4720 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
4721 array. The two dimensions of the array are subscripted into, yielding an
4722 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
4723 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004724
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004725<p>Note that it is perfectly legal to index partially through a structure,
4726 returning a pointer to an inner element. Because of this, the LLVM code for
4727 the given testcase is equivalent to:</p>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004728
4729<pre>
Dan Gohman4df605b2009-07-25 02:23:48 +00004730 define i32* @foo(%ST* %s) {
Reid Spencerca86e162006-12-31 07:07:53 +00004731 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
Jeff Cohen6f1cc772007-04-22 01:17:39 +00004732 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
4733 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
Reid Spencerca86e162006-12-31 07:07:53 +00004734 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
4735 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
4736 ret i32* %t5
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004737 }
Chris Lattner6536cfe2002-05-06 22:08:29 +00004738</pre>
Chris Lattnere67a9512005-06-24 17:22:57 +00004739
Dan Gohmandd8004d2009-07-27 21:53:46 +00004740<p>If the <tt>inbounds</tt> keyword is present, the result value of the
Dan Gohman27ef9972010-04-23 15:23:32 +00004741 <tt>getelementptr</tt> is a <a href="#trapvalues">trap value</a> if the
4742 base pointer is not an <i>in bounds</i> address of an allocated object,
4743 or if any of the addresses that would be formed by successive addition of
4744 the offsets implied by the indices to the base address with infinitely
4745 precise arithmetic are not an <i>in bounds</i> address of that allocated
4746 object. The <i>in bounds</i> addresses for an allocated object are all
4747 the addresses that point into the object, plus the address one byte past
4748 the end.</p>
Dan Gohmandd8004d2009-07-27 21:53:46 +00004749
4750<p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
4751 the base address with silently-wrapping two's complement arithmetic, and
4752 the result value of the <tt>getelementptr</tt> may be outside the object
4753 pointed to by the base pointer. The result value may not necessarily be
4754 used to access memory though, even if it happens to point into allocated
4755 storage. See the <a href="#pointeraliasing">Pointer Aliasing Rules</a>
4756 section for more information.</p>
4757
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004758<p>The getelementptr instruction is often confusing. For some more insight into
4759 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
Chris Lattner884a9702006-08-15 00:45:58 +00004760
Chris Lattner7faa8832002-04-14 06:13:44 +00004761<h5>Example:</h5>
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004762<pre>
Reid Spencerca86e162006-12-31 07:07:53 +00004763 <i>; yields [12 x i8]*:aptr</i>
Matthijs Kooijmane49d0bc2008-10-13 13:44:15 +00004764 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
4765 <i>; yields i8*:vptr</i>
Dan Gohman0e451ce2008-10-14 16:51:45 +00004766 %vptr = getelementptr {i32, &lt;2 x i8&gt;}* %svptr, i64 0, i32 1, i32 1
Matthijs Kooijmane49d0bc2008-10-13 13:44:15 +00004767 <i>; yields i8*:eptr</i>
4768 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
Sanjiv Gupta9f805c22009-04-25 07:27:44 +00004769 <i>; yields i32*:iptr</i>
Sanjiv Gupta16ffa802009-04-24 16:38:13 +00004770 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004771</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004772
Chris Lattnerf74d5c72004-04-05 01:30:49 +00004773</div>
Reid Spencer47ce1792006-11-09 21:15:49 +00004774
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004775</div>
4776
Chris Lattner00950542001-06-06 20:29:01 +00004777<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004778<h3>
4779 <a name="convertops">Conversion Operations</a>
4780</h3>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004781
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004782<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004783
Reid Spencer2fd21e62006-11-08 01:18:52 +00004784<p>The instructions in this category are the conversion instructions (casting)
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004785 which all take a single operand and a type. They perform various bit
4786 conversions on the operand.</p>
4787
Chris Lattner6536cfe2002-05-06 22:08:29 +00004788<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004789<h4>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004790 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004791</h4>
4792
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004793<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004794
4795<h5>Syntax:</h5>
4796<pre>
4797 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4798</pre>
4799
4800<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004801<p>The '<tt>trunc</tt>' instruction truncates its operand to the
4802 type <tt>ty2</tt>.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004803
4804<h5>Arguments:</h5>
Nadav Rotem8c20ec52011-02-24 21:01:34 +00004805<p>The '<tt>trunc</tt>' instruction takes a value to trunc, and a type to trunc it to.
4806 Both types must be of <a href="#t_integer">integer</a> types, or vectors
4807 of the same number of integers.
4808 The bit size of the <tt>value</tt> must be larger than
4809 the bit size of the destination type, <tt>ty2</tt>.
4810 Equal sized types are not allowed.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004811
4812<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004813<p>The '<tt>trunc</tt>' instruction truncates the high order bits
4814 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
4815 source size must be larger than the destination size, <tt>trunc</tt> cannot
4816 be a <i>no-op cast</i>. It will always truncate bits.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004817
4818<h5>Example:</h5>
4819<pre>
Nadav Rotem8c20ec52011-02-24 21:01:34 +00004820 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
4821 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
4822 %Z = trunc i32 122 to i1 <i>; yields i1:false</i>
4823 %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 +00004824</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004825
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004826</div>
4827
4828<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004829<h4>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004830 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004831</h4>
4832
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004833<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004834
4835<h5>Syntax:</h5>
4836<pre>
4837 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4838</pre>
4839
4840<h5>Overview:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004841<p>The '<tt>zext</tt>' instruction zero extends its operand to type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004842 <tt>ty2</tt>.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004843
4844
4845<h5>Arguments:</h5>
Nadav Rotemed9b9342011-02-20 12:37:50 +00004846<p>The '<tt>zext</tt>' instruction takes a value to cast, and a type to cast it to.
4847 Both types must be of <a href="#t_integer">integer</a> types, or vectors
4848 of the same number of integers.
4849 The bit size of the <tt>value</tt> must be smaller than
4850 the bit size of the destination type,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004851 <tt>ty2</tt>.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004852
4853<h5>Semantics:</h5>
4854<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004855 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004856
Reid Spencerb5929522007-01-12 15:46:11 +00004857<p>When zero extending from i1, the result will always be either 0 or 1.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004858
4859<h5>Example:</h5>
4860<pre>
Reid Spencerca86e162006-12-31 07:07:53 +00004861 %X = zext i32 257 to i64 <i>; yields i64:257</i>
Reid Spencerc78f3372007-01-12 03:35:51 +00004862 %Y = zext i1 true to i32 <i>; yields i32:1</i>
Nadav Rotemed9b9342011-02-20 12:37:50 +00004863 %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 +00004864</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004865
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004866</div>
4867
4868<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004869<h4>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004870 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004871</h4>
4872
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004873<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004874
4875<h5>Syntax:</h5>
4876<pre>
4877 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4878</pre>
4879
4880<h5>Overview:</h5>
4881<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
4882
4883<h5>Arguments:</h5>
Nadav Rotem8c20ec52011-02-24 21:01:34 +00004884<p>The '<tt>sext</tt>' instruction takes a value to cast, and a type to cast it to.
4885 Both types must be of <a href="#t_integer">integer</a> types, or vectors
4886 of the same number of integers.
4887 The bit size of the <tt>value</tt> must be smaller than
4888 the bit size of the destination type,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004889 <tt>ty2</tt>.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004890
4891<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004892<p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
4893 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
4894 of the type <tt>ty2</tt>.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004895
Reid Spencerc78f3372007-01-12 03:35:51 +00004896<p>When sign extending from i1, the extension always results in -1 or 0.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004897
4898<h5>Example:</h5>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004899<pre>
Reid Spencerca86e162006-12-31 07:07:53 +00004900 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
Reid Spencerc78f3372007-01-12 03:35:51 +00004901 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
Nadav Rotem8c20ec52011-02-24 21:01:34 +00004902 %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 +00004903</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004904
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004905</div>
4906
4907<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004908<h4>
Reid Spencer3fa91b02006-11-09 21:48:10 +00004909 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004910</h4>
Reid Spencer3fa91b02006-11-09 21:48:10 +00004911
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004912<div>
Reid Spencer3fa91b02006-11-09 21:48:10 +00004913
4914<h5>Syntax:</h5>
Reid Spencer3fa91b02006-11-09 21:48:10 +00004915<pre>
4916 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4917</pre>
4918
4919<h5>Overview:</h5>
4920<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004921 <tt>ty2</tt>.</p>
Reid Spencer3fa91b02006-11-09 21:48:10 +00004922
4923<h5>Arguments:</h5>
4924<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004925 point</a> value to cast and a <a href="#t_floating">floating point</a> type
4926 to cast it to. The size of <tt>value</tt> must be larger than the size of
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004927 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004928 <i>no-op cast</i>.</p>
Reid Spencer3fa91b02006-11-09 21:48:10 +00004929
4930<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004931<p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004932 <a href="#t_floating">floating point</a> type to a smaller
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004933 <a href="#t_floating">floating point</a> type. If the value cannot fit
4934 within the destination type, <tt>ty2</tt>, then the results are
4935 undefined.</p>
Reid Spencer3fa91b02006-11-09 21:48:10 +00004936
4937<h5>Example:</h5>
4938<pre>
4939 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4940 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4941</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004942
Reid Spencer3fa91b02006-11-09 21:48:10 +00004943</div>
4944
4945<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004946<h4>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004947 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004948</h4>
4949
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004950<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004951
4952<h5>Syntax:</h5>
4953<pre>
4954 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4955</pre>
4956
4957<h5>Overview:</h5>
4958<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004959 floating point value.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004960
4961<h5>Arguments:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00004962<p>The '<tt>fpext</tt>' instruction takes a
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004963 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
4964 a <a href="#t_floating">floating point</a> type to cast it to. The source
4965 type must be smaller than the destination type.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004966
4967<h5>Semantics:</h5>
Reid Spencerd4448792006-11-09 23:03:26 +00004968<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004969 <a href="#t_floating">floating point</a> type to a larger
4970 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4971 used to make a <i>no-op cast</i> because it always changes bits. Use
4972 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004973
4974<h5>Example:</h5>
4975<pre>
Nick Lewycky5bb3ece2011-03-31 18:20:19 +00004976 %X = fpext float 3.125 to double <i>; yields double:3.125000e+00</i>
4977 %Y = fpext double %X to fp128 <i>; yields fp128:0xL00000000000000004000900000000000</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004978</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004979
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004980</div>
4981
4982<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004983<h4>
Reid Spencer24d6da52007-01-21 00:29:26 +00004984 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00004985</h4>
4986
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00004987<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004988
4989<h5>Syntax:</h5>
4990<pre>
Reid Spencer1539a1c2007-07-31 14:40:14 +00004991 &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 +00004992</pre>
4993
4994<h5>Overview:</h5>
Reid Spencer1539a1c2007-07-31 14:40:14 +00004995<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004996 unsigned integer equivalent of type <tt>ty2</tt>.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00004997
4998<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00004999<p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
5000 scalar or vector <a href="#t_floating">floating point</a> value, and a type
5001 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
5002 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
5003 vector integer type with the same number of elements as <tt>ty</tt></p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005004
5005<h5>Semantics:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005006<p>The '<tt>fptoui</tt>' instruction converts its
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005007 <a href="#t_floating">floating point</a> operand into the nearest (rounding
5008 towards zero) unsigned integer value. If the value cannot fit
5009 in <tt>ty2</tt>, the results are undefined.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005010
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005011<h5>Example:</h5>
5012<pre>
Reid Spencer1539a1c2007-07-31 14:40:14 +00005013 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner88519042007-09-22 03:17:52 +00005014 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Gabor Greif6a292012009-10-28 09:21:30 +00005015 %Z = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005016</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005017
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005018</div>
5019
5020<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005021<h4>
Reid Spencerd4448792006-11-09 23:03:26 +00005022 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005023</h4>
5024
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005025<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005026
5027<h5>Syntax:</h5>
5028<pre>
Reid Spencerd4448792006-11-09 23:03:26 +00005029 &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 +00005030</pre>
5031
5032<h5>Overview:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005033<p>The '<tt>fptosi</tt>' instruction converts
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005034 <a href="#t_floating">floating point</a> <tt>value</tt> to
5035 type <tt>ty2</tt>.</p>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005036
Chris Lattner6536cfe2002-05-06 22:08:29 +00005037<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005038<p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
5039 scalar or vector <a href="#t_floating">floating point</a> value, and a type
5040 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
5041 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
5042 vector integer type with the same number of elements as <tt>ty</tt></p>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005043
Chris Lattner6536cfe2002-05-06 22:08:29 +00005044<h5>Semantics:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005045<p>The '<tt>fptosi</tt>' instruction converts its
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005046 <a href="#t_floating">floating point</a> operand into the nearest (rounding
5047 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
5048 the results are undefined.</p>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005049
Chris Lattner33ba0d92001-07-09 00:26:23 +00005050<h5>Example:</h5>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005051<pre>
Reid Spencerc78f3372007-01-12 03:35:51 +00005052 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner88519042007-09-22 03:17:52 +00005053 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Gabor Greif6a292012009-10-28 09:21:30 +00005054 %Z = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005055</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005056
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005057</div>
5058
5059<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005060<h4>
Reid Spencerd4448792006-11-09 23:03:26 +00005061 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005062</h4>
5063
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005064<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005065
5066<h5>Syntax:</h5>
5067<pre>
Reid Spencerd4448792006-11-09 23:03:26 +00005068 &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 +00005069</pre>
5070
5071<h5>Overview:</h5>
Reid Spencerd4448792006-11-09 23:03:26 +00005072<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005073 integer and converts that value to the <tt>ty2</tt> type.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005074
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005075<h5>Arguments:</h5>
Nate Begemanb348d182007-11-17 03:58:34 +00005076<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005077 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
5078 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
5079 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
5080 floating point type with the same number of elements as <tt>ty</tt></p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005081
5082<h5>Semantics:</h5>
Reid Spencerd4448792006-11-09 23:03:26 +00005083<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005084 integer quantity and converts it to the corresponding floating point
5085 value. If the value cannot fit in the floating point value, the results are
5086 undefined.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005087
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005088<h5>Example:</h5>
5089<pre>
Reid Spencerca86e162006-12-31 07:07:53 +00005090 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohman0e451ce2008-10-14 16:51:45 +00005091 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005092</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005093
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005094</div>
5095
5096<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005097<h4>
Reid Spencerd4448792006-11-09 23:03:26 +00005098 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005099</h4>
5100
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005101<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005102
5103<h5>Syntax:</h5>
5104<pre>
Reid Spencerd4448792006-11-09 23:03:26 +00005105 &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 +00005106</pre>
5107
5108<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005109<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
5110 and converts that value to the <tt>ty2</tt> type.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005111
5112<h5>Arguments:</h5>
Nate Begemanb348d182007-11-17 03:58:34 +00005113<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005114 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
5115 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
5116 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
5117 floating point type with the same number of elements as <tt>ty</tt></p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005118
5119<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005120<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
5121 quantity and converts it to the corresponding floating point value. If the
5122 value cannot fit in the floating point value, the results are undefined.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005123
5124<h5>Example:</h5>
5125<pre>
Reid Spencerca86e162006-12-31 07:07:53 +00005126 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohman0e451ce2008-10-14 16:51:45 +00005127 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005128</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005129
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005130</div>
5131
5132<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005133<h4>
Reid Spencer72679252006-11-11 21:00:47 +00005134 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005135</h4>
5136
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005137<div>
Reid Spencer72679252006-11-11 21:00:47 +00005138
5139<h5>Syntax:</h5>
5140<pre>
5141 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
5142</pre>
5143
5144<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005145<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
5146 the integer type <tt>ty2</tt>.</p>
Reid Spencer72679252006-11-11 21:00:47 +00005147
5148<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005149<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
5150 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
5151 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
Reid Spencer72679252006-11-11 21:00:47 +00005152
5153<h5>Semantics:</h5>
5154<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005155 <tt>ty2</tt> by interpreting the pointer value as an integer and either
5156 truncating or zero extending that value to the size of the integer type. If
5157 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
5158 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
5159 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
5160 change.</p>
Reid Spencer72679252006-11-11 21:00:47 +00005161
5162<h5>Example:</h5>
5163<pre>
Jeff Cohenb627eab2007-04-29 01:07:00 +00005164 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
5165 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
Reid Spencer72679252006-11-11 21:00:47 +00005166</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005167
Reid Spencer72679252006-11-11 21:00:47 +00005168</div>
5169
5170<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005171<h4>
Reid Spencer72679252006-11-11 21:00:47 +00005172 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005173</h4>
5174
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005175<div>
Reid Spencer72679252006-11-11 21:00:47 +00005176
5177<h5>Syntax:</h5>
5178<pre>
5179 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
5180</pre>
5181
5182<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005183<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
5184 pointer type, <tt>ty2</tt>.</p>
Reid Spencer72679252006-11-11 21:00:47 +00005185
5186<h5>Arguments:</h5>
Duncan Sands8036ca42007-03-30 12:22:09 +00005187<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005188 value to cast, and a type to cast it to, which must be a
5189 <a href="#t_pointer">pointer</a> type.</p>
Reid Spencer72679252006-11-11 21:00:47 +00005190
5191<h5>Semantics:</h5>
5192<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005193 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
5194 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
5195 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
5196 than the size of a pointer then a zero extension is done. If they are the
5197 same size, nothing is done (<i>no-op cast</i>).</p>
Reid Spencer72679252006-11-11 21:00:47 +00005198
5199<h5>Example:</h5>
5200<pre>
Jeff Cohenb627eab2007-04-29 01:07:00 +00005201 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
Gabor Greif6a292012009-10-28 09:21:30 +00005202 %Y = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
5203 %Z = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
Reid Spencer72679252006-11-11 21:00:47 +00005204</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005205
Reid Spencer72679252006-11-11 21:00:47 +00005206</div>
5207
5208<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005209<h4>
Reid Spencer5c0ef472006-11-11 23:08:07 +00005210 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005211</h4>
5212
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005213<div>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005214
5215<h5>Syntax:</h5>
5216<pre>
Reid Spencer5c0ef472006-11-11 23:08:07 +00005217 &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 +00005218</pre>
5219
5220<h5>Overview:</h5>
Reid Spencer5c0ef472006-11-11 23:08:07 +00005221<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005222 <tt>ty2</tt> without changing any bits.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005223
5224<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005225<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
5226 non-aggregate first class value, and a type to cast it to, which must also be
5227 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
5228 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
5229 identical. If the source type is a pointer, the destination type must also be
5230 a pointer. This instruction supports bitwise conversion of vectors to
5231 integers and to vectors of other types (as long as they have the same
5232 size).</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005233
5234<h5>Semantics:</h5>
Reid Spencer5c0ef472006-11-11 23:08:07 +00005235<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005236 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
5237 this conversion. The conversion is done as if the <tt>value</tt> had been
5238 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
5239 be converted to other pointer types with this instruction. To convert
5240 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
5241 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
Reid Spencer9dee3ac2006-11-08 01:11:31 +00005242
5243<h5>Example:</h5>
5244<pre>
Jeff Cohenb627eab2007-04-29 01:07:00 +00005245 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
Reid Spencerca86e162006-12-31 07:07:53 +00005246 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005247 %Z = bitcast &lt;2 x int&gt; %V to i64; <i>; yields i64: %V</i>
Chris Lattner33ba0d92001-07-09 00:26:23 +00005248</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005249
Misha Brukman9d0919f2003-11-08 01:05:38 +00005250</div>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005251
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005252</div>
5253
Reid Spencer2fd21e62006-11-08 01:18:52 +00005254<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005255<h3>
5256 <a name="otherops">Other Operations</a>
5257</h3>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005258
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005259<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005260
5261<p>The instructions in this category are the "miscellaneous" instructions, which
5262 defy better classification.</p>
5263
Reid Spencerf3a70a62006-11-18 21:50:54 +00005264<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005265<h4>
5266 <a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
5267</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005268
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005269<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005270
Reid Spencerf3a70a62006-11-18 21:50:54 +00005271<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005272<pre>
5273 &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 +00005274</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005275
Reid Spencerf3a70a62006-11-18 21:50:54 +00005276<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005277<p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
5278 boolean values based on comparison of its two integer, integer vector, or
5279 pointer operands.</p>
5280
Reid Spencerf3a70a62006-11-18 21:50:54 +00005281<h5>Arguments:</h5>
5282<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005283 the condition code indicating the kind of comparison to perform. It is not a
5284 value, just a keyword. The possible condition code are:</p>
5285
Reid Spencerf3a70a62006-11-18 21:50:54 +00005286<ol>
5287 <li><tt>eq</tt>: equal</li>
5288 <li><tt>ne</tt>: not equal </li>
5289 <li><tt>ugt</tt>: unsigned greater than</li>
5290 <li><tt>uge</tt>: unsigned greater or equal</li>
5291 <li><tt>ult</tt>: unsigned less than</li>
5292 <li><tt>ule</tt>: unsigned less or equal</li>
5293 <li><tt>sgt</tt>: signed greater than</li>
5294 <li><tt>sge</tt>: signed greater or equal</li>
5295 <li><tt>slt</tt>: signed less than</li>
5296 <li><tt>sle</tt>: signed less or equal</li>
5297</ol>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005298
Chris Lattner3b19d652007-01-15 01:54:13 +00005299<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005300 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
5301 typed. They must also be identical types.</p>
5302
Reid Spencerf3a70a62006-11-18 21:50:54 +00005303<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005304<p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
5305 condition code given as <tt>cond</tt>. The comparison performed always yields
Nick Lewyckyec38da42009-09-27 00:45:11 +00005306 either an <a href="#t_integer"><tt>i1</tt></a> or vector of <tt>i1</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005307 result, as follows:</p>
5308
Reid Spencerf3a70a62006-11-18 21:50:54 +00005309<ol>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005310 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005311 <tt>false</tt> otherwise. No sign interpretation is necessary or
5312 performed.</li>
5313
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005314 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005315 <tt>false</tt> otherwise. No sign interpretation is necessary or
5316 performed.</li>
5317
Reid Spencerf3a70a62006-11-18 21:50:54 +00005318 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005319 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5320
Reid Spencerf3a70a62006-11-18 21:50:54 +00005321 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005322 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5323 to <tt>op2</tt>.</li>
5324
Reid Spencerf3a70a62006-11-18 21:50:54 +00005325 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005326 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5327
Reid Spencerf3a70a62006-11-18 21:50:54 +00005328 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005329 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5330
Reid Spencerf3a70a62006-11-18 21:50:54 +00005331 <li><tt>sgt</tt>: interprets the operands as signed values and yields
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005332 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5333
Reid Spencerf3a70a62006-11-18 21:50:54 +00005334 <li><tt>sge</tt>: interprets the operands as signed values and yields
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005335 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5336 to <tt>op2</tt>.</li>
5337
Reid Spencerf3a70a62006-11-18 21:50:54 +00005338 <li><tt>slt</tt>: interprets the operands as signed values and yields
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005339 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5340
Reid Spencerf3a70a62006-11-18 21:50:54 +00005341 <li><tt>sle</tt>: interprets the operands as signed values and yields
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005342 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Reid Spencerf3a70a62006-11-18 21:50:54 +00005343</ol>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005344
Reid Spencerf3a70a62006-11-18 21:50:54 +00005345<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005346 values are compared as if they were integers.</p>
5347
5348<p>If the operands are integer vectors, then they are compared element by
5349 element. The result is an <tt>i1</tt> vector with the same number of elements
5350 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
Reid Spencerf3a70a62006-11-18 21:50:54 +00005351
5352<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005353<pre>
5354 &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
Reid Spencerca86e162006-12-31 07:07:53 +00005355 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
5356 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
5357 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
5358 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
5359 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
Reid Spencerf3a70a62006-11-18 21:50:54 +00005360</pre>
Dan Gohmanc766f722009-01-22 01:39:38 +00005361
5362<p>Note that the code generator does not yet support vector types with
5363 the <tt>icmp</tt> instruction.</p>
5364
Reid Spencerf3a70a62006-11-18 21:50:54 +00005365</div>
5366
5367<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005368<h4>
5369 <a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
5370</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005371
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005372<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005373
Reid Spencerf3a70a62006-11-18 21:50:54 +00005374<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005375<pre>
5376 &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 +00005377</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005378
Reid Spencerf3a70a62006-11-18 21:50:54 +00005379<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005380<p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
5381 values based on comparison of its operands.</p>
5382
5383<p>If the operands are floating point scalars, then the result type is a boolean
Nick Lewyckyec38da42009-09-27 00:45:11 +00005384(<a href="#t_integer"><tt>i1</tt></a>).</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005385
5386<p>If the operands are floating point vectors, then the result type is a vector
5387 of boolean with the same number of elements as the operands being
5388 compared.</p>
5389
Reid Spencerf3a70a62006-11-18 21:50:54 +00005390<h5>Arguments:</h5>
5391<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005392 the condition code indicating the kind of comparison to perform. It is not a
5393 value, just a keyword. The possible condition code are:</p>
5394
Reid Spencerf3a70a62006-11-18 21:50:54 +00005395<ol>
Reid Spencerb7f26282006-11-19 03:00:14 +00005396 <li><tt>false</tt>: no comparison, always returns false</li>
Reid Spencerf3a70a62006-11-18 21:50:54 +00005397 <li><tt>oeq</tt>: ordered and equal</li>
5398 <li><tt>ogt</tt>: ordered and greater than </li>
5399 <li><tt>oge</tt>: ordered and greater than or equal</li>
5400 <li><tt>olt</tt>: ordered and less than </li>
5401 <li><tt>ole</tt>: ordered and less than or equal</li>
5402 <li><tt>one</tt>: ordered and not equal</li>
5403 <li><tt>ord</tt>: ordered (no nans)</li>
5404 <li><tt>ueq</tt>: unordered or equal</li>
5405 <li><tt>ugt</tt>: unordered or greater than </li>
5406 <li><tt>uge</tt>: unordered or greater than or equal</li>
5407 <li><tt>ult</tt>: unordered or less than </li>
5408 <li><tt>ule</tt>: unordered or less than or equal</li>
5409 <li><tt>une</tt>: unordered or not equal</li>
5410 <li><tt>uno</tt>: unordered (either nans)</li>
Reid Spencerb7f26282006-11-19 03:00:14 +00005411 <li><tt>true</tt>: no comparison, always returns true</li>
Reid Spencerf3a70a62006-11-18 21:50:54 +00005412</ol>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005413
Jeff Cohenb627eab2007-04-29 01:07:00 +00005414<p><i>Ordered</i> means that neither operand is a QNAN while
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005415 <i>unordered</i> means that either operand may be a QNAN.</p>
5416
5417<p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
5418 a <a href="#t_floating">floating point</a> type or
5419 a <a href="#t_vector">vector</a> of floating point type. They must have
5420 identical types.</p>
5421
Reid Spencerf3a70a62006-11-18 21:50:54 +00005422<h5>Semantics:</h5>
Gabor Greiffb224a22008-08-07 21:46:00 +00005423<p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005424 according to the condition code given as <tt>cond</tt>. If the operands are
5425 vectors, then the vectors are compared element by element. Each comparison
Nick Lewyckyec38da42009-09-27 00:45:11 +00005426 performed always yields an <a href="#t_integer">i1</a> result, as
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005427 follows:</p>
5428
Reid Spencerf3a70a62006-11-18 21:50:54 +00005429<ol>
5430 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005431
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005432 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005433 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5434
Reid Spencerb7f26282006-11-19 03:00:14 +00005435 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00005436 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005437
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005438 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005439 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5440
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005441 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005442 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5443
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005444 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005445 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5446
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005447 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005448 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5449
Reid Spencerb7f26282006-11-19 03:00:14 +00005450 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005451
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005452 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005453 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5454
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005455 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005456 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5457
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005458 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005459 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5460
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005461 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005462 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5463
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005464 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005465 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5466
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005467 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005468 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5469
Reid Spencerb7f26282006-11-19 03:00:14 +00005470 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005471
Reid Spencerf3a70a62006-11-18 21:50:54 +00005472 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
5473</ol>
Reid Spencerf3a70a62006-11-18 21:50:54 +00005474
5475<h5>Example:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005476<pre>
5477 &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
Dan Gohmanf72fb672008-09-09 01:02:47 +00005478 &lt;result&gt; = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
5479 &lt;result&gt; = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
5480 &lt;result&gt; = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
Reid Spencerf3a70a62006-11-18 21:50:54 +00005481</pre>
Dan Gohmanc766f722009-01-22 01:39:38 +00005482
5483<p>Note that the code generator does not yet support vector types with
5484 the <tt>fcmp</tt> instruction.</p>
5485
Reid Spencerf3a70a62006-11-18 21:50:54 +00005486</div>
5487
Reid Spencer2fd21e62006-11-08 01:18:52 +00005488<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005489<h4>
Chris Lattner5568e942008-05-20 20:48:21 +00005490 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005491</h4>
Chris Lattner5568e942008-05-20 20:48:21 +00005492
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005493<div>
Chris Lattner5568e942008-05-20 20:48:21 +00005494
Reid Spencer2fd21e62006-11-08 01:18:52 +00005495<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005496<pre>
5497 &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...
5498</pre>
Chris Lattner5568e942008-05-20 20:48:21 +00005499
Reid Spencer2fd21e62006-11-08 01:18:52 +00005500<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005501<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in the
5502 SSA graph representing the function.</p>
5503
Reid Spencer2fd21e62006-11-08 01:18:52 +00005504<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005505<p>The type of the incoming values is specified with the first type field. After
5506 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
5507 one pair for each predecessor basic block of the current block. Only values
5508 of <a href="#t_firstclass">first class</a> type may be used as the value
5509 arguments to the PHI node. Only labels may be used as the label
5510 arguments.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00005511
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005512<p>There must be no non-phi instructions between the start of a basic block and
5513 the PHI instructions: i.e. PHI instructions must be first in a basic
5514 block.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00005515
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005516<p>For the purposes of the SSA form, the use of each incoming value is deemed to
5517 occur on the edge from the corresponding predecessor block to the current
5518 block (but after any definition of an '<tt>invoke</tt>' instruction's return
5519 value on the same edge).</p>
Jay Foadd2449092009-06-03 10:20:10 +00005520
Reid Spencer2fd21e62006-11-08 01:18:52 +00005521<h5>Semantics:</h5>
Jeff Cohenb627eab2007-04-29 01:07:00 +00005522<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005523 specified by the pair corresponding to the predecessor basic block that
5524 executed just prior to the current block.</p>
Chris Lattner5568e942008-05-20 20:48:21 +00005525
Reid Spencer2fd21e62006-11-08 01:18:52 +00005526<h5>Example:</h5>
Chris Lattner5568e942008-05-20 20:48:21 +00005527<pre>
5528Loop: ; Infinite loop that counts from 0 on up...
5529 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
5530 %nextindvar = add i32 %indvar, 1
5531 br label %Loop
5532</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005533
Reid Spencer2fd21e62006-11-08 01:18:52 +00005534</div>
5535
Chris Lattnercc37aae2004-03-12 05:50:16 +00005536<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005537<h4>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005538 <a name="i_select">'<tt>select</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005539</h4>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005540
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005541<div>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005542
5543<h5>Syntax:</h5>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005544<pre>
Dan Gohmanf72fb672008-09-09 01:02:47 +00005545 &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>
5546
Dan Gohman0e451ce2008-10-14 16:51:45 +00005547 <i>selty</i> is either i1 or {&lt;N x i1&gt;}
Chris Lattnercc37aae2004-03-12 05:50:16 +00005548</pre>
5549
5550<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005551<p>The '<tt>select</tt>' instruction is used to choose one value based on a
5552 condition, without branching.</p>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005553
5554
5555<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005556<p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
5557 values indicating the condition, and two values of the
5558 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
5559 vectors and the condition is a scalar, then entire vectors are selected, not
5560 individual elements.</p>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005561
5562<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005563<p>If the condition is an i1 and it evaluates to 1, the instruction returns the
5564 first value argument; otherwise, it returns the second value argument.</p>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005565
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005566<p>If the condition is a vector of i1, then the value arguments must be vectors
5567 of the same size, and the selection is done element by element.</p>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005568
5569<h5>Example:</h5>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005570<pre>
Reid Spencerc78f3372007-01-12 03:35:51 +00005571 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
Chris Lattnercc37aae2004-03-12 05:50:16 +00005572</pre>
Dan Gohmanc766f722009-01-22 01:39:38 +00005573
5574<p>Note that the code generator does not yet support conditions
5575 with vector type.</p>
5576
Chris Lattnercc37aae2004-03-12 05:50:16 +00005577</div>
5578
Robert Bocchino05ccd702006-01-15 20:48:27 +00005579<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005580<h4>
Chris Lattner2bff5242005-05-06 05:47:36 +00005581 <a name="i_call">'<tt>call</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005582</h4>
Chris Lattner2bff5242005-05-06 05:47:36 +00005583
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005584<div>
Chris Lattner2bff5242005-05-06 05:47:36 +00005585
Chris Lattner00950542001-06-06 20:29:01 +00005586<h5>Syntax:</h5>
Chris Lattner2bff5242005-05-06 05:47:36 +00005587<pre>
Devang Patel307e8ab2008-10-07 17:48:33 +00005588 &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 +00005589</pre>
5590
Chris Lattner00950542001-06-06 20:29:01 +00005591<h5>Overview:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00005592<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
Chris Lattner2bff5242005-05-06 05:47:36 +00005593
Chris Lattner00950542001-06-06 20:29:01 +00005594<h5>Arguments:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00005595<p>This instruction requires several arguments:</p>
Chris Lattner2bff5242005-05-06 05:47:36 +00005596
Chris Lattner6536cfe2002-05-06 22:08:29 +00005597<ol>
Jeffrey Yasskin95fa80a2010-01-09 19:44:16 +00005598 <li>The optional "tail" marker indicates that the callee function does not
5599 access any allocas or varargs in the caller. Note that calls may be
5600 marked "tail" even if they do not occur before
5601 a <a href="#i_ret"><tt>ret</tt></a> instruction. If the "tail" marker is
5602 present, the function call is eligible for tail call optimization,
5603 but <a href="CodeGenerator.html#tailcallopt">might not in fact be
Evan Chengdc444e92010-03-08 21:05:02 +00005604 optimized into a jump</a>. The code generator may optimize calls marked
5605 "tail" with either 1) automatic <a href="CodeGenerator.html#sibcallopt">
5606 sibling call optimization</a> when the caller and callee have
5607 matching signatures, or 2) forced tail call optimization when the
5608 following extra requirements are met:
Jeffrey Yasskin95fa80a2010-01-09 19:44:16 +00005609 <ul>
5610 <li>Caller and callee both have the calling
5611 convention <tt>fastcc</tt>.</li>
5612 <li>The call is in tail position (ret immediately follows call and ret
5613 uses value of call or is void).</li>
5614 <li>Option <tt>-tailcallopt</tt> is enabled,
Dan Gohmanfbbee8d2010-03-02 01:08:11 +00005615 or <code>llvm::GuaranteedTailCallOpt</code> is <code>true</code>.</li>
Jeffrey Yasskin95fa80a2010-01-09 19:44:16 +00005616 <li><a href="CodeGenerator.html#tailcallopt">Platform specific
5617 constraints are met.</a></li>
5618 </ul>
5619 </li>
Devang Patelf642f472008-10-06 18:50:38 +00005620
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005621 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
5622 convention</a> the call should use. If none is specified, the call
Jeffrey Yasskin95fa80a2010-01-09 19:44:16 +00005623 defaults to using C calling conventions. The calling convention of the
5624 call must match the calling convention of the target function, or else the
5625 behavior is undefined.</li>
Devang Patelf642f472008-10-06 18:50:38 +00005626
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005627 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
5628 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
5629 '<tt>inreg</tt>' attributes are valid here.</li>
5630
5631 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
5632 type of the return value. Functions that return no value are marked
5633 <tt><a href="#t_void">void</a></tt>.</li>
5634
5635 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
5636 being invoked. The argument types must match the types implied by this
5637 signature. This type can be omitted if the function is not varargs and if
5638 the function type does not return a pointer to a function.</li>
5639
5640 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
5641 be invoked. In most cases, this is a direct function invocation, but
5642 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
5643 to function value.</li>
5644
5645 <li>'<tt>function args</tt>': argument list whose types match the function
Chris Lattner0724fbd2010-03-02 06:36:51 +00005646 signature argument types and parameter attributes. All arguments must be
5647 of <a href="#t_firstclass">first class</a> type. If the function
5648 signature indicates the function accepts a variable number of arguments,
5649 the extra arguments can be specified.</li>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005650
5651 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
5652 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
5653 '<tt>readnone</tt>' attributes are valid here.</li>
Chris Lattner6536cfe2002-05-06 22:08:29 +00005654</ol>
Chris Lattner2bff5242005-05-06 05:47:36 +00005655
Chris Lattner00950542001-06-06 20:29:01 +00005656<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005657<p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
5658 a specified function, with its incoming arguments bound to the specified
5659 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
5660 function, control flow continues with the instruction after the function
5661 call, and the return value of the function is bound to the result
5662 argument.</p>
Chris Lattner2bff5242005-05-06 05:47:36 +00005663
Chris Lattner00950542001-06-06 20:29:01 +00005664<h5>Example:</h5>
Chris Lattner2bff5242005-05-06 05:47:36 +00005665<pre>
Nick Lewyckydb7e3c92007-09-08 13:57:50 +00005666 %retval = call i32 @test(i32 %argc)
Dan Gohmanfe47aae2010-05-28 17:13:49 +00005667 call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42) <i>; yields i32</i>
Chris Lattner772fccf2008-03-21 17:24:17 +00005668 %X = tail call i32 @foo() <i>; yields i32</i>
5669 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
5670 call void %foo(i8 97 signext)
Devang Patelc3fc6df2008-03-10 20:49:15 +00005671
5672 %struct.A = type { i32, i8 }
Devang Patelf642f472008-10-06 18:50:38 +00005673 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
Dan Gohmanb1e6b962008-10-04 19:00:07 +00005674 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
5675 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
Chris Lattner85a350f2008-10-08 06:26:11 +00005676 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
Matthijs Kooijmancb73d192008-10-07 10:03:45 +00005677 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
Chris Lattner2bff5242005-05-06 05:47:36 +00005678</pre>
5679
Dale Johannesen07de8d12009-09-24 18:38:21 +00005680<p>llvm treats calls to some functions with names and arguments that match the
Dale Johannesen9f8380b2009-09-25 17:04:42 +00005681standard C99 library as being the C99 library functions, and may perform
5682optimizations or generate code for them under that assumption. This is
5683something we'd like to change in the future to provide better support for
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00005684freestanding environments and non-C-based languages.</p>
Dale Johannesen07de8d12009-09-24 18:38:21 +00005685
Misha Brukman9d0919f2003-11-08 01:05:38 +00005686</div>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005687
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005688<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005689<h4>
Chris Lattnerfb6977d2006-01-13 23:26:01 +00005690 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005691</h4>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005692
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005693<div>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005694
Chris Lattner8d1a81d2003-10-18 05:51:36 +00005695<h5>Syntax:</h5>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005696<pre>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00005697 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
Chris Lattnere19d7a72004-09-27 21:51:25 +00005698</pre>
5699
Chris Lattner8d1a81d2003-10-18 05:51:36 +00005700<h5>Overview:</h5>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00005701<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005702 the "variable argument" area of a function call. It is used to implement the
5703 <tt>va_arg</tt> macro in C.</p>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005704
Chris Lattner8d1a81d2003-10-18 05:51:36 +00005705<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005706<p>This instruction takes a <tt>va_list*</tt> value and the type of the
5707 argument. It returns a value of the specified argument type and increments
5708 the <tt>va_list</tt> to point to the next argument. The actual type
5709 of <tt>va_list</tt> is target specific.</p>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005710
Chris Lattner8d1a81d2003-10-18 05:51:36 +00005711<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005712<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
5713 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
5714 to the next argument. For more information, see the variable argument
5715 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005716
5717<p>It is legal for this instruction to be called in a function which does not
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005718 take a variable number of arguments, for example, the <tt>vfprintf</tt>
5719 function.</p>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005720
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005721<p><tt>va_arg</tt> is an LLVM instruction instead of
5722 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
5723 argument.</p>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005724
Chris Lattner8d1a81d2003-10-18 05:51:36 +00005725<h5>Example:</h5>
Chris Lattnere19d7a72004-09-27 21:51:25 +00005726<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
5727
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005728<p>Note that the code generator does not yet fully support va_arg on many
5729 targets. Also, it does not currently support va_arg with aggregate types on
5730 any target.</p>
Dan Gohmanf3e60bd2009-01-12 23:12:39 +00005731
Misha Brukman9d0919f2003-11-08 01:05:38 +00005732</div>
Chris Lattner8ff75902004-01-06 05:31:32 +00005733
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005734</div>
5735
5736</div>
5737
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005738<!-- *********************************************************************** -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005739<h2><a name="intrinsics">Intrinsic Functions</a></h2>
Chris Lattner261efe92003-11-25 01:02:51 +00005740<!-- *********************************************************************** -->
Chris Lattner8ff75902004-01-06 05:31:32 +00005741
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005742<div>
Chris Lattner33aec9e2004-02-12 17:01:32 +00005743
5744<p>LLVM supports the notion of an "intrinsic function". These functions have
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005745 well known names and semantics and are required to follow certain
5746 restrictions. Overall, these intrinsics represent an extension mechanism for
5747 the LLVM language that does not require changing all of the transformations
5748 in LLVM when adding to the language (or the bitcode reader/writer, the
5749 parser, etc...).</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00005750
John Criswellfc6b8952005-05-16 16:17:45 +00005751<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005752 prefix is reserved in LLVM for intrinsic names; thus, function names may not
5753 begin with this prefix. Intrinsic functions must always be external
5754 functions: you cannot define the body of intrinsic functions. Intrinsic
5755 functions may only be used in call or invoke instructions: it is illegal to
5756 take the address of an intrinsic function. Additionally, because intrinsic
5757 functions are part of the LLVM language, it is required if any are added that
5758 they be documented here.</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00005759
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005760<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
5761 family of functions that perform the same operation but on different data
5762 types. Because LLVM can represent over 8 million different integer types,
5763 overloading is used commonly to allow an intrinsic function to operate on any
5764 integer type. One or more of the argument types or the result type can be
5765 overloaded to accept any integer type. Argument types may also be defined as
5766 exactly matching a previous argument's type or the result type. This allows
5767 an intrinsic function which accepts multiple arguments, but needs all of them
5768 to be of the same type, to only be overloaded with respect to a single
5769 argument or the result.</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00005770
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005771<p>Overloaded intrinsics will have the names of its overloaded argument types
5772 encoded into its function name, each preceded by a period. Only those types
5773 which are overloaded result in a name suffix. Arguments whose type is matched
5774 against another type do not. For example, the <tt>llvm.ctpop</tt> function
5775 can take an integer of any width and returns an integer of exactly the same
5776 integer width. This leads to a family of functions such as
5777 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
5778 %val)</tt>. Only one type, the return type, is overloaded, and only one type
5779 suffix is required. Because the argument's type is matched against the return
5780 type, it does not require its own name suffix.</p>
Reid Spencer409e28f2007-04-01 08:04:23 +00005781
Eric Christopher6c7e8a02009-12-05 02:46:03 +00005782<p>To learn how to add an intrinsic function, please see the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005783 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00005784
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005785<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005786<h3>
Chris Lattner8ff75902004-01-06 05:31:32 +00005787 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005788</h3>
Chris Lattner8ff75902004-01-06 05:31:32 +00005789
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005790<div>
Chris Lattnerd7923912004-05-23 21:06:01 +00005791
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005792<p>Variable argument support is defined in LLVM with
5793 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
5794 intrinsic functions. These functions are related to the similarly named
5795 macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005796
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005797<p>All of these functions operate on arguments that use a target-specific value
5798 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
5799 not define what this type is, so all transformations should be prepared to
5800 handle these functions regardless of the type used.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005801
Chris Lattner374ab302006-05-15 17:26:46 +00005802<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005803 instruction and the variable argument handling intrinsic functions are
5804 used.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005805
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00005806<pre class="doc_code">
Anton Korobeynikov5d522f32007-03-21 23:58:04 +00005807define i32 @test(i32 %X, ...) {
Chris Lattner33aec9e2004-02-12 17:01:32 +00005808 ; Initialize variable argument processing
Jeff Cohenb627eab2007-04-29 01:07:00 +00005809 %ap = alloca i8*
Chris Lattnerb75137d2007-01-08 07:55:15 +00005810 %ap2 = bitcast i8** %ap to i8*
Anton Korobeynikov5d522f32007-03-21 23:58:04 +00005811 call void @llvm.va_start(i8* %ap2)
Chris Lattner33aec9e2004-02-12 17:01:32 +00005812
5813 ; Read a single integer argument
Jeff Cohenb627eab2007-04-29 01:07:00 +00005814 %tmp = va_arg i8** %ap, i32
Chris Lattner33aec9e2004-02-12 17:01:32 +00005815
5816 ; Demonstrate usage of llvm.va_copy and llvm.va_end
Jeff Cohenb627eab2007-04-29 01:07:00 +00005817 %aq = alloca i8*
Chris Lattnerb75137d2007-01-08 07:55:15 +00005818 %aq2 = bitcast i8** %aq to i8*
Jeff Cohenb627eab2007-04-29 01:07:00 +00005819 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
Anton Korobeynikov5d522f32007-03-21 23:58:04 +00005820 call void @llvm.va_end(i8* %aq2)
Chris Lattner33aec9e2004-02-12 17:01:32 +00005821
5822 ; Stop processing of arguments.
Anton Korobeynikov5d522f32007-03-21 23:58:04 +00005823 call void @llvm.va_end(i8* %ap2)
Reid Spencerca86e162006-12-31 07:07:53 +00005824 ret i32 %tmp
Chris Lattner33aec9e2004-02-12 17:01:32 +00005825}
Anton Korobeynikov5d522f32007-03-21 23:58:04 +00005826
5827declare void @llvm.va_start(i8*)
5828declare void @llvm.va_copy(i8*, i8*)
5829declare void @llvm.va_end(i8*)
Chris Lattner33aec9e2004-02-12 17:01:32 +00005830</pre>
Chris Lattner8ff75902004-01-06 05:31:32 +00005831
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005832<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005833<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00005834 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005835</h4>
Chris Lattner8ff75902004-01-06 05:31:32 +00005836
5837
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005838<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005839
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005840<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005841<pre>
5842 declare void %llvm.va_start(i8* &lt;arglist&gt;)
5843</pre>
5844
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005845<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005846<p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*&lt;arglist&gt;</tt>
5847 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00005848
5849<h5>Arguments:</h5>
Dan Gohman0e451ce2008-10-14 16:51:45 +00005850<p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00005851
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005852<h5>Semantics:</h5>
Dan Gohman0e451ce2008-10-14 16:51:45 +00005853<p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005854 macro available in C. In a target-dependent way, it initializes
5855 the <tt>va_list</tt> element to which the argument points, so that the next
5856 call to <tt>va_arg</tt> will produce the first variable argument passed to
5857 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
5858 need to know the last argument of the function as the compiler can figure
5859 that out.</p>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00005860
Misha Brukman9d0919f2003-11-08 01:05:38 +00005861</div>
Chris Lattner8ff75902004-01-06 05:31:32 +00005862
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005863<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005864<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00005865 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005866</h4>
Chris Lattner8ff75902004-01-06 05:31:32 +00005867
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005868<div>
Chris Lattnerb75137d2007-01-08 07:55:15 +00005869
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005870<h5>Syntax:</h5>
5871<pre>
5872 declare void @llvm.va_end(i8* &lt;arglist&gt;)
5873</pre>
5874
5875<h5>Overview:</h5>
Jeff Cohenb627eab2007-04-29 01:07:00 +00005876<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005877 which has been initialized previously
5878 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
5879 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
Chris Lattnerb75137d2007-01-08 07:55:15 +00005880
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005881<h5>Arguments:</h5>
Jeff Cohenb627eab2007-04-29 01:07:00 +00005882<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
Chris Lattnerb75137d2007-01-08 07:55:15 +00005883
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005884<h5>Semantics:</h5>
Misha Brukman9d0919f2003-11-08 01:05:38 +00005885<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005886 macro available in C. In a target-dependent way, it destroys
5887 the <tt>va_list</tt> element to which the argument points. Calls
5888 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
5889 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
5890 with calls to <tt>llvm.va_end</tt>.</p>
Chris Lattnerb75137d2007-01-08 07:55:15 +00005891
Misha Brukman9d0919f2003-11-08 01:05:38 +00005892</div>
Chris Lattner8ff75902004-01-06 05:31:32 +00005893
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005894<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005895<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00005896 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005897</h4>
Chris Lattner8ff75902004-01-06 05:31:32 +00005898
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005899<div>
Chris Lattnerd7923912004-05-23 21:06:01 +00005900
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005901<h5>Syntax:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00005902<pre>
Anton Korobeynikovec43a062007-03-22 00:02:17 +00005903 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
Chris Lattnerd7923912004-05-23 21:06:01 +00005904</pre>
5905
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005906<h5>Overview:</h5>
Jeff Cohenb627eab2007-04-29 01:07:00 +00005907<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005908 from the source argument list to the destination argument list.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005909
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005910<h5>Arguments:</h5>
Andrew Lenharth8bf607a2005-06-18 18:28:17 +00005911<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005912 The second argument is a pointer to a <tt>va_list</tt> element to copy
5913 from.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005914
Chris Lattnerd9ad5b32003-05-08 04:57:36 +00005915<h5>Semantics:</h5>
Jeff Cohenb627eab2007-04-29 01:07:00 +00005916<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005917 macro available in C. In a target-dependent way, it copies the
5918 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
5919 element. This intrinsic is necessary because
5920 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
5921 arbitrarily complex and require, for example, memory allocation.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005922
Misha Brukman9d0919f2003-11-08 01:05:38 +00005923</div>
Chris Lattner8ff75902004-01-06 05:31:32 +00005924
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005925</div>
5926
Chris Lattner33aec9e2004-02-12 17:01:32 +00005927<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005928<h3>
Chris Lattnerd7923912004-05-23 21:06:01 +00005929 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005930</h3>
Chris Lattnerd7923912004-05-23 21:06:01 +00005931
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005932<div>
Chris Lattnerd7923912004-05-23 21:06:01 +00005933
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005934<p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
Chris Lattnerd3eda892008-08-05 18:29:16 +00005935Collection</a> (GC) requires the implementation and generation of these
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005936intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
5937roots on the stack</a>, as well as garbage collector implementations that
5938require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
5939barriers. Front-ends for type-safe garbage collected languages should generate
5940these intrinsics to make use of the LLVM garbage collectors. For more details,
5941see <a href="GarbageCollection.html">Accurate Garbage Collection with
5942LLVM</a>.</p>
Christopher Lamb303dae92007-12-17 01:00:21 +00005943
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005944<p>The garbage collection intrinsics only operate on objects in the generic
5945 address space (address space zero).</p>
Christopher Lamb303dae92007-12-17 01:00:21 +00005946
Chris Lattnerd7923912004-05-23 21:06:01 +00005947<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005948<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00005949 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005950</h4>
Chris Lattnerd7923912004-05-23 21:06:01 +00005951
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005952<div>
Chris Lattnerd7923912004-05-23 21:06:01 +00005953
5954<h5>Syntax:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00005955<pre>
Chris Lattner1df4f752007-09-21 17:30:40 +00005956 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Chris Lattnerd7923912004-05-23 21:06:01 +00005957</pre>
5958
5959<h5>Overview:</h5>
John Criswell9e2485c2004-12-10 15:51:16 +00005960<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005961 the code generator, and allows some metadata to be associated with it.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005962
5963<h5>Arguments:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00005964<p>The first argument specifies the address of a stack object that contains the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005965 root pointer. The second pointer (which must be either a constant or a
5966 global value address) contains the meta-data to be associated with the
5967 root.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005968
5969<h5>Semantics:</h5>
Chris Lattner05d67092008-04-24 05:59:56 +00005970<p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005971 location. At compile-time, the code generator generates information to allow
5972 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5973 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5974 algorithm</a>.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005975
5976</div>
5977
Chris Lattnerd7923912004-05-23 21:06:01 +00005978<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005979<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00005980 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00005981</h4>
Chris Lattnerd7923912004-05-23 21:06:01 +00005982
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00005983<div>
Chris Lattnerd7923912004-05-23 21:06:01 +00005984
5985<h5>Syntax:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00005986<pre>
Chris Lattner1df4f752007-09-21 17:30:40 +00005987 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Chris Lattnerd7923912004-05-23 21:06:01 +00005988</pre>
5989
5990<h5>Overview:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00005991<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005992 locations, allowing garbage collector implementations that require read
5993 barriers.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00005994
5995<h5>Arguments:</h5>
Chris Lattner80626e92006-03-14 20:02:51 +00005996<p>The second argument is the address to read from, which should be an address
Bill Wendlinge910b4c2009-07-20 02:29:24 +00005997 allocated from the garbage collector. The first object is a pointer to the
5998 start of the referenced object, if needed by the language runtime (otherwise
5999 null).</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00006000
6001<h5>Semantics:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00006002<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006003 instruction, but may be replaced with substantially more complex code by the
6004 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
6005 may only be used in a function which <a href="#gc">specifies a GC
6006 algorithm</a>.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00006007
6008</div>
6009
Chris Lattnerd7923912004-05-23 21:06:01 +00006010<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006011<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006012 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006013</h4>
Chris Lattnerd7923912004-05-23 21:06:01 +00006014
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006015<div>
Chris Lattnerd7923912004-05-23 21:06:01 +00006016
6017<h5>Syntax:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00006018<pre>
Chris Lattner1df4f752007-09-21 17:30:40 +00006019 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Chris Lattnerd7923912004-05-23 21:06:01 +00006020</pre>
6021
6022<h5>Overview:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00006023<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006024 locations, allowing garbage collector implementations that require write
6025 barriers (such as generational or reference counting collectors).</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00006026
6027<h5>Arguments:</h5>
Chris Lattner80626e92006-03-14 20:02:51 +00006028<p>The first argument is the reference to store, the second is the start of the
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006029 object to store it to, and the third is the address of the field of Obj to
6030 store to. If the runtime does not require a pointer to the object, Obj may
6031 be null.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00006032
6033<h5>Semantics:</h5>
Chris Lattnerd7923912004-05-23 21:06:01 +00006034<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006035 instruction, but may be replaced with substantially more complex code by the
6036 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
6037 may only be used in a function which <a href="#gc">specifies a GC
6038 algorithm</a>.</p>
Chris Lattnerd7923912004-05-23 21:06:01 +00006039
6040</div>
6041
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006042</div>
6043
Chris Lattnerd7923912004-05-23 21:06:01 +00006044<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006045<h3>
Chris Lattner10610642004-02-14 04:08:35 +00006046 <a name="int_codegen">Code Generator Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006047</h3>
Chris Lattner10610642004-02-14 04:08:35 +00006048
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006049<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006050
6051<p>These intrinsics are provided by LLVM to expose special features that may
6052 only be implemented with code generator support.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006053
Chris Lattner10610642004-02-14 04:08:35 +00006054<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006055<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006056 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006057</h4>
Chris Lattner10610642004-02-14 04:08:35 +00006058
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006059<div>
Chris Lattner10610642004-02-14 04:08:35 +00006060
6061<h5>Syntax:</h5>
6062<pre>
Anton Korobeynikovec43a062007-03-22 00:02:17 +00006063 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
Chris Lattner10610642004-02-14 04:08:35 +00006064</pre>
6065
6066<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006067<p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
6068 target-specific value indicating the return address of the current function
6069 or one of its callers.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006070
6071<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006072<p>The argument to this intrinsic indicates which function to return the address
6073 for. Zero indicates the calling function, one indicates its caller, etc.
6074 The argument is <b>required</b> to be a constant integer value.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006075
6076<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006077<p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
6078 indicating the return address of the specified call frame, or zero if it
6079 cannot be identified. The value returned by this intrinsic is likely to be
6080 incorrect or 0 for arguments other than zero, so it should only be used for
6081 debugging purposes.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006082
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006083<p>Note that calling this intrinsic does not prevent function inlining or other
6084 aggressive transformations, so the value returned may not be that of the
6085 obvious source-language caller.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006086
Chris Lattner10610642004-02-14 04:08:35 +00006087</div>
6088
Chris Lattner10610642004-02-14 04:08:35 +00006089<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006090<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006091 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006092</h4>
Chris Lattner10610642004-02-14 04:08:35 +00006093
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006094<div>
Chris Lattner10610642004-02-14 04:08:35 +00006095
6096<h5>Syntax:</h5>
6097<pre>
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006098 declare i8* @llvm.frameaddress(i32 &lt;level&gt;)
Chris Lattner10610642004-02-14 04:08:35 +00006099</pre>
6100
6101<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006102<p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
6103 target-specific frame pointer value for the specified stack frame.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006104
6105<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006106<p>The argument to this intrinsic indicates which function to return the frame
6107 pointer for. Zero indicates the calling function, one indicates its caller,
6108 etc. The argument is <b>required</b> to be a constant integer value.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006109
6110<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006111<p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
6112 indicating the frame address of the specified call frame, or zero if it
6113 cannot be identified. The value returned by this intrinsic is likely to be
6114 incorrect or 0 for arguments other than zero, so it should only be used for
6115 debugging purposes.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006116
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006117<p>Note that calling this intrinsic does not prevent function inlining or other
6118 aggressive transformations, so the value returned may not be that of the
6119 obvious source-language caller.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006120
Chris Lattner10610642004-02-14 04:08:35 +00006121</div>
6122
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00006123<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006124<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006125 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006126</h4>
Chris Lattner57e1f392006-01-13 02:03:13 +00006127
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006128<div>
Chris Lattner57e1f392006-01-13 02:03:13 +00006129
6130<h5>Syntax:</h5>
6131<pre>
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006132 declare i8* @llvm.stacksave()
Chris Lattner57e1f392006-01-13 02:03:13 +00006133</pre>
6134
6135<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006136<p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
6137 of the function stack, for use
6138 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
6139 useful for implementing language features like scoped automatic variable
6140 sized arrays in C99.</p>
Chris Lattner57e1f392006-01-13 02:03:13 +00006141
6142<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006143<p>This intrinsic returns a opaque pointer value that can be passed
6144 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
6145 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
6146 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
6147 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
6148 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
6149 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
Chris Lattner57e1f392006-01-13 02:03:13 +00006150
6151</div>
6152
6153<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006154<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006155 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006156</h4>
Chris Lattner57e1f392006-01-13 02:03:13 +00006157
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006158<div>
Chris Lattner57e1f392006-01-13 02:03:13 +00006159
6160<h5>Syntax:</h5>
6161<pre>
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006162 declare void @llvm.stackrestore(i8* %ptr)
Chris Lattner57e1f392006-01-13 02:03:13 +00006163</pre>
6164
6165<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006166<p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
6167 the function stack to the state it was in when the
6168 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
6169 executed. This is useful for implementing language features like scoped
6170 automatic variable sized arrays in C99.</p>
Chris Lattner57e1f392006-01-13 02:03:13 +00006171
6172<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006173<p>See the description
6174 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
Chris Lattner57e1f392006-01-13 02:03:13 +00006175
6176</div>
6177
Chris Lattner57e1f392006-01-13 02:03:13 +00006178<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006179<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006180 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006181</h4>
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00006182
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006183<div>
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00006184
6185<h5>Syntax:</h5>
6186<pre>
Bruno Cardoso Lopes9a767332011-06-14 04:58:37 +00006187 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 +00006188</pre>
6189
6190<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006191<p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
6192 insert a prefetch instruction if supported; otherwise, it is a noop.
6193 Prefetches have no effect on the behavior of the program but can change its
6194 performance characteristics.</p>
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00006195
6196<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006197<p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
6198 specifier determining if the fetch should be for a read (0) or write (1),
6199 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
Bruno Cardoso Lopes9a767332011-06-14 04:58:37 +00006200 locality, to (3) - extremely local keep in cache. The <tt>cache type</tt>
6201 specifies whether the prefetch is performed on the data (1) or instruction (0)
6202 cache. The <tt>rw</tt>, <tt>locality</tt> and <tt>cache type</tt> arguments
6203 must be constant integers.</p>
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00006204
6205<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006206<p>This intrinsic does not modify the behavior of the program. In particular,
6207 prefetches cannot trap and do not produce a value. On targets that support
6208 this intrinsic, the prefetch can provide hints to the processor cache for
6209 better performance.</p>
Chris Lattner9a9d7ac2005-02-28 19:24:19 +00006210
6211</div>
6212
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00006213<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006214<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006215 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006216</h4>
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00006217
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006218<div>
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00006219
6220<h5>Syntax:</h5>
6221<pre>
Chris Lattner1df4f752007-09-21 17:30:40 +00006222 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00006223</pre>
6224
6225<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006226<p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
6227 Counter (PC) in a region of code to simulators and other tools. The method
6228 is target specific, but it is expected that the marker will use exported
6229 symbols to transmit the PC of the marker. The marker makes no guarantees
6230 that it will remain with any specific instruction after optimizations. It is
6231 possible that the presence of a marker will inhibit optimizations. The
6232 intended use is to be inserted after optimizations to allow correlations of
6233 simulation runs.</p>
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00006234
6235<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006236<p><tt>id</tt> is a numerical id identifying the marker.</p>
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00006237
6238<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006239<p>This intrinsic does not modify the behavior of the program. Backends that do
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00006240 not support this intrinsic may ignore it.</p>
Andrew Lenharth7f4ec3b2005-03-28 20:05:49 +00006241
6242</div>
6243
Andrew Lenharth51b8d542005-11-11 16:47:30 +00006244<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006245<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006246 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006247</h4>
Andrew Lenharth51b8d542005-11-11 16:47:30 +00006248
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006249<div>
Andrew Lenharth51b8d542005-11-11 16:47:30 +00006250
6251<h5>Syntax:</h5>
6252<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00006253 declare i64 @llvm.readcyclecounter()
Andrew Lenharth51b8d542005-11-11 16:47:30 +00006254</pre>
6255
6256<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006257<p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
6258 counter register (or similar low latency, high accuracy clocks) on those
6259 targets that support it. On X86, it should map to RDTSC. On Alpha, it
6260 should map to RPCC. As the backing counters overflow quickly (on the order
6261 of 9 seconds on alpha), this should only be used for small timings.</p>
Andrew Lenharth51b8d542005-11-11 16:47:30 +00006262
6263<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006264<p>When directly supported, reading the cycle counter should not modify any
6265 memory. Implementations are allowed to either return a application specific
6266 value or a system wide value. On backends without support, this is lowered
6267 to a constant 0.</p>
Andrew Lenharth51b8d542005-11-11 16:47:30 +00006268
6269</div>
6270
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006271</div>
6272
Chris Lattner10610642004-02-14 04:08:35 +00006273<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006274<h3>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006275 <a name="int_libc">Standard C Library Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006276</h3>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006277
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006278<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006279
6280<p>LLVM provides intrinsics for a few important standard C library functions.
6281 These intrinsics allow source-language front-ends to pass information about
6282 the alignment of the pointer arguments to the code generator, providing
6283 opportunity for more efficient code generation.</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006284
Chris Lattner33aec9e2004-02-12 17:01:32 +00006285<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006286<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006287 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006288</h4>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006289
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006290<div>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006291
6292<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006293<p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
Mon P Wange88909b2010-04-07 06:35:53 +00006294 integer bit width and for different address spaces. Not all targets support
6295 all bit widths however.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006296
Chris Lattner33aec9e2004-02-12 17:01:32 +00006297<pre>
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006298 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* &lt;dest&gt;, i8* &lt;src&gt;,
Chris Lattner9f636de2010-04-08 00:53:57 +00006299 i32 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006300 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* &lt;dest&gt;, i8* &lt;src&gt;,
Chris Lattner9f636de2010-04-08 00:53:57 +00006301 i64 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Chris Lattner33aec9e2004-02-12 17:01:32 +00006302</pre>
6303
6304<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006305<p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6306 source location to the destination location.</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006307
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006308<p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
Chris Lattner9f636de2010-04-08 00:53:57 +00006309 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6310 and the pointers can be in specified address spaces.</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006311
6312<h5>Arguments:</h5>
Chris Lattner9f636de2010-04-08 00:53:57 +00006313
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006314<p>The first argument is a pointer to the destination, the second is a pointer
6315 to the source. The third argument is an integer argument specifying the
Chris Lattner9f636de2010-04-08 00:53:57 +00006316 number of bytes to copy, the fourth argument is the alignment of the
6317 source and destination locations, and the fifth is a boolean indicating a
6318 volatile access.</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006319
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00006320<p>If the call to this intrinsic has an alignment value that is not 0 or 1,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006321 then the caller guarantees that both the source and destination pointers are
6322 aligned to that boundary.</p>
Chris Lattner3301ced2004-02-12 21:18:15 +00006323
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00006324<p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6325 <tt>llvm.memcpy</tt> call is a <a href="#volatile">volatile operation</a>.
6326 The detailed access behavior is not very cleanly specified and it is unwise
6327 to depend on it.</p>
Chris Lattner9f636de2010-04-08 00:53:57 +00006328
Chris Lattner33aec9e2004-02-12 17:01:32 +00006329<h5>Semantics:</h5>
Chris Lattner9f636de2010-04-08 00:53:57 +00006330
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006331<p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6332 source location to the destination location, which are not allowed to
6333 overlap. It copies "len" bytes of memory over. If the argument is known to
6334 be aligned to some boundary, this can be specified as the fourth argument,
6335 otherwise it should be set to 0 or 1.</p>
Chris Lattner33aec9e2004-02-12 17:01:32 +00006336
Chris Lattner33aec9e2004-02-12 17:01:32 +00006337</div>
6338
Chris Lattner0eb51b42004-02-12 18:10:10 +00006339<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006340<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006341 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006342</h4>
Chris Lattner0eb51b42004-02-12 18:10:10 +00006343
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006344<div>
Chris Lattner0eb51b42004-02-12 18:10:10 +00006345
6346<h5>Syntax:</h5>
Chris Lattner824b9582008-11-21 16:42:48 +00006347<p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
Mon P Wange88909b2010-04-07 06:35:53 +00006348 width and for different address space. Not all targets support all bit
6349 widths however.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006350
Chris Lattner0eb51b42004-02-12 18:10:10 +00006351<pre>
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006352 declare void @llvm.memmove.p0i8.p0i8.i32(i8* &lt;dest&gt;, i8* &lt;src&gt;,
Chris Lattner9f636de2010-04-08 00:53:57 +00006353 i32 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006354 declare void @llvm.memmove.p0i8.p0i8.i64(i8* &lt;dest&gt;, i8* &lt;src&gt;,
Chris Lattner9f636de2010-04-08 00:53:57 +00006355 i64 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Chris Lattner0eb51b42004-02-12 18:10:10 +00006356</pre>
6357
6358<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006359<p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
6360 source location to the destination location. It is similar to the
6361 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
6362 overlap.</p>
Chris Lattner0eb51b42004-02-12 18:10:10 +00006363
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006364<p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
Chris Lattner9f636de2010-04-08 00:53:57 +00006365 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6366 and the pointers can be in specified address spaces.</p>
Chris Lattner0eb51b42004-02-12 18:10:10 +00006367
6368<h5>Arguments:</h5>
Chris Lattner9f636de2010-04-08 00:53:57 +00006369
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006370<p>The first argument is a pointer to the destination, the second is a pointer
6371 to the source. The third argument is an integer argument specifying the
Chris Lattner9f636de2010-04-08 00:53:57 +00006372 number of bytes to copy, the fourth argument is the alignment of the
6373 source and destination locations, and the fifth is a boolean indicating a
6374 volatile access.</p>
Chris Lattner0eb51b42004-02-12 18:10:10 +00006375
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00006376<p>If the call to this intrinsic has an alignment value that is not 0 or 1,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006377 then the caller guarantees that the source and destination pointers are
6378 aligned to that boundary.</p>
Chris Lattner3301ced2004-02-12 21:18:15 +00006379
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00006380<p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6381 <tt>llvm.memmove</tt> call is a <a href="#volatile">volatile operation</a>.
6382 The detailed access behavior is not very cleanly specified and it is unwise
6383 to depend on it.</p>
Chris Lattner9f636de2010-04-08 00:53:57 +00006384
Chris Lattner0eb51b42004-02-12 18:10:10 +00006385<h5>Semantics:</h5>
Chris Lattner9f636de2010-04-08 00:53:57 +00006386
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006387<p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
6388 source location to the destination location, which may overlap. It copies
6389 "len" bytes of memory over. If the argument is known to be aligned to some
6390 boundary, this can be specified as the fourth argument, otherwise it should
6391 be set to 0 or 1.</p>
Chris Lattner0eb51b42004-02-12 18:10:10 +00006392
Chris Lattner0eb51b42004-02-12 18:10:10 +00006393</div>
6394
Chris Lattner10610642004-02-14 04:08:35 +00006395<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006396<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006397 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006398</h4>
Chris Lattner10610642004-02-14 04:08:35 +00006399
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006400<div>
Chris Lattner10610642004-02-14 04:08:35 +00006401
6402<h5>Syntax:</h5>
Chris Lattner824b9582008-11-21 16:42:48 +00006403<p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
John Criswellcdcbbfc2010-07-30 16:30:28 +00006404 width and for different address spaces. However, not all targets support all
6405 bit widths.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006406
Chris Lattner10610642004-02-14 04:08:35 +00006407<pre>
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006408 declare void @llvm.memset.p0i8.i32(i8* &lt;dest&gt;, i8 &lt;val&gt;,
Chris Lattnerff35c3f2010-04-08 00:54:34 +00006409 i32 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Dan Gohmanfe47aae2010-05-28 17:13:49 +00006410 declare void @llvm.memset.p0i8.i64(i8* &lt;dest&gt;, i8 &lt;val&gt;,
Chris Lattnerff35c3f2010-04-08 00:54:34 +00006411 i64 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Chris Lattner10610642004-02-14 04:08:35 +00006412</pre>
6413
6414<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006415<p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
6416 particular byte value.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006417
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006418<p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
John Criswellcdcbbfc2010-07-30 16:30:28 +00006419 intrinsic does not return a value and takes extra alignment/volatile
6420 arguments. Also, the destination can be in an arbitrary address space.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006421
6422<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006423<p>The first argument is a pointer to the destination to fill, the second is the
John Criswellcdcbbfc2010-07-30 16:30:28 +00006424 byte value with which to fill it, the third argument is an integer argument
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006425 specifying the number of bytes to fill, and the fourth argument is the known
John Criswellcdcbbfc2010-07-30 16:30:28 +00006426 alignment of the destination location.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006427
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00006428<p>If the call to this intrinsic has an alignment value that is not 0 or 1,
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006429 then the caller guarantees that the destination pointer is aligned to that
6430 boundary.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006431
Jeffrey Yasskin93e066d2010-04-26 21:21:24 +00006432<p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6433 <tt>llvm.memset</tt> call is a <a href="#volatile">volatile operation</a>.
6434 The detailed access behavior is not very cleanly specified and it is unwise
6435 to depend on it.</p>
Chris Lattner9f636de2010-04-08 00:53:57 +00006436
Chris Lattner10610642004-02-14 04:08:35 +00006437<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006438<p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
6439 at the destination location. If the argument is known to be aligned to some
6440 boundary, this can be specified as the fourth argument, otherwise it should
6441 be set to 0 or 1.</p>
Chris Lattner10610642004-02-14 04:08:35 +00006442
Chris Lattner10610642004-02-14 04:08:35 +00006443</div>
6444
Chris Lattner32006282004-06-11 02:28:03 +00006445<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006446<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006447 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006448</h4>
Chris Lattnera4d74142005-07-21 01:29:16 +00006449
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006450<div>
Chris Lattnera4d74142005-07-21 01:29:16 +00006451
6452<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006453<p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
6454 floating point or vector of floating point type. Not all targets support all
6455 types however.</p>
6456
Chris Lattnera4d74142005-07-21 01:29:16 +00006457<pre>
Dale Johannesen408f9c12007-10-02 17:47:38 +00006458 declare float @llvm.sqrt.f32(float %Val)
6459 declare double @llvm.sqrt.f64(double %Val)
6460 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
6461 declare fp128 @llvm.sqrt.f128(fp128 %Val)
6462 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Chris Lattnera4d74142005-07-21 01:29:16 +00006463</pre>
6464
6465<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006466<p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
6467 returning the same value as the libm '<tt>sqrt</tt>' functions would.
6468 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
6469 behavior for negative numbers other than -0.0 (which allows for better
6470 optimization, because there is no need to worry about errno being
6471 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
Chris Lattnera4d74142005-07-21 01:29:16 +00006472
6473<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006474<p>The argument and return value are floating point numbers of the same
6475 type.</p>
Chris Lattnera4d74142005-07-21 01:29:16 +00006476
6477<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006478<p>This function returns the sqrt of the specified operand if it is a
6479 nonnegative floating point number.</p>
Chris Lattnera4d74142005-07-21 01:29:16 +00006480
Chris Lattnera4d74142005-07-21 01:29:16 +00006481</div>
6482
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006483<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006484<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006485 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006486</h4>
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006487
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006488<div>
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006489
6490<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006491<p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
6492 floating point or vector of floating point type. Not all targets support all
6493 types however.</p>
6494
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006495<pre>
Dale Johannesen408f9c12007-10-02 17:47:38 +00006496 declare float @llvm.powi.f32(float %Val, i32 %power)
6497 declare double @llvm.powi.f64(double %Val, i32 %power)
6498 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
6499 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
6500 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006501</pre>
6502
6503<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006504<p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
6505 specified (positive or negative) power. The order of evaluation of
6506 multiplications is not defined. When a vector of floating point type is
6507 used, the second argument remains a scalar integer value.</p>
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006508
6509<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006510<p>The second argument is an integer power, and the first is a value to raise to
6511 that power.</p>
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006512
6513<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006514<p>This function returns the first value raised to the second power with an
6515 unspecified sequence of rounding operations.</p>
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006516
Chris Lattnerf4d252d2006-09-08 06:34:02 +00006517</div>
6518
Dan Gohman91c284c2007-10-15 20:30:11 +00006519<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006520<h4>
Dan Gohman91c284c2007-10-15 20:30:11 +00006521 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006522</h4>
Dan Gohman91c284c2007-10-15 20:30:11 +00006523
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006524<div>
Dan Gohman91c284c2007-10-15 20:30:11 +00006525
6526<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006527<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
6528 floating point or vector of floating point type. Not all targets support all
6529 types however.</p>
6530
Dan Gohman91c284c2007-10-15 20:30:11 +00006531<pre>
6532 declare float @llvm.sin.f32(float %Val)
6533 declare double @llvm.sin.f64(double %Val)
6534 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
6535 declare fp128 @llvm.sin.f128(fp128 %Val)
6536 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
6537</pre>
6538
6539<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006540<p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006541
6542<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006543<p>The argument and return value are floating point numbers of the same
6544 type.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006545
6546<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006547<p>This function returns the sine of the specified operand, returning the same
6548 values as the libm <tt>sin</tt> functions would, and handles error conditions
6549 in the same way.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006550
Dan Gohman91c284c2007-10-15 20:30:11 +00006551</div>
6552
6553<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006554<h4>
Dan Gohman91c284c2007-10-15 20:30:11 +00006555 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006556</h4>
Dan Gohman91c284c2007-10-15 20:30:11 +00006557
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006558<div>
Dan Gohman91c284c2007-10-15 20:30:11 +00006559
6560<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006561<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
6562 floating point or vector of floating point type. Not all targets support all
6563 types however.</p>
6564
Dan Gohman91c284c2007-10-15 20:30:11 +00006565<pre>
6566 declare float @llvm.cos.f32(float %Val)
6567 declare double @llvm.cos.f64(double %Val)
6568 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
6569 declare fp128 @llvm.cos.f128(fp128 %Val)
6570 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
6571</pre>
6572
6573<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006574<p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006575
6576<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006577<p>The argument and return value are floating point numbers of the same
6578 type.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006579
6580<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006581<p>This function returns the cosine of the specified operand, returning the same
6582 values as the libm <tt>cos</tt> functions would, and handles error conditions
6583 in the same way.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006584
Dan Gohman91c284c2007-10-15 20:30:11 +00006585</div>
6586
6587<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006588<h4>
Dan Gohman91c284c2007-10-15 20:30:11 +00006589 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006590</h4>
Dan Gohman91c284c2007-10-15 20:30:11 +00006591
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006592<div>
Dan Gohman91c284c2007-10-15 20:30:11 +00006593
6594<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006595<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
6596 floating point or vector of floating point type. Not all targets support all
6597 types however.</p>
6598
Dan Gohman91c284c2007-10-15 20:30:11 +00006599<pre>
6600 declare float @llvm.pow.f32(float %Val, float %Power)
6601 declare double @llvm.pow.f64(double %Val, double %Power)
6602 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
6603 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
6604 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
6605</pre>
6606
6607<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006608<p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
6609 specified (positive or negative) power.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006610
6611<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006612<p>The second argument is a floating point power, and the first is a value to
6613 raise to that power.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006614
6615<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006616<p>This function returns the first value raised to the second power, returning
6617 the same values as the libm <tt>pow</tt> functions would, and handles error
6618 conditions in the same way.</p>
Dan Gohman91c284c2007-10-15 20:30:11 +00006619
Dan Gohman91c284c2007-10-15 20:30:11 +00006620</div>
6621
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006622</div>
6623
Dan Gohman4e9011c2011-05-23 21:13:03 +00006624<!-- _______________________________________________________________________ -->
6625<h4>
6626 <a name="int_exp">'<tt>llvm.exp.*</tt>' Intrinsic</a>
6627</h4>
6628
6629<div>
6630
6631<h5>Syntax:</h5>
6632<p>This is an overloaded intrinsic. You can use <tt>llvm.exp</tt> on any
6633 floating point or vector of floating point type. Not all targets support all
6634 types however.</p>
6635
6636<pre>
6637 declare float @llvm.exp.f32(float %Val)
6638 declare double @llvm.exp.f64(double %Val)
6639 declare x86_fp80 @llvm.exp.f80(x86_fp80 %Val)
6640 declare fp128 @llvm.exp.f128(fp128 %Val)
6641 declare ppc_fp128 @llvm.exp.ppcf128(ppc_fp128 %Val)
6642</pre>
6643
6644<h5>Overview:</h5>
6645<p>The '<tt>llvm.exp.*</tt>' intrinsics perform the exp function.</p>
6646
6647<h5>Arguments:</h5>
6648<p>The argument and return value are floating point numbers of the same
6649 type.</p>
6650
6651<h5>Semantics:</h5>
6652<p>This function returns the same values as the libm <tt>exp</tt> functions
6653 would, and handles error conditions in the same way.</p>
6654
6655</div>
6656
6657<!-- _______________________________________________________________________ -->
6658<h4>
6659 <a name="int_log">'<tt>llvm.log.*</tt>' Intrinsic</a>
6660</h4>
6661
6662<div>
6663
6664<h5>Syntax:</h5>
6665<p>This is an overloaded intrinsic. You can use <tt>llvm.log</tt> on any
6666 floating point or vector of floating point type. Not all targets support all
6667 types however.</p>
6668
6669<pre>
6670 declare float @llvm.log.f32(float %Val)
6671 declare double @llvm.log.f64(double %Val)
6672 declare x86_fp80 @llvm.log.f80(x86_fp80 %Val)
6673 declare fp128 @llvm.log.f128(fp128 %Val)
6674 declare ppc_fp128 @llvm.log.ppcf128(ppc_fp128 %Val)
6675</pre>
6676
6677<h5>Overview:</h5>
6678<p>The '<tt>llvm.log.*</tt>' intrinsics perform the log function.</p>
6679
6680<h5>Arguments:</h5>
6681<p>The argument and return value are floating point numbers of the same
6682 type.</p>
6683
6684<h5>Semantics:</h5>
6685<p>This function returns the same values as the libm <tt>log</tt> functions
6686 would, and handles error conditions in the same way.</p>
6687
Cameron Zwarich33390842011-07-08 21:39:21 +00006688<h4>
6689 <a name="int_fma">'<tt>llvm.fma.*</tt>' Intrinsic</a>
6690</h4>
6691
6692<div>
6693
6694<h5>Syntax:</h5>
6695<p>This is an overloaded intrinsic. You can use <tt>llvm.fma</tt> on any
6696 floating point or vector of floating point type. Not all targets support all
6697 types however.</p>
6698
6699<pre>
6700 declare float @llvm.fma.f32(float %a, float %b, float %c)
6701 declare double @llvm.fma.f64(double %a, double %b, double %c)
6702 declare x86_fp80 @llvm.fma.f80(x86_fp80 %a, x86_fp80 %b, x86_fp80 %c)
6703 declare fp128 @llvm.fma.f128(fp128 %a, fp128 %b, fp128 %c)
6704 declare ppc_fp128 @llvm.fma.ppcf128(ppc_fp128 %a, ppc_fp128 %b, ppc_fp128 %c)
6705</pre>
6706
6707<h5>Overview:</h5>
Cameron Zwarichabc43e62011-07-08 22:13:55 +00006708<p>The '<tt>llvm.fma.*</tt>' intrinsics perform the fused multiply-add
Cameron Zwarich33390842011-07-08 21:39:21 +00006709 operation.</p>
6710
6711<h5>Arguments:</h5>
6712<p>The argument and return value are floating point numbers of the same
6713 type.</p>
6714
6715<h5>Semantics:</h5>
6716<p>This function returns the same values as the libm <tt>fma</tt> functions
6717 would.</p>
6718
Dan Gohman4e9011c2011-05-23 21:13:03 +00006719</div>
6720
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006721<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006722<h3>
Nate Begeman7e36c472006-01-13 23:26:38 +00006723 <a name="int_manip">Bit Manipulation Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006724</h3>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006725
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006726<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006727
6728<p>LLVM provides intrinsics for a few important bit manipulation operations.
6729 These allow efficient code generation for some algorithms.</p>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006730
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006731<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006732<h4>
Reid Spencera3e435f2007-04-04 02:42:35 +00006733 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006734</h4>
Nate Begeman7e36c472006-01-13 23:26:38 +00006735
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006736<div>
Nate Begeman7e36c472006-01-13 23:26:38 +00006737
6738<h5>Syntax:</h5>
Reid Spencer409e28f2007-04-01 08:04:23 +00006739<p>This is an overloaded intrinsic function. You can use bswap on any integer
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006740 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
6741
Nate Begeman7e36c472006-01-13 23:26:38 +00006742<pre>
Chandler Carruth69940402007-08-04 01:51:18 +00006743 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
6744 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
6745 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Nate Begeman7e36c472006-01-13 23:26:38 +00006746</pre>
6747
6748<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006749<p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
6750 values with an even number of bytes (positive multiple of 16 bits). These
6751 are useful for performing operations on data that is not in the target's
6752 native byte order.</p>
Nate Begeman7e36c472006-01-13 23:26:38 +00006753
6754<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006755<p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
6756 and low byte of the input i16 swapped. Similarly,
6757 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
6758 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
6759 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
6760 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
6761 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
6762 more, respectively).</p>
Nate Begeman7e36c472006-01-13 23:26:38 +00006763
6764</div>
6765
6766<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006767<h4>
Reid Spencer0b118202006-01-16 21:12:35 +00006768 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006769</h4>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006770
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006771<div>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006772
6773<h5>Syntax:</h5>
Reid Spencer409e28f2007-04-01 08:04:23 +00006774<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
Owen Andersonf1ac4652011-07-01 21:52:38 +00006775 width, or on any vector with integer elements. Not all targets support all
6776 bit widths or vector types, however.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006777
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006778<pre>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006779 declare i8 @llvm.ctpop.i8(i8 &lt;src&gt;)
Chandler Carruth69940402007-08-04 01:51:18 +00006780 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Anton Korobeynikovec43a062007-03-22 00:02:17 +00006781 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carruth69940402007-08-04 01:51:18 +00006782 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
6783 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Owen Andersonf1ac4652011-07-01 21:52:38 +00006784 declare &lt;2 x i32&gt; @llvm.ctpop.v2i32(&lt;2 x i32&gt; &lt;src&gt;)
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006785</pre>
6786
6787<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006788<p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
6789 in a value.</p>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006790
6791<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006792<p>The only argument is the value to be counted. The argument may be of any
Owen Andersonf1ac4652011-07-01 21:52:38 +00006793 integer type, or a vector with integer elements.
6794 The return type must match the argument type.</p>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006795
6796<h5>Semantics:</h5>
Owen Andersonf1ac4652011-07-01 21:52:38 +00006797<p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable, or within each
6798 element of a vector.</p>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006799
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006800</div>
6801
6802<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006803<h4>
Chris Lattner8a886be2006-01-16 22:34:14 +00006804 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006805</h4>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006806
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006807<div>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006808
6809<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006810<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
Owen Andersonf1ac4652011-07-01 21:52:38 +00006811 integer bit width, or any vector whose elements are integers. Not all
6812 targets support all bit widths or vector types, however.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006813
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006814<pre>
Chandler Carruth69940402007-08-04 01:51:18 +00006815 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
6816 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Anton Korobeynikovec43a062007-03-22 00:02:17 +00006817 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carruth69940402007-08-04 01:51:18 +00006818 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
6819 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Owen Andersonf1ac4652011-07-01 21:52:38 +00006820 declare &lt;2 x i32&gt; @llvm.ctlz.v2i32(&lt;2 x i32&gt; &lt;src;gt)
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006821</pre>
6822
6823<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006824<p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
6825 leading zeros in a variable.</p>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006826
6827<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006828<p>The only argument is the value to be counted. The argument may be of any
Owen Andersonf1ac4652011-07-01 21:52:38 +00006829 integer type, or any vector type with integer element type.
6830 The return type must match the argument type.</p>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006831
6832<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006833<p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
Owen Andersonf1ac4652011-07-01 21:52:38 +00006834 zeros in a variable, or within each element of the vector if the operation
6835 is of vector type. If the src == 0 then the result is the size in bits of
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006836 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006837
Andrew Lenharthec370fd2005-05-03 18:01:48 +00006838</div>
Chris Lattner32006282004-06-11 02:28:03 +00006839
Chris Lattnereff29ab2005-05-15 19:39:26 +00006840<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006841<h4>
Chris Lattner8a886be2006-01-16 22:34:14 +00006842 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006843</h4>
Chris Lattnereff29ab2005-05-15 19:39:26 +00006844
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006845<div>
Chris Lattnereff29ab2005-05-15 19:39:26 +00006846
6847<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006848<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
Owen Andersonf1ac4652011-07-01 21:52:38 +00006849 integer bit width, or any vector of integer elements. Not all targets
6850 support all bit widths or vector types, however.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006851
Chris Lattnereff29ab2005-05-15 19:39:26 +00006852<pre>
Chandler Carruth69940402007-08-04 01:51:18 +00006853 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
6854 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Anton Korobeynikovec43a062007-03-22 00:02:17 +00006855 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carruth69940402007-08-04 01:51:18 +00006856 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
6857 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Owen Andersonf1ac4652011-07-01 21:52:38 +00006858 declase &lt;2 x i32&gt; @llvm.cttz.v2i32(&lt;2 x i32&gt; &lt;src&gt;)
Chris Lattnereff29ab2005-05-15 19:39:26 +00006859</pre>
6860
6861<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006862<p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
6863 trailing zeros.</p>
Chris Lattnereff29ab2005-05-15 19:39:26 +00006864
6865<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006866<p>The only argument is the value to be counted. The argument may be of any
Owen Andersonf1ac4652011-07-01 21:52:38 +00006867 integer type, or a vectory with integer element type.. The return type
6868 must match the argument type.</p>
Chris Lattnereff29ab2005-05-15 19:39:26 +00006869
6870<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006871<p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
Owen Andersonf1ac4652011-07-01 21:52:38 +00006872 zeros in a variable, or within each element of a vector.
6873 If the src == 0 then the result is the size in bits of
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006874 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
Chris Lattnereff29ab2005-05-15 19:39:26 +00006875
Chris Lattnereff29ab2005-05-15 19:39:26 +00006876</div>
6877
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006878</div>
6879
Bill Wendlingda01af72009-02-08 04:04:40 +00006880<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006881<h3>
Bill Wendlingda01af72009-02-08 04:04:40 +00006882 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006883</h3>
Bill Wendlingda01af72009-02-08 04:04:40 +00006884
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006885<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006886
6887<p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
Bill Wendlingda01af72009-02-08 04:04:40 +00006888
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006889<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006890<h4>
6891 <a name="int_sadd_overflow">
6892 '<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics
6893 </a>
6894</h4>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006895
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006896<div>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006897
6898<h5>Syntax:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006899<p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006900 on any integer bit width.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006901
6902<pre>
6903 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
6904 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6905 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6906</pre>
6907
6908<h5>Overview:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006909<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006910 a signed addition of the two arguments, and indicate whether an overflow
6911 occurred during the signed summation.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006912
6913<h5>Arguments:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006914<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006915 be of integer types of any bit width, but they must have the same bit
6916 width. The second element of the result structure must be of
6917 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6918 undergo signed addition.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006919
6920<h5>Semantics:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006921<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006922 a signed addition of the two variables. They return a structure &mdash; the
6923 first element of which is the signed summation, and the second element of
6924 which is a bit specifying if the signed summation resulted in an
6925 overflow.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006926
6927<h5>Examples:</h5>
6928<pre>
6929 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6930 %sum = extractvalue {i32, i1} %res, 0
6931 %obit = extractvalue {i32, i1} %res, 1
6932 br i1 %obit, label %overflow, label %normal
6933</pre>
6934
6935</div>
6936
6937<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006938<h4>
6939 <a name="int_uadd_overflow">
6940 '<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics
6941 </a>
6942</h4>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006943
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006944<div>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006945
6946<h5>Syntax:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006947<p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006948 on any integer bit width.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006949
6950<pre>
6951 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6952 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6953 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6954</pre>
6955
6956<h5>Overview:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006957<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006958 an unsigned addition of the two arguments, and indicate whether a carry
6959 occurred during the unsigned summation.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006960
6961<h5>Arguments:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006962<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006963 be of integer types of any bit width, but they must have the same bit
6964 width. The second element of the result structure must be of
6965 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6966 undergo unsigned addition.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006967
6968<h5>Semantics:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006969<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006970 an unsigned addition of the two arguments. They return a structure &mdash;
6971 the first element of which is the sum, and the second element of which is a
6972 bit specifying if the unsigned summation resulted in a carry.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006973
6974<h5>Examples:</h5>
6975<pre>
6976 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6977 %sum = extractvalue {i32, i1} %res, 0
6978 %obit = extractvalue {i32, i1} %res, 1
6979 br i1 %obit, label %carry, label %normal
6980</pre>
6981
6982</div>
6983
6984<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00006985<h4>
6986 <a name="int_ssub_overflow">
6987 '<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics
6988 </a>
6989</h4>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006990
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00006991<div>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006992
6993<h5>Syntax:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006994<p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00006995 on any integer bit width.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00006996
6997<pre>
6998 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6999 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
7000 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
7001</pre>
7002
7003<h5>Overview:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007004<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007005 a signed subtraction of the two arguments, and indicate whether an overflow
7006 occurred during the signed subtraction.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007007
7008<h5>Arguments:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007009<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007010 be of integer types of any bit width, but they must have the same bit
7011 width. The second element of the result structure must be of
7012 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7013 undergo signed subtraction.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007014
7015<h5>Semantics:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007016<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007017 a signed subtraction of the two arguments. They return a structure &mdash;
7018 the first element of which is the subtraction, and the second element of
7019 which is a bit specifying if the signed subtraction resulted in an
7020 overflow.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007021
7022<h5>Examples:</h5>
7023<pre>
7024 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
7025 %sum = extractvalue {i32, i1} %res, 0
7026 %obit = extractvalue {i32, i1} %res, 1
7027 br i1 %obit, label %overflow, label %normal
7028</pre>
7029
7030</div>
7031
7032<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007033<h4>
7034 <a name="int_usub_overflow">
7035 '<tt>llvm.usub.with.overflow.*</tt>' Intrinsics
7036 </a>
7037</h4>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007038
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007039<div>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007040
7041<h5>Syntax:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007042<p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007043 on any integer bit width.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007044
7045<pre>
7046 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
7047 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
7048 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
7049</pre>
7050
7051<h5>Overview:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007052<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007053 an unsigned subtraction of the two arguments, and indicate whether an
7054 overflow occurred during the unsigned subtraction.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007055
7056<h5>Arguments:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007057<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007058 be of integer types of any bit width, but they must have the same bit
7059 width. The second element of the result structure must be of
7060 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7061 undergo unsigned subtraction.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007062
7063<h5>Semantics:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007064<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007065 an unsigned subtraction of the two arguments. They return a structure &mdash;
7066 the first element of which is the subtraction, and the second element of
7067 which is a bit specifying if the unsigned subtraction resulted in an
7068 overflow.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007069
7070<h5>Examples:</h5>
7071<pre>
7072 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
7073 %sum = extractvalue {i32, i1} %res, 0
7074 %obit = extractvalue {i32, i1} %res, 1
7075 br i1 %obit, label %overflow, label %normal
7076</pre>
7077
7078</div>
7079
7080<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007081<h4>
7082 <a name="int_smul_overflow">
7083 '<tt>llvm.smul.with.overflow.*</tt>' Intrinsics
7084 </a>
7085</h4>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007086
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007087<div>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007088
7089<h5>Syntax:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007090<p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007091 on any integer bit width.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007092
7093<pre>
7094 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
7095 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
7096 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
7097</pre>
7098
7099<h5>Overview:</h5>
7100
7101<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007102 a signed multiplication of the two arguments, and indicate whether an
7103 overflow occurred during the signed multiplication.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007104
7105<h5>Arguments:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007106<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007107 be of integer types of any bit width, but they must have the same bit
7108 width. The second element of the result structure must be of
7109 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7110 undergo signed multiplication.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007111
7112<h5>Semantics:</h5>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007113<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007114 a signed multiplication of the two arguments. They return a structure &mdash;
7115 the first element of which is the multiplication, and the second element of
7116 which is a bit specifying if the signed multiplication resulted in an
7117 overflow.</p>
Bill Wendlingac1df8e2009-02-08 01:40:31 +00007118
7119<h5>Examples:</h5>
7120<pre>
7121 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
7122 %sum = extractvalue {i32, i1} %res, 0
7123 %obit = extractvalue {i32, i1} %res, 1
7124 br i1 %obit, label %overflow, label %normal
7125</pre>
7126
Reid Spencerf86037f2007-04-11 23:23:49 +00007127</div>
7128
Bill Wendling41b485c2009-02-08 23:00:09 +00007129<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007130<h4>
7131 <a name="int_umul_overflow">
7132 '<tt>llvm.umul.with.overflow.*</tt>' Intrinsics
7133 </a>
7134</h4>
Bill Wendling41b485c2009-02-08 23:00:09 +00007135
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007136<div>
Bill Wendling41b485c2009-02-08 23:00:09 +00007137
7138<h5>Syntax:</h5>
Bill Wendling41b485c2009-02-08 23:00:09 +00007139<p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007140 on any integer bit width.</p>
Bill Wendling41b485c2009-02-08 23:00:09 +00007141
7142<pre>
7143 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
7144 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
7145 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
7146</pre>
7147
7148<h5>Overview:</h5>
Bill Wendling41b485c2009-02-08 23:00:09 +00007149<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007150 a unsigned multiplication of the two arguments, and indicate whether an
7151 overflow occurred during the unsigned multiplication.</p>
Bill Wendling41b485c2009-02-08 23:00:09 +00007152
7153<h5>Arguments:</h5>
Bill Wendling41b485c2009-02-08 23:00:09 +00007154<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007155 be of integer types of any bit width, but they must have the same bit
7156 width. The second element of the result structure must be of
7157 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7158 undergo unsigned multiplication.</p>
Bill Wendling41b485c2009-02-08 23:00:09 +00007159
7160<h5>Semantics:</h5>
Bill Wendling41b485c2009-02-08 23:00:09 +00007161<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007162 an unsigned multiplication of the two arguments. They return a structure
7163 &mdash; the first element of which is the multiplication, and the second
7164 element of which is a bit specifying if the unsigned multiplication resulted
7165 in an overflow.</p>
Bill Wendling41b485c2009-02-08 23:00:09 +00007166
7167<h5>Examples:</h5>
7168<pre>
7169 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
7170 %sum = extractvalue {i32, i1} %res, 0
7171 %obit = extractvalue {i32, i1} %res, 1
7172 br i1 %obit, label %overflow, label %normal
7173</pre>
7174
7175</div>
7176
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007177</div>
7178
Chris Lattner8ff75902004-01-06 05:31:32 +00007179<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007180<h3>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007181 <a name="int_fp16">Half Precision Floating Point Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007182</h3>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007183
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007184<div>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007185
Chris Lattner0cec9c82010-03-15 04:12:21 +00007186<p>Half precision floating point is a storage-only format. This means that it is
7187 a dense encoding (in memory) but does not support computation in the
7188 format.</p>
Chris Lattner82c3dc62010-03-14 23:03:31 +00007189
Chris Lattner0cec9c82010-03-15 04:12:21 +00007190<p>This means that code must first load the half-precision floating point
Chris Lattner82c3dc62010-03-14 23:03:31 +00007191 value as an i16, then convert it to float with <a
7192 href="#int_convert_from_fp16"><tt>llvm.convert.from.fp16</tt></a>.
7193 Computation can then be performed on the float value (including extending to
Chris Lattner0cec9c82010-03-15 04:12:21 +00007194 double etc). To store the value back to memory, it is first converted to
7195 float if needed, then converted to i16 with
Chris Lattner82c3dc62010-03-14 23:03:31 +00007196 <a href="#int_convert_to_fp16"><tt>llvm.convert.to.fp16</tt></a>, then
7197 storing as an i16 value.</p>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007198
7199<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007200<h4>
7201 <a name="int_convert_to_fp16">
7202 '<tt>llvm.convert.to.fp16</tt>' Intrinsic
7203 </a>
7204</h4>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007205
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007206<div>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007207
7208<h5>Syntax:</h5>
7209<pre>
7210 declare i16 @llvm.convert.to.fp16(f32 %a)
7211</pre>
7212
7213<h5>Overview:</h5>
7214<p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
7215 a conversion from single precision floating point format to half precision
7216 floating point format.</p>
7217
7218<h5>Arguments:</h5>
7219<p>The intrinsic function contains single argument - the value to be
7220 converted.</p>
7221
7222<h5>Semantics:</h5>
7223<p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
7224 a conversion from single precision floating point format to half precision
Chris Lattner0cec9c82010-03-15 04:12:21 +00007225 floating point format. The return value is an <tt>i16</tt> which
Chris Lattner82c3dc62010-03-14 23:03:31 +00007226 contains the converted number.</p>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007227
7228<h5>Examples:</h5>
7229<pre>
7230 %res = call i16 @llvm.convert.to.fp16(f32 %a)
7231 store i16 %res, i16* @x, align 2
7232</pre>
7233
7234</div>
7235
7236<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007237<h4>
7238 <a name="int_convert_from_fp16">
7239 '<tt>llvm.convert.from.fp16</tt>' Intrinsic
7240 </a>
7241</h4>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007242
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007243<div>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007244
7245<h5>Syntax:</h5>
7246<pre>
7247 declare f32 @llvm.convert.from.fp16(i16 %a)
7248</pre>
7249
7250<h5>Overview:</h5>
7251<p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs
7252 a conversion from half precision floating point format to single precision
7253 floating point format.</p>
7254
7255<h5>Arguments:</h5>
7256<p>The intrinsic function contains single argument - the value to be
7257 converted.</p>
7258
7259<h5>Semantics:</h5>
7260<p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs a
Chris Lattner0cec9c82010-03-15 04:12:21 +00007261 conversion from half single precision floating point format to single
Chris Lattner82c3dc62010-03-14 23:03:31 +00007262 precision floating point format. The input half-float value is represented by
7263 an <tt>i16</tt> value.</p>
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007264
7265<h5>Examples:</h5>
7266<pre>
7267 %a = load i16* @x, align 2
7268 %res = call f32 @llvm.convert.from.fp16(i16 %a)
7269</pre>
7270
7271</div>
7272
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007273</div>
7274
Anton Korobeynikovf02e7302010-03-14 18:42:47 +00007275<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007276<h3>
Chris Lattner8ff75902004-01-06 05:31:32 +00007277 <a name="int_debugger">Debugger Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007278</h3>
Chris Lattner8ff75902004-01-06 05:31:32 +00007279
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007280<div>
Chris Lattner8ff75902004-01-06 05:31:32 +00007281
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007282<p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
7283 prefix), are described in
7284 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
7285 Level Debugging</a> document.</p>
7286
7287</div>
Chris Lattner8ff75902004-01-06 05:31:32 +00007288
Jim Laskeydd4ef1b2007-03-14 19:31:19 +00007289<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007290<h3>
Jim Laskeydd4ef1b2007-03-14 19:31:19 +00007291 <a name="int_eh">Exception Handling Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007292</h3>
Jim Laskeydd4ef1b2007-03-14 19:31:19 +00007293
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007294<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007295
7296<p>The LLVM exception handling intrinsics (which all start with
7297 <tt>llvm.eh.</tt> prefix), are described in
7298 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
7299 Handling</a> document.</p>
7300
Jim Laskeydd4ef1b2007-03-14 19:31:19 +00007301</div>
7302
Tanya Lattner6d806e92007-06-15 20:50:54 +00007303<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007304<h3>
Duncan Sandsf7331b32007-09-11 14:10:23 +00007305 <a name="int_trampoline">Trampoline Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007306</h3>
Duncan Sands36397f52007-07-27 12:58:54 +00007307
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007308<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007309
7310<p>This intrinsic makes it possible to excise one parameter, marked with
Dan Gohmanff235352010-07-02 23:18:08 +00007311 the <a href="#nest"><tt>nest</tt></a> attribute, from a function.
7312 The result is a callable
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007313 function pointer lacking the nest parameter - the caller does not need to
7314 provide a value for it. Instead, the value to use is stored in advance in a
7315 "trampoline", a block of memory usually allocated on the stack, which also
7316 contains code to splice the nest value into the argument list. This is used
7317 to implement the GCC nested function address extension.</p>
7318
7319<p>For example, if the function is
7320 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
7321 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
7322 follows:</p>
7323
Benjamin Kramer26fe25f2010-07-13 12:26:09 +00007324<pre class="doc_code">
Duncan Sandsf7331b32007-09-11 14:10:23 +00007325 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
7326 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007327 %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 +00007328 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands36397f52007-07-27 12:58:54 +00007329</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007330
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007331<p>The call <tt>%val = call i32 %fp(i32 %x, i32 %y)</tt> is then equivalent
7332 to <tt>%val = call i32 %f(i8* %nval, i32 %x, i32 %y)</tt>.</p>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007333
Duncan Sands36397f52007-07-27 12:58:54 +00007334<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007335<h4>
7336 <a name="int_it">
7337 '<tt>llvm.init.trampoline</tt>' Intrinsic
7338 </a>
7339</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007340
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007341<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007342
Duncan Sands36397f52007-07-27 12:58:54 +00007343<h5>Syntax:</h5>
7344<pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007345 declare i8* @llvm.init.trampoline(i8* &lt;tramp&gt;, i8* &lt;func&gt;, i8* &lt;nval&gt;)
Duncan Sands36397f52007-07-27 12:58:54 +00007346</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007347
Duncan Sands36397f52007-07-27 12:58:54 +00007348<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007349<p>This fills the memory pointed to by <tt>tramp</tt> with code and returns a
7350 function pointer suitable for executing it.</p>
7351
Duncan Sands36397f52007-07-27 12:58:54 +00007352<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007353<p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
7354 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
7355 sufficiently aligned block of memory; this memory is written to by the
7356 intrinsic. Note that the size and the alignment are target-specific - LLVM
7357 currently provides no portable way of determining them, so a front-end that
7358 generates this intrinsic needs to have some target-specific knowledge.
7359 The <tt>func</tt> argument must hold a function bitcast to
7360 an <tt>i8*</tt>.</p>
7361
Duncan Sands36397f52007-07-27 12:58:54 +00007362<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007363<p>The block of memory pointed to by <tt>tramp</tt> is filled with target
7364 dependent code, turning it into a function. A pointer to this function is
7365 returned, but needs to be bitcast to an <a href="#int_trampoline">appropriate
7366 function pointer type</a> before being called. The new function's signature
7367 is the same as that of <tt>func</tt> with any arguments marked with
7368 the <tt>nest</tt> attribute removed. At most one such <tt>nest</tt> argument
7369 is allowed, and it must be of pointer type. Calling the new function is
7370 equivalent to calling <tt>func</tt> with the same argument list, but
7371 with <tt>nval</tt> used for the missing <tt>nest</tt> argument. If, after
7372 calling <tt>llvm.init.trampoline</tt>, the memory pointed to
7373 by <tt>tramp</tt> is modified, then the effect of any later call to the
7374 returned function pointer is undefined.</p>
7375
Duncan Sands36397f52007-07-27 12:58:54 +00007376</div>
7377
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007378</div>
7379
Duncan Sands36397f52007-07-27 12:58:54 +00007380<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007381<h3>
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007382 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007383</h3>
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007384
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007385<div>
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007386
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007387<p>These intrinsic functions expand the "universal IR" of LLVM to represent
7388 hardware constructs for atomic operations and memory synchronization. This
7389 provides an interface to the hardware, not an interface to the programmer. It
7390 is aimed at a low enough level to allow any programming models or APIs
7391 (Application Programming Interfaces) which need atomic behaviors to map
7392 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
7393 hardware provides a "universal IR" for source languages, it also provides a
7394 starting point for developing a "universal" atomic operation and
7395 synchronization IR.</p>
7396
7397<p>These do <em>not</em> form an API such as high-level threading libraries,
7398 software transaction memory systems, atomic primitives, and intrinsic
7399 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
7400 application libraries. The hardware interface provided by LLVM should allow
7401 a clean implementation of all of these APIs and parallel programming models.
7402 No one model or paradigm should be selected above others unless the hardware
7403 itself ubiquitously does so.</p>
7404
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007405<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007406<h4>
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007407 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007408</h4>
7409
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007410<div>
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007411<h5>Syntax:</h5>
7412<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007413 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 +00007414</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007415
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007416<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007417<p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
7418 specific pairs of memory access types.</p>
7419
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007420<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007421<p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
7422 The first four arguments enables a specific barrier as listed below. The
Dan Gohmanb55a1ee2010-03-01 17:41:39 +00007423 fifth argument specifies that the barrier applies to io or device or uncached
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007424 memory.</p>
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007425
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007426<ul>
7427 <li><tt>ll</tt>: load-load barrier</li>
7428 <li><tt>ls</tt>: load-store barrier</li>
7429 <li><tt>sl</tt>: store-load barrier</li>
7430 <li><tt>ss</tt>: store-store barrier</li>
7431 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
7432</ul>
7433
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007434<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007435<p>This intrinsic causes the system to enforce some ordering constraints upon
7436 the loads and stores of the program. This barrier does not
7437 indicate <em>when</em> any events will occur, it only enforces
7438 an <em>order</em> in which they occur. For any of the specified pairs of load
7439 and store operations (f.ex. load-load, or store-load), all of the first
7440 operations preceding the barrier will complete before any of the second
7441 operations succeeding the barrier begin. Specifically the semantics for each
7442 pairing is as follows:</p>
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007443
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007444<ul>
7445 <li><tt>ll</tt>: All loads before the barrier must complete before any load
7446 after the barrier begins.</li>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00007447 <li><tt>ls</tt>: All loads before the barrier must complete before any
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007448 store after the barrier begins.</li>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00007449 <li><tt>ss</tt>: All stores before the barrier must complete before any
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007450 store after the barrier begins.</li>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00007451 <li><tt>sl</tt>: All stores before the barrier must complete before any
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007452 load after the barrier begins.</li>
7453</ul>
7454
7455<p>These semantics are applied with a logical "and" behavior when more than one
7456 is enabled in a single memory barrier intrinsic.</p>
7457
7458<p>Backends may implement stronger barriers than those requested when they do
7459 not support as fine grained a barrier as requested. Some architectures do
7460 not need all types of barriers and on such architectures, these become
7461 noops.</p>
7462
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007463<h5>Example:</h5>
7464<pre>
Victor Hernandez2fee2942009-10-26 23:44:29 +00007465%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7466%ptr = bitcast i8* %mallocP to i32*
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007467 store i32 4, %ptr
7468
7469%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
Evan Cheng0b0669a2011-06-29 17:14:00 +00007470 call void @llvm.memory.barrier(i1 false, i1 true, i1 false, i1 false, i1 true)
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007471 <i>; guarantee the above finishes</i>
7472 store i32 8, %ptr <i>; before this begins</i>
7473</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007474
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007475</div>
7476
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007477<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007478<h4>
Mon P Wang28873102008-06-25 08:15:39 +00007479 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007480</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007481
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007482<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007483
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007484<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007485<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
7486 any integer bit width and for different address spaces. Not all targets
7487 support all bit widths however.</p>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007488
7489<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007490 declare i8 @llvm.atomic.cmp.swap.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt;)
7491 declare i16 @llvm.atomic.cmp.swap.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt;)
7492 declare i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt;)
7493 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 +00007494</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007495
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007496<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007497<p>This loads a value in memory and compares it to a given value. If they are
7498 equal, it stores a new value into the memory.</p>
7499
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007500<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007501<p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
7502 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
7503 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
7504 this integer type. While any bit width integer may be used, targets may only
7505 lower representations they support in hardware.</p>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007506
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007507<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007508<p>This entire intrinsic must be executed atomically. It first loads the value
7509 in memory pointed to by <tt>ptr</tt> and compares it with the
7510 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
7511 memory. The loaded value is yielded in all cases. This provides the
7512 equivalent of an atomic compare-and-swap operation within the SSA
7513 framework.</p>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007514
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007515<h5>Examples:</h5>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007516<pre>
Victor Hernandez2fee2942009-10-26 23:44:29 +00007517%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7518%ptr = bitcast i8* %mallocP to i32*
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007519 store i32 4, %ptr
7520
7521%val1 = add i32 4, 4
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007522%result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 4, %val1)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007523 <i>; yields {i32}:result1 = 4</i>
7524%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7525%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7526
7527%val2 = add i32 1, 1
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007528%result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 5, %val2)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007529 <i>; yields {i32}:result2 = 8</i>
7530%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
7531
7532%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
7533</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007534
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007535</div>
7536
7537<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007538<h4>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007539 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007540</h4>
7541
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007542<div>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007543<h5>Syntax:</h5>
7544
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007545<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
7546 integer bit width. Not all targets support all bit widths however.</p>
7547
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007548<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007549 declare i8 @llvm.atomic.swap.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;val&gt;)
7550 declare i16 @llvm.atomic.swap.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;val&gt;)
7551 declare i32 @llvm.atomic.swap.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;val&gt;)
7552 declare i64 @llvm.atomic.swap.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;val&gt;)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007553</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007554
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007555<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007556<p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
7557 the value from memory. It then stores the value in <tt>val</tt> in the memory
7558 at <tt>ptr</tt>.</p>
7559
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007560<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007561<p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
7562 the <tt>val</tt> argument and the result must be integers of the same bit
7563 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
7564 integer type. The targets may only lower integer representations they
7565 support.</p>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007566
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007567<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007568<p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
7569 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
7570 equivalent of an atomic swap operation within the SSA framework.</p>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007571
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007572<h5>Examples:</h5>
7573<pre>
Victor Hernandez2fee2942009-10-26 23:44:29 +00007574%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7575%ptr = bitcast i8* %mallocP to i32*
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007576 store i32 4, %ptr
7577
7578%val1 = add i32 4, 4
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007579%result1 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val1)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007580 <i>; yields {i32}:result1 = 4</i>
7581%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7582%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7583
7584%val2 = add i32 1, 1
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007585%result2 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val2)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007586 <i>; yields {i32}:result2 = 8</i>
7587
7588%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
7589%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
7590</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007591
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007592</div>
7593
7594<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007595<h4>
Mon P Wang28873102008-06-25 08:15:39 +00007596 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007597</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007598
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007599<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007600
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007601<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007602<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
7603 any integer bit width. Not all targets support all bit widths however.</p>
7604
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007605<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007606 declare i8 @llvm.atomic.load.add.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7607 declare i16 @llvm.atomic.load.add.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7608 declare i32 @llvm.atomic.load.add.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7609 declare i64 @llvm.atomic.load.add.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007610</pre>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007611
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007612<h5>Overview:</h5>
7613<p>This intrinsic adds <tt>delta</tt> to the value stored in memory
7614 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7615
7616<h5>Arguments:</h5>
7617<p>The intrinsic takes two arguments, the first a pointer to an integer value
7618 and the second an integer value. The result is also an integer value. These
7619 integer types can have any bit width, but they must all have the same bit
7620 width. The targets may only lower integer representations they support.</p>
7621
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007622<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007623<p>This intrinsic does a series of operations atomically. It first loads the
7624 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
7625 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007626
7627<h5>Examples:</h5>
7628<pre>
Victor Hernandez2fee2942009-10-26 23:44:29 +00007629%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7630%ptr = bitcast i8* %mallocP to i32*
7631 store i32 4, %ptr
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007632%result1 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 4)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007633 <i>; yields {i32}:result1 = 4</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007634%result2 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 2)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007635 <i>; yields {i32}:result2 = 8</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007636%result3 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 5)
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007637 <i>; yields {i32}:result3 = 10</i>
Mon P Wang28873102008-06-25 08:15:39 +00007638%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007639</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007640
Andrew Lenharthab0b9492008-02-21 06:45:13 +00007641</div>
7642
Mon P Wang28873102008-06-25 08:15:39 +00007643<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007644<h4>
Mon P Wang28873102008-06-25 08:15:39 +00007645 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007646</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007647
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007648<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007649
Mon P Wang28873102008-06-25 08:15:39 +00007650<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007651<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
7652 any integer bit width and for different address spaces. Not all targets
7653 support all bit widths however.</p>
7654
Mon P Wang28873102008-06-25 08:15:39 +00007655<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007656 declare i8 @llvm.atomic.load.sub.i8.p0i32(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7657 declare i16 @llvm.atomic.load.sub.i16.p0i32(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7658 declare i32 @llvm.atomic.load.sub.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7659 declare i64 @llvm.atomic.load.sub.i64.p0i32(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007660</pre>
Mon P Wang28873102008-06-25 08:15:39 +00007661
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007662<h5>Overview:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00007663<p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007664 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7665
7666<h5>Arguments:</h5>
7667<p>The intrinsic takes two arguments, the first a pointer to an integer value
7668 and the second an integer value. The result is also an integer value. These
7669 integer types can have any bit width, but they must all have the same bit
7670 width. The targets may only lower integer representations they support.</p>
7671
Mon P Wang28873102008-06-25 08:15:39 +00007672<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007673<p>This intrinsic does a series of operations atomically. It first loads the
7674 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
7675 result to <tt>ptr</tt>. It yields the original value stored
7676 at <tt>ptr</tt>.</p>
Mon P Wang28873102008-06-25 08:15:39 +00007677
7678<h5>Examples:</h5>
7679<pre>
Victor Hernandez2fee2942009-10-26 23:44:29 +00007680%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7681%ptr = bitcast i8* %mallocP to i32*
7682 store i32 8, %ptr
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007683%result1 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 4)
Mon P Wang28873102008-06-25 08:15:39 +00007684 <i>; yields {i32}:result1 = 8</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007685%result2 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 2)
Mon P Wang28873102008-06-25 08:15:39 +00007686 <i>; yields {i32}:result2 = 4</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007687%result3 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 5)
Mon P Wang28873102008-06-25 08:15:39 +00007688 <i>; yields {i32}:result3 = 2</i>
7689%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
7690</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007691
Mon P Wang28873102008-06-25 08:15:39 +00007692</div>
7693
7694<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007695<h4>
7696 <a name="int_atomic_load_and">
7697 '<tt>llvm.atomic.load.and.*</tt>' Intrinsic
7698 </a>
7699 <br>
7700 <a name="int_atomic_load_nand">
7701 '<tt>llvm.atomic.load.nand.*</tt>' Intrinsic
7702 </a>
7703 <br>
7704 <a name="int_atomic_load_or">
7705 '<tt>llvm.atomic.load.or.*</tt>' Intrinsic
7706 </a>
7707 <br>
7708 <a name="int_atomic_load_xor">
7709 '<tt>llvm.atomic.load.xor.*</tt>' Intrinsic
7710 </a>
7711</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007712
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007713<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007714
Mon P Wang28873102008-06-25 08:15:39 +00007715<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007716<p>These are overloaded intrinsics. You can
7717 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
7718 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
7719 bit width and for different address spaces. Not all targets support all bit
7720 widths however.</p>
Mon P Wang28873102008-06-25 08:15:39 +00007721
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007722<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007723 declare i8 @llvm.atomic.load.and.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7724 declare i16 @llvm.atomic.load.and.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7725 declare i32 @llvm.atomic.load.and.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7726 declare i64 @llvm.atomic.load.and.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007727</pre>
7728
7729<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007730 declare i8 @llvm.atomic.load.or.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7731 declare i16 @llvm.atomic.load.or.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7732 declare i32 @llvm.atomic.load.or.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7733 declare i64 @llvm.atomic.load.or.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007734</pre>
7735
7736<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007737 declare i8 @llvm.atomic.load.nand.i8.p0i32(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7738 declare i16 @llvm.atomic.load.nand.i16.p0i32(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7739 declare i32 @llvm.atomic.load.nand.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7740 declare i64 @llvm.atomic.load.nand.i64.p0i32(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007741</pre>
7742
7743<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007744 declare i8 @llvm.atomic.load.xor.i8.p0i32(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7745 declare i16 @llvm.atomic.load.xor.i16.p0i32(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7746 declare i32 @llvm.atomic.load.xor.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7747 declare i64 @llvm.atomic.load.xor.i64.p0i32(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007748</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007749
Mon P Wang28873102008-06-25 08:15:39 +00007750<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007751<p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
7752 the value stored in memory at <tt>ptr</tt>. It yields the original value
7753 at <tt>ptr</tt>.</p>
Mon P Wang28873102008-06-25 08:15:39 +00007754
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007755<h5>Arguments:</h5>
7756<p>These intrinsics take two arguments, the first a pointer to an integer value
7757 and the second an integer value. The result is also an integer value. These
7758 integer types can have any bit width, but they must all have the same bit
7759 width. The targets may only lower integer representations they support.</p>
7760
Mon P Wang28873102008-06-25 08:15:39 +00007761<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007762<p>These intrinsics does a series of operations atomically. They first load the
7763 value stored at <tt>ptr</tt>. They then do the bitwise
7764 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
7765 original value stored at <tt>ptr</tt>.</p>
Mon P Wang28873102008-06-25 08:15:39 +00007766
7767<h5>Examples:</h5>
7768<pre>
Victor Hernandez2fee2942009-10-26 23:44:29 +00007769%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7770%ptr = bitcast i8* %mallocP to i32*
7771 store i32 0x0F0F, %ptr
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007772%result0 = call i32 @llvm.atomic.load.nand.i32.p0i32(i32* %ptr, i32 0xFF)
Mon P Wang28873102008-06-25 08:15:39 +00007773 <i>; yields {i32}:result0 = 0x0F0F</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007774%result1 = call i32 @llvm.atomic.load.and.i32.p0i32(i32* %ptr, i32 0xFF)
Mon P Wang28873102008-06-25 08:15:39 +00007775 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007776%result2 = call i32 @llvm.atomic.load.or.i32.p0i32(i32* %ptr, i32 0F)
Mon P Wang28873102008-06-25 08:15:39 +00007777 <i>; yields {i32}:result2 = 0xF0</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007778%result3 = call i32 @llvm.atomic.load.xor.i32.p0i32(i32* %ptr, i32 0F)
Mon P Wang28873102008-06-25 08:15:39 +00007779 <i>; yields {i32}:result3 = FF</i>
7780%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
7781</pre>
Mon P Wang28873102008-06-25 08:15:39 +00007782
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007783</div>
Mon P Wang28873102008-06-25 08:15:39 +00007784
7785<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007786<h4>
7787 <a name="int_atomic_load_max">
7788 '<tt>llvm.atomic.load.max.*</tt>' Intrinsic
7789 </a>
7790 <br>
7791 <a name="int_atomic_load_min">
7792 '<tt>llvm.atomic.load.min.*</tt>' Intrinsic
7793 </a>
7794 <br>
7795 <a name="int_atomic_load_umax">
7796 '<tt>llvm.atomic.load.umax.*</tt>' Intrinsic
7797 </a>
7798 <br>
7799 <a name="int_atomic_load_umin">
7800 '<tt>llvm.atomic.load.umin.*</tt>' Intrinsic
7801 </a>
7802</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007803
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007804<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007805
Mon P Wang28873102008-06-25 08:15:39 +00007806<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007807<p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
7808 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
7809 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
7810 address spaces. Not all targets support all bit widths however.</p>
Mon P Wang28873102008-06-25 08:15:39 +00007811
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007812<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007813 declare i8 @llvm.atomic.load.max.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7814 declare i16 @llvm.atomic.load.max.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7815 declare i32 @llvm.atomic.load.max.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7816 declare i64 @llvm.atomic.load.max.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007817</pre>
7818
7819<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007820 declare i8 @llvm.atomic.load.min.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7821 declare i16 @llvm.atomic.load.min.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7822 declare i32 @llvm.atomic.load.min.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7823 declare i64 @llvm.atomic.load.min.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007824</pre>
7825
7826<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007827 declare i8 @llvm.atomic.load.umax.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7828 declare i16 @llvm.atomic.load.umax.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7829 declare i32 @llvm.atomic.load.umax.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7830 declare i64 @llvm.atomic.load.umax.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007831</pre>
7832
7833<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007834 declare i8 @llvm.atomic.load.umin.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7835 declare i16 @llvm.atomic.load.umin.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7836 declare i32 @llvm.atomic.load.umin.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7837 declare i64 @llvm.atomic.load.umin.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang28873102008-06-25 08:15:39 +00007838</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007839
Mon P Wang28873102008-06-25 08:15:39 +00007840<h5>Overview:</h5>
Eric Christopher6c7e8a02009-12-05 02:46:03 +00007841<p>These intrinsics takes the signed or unsigned minimum or maximum of
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007842 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
7843 original value at <tt>ptr</tt>.</p>
Mon P Wang28873102008-06-25 08:15:39 +00007844
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007845<h5>Arguments:</h5>
7846<p>These intrinsics take two arguments, the first a pointer to an integer value
7847 and the second an integer value. The result is also an integer value. These
7848 integer types can have any bit width, but they must all have the same bit
7849 width. The targets may only lower integer representations they support.</p>
7850
Mon P Wang28873102008-06-25 08:15:39 +00007851<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007852<p>These intrinsics does a series of operations atomically. They first load the
7853 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
7854 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
7855 yield the original value stored at <tt>ptr</tt>.</p>
Mon P Wang28873102008-06-25 08:15:39 +00007856
7857<h5>Examples:</h5>
7858<pre>
Victor Hernandez2fee2942009-10-26 23:44:29 +00007859%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7860%ptr = bitcast i8* %mallocP to i32*
7861 store i32 7, %ptr
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007862%result0 = call i32 @llvm.atomic.load.min.i32.p0i32(i32* %ptr, i32 -2)
Mon P Wang28873102008-06-25 08:15:39 +00007863 <i>; yields {i32}:result0 = 7</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007864%result1 = call i32 @llvm.atomic.load.max.i32.p0i32(i32* %ptr, i32 8)
Mon P Wang28873102008-06-25 08:15:39 +00007865 <i>; yields {i32}:result1 = -2</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007866%result2 = call i32 @llvm.atomic.load.umin.i32.p0i32(i32* %ptr, i32 10)
Mon P Wang28873102008-06-25 08:15:39 +00007867 <i>; yields {i32}:result2 = 8</i>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00007868%result3 = call i32 @llvm.atomic.load.umax.i32.p0i32(i32* %ptr, i32 30)
Mon P Wang28873102008-06-25 08:15:39 +00007869 <i>; yields {i32}:result3 = 8</i>
7870%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
7871</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00007872
Mon P Wang28873102008-06-25 08:15:39 +00007873</div>
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00007874
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007875</div>
Nick Lewyckycc271862009-10-13 07:03:23 +00007876
7877<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007878<h3>
Nick Lewyckycc271862009-10-13 07:03:23 +00007879 <a name="int_memorymarkers">Memory Use Markers</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007880</h3>
Nick Lewyckycc271862009-10-13 07:03:23 +00007881
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007882<div>
Nick Lewyckycc271862009-10-13 07:03:23 +00007883
7884<p>This class of intrinsics exists to information about the lifetime of memory
7885 objects and ranges where variables are immutable.</p>
7886
Nick Lewyckycc271862009-10-13 07:03:23 +00007887<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007888<h4>
Nick Lewyckycc271862009-10-13 07:03:23 +00007889 <a name="int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007890</h4>
Nick Lewyckycc271862009-10-13 07:03:23 +00007891
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007892<div>
Nick Lewyckycc271862009-10-13 07:03:23 +00007893
7894<h5>Syntax:</h5>
7895<pre>
7896 declare void @llvm.lifetime.start(i64 &lt;size&gt;, i8* nocapture &lt;ptr&gt;)
7897</pre>
7898
7899<h5>Overview:</h5>
7900<p>The '<tt>llvm.lifetime.start</tt>' intrinsic specifies the start of a memory
7901 object's lifetime.</p>
7902
7903<h5>Arguments:</h5>
Nick Lewycky321333e2009-10-13 07:57:33 +00007904<p>The first argument is a constant integer representing the size of the
7905 object, or -1 if it is variable sized. The second argument is a pointer to
7906 the object.</p>
Nick Lewyckycc271862009-10-13 07:03:23 +00007907
7908<h5>Semantics:</h5>
7909<p>This intrinsic indicates that before this point in the code, the value of the
7910 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
Nick Lewycky8d336592009-10-27 16:56:58 +00007911 never be used and has an undefined value. A load from the pointer that
7912 precedes this intrinsic can be replaced with
Nick Lewyckycc271862009-10-13 07:03:23 +00007913 <tt>'<a href="#undefvalues">undef</a>'</tt>.</p>
7914
7915</div>
7916
7917<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007918<h4>
Nick Lewyckycc271862009-10-13 07:03:23 +00007919 <a name="int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007920</h4>
Nick Lewyckycc271862009-10-13 07:03:23 +00007921
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007922<div>
Nick Lewyckycc271862009-10-13 07:03:23 +00007923
7924<h5>Syntax:</h5>
7925<pre>
7926 declare void @llvm.lifetime.end(i64 &lt;size&gt;, i8* nocapture &lt;ptr&gt;)
7927</pre>
7928
7929<h5>Overview:</h5>
7930<p>The '<tt>llvm.lifetime.end</tt>' intrinsic specifies the end of a memory
7931 object's lifetime.</p>
7932
7933<h5>Arguments:</h5>
Nick Lewycky321333e2009-10-13 07:57:33 +00007934<p>The first argument is a constant integer representing the size of the
7935 object, or -1 if it is variable sized. The second argument is a pointer to
7936 the object.</p>
Nick Lewyckycc271862009-10-13 07:03:23 +00007937
7938<h5>Semantics:</h5>
7939<p>This intrinsic indicates that after this point in the code, the value of the
7940 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7941 never be used and has an undefined value. Any stores into the memory object
7942 following this intrinsic may be removed as dead.
7943
7944</div>
7945
7946<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007947<h4>
Nick Lewyckycc271862009-10-13 07:03:23 +00007948 <a name="int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007949</h4>
Nick Lewyckycc271862009-10-13 07:03:23 +00007950
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007951<div>
Nick Lewyckycc271862009-10-13 07:03:23 +00007952
7953<h5>Syntax:</h5>
7954<pre>
Nick Lewycky29b6cb42010-11-30 04:13:41 +00007955 declare {}* @llvm.invariant.start(i64 &lt;size&gt;, i8* nocapture &lt;ptr&gt;)
Nick Lewyckycc271862009-10-13 07:03:23 +00007956</pre>
7957
7958<h5>Overview:</h5>
7959<p>The '<tt>llvm.invariant.start</tt>' intrinsic specifies that the contents of
7960 a memory object will not change.</p>
7961
7962<h5>Arguments:</h5>
Nick Lewycky321333e2009-10-13 07:57:33 +00007963<p>The first argument is a constant integer representing the size of the
7964 object, or -1 if it is variable sized. The second argument is a pointer to
7965 the object.</p>
Nick Lewyckycc271862009-10-13 07:03:23 +00007966
7967<h5>Semantics:</h5>
7968<p>This intrinsic indicates that until an <tt>llvm.invariant.end</tt> that uses
7969 the return value, the referenced memory location is constant and
7970 unchanging.</p>
7971
7972</div>
7973
7974<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007975<h4>
Nick Lewyckycc271862009-10-13 07:03:23 +00007976 <a name="int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00007977</h4>
Nick Lewyckycc271862009-10-13 07:03:23 +00007978
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00007979<div>
Nick Lewyckycc271862009-10-13 07:03:23 +00007980
7981<h5>Syntax:</h5>
7982<pre>
7983 declare void @llvm.invariant.end({}* &lt;start&gt;, i64 &lt;size&gt;, i8* nocapture &lt;ptr&gt;)
7984</pre>
7985
7986<h5>Overview:</h5>
7987<p>The '<tt>llvm.invariant.end</tt>' intrinsic specifies that the contents of
7988 a memory object are mutable.</p>
7989
7990<h5>Arguments:</h5>
7991<p>The first argument is the matching <tt>llvm.invariant.start</tt> intrinsic.
Nick Lewycky321333e2009-10-13 07:57:33 +00007992 The second argument is a constant integer representing the size of the
7993 object, or -1 if it is variable sized and the third argument is a pointer
7994 to the object.</p>
Nick Lewyckycc271862009-10-13 07:03:23 +00007995
7996<h5>Semantics:</h5>
7997<p>This intrinsic indicates that the memory is mutable again.</p>
7998
7999</div>
8000
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00008001</div>
8002
Andrew Lenharth22c5c1b2008-02-16 01:24:58 +00008003<!-- ======================================================================= -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00008004<h3>
Tanya Lattner6d806e92007-06-15 20:50:54 +00008005 <a name="int_general">General Intrinsics</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00008006</h3>
Tanya Lattner6d806e92007-06-15 20:50:54 +00008007
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00008008<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008009
8010<p>This class of intrinsics is designed to be generic and has no specific
8011 purpose.</p>
8012
Tanya Lattner6d806e92007-06-15 20:50:54 +00008013<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00008014<h4>
Tanya Lattner6d806e92007-06-15 20:50:54 +00008015 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00008016</h4>
Tanya Lattner6d806e92007-06-15 20:50:54 +00008017
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00008018<div>
Tanya Lattner6d806e92007-06-15 20:50:54 +00008019
8020<h5>Syntax:</h5>
8021<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00008022 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 +00008023</pre>
8024
8025<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008026<p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
Tanya Lattner6d806e92007-06-15 20:50:54 +00008027
8028<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008029<p>The first argument is a pointer to a value, the second is a pointer to a
8030 global string, the third is a pointer to a global string which is the source
8031 file name, and the last argument is the line number.</p>
Tanya Lattner6d806e92007-06-15 20:50:54 +00008032
8033<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008034<p>This intrinsic allows annotation of local variables with arbitrary strings.
8035 This can be useful for special purpose optimizations that want to look for
8036 these annotations. These have no other defined use, they are ignored by code
8037 generation and optimization.</p>
Tanya Lattner6d806e92007-06-15 20:50:54 +00008038
Tanya Lattner6d806e92007-06-15 20:50:54 +00008039</div>
8040
Tanya Lattnerb6367882007-09-21 22:59:12 +00008041<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00008042<h4>
Tanya Lattnere1a8da02007-09-21 23:57:59 +00008043 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00008044</h4>
Tanya Lattnerb6367882007-09-21 22:59:12 +00008045
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00008046<div>
Tanya Lattnerb6367882007-09-21 22:59:12 +00008047
8048<h5>Syntax:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008049<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
8050 any integer bit width.</p>
8051
Tanya Lattnerb6367882007-09-21 22:59:12 +00008052<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00008053 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
8054 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
8055 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
8056 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
8057 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 +00008058</pre>
8059
8060<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008061<p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
Tanya Lattnerb6367882007-09-21 22:59:12 +00008062
8063<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008064<p>The first argument is an integer value (result of some expression), the
8065 second is a pointer to a global string, the third is a pointer to a global
8066 string which is the source file name, and the last argument is the line
8067 number. It returns the value of the first argument.</p>
Tanya Lattnerb6367882007-09-21 22:59:12 +00008068
8069<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008070<p>This intrinsic allows annotations to be put on arbitrary expressions with
8071 arbitrary strings. This can be useful for special purpose optimizations that
8072 want to look for these annotations. These have no other defined use, they
8073 are ignored by code generation and optimization.</p>
Tanya Lattnerb6367882007-09-21 22:59:12 +00008074
Tanya Lattnerb6367882007-09-21 22:59:12 +00008075</div>
Jim Laskeydd4ef1b2007-03-14 19:31:19 +00008076
Anton Korobeynikov4cb86182008-01-15 22:31:34 +00008077<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00008078<h4>
Anton Korobeynikov4cb86182008-01-15 22:31:34 +00008079 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00008080</h4>
Anton Korobeynikov4cb86182008-01-15 22:31:34 +00008081
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00008082<div>
Anton Korobeynikov4cb86182008-01-15 22:31:34 +00008083
8084<h5>Syntax:</h5>
8085<pre>
8086 declare void @llvm.trap()
8087</pre>
8088
8089<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008090<p>The '<tt>llvm.trap</tt>' intrinsic.</p>
Anton Korobeynikov4cb86182008-01-15 22:31:34 +00008091
8092<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008093<p>None.</p>
Anton Korobeynikov4cb86182008-01-15 22:31:34 +00008094
8095<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008096<p>This intrinsics is lowered to the target dependent trap instruction. If the
8097 target does not have a trap instruction, this intrinsic will be lowered to
8098 the call of the <tt>abort()</tt> function.</p>
Anton Korobeynikov4cb86182008-01-15 22:31:34 +00008099
Anton Korobeynikov4cb86182008-01-15 22:31:34 +00008100</div>
8101
Bill Wendling69e4adb2008-11-19 05:56:17 +00008102<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00008103<h4>
Misha Brukmandccb0252008-11-22 23:55:29 +00008104 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00008105</h4>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008106
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00008107<div>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008108
Bill Wendling69e4adb2008-11-19 05:56:17 +00008109<h5>Syntax:</h5>
8110<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00008111 declare void @llvm.stackprotector(i8* &lt;guard&gt;, i8** &lt;slot&gt;)
Bill Wendling69e4adb2008-11-19 05:56:17 +00008112</pre>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008113
Bill Wendling69e4adb2008-11-19 05:56:17 +00008114<h5>Overview:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008115<p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
8116 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
8117 ensure that it is placed on the stack before local variables.</p>
8118
Bill Wendling69e4adb2008-11-19 05:56:17 +00008119<h5>Arguments:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008120<p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
8121 arguments. The first argument is the value loaded from the stack
8122 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
8123 that has enough space to hold the value of the guard.</p>
8124
Bill Wendling69e4adb2008-11-19 05:56:17 +00008125<h5>Semantics:</h5>
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008126<p>This intrinsic causes the prologue/epilogue inserter to force the position of
8127 the <tt>AllocaInst</tt> stack slot to be before local variables on the
8128 stack. This is to ensure that if a local variable on the stack is
8129 overwritten, it will destroy the value of the guard. When the function exits,
Bill Wendling1b383ba2010-10-27 01:07:41 +00008130 the guard on the stack is checked against the original guard. If they are
Bill Wendlinge910b4c2009-07-20 02:29:24 +00008131 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
8132 function.</p>
8133
Bill Wendling69e4adb2008-11-19 05:56:17 +00008134</div>
8135
Eric Christopher0e671492009-11-30 08:03:53 +00008136<!-- _______________________________________________________________________ -->
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00008137<h4>
Eric Christopher0e671492009-11-30 08:03:53 +00008138 <a name="int_objectsize">'<tt>llvm.objectsize</tt>' Intrinsic</a>
NAKAMURA Takumi05d02652011-04-18 23:59:50 +00008139</h4>
Eric Christopher0e671492009-11-30 08:03:53 +00008140
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00008141<div>
Eric Christopher0e671492009-11-30 08:03:53 +00008142
8143<h5>Syntax:</h5>
8144<pre>
Dan Gohman3dfb3cf2010-05-28 17:07:41 +00008145 declare i32 @llvm.objectsize.i32(i8* &lt;object&gt;, i1 &lt;type&gt;)
8146 declare i64 @llvm.objectsize.i64(i8* &lt;object&gt;, i1 &lt;type&gt;)
Eric Christopher0e671492009-11-30 08:03:53 +00008147</pre>
8148
8149<h5>Overview:</h5>
Bill Wendling1b383ba2010-10-27 01:07:41 +00008150<p>The <tt>llvm.objectsize</tt> intrinsic is designed to provide information to
8151 the optimizers to determine at compile time whether a) an operation (like
8152 memcpy) will overflow a buffer that corresponds to an object, or b) that a
8153 runtime check for overflow isn't necessary. An object in this context means
8154 an allocation of a specific class, structure, array, or other object.</p>
Eric Christopher0e671492009-11-30 08:03:53 +00008155
8156<h5>Arguments:</h5>
Bill Wendling1b383ba2010-10-27 01:07:41 +00008157<p>The <tt>llvm.objectsize</tt> intrinsic takes two arguments. The first
Eric Christopher8295a0a2009-12-23 00:29:49 +00008158 argument is a pointer to or into the <tt>object</tt>. The second argument
Bill Wendling1b383ba2010-10-27 01:07:41 +00008159 is a boolean 0 or 1. This argument determines whether you want the
8160 maximum (0) or minimum (1) bytes remaining. This needs to be a literal 0 or
Eric Christopher8295a0a2009-12-23 00:29:49 +00008161 1, variables are not allowed.</p>
8162
Eric Christopher0e671492009-11-30 08:03:53 +00008163<h5>Semantics:</h5>
8164<p>The <tt>llvm.objectsize</tt> intrinsic is lowered to either a constant
Bill Wendling1b383ba2010-10-27 01:07:41 +00008165 representing the size of the object concerned, or <tt>i32/i64 -1 or 0</tt>,
8166 depending on the <tt>type</tt> argument, if the size cannot be determined at
8167 compile time.</p>
Eric Christopher0e671492009-11-30 08:03:53 +00008168
8169</div>
8170
NAKAMURA Takumif5af6ad2011-04-23 00:30:22 +00008171</div>
8172
8173</div>
8174
Chris Lattner00950542001-06-06 20:29:01 +00008175<!-- *********************************************************************** -->
Chris Lattner00950542001-06-06 20:29:01 +00008176<hr>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00008177<address>
8178 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman44408702008-12-11 17:34:48 +00008179 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00008180 <a href="http://validator.w3.org/check/referer"><img
Misha Brukman44408702008-12-11 17:34:48 +00008181 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00008182
8183 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
NAKAMURA Takumib9a33632011-04-09 02:13:37 +00008184 <a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br>
Misha Brukmandaa4cb02004-03-01 17:47:27 +00008185 Last modified: $Date$
8186</address>
Chris Lattnerc7d3ab32008-01-04 04:33:49 +00008187
Misha Brukman9d0919f2003-11-08 01:05:38 +00008188</body>
8189</html>