blob: 7422d72a59d367c938df8e6320c1a804b5f010ea [file] [log] [blame]
Misha Brukmanc501f552004-03-01 17:47:27 +00001<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
Misha Brukman76307852003-11-08 01:05:38 +00003<html>
4<head>
5 <title>LLVM Assembly Language Reference Manual</title>
Reid Spencercb84e432004-08-26 20:44:00 +00006 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
Eric Christopher455c5772009-12-05 02:46:03 +00008 <meta name="description"
Reid Spencercb84e432004-08-26 20:44:00 +00009 content="LLVM Assembly Language Reference Manual.">
Misha Brukman76307852003-11-08 01:05:38 +000010 <link rel="stylesheet" href="llvm.css" type="text/css">
11</head>
Chris Lattner757528b0b2004-05-23 21:06:01 +000012
Misha Brukman76307852003-11-08 01:05:38 +000013<body>
Chris Lattner757528b0b2004-05-23 21:06:01 +000014
Chris Lattner48b383b02003-11-25 01:02:51 +000015<div class="doc_title"> LLVM Language Reference Manual </div>
Chris Lattner2f7c9632001-06-06 20:29:01 +000016<ol>
Misha Brukman76307852003-11-08 01:05:38 +000017 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
Chris Lattner6af02f32004-12-09 16:11:40 +000020 <li><a href="#highlevel">High Level Structure</a>
21 <ol>
22 <li><a href="#modulestructure">Module Structure</a></li>
Bill Wendlinga3c6f6b2009-07-20 01:03:30 +000023 <li><a href="#linkage">Linkage Types</a>
24 <ol>
Bill Wendling8693ef82009-07-20 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 Wendling03bcd6e2010-07-01 21:55:59 +000027 <li><a href="#linkage_linker_private_weak">'<tt>linker_private_weak</tt>' Linkage</a></li>
Bill Wendling8693ef82009-07-20 02:41:50 +000028 <li><a href="#linkage_internal">'<tt>internal</tt>' Linkage</a></li>
29 <li><a href="#linkage_available_externally">'<tt>available_externally</tt>' Linkage</a></li>
30 <li><a href="#linkage_linkonce">'<tt>linkonce</tt>' Linkage</a></li>
31 <li><a href="#linkage_common">'<tt>common</tt>' Linkage</a></li>
32 <li><a href="#linkage_weak">'<tt>weak</tt>' Linkage</a></li>
33 <li><a href="#linkage_appending">'<tt>appending</tt>' Linkage</a></li>
34 <li><a href="#linkage_externweak">'<tt>extern_weak</tt>' Linkage</a></li>
Chris Lattner80d73c72009-10-10 18:26:06 +000035 <li><a href="#linkage_linkonce_odr">'<tt>linkonce_odr</tt>' Linkage</a></li>
Bill Wendling8693ef82009-07-20 02:41:50 +000036 <li><a href="#linkage_weak">'<tt>weak_odr</tt>' Linkage</a></li>
37 <li><a href="#linkage_external">'<tt>externally visible</tt>' Linkage</a></li>
38 <li><a href="#linkage_dllimport">'<tt>dllimport</tt>' Linkage</a></li>
39 <li><a href="#linkage_dllexport">'<tt>dllexport</tt>' Linkage</a></li>
Bill Wendlinga3c6f6b2009-07-20 01:03:30 +000040 </ol>
41 </li>
Chris Lattner0132aff2005-05-06 22:57:40 +000042 <li><a href="#callingconv">Calling Conventions</a></li>
Chris Lattnerbc088212009-01-11 20:53:49 +000043 <li><a href="#namedtypes">Named Types</a></li>
Chris Lattner6af02f32004-12-09 16:11:40 +000044 <li><a href="#globalvars">Global Variables</a></li>
Chris Lattner91c15c42006-01-23 23:23:47 +000045 <li><a href="#functionstructure">Functions</a></li>
Dan Gohmanef9462f2008-10-14 16:51:45 +000046 <li><a href="#aliasstructure">Aliases</a></li>
Devang Pateld1a89692010-01-11 19:35:55 +000047 <li><a href="#namedmetadatastructure">Named Metadata</a></li>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +000048 <li><a href="#paramattrs">Parameter Attributes</a></li>
Devang Patel9eb525d2008-09-26 23:51:19 +000049 <li><a href="#fnattrs">Function Attributes</a></li>
Gordon Henriksen71183b62007-12-10 03:18:06 +000050 <li><a href="#gc">Garbage Collector Names</a></li>
Chris Lattner91c15c42006-01-23 23:23:47 +000051 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
Reid Spencer50c723a2007-02-19 23:54:10 +000052 <li><a href="#datalayout">Data Layout</a></li>
Dan Gohman6154a012009-07-27 18:07:55 +000053 <li><a href="#pointeraliasing">Pointer Aliasing Rules</a></li>
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +000054 <li><a href="#volatile">Volatile Memory Accesses</a></li>
Chris Lattner6af02f32004-12-09 16:11:40 +000055 </ol>
56 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +000057 <li><a href="#typesystem">Type System</a>
58 <ol>
Chris Lattner7824d182008-01-04 04:32:38 +000059 <li><a href="#t_classifications">Type Classifications</a></li>
Eric Christopher455c5772009-12-05 02:46:03 +000060 <li><a href="#t_primitive">Primitive Types</a>
Chris Lattner48b383b02003-11-25 01:02:51 +000061 <ol>
Nick Lewycky84a1eeb2009-09-27 00:45:11 +000062 <li><a href="#t_integer">Integer Type</a></li>
Chris Lattner7824d182008-01-04 04:32:38 +000063 <li><a href="#t_floating">Floating Point Types</a></li>
64 <li><a href="#t_void">Void Type</a></li>
65 <li><a href="#t_label">Label Type</a></li>
Nick Lewyckyadbc2842009-05-30 05:06:04 +000066 <li><a href="#t_metadata">Metadata Type</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000067 </ol>
68 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +000069 <li><a href="#t_derived">Derived Types</a>
70 <ol>
Chris Lattner392be582010-02-12 20:49:41 +000071 <li><a href="#t_aggregate">Aggregate Types</a>
72 <ol>
73 <li><a href="#t_array">Array Type</a></li>
74 <li><a href="#t_struct">Structure Type</a></li>
75 <li><a href="#t_pstruct">Packed Structure Type</a></li>
76 <li><a href="#t_union">Union Type</a></li>
77 <li><a href="#t_vector">Vector Type</a></li>
78 </ol>
79 </li>
Misha Brukman76307852003-11-08 01:05:38 +000080 <li><a href="#t_function">Function Type</a></li>
81 <li><a href="#t_pointer">Pointer Type</a></li>
Chris Lattner37b6b092005-04-25 17:34:15 +000082 <li><a href="#t_opaque">Opaque Type</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000083 </ol>
84 </li>
Chris Lattnercf7a5842009-02-02 07:32:36 +000085 <li><a href="#t_uprefs">Type Up-references</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000086 </ol>
87 </li>
Chris Lattner6af02f32004-12-09 16:11:40 +000088 <li><a href="#constants">Constants</a>
Chris Lattner74d3f822004-12-09 17:30:23 +000089 <ol>
Dan Gohmanef9462f2008-10-14 16:51:45 +000090 <li><a href="#simpleconstants">Simple Constants</a></li>
Chris Lattner361bfcd2009-02-28 18:32:25 +000091 <li><a href="#complexconstants">Complex Constants</a></li>
Dan Gohmanef9462f2008-10-14 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 Gohmanffc9a6b2010-04-22 23:14:21 +000094 <li><a href="#trapvalues">Trap Values</a></li>
Chris Lattner2bfd3202009-10-27 21:19:13 +000095 <li><a href="#blockaddress">Addresses of Basic Blocks</a></li>
Dan Gohmanef9462f2008-10-14 16:51:45 +000096 <li><a href="#constantexprs">Constant Expressions</a></li>
Chris Lattner74d3f822004-12-09 17:30:23 +000097 </ol>
Chris Lattner48b383b02003-11-25 01:02:51 +000098 </li>
Chris Lattner98f013c2006-01-25 23:47:57 +000099 <li><a href="#othervalues">Other Values</a>
100 <ol>
Dan Gohmanef9462f2008-10-14 16:51:45 +0000101 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
Devang Pateld1a89692010-01-11 19:35:55 +0000102 <li><a href="#metadata">Metadata Nodes and Metadata Strings</a></li>
Chris Lattner98f013c2006-01-25 23:47:57 +0000103 </ol>
104 </li>
Chris Lattnerae76db52009-07-20 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 Lattner58f9bb22009-07-20 06:14:25 +0000108 <li><a href="#intg_compiler_used">The '<tt>llvm.compiler.used</tt>'
109 Global Variable</a></li>
Chris Lattnerae76db52009-07-20 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 Lattner2f7c9632001-06-06 20:29:01 +0000116 <li><a href="#instref">Instruction Reference</a>
117 <ol>
118 <li><a href="#terminators">Terminator Instructions</a>
119 <ol>
Chris Lattner48b383b02003-11-25 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 Brukman76307852003-11-08 01:05:38 +0000122 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
Chris Lattnerd04cb6d2009-10-28 00:19:10 +0000123 <li><a href="#i_indirectbr">'<tt>indirectbr</tt>' Instruction</a></li>
Misha Brukman76307852003-11-08 01:05:38 +0000124 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000125 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000126 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000127 </ol>
128 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000129 <li><a href="#binaryops">Binary Operations</a>
130 <ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000131 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
Dan Gohmana5b96452009-06-04 22:49:04 +0000132 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000133 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
Dan Gohmana5b96452009-06-04 22:49:04 +0000134 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000135 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
Dan Gohmana5b96452009-06-04 22:49:04 +0000136 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000137 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
138 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
139 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
Reid Spencer7eb55b32006-11-02 01:53:59 +0000140 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
141 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
142 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000143 </ol>
144 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000145 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
146 <ol>
Reid Spencer2ab01932007-02-02 13:57:07 +0000147 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
148 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
149 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
Misha Brukman76307852003-11-08 01:05:38 +0000150 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000151 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
Misha Brukman76307852003-11-08 01:05:38 +0000152 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000153 </ol>
154 </li>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000155 <li><a href="#vectorops">Vector Operations</a>
156 <ol>
157 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
158 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
159 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000160 </ol>
161 </li>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000162 <li><a href="#aggregateops">Aggregate Operations</a>
163 <ol>
164 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
165 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
166 </ol>
167 </li>
Chris Lattner6ab66722006-08-15 00:45:58 +0000168 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000169 <ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000170 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
Robert Bocchino820bc75b2006-02-17 21:18:08 +0000171 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
172 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
173 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000174 </ol>
175 </li>
Reid Spencer97c5fa42006-11-08 01:18:52 +0000176 <li><a href="#convertops">Conversion Operations</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000177 <ol>
178 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
179 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
180 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
181 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
182 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
Reid Spencer51b07252006-11-09 23:03:26 +0000183 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
184 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
185 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
186 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
Reid Spencerb7344ff2006-11-11 21:00:47 +0000187 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
188 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
Reid Spencer5b950642006-11-11 23:08:07 +0000189 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000190 </ol>
Dan Gohmanef9462f2008-10-14 16:51:45 +0000191 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000192 <li><a href="#otherops">Other Operations</a>
193 <ol>
Reid Spencerc828a0e2006-11-18 21:50:54 +0000194 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
195 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000196 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000197 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000198 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
Chris Lattner33337472006-01-13 23:26:01 +0000199 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000200 </ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000201 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000202 </ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000203 </li>
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000204 <li><a href="#intrinsics">Intrinsic Functions</a>
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000205 <ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000206 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
207 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000208 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
209 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
210 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000211 </ol>
212 </li>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000213 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
214 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000215 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
216 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
217 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000218 </ol>
219 </li>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000220 <li><a href="#int_codegen">Code Generator Intrinsics</a>
221 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000222 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
223 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
224 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
225 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
226 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
227 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
Dan Gohmane58f7b32010-05-26 21:56:15 +0000228 <li><a href="#int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
John Criswellaa1c3c12004-04-09 16:43:20 +0000229 </ol>
230 </li>
Chris Lattnerfee11462004-02-12 17:01:32 +0000231 <li><a href="#int_libc">Standard C Library Intrinsics</a>
232 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000233 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
234 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
235 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
236 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
237 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000238 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
239 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
240 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Chris Lattnerfee11462004-02-12 17:01:32 +0000241 </ol>
242 </li>
Nate Begeman0f223bb2006-01-13 23:26:38 +0000243 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000244 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000245 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
Chris Lattnerb748c672006-01-16 22:34:14 +0000246 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
247 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
248 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000249 </ol>
250 </li>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000251 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
252 <ol>
Bill Wendlingfd2bd722009-02-08 04:04:40 +0000253 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
254 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
255 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
256 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
257 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
Bill Wendlingb9a73272009-02-08 23:00:09 +0000258 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000259 </ol>
260 </li>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +0000261 <li><a href="#int_fp16">Half Precision Floating Point Intrinsics</a>
262 <ol>
Chris Lattnerbbd8bd32010-03-14 23:03:31 +0000263 <li><a href="#int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a></li>
264 <li><a href="#int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a></li>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +0000265 </ol>
266 </li>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000267 <li><a href="#int_debugger">Debugger intrinsics</a></li>
Jim Laskey2211f492007-03-14 19:31:19 +0000268 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sands86e01192007-09-11 14:10:23 +0000269 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands644f9172007-07-27 12:58:54 +0000270 <ol>
271 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands644f9172007-07-27 12:58:54 +0000272 </ol>
273 </li>
Bill Wendlingf85850f2008-11-18 22:10:53 +0000274 <li><a href="#int_atomics">Atomic intrinsics</a>
275 <ol>
276 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
277 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
278 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
279 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
280 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
281 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
282 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
283 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
284 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
285 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
286 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
287 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
288 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
289 </ol>
290 </li>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000291 <li><a href="#int_memorymarkers">Memory Use Markers</a>
292 <ol>
293 <li><a href="#int_lifetime_start"><tt>llvm.lifetime.start</tt></a></li>
294 <li><a href="#int_lifetime_end"><tt>llvm.lifetime.end</tt></a></li>
295 <li><a href="#int_invariant_start"><tt>llvm.invariant.start</tt></a></li>
296 <li><a href="#int_invariant_end"><tt>llvm.invariant.end</tt></a></li>
297 </ol>
298 </li>
Reid Spencer5b2cb0f2007-07-20 19:59:11 +0000299 <li><a href="#int_general">General intrinsics</a>
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000300 <ol>
Reid Spencer5b2cb0f2007-07-20 19:59:11 +0000301 <li><a href="#int_var_annotation">
Bill Wendling14313312008-11-19 05:56:17 +0000302 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Tanya Lattner293c0372007-09-21 22:59:12 +0000303 <li><a href="#int_annotation">
Bill Wendling14313312008-11-19 05:56:17 +0000304 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Anton Korobeynikov06cbb652008-01-15 22:31:34 +0000305 <li><a href="#int_trap">
Bill Wendling14313312008-11-19 05:56:17 +0000306 '<tt>llvm.trap</tt>' Intrinsic</a></li>
307 <li><a href="#int_stackprotector">
308 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
Eric Christopher73484322009-11-30 08:03:53 +0000309 <li><a href="#int_objectsize">
310 '<tt>llvm.objectsize</tt>' Intrinsic</a></li>
Tanya Lattner293c0372007-09-21 22:59:12 +0000311 </ol>
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000312 </li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000313 </ol>
314 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000315</ol>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000316
317<div class="doc_author">
318 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
319 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
Misha Brukman76307852003-11-08 01:05:38 +0000320</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000321
Chris Lattner2f7c9632001-06-06 20:29:01 +0000322<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +0000323<div class="doc_section"> <a name="abstract">Abstract </a></div>
324<!-- *********************************************************************** -->
Chris Lattner757528b0b2004-05-23 21:06:01 +0000325
Misha Brukman76307852003-11-08 01:05:38 +0000326<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000327
328<p>This document is a reference manual for the LLVM assembly language. LLVM is
329 a Static Single Assignment (SSA) based representation that provides type
330 safety, low-level operations, flexibility, and the capability of representing
331 'all' high-level languages cleanly. It is the common code representation
332 used throughout all phases of the LLVM compilation strategy.</p>
333
Misha Brukman76307852003-11-08 01:05:38 +0000334</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000335
Chris Lattner2f7c9632001-06-06 20:29:01 +0000336<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +0000337<div class="doc_section"> <a name="introduction">Introduction</a> </div>
338<!-- *********************************************************************** -->
Chris Lattner757528b0b2004-05-23 21:06:01 +0000339
Misha Brukman76307852003-11-08 01:05:38 +0000340<div class="doc_text">
Chris Lattner757528b0b2004-05-23 21:06:01 +0000341
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000342<p>The LLVM code representation is designed to be used in three different forms:
343 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
344 for fast loading by a Just-In-Time compiler), and as a human readable
345 assembly language representation. This allows LLVM to provide a powerful
346 intermediate representation for efficient compiler transformations and
347 analysis, while providing a natural means to debug and visualize the
348 transformations. The three different forms of LLVM are all equivalent. This
349 document describes the human readable representation and notation.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000350
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000351<p>The LLVM representation aims to be light-weight and low-level while being
352 expressive, typed, and extensible at the same time. It aims to be a
353 "universal IR" of sorts, by being at a low enough level that high-level ideas
354 may be cleanly mapped to it (similar to how microprocessors are "universal
355 IR's", allowing many source languages to be mapped to them). By providing
356 type information, LLVM can be used as the target of optimizations: for
357 example, through pointer analysis, it can be proven that a C automatic
Bill Wendling7f4a3362009-11-02 00:24:16 +0000358 variable is never accessed outside of the current function, allowing it to
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000359 be promoted to a simple SSA value instead of a memory location.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000360
Misha Brukman76307852003-11-08 01:05:38 +0000361</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000362
Chris Lattner2f7c9632001-06-06 20:29:01 +0000363<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +0000364<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000365
Misha Brukman76307852003-11-08 01:05:38 +0000366<div class="doc_text">
Chris Lattner757528b0b2004-05-23 21:06:01 +0000367
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000368<p>It is important to note that this document describes 'well formed' LLVM
369 assembly language. There is a difference between what the parser accepts and
370 what is considered 'well formed'. For example, the following instruction is
371 syntactically okay, but not well formed:</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000372
Benjamin Kramer79698be2010-07-13 12:26:09 +0000373<pre class="doc_code">
Bill Wendling3716c5d2007-05-29 09:04:49 +0000374%x = <a href="#i_add">add</a> i32 1, %x
Chris Lattner757528b0b2004-05-23 21:06:01 +0000375</pre>
376
Bill Wendling7f4a3362009-11-02 00:24:16 +0000377<p>because the definition of <tt>%x</tt> does not dominate all of its uses. The
378 LLVM infrastructure provides a verification pass that may be used to verify
379 that an LLVM module is well formed. This pass is automatically run by the
380 parser after parsing input assembly and by the optimizer before it outputs
381 bitcode. The violations pointed out by the verifier pass indicate bugs in
382 transformation passes or input to the parser.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000383
Bill Wendling3716c5d2007-05-29 09:04:49 +0000384</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000385
Chris Lattner87a3dbe2007-10-03 17:34:29 +0000386<!-- Describe the typesetting conventions here. -->
Chris Lattner757528b0b2004-05-23 21:06:01 +0000387
Chris Lattner2f7c9632001-06-06 20:29:01 +0000388<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +0000389<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000390<!-- *********************************************************************** -->
Chris Lattner757528b0b2004-05-23 21:06:01 +0000391
Misha Brukman76307852003-11-08 01:05:38 +0000392<div class="doc_text">
Chris Lattner757528b0b2004-05-23 21:06:01 +0000393
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000394<p>LLVM identifiers come in two basic types: global and local. Global
395 identifiers (functions, global variables) begin with the <tt>'@'</tt>
396 character. Local identifiers (register names, types) begin with
397 the <tt>'%'</tt> character. Additionally, there are three different formats
398 for identifiers, for different purposes:</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000399
Chris Lattner2f7c9632001-06-06 20:29:01 +0000400<ol>
Reid Spencerb23b65f2007-08-07 14:34:28 +0000401 <li>Named values are represented as a string of characters with their prefix.
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000402 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
403 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
404 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
405 other characters in their names can be surrounded with quotes. Special
406 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
407 ASCII code for the character in hexadecimal. In this way, any character
408 can be used in a name value, even quotes themselves.</li>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000409
Reid Spencerb23b65f2007-08-07 14:34:28 +0000410 <li>Unnamed values are represented as an unsigned numeric value with their
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000411 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000412
Reid Spencer8f08d802004-12-09 18:02:53 +0000413 <li>Constants, which are described in a <a href="#constants">section about
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000414 constants</a>, below.</li>
Misha Brukman76307852003-11-08 01:05:38 +0000415</ol>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000416
Reid Spencerb23b65f2007-08-07 14:34:28 +0000417<p>LLVM requires that values start with a prefix for two reasons: Compilers
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000418 don't need to worry about name clashes with reserved words, and the set of
419 reserved words may be expanded in the future without penalty. Additionally,
420 unnamed identifiers allow a compiler to quickly come up with a temporary
421 variable without having to avoid symbol table conflicts.</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000422
Chris Lattner48b383b02003-11-25 01:02:51 +0000423<p>Reserved words in LLVM are very similar to reserved words in other
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000424 languages. There are keywords for different opcodes
425 ('<tt><a href="#i_add">add</a></tt>',
426 '<tt><a href="#i_bitcast">bitcast</a></tt>',
427 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
428 ('<tt><a href="#t_void">void</a></tt>',
429 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
430 reserved words cannot conflict with variable names, because none of them
431 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000432
433<p>Here is an example of LLVM code to multiply the integer variable
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000434 '<tt>%X</tt>' by 8:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000435
Misha Brukman76307852003-11-08 01:05:38 +0000436<p>The easy way:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000437
Benjamin Kramer79698be2010-07-13 12:26:09 +0000438<pre class="doc_code">
Bill Wendling3716c5d2007-05-29 09:04:49 +0000439%result = <a href="#i_mul">mul</a> i32 %X, 8
Chris Lattnerd79749a2004-12-09 16:36:40 +0000440</pre>
441
Misha Brukman76307852003-11-08 01:05:38 +0000442<p>After strength reduction:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000443
Benjamin Kramer79698be2010-07-13 12:26:09 +0000444<pre class="doc_code">
Bill Wendling3716c5d2007-05-29 09:04:49 +0000445%result = <a href="#i_shl">shl</a> i32 %X, i8 3
Chris Lattnerd79749a2004-12-09 16:36:40 +0000446</pre>
447
Misha Brukman76307852003-11-08 01:05:38 +0000448<p>And the hard way:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000449
Benjamin Kramer79698be2010-07-13 12:26:09 +0000450<pre class="doc_code">
Gabor Greifbd0328f2009-10-28 13:05:07 +0000451%0 = <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
452%1 = <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000453%result = <a href="#i_add">add</a> i32 %1, %1
Chris Lattnerd79749a2004-12-09 16:36:40 +0000454</pre>
455
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000456<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
457 lexical features of LLVM:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000458
Chris Lattner2f7c9632001-06-06 20:29:01 +0000459<ol>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000460 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000461 line.</li>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000462
463 <li>Unnamed temporaries are created when the result of a computation is not
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000464 assigned to a named value.</li>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000465
Misha Brukman76307852003-11-08 01:05:38 +0000466 <li>Unnamed temporaries are numbered sequentially</li>
467</ol>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000468
Bill Wendling7f4a3362009-11-02 00:24:16 +0000469<p>It also shows a convention that we follow in this document. When
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000470 demonstrating instructions, we will follow an instruction with a comment that
471 defines the type and name of value produced. Comments are shown in italic
472 text.</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000473
Misha Brukman76307852003-11-08 01:05:38 +0000474</div>
Chris Lattner6af02f32004-12-09 16:11:40 +0000475
476<!-- *********************************************************************** -->
477<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
478<!-- *********************************************************************** -->
479
480<!-- ======================================================================= -->
481<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
482</div>
483
484<div class="doc_text">
485
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000486<p>LLVM programs are composed of "Module"s, each of which is a translation unit
487 of the input programs. Each module consists of functions, global variables,
488 and symbol table entries. Modules may be combined together with the LLVM
489 linker, which merges function (and global variable) definitions, resolves
490 forward declarations, and merges symbol table entries. Here is an example of
491 the "hello world" module:</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000492
Benjamin Kramer79698be2010-07-13 12:26:09 +0000493<pre class="doc_code">
Chris Lattner54a7be72010-08-17 17:13:42 +0000494<i>; Declare the string constant as a global constant.</i>&nbsp;
495<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>&nbsp;
Chris Lattner6af02f32004-12-09 16:11:40 +0000496
Chris Lattner54a7be72010-08-17 17:13:42 +0000497<i>; External declaration of the puts function</i>&nbsp;
498<a href="#functionstructure">declare</a> i32 @puts(i8*) <i>; i32 (i8*)* </i>&nbsp;
Chris Lattner6af02f32004-12-09 16:11:40 +0000499
500<i>; Definition of main function</i>
Chris Lattner54a7be72010-08-17 17:13:42 +0000501define i32 @main() { <i>; i32()* </i>&nbsp;
502 <i>; Convert [13 x i8]* to i8 *...</i>&nbsp;
503 %cast210 = <a href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8*</i>&nbsp;
Chris Lattner6af02f32004-12-09 16:11:40 +0000504
Chris Lattner54a7be72010-08-17 17:13:42 +0000505 <i>; Call puts function to write out the string to stdout.</i>&nbsp;
506 <a href="#i_call">call</a> i32 @puts(i8* %cast210) <i>; i32</i>&nbsp;
507 <a href="#i_ret">ret</a> i32 0&nbsp;
508}
Devang Pateld1a89692010-01-11 19:35:55 +0000509
510<i>; Named metadata</i>
511!1 = metadata !{i32 41}
512!foo = !{!1, null}
Bill Wendling3716c5d2007-05-29 09:04:49 +0000513</pre>
Chris Lattner6af02f32004-12-09 16:11:40 +0000514
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000515<p>This example is made up of a <a href="#globalvars">global variable</a> named
Devang Pateld1a89692010-01-11 19:35:55 +0000516 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function,
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000517 a <a href="#functionstructure">function definition</a> for
Devang Pateld1a89692010-01-11 19:35:55 +0000518 "<tt>main</tt>" and <a href="#namedmetadatastructure">named metadata</a>
519 "<tt>foo"</tt>.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000520
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000521<p>In general, a module is made up of a list of global values, where both
522 functions and global variables are global values. Global values are
523 represented by a pointer to a memory location (in this case, a pointer to an
524 array of char, and a pointer to a function), and have one of the
525 following <a href="#linkage">linkage types</a>.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000526
Chris Lattnerd79749a2004-12-09 16:36:40 +0000527</div>
528
529<!-- ======================================================================= -->
530<div class="doc_subsection">
531 <a name="linkage">Linkage Types</a>
532</div>
533
534<div class="doc_text">
535
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000536<p>All Global Variables and Functions have one of the following types of
537 linkage:</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000538
539<dl>
Bill Wendling7f4a3362009-11-02 00:24:16 +0000540 <dt><tt><b><a name="linkage_private">private</a></b></tt></dt>
Bill Wendling03bcd6e2010-07-01 21:55:59 +0000541 <dd>Global values with "<tt>private</tt>" linkage are only directly accessible
542 by objects in the current module. In particular, linking code into a
543 module with an private global value may cause the private to be renamed as
544 necessary to avoid collisions. Because the symbol is private to the
545 module, all references can be updated. This doesn't show up in any symbol
546 table in the object file.</dd>
Rafael Espindola6de96a12009-01-15 20:18:42 +0000547
Bill Wendling7f4a3362009-11-02 00:24:16 +0000548 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt></dt>
Bill Wendling03bcd6e2010-07-01 21:55:59 +0000549 <dd>Similar to <tt>private</tt>, but the symbol is passed through the
550 assembler and evaluated by the linker. Unlike normal strong symbols, they
551 are removed by the linker from the final linked image (executable or
552 dynamic library).</dd>
553
554 <dt><tt><b><a name="linkage_linker_private_weak">linker_private_weak</a></b></tt></dt>
555 <dd>Similar to "<tt>linker_private</tt>", but the symbol is weak. Note that
556 <tt>linker_private_weak</tt> symbols are subject to coalescing by the
557 linker. The symbols are removed by the linker from the final linked image
558 (executable or dynamic library).</dd>
Bill Wendlinga3c6f6b2009-07-20 01:03:30 +0000559
Bill Wendling7f4a3362009-11-02 00:24:16 +0000560 <dt><tt><b><a name="linkage_internal">internal</a></b></tt></dt>
Bill Wendling36321712010-06-29 22:34:52 +0000561 <dd>Similar to private, but the value shows as a local symbol
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000562 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
563 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000564
Bill Wendling7f4a3362009-11-02 00:24:16 +0000565 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt></dt>
Chris Lattner184f1be2009-04-13 05:44:34 +0000566 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000567 into the object file corresponding to the LLVM module. They exist to
568 allow inlining and other optimizations to take place given knowledge of
569 the definition of the global, which is known to be somewhere outside the
570 module. Globals with <tt>available_externally</tt> linkage are allowed to
571 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
572 This linkage type is only allowed on definitions, not declarations.</dd>
Chris Lattner184f1be2009-04-13 05:44:34 +0000573
Bill Wendling7f4a3362009-11-02 00:24:16 +0000574 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt></dt>
Chris Lattnere20b4702007-01-14 06:51:48 +0000575 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
Chris Lattner0de4caa2010-01-09 19:15:14 +0000576 the same name when linkage occurs. This can be used to implement
577 some forms of inline functions, templates, or other code which must be
578 generated in each translation unit that uses it, but where the body may
579 be overridden with a more definitive definition later. Unreferenced
580 <tt>linkonce</tt> globals are allowed to be discarded. Note that
581 <tt>linkonce</tt> linkage does not actually allow the optimizer to
582 inline the body of this function into callers because it doesn't know if
583 this definition of the function is the definitive definition within the
584 program or whether it will be overridden by a stronger definition.
585 To enable inlining and other optimizations, use "<tt>linkonce_odr</tt>"
586 linkage.</dd>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000587
Bill Wendling7f4a3362009-11-02 00:24:16 +0000588 <dt><tt><b><a name="linkage_weak">weak</a></b></tt></dt>
Chris Lattnerd0554882009-08-05 05:21:07 +0000589 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
590 <tt>linkonce</tt> linkage, except that unreferenced globals with
591 <tt>weak</tt> linkage may not be discarded. This is used for globals that
592 are declared "weak" in C source code.</dd>
593
Bill Wendling7f4a3362009-11-02 00:24:16 +0000594 <dt><tt><b><a name="linkage_common">common</a></b></tt></dt>
Chris Lattnerd0554882009-08-05 05:21:07 +0000595 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
596 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
597 global scope.
598 Symbols with "<tt>common</tt>" linkage are merged in the same way as
599 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
Chris Lattner0aff0b22009-08-05 05:41:44 +0000600 <tt>common</tt> symbols may not have an explicit section,
Eric Christopher455c5772009-12-05 02:46:03 +0000601 must have a zero initializer, and may not be marked '<a
Chris Lattner0aff0b22009-08-05 05:41:44 +0000602 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
603 have common linkage.</dd>
Chris Lattnerd0554882009-08-05 05:21:07 +0000604
Chris Lattnerd79749a2004-12-09 16:36:40 +0000605
Bill Wendling7f4a3362009-11-02 00:24:16 +0000606 <dt><tt><b><a name="linkage_appending">appending</a></b></tt></dt>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000607 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000608 pointer to array type. When two global variables with appending linkage
609 are linked together, the two global arrays are appended together. This is
610 the LLVM, typesafe, equivalent of having the system linker append together
611 "sections" with identical names when .o files are linked.</dd>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000612
Bill Wendling7f4a3362009-11-02 00:24:16 +0000613 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000614 <dd>The semantics of this linkage follow the ELF object file model: the symbol
615 is weak until linked, if not linked, the symbol becomes null instead of
616 being an undefined reference.</dd>
Anton Korobeynikova0554d92007-01-12 19:20:47 +0000617
Bill Wendling7f4a3362009-11-02 00:24:16 +0000618 <dt><tt><b><a name="linkage_linkonce_odr">linkonce_odr</a></b></tt></dt>
619 <dt><tt><b><a name="linkage_weak_odr">weak_odr</a></b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000620 <dd>Some languages allow differing globals to be merged, such as two functions
621 with different semantics. Other languages, such as <tt>C++</tt>, ensure
Bill Wendling03bcd6e2010-07-01 21:55:59 +0000622 that only equivalent globals are ever merged (the "one definition rule"
623 &mdash; "ODR"). Such languages can use the <tt>linkonce_odr</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000624 and <tt>weak_odr</tt> linkage types to indicate that the global will only
625 be merged with equivalent globals. These linkage types are otherwise the
626 same as their non-<tt>odr</tt> versions.</dd>
Duncan Sands12da8ce2009-03-07 15:45:40 +0000627
Chris Lattner6af02f32004-12-09 16:11:40 +0000628 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000629 <dd>If none of the above identifiers are used, the global is externally
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000630 visible, meaning that it participates in linkage and can be used to
631 resolve external symbol references.</dd>
Reid Spencer7972c472007-04-11 23:49:50 +0000632</dl>
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000633
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000634<p>The next two types of linkage are targeted for Microsoft Windows platform
635 only. They are designed to support importing (exporting) symbols from (to)
636 DLLs (Dynamic Link Libraries).</p>
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000637
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000638<dl>
Bill Wendling7f4a3362009-11-02 00:24:16 +0000639 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt></dt>
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000640 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000641 or variable via a global pointer to a pointer that is set up by the DLL
642 exporting the symbol. On Microsoft Windows targets, the pointer name is
643 formed by combining <code>__imp_</code> and the function or variable
644 name.</dd>
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000645
Bill Wendling7f4a3362009-11-02 00:24:16 +0000646 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt></dt>
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000647 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000648 pointer to a pointer in a DLL, so that it can be referenced with the
649 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
650 name is formed by combining <code>__imp_</code> and the function or
651 variable name.</dd>
Chris Lattner6af02f32004-12-09 16:11:40 +0000652</dl>
653
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000654<p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
655 another module defined a "<tt>.LC0</tt>" variable and was linked with this
656 one, one of the two would be renamed, preventing a collision. Since
657 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
658 declarations), they are accessible outside of the current module.</p>
659
660<p>It is illegal for a function <i>declaration</i> to have any linkage type
661 other than "externally visible", <tt>dllimport</tt>
662 or <tt>extern_weak</tt>.</p>
663
Duncan Sands12da8ce2009-03-07 15:45:40 +0000664<p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000665 or <tt>weak_odr</tt> linkages.</p>
666
Chris Lattner6af02f32004-12-09 16:11:40 +0000667</div>
668
669<!-- ======================================================================= -->
670<div class="doc_subsection">
Chris Lattner0132aff2005-05-06 22:57:40 +0000671 <a name="callingconv">Calling Conventions</a>
672</div>
673
674<div class="doc_text">
675
676<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000677 and <a href="#i_invoke">invokes</a> can all have an optional calling
678 convention specified for the call. The calling convention of any pair of
679 dynamic caller/callee must match, or the behavior of the program is
680 undefined. The following calling conventions are supported by LLVM, and more
681 may be added in the future:</p>
Chris Lattner0132aff2005-05-06 22:57:40 +0000682
683<dl>
684 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
Chris Lattner0132aff2005-05-06 22:57:40 +0000685 <dd>This calling convention (the default if no other calling convention is
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000686 specified) matches the target C calling conventions. This calling
687 convention supports varargs function calls and tolerates some mismatch in
688 the declared prototype and implemented declaration of the function (as
689 does normal C).</dd>
Chris Lattner0132aff2005-05-06 22:57:40 +0000690
691 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
Chris Lattner0132aff2005-05-06 22:57:40 +0000692 <dd>This calling convention attempts to make calls as fast as possible
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000693 (e.g. by passing things in registers). This calling convention allows the
694 target to use whatever tricks it wants to produce fast code for the
695 target, without having to conform to an externally specified ABI
Jeffrey Yasskinb8677462010-01-09 19:44:16 +0000696 (Application Binary Interface).
697 <a href="CodeGenerator.html#tailcallopt">Tail calls can only be optimized
Chris Lattnera179e4d2010-03-11 00:22:57 +0000698 when this or the GHC convention is used.</a> This calling convention
699 does not support varargs and requires the prototype of all callees to
700 exactly match the prototype of the function definition.</dd>
Chris Lattner0132aff2005-05-06 22:57:40 +0000701
702 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
Chris Lattner0132aff2005-05-06 22:57:40 +0000703 <dd>This calling convention attempts to make code in the caller as efficient
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000704 as possible under the assumption that the call is not commonly executed.
705 As such, these calls often preserve all registers so that the call does
706 not break any live ranges in the caller side. This calling convention
707 does not support varargs and requires the prototype of all callees to
708 exactly match the prototype of the function definition.</dd>
Chris Lattner0132aff2005-05-06 22:57:40 +0000709
Chris Lattnera179e4d2010-03-11 00:22:57 +0000710 <dt><b>"<tt>cc <em>10</em></tt>" - GHC convention</b>:</dt>
711 <dd>This calling convention has been implemented specifically for use by the
712 <a href="http://www.haskell.org/ghc">Glasgow Haskell Compiler (GHC)</a>.
713 It passes everything in registers, going to extremes to achieve this by
714 disabling callee save registers. This calling convention should not be
715 used lightly but only for specific situations such as an alternative to
716 the <em>register pinning</em> performance technique often used when
717 implementing functional programming languages.At the moment only X86
718 supports this convention and it has the following limitations:
719 <ul>
720 <li>On <em>X86-32</em> only supports up to 4 bit type parameters. No
721 floating point types are supported.</li>
722 <li>On <em>X86-64</em> only supports up to 10 bit type parameters and
723 6 floating point parameters.</li>
724 </ul>
725 This calling convention supports
726 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> but
727 requires both the caller and callee are using it.
728 </dd>
729
Chris Lattner573f64e2005-05-07 01:46:40 +0000730 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
Chris Lattner0132aff2005-05-06 22:57:40 +0000731 <dd>Any calling convention may be specified by number, allowing
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000732 target-specific calling conventions to be used. Target specific calling
733 conventions start at 64.</dd>
Chris Lattner573f64e2005-05-07 01:46:40 +0000734</dl>
Chris Lattner0132aff2005-05-06 22:57:40 +0000735
736<p>More calling conventions can be added/defined on an as-needed basis, to
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000737 support Pascal conventions or any other well-known target-independent
738 convention.</p>
Chris Lattner0132aff2005-05-06 22:57:40 +0000739
740</div>
741
742<!-- ======================================================================= -->
743<div class="doc_subsection">
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000744 <a name="visibility">Visibility Styles</a>
745</div>
746
747<div class="doc_text">
748
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000749<p>All Global Variables and Functions have one of the following visibility
750 styles:</p>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000751
752<dl>
753 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
Chris Lattner67c37d12008-08-05 18:29:16 +0000754 <dd>On targets that use the ELF object file format, default visibility means
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000755 that the declaration is visible to other modules and, in shared libraries,
756 means that the declared entity may be overridden. On Darwin, default
757 visibility means that the declaration is visible to other modules. Default
758 visibility corresponds to "external linkage" in the language.</dd>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000759
760 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000761 <dd>Two declarations of an object with hidden visibility refer to the same
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000762 object if they are in the same shared object. Usually, hidden visibility
763 indicates that the symbol will not be placed into the dynamic symbol
764 table, so no other module (executable or shared library) can reference it
765 directly.</dd>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000766
Anton Korobeynikov39f3cff2007-04-29 18:35:00 +0000767 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
Anton Korobeynikov39f3cff2007-04-29 18:35:00 +0000768 <dd>On ELF, protected visibility indicates that the symbol will be placed in
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000769 the dynamic symbol table, but that references within the defining module
770 will bind to the local symbol. That is, the symbol cannot be overridden by
771 another module.</dd>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000772</dl>
773
774</div>
775
776<!-- ======================================================================= -->
777<div class="doc_subsection">
Chris Lattnerbc088212009-01-11 20:53:49 +0000778 <a name="namedtypes">Named Types</a>
779</div>
780
781<div class="doc_text">
782
783<p>LLVM IR allows you to specify name aliases for certain types. This can make
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000784 it easier to read the IR and make the IR more condensed (particularly when
785 recursive types are involved). An example of a name specification is:</p>
Chris Lattnerbc088212009-01-11 20:53:49 +0000786
Benjamin Kramer79698be2010-07-13 12:26:09 +0000787<pre class="doc_code">
Chris Lattnerbc088212009-01-11 20:53:49 +0000788%mytype = type { %mytype*, i32 }
789</pre>
Chris Lattnerbc088212009-01-11 20:53:49 +0000790
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000791<p>You may give a name to any <a href="#typesystem">type</a> except
Chris Lattner249b9762010-08-17 23:26:04 +0000792 "<a href="#t_void">void</a>". Type name aliases may be used anywhere a type
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000793 is expected with the syntax "%mytype".</p>
Chris Lattnerbc088212009-01-11 20:53:49 +0000794
795<p>Note that type names are aliases for the structural type that they indicate,
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000796 and that you can therefore specify multiple names for the same type. This
797 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
798 uses structural typing, the name is not part of the type. When printing out
799 LLVM IR, the printer will pick <em>one name</em> to render all types of a
800 particular shape. This means that if you have code where two different
801 source types end up having the same LLVM type, that the dumper will sometimes
802 print the "wrong" or unexpected type. This is an important design point and
803 isn't going to change.</p>
Chris Lattnerbc088212009-01-11 20:53:49 +0000804
805</div>
806
Chris Lattnerbc088212009-01-11 20:53:49 +0000807<!-- ======================================================================= -->
808<div class="doc_subsection">
Chris Lattner6af02f32004-12-09 16:11:40 +0000809 <a name="globalvars">Global Variables</a>
810</div>
811
812<div class="doc_text">
813
Chris Lattner5d5aede2005-02-12 19:30:21 +0000814<p>Global variables define regions of memory allocated at compilation time
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000815 instead of run-time. Global variables may optionally be initialized, may
816 have an explicit section to be placed in, and may have an optional explicit
817 alignment specified. A variable may be defined as "thread_local", which
818 means that it will not be shared by threads (each thread will have a
819 separated copy of the variable). A variable may be defined as a global
820 "constant," which indicates that the contents of the variable
821 will <b>never</b> be modified (enabling better optimization, allowing the
822 global data to be placed in the read-only section of an executable, etc).
823 Note that variables that need runtime initialization cannot be marked
824 "constant" as there is a store to the variable.</p>
Chris Lattner5d5aede2005-02-12 19:30:21 +0000825
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000826<p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
827 constant, even if the final definition of the global is not. This capability
828 can be used to enable slightly better optimization of the program, but
829 requires the language definition to guarantee that optimizations based on the
830 'constantness' are valid for the translation units that do not include the
831 definition.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000832
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000833<p>As SSA values, global variables define pointer values that are in scope
834 (i.e. they dominate) all basic blocks in the program. Global variables
835 always define a pointer to their "content" type because they describe a
836 region of memory, and all memory objects in LLVM are accessed through
837 pointers.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000838
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000839<p>A global variable may be declared to reside in a target-specific numbered
840 address space. For targets that support them, address spaces may affect how
841 optimizations are performed and/or what target instructions are used to
842 access the variable. The default address space is zero. The address space
843 qualifier must precede any other attributes.</p>
Christopher Lamb308121c2007-12-11 09:31:00 +0000844
Chris Lattner662c8722005-11-12 00:45:07 +0000845<p>LLVM allows an explicit section to be specified for globals. If the target
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000846 supports it, it will emit globals to the section specified.</p>
Chris Lattner662c8722005-11-12 00:45:07 +0000847
Chris Lattner78e00bc2010-04-28 00:13:42 +0000848<p>An explicit alignment may be specified for a global, which must be a power
849 of 2. If not present, or if the alignment is set to zero, the alignment of
850 the global is set by the target to whatever it feels convenient. If an
851 explicit alignment is specified, the global is forced to have exactly that
Chris Lattner4bd85e42010-04-28 00:31:12 +0000852 alignment. Targets and optimizers are not allowed to over-align the global
853 if the global has an assigned section. In this case, the extra alignment
854 could be observable: for example, code could assume that the globals are
855 densely packed in their section and try to iterate over them as an array,
856 alignment padding would break this iteration.</p>
Chris Lattner54611b42005-11-06 08:02:57 +0000857
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000858<p>For example, the following defines a global in a numbered address space with
859 an initializer, section, and alignment:</p>
Chris Lattner5760c502007-01-14 00:27:09 +0000860
Benjamin Kramer79698be2010-07-13 12:26:09 +0000861<pre class="doc_code">
Dan Gohmanaaa679b2009-01-11 00:40:00 +0000862@G = addrspace(5) constant float 1.0, section "foo", align 4
Chris Lattner5760c502007-01-14 00:27:09 +0000863</pre>
864
Chris Lattner6af02f32004-12-09 16:11:40 +0000865</div>
866
867
868<!-- ======================================================================= -->
869<div class="doc_subsection">
870 <a name="functionstructure">Functions</a>
871</div>
872
873<div class="doc_text">
874
Dan Gohmana269a0a2010-03-01 17:41:39 +0000875<p>LLVM function definitions consist of the "<tt>define</tt>" keyword, an
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000876 optional <a href="#linkage">linkage type</a>, an optional
877 <a href="#visibility">visibility style</a>, an optional
878 <a href="#callingconv">calling convention</a>, a return type, an optional
879 <a href="#paramattrs">parameter attribute</a> for the return type, a function
880 name, a (possibly empty) argument list (each with optional
881 <a href="#paramattrs">parameter attributes</a>), optional
882 <a href="#fnattrs">function attributes</a>, an optional section, an optional
883 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
884 curly brace, a list of basic blocks, and a closing curly brace.</p>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000885
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000886<p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
887 optional <a href="#linkage">linkage type</a>, an optional
Eric Christopher455c5772009-12-05 02:46:03 +0000888 <a href="#visibility">visibility style</a>, an optional
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000889 <a href="#callingconv">calling convention</a>, a return type, an optional
890 <a href="#paramattrs">parameter attribute</a> for the return type, a function
891 name, a possibly empty list of arguments, an optional alignment, and an
892 optional <a href="#gc">garbage collector name</a>.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000893
Chris Lattner67c37d12008-08-05 18:29:16 +0000894<p>A function definition contains a list of basic blocks, forming the CFG
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000895 (Control Flow Graph) for the function. Each basic block may optionally start
896 with a label (giving the basic block a symbol table entry), contains a list
897 of instructions, and ends with a <a href="#terminators">terminator</a>
898 instruction (such as a branch or function return).</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000899
Chris Lattnera59fb102007-06-08 16:52:14 +0000900<p>The first basic block in a function is special in two ways: it is immediately
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000901 executed on entrance to the function, and it is not allowed to have
902 predecessor basic blocks (i.e. there can not be any branches to the entry
903 block of a function). Because the block can have no predecessors, it also
904 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000905
Chris Lattner662c8722005-11-12 00:45:07 +0000906<p>LLVM allows an explicit section to be specified for functions. If the target
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000907 supports it, it will emit functions to the section specified.</p>
Chris Lattner662c8722005-11-12 00:45:07 +0000908
Chris Lattner54611b42005-11-06 08:02:57 +0000909<p>An explicit alignment may be specified for a function. If not present, or if
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000910 the alignment is set to zero, the alignment of the function is set by the
911 target to whatever it feels convenient. If an explicit alignment is
912 specified, the function is forced to have at least that much alignment. All
913 alignments must be a power of 2.</p>
Chris Lattner54611b42005-11-06 08:02:57 +0000914
Bill Wendling30235112009-07-20 02:39:26 +0000915<h5>Syntax:</h5>
Benjamin Kramer79698be2010-07-13 12:26:09 +0000916<pre class="doc_code">
Chris Lattner0ae02092008-10-13 16:55:18 +0000917define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000918 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
919 &lt;ResultType&gt; @&lt;FunctionName&gt; ([argument list])
920 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
921 [<a href="#gc">gc</a>] { ... }
922</pre>
Devang Patel02256232008-10-07 17:48:33 +0000923
Chris Lattner6af02f32004-12-09 16:11:40 +0000924</div>
925
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000926<!-- ======================================================================= -->
927<div class="doc_subsection">
928 <a name="aliasstructure">Aliases</a>
929</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000930
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000931<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000932
933<p>Aliases act as "second name" for the aliasee value (which can be either
934 function, global variable, another alias or bitcast of global value). Aliases
935 may have an optional <a href="#linkage">linkage type</a>, and an
936 optional <a href="#visibility">visibility style</a>.</p>
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000937
Bill Wendling30235112009-07-20 02:39:26 +0000938<h5>Syntax:</h5>
Benjamin Kramer79698be2010-07-13 12:26:09 +0000939<pre class="doc_code">
Duncan Sands7e99a942008-09-12 20:48:21 +0000940@&lt;Name&gt; = alias [Linkage] [Visibility] &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
Bill Wendling2d8b9a82007-05-29 09:42:13 +0000941</pre>
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000942
943</div>
944
Chris Lattner91c15c42006-01-23 23:23:47 +0000945<!-- ======================================================================= -->
Devang Pateld1a89692010-01-11 19:35:55 +0000946<div class="doc_subsection">
947 <a name="namedmetadatastructure">Named Metadata</a>
948</div>
949
950<div class="doc_text">
951
Chris Lattnerc2f8f162010-01-15 21:50:19 +0000952<p>Named metadata is a collection of metadata. <a href="#metadata">Metadata
Dan Gohman093cb792010-07-21 18:54:18 +0000953 nodes</a> (but not metadata strings) are the only valid operands for
Chris Lattnerc2f8f162010-01-15 21:50:19 +0000954 a named metadata.</p>
Devang Pateld1a89692010-01-11 19:35:55 +0000955
956<h5>Syntax:</h5>
Benjamin Kramer79698be2010-07-13 12:26:09 +0000957<pre class="doc_code">
Dan Gohman093cb792010-07-21 18:54:18 +0000958; Some unnamed metadata nodes, which are referenced by the named metadata.
959!0 = metadata !{metadata !"zero"}
Devang Pateld1a89692010-01-11 19:35:55 +0000960!1 = metadata !{metadata !"one"}
Dan Gohman093cb792010-07-21 18:54:18 +0000961!2 = metadata !{metadata !"two"}
Dan Gohman58cd65f2010-07-13 19:48:13 +0000962; A named metadata.
Dan Gohman093cb792010-07-21 18:54:18 +0000963!name = !{!0, !1, !2}
Devang Pateld1a89692010-01-11 19:35:55 +0000964</pre>
Devang Pateld1a89692010-01-11 19:35:55 +0000965
966</div>
967
968<!-- ======================================================================= -->
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000969<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000970
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000971<div class="doc_text">
972
973<p>The return type and each parameter of a function type may have a set of
974 <i>parameter attributes</i> associated with them. Parameter attributes are
975 used to communicate additional information about the result or parameters of
976 a function. Parameter attributes are considered to be part of the function,
977 not of the function type, so functions with different parameter attributes
978 can have the same function type.</p>
979
980<p>Parameter attributes are simple keywords that follow the type specified. If
981 multiple parameter attributes are needed, they are space separated. For
982 example:</p>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000983
Benjamin Kramer79698be2010-07-13 12:26:09 +0000984<pre class="doc_code">
Nick Lewyckydac78d82009-02-15 23:06:14 +0000985declare i32 @printf(i8* noalias nocapture, ...)
Chris Lattnerd2597d72008-10-04 18:33:34 +0000986declare i32 @atoi(i8 zeroext)
987declare signext i8 @returns_signed_char()
Bill Wendling3716c5d2007-05-29 09:04:49 +0000988</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000989
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000990<p>Note that any attributes for the function result (<tt>nounwind</tt>,
991 <tt>readonly</tt>) come immediately after the argument list.</p>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000992
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000993<p>Currently, only the following parameter attributes are defined:</p>
Chris Lattner5cee13f2008-01-11 06:20:47 +0000994
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000995<dl>
Bill Wendling7f4a3362009-11-02 00:24:16 +0000996 <dt><tt><b>zeroext</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000997 <dd>This indicates to the code generator that the parameter or return value
998 should be zero-extended to a 32-bit value by the caller (for a parameter)
999 or the callee (for a return value).</dd>
Chris Lattner5cee13f2008-01-11 06:20:47 +00001000
Bill Wendling7f4a3362009-11-02 00:24:16 +00001001 <dt><tt><b>signext</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001002 <dd>This indicates to the code generator that the parameter or return value
1003 should be sign-extended to a 32-bit value by the caller (for a parameter)
1004 or the callee (for a return value).</dd>
Chris Lattner5cee13f2008-01-11 06:20:47 +00001005
Bill Wendling7f4a3362009-11-02 00:24:16 +00001006 <dt><tt><b>inreg</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001007 <dd>This indicates that this parameter or return value should be treated in a
1008 special target-dependent fashion during while emitting code for a function
1009 call or return (usually, by putting it in a register as opposed to memory,
1010 though some targets use it to distinguish between two different kinds of
1011 registers). Use of this attribute is target-specific.</dd>
Chris Lattner5cee13f2008-01-11 06:20:47 +00001012
Bill Wendling7f4a3362009-11-02 00:24:16 +00001013 <dt><tt><b><a name="byval">byval</a></b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001014 <dd>This indicates that the pointer parameter should really be passed by value
1015 to the function. The attribute implies that a hidden copy of the pointee
1016 is made between the caller and the callee, so the callee is unable to
1017 modify the value in the callee. This attribute is only valid on LLVM
1018 pointer arguments. It is generally used to pass structs and arrays by
1019 value, but is also valid on pointers to scalars. The copy is considered
1020 to belong to the caller not the callee (for example,
1021 <tt><a href="#readonly">readonly</a></tt> functions should not write to
1022 <tt>byval</tt> parameters). This is not a valid attribute for return
1023 values. The byval attribute also supports specifying an alignment with
1024 the align attribute. This has a target-specific effect on the code
1025 generator that usually indicates a desired alignment for the synthesized
1026 stack slot.</dd>
1027
Dan Gohman3770af52010-07-02 23:18:08 +00001028 <dt><tt><b><a name="sret">sret</a></b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001029 <dd>This indicates that the pointer parameter specifies the address of a
1030 structure that is the return value of the function in the source program.
1031 This pointer must be guaranteed by the caller to be valid: loads and
1032 stores to the structure may be assumed by the callee to not to trap. This
1033 may only be applied to the first parameter. This is not a valid attribute
1034 for return values. </dd>
1035
Dan Gohman3770af52010-07-02 23:18:08 +00001036 <dt><tt><b><a name="noalias">noalias</a></b></tt></dt>
Dan Gohmandf12d082010-07-02 18:41:32 +00001037 <dd>This indicates that pointer values
1038 <a href="#pointeraliasing"><i>based</i></a> on the argument or return
Dan Gohmande256292010-07-02 23:46:54 +00001039 value do not alias pointer values which are not <i>based</i> on it,
1040 ignoring certain "irrelevant" dependencies.
1041 For a call to the parent function, dependencies between memory
1042 references from before or after the call and from those during the call
1043 are "irrelevant" to the <tt>noalias</tt> keyword for the arguments and
1044 return value used in that call.
Dan Gohmandf12d082010-07-02 18:41:32 +00001045 The caller shares the responsibility with the callee for ensuring that
1046 these requirements are met.
1047 For further details, please see the discussion of the NoAlias response in
Dan Gohman6c858db2010-07-06 15:26:33 +00001048 <a href="AliasAnalysis.html#MustMayNo">alias analysis</a>.<br>
1049<br>
John McCall72ed8902010-07-06 21:07:14 +00001050 Note that this definition of <tt>noalias</tt> is intentionally
1051 similar to the definition of <tt>restrict</tt> in C99 for function
Chris Lattner5eff9ca2010-07-06 20:51:35 +00001052 arguments, though it is slightly weaker.
Dan Gohman6c858db2010-07-06 15:26:33 +00001053<br>
1054 For function return values, C99's <tt>restrict</tt> is not meaningful,
1055 while LLVM's <tt>noalias</tt> is.
1056 </dd>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001057
Dan Gohman3770af52010-07-02 23:18:08 +00001058 <dt><tt><b><a name="nocapture">nocapture</a></b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001059 <dd>This indicates that the callee does not make any copies of the pointer
1060 that outlive the callee itself. This is not a valid attribute for return
1061 values.</dd>
1062
Dan Gohman3770af52010-07-02 23:18:08 +00001063 <dt><tt><b><a name="nest">nest</a></b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001064 <dd>This indicates that the pointer parameter can be excised using the
1065 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
1066 attribute for return values.</dd>
1067</dl>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001068
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001069</div>
1070
1071<!-- ======================================================================= -->
Chris Lattner91c15c42006-01-23 23:23:47 +00001072<div class="doc_subsection">
Gordon Henriksen71183b62007-12-10 03:18:06 +00001073 <a name="gc">Garbage Collector Names</a>
1074</div>
1075
1076<div class="doc_text">
Gordon Henriksen71183b62007-12-10 03:18:06 +00001077
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001078<p>Each function may specify a garbage collector name, which is simply a
1079 string:</p>
1080
Benjamin Kramer79698be2010-07-13 12:26:09 +00001081<pre class="doc_code">
Bill Wendling7f4a3362009-11-02 00:24:16 +00001082define void @f() gc "name" { ... }
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001083</pre>
Gordon Henriksen71183b62007-12-10 03:18:06 +00001084
1085<p>The compiler declares the supported values of <i>name</i>. Specifying a
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001086 collector which will cause the compiler to alter its output in order to
1087 support the named garbage collection algorithm.</p>
1088
Gordon Henriksen71183b62007-12-10 03:18:06 +00001089</div>
1090
1091<!-- ======================================================================= -->
1092<div class="doc_subsection">
Devang Patel9eb525d2008-09-26 23:51:19 +00001093 <a name="fnattrs">Function Attributes</a>
Devang Patelcaacdba2008-09-04 23:05:13 +00001094</div>
1095
1096<div class="doc_text">
Devang Patel9eb525d2008-09-26 23:51:19 +00001097
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001098<p>Function attributes are set to communicate additional information about a
1099 function. Function attributes are considered to be part of the function, not
1100 of the function type, so functions with different parameter attributes can
1101 have the same function type.</p>
Devang Patel9eb525d2008-09-26 23:51:19 +00001102
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001103<p>Function attributes are simple keywords that follow the type specified. If
1104 multiple attributes are needed, they are space separated. For example:</p>
Devang Patelcaacdba2008-09-04 23:05:13 +00001105
Benjamin Kramer79698be2010-07-13 12:26:09 +00001106<pre class="doc_code">
Devang Patel9eb525d2008-09-26 23:51:19 +00001107define void @f() noinline { ... }
1108define void @f() alwaysinline { ... }
1109define void @f() alwaysinline optsize { ... }
Bill Wendling7f4a3362009-11-02 00:24:16 +00001110define void @f() optsize { ... }
Bill Wendlingb175fa42008-09-07 10:26:33 +00001111</pre>
Devang Patelcaacdba2008-09-04 23:05:13 +00001112
Bill Wendlingb175fa42008-09-07 10:26:33 +00001113<dl>
Charles Davisbe5557e2010-02-12 00:31:15 +00001114 <dt><tt><b>alignstack(&lt;<em>n</em>&gt;)</b></tt></dt>
1115 <dd>This attribute indicates that, when emitting the prologue and epilogue,
1116 the backend should forcibly align the stack pointer. Specify the
1117 desired alignment, which must be a power of two, in parentheses.
1118
Bill Wendling7f4a3362009-11-02 00:24:16 +00001119 <dt><tt><b>alwaysinline</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001120 <dd>This attribute indicates that the inliner should attempt to inline this
1121 function into callers whenever possible, ignoring any active inlining size
1122 threshold for this caller.</dd>
Bill Wendlingb175fa42008-09-07 10:26:33 +00001123
Jakob Stoklund Olesen74bb06c2010-02-06 01:16:28 +00001124 <dt><tt><b>inlinehint</b></tt></dt>
1125 <dd>This attribute indicates that the source code contained a hint that inlining
1126 this function is desirable (such as the "inline" keyword in C/C++). It
1127 is just a hint; it imposes no requirements on the inliner.</dd>
1128
Nick Lewycky14b58da2010-07-06 18:24:09 +00001129 <dt><tt><b>naked</b></tt></dt>
1130 <dd>This attribute disables prologue / epilogue emission for the function.
1131 This can have very system-specific consequences.</dd>
1132
1133 <dt><tt><b>noimplicitfloat</b></tt></dt>
1134 <dd>This attributes disables implicit floating point instructions.</dd>
1135
Bill Wendling7f4a3362009-11-02 00:24:16 +00001136 <dt><tt><b>noinline</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001137 <dd>This attribute indicates that the inliner should never inline this
1138 function in any situation. This attribute may not be used together with
1139 the <tt>alwaysinline</tt> attribute.</dd>
Devang Patel9eb525d2008-09-26 23:51:19 +00001140
Nick Lewycky14b58da2010-07-06 18:24:09 +00001141 <dt><tt><b>noredzone</b></tt></dt>
1142 <dd>This attribute indicates that the code generator should not use a red
1143 zone, even if the target-specific ABI normally permits it.</dd>
Devang Patel9eb525d2008-09-26 23:51:19 +00001144
Bill Wendling7f4a3362009-11-02 00:24:16 +00001145 <dt><tt><b>noreturn</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001146 <dd>This function attribute indicates that the function never returns
1147 normally. This produces undefined behavior at runtime if the function
1148 ever does dynamically return.</dd>
Bill Wendlinga8130172008-11-13 01:02:51 +00001149
Bill Wendling7f4a3362009-11-02 00:24:16 +00001150 <dt><tt><b>nounwind</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001151 <dd>This function attribute indicates that the function never returns with an
1152 unwind or exceptional control flow. If the function does unwind, its
1153 runtime behavior is undefined.</dd>
Bill Wendling0f5541e2008-11-26 19:07:40 +00001154
Nick Lewycky14b58da2010-07-06 18:24:09 +00001155 <dt><tt><b>optsize</b></tt></dt>
1156 <dd>This attribute suggests that optimization passes and code generator passes
1157 make choices that keep the code size of this function low, and otherwise
1158 do optimizations specifically to reduce code size.</dd>
1159
Bill Wendling7f4a3362009-11-02 00:24:16 +00001160 <dt><tt><b>readnone</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001161 <dd>This attribute indicates that the function computes its result (or decides
1162 to unwind an exception) based strictly on its arguments, without
1163 dereferencing any pointer arguments or otherwise accessing any mutable
1164 state (e.g. memory, control registers, etc) visible to caller functions.
1165 It does not write through any pointer arguments
1166 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1167 changes any state visible to callers. This means that it cannot unwind
1168 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1169 could use the <tt>unwind</tt> instruction.</dd>
Devang Patel310fd4a2009-06-12 19:45:19 +00001170
Bill Wendling7f4a3362009-11-02 00:24:16 +00001171 <dt><tt><b><a name="readonly">readonly</a></b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001172 <dd>This attribute indicates that the function does not write through any
1173 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1174 arguments) or otherwise modify any state (e.g. memory, control registers,
1175 etc) visible to caller functions. It may dereference pointer arguments
1176 and read state that may be set in the caller. A readonly function always
1177 returns the same value (or unwinds an exception identically) when called
1178 with the same set of arguments and global state. It cannot unwind an
1179 exception by calling the <tt>C++</tt> exception throwing methods, but may
1180 use the <tt>unwind</tt> instruction.</dd>
Anton Korobeynikovc8ce7b082009-07-17 18:07:26 +00001181
Bill Wendling7f4a3362009-11-02 00:24:16 +00001182 <dt><tt><b><a name="ssp">ssp</a></b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001183 <dd>This attribute indicates that the function should emit a stack smashing
1184 protector. It is in the form of a "canary"&mdash;a random value placed on
1185 the stack before the local variables that's checked upon return from the
1186 function to see if it has been overwritten. A heuristic is used to
1187 determine if a function needs stack protectors or not.<br>
1188<br>
1189 If a function that has an <tt>ssp</tt> attribute is inlined into a
1190 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1191 function will have an <tt>ssp</tt> attribute.</dd>
1192
Bill Wendling7f4a3362009-11-02 00:24:16 +00001193 <dt><tt><b>sspreq</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001194 <dd>This attribute indicates that the function should <em>always</em> emit a
1195 stack smashing protector. This overrides
Bill Wendling30235112009-07-20 02:39:26 +00001196 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1197<br>
1198 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1199 function that doesn't have an <tt>sspreq</tt> attribute or which has
1200 an <tt>ssp</tt> attribute, then the resulting function will have
1201 an <tt>sspreq</tt> attribute.</dd>
Bill Wendlingb175fa42008-09-07 10:26:33 +00001202</dl>
1203
Devang Patelcaacdba2008-09-04 23:05:13 +00001204</div>
1205
1206<!-- ======================================================================= -->
1207<div class="doc_subsection">
Chris Lattner93564892006-04-08 04:40:53 +00001208 <a name="moduleasm">Module-Level Inline Assembly</a>
Chris Lattner91c15c42006-01-23 23:23:47 +00001209</div>
1210
1211<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001212
1213<p>Modules may contain "module-level inline asm" blocks, which corresponds to
1214 the GCC "file scope inline asm" blocks. These blocks are internally
1215 concatenated by LLVM and treated as a single unit, but may be separated in
1216 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
Chris Lattner91c15c42006-01-23 23:23:47 +00001217
Benjamin Kramer79698be2010-07-13 12:26:09 +00001218<pre class="doc_code">
Bill Wendling3716c5d2007-05-29 09:04:49 +00001219module asm "inline asm code goes here"
1220module asm "more can go here"
1221</pre>
Chris Lattner91c15c42006-01-23 23:23:47 +00001222
1223<p>The strings can contain any character by escaping non-printable characters.
1224 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001225 for the number.</p>
Chris Lattner91c15c42006-01-23 23:23:47 +00001226
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001227<p>The inline asm code is simply printed to the machine code .s file when
1228 assembly code is generated.</p>
1229
Chris Lattner91c15c42006-01-23 23:23:47 +00001230</div>
Chris Lattner6af02f32004-12-09 16:11:40 +00001231
Reid Spencer50c723a2007-02-19 23:54:10 +00001232<!-- ======================================================================= -->
1233<div class="doc_subsection">
1234 <a name="datalayout">Data Layout</a>
1235</div>
1236
1237<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001238
Reid Spencer50c723a2007-02-19 23:54:10 +00001239<p>A module may specify a target specific data layout string that specifies how
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001240 data is to be laid out in memory. The syntax for the data layout is
1241 simply:</p>
1242
Benjamin Kramer79698be2010-07-13 12:26:09 +00001243<pre class="doc_code">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001244target datalayout = "<i>layout specification</i>"
1245</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001246
1247<p>The <i>layout specification</i> consists of a list of specifications
1248 separated by the minus sign character ('-'). Each specification starts with
1249 a letter and may include other information after the letter to define some
1250 aspect of the data layout. The specifications accepted are as follows:</p>
1251
Reid Spencer50c723a2007-02-19 23:54:10 +00001252<dl>
1253 <dt><tt>E</tt></dt>
1254 <dd>Specifies that the target lays out data in big-endian form. That is, the
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001255 bits with the most significance have the lowest address location.</dd>
1256
Reid Spencer50c723a2007-02-19 23:54:10 +00001257 <dt><tt>e</tt></dt>
Chris Lattner67c37d12008-08-05 18:29:16 +00001258 <dd>Specifies that the target lays out data in little-endian form. That is,
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001259 the bits with the least significance have the lowest address
1260 location.</dd>
1261
Reid Spencer50c723a2007-02-19 23:54:10 +00001262 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
Eric Christopher455c5772009-12-05 02:46:03 +00001263 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001264 <i>preferred</i> alignments. All sizes are in bits. Specifying
1265 the <i>pref</i> alignment is optional. If omitted, the
1266 preceding <tt>:</tt> should be omitted too.</dd>
1267
Reid Spencer50c723a2007-02-19 23:54:10 +00001268 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1269 <dd>This specifies the alignment for an integer type of a given bit
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001270 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1271
Reid Spencer50c723a2007-02-19 23:54:10 +00001272 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
Eric Christopher455c5772009-12-05 02:46:03 +00001273 <dd>This specifies the alignment for a vector type of a given bit
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001274 <i>size</i>.</dd>
1275
Reid Spencer50c723a2007-02-19 23:54:10 +00001276 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
Eric Christopher455c5772009-12-05 02:46:03 +00001277 <dd>This specifies the alignment for a floating point type of a given bit
Dale Johannesence522852010-05-28 18:54:47 +00001278 <i>size</i>. Only values of <i>size</i> that are supported by the target
1279 will work. 32 (float) and 64 (double) are supported on all targets;
1280 80 or 128 (different flavors of long double) are also supported on some
1281 targets.
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001282
Reid Spencer50c723a2007-02-19 23:54:10 +00001283 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1284 <dd>This specifies the alignment for an aggregate type of a given bit
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001285 <i>size</i>.</dd>
1286
Daniel Dunbar7921a592009-06-08 22:17:53 +00001287 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1288 <dd>This specifies the alignment for a stack object of a given bit
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001289 <i>size</i>.</dd>
Chris Lattnera381eff2009-11-07 09:35:34 +00001290
1291 <dt><tt>n<i>size1</i>:<i>size2</i>:<i>size3</i>...</tt></dt>
1292 <dd>This specifies a set of native integer widths for the target CPU
1293 in bits. For example, it might contain "n32" for 32-bit PowerPC,
1294 "n32:64" for PowerPC 64, or "n8:16:32:64" for X86-64. Elements of
Eric Christopher455c5772009-12-05 02:46:03 +00001295 this set are considered to support most general arithmetic
Chris Lattnera381eff2009-11-07 09:35:34 +00001296 operations efficiently.</dd>
Reid Spencer50c723a2007-02-19 23:54:10 +00001297</dl>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001298
Reid Spencer50c723a2007-02-19 23:54:10 +00001299<p>When constructing the data layout for a given target, LLVM starts with a
Dan Gohman61110ae2010-04-28 00:36:01 +00001300 default set of specifications which are then (possibly) overridden by the
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001301 specifications in the <tt>datalayout</tt> keyword. The default specifications
1302 are given in this list:</p>
1303
Reid Spencer50c723a2007-02-19 23:54:10 +00001304<ul>
1305 <li><tt>E</tt> - big endian</li>
Dan Gohman8ad777d2010-02-23 02:44:03 +00001306 <li><tt>p:64:64:64</tt> - 64-bit pointers with 64-bit alignment</li>
Reid Spencer50c723a2007-02-19 23:54:10 +00001307 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1308 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1309 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1310 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
Chris Lattner67c37d12008-08-05 18:29:16 +00001311 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
Reid Spencer50c723a2007-02-19 23:54:10 +00001312 alignment of 64-bits</li>
1313 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1314 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1315 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1316 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1317 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
Daniel Dunbar7921a592009-06-08 22:17:53 +00001318 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
Reid Spencer50c723a2007-02-19 23:54:10 +00001319</ul>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001320
1321<p>When LLVM is determining the alignment for a given type, it uses the
1322 following rules:</p>
1323
Reid Spencer50c723a2007-02-19 23:54:10 +00001324<ol>
1325 <li>If the type sought is an exact match for one of the specifications, that
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001326 specification is used.</li>
1327
Reid Spencer50c723a2007-02-19 23:54:10 +00001328 <li>If no match is found, and the type sought is an integer type, then the
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001329 smallest integer type that is larger than the bitwidth of the sought type
1330 is used. If none of the specifications are larger than the bitwidth then
1331 the the largest integer type is used. For example, given the default
1332 specifications above, the i7 type will use the alignment of i8 (next
1333 largest) while both i65 and i256 will use the alignment of i64 (largest
1334 specified).</li>
1335
Reid Spencer50c723a2007-02-19 23:54:10 +00001336 <li>If no match is found, and the type sought is a vector type, then the
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001337 largest vector type that is smaller than the sought vector type will be
1338 used as a fall back. This happens because &lt;128 x double&gt; can be
1339 implemented in terms of 64 &lt;2 x double&gt;, for example.</li>
Reid Spencer50c723a2007-02-19 23:54:10 +00001340</ol>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001341
Reid Spencer50c723a2007-02-19 23:54:10 +00001342</div>
Chris Lattner6af02f32004-12-09 16:11:40 +00001343
Dan Gohman6154a012009-07-27 18:07:55 +00001344<!-- ======================================================================= -->
1345<div class="doc_subsection">
1346 <a name="pointeraliasing">Pointer Aliasing Rules</a>
1347</div>
1348
1349<div class="doc_text">
1350
Andreas Bolka8ae4e242009-07-29 00:02:05 +00001351<p>Any memory access must be done through a pointer value associated
Andreas Bolkae39f0332009-07-27 20:37:10 +00001352with an address range of the memory access, otherwise the behavior
Dan Gohman6154a012009-07-27 18:07:55 +00001353is undefined. Pointer values are associated with address ranges
1354according to the following rules:</p>
1355
1356<ul>
Dan Gohmandf12d082010-07-02 18:41:32 +00001357 <li>A pointer value is associated with the addresses associated with
1358 any value it is <i>based</i> on.
Andreas Bolka8ae4e242009-07-29 00:02:05 +00001359 <li>An address of a global variable is associated with the address
Dan Gohman6154a012009-07-27 18:07:55 +00001360 range of the variable's storage.</li>
1361 <li>The result value of an allocation instruction is associated with
1362 the address range of the allocated storage.</li>
1363 <li>A null pointer in the default address-space is associated with
Andreas Bolka8ae4e242009-07-29 00:02:05 +00001364 no address.</li>
Dan Gohman6154a012009-07-27 18:07:55 +00001365 <li>An integer constant other than zero or a pointer value returned
1366 from a function not defined within LLVM may be associated with address
1367 ranges allocated through mechanisms other than those provided by
Andreas Bolka8ae4e242009-07-29 00:02:05 +00001368 LLVM. Such ranges shall not overlap with any ranges of addresses
Dan Gohman6154a012009-07-27 18:07:55 +00001369 allocated by mechanisms provided by LLVM.</li>
Dan Gohmandf12d082010-07-02 18:41:32 +00001370</ul>
1371
1372<p>A pointer value is <i>based</i> on another pointer value according
1373 to the following rules:</p>
1374
1375<ul>
1376 <li>A pointer value formed from a
1377 <tt><a href="#i_getelementptr">getelementptr</a></tt> operation
1378 is <i>based</i> on the first operand of the <tt>getelementptr</tt>.</li>
1379 <li>The result value of a
1380 <tt><a href="#i_bitcast">bitcast</a></tt> is <i>based</i> on the operand
1381 of the <tt>bitcast</tt>.</li>
1382 <li>A pointer value formed by an
1383 <tt><a href="#i_inttoptr">inttoptr</a></tt> is <i>based</i> on all
1384 pointer values that contribute (directly or indirectly) to the
1385 computation of the pointer's value.</li>
1386 <li>The "<i>based</i> on" relationship is transitive.</li>
1387</ul>
1388
1389<p>Note that this definition of <i>"based"</i> is intentionally
1390 similar to the definition of <i>"based"</i> in C99, though it is
1391 slightly weaker.</p>
Dan Gohman6154a012009-07-27 18:07:55 +00001392
1393<p>LLVM IR does not associate types with memory. The result type of a
Andreas Bolka8ae4e242009-07-29 00:02:05 +00001394<tt><a href="#i_load">load</a></tt> merely indicates the size and
1395alignment of the memory from which to load, as well as the
Dan Gohman4eb47192010-06-17 19:23:50 +00001396interpretation of the value. The first operand type of a
Andreas Bolka8ae4e242009-07-29 00:02:05 +00001397<tt><a href="#i_store">store</a></tt> similarly only indicates the size
1398and alignment of the store.</p>
Dan Gohman6154a012009-07-27 18:07:55 +00001399
1400<p>Consequently, type-based alias analysis, aka TBAA, aka
1401<tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1402LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1403additional information which specialized optimization passes may use
1404to implement type-based alias analysis.</p>
1405
1406</div>
1407
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +00001408<!-- ======================================================================= -->
1409<div class="doc_subsection">
1410 <a name="volatile">Volatile Memory Accesses</a>
1411</div>
1412
1413<div class="doc_text">
1414
1415<p>Certain memory accesses, such as <a href="#i_load"><tt>load</tt></a>s, <a
1416href="#i_store"><tt>store</tt></a>s, and <a
1417href="#int_memcpy"><tt>llvm.memcpy</tt></a>s may be marked <tt>volatile</tt>.
1418The optimizers must not change the number of volatile operations or change their
1419order of execution relative to other volatile operations. The optimizers
1420<i>may</i> change the order of volatile operations relative to non-volatile
1421operations. This is not Java's "volatile" and has no cross-thread
1422synchronization behavior.</p>
1423
1424</div>
1425
Chris Lattner2f7c9632001-06-06 20:29:01 +00001426<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001427<div class="doc_section"> <a name="typesystem">Type System</a> </div>
1428<!-- *********************************************************************** -->
Chris Lattner6af02f32004-12-09 16:11:40 +00001429
Misha Brukman76307852003-11-08 01:05:38 +00001430<div class="doc_text">
Chris Lattner6af02f32004-12-09 16:11:40 +00001431
Misha Brukman76307852003-11-08 01:05:38 +00001432<p>The LLVM type system is one of the most important features of the
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001433 intermediate representation. Being typed enables a number of optimizations
1434 to be performed on the intermediate representation directly, without having
1435 to do extra analyses on the side before the transformation. A strong type
1436 system makes it easier to read the generated code and enables novel analyses
1437 and transformations that are not feasible to perform on normal three address
1438 code representations.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +00001439
1440</div>
1441
Chris Lattner2f7c9632001-06-06 20:29:01 +00001442<!-- ======================================================================= -->
Chris Lattner7824d182008-01-04 04:32:38 +00001443<div class="doc_subsection"> <a name="t_classifications">Type
Chris Lattner48b383b02003-11-25 01:02:51 +00001444Classifications</a> </div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001445
Misha Brukman76307852003-11-08 01:05:38 +00001446<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001447
1448<p>The types fall into a few useful classifications:</p>
Misha Brukmanc501f552004-03-01 17:47:27 +00001449
1450<table border="1" cellspacing="0" cellpadding="4">
Chris Lattner48b383b02003-11-25 01:02:51 +00001451 <tbody>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001452 <tr><th>Classification</th><th>Types</th></tr>
Chris Lattner48b383b02003-11-25 01:02:51 +00001453 <tr>
Chris Lattner7824d182008-01-04 04:32:38 +00001454 <td><a href="#t_integer">integer</a></td>
Reid Spencer138249b2007-05-16 18:44:01 +00001455 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
Chris Lattner48b383b02003-11-25 01:02:51 +00001456 </tr>
1457 <tr>
Chris Lattner7824d182008-01-04 04:32:38 +00001458 <td><a href="#t_floating">floating point</a></td>
1459 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Chris Lattner48b383b02003-11-25 01:02:51 +00001460 </tr>
1461 <tr>
1462 <td><a name="t_firstclass">first class</a></td>
Chris Lattner7824d182008-01-04 04:32:38 +00001463 <td><a href="#t_integer">integer</a>,
1464 <a href="#t_floating">floating point</a>,
1465 <a href="#t_pointer">pointer</a>,
Dan Gohman08783a882008-06-18 18:42:13 +00001466 <a href="#t_vector">vector</a>,
Dan Gohmanb9d66602008-05-12 23:51:09 +00001467 <a href="#t_struct">structure</a>,
Chris Lattner392be582010-02-12 20:49:41 +00001468 <a href="#t_union">union</a>,
Dan Gohmanb9d66602008-05-12 23:51:09 +00001469 <a href="#t_array">array</a>,
Nick Lewyckyadbc2842009-05-30 05:06:04 +00001470 <a href="#t_label">label</a>,
1471 <a href="#t_metadata">metadata</a>.
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001472 </td>
Chris Lattner48b383b02003-11-25 01:02:51 +00001473 </tr>
Chris Lattner7824d182008-01-04 04:32:38 +00001474 <tr>
1475 <td><a href="#t_primitive">primitive</a></td>
1476 <td><a href="#t_label">label</a>,
1477 <a href="#t_void">void</a>,
Nick Lewyckyadbc2842009-05-30 05:06:04 +00001478 <a href="#t_floating">floating point</a>,
1479 <a href="#t_metadata">metadata</a>.</td>
Chris Lattner7824d182008-01-04 04:32:38 +00001480 </tr>
1481 <tr>
1482 <td><a href="#t_derived">derived</a></td>
Chris Lattner392be582010-02-12 20:49:41 +00001483 <td><a href="#t_array">array</a>,
Chris Lattner7824d182008-01-04 04:32:38 +00001484 <a href="#t_function">function</a>,
1485 <a href="#t_pointer">pointer</a>,
1486 <a href="#t_struct">structure</a>,
1487 <a href="#t_pstruct">packed structure</a>,
Chris Lattner392be582010-02-12 20:49:41 +00001488 <a href="#t_union">union</a>,
Chris Lattner7824d182008-01-04 04:32:38 +00001489 <a href="#t_vector">vector</a>,
1490 <a href="#t_opaque">opaque</a>.
Dan Gohman93bf60d2008-10-14 16:32:04 +00001491 </td>
Chris Lattner7824d182008-01-04 04:32:38 +00001492 </tr>
Chris Lattner48b383b02003-11-25 01:02:51 +00001493 </tbody>
Misha Brukman76307852003-11-08 01:05:38 +00001494</table>
Misha Brukmanc501f552004-03-01 17:47:27 +00001495
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001496<p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1497 important. Values of these types are the only ones which can be produced by
Nick Lewycky84a1eeb2009-09-27 00:45:11 +00001498 instructions.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001499
Misha Brukman76307852003-11-08 01:05:38 +00001500</div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001501
Chris Lattner2f7c9632001-06-06 20:29:01 +00001502<!-- ======================================================================= -->
Chris Lattner7824d182008-01-04 04:32:38 +00001503<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner43542b32008-01-04 04:34:14 +00001504
Chris Lattner7824d182008-01-04 04:32:38 +00001505<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001506
Chris Lattner7824d182008-01-04 04:32:38 +00001507<p>The primitive types are the fundamental building blocks of the LLVM
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001508 system.</p>
Chris Lattner7824d182008-01-04 04:32:38 +00001509
Chris Lattner43542b32008-01-04 04:34:14 +00001510</div>
1511
Chris Lattner7824d182008-01-04 04:32:38 +00001512<!-- _______________________________________________________________________ -->
Nick Lewycky84a1eeb2009-09-27 00:45:11 +00001513<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1514
1515<div class="doc_text">
1516
1517<h5>Overview:</h5>
1518<p>The integer type is a very simple type that simply specifies an arbitrary
1519 bit width for the integer type desired. Any bit width from 1 bit to
1520 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
1521
1522<h5>Syntax:</h5>
1523<pre>
1524 iN
1525</pre>
1526
1527<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1528 value.</p>
1529
1530<h5>Examples:</h5>
1531<table class="layout">
1532 <tr class="layout">
1533 <td class="left"><tt>i1</tt></td>
1534 <td class="left">a single-bit integer.</td>
1535 </tr>
1536 <tr class="layout">
1537 <td class="left"><tt>i32</tt></td>
1538 <td class="left">a 32-bit integer.</td>
1539 </tr>
1540 <tr class="layout">
1541 <td class="left"><tt>i1942652</tt></td>
1542 <td class="left">a really big integer of over 1 million bits.</td>
1543 </tr>
1544</table>
1545
Nick Lewycky84a1eeb2009-09-27 00:45:11 +00001546</div>
1547
1548<!-- _______________________________________________________________________ -->
Chris Lattner7824d182008-01-04 04:32:38 +00001549<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1550
1551<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001552
1553<table>
1554 <tbody>
1555 <tr><th>Type</th><th>Description</th></tr>
1556 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1557 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1558 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1559 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1560 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1561 </tbody>
1562</table>
1563
Chris Lattner7824d182008-01-04 04:32:38 +00001564</div>
1565
1566<!-- _______________________________________________________________________ -->
1567<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1568
1569<div class="doc_text">
Bill Wendling30235112009-07-20 02:39:26 +00001570
Chris Lattner7824d182008-01-04 04:32:38 +00001571<h5>Overview:</h5>
1572<p>The void type does not represent any value and has no size.</p>
1573
1574<h5>Syntax:</h5>
Chris Lattner7824d182008-01-04 04:32:38 +00001575<pre>
1576 void
1577</pre>
Bill Wendling30235112009-07-20 02:39:26 +00001578
Chris Lattner7824d182008-01-04 04:32:38 +00001579</div>
1580
1581<!-- _______________________________________________________________________ -->
1582<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1583
1584<div class="doc_text">
Bill Wendling30235112009-07-20 02:39:26 +00001585
Chris Lattner7824d182008-01-04 04:32:38 +00001586<h5>Overview:</h5>
1587<p>The label type represents code labels.</p>
1588
1589<h5>Syntax:</h5>
Chris Lattner7824d182008-01-04 04:32:38 +00001590<pre>
1591 label
1592</pre>
Bill Wendling30235112009-07-20 02:39:26 +00001593
Chris Lattner7824d182008-01-04 04:32:38 +00001594</div>
1595
Nick Lewyckyadbc2842009-05-30 05:06:04 +00001596<!-- _______________________________________________________________________ -->
1597<div class="doc_subsubsection"> <a name="t_metadata">Metadata Type</a> </div>
1598
1599<div class="doc_text">
Bill Wendling30235112009-07-20 02:39:26 +00001600
Nick Lewyckyadbc2842009-05-30 05:06:04 +00001601<h5>Overview:</h5>
Nick Lewycky93e06a52009-09-27 23:27:42 +00001602<p>The metadata type represents embedded metadata. No derived types may be
1603 created from metadata except for <a href="#t_function">function</a>
1604 arguments.
Nick Lewyckyadbc2842009-05-30 05:06:04 +00001605
1606<h5>Syntax:</h5>
Nick Lewyckyadbc2842009-05-30 05:06:04 +00001607<pre>
1608 metadata
1609</pre>
Bill Wendling30235112009-07-20 02:39:26 +00001610
Nick Lewyckyadbc2842009-05-30 05:06:04 +00001611</div>
1612
Chris Lattner7824d182008-01-04 04:32:38 +00001613
1614<!-- ======================================================================= -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001615<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001616
Misha Brukman76307852003-11-08 01:05:38 +00001617<div class="doc_text">
Chris Lattner74d3f822004-12-09 17:30:23 +00001618
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001619<p>The real power in LLVM comes from the derived types in the system. This is
1620 what allows a programmer to represent arrays, functions, pointers, and other
Nick Lewycky84a1eeb2009-09-27 00:45:11 +00001621 useful types. Each of these types contain one or more element types which
1622 may be a primitive type, or another derived type. For example, it is
1623 possible to have a two dimensional array, using an array as the element type
1624 of another array.</p>
Dan Gohman142ccc02009-01-24 15:58:40 +00001625
Chris Lattner392be582010-02-12 20:49:41 +00001626
1627</div>
1628
1629<!-- _______________________________________________________________________ -->
1630<div class="doc_subsubsection"> <a name="t_aggregate">Aggregate Types</a> </div>
1631
1632<div class="doc_text">
1633
1634<p>Aggregate Types are a subset of derived types that can contain multiple
1635 member types. <a href="#t_array">Arrays</a>,
1636 <a href="#t_struct">structs</a>, <a href="#t_vector">vectors</a> and
1637 <a href="#t_union">unions</a> are aggregate types.</p>
1638
1639</div>
1640
Reid Spencer138249b2007-05-16 18:44:01 +00001641<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001642<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001643
Misha Brukman76307852003-11-08 01:05:38 +00001644<div class="doc_text">
Chris Lattner74d3f822004-12-09 17:30:23 +00001645
Chris Lattner2f7c9632001-06-06 20:29:01 +00001646<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00001647<p>The array type is a very simple derived type that arranges elements
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001648 sequentially in memory. The array type requires a size (number of elements)
1649 and an underlying data type.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001650
Chris Lattner590645f2002-04-14 06:13:44 +00001651<h5>Syntax:</h5>
Chris Lattner74d3f822004-12-09 17:30:23 +00001652<pre>
1653 [&lt;# elements&gt; x &lt;elementtype&gt;]
1654</pre>
1655
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001656<p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1657 be any type with a size.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001658
Chris Lattner590645f2002-04-14 06:13:44 +00001659<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001660<table class="layout">
1661 <tr class="layout">
Chris Lattner747359f2007-12-19 05:04:11 +00001662 <td class="left"><tt>[40 x i32]</tt></td>
1663 <td class="left">Array of 40 32-bit integer values.</td>
1664 </tr>
1665 <tr class="layout">
1666 <td class="left"><tt>[41 x i32]</tt></td>
1667 <td class="left">Array of 41 32-bit integer values.</td>
1668 </tr>
1669 <tr class="layout">
1670 <td class="left"><tt>[4 x i8]</tt></td>
1671 <td class="left">Array of 4 8-bit integer values.</td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001672 </tr>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001673</table>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001674<p>Here are some examples of multidimensional arrays:</p>
1675<table class="layout">
1676 <tr class="layout">
Chris Lattner747359f2007-12-19 05:04:11 +00001677 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1678 <td class="left">3x4 array of 32-bit integer values.</td>
1679 </tr>
1680 <tr class="layout">
1681 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1682 <td class="left">12x10 array of single precision floating point values.</td>
1683 </tr>
1684 <tr class="layout">
1685 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1686 <td class="left">2x3x4 array of 16-bit integer values.</td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001687 </tr>
1688</table>
Chris Lattnerc0ad71e2005-06-24 17:22:57 +00001689
Dan Gohmanc74bc282009-11-09 19:01:53 +00001690<p>There is no restriction on indexing beyond the end of the array implied by
1691 a static type (though there are restrictions on indexing beyond the bounds
1692 of an allocated object in some cases). This means that single-dimension
1693 'variable sized array' addressing can be implemented in LLVM with a zero
1694 length array type. An implementation of 'pascal style arrays' in LLVM could
1695 use the type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
Chris Lattnerc0ad71e2005-06-24 17:22:57 +00001696
Misha Brukman76307852003-11-08 01:05:38 +00001697</div>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001698
Chris Lattner2f7c9632001-06-06 20:29:01 +00001699<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001700<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001701
Misha Brukman76307852003-11-08 01:05:38 +00001702<div class="doc_text">
Chris Lattnerda508ac2008-04-23 04:59:35 +00001703
Chris Lattner2f7c9632001-06-06 20:29:01 +00001704<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001705<p>The function type can be thought of as a function signature. It consists of
1706 a return type and a list of formal parameter types. The return type of a
Chris Lattner392be582010-02-12 20:49:41 +00001707 function type is a scalar type, a void type, a struct type, or a union
1708 type. If the return type is a struct type then all struct elements must be
1709 of first class types, and the struct must have at least one element.</p>
Devang Pateld6cff512008-03-10 20:49:15 +00001710
Chris Lattner2f7c9632001-06-06 20:29:01 +00001711<h5>Syntax:</h5>
Chris Lattnerda508ac2008-04-23 04:59:35 +00001712<pre>
Nick Lewycky14d1ccc2009-09-27 07:55:32 +00001713 &lt;returntype&gt; (&lt;parameter list&gt;)
Chris Lattnerda508ac2008-04-23 04:59:35 +00001714</pre>
1715
John Criswell4c0cf7f2005-10-24 16:17:18 +00001716<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001717 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1718 which indicates that the function takes a variable number of arguments.
1719 Variable argument functions can access their arguments with
1720 the <a href="#int_varargs">variable argument handling intrinsic</a>
Chris Lattner47f2a832010-03-02 06:36:51 +00001721 functions. '<tt>&lt;returntype&gt;</tt>' is any type except
Nick Lewycky93e06a52009-09-27 23:27:42 +00001722 <a href="#t_label">label</a>.</p>
Chris Lattnerda508ac2008-04-23 04:59:35 +00001723
Chris Lattner2f7c9632001-06-06 20:29:01 +00001724<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001725<table class="layout">
1726 <tr class="layout">
Reid Spencer58c08712006-12-31 07:18:34 +00001727 <td class="left"><tt>i32 (i32)</tt></td>
1728 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001729 </td>
Reid Spencer58c08712006-12-31 07:18:34 +00001730 </tr><tr class="layout">
Chris Lattner47f2a832010-03-02 06:36:51 +00001731 <td class="left"><tt>float&nbsp;(i16,&nbsp;i32&nbsp;*)&nbsp;*
Reid Spencer655dcc62006-12-31 07:20:23 +00001732 </tt></td>
Eric Christopher455c5772009-12-05 02:46:03 +00001733 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
Chris Lattner47f2a832010-03-02 06:36:51 +00001734 an <tt>i16</tt> and a <a href="#t_pointer">pointer</a> to <tt>i32</tt>,
1735 returning <tt>float</tt>.
Reid Spencer58c08712006-12-31 07:18:34 +00001736 </td>
1737 </tr><tr class="layout">
1738 <td class="left"><tt>i32 (i8*, ...)</tt></td>
Eric Christopher455c5772009-12-05 02:46:03 +00001739 <td class="left">A vararg function that takes at least one
1740 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1741 which returns an integer. This is the signature for <tt>printf</tt> in
Reid Spencer58c08712006-12-31 07:18:34 +00001742 LLVM.
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001743 </td>
Devang Patele3dfc1c2008-03-24 05:35:41 +00001744 </tr><tr class="layout">
1745 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Nick Lewycky14d1ccc2009-09-27 07:55:32 +00001746 <td class="left">A function taking an <tt>i32</tt>, returning a
1747 <a href="#t_struct">structure</a> containing two <tt>i32</tt> values
Devang Patele3dfc1c2008-03-24 05:35:41 +00001748 </td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001749 </tr>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001750</table>
Misha Brukmanc501f552004-03-01 17:47:27 +00001751
Misha Brukman76307852003-11-08 01:05:38 +00001752</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001753
Chris Lattner2f7c9632001-06-06 20:29:01 +00001754<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001755<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001756
Misha Brukman76307852003-11-08 01:05:38 +00001757<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001758
Chris Lattner2f7c9632001-06-06 20:29:01 +00001759<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001760<p>The structure type is used to represent a collection of data members together
1761 in memory. The packing of the field types is defined to match the ABI of the
1762 underlying processor. The elements of a structure may be any type that has a
1763 size.</p>
1764
Jeffrey Yasskinf991bbb2010-01-11 19:19:26 +00001765<p>Structures in memory are accessed using '<tt><a href="#i_load">load</a></tt>'
1766 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field
1767 with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
1768 Structures in registers are accessed using the
1769 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' and
1770 '<tt><a href="#i_insertvalue">insertvalue</a></tt>' instructions.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001771<h5>Syntax:</h5>
Bill Wendling30235112009-07-20 02:39:26 +00001772<pre>
1773 { &lt;type list&gt; }
1774</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001775
Chris Lattner2f7c9632001-06-06 20:29:01 +00001776<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001777<table class="layout">
1778 <tr class="layout">
Jeff Cohen5819f182007-04-22 01:17:39 +00001779 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1780 <td class="left">A triple of three <tt>i32</tt> values</td>
1781 </tr><tr class="layout">
1782 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1783 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1784 second element is a <a href="#t_pointer">pointer</a> to a
1785 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1786 an <tt>i32</tt>.</td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001787 </tr>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001788</table>
Dan Gohman142ccc02009-01-24 15:58:40 +00001789
Misha Brukman76307852003-11-08 01:05:38 +00001790</div>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001791
Chris Lattner2f7c9632001-06-06 20:29:01 +00001792<!-- _______________________________________________________________________ -->
Andrew Lenharth8df88e22006-12-08 17:13:00 +00001793<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1794</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001795
Andrew Lenharth8df88e22006-12-08 17:13:00 +00001796<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001797
Andrew Lenharth8df88e22006-12-08 17:13:00 +00001798<h5>Overview:</h5>
1799<p>The packed structure type is used to represent a collection of data members
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001800 together in memory. There is no padding between fields. Further, the
1801 alignment of a packed structure is 1 byte. The elements of a packed
1802 structure may be any type that has a size.</p>
1803
1804<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1805 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1806 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1807
Andrew Lenharth8df88e22006-12-08 17:13:00 +00001808<h5>Syntax:</h5>
Bill Wendling30235112009-07-20 02:39:26 +00001809<pre>
1810 &lt; { &lt;type list&gt; } &gt;
1811</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001812
Andrew Lenharth8df88e22006-12-08 17:13:00 +00001813<h5>Examples:</h5>
1814<table class="layout">
1815 <tr class="layout">
Jeff Cohen5819f182007-04-22 01:17:39 +00001816 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1817 <td class="left">A triple of three <tt>i32</tt> values</td>
1818 </tr><tr class="layout">
Bill Wendlingb175fa42008-09-07 10:26:33 +00001819 <td class="left">
1820<tt>&lt;&nbsp;{&nbsp;float,&nbsp;i32&nbsp;(i32)*&nbsp;}&nbsp;&gt;</tt></td>
Jeff Cohen5819f182007-04-22 01:17:39 +00001821 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1822 second element is a <a href="#t_pointer">pointer</a> to a
1823 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1824 an <tt>i32</tt>.</td>
Andrew Lenharth8df88e22006-12-08 17:13:00 +00001825 </tr>
1826</table>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001827
Andrew Lenharth8df88e22006-12-08 17:13:00 +00001828</div>
1829
1830<!-- _______________________________________________________________________ -->
Chris Lattner392be582010-02-12 20:49:41 +00001831<div class="doc_subsubsection"> <a name="t_union">Union Type</a> </div>
1832
1833<div class="doc_text">
1834
1835<h5>Overview:</h5>
1836<p>A union type describes an object with size and alignment suitable for
1837 an object of any one of a given set of types (also known as an "untagged"
1838 union). It is similar in concept and usage to a
1839 <a href="#t_struct">struct</a>, except that all members of the union
1840 have an offset of zero. The elements of a union may be any type that has a
1841 size. Unions must have at least one member - empty unions are not allowed.
1842 </p>
1843
1844<p>The size of the union as a whole will be the size of its largest member,
1845 and the alignment requirements of the union as a whole will be the largest
1846 alignment requirement of any member.</p>
1847
Dan Gohman1ad14992010-02-25 16:51:31 +00001848<p>Union members are accessed using '<tt><a href="#i_load">load</a></tt> and
Chris Lattner392be582010-02-12 20:49:41 +00001849 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1850 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
1851 Since all members are at offset zero, the getelementptr instruction does
1852 not affect the address, only the type of the resulting pointer.</p>
1853
1854<h5>Syntax:</h5>
1855<pre>
1856 union { &lt;type list&gt; }
1857</pre>
1858
1859<h5>Examples:</h5>
1860<table class="layout">
1861 <tr class="layout">
1862 <td class="left"><tt>union { i32, i32*, float }</tt></td>
1863 <td class="left">A union of three types: an <tt>i32</tt>, a pointer to
1864 an <tt>i32</tt>, and a <tt>float</tt>.</td>
1865 </tr><tr class="layout">
1866 <td class="left">
1867 <tt>union {&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1868 <td class="left">A union, where the first element is a <tt>float</tt> and the
1869 second element is a <a href="#t_pointer">pointer</a> to a
1870 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1871 an <tt>i32</tt>.</td>
1872 </tr>
1873</table>
1874
1875</div>
1876
1877<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001878<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
Chris Lattner4a67c912009-02-08 19:53:29 +00001879
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001880<div class="doc_text">
1881
1882<h5>Overview:</h5>
Dan Gohman88481112010-02-25 16:50:07 +00001883<p>The pointer type is used to specify memory locations.
1884 Pointers are commonly used to reference objects in memory.</p>
1885
1886<p>Pointer types may have an optional address space attribute defining the
1887 numbered address space where the pointed-to object resides. The default
1888 address space is number zero. The semantics of non-zero address
1889 spaces are target-specific.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001890
1891<p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
1892 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
Chris Lattner4a67c912009-02-08 19:53:29 +00001893
Chris Lattner590645f2002-04-14 06:13:44 +00001894<h5>Syntax:</h5>
Bill Wendling30235112009-07-20 02:39:26 +00001895<pre>
1896 &lt;type&gt; *
1897</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001898
Chris Lattner590645f2002-04-14 06:13:44 +00001899<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001900<table class="layout">
1901 <tr class="layout">
Dan Gohman623806e2009-01-04 23:44:43 +00001902 <td class="left"><tt>[4 x i32]*</tt></td>
Chris Lattner747359f2007-12-19 05:04:11 +00001903 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1904 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1905 </tr>
1906 <tr class="layout">
Dan Gohmanaabfdb32010-05-28 17:13:49 +00001907 <td class="left"><tt>i32 (i32*) *</tt></td>
Chris Lattner747359f2007-12-19 05:04:11 +00001908 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001909 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner747359f2007-12-19 05:04:11 +00001910 <tt>i32</tt>.</td>
1911 </tr>
1912 <tr class="layout">
1913 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1914 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1915 that resides in address space #5.</td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001916 </tr>
Misha Brukman76307852003-11-08 01:05:38 +00001917</table>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001918
Misha Brukman76307852003-11-08 01:05:38 +00001919</div>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001920
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001921<!-- _______________________________________________________________________ -->
Reid Spencer404a3252007-02-15 03:07:05 +00001922<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001923
Misha Brukman76307852003-11-08 01:05:38 +00001924<div class="doc_text">
Chris Lattner37b6b092005-04-25 17:34:15 +00001925
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001926<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001927<p>A vector type is a simple derived type that represents a vector of elements.
1928 Vector types are used when multiple primitive data are operated in parallel
1929 using a single instruction (SIMD). A vector type requires a size (number of
Duncan Sands31c0e0e2009-11-27 13:38:03 +00001930 elements) and an underlying primitive data type. Vector types are considered
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001931 <a href="#t_firstclass">first class</a>.</p>
Chris Lattner37b6b092005-04-25 17:34:15 +00001932
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001933<h5>Syntax:</h5>
Chris Lattner37b6b092005-04-25 17:34:15 +00001934<pre>
1935 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1936</pre>
1937
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001938<p>The number of elements is a constant integer value; elementtype may be any
1939 integer or floating point type.</p>
Chris Lattner37b6b092005-04-25 17:34:15 +00001940
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001941<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001942<table class="layout">
1943 <tr class="layout">
Chris Lattner747359f2007-12-19 05:04:11 +00001944 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1945 <td class="left">Vector of 4 32-bit integer values.</td>
1946 </tr>
1947 <tr class="layout">
1948 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1949 <td class="left">Vector of 8 32-bit floating-point values.</td>
1950 </tr>
1951 <tr class="layout">
1952 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1953 <td class="left">Vector of 2 64-bit integer values.</td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001954 </tr>
1955</table>
Dan Gohman142ccc02009-01-24 15:58:40 +00001956
Misha Brukman76307852003-11-08 01:05:38 +00001957</div>
1958
Chris Lattner37b6b092005-04-25 17:34:15 +00001959<!-- _______________________________________________________________________ -->
1960<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1961<div class="doc_text">
1962
1963<h5>Overview:</h5>
Chris Lattner37b6b092005-04-25 17:34:15 +00001964<p>Opaque types are used to represent unknown types in the system. This
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001965 corresponds (for example) to the C notion of a forward declared structure
1966 type. In LLVM, opaque types can eventually be resolved to any type (not just
1967 a structure type).</p>
Chris Lattner37b6b092005-04-25 17:34:15 +00001968
1969<h5>Syntax:</h5>
Chris Lattner37b6b092005-04-25 17:34:15 +00001970<pre>
1971 opaque
1972</pre>
1973
1974<h5>Examples:</h5>
Chris Lattner37b6b092005-04-25 17:34:15 +00001975<table class="layout">
1976 <tr class="layout">
Chris Lattner747359f2007-12-19 05:04:11 +00001977 <td class="left"><tt>opaque</tt></td>
1978 <td class="left">An opaque type.</td>
Chris Lattner37b6b092005-04-25 17:34:15 +00001979 </tr>
1980</table>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001981
Chris Lattner37b6b092005-04-25 17:34:15 +00001982</div>
1983
Chris Lattnercf7a5842009-02-02 07:32:36 +00001984<!-- ======================================================================= -->
1985<div class="doc_subsection">
1986 <a name="t_uprefs">Type Up-references</a>
1987</div>
1988
1989<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001990
Chris Lattnercf7a5842009-02-02 07:32:36 +00001991<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00001992<p>An "up reference" allows you to refer to a lexically enclosing type without
1993 requiring it to have a name. For instance, a structure declaration may
1994 contain a pointer to any of the types it is lexically a member of. Example
1995 of up references (with their equivalent as named type declarations)
1996 include:</p>
Chris Lattnercf7a5842009-02-02 07:32:36 +00001997
1998<pre>
Chris Lattnerbf1d5452009-02-09 10:00:56 +00001999 { \2 * } %x = type { %x* }
Chris Lattnercf7a5842009-02-02 07:32:36 +00002000 { \2 }* %y = type { %y }*
2001 \1* %z = type %z*
2002</pre>
2003
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002004<p>An up reference is needed by the asmprinter for printing out cyclic types
2005 when there is no declared name for a type in the cycle. Because the
2006 asmprinter does not want to print out an infinite type string, it needs a
2007 syntax to handle recursive types that have no names (all names are optional
2008 in llvm IR).</p>
Chris Lattnercf7a5842009-02-02 07:32:36 +00002009
2010<h5>Syntax:</h5>
2011<pre>
2012 \&lt;level&gt;
2013</pre>
2014
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002015<p>The level is the count of the lexical type that is being referred to.</p>
Chris Lattnercf7a5842009-02-02 07:32:36 +00002016
2017<h5>Examples:</h5>
Chris Lattnercf7a5842009-02-02 07:32:36 +00002018<table class="layout">
2019 <tr class="layout">
2020 <td class="left"><tt>\1*</tt></td>
2021 <td class="left">Self-referential pointer.</td>
2022 </tr>
2023 <tr class="layout">
2024 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
2025 <td class="left">Recursive structure where the upref refers to the out-most
2026 structure.</td>
2027 </tr>
2028</table>
Chris Lattnercf7a5842009-02-02 07:32:36 +00002029
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002030</div>
Chris Lattner37b6b092005-04-25 17:34:15 +00002031
Chris Lattner74d3f822004-12-09 17:30:23 +00002032<!-- *********************************************************************** -->
2033<div class="doc_section"> <a name="constants">Constants</a> </div>
2034<!-- *********************************************************************** -->
2035
2036<div class="doc_text">
2037
2038<p>LLVM has several different basic types of constants. This section describes
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002039 them all and their syntax.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00002040
2041</div>
2042
2043<!-- ======================================================================= -->
Reid Spencer8f08d802004-12-09 18:02:53 +00002044<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
Chris Lattner74d3f822004-12-09 17:30:23 +00002045
2046<div class="doc_text">
2047
2048<dl>
2049 <dt><b>Boolean constants</b></dt>
Chris Lattner74d3f822004-12-09 17:30:23 +00002050 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
Nick Lewycky84a1eeb2009-09-27 00:45:11 +00002051 constants of the <tt><a href="#t_integer">i1</a></tt> type.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00002052
2053 <dt><b>Integer constants</b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002054 <dd>Standard integers (such as '4') are constants of
2055 the <a href="#t_integer">integer</a> type. Negative numbers may be used
2056 with integer types.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00002057
2058 <dt><b>Floating point constants</b></dt>
Chris Lattner74d3f822004-12-09 17:30:23 +00002059 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002060 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
2061 notation (see below). The assembler requires the exact decimal value of a
2062 floating-point constant. For example, the assembler accepts 1.25 but
2063 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
2064 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00002065
2066 <dt><b>Null pointer constants</b></dt>
John Criswelldfe6a862004-12-10 15:51:16 +00002067 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002068 and must be of <a href="#t_pointer">pointer type</a>.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00002069</dl>
2070
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002071<p>The one non-intuitive notation for constants is the hexadecimal form of
2072 floating point constants. For example, the form '<tt>double
2073 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
2074 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
2075 constants are required (and the only time that they are generated by the
2076 disassembler) is when a floating point constant must be emitted but it cannot
2077 be represented as a decimal floating point number in a reasonable number of
2078 digits. For example, NaN's, infinities, and other special values are
2079 represented in their IEEE hexadecimal format so that assembly and disassembly
2080 do not cause any bits to change in the constants.</p>
2081
Dale Johannesencd4a3012009-02-11 22:14:51 +00002082<p>When using the hexadecimal form, constants of types float and double are
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002083 represented using the 16-digit form shown above (which matches the IEEE754
2084 representation for double); float values must, however, be exactly
2085 representable as IEE754 single precision. Hexadecimal format is always used
2086 for long double, and there are three forms of long double. The 80-bit format
2087 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
2088 The 128-bit format used by PowerPC (two adjacent doubles) is represented
2089 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
2090 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
2091 currently supported target uses this format. Long doubles will only work if
2092 they match the long double format on your target. All hexadecimal formats
2093 are big-endian (sign bit at the left).</p>
2094
Chris Lattner74d3f822004-12-09 17:30:23 +00002095</div>
2096
2097<!-- ======================================================================= -->
Chris Lattner361bfcd2009-02-28 18:32:25 +00002098<div class="doc_subsection">
Bill Wendling972b7202009-07-20 02:32:41 +00002099<a name="aggregateconstants"></a> <!-- old anchor -->
2100<a name="complexconstants">Complex Constants</a>
Chris Lattner74d3f822004-12-09 17:30:23 +00002101</div>
2102
2103<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002104
Chris Lattner361bfcd2009-02-28 18:32:25 +00002105<p>Complex constants are a (potentially recursive) combination of simple
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002106 constants and smaller complex constants.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00002107
2108<dl>
2109 <dt><b>Structure constants</b></dt>
Chris Lattner74d3f822004-12-09 17:30:23 +00002110 <dd>Structure constants are represented with notation similar to structure
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002111 type definitions (a comma separated list of elements, surrounded by braces
2112 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
2113 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
2114 Structure constants must have <a href="#t_struct">structure type</a>, and
2115 the number and types of elements must match those specified by the
2116 type.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00002117
Chris Lattner392be582010-02-12 20:49:41 +00002118 <dt><b>Union constants</b></dt>
2119 <dd>Union constants are represented with notation similar to a structure with
2120 a single element - that is, a single typed element surrounded
2121 by braces (<tt>{}</tt>)). For example: "<tt>{ i32 4 }</tt>". The
2122 <a href="#t_union">union type</a> can be initialized with a single-element
2123 struct as long as the type of the struct element matches the type of
2124 one of the union members.</dd>
2125
Chris Lattner74d3f822004-12-09 17:30:23 +00002126 <dt><b>Array constants</b></dt>
Chris Lattner74d3f822004-12-09 17:30:23 +00002127 <dd>Array constants are represented with notation similar to array type
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002128 definitions (a comma separated list of elements, surrounded by square
2129 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
2130 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
2131 the number and types of elements must match those specified by the
2132 type.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00002133
Reid Spencer404a3252007-02-15 03:07:05 +00002134 <dt><b>Vector constants</b></dt>
Reid Spencer404a3252007-02-15 03:07:05 +00002135 <dd>Vector constants are represented with notation similar to vector type
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002136 definitions (a comma separated list of elements, surrounded by
2137 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32
2138 42, i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must
2139 have <a href="#t_vector">vector type</a>, and the number and types of
2140 elements must match those specified by the type.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00002141
2142 <dt><b>Zero initialization</b></dt>
Chris Lattner74d3f822004-12-09 17:30:23 +00002143 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
Chris Lattner392be582010-02-12 20:49:41 +00002144 value to zero of <em>any</em> type, including scalar and
2145 <a href="#t_aggregate">aggregate</a> types.
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002146 This is often used to avoid having to print large zero initializers
2147 (e.g. for large arrays) and is always exactly equivalent to using explicit
2148 zero initializers.</dd>
Nick Lewycky49f89192009-04-04 07:22:01 +00002149
2150 <dt><b>Metadata node</b></dt>
Nick Lewycky8e2c4f42009-05-30 16:08:30 +00002151 <dd>A metadata node is a structure-like constant with
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002152 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
2153 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
2154 be interpreted as part of the instruction stream, metadata is a place to
2155 attach additional information such as debug info.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00002156</dl>
2157
2158</div>
2159
2160<!-- ======================================================================= -->
2161<div class="doc_subsection">
2162 <a name="globalconstants">Global Variable and Function Addresses</a>
2163</div>
2164
2165<div class="doc_text">
2166
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002167<p>The addresses of <a href="#globalvars">global variables</a>
2168 and <a href="#functionstructure">functions</a> are always implicitly valid
2169 (link-time) constants. These constants are explicitly referenced when
2170 the <a href="#identifiers">identifier for the global</a> is used and always
2171 have <a href="#t_pointer">pointer</a> type. For example, the following is a
2172 legal LLVM file:</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00002173
Benjamin Kramer79698be2010-07-13 12:26:09 +00002174<pre class="doc_code">
Chris Lattner00538a12007-06-06 18:28:13 +00002175@X = global i32 17
2176@Y = global i32 42
2177@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
Chris Lattner74d3f822004-12-09 17:30:23 +00002178</pre>
2179
2180</div>
2181
2182<!-- ======================================================================= -->
Reid Spencer641f5c92004-12-09 18:13:12 +00002183<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
Chris Lattner74d3f822004-12-09 17:30:23 +00002184<div class="doc_text">
2185
Chris Lattnerec72b9b2009-09-07 22:52:39 +00002186<p>The string '<tt>undef</tt>' can be used anywhere a constant is expected, and
Benjamin Kramer0f420382009-10-12 14:46:08 +00002187 indicates that the user of the value may receive an unspecified bit-pattern.
Chris Lattnerec72b9b2009-09-07 22:52:39 +00002188 Undefined values may be of any type (other than label or void) and be used
2189 anywhere a constant is permitted.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002190
Chris Lattner92ada5d2009-09-11 01:49:31 +00002191<p>Undefined values are useful because they indicate to the compiler that the
Chris Lattnerec72b9b2009-09-07 22:52:39 +00002192 program is well defined no matter what value is used. This gives the
2193 compiler more freedom to optimize. Here are some examples of (potentially
2194 surprising) transformations that are valid (in pseudo IR):</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002195
Chris Lattnerec72b9b2009-09-07 22:52:39 +00002196
Benjamin Kramer79698be2010-07-13 12:26:09 +00002197<pre class="doc_code">
Chris Lattnerec72b9b2009-09-07 22:52:39 +00002198 %A = add %X, undef
2199 %B = sub %X, undef
2200 %C = xor %X, undef
2201Safe:
2202 %A = undef
2203 %B = undef
2204 %C = undef
2205</pre>
Chris Lattnerec72b9b2009-09-07 22:52:39 +00002206
2207<p>This is safe because all of the output bits are affected by the undef bits.
2208Any output bit can have a zero or one depending on the input bits.</p>
2209
Benjamin Kramer79698be2010-07-13 12:26:09 +00002210<pre class="doc_code">
Chris Lattnerec72b9b2009-09-07 22:52:39 +00002211 %A = or %X, undef
2212 %B = and %X, undef
2213Safe:
2214 %A = -1
2215 %B = 0
2216Unsafe:
2217 %A = undef
2218 %B = undef
2219</pre>
Chris Lattnerec72b9b2009-09-07 22:52:39 +00002220
2221<p>These logical operations have bits that are not always affected by the input.
2222For example, if "%X" has a zero bit, then the output of the 'and' operation will
2223always be a zero, no matter what the corresponding bit from the undef is. As
Chris Lattner92ada5d2009-09-11 01:49:31 +00002224such, it is unsafe to optimize or assume that the result of the and is undef.
Eric Christopher455c5772009-12-05 02:46:03 +00002225However, it is safe to assume that all bits of the undef could be 0, and
2226optimize the and to 0. Likewise, it is safe to assume that all the bits of
2227the undef operand to the or could be set, allowing the or to be folded to
Chris Lattner92ada5d2009-09-11 01:49:31 +00002228-1.</p>
Chris Lattnerec72b9b2009-09-07 22:52:39 +00002229
Benjamin Kramer79698be2010-07-13 12:26:09 +00002230<pre class="doc_code">
Chris Lattnerec72b9b2009-09-07 22:52:39 +00002231 %A = select undef, %X, %Y
2232 %B = select undef, 42, %Y
2233 %C = select %X, %Y, undef
2234Safe:
2235 %A = %X (or %Y)
2236 %B = 42 (or %Y)
2237 %C = %Y
2238Unsafe:
2239 %A = undef
2240 %B = undef
2241 %C = undef
2242</pre>
Chris Lattnerec72b9b2009-09-07 22:52:39 +00002243
2244<p>This set of examples show that undefined select (and conditional branch)
2245conditions can go "either way" but they have to come from one of the two
2246operands. In the %A example, if %X and %Y were both known to have a clear low
2247bit, then %A would have to have a cleared low bit. However, in the %C example,
2248the optimizer is allowed to assume that the undef operand could be the same as
2249%Y, allowing the whole select to be eliminated.</p>
2250
2251
Benjamin Kramer79698be2010-07-13 12:26:09 +00002252<pre class="doc_code">
Chris Lattnerec72b9b2009-09-07 22:52:39 +00002253 %A = xor undef, undef
Eric Christopher455c5772009-12-05 02:46:03 +00002254
Chris Lattnerec72b9b2009-09-07 22:52:39 +00002255 %B = undef
2256 %C = xor %B, %B
2257
2258 %D = undef
2259 %E = icmp lt %D, 4
2260 %F = icmp gte %D, 4
2261
2262Safe:
2263 %A = undef
2264 %B = undef
2265 %C = undef
2266 %D = undef
2267 %E = undef
2268 %F = undef
2269</pre>
Chris Lattnerec72b9b2009-09-07 22:52:39 +00002270
2271<p>This example points out that two undef operands are not necessarily the same.
2272This can be surprising to people (and also matches C semantics) where they
2273assume that "X^X" is always zero, even if X is undef. This isn't true for a
2274number of reasons, but the short answer is that an undef "variable" can
2275arbitrarily change its value over its "live range". This is true because the
2276"variable" doesn't actually <em>have a live range</em>. Instead, the value is
2277logically read from arbitrary registers that happen to be around when needed,
Benjamin Kramer0f420382009-10-12 14:46:08 +00002278so the value is not necessarily consistent over time. In fact, %A and %C need
Chris Lattner6760e542009-09-08 15:13:16 +00002279to have the same semantics or the core LLVM "replace all uses with" concept
Chris Lattnerec72b9b2009-09-07 22:52:39 +00002280would not hold.</p>
Chris Lattnera34a7182009-09-07 23:33:52 +00002281
Benjamin Kramer79698be2010-07-13 12:26:09 +00002282<pre class="doc_code">
Chris Lattnera34a7182009-09-07 23:33:52 +00002283 %A = fdiv undef, %X
2284 %B = fdiv %X, undef
2285Safe:
2286 %A = undef
2287b: unreachable
2288</pre>
Chris Lattnera34a7182009-09-07 23:33:52 +00002289
2290<p>These examples show the crucial difference between an <em>undefined
2291value</em> and <em>undefined behavior</em>. An undefined value (like undef) is
2292allowed to have an arbitrary bit-pattern. This means that the %A operation
2293can be constant folded to undef because the undef could be an SNaN, and fdiv is
2294not (currently) defined on SNaN's. However, in the second example, we can make
2295a more aggressive assumption: because the undef is allowed to be an arbitrary
2296value, we are allowed to assume that it could be zero. Since a divide by zero
Chris Lattner10ff0c12009-09-08 19:45:34 +00002297has <em>undefined behavior</em>, we are allowed to assume that the operation
Chris Lattnera34a7182009-09-07 23:33:52 +00002298does not execute at all. This allows us to delete the divide and all code after
2299it: since the undefined operation "can't happen", the optimizer can assume that
2300it occurs in dead code.
2301</p>
Eric Christopher455c5772009-12-05 02:46:03 +00002302
Benjamin Kramer79698be2010-07-13 12:26:09 +00002303<pre class="doc_code">
Chris Lattnera34a7182009-09-07 23:33:52 +00002304a: store undef -> %X
2305b: store %X -> undef
2306Safe:
2307a: &lt;deleted&gt;
2308b: unreachable
2309</pre>
Chris Lattnera34a7182009-09-07 23:33:52 +00002310
2311<p>These examples reiterate the fdiv example: a store "of" an undefined value
Eric Christopher455c5772009-12-05 02:46:03 +00002312can be assumed to not have any effect: we can assume that the value is
Chris Lattnera34a7182009-09-07 23:33:52 +00002313overwritten with bits that happen to match what was already there. However, a
2314store "to" an undefined location could clobber arbitrary memory, therefore, it
2315has undefined behavior.</p>
2316
Chris Lattner74d3f822004-12-09 17:30:23 +00002317</div>
2318
2319<!-- ======================================================================= -->
Dan Gohmanffc9a6b2010-04-22 23:14:21 +00002320<div class="doc_subsection"><a name="trapvalues">Trap Values</a></div>
2321<div class="doc_text">
2322
Dan Gohmanb2a709b2010-04-26 20:21:21 +00002323<p>Trap values are similar to <a href="#undefvalues">undef values</a>, however
Dan Gohmanffc9a6b2010-04-22 23:14:21 +00002324 instead of representing an unspecified bit pattern, they represent the
2325 fact that an instruction or constant expression which cannot evoke side
2326 effects has nevertheless detected a condition which results in undefined
Dan Gohmanb2a709b2010-04-26 20:21:21 +00002327 behavior.</p>
Dan Gohmanffc9a6b2010-04-22 23:14:21 +00002328
Dan Gohman2f1ae062010-04-28 00:49:41 +00002329<p>There is currently no way of representing a trap value in the IR; they
Dan Gohmanac355aa2010-05-03 14:51:43 +00002330 only exist when produced by operations such as
Dan Gohman2f1ae062010-04-28 00:49:41 +00002331 <a href="#i_add"><tt>add</tt></a> with the <tt>nsw</tt> flag.</p>
Dan Gohmanb8b85c12010-04-26 23:36:52 +00002332
Dan Gohman2f1ae062010-04-28 00:49:41 +00002333<p>Trap value behavior is defined in terms of value <i>dependence</i>:</p>
Dan Gohmanb8b85c12010-04-26 23:36:52 +00002334
Dan Gohman2f1ae062010-04-28 00:49:41 +00002335<ul>
2336<li>Values other than <a href="#i_phi"><tt>phi</tt></a> nodes depend on
2337 their operands.</li>
2338
2339<li><a href="#i_phi"><tt>Phi</tt></a> nodes depend on the operand corresponding
2340 to their dynamic predecessor basic block.</li>
2341
2342<li>Function arguments depend on the corresponding actual argument values in
2343 the dynamic callers of their functions.</li>
2344
2345<li><a href="#i_call"><tt>Call</tt></a> instructions depend on the
2346 <a href="#i_ret"><tt>ret</tt></a> instructions that dynamically transfer
2347 control back to them.</li>
2348
Dan Gohman7292a752010-05-03 14:55:22 +00002349<li><a href="#i_invoke"><tt>Invoke</tt></a> instructions depend on the
2350 <a href="#i_ret"><tt>ret</tt></a>, <a href="#i_unwind"><tt>unwind</tt></a>,
2351 or exception-throwing call instructions that dynamically transfer control
2352 back to them.</li>
2353
Dan Gohman2f1ae062010-04-28 00:49:41 +00002354<li>Non-volatile loads and stores depend on the most recent stores to all of the
2355 referenced memory addresses, following the order in the IR
2356 (including loads and stores implied by intrinsics such as
2357 <a href="#int_memcpy"><tt>@llvm.memcpy</tt></a>.)</li>
2358
Dan Gohman3513ea52010-05-03 14:59:34 +00002359<!-- TODO: In the case of multiple threads, this only applies if the store
2360 "happens-before" the load or store. -->
Dan Gohmanb8b85c12010-04-26 23:36:52 +00002361
Dan Gohman2f1ae062010-04-28 00:49:41 +00002362<!-- TODO: floating-point exception state -->
Dan Gohmanb8b85c12010-04-26 23:36:52 +00002363
Dan Gohman2f1ae062010-04-28 00:49:41 +00002364<li>An instruction with externally visible side effects depends on the most
2365 recent preceding instruction with externally visible side effects, following
Dan Gohman6c858db2010-07-06 15:26:33 +00002366 the order in the IR. (This includes
2367 <a href="#volatile">volatile operations</a>.)</li>
Dan Gohmanffc9a6b2010-04-22 23:14:21 +00002368
Dan Gohman7292a752010-05-03 14:55:22 +00002369<li>An instruction <i>control-depends</i> on a
2370 <a href="#terminators">terminator instruction</a>
2371 if the terminator instruction has multiple successors and the instruction
2372 is always executed when control transfers to one of the successors, and
2373 may not be executed when control is transfered to another.</li>
Dan Gohman2f1ae062010-04-28 00:49:41 +00002374
2375<li>Dependence is transitive.</li>
2376
2377</ul>
Dan Gohman2f1ae062010-04-28 00:49:41 +00002378
2379<p>Whenever a trap value is generated, all values which depend on it evaluate
2380 to trap. If they have side effects, the evoke their side effects as if each
2381 operand with a trap value were undef. If they have externally-visible side
2382 effects, the behavior is undefined.</p>
2383
2384<p>Here are some examples:</p>
Dan Gohman48a25882010-04-26 20:54:53 +00002385
Benjamin Kramer79698be2010-07-13 12:26:09 +00002386<pre class="doc_code">
Dan Gohmanb8b85c12010-04-26 23:36:52 +00002387entry:
2388 %trap = sub nuw i32 0, 1 ; Results in a trap value.
Dan Gohman2f1ae062010-04-28 00:49:41 +00002389 %still_trap = and i32 %trap, 0 ; Whereas (and i32 undef, 0) would return 0.
2390 %trap_yet_again = getelementptr i32* @h, i32 %still_trap
2391 store i32 0, i32* %trap_yet_again ; undefined behavior
2392
2393 store i32 %trap, i32* @g ; Trap value conceptually stored to memory.
2394 %trap2 = load i32* @g ; Returns a trap value, not just undef.
2395
2396 volatile store i32 %trap, i32* @g ; External observation; undefined behavior.
2397
2398 %narrowaddr = bitcast i32* @g to i16*
2399 %wideaddr = bitcast i32* @g to i64*
2400 %trap3 = load 16* %narrowaddr ; Returns a trap value.
2401 %trap4 = load i64* %widaddr ; Returns a trap value.
2402
2403 %cmp = icmp i32 slt %trap, 0 ; Returns a trap value.
Dan Gohmanb8b85c12010-04-26 23:36:52 +00002404 %br i1 %cmp, %true, %end ; Branch to either destination.
2405
2406true:
Dan Gohman2f1ae062010-04-28 00:49:41 +00002407 volatile store i32 0, i32* @g ; This is control-dependent on %cmp, so
2408 ; it has undefined behavior.
Dan Gohmanb8b85c12010-04-26 23:36:52 +00002409 br label %end
2410
2411end:
2412 %p = phi i32 [ 0, %entry ], [ 1, %true ]
2413 ; Both edges into this PHI are
2414 ; control-dependent on %cmp, so this
Dan Gohman2f1ae062010-04-28 00:49:41 +00002415 ; always results in a trap value.
Dan Gohmanb8b85c12010-04-26 23:36:52 +00002416
2417 volatile store i32 0, i32* @g ; %end is control-equivalent to %entry
2418 ; so this is defined (ignoring earlier
2419 ; undefined behavior in this example).
Dan Gohmanb8b85c12010-04-26 23:36:52 +00002420</pre>
Dan Gohmanffc9a6b2010-04-22 23:14:21 +00002421
Dan Gohmanffc9a6b2010-04-22 23:14:21 +00002422</div>
2423
2424<!-- ======================================================================= -->
Chris Lattner2bfd3202009-10-27 21:19:13 +00002425<div class="doc_subsection"><a name="blockaddress">Addresses of Basic
2426 Blocks</a></div>
Chris Lattnere4801f72009-10-27 21:01:34 +00002427<div class="doc_text">
2428
Chris Lattneraa99c942009-11-01 01:27:45 +00002429<p><b><tt>blockaddress(@function, %block)</tt></b></p>
Chris Lattnere4801f72009-10-27 21:01:34 +00002430
2431<p>The '<tt>blockaddress</tt>' constant computes the address of the specified
Chris Lattner5c5f0ac2009-10-27 21:49:40 +00002432 basic block in the specified function, and always has an i8* type. Taking
Chris Lattneraa99c942009-11-01 01:27:45 +00002433 the address of the entry block is illegal.</p>
Eric Christopher455c5772009-12-05 02:46:03 +00002434
Chris Lattnere4801f72009-10-27 21:01:34 +00002435<p>This value only has defined behavior when used as an operand to the
Chris Lattnerd04cb6d2009-10-28 00:19:10 +00002436 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>' instruction or for comparisons
Chris Lattnere4801f72009-10-27 21:01:34 +00002437 against null. Pointer equality tests between labels addresses is undefined
2438 behavior - though, again, comparison against null is ok, and no label is
Chris Lattner2bfd3202009-10-27 21:19:13 +00002439 equal to the null pointer. This may also be passed around as an opaque
2440 pointer sized value as long as the bits are not inspected. This allows
Chris Lattnerda37b302009-10-27 21:44:20 +00002441 <tt>ptrtoint</tt> and arithmetic to be performed on these values so long as
Chris Lattnerd04cb6d2009-10-28 00:19:10 +00002442 the original value is reconstituted before the <tt>indirectbr</tt>.</p>
Eric Christopher455c5772009-12-05 02:46:03 +00002443
Chris Lattner2bfd3202009-10-27 21:19:13 +00002444<p>Finally, some targets may provide defined semantics when
Chris Lattnere4801f72009-10-27 21:01:34 +00002445 using the value as the operand to an inline assembly, but that is target
2446 specific.
2447 </p>
2448
2449</div>
2450
2451
2452<!-- ======================================================================= -->
Chris Lattner74d3f822004-12-09 17:30:23 +00002453<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
2454</div>
2455
2456<div class="doc_text">
2457
2458<p>Constant expressions are used to allow expressions involving other constants
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002459 to be used as constants. Constant expressions may be of
2460 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2461 operation that does not have side effects (e.g. load and call are not
2462 supported). The following is the syntax for constant expressions:</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00002463
2464<dl>
Dan Gohmand6a6f612010-05-28 17:07:41 +00002465 <dt><b><tt>trunc (CST to TYPE)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002466 <dd>Truncate a constant to another type. The bit size of CST must be larger
2467 than the bit size of TYPE. Both types must be integers.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00002468
Dan Gohmand6a6f612010-05-28 17:07:41 +00002469 <dt><b><tt>zext (CST to TYPE)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002470 <dd>Zero extend a constant to another type. The bit size of CST must be
Duncan Sandsa522e562010-07-13 12:06:14 +00002471 smaller than the bit size of TYPE. Both types must be integers.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002472
Dan Gohmand6a6f612010-05-28 17:07:41 +00002473 <dt><b><tt>sext (CST to TYPE)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002474 <dd>Sign extend a constant to another type. The bit size of CST must be
Duncan Sandsa522e562010-07-13 12:06:14 +00002475 smaller than the bit size of TYPE. Both types must be integers.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002476
Dan Gohmand6a6f612010-05-28 17:07:41 +00002477 <dt><b><tt>fptrunc (CST to TYPE)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002478 <dd>Truncate a floating point constant to another floating point type. The
2479 size of CST must be larger than the size of TYPE. Both types must be
2480 floating point.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002481
Dan Gohmand6a6f612010-05-28 17:07:41 +00002482 <dt><b><tt>fpext (CST to TYPE)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002483 <dd>Floating point extend a constant to another type. The size of CST must be
2484 smaller or equal to the size of TYPE. Both types must be floating
2485 point.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002486
Dan Gohmand6a6f612010-05-28 17:07:41 +00002487 <dt><b><tt>fptoui (CST to TYPE)</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002488 <dd>Convert a floating point constant to the corresponding unsigned integer
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002489 constant. TYPE must be a scalar or vector integer type. CST must be of
2490 scalar or vector floating point type. Both CST and TYPE must be scalars,
2491 or vectors of the same number of elements. If the value won't fit in the
2492 integer type, the results are undefined.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002493
Dan Gohmand6a6f612010-05-28 17:07:41 +00002494 <dt><b><tt>fptosi (CST to TYPE)</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002495 <dd>Convert a floating point constant to the corresponding signed integer
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002496 constant. TYPE must be a scalar or vector integer type. CST must be of
2497 scalar or vector floating point type. Both CST and TYPE must be scalars,
2498 or vectors of the same number of elements. If the value won't fit in the
2499 integer type, the results are undefined.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002500
Dan Gohmand6a6f612010-05-28 17:07:41 +00002501 <dt><b><tt>uitofp (CST to TYPE)</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002502 <dd>Convert an unsigned integer constant to the corresponding floating point
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002503 constant. TYPE must be a scalar or vector floating point type. CST must be
2504 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2505 vectors of the same number of elements. If the value won't fit in the
2506 floating point type, the results are undefined.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002507
Dan Gohmand6a6f612010-05-28 17:07:41 +00002508 <dt><b><tt>sitofp (CST to TYPE)</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002509 <dd>Convert a signed integer constant to the corresponding floating point
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002510 constant. TYPE must be a scalar or vector floating point type. CST must be
2511 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2512 vectors of the same number of elements. If the value won't fit in the
2513 floating point type, the results are undefined.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00002514
Dan Gohmand6a6f612010-05-28 17:07:41 +00002515 <dt><b><tt>ptrtoint (CST to TYPE)</tt></b></dt>
Reid Spencer5b950642006-11-11 23:08:07 +00002516 <dd>Convert a pointer typed constant to the corresponding integer constant
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002517 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2518 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2519 make it fit in <tt>TYPE</tt>.</dd>
Reid Spencer5b950642006-11-11 23:08:07 +00002520
Dan Gohmand6a6f612010-05-28 17:07:41 +00002521 <dt><b><tt>inttoptr (CST to TYPE)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002522 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2523 type. CST must be of integer type. The CST value is zero extended,
2524 truncated, or unchanged to make it fit in a pointer size. This one is
2525 <i>really</i> dangerous!</dd>
Reid Spencer5b950642006-11-11 23:08:07 +00002526
Dan Gohmand6a6f612010-05-28 17:07:41 +00002527 <dt><b><tt>bitcast (CST to TYPE)</tt></b></dt>
Chris Lattner789dee32009-02-28 18:27:03 +00002528 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2529 are the same as those for the <a href="#i_bitcast">bitcast
2530 instruction</a>.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00002531
Dan Gohmand6a6f612010-05-28 17:07:41 +00002532 <dt><b><tt>getelementptr (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2533 <dt><b><tt>getelementptr inbounds (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
Chris Lattner74d3f822004-12-09 17:30:23 +00002534 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002535 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2536 instruction, the index list may have zero or more indexes, which are
2537 required to make sense for the type of "CSTPTR".</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00002538
Dan Gohmand6a6f612010-05-28 17:07:41 +00002539 <dt><b><tt>select (COND, VAL1, VAL2)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002540 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
Reid Spencer9965ee72006-12-04 19:23:19 +00002541
Dan Gohmand6a6f612010-05-28 17:07:41 +00002542 <dt><b><tt>icmp COND (VAL1, VAL2)</tt></b></dt>
Reid Spencer9965ee72006-12-04 19:23:19 +00002543 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2544
Dan Gohmand6a6f612010-05-28 17:07:41 +00002545 <dt><b><tt>fcmp COND (VAL1, VAL2)</tt></b></dt>
Reid Spencer9965ee72006-12-04 19:23:19 +00002546 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
Robert Bocchino7e97a6d2006-01-10 19:31:34 +00002547
Dan Gohmand6a6f612010-05-28 17:07:41 +00002548 <dt><b><tt>extractelement (VAL, IDX)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002549 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2550 constants.</dd>
Robert Bocchino7e97a6d2006-01-10 19:31:34 +00002551
Dan Gohmand6a6f612010-05-28 17:07:41 +00002552 <dt><b><tt>insertelement (VAL, ELT, IDX)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002553 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2554 constants.</dd>
Chris Lattner016a0e52006-04-08 00:13:41 +00002555
Dan Gohmand6a6f612010-05-28 17:07:41 +00002556 <dt><b><tt>shufflevector (VEC1, VEC2, IDXMASK)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002557 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2558 constants.</dd>
Chris Lattner016a0e52006-04-08 00:13:41 +00002559
Nick Lewycky9ab9a7f2010-05-29 06:44:15 +00002560 <dt><b><tt>extractvalue (VAL, IDX0, IDX1, ...)</tt></b></dt>
2561 <dd>Perform the <a href="#i_extractvalue">extractvalue operation</a> on
2562 constants. The index list is interpreted in a similar manner as indices in
2563 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2564 index value must be specified.</dd>
2565
2566 <dt><b><tt>insertvalue (VAL, ELT, IDX0, IDX1, ...)</tt></b></dt>
2567 <dd>Perform the <a href="#i_insertvalue">insertvalue operation</a> on
2568 constants. The index list is interpreted in a similar manner as indices in
2569 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2570 index value must be specified.</dd>
2571
Dan Gohmand6a6f612010-05-28 17:07:41 +00002572 <dt><b><tt>OPCODE (LHS, RHS)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002573 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2574 be any of the <a href="#binaryops">binary</a>
2575 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2576 on operands are the same as those for the corresponding instruction
2577 (e.g. no bitwise operations on floating point values are allowed).</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00002578</dl>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002579
Chris Lattner74d3f822004-12-09 17:30:23 +00002580</div>
Chris Lattnerb1652612004-03-08 16:49:10 +00002581
Chris Lattner2f7c9632001-06-06 20:29:01 +00002582<!-- *********************************************************************** -->
Chris Lattner98f013c2006-01-25 23:47:57 +00002583<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2584<!-- *********************************************************************** -->
2585
2586<!-- ======================================================================= -->
2587<div class="doc_subsection">
2588<a name="inlineasm">Inline Assembler Expressions</a>
2589</div>
2590
2591<div class="doc_text">
2592
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002593<p>LLVM supports inline assembler expressions (as opposed
2594 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2595 a special value. This value represents the inline assembler as a string
2596 (containing the instructions to emit), a list of operand constraints (stored
Dale Johannesen63c94fe2009-10-13 21:56:55 +00002597 as a string), a flag that indicates whether or not the inline asm
Dale Johannesen1cfb9582009-10-21 23:28:00 +00002598 expression has side effects, and a flag indicating whether the function
2599 containing the asm needs to align its stack conservatively. An example
2600 inline assembler expression is:</p>
Chris Lattner98f013c2006-01-25 23:47:57 +00002601
Benjamin Kramer79698be2010-07-13 12:26:09 +00002602<pre class="doc_code">
Bill Wendling3716c5d2007-05-29 09:04:49 +00002603i32 (i32) asm "bswap $0", "=r,r"
Chris Lattner98f013c2006-01-25 23:47:57 +00002604</pre>
2605
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002606<p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2607 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2608 have:</p>
Chris Lattner98f013c2006-01-25 23:47:57 +00002609
Benjamin Kramer79698be2010-07-13 12:26:09 +00002610<pre class="doc_code">
Bill Wendling3716c5d2007-05-29 09:04:49 +00002611%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
Chris Lattner98f013c2006-01-25 23:47:57 +00002612</pre>
2613
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002614<p>Inline asms with side effects not visible in the constraint list must be
2615 marked as having side effects. This is done through the use of the
2616 '<tt>sideeffect</tt>' keyword, like so:</p>
Chris Lattner98f013c2006-01-25 23:47:57 +00002617
Benjamin Kramer79698be2010-07-13 12:26:09 +00002618<pre class="doc_code">
Bill Wendling3716c5d2007-05-29 09:04:49 +00002619call void asm sideeffect "eieio", ""()
Chris Lattner98f013c2006-01-25 23:47:57 +00002620</pre>
2621
Dale Johannesen1cfb9582009-10-21 23:28:00 +00002622<p>In some cases inline asms will contain code that will not work unless the
2623 stack is aligned in some way, such as calls or SSE instructions on x86,
2624 yet will not contain code that does that alignment within the asm.
2625 The compiler should make conservative assumptions about what the asm might
2626 contain and should generate its usual stack alignment code in the prologue
2627 if the '<tt>alignstack</tt>' keyword is present:</p>
Dale Johannesen63c94fe2009-10-13 21:56:55 +00002628
Benjamin Kramer79698be2010-07-13 12:26:09 +00002629<pre class="doc_code">
Dale Johannesen1cfb9582009-10-21 23:28:00 +00002630call void asm alignstack "eieio", ""()
Dale Johannesen63c94fe2009-10-13 21:56:55 +00002631</pre>
Dale Johannesen63c94fe2009-10-13 21:56:55 +00002632
2633<p>If both keywords appear the '<tt>sideeffect</tt>' keyword must come
2634 first.</p>
2635
Chris Lattner98f013c2006-01-25 23:47:57 +00002636<p>TODO: The format of the asm and constraints string still need to be
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002637 documented here. Constraints on what can be done (e.g. duplication, moving,
2638 etc need to be documented). This is probably best done by reference to
2639 another document that covers inline asm from a holistic perspective.</p>
Chris Lattner51065562010-04-07 05:38:05 +00002640</div>
2641
2642<div class="doc_subsubsection">
2643<a name="inlineasm_md">Inline Asm Metadata</a>
2644</div>
2645
2646<div class="doc_text">
2647
2648<p>The call instructions that wrap inline asm nodes may have a "!srcloc" MDNode
2649 attached to it that contains a constant integer. If present, the code
2650 generator will use the integer as the location cookie value when report
2651 errors through the LLVMContext error reporting mechanisms. This allows a
Dan Gohman61110ae2010-04-28 00:36:01 +00002652 front-end to correlate backend errors that occur with inline asm back to the
Chris Lattner51065562010-04-07 05:38:05 +00002653 source code that produced it. For example:</p>
2654
Benjamin Kramer79698be2010-07-13 12:26:09 +00002655<pre class="doc_code">
Chris Lattner51065562010-04-07 05:38:05 +00002656call void asm sideeffect "something bad", ""()<b>, !srcloc !42</b>
2657...
2658!42 = !{ i32 1234567 }
2659</pre>
Chris Lattner51065562010-04-07 05:38:05 +00002660
2661<p>It is up to the front-end to make sense of the magic numbers it places in the
2662 IR.</p>
Chris Lattner98f013c2006-01-25 23:47:57 +00002663
2664</div>
2665
Chris Lattnerc2f8f162010-01-15 21:50:19 +00002666<!-- ======================================================================= -->
2667<div class="doc_subsection"><a name="metadata">Metadata Nodes and Metadata
2668 Strings</a>
2669</div>
2670
2671<div class="doc_text">
2672
2673<p>LLVM IR allows metadata to be attached to instructions in the program that
2674 can convey extra information about the code to the optimizers and code
2675 generator. One example application of metadata is source-level debug
2676 information. There are two metadata primitives: strings and nodes. All
2677 metadata has the <tt>metadata</tt> type and is identified in syntax by a
2678 preceding exclamation point ('<tt>!</tt>').</p>
2679
2680<p>A metadata string is a string surrounded by double quotes. It can contain
2681 any character by escaping non-printable characters with "\xx" where "xx" is
2682 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
2683
2684<p>Metadata nodes are represented with notation similar to structure constants
2685 (a comma separated list of elements, surrounded by braces and preceded by an
2686 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2687 10}</tt>". Metadata nodes can have any values as their operand.</p>
2688
2689<p>A <a href="#namedmetadatastructure">named metadata</a> is a collection of
2690 metadata nodes, which can be looked up in the module symbol table. For
2691 example: "<tt>!foo = metadata !{!4, !3}</tt>".
2692
Devang Patel9984bd62010-03-04 23:44:48 +00002693<p>Metadata can be used as function arguments. Here <tt>llvm.dbg.value</tt>
Benjamin Kramer79698be2010-07-13 12:26:09 +00002694 function is using two metadata arguments.</p>
Devang Patel9984bd62010-03-04 23:44:48 +00002695
Benjamin Kramer79698be2010-07-13 12:26:09 +00002696 <pre class="doc_code">
Devang Patel9984bd62010-03-04 23:44:48 +00002697 call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
2698 </pre>
Devang Patel9984bd62010-03-04 23:44:48 +00002699
2700<p>Metadata can be attached with an instruction. Here metadata <tt>!21</tt> is
Benjamin Kramer79698be2010-07-13 12:26:09 +00002701 attached with <tt>add</tt> instruction using <tt>!dbg</tt> identifier.</p>
Devang Patel9984bd62010-03-04 23:44:48 +00002702
Benjamin Kramer79698be2010-07-13 12:26:09 +00002703 <pre class="doc_code">
Devang Patel9984bd62010-03-04 23:44:48 +00002704 %indvar.next = add i64 %indvar, 1, !dbg !21
2705 </pre>
Chris Lattnerc2f8f162010-01-15 21:50:19 +00002706</div>
2707
Chris Lattnerae76db52009-07-20 05:55:19 +00002708
2709<!-- *********************************************************************** -->
2710<div class="doc_section">
2711 <a name="intrinsic_globals">Intrinsic Global Variables</a>
2712</div>
2713<!-- *********************************************************************** -->
2714
2715<p>LLVM has a number of "magic" global variables that contain data that affect
2716code generation or other IR semantics. These are documented here. All globals
Chris Lattner58f9bb22009-07-20 06:14:25 +00002717of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2718section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2719by LLVM.</p>
Chris Lattnerae76db52009-07-20 05:55:19 +00002720
2721<!-- ======================================================================= -->
2722<div class="doc_subsection">
2723<a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
2724</div>
2725
2726<div class="doc_text">
2727
2728<p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2729href="#linkage_appending">appending linkage</a>. This array contains a list of
2730pointers to global variables and functions which may optionally have a pointer
2731cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2732
2733<pre>
2734 @X = global i8 4
2735 @Y = global i32 123
2736
2737 @llvm.used = appending global [2 x i8*] [
2738 i8* @X,
2739 i8* bitcast (i32* @Y to i8*)
2740 ], section "llvm.metadata"
2741</pre>
2742
2743<p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2744compiler, assembler, and linker are required to treat the symbol as if there is
2745a reference to the global that it cannot see. For example, if a variable has
2746internal linkage and no references other than that from the <tt>@llvm.used</tt>
2747list, it cannot be deleted. This is commonly used to represent references from
2748inline asms and other things the compiler cannot "see", and corresponds to
2749"attribute((used))" in GNU C.</p>
2750
2751<p>On some targets, the code generator must emit a directive to the assembler or
2752object file to prevent the assembler and linker from molesting the symbol.</p>
2753
2754</div>
2755
2756<!-- ======================================================================= -->
2757<div class="doc_subsection">
Chris Lattner58f9bb22009-07-20 06:14:25 +00002758<a name="intg_compiler_used">The '<tt>llvm.compiler.used</tt>' Global Variable</a>
2759</div>
2760
2761<div class="doc_text">
2762
2763<p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2764<tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2765touching the symbol. On targets that support it, this allows an intelligent
2766linker to optimize references to the symbol without being impeded as it would be
2767by <tt>@llvm.used</tt>.</p>
2768
2769<p>This is a rare construct that should only be used in rare circumstances, and
2770should not be exposed to source languages.</p>
2771
2772</div>
2773
2774<!-- ======================================================================= -->
2775<div class="doc_subsection">
Chris Lattnerae76db52009-07-20 05:55:19 +00002776<a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
2777</div>
2778
2779<div class="doc_text">
David Chisnalla9d4a6f2010-04-30 19:23:49 +00002780<pre>
2781%0 = type { i32, void ()* }
David Chisnallb492b812010-04-30 19:27:35 +00002782@llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor }]
David Chisnalla9d4a6f2010-04-30 19:23:49 +00002783</pre>
2784<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.
2785</p>
Chris Lattnerae76db52009-07-20 05:55:19 +00002786
2787</div>
2788
2789<!-- ======================================================================= -->
2790<div class="doc_subsection">
2791<a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
2792</div>
2793
2794<div class="doc_text">
David Chisnalla9d4a6f2010-04-30 19:23:49 +00002795<pre>
2796%0 = type { i32, void ()* }
David Chisnallb492b812010-04-30 19:27:35 +00002797@llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor }]
David Chisnalla9d4a6f2010-04-30 19:23:49 +00002798</pre>
Chris Lattnerae76db52009-07-20 05:55:19 +00002799
David Chisnalla9d4a6f2010-04-30 19:23:49 +00002800<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.
2801</p>
Chris Lattnerae76db52009-07-20 05:55:19 +00002802
2803</div>
2804
2805
Chris Lattner98f013c2006-01-25 23:47:57 +00002806<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +00002807<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2808<!-- *********************************************************************** -->
Chris Lattner74d3f822004-12-09 17:30:23 +00002809
Misha Brukman76307852003-11-08 01:05:38 +00002810<div class="doc_text">
Chris Lattner74d3f822004-12-09 17:30:23 +00002811
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002812<p>The LLVM instruction set consists of several different classifications of
2813 instructions: <a href="#terminators">terminator
2814 instructions</a>, <a href="#binaryops">binary instructions</a>,
2815 <a href="#bitwiseops">bitwise binary instructions</a>,
2816 <a href="#memoryops">memory instructions</a>, and
2817 <a href="#otherops">other instructions</a>.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00002818
Misha Brukman76307852003-11-08 01:05:38 +00002819</div>
Chris Lattner74d3f822004-12-09 17:30:23 +00002820
Chris Lattner2f7c9632001-06-06 20:29:01 +00002821<!-- ======================================================================= -->
Chris Lattner48b383b02003-11-25 01:02:51 +00002822<div class="doc_subsection"> <a name="terminators">Terminator
2823Instructions</a> </div>
Chris Lattner74d3f822004-12-09 17:30:23 +00002824
Misha Brukman76307852003-11-08 01:05:38 +00002825<div class="doc_text">
Chris Lattner74d3f822004-12-09 17:30:23 +00002826
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002827<p>As mentioned <a href="#functionstructure">previously</a>, every basic block
2828 in a program ends with a "Terminator" instruction, which indicates which
2829 block should be executed after the current block is finished. These
2830 terminator instructions typically yield a '<tt>void</tt>' value: they produce
2831 control flow, not values (the one exception being the
2832 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2833
Duncan Sands626b0242010-04-15 20:35:54 +00002834<p>There are seven different terminator instructions: the
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002835 '<a href="#i_ret"><tt>ret</tt></a>' instruction, the
2836 '<a href="#i_br"><tt>br</tt></a>' instruction, the
2837 '<a href="#i_switch"><tt>switch</tt></a>' instruction, the
Bill Wendling33fef7e2009-11-02 00:25:26 +00002838 '<a href="#i_indirectbr">'<tt>indirectbr</tt></a>' Instruction, the
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002839 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the
2840 '<a href="#i_unwind"><tt>unwind</tt></a>' instruction, and the
2841 '<a href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00002842
Misha Brukman76307852003-11-08 01:05:38 +00002843</div>
Chris Lattner74d3f822004-12-09 17:30:23 +00002844
Chris Lattner2f7c9632001-06-06 20:29:01 +00002845<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00002846<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2847Instruction</a> </div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002848
Misha Brukman76307852003-11-08 01:05:38 +00002849<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002850
Chris Lattner2f7c9632001-06-06 20:29:01 +00002851<h5>Syntax:</h5>
Dan Gohmancc3132e2008-10-04 19:00:07 +00002852<pre>
2853 ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
Chris Lattner590645f2002-04-14 06:13:44 +00002854 ret void <i>; Return from void function</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002855</pre>
Chris Lattnerda508ac2008-04-23 04:59:35 +00002856
Chris Lattner2f7c9632001-06-06 20:29:01 +00002857<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002858<p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
2859 a value) from a function back to the caller.</p>
2860
2861<p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
2862 value and then causes control flow, and one that just causes control flow to
2863 occur.</p>
Chris Lattnerda508ac2008-04-23 04:59:35 +00002864
Chris Lattner2f7c9632001-06-06 20:29:01 +00002865<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002866<p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
2867 return value. The type of the return value must be a
2868 '<a href="#t_firstclass">first class</a>' type.</p>
Dan Gohmancc3132e2008-10-04 19:00:07 +00002869
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002870<p>A function is not <a href="#wellformed">well formed</a> if it it has a
2871 non-void return type and contains a '<tt>ret</tt>' instruction with no return
2872 value or a return value with a type that does not match its type, or if it
2873 has a void return type and contains a '<tt>ret</tt>' instruction with a
2874 return value.</p>
Chris Lattnerda508ac2008-04-23 04:59:35 +00002875
Chris Lattner2f7c9632001-06-06 20:29:01 +00002876<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002877<p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
2878 the calling function's context. If the caller is a
2879 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
2880 instruction after the call. If the caller was an
2881 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
2882 the beginning of the "normal" destination block. If the instruction returns
2883 a value, that value shall set the call or invoke instruction's return
2884 value.</p>
Chris Lattnerda508ac2008-04-23 04:59:35 +00002885
Chris Lattner2f7c9632001-06-06 20:29:01 +00002886<h5>Example:</h5>
Chris Lattnerda508ac2008-04-23 04:59:35 +00002887<pre>
2888 ret i32 5 <i>; Return an integer value of 5</i>
Chris Lattner590645f2002-04-14 06:13:44 +00002889 ret void <i>; Return from a void function</i>
Bill Wendling050ee8f2009-02-28 22:12:54 +00002890 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002891</pre>
Dan Gohman3065b612009-01-12 23:12:39 +00002892
Misha Brukman76307852003-11-08 01:05:38 +00002893</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002894<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00002895<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002896
Misha Brukman76307852003-11-08 01:05:38 +00002897<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002898
Chris Lattner2f7c9632001-06-06 20:29:01 +00002899<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002900<pre>
2901 br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002902</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002903
Chris Lattner2f7c9632001-06-06 20:29:01 +00002904<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002905<p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
2906 different basic block in the current function. There are two forms of this
2907 instruction, corresponding to a conditional branch and an unconditional
2908 branch.</p>
2909
Chris Lattner2f7c9632001-06-06 20:29:01 +00002910<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002911<p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
2912 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
2913 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
2914 target.</p>
2915
Chris Lattner2f7c9632001-06-06 20:29:01 +00002916<h5>Semantics:</h5>
Reid Spencer36a15422007-01-12 03:35:51 +00002917<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002918 argument is evaluated. If the value is <tt>true</tt>, control flows to the
2919 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2920 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2921
Chris Lattner2f7c9632001-06-06 20:29:01 +00002922<h5>Example:</h5>
Bill Wendling30235112009-07-20 02:39:26 +00002923<pre>
2924Test:
2925 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
2926 br i1 %cond, label %IfEqual, label %IfUnequal
2927IfEqual:
2928 <a href="#i_ret">ret</a> i32 1
2929IfUnequal:
2930 <a href="#i_ret">ret</a> i32 0
2931</pre>
2932
Misha Brukman76307852003-11-08 01:05:38 +00002933</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002934
Chris Lattner2f7c9632001-06-06 20:29:01 +00002935<!-- _______________________________________________________________________ -->
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002936<div class="doc_subsubsection">
2937 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2938</div>
2939
Misha Brukman76307852003-11-08 01:05:38 +00002940<div class="doc_text">
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002941
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002942<h5>Syntax:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002943<pre>
2944 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
2945</pre>
2946
Chris Lattner2f7c9632001-06-06 20:29:01 +00002947<h5>Overview:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002948<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002949 several different places. It is a generalization of the '<tt>br</tt>'
2950 instruction, allowing a branch to occur to one of many possible
2951 destinations.</p>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002952
Chris Lattner2f7c9632001-06-06 20:29:01 +00002953<h5>Arguments:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002954<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002955 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
2956 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
2957 The table is not allowed to contain duplicate constant entries.</p>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002958
Chris Lattner2f7c9632001-06-06 20:29:01 +00002959<h5>Semantics:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00002960<p>The <tt>switch</tt> instruction specifies a table of values and
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002961 destinations. When the '<tt>switch</tt>' instruction is executed, this table
2962 is searched for the given value. If the value is found, control flow is
Benjamin Kramer0f420382009-10-12 14:46:08 +00002963 transferred to the corresponding destination; otherwise, control flow is
2964 transferred to the default destination.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002965
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002966<h5>Implementation:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002967<p>Depending on properties of the target machine and the particular
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002968 <tt>switch</tt> instruction, this instruction may be code generated in
2969 different ways. For example, it could be generated as a series of chained
2970 conditional branches or with a lookup table.</p>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002971
2972<h5>Example:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002973<pre>
2974 <i>; Emulate a conditional br instruction</i>
Reid Spencer36a15422007-01-12 03:35:51 +00002975 %Val = <a href="#i_zext">zext</a> i1 %value to i32
Dan Gohman623806e2009-01-04 23:44:43 +00002976 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002977
2978 <i>; Emulate an unconditional br instruction</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002979 switch i32 0, label %dest [ ]
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002980
2981 <i>; Implement a jump table:</i>
Dan Gohman623806e2009-01-04 23:44:43 +00002982 switch i32 %val, label %otherwise [ i32 0, label %onzero
2983 i32 1, label %onone
2984 i32 2, label %ontwo ]
Chris Lattner2f7c9632001-06-06 20:29:01 +00002985</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00002986
Misha Brukman76307852003-11-08 01:05:38 +00002987</div>
Chris Lattner0132aff2005-05-06 22:57:40 +00002988
Chris Lattner3ed871f2009-10-27 19:13:16 +00002989
2990<!-- _______________________________________________________________________ -->
2991<div class="doc_subsubsection">
Chris Lattnerd04cb6d2009-10-28 00:19:10 +00002992 <a name="i_indirectbr">'<tt>indirectbr</tt>' Instruction</a>
Chris Lattner3ed871f2009-10-27 19:13:16 +00002993</div>
2994
2995<div class="doc_text">
2996
2997<h5>Syntax:</h5>
2998<pre>
Chris Lattnerd04cb6d2009-10-28 00:19:10 +00002999 indirectbr &lt;somety&gt;* &lt;address&gt;, [ label &lt;dest1&gt;, label &lt;dest2&gt;, ... ]
Chris Lattner3ed871f2009-10-27 19:13:16 +00003000</pre>
3001
3002<h5>Overview:</h5>
3003
Chris Lattnerd04cb6d2009-10-28 00:19:10 +00003004<p>The '<tt>indirectbr</tt>' instruction implements an indirect branch to a label
Chris Lattner3ed871f2009-10-27 19:13:16 +00003005 within the current function, whose address is specified by
Chris Lattnere4801f72009-10-27 21:01:34 +00003006 "<tt>address</tt>". Address must be derived from a <a
3007 href="#blockaddress">blockaddress</a> constant.</p>
Chris Lattner3ed871f2009-10-27 19:13:16 +00003008
3009<h5>Arguments:</h5>
3010
3011<p>The '<tt>address</tt>' argument is the address of the label to jump to. The
3012 rest of the arguments indicate the full set of possible destinations that the
3013 address may point to. Blocks are allowed to occur multiple times in the
3014 destination list, though this isn't particularly useful.</p>
Eric Christopher455c5772009-12-05 02:46:03 +00003015
Chris Lattner3ed871f2009-10-27 19:13:16 +00003016<p>This destination list is required so that dataflow analysis has an accurate
3017 understanding of the CFG.</p>
3018
3019<h5>Semantics:</h5>
3020
3021<p>Control transfers to the block specified in the address argument. All
3022 possible destination blocks must be listed in the label list, otherwise this
3023 instruction has undefined behavior. This implies that jumps to labels
3024 defined in other functions have undefined behavior as well.</p>
3025
3026<h5>Implementation:</h5>
3027
3028<p>This is typically implemented with a jump through a register.</p>
3029
3030<h5>Example:</h5>
3031<pre>
Chris Lattnerd04cb6d2009-10-28 00:19:10 +00003032 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
Chris Lattner3ed871f2009-10-27 19:13:16 +00003033</pre>
3034
3035</div>
3036
3037
Chris Lattner2f7c9632001-06-06 20:29:01 +00003038<!-- _______________________________________________________________________ -->
Chris Lattner0132aff2005-05-06 22:57:40 +00003039<div class="doc_subsubsection">
3040 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
3041</div>
3042
Misha Brukman76307852003-11-08 01:05:38 +00003043<div class="doc_text">
Chris Lattner0132aff2005-05-06 22:57:40 +00003044
Chris Lattner2f7c9632001-06-06 20:29:01 +00003045<h5>Syntax:</h5>
Chris Lattner0132aff2005-05-06 22:57:40 +00003046<pre>
Devang Patel02256232008-10-07 17:48:33 +00003047 &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 Lattner6b7a0082006-05-14 18:23:06 +00003048 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
Chris Lattner0132aff2005-05-06 22:57:40 +00003049</pre>
3050
Chris Lattnera8292f32002-05-06 22:08:29 +00003051<h5>Overview:</h5>
Chris Lattner0132aff2005-05-06 22:57:40 +00003052<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003053 function, with the possibility of control flow transfer to either the
3054 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
3055 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
3056 control flow will return to the "normal" label. If the callee (or any
3057 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
3058 instruction, control is interrupted and continued at the dynamically nearest
3059 "exception" label.</p>
Chris Lattner0132aff2005-05-06 22:57:40 +00003060
Chris Lattner2f7c9632001-06-06 20:29:01 +00003061<h5>Arguments:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00003062<p>This instruction requires several arguments:</p>
Chris Lattner0132aff2005-05-06 22:57:40 +00003063
Chris Lattner2f7c9632001-06-06 20:29:01 +00003064<ol>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003065 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
3066 convention</a> the call should use. If none is specified, the call
3067 defaults to using C calling conventions.</li>
Devang Patel7e9b05e2008-10-06 18:50:38 +00003068
3069 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003070 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
3071 '<tt>inreg</tt>' attributes are valid here.</li>
Devang Patel7e9b05e2008-10-06 18:50:38 +00003072
Chris Lattner0132aff2005-05-06 22:57:40 +00003073 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003074 function value being invoked. In most cases, this is a direct function
3075 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
3076 off an arbitrary pointer to function value.</li>
Chris Lattner0132aff2005-05-06 22:57:40 +00003077
3078 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003079 function to be invoked. </li>
Chris Lattner0132aff2005-05-06 22:57:40 +00003080
3081 <li>'<tt>function args</tt>': argument list whose types match the function
Chris Lattner47f2a832010-03-02 06:36:51 +00003082 signature argument types and parameter attributes. All arguments must be
3083 of <a href="#t_firstclass">first class</a> type. If the function
3084 signature indicates the function accepts a variable number of arguments,
3085 the extra arguments can be specified.</li>
Chris Lattner0132aff2005-05-06 22:57:40 +00003086
3087 <li>'<tt>normal label</tt>': the label reached when the called function
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003088 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
Chris Lattner0132aff2005-05-06 22:57:40 +00003089
3090 <li>'<tt>exception label</tt>': the label reached when a callee returns with
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003091 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
Chris Lattner0132aff2005-05-06 22:57:40 +00003092
Devang Patel02256232008-10-07 17:48:33 +00003093 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003094 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
3095 '<tt>readnone</tt>' attributes are valid here.</li>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003096</ol>
Chris Lattner0132aff2005-05-06 22:57:40 +00003097
Chris Lattner2f7c9632001-06-06 20:29:01 +00003098<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003099<p>This instruction is designed to operate as a standard
3100 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
3101 primary difference is that it establishes an association with a label, which
3102 is used by the runtime library to unwind the stack.</p>
Chris Lattner0132aff2005-05-06 22:57:40 +00003103
3104<p>This instruction is used in languages with destructors to ensure that proper
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003105 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
3106 exception. Additionally, this is important for implementation of
3107 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
Chris Lattner0132aff2005-05-06 22:57:40 +00003108
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003109<p>For the purposes of the SSA form, the definition of the value returned by the
3110 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
3111 block to the "normal" label. If the callee unwinds then no return value is
3112 available.</p>
Dan Gohman9069d892009-05-22 21:47:08 +00003113
Chris Lattner97257f82010-01-15 18:08:37 +00003114<p>Note that the code generator does not yet completely support unwind, and
3115that the invoke/unwind semantics are likely to change in future versions.</p>
3116
Chris Lattner2f7c9632001-06-06 20:29:01 +00003117<h5>Example:</h5>
Chris Lattner0132aff2005-05-06 22:57:40 +00003118<pre>
Nick Lewycky084ab472008-03-16 07:18:12 +00003119 %retval = invoke i32 @Test(i32 15) to label %Continue
Jeff Cohen5819f182007-04-22 01:17:39 +00003120 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewycky084ab472008-03-16 07:18:12 +00003121 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Jeff Cohen5819f182007-04-22 01:17:39 +00003122 unwind label %TestCleanup <i>; {i32}:retval set</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003123</pre>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00003124
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003125</div>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00003126
Chris Lattner5ed60612003-09-03 00:41:47 +00003127<!-- _______________________________________________________________________ -->
Chris Lattner08b7d5b2004-10-16 18:04:13 +00003128
Chris Lattner48b383b02003-11-25 01:02:51 +00003129<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
3130Instruction</a> </div>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00003131
Misha Brukman76307852003-11-08 01:05:38 +00003132<div class="doc_text">
Chris Lattner08b7d5b2004-10-16 18:04:13 +00003133
Chris Lattner5ed60612003-09-03 00:41:47 +00003134<h5>Syntax:</h5>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00003135<pre>
3136 unwind
3137</pre>
3138
Chris Lattner5ed60612003-09-03 00:41:47 +00003139<h5>Overview:</h5>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00003140<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003141 at the first callee in the dynamic call stack which used
3142 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
3143 This is primarily used to implement exception handling.</p>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00003144
Chris Lattner5ed60612003-09-03 00:41:47 +00003145<h5>Semantics:</h5>
Chris Lattnerfe8519c2008-04-19 21:01:16 +00003146<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003147 immediately halt. The dynamic call stack is then searched for the
3148 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
3149 Once found, execution continues at the "exceptional" destination block
3150 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
3151 instruction in the dynamic call chain, undefined behavior results.</p>
3152
Chris Lattner97257f82010-01-15 18:08:37 +00003153<p>Note that the code generator does not yet completely support unwind, and
3154that the invoke/unwind semantics are likely to change in future versions.</p>
3155
Misha Brukman76307852003-11-08 01:05:38 +00003156</div>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00003157
3158<!-- _______________________________________________________________________ -->
3159
3160<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
3161Instruction</a> </div>
3162
3163<div class="doc_text">
3164
3165<h5>Syntax:</h5>
3166<pre>
3167 unreachable
3168</pre>
3169
3170<h5>Overview:</h5>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00003171<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003172 instruction is used to inform the optimizer that a particular portion of the
3173 code is not reachable. This can be used to indicate that the code after a
3174 no-return function cannot be reached, and other facts.</p>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00003175
3176<h5>Semantics:</h5>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00003177<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003178
Chris Lattner08b7d5b2004-10-16 18:04:13 +00003179</div>
3180
Chris Lattner2f7c9632001-06-06 20:29:01 +00003181<!-- ======================================================================= -->
Chris Lattner48b383b02003-11-25 01:02:51 +00003182<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003183
Misha Brukman76307852003-11-08 01:05:38 +00003184<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003185
3186<p>Binary operators are used to do most of the computation in a program. They
3187 require two operands of the same type, execute an operation on them, and
3188 produce a single value. The operands might represent multiple data, as is
3189 the case with the <a href="#t_vector">vector</a> data type. The result value
3190 has the same type as its operands.</p>
3191
Misha Brukman76307852003-11-08 01:05:38 +00003192<p>There are several different binary operators:</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003193
Misha Brukman76307852003-11-08 01:05:38 +00003194</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003195
Chris Lattner2f7c9632001-06-06 20:29:01 +00003196<!-- _______________________________________________________________________ -->
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003197<div class="doc_subsubsection">
3198 <a name="i_add">'<tt>add</tt>' Instruction</a>
3199</div>
3200
Misha Brukman76307852003-11-08 01:05:38 +00003201<div class="doc_text">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003202
Chris Lattner2f7c9632001-06-06 20:29:01 +00003203<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003204<pre>
Dan Gohmanb07de442009-07-20 22:41:19 +00003205 &lt;result&gt; = add &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohman957b1312009-09-02 17:31:42 +00003206 &lt;result&gt; = add nuw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3207 &lt;result&gt; = add nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3208 &lt;result&gt; = add nuw nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003209</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003210
Chris Lattner2f7c9632001-06-06 20:29:01 +00003211<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00003212<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003213
Chris Lattner2f7c9632001-06-06 20:29:01 +00003214<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003215<p>The two arguments to the '<tt>add</tt>' instruction must
3216 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3217 integer values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003218
Chris Lattner2f7c9632001-06-06 20:29:01 +00003219<h5>Semantics:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +00003220<p>The value produced is the integer sum of the two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003221
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003222<p>If the sum has unsigned overflow, the result returned is the mathematical
3223 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003224
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003225<p>Because LLVM integers use a two's complement representation, this instruction
3226 is appropriate for both signed and unsigned integers.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003227
Dan Gohman902dfff2009-07-22 22:44:56 +00003228<p><tt>nuw</tt> and <tt>nsw</tt> stand for &quot;No Unsigned Wrap&quot;
3229 and &quot;No Signed Wrap&quot;, respectively. If the <tt>nuw</tt> and/or
3230 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
Dan Gohmanffc9a6b2010-04-22 23:14:21 +00003231 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3232 respectively, occurs.</p>
Dan Gohmanb07de442009-07-20 22:41:19 +00003233
Chris Lattner2f7c9632001-06-06 20:29:01 +00003234<h5>Example:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003235<pre>
3236 &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003237</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003238
Misha Brukman76307852003-11-08 01:05:38 +00003239</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003240
Chris Lattner2f7c9632001-06-06 20:29:01 +00003241<!-- _______________________________________________________________________ -->
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003242<div class="doc_subsubsection">
Dan Gohmana5b96452009-06-04 22:49:04 +00003243 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
3244</div>
3245
3246<div class="doc_text">
3247
3248<h5>Syntax:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +00003249<pre>
3250 &lt;result&gt; = fadd &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3251</pre>
3252
3253<h5>Overview:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +00003254<p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
3255
3256<h5>Arguments:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +00003257<p>The two arguments to the '<tt>fadd</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003258 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3259 floating point values. Both arguments must have identical types.</p>
Dan Gohmana5b96452009-06-04 22:49:04 +00003260
3261<h5>Semantics:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +00003262<p>The value produced is the floating point sum of the two operands.</p>
3263
3264<h5>Example:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +00003265<pre>
3266 &lt;result&gt; = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
3267</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003268
Dan Gohmana5b96452009-06-04 22:49:04 +00003269</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003270
Dan Gohmana5b96452009-06-04 22:49:04 +00003271<!-- _______________________________________________________________________ -->
3272<div class="doc_subsubsection">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003273 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
3274</div>
3275
Misha Brukman76307852003-11-08 01:05:38 +00003276<div class="doc_text">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003277
Chris Lattner2f7c9632001-06-06 20:29:01 +00003278<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003279<pre>
Dan Gohman902dfff2009-07-22 22:44:56 +00003280 &lt;result&gt; = sub &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohman957b1312009-09-02 17:31:42 +00003281 &lt;result&gt; = sub nuw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3282 &lt;result&gt; = sub nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3283 &lt;result&gt; = sub nuw nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003284</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003285
Chris Lattner2f7c9632001-06-06 20:29:01 +00003286<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00003287<p>The '<tt>sub</tt>' instruction returns the difference of its two
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003288 operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003289
3290<p>Note that the '<tt>sub</tt>' instruction is used to represent the
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003291 '<tt>neg</tt>' instruction present in most other intermediate
3292 representations.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003293
Chris Lattner2f7c9632001-06-06 20:29:01 +00003294<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003295<p>The two arguments to the '<tt>sub</tt>' instruction must
3296 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3297 integer values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003298
Chris Lattner2f7c9632001-06-06 20:29:01 +00003299<h5>Semantics:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +00003300<p>The value produced is the integer difference of the two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003301
Dan Gohmana5b96452009-06-04 22:49:04 +00003302<p>If the difference has unsigned overflow, the result returned is the
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003303 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
3304 result.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003305
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003306<p>Because LLVM integers use a two's complement representation, this instruction
3307 is appropriate for both signed and unsigned integers.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003308
Dan Gohman902dfff2009-07-22 22:44:56 +00003309<p><tt>nuw</tt> and <tt>nsw</tt> stand for &quot;No Unsigned Wrap&quot;
3310 and &quot;No Signed Wrap&quot;, respectively. If the <tt>nuw</tt> and/or
3311 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
Dan Gohmanffc9a6b2010-04-22 23:14:21 +00003312 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3313 respectively, occurs.</p>
Dan Gohmanb07de442009-07-20 22:41:19 +00003314
Chris Lattner2f7c9632001-06-06 20:29:01 +00003315<h5>Example:</h5>
Bill Wendling2d8b9a82007-05-29 09:42:13 +00003316<pre>
3317 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003318 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003319</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003320
Misha Brukman76307852003-11-08 01:05:38 +00003321</div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003322
Chris Lattner2f7c9632001-06-06 20:29:01 +00003323<!-- _______________________________________________________________________ -->
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003324<div class="doc_subsubsection">
Dan Gohmana5b96452009-06-04 22:49:04 +00003325 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
3326</div>
3327
3328<div class="doc_text">
3329
3330<h5>Syntax:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +00003331<pre>
3332 &lt;result&gt; = fsub &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3333</pre>
3334
3335<h5>Overview:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +00003336<p>The '<tt>fsub</tt>' instruction returns the difference of its two
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003337 operands.</p>
Dan Gohmana5b96452009-06-04 22:49:04 +00003338
3339<p>Note that the '<tt>fsub</tt>' instruction is used to represent the
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003340 '<tt>fneg</tt>' instruction present in most other intermediate
3341 representations.</p>
Dan Gohmana5b96452009-06-04 22:49:04 +00003342
3343<h5>Arguments:</h5>
Bill Wendling972b7202009-07-20 02:32:41 +00003344<p>The two arguments to the '<tt>fsub</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003345 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3346 floating point values. Both arguments must have identical types.</p>
Dan Gohmana5b96452009-06-04 22:49:04 +00003347
3348<h5>Semantics:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +00003349<p>The value produced is the floating point difference of the two operands.</p>
3350
3351<h5>Example:</h5>
3352<pre>
3353 &lt;result&gt; = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
3354 &lt;result&gt; = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
3355</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003356
Dan Gohmana5b96452009-06-04 22:49:04 +00003357</div>
3358
3359<!-- _______________________________________________________________________ -->
3360<div class="doc_subsubsection">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003361 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
3362</div>
3363
Misha Brukman76307852003-11-08 01:05:38 +00003364<div class="doc_text">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003365
Chris Lattner2f7c9632001-06-06 20:29:01 +00003366<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003367<pre>
Dan Gohman902dfff2009-07-22 22:44:56 +00003368 &lt;result&gt; = mul &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohman957b1312009-09-02 17:31:42 +00003369 &lt;result&gt; = mul nuw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3370 &lt;result&gt; = mul nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3371 &lt;result&gt; = mul nuw nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003372</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003373
Chris Lattner2f7c9632001-06-06 20:29:01 +00003374<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003375<p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003376
Chris Lattner2f7c9632001-06-06 20:29:01 +00003377<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003378<p>The two arguments to the '<tt>mul</tt>' instruction must
3379 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3380 integer values. Both arguments must have identical types.</p>
Eric Christopher455c5772009-12-05 02:46:03 +00003381
Chris Lattner2f7c9632001-06-06 20:29:01 +00003382<h5>Semantics:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +00003383<p>The value produced is the integer product of the two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003384
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003385<p>If the result of the multiplication has unsigned overflow, the result
3386 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
3387 width of the result.</p>
3388
3389<p>Because LLVM integers use a two's complement representation, and the result
3390 is the same width as the operands, this instruction returns the correct
3391 result for both signed and unsigned integers. If a full product
3392 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
3393 be sign-extended or zero-extended as appropriate to the width of the full
3394 product.</p>
3395
Dan Gohman902dfff2009-07-22 22:44:56 +00003396<p><tt>nuw</tt> and <tt>nsw</tt> stand for &quot;No Unsigned Wrap&quot;
3397 and &quot;No Signed Wrap&quot;, respectively. If the <tt>nuw</tt> and/or
3398 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
Dan Gohmanffc9a6b2010-04-22 23:14:21 +00003399 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3400 respectively, occurs.</p>
Dan Gohmanb07de442009-07-20 22:41:19 +00003401
Chris Lattner2f7c9632001-06-06 20:29:01 +00003402<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003403<pre>
3404 &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003405</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003406
Misha Brukman76307852003-11-08 01:05:38 +00003407</div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003408
Chris Lattner2f7c9632001-06-06 20:29:01 +00003409<!-- _______________________________________________________________________ -->
Dan Gohmana5b96452009-06-04 22:49:04 +00003410<div class="doc_subsubsection">
3411 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
3412</div>
3413
3414<div class="doc_text">
3415
3416<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003417<pre>
3418 &lt;result&gt; = fmul &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohmana5b96452009-06-04 22:49:04 +00003419</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003420
Dan Gohmana5b96452009-06-04 22:49:04 +00003421<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003422<p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
Dan Gohmana5b96452009-06-04 22:49:04 +00003423
3424<h5>Arguments:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +00003425<p>The two arguments to the '<tt>fmul</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003426 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3427 floating point values. Both arguments must have identical types.</p>
Dan Gohmana5b96452009-06-04 22:49:04 +00003428
3429<h5>Semantics:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +00003430<p>The value produced is the floating point product of the two operands.</p>
3431
3432<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003433<pre>
3434 &lt;result&gt; = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
Dan Gohmana5b96452009-06-04 22:49:04 +00003435</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003436
Dan Gohmana5b96452009-06-04 22:49:04 +00003437</div>
3438
3439<!-- _______________________________________________________________________ -->
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003440<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
3441</a></div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003442
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003443<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003444
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003445<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003446<pre>
3447 &lt;result&gt; = udiv &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003448</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003449
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003450<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003451<p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003452
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003453<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00003454<p>The two arguments to the '<tt>udiv</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003455 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3456 values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003457
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003458<h5>Semantics:</h5>
Chris Lattner2f2427e2008-01-28 00:36:27 +00003459<p>The value produced is the unsigned integer quotient of the two operands.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003460
Chris Lattner2f2427e2008-01-28 00:36:27 +00003461<p>Note that unsigned integer division and signed integer division are distinct
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003462 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
3463
Chris Lattner2f2427e2008-01-28 00:36:27 +00003464<p>Division by zero leads to undefined behavior.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003465
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003466<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003467<pre>
3468 &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003469</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003470
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003471</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003472
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003473<!-- _______________________________________________________________________ -->
3474<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
3475</a> </div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003476
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003477<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003478
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003479<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003480<pre>
Dan Gohmanb07de442009-07-20 22:41:19 +00003481 &lt;result&gt; = sdiv &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Dan Gohman957b1312009-09-02 17:31:42 +00003482 &lt;result&gt; = sdiv exact &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003483</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003484
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003485<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003486<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003487
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003488<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00003489<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003490 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3491 values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003492
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003493<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003494<p>The value produced is the signed integer quotient of the two operands rounded
3495 towards zero.</p>
3496
Chris Lattner2f2427e2008-01-28 00:36:27 +00003497<p>Note that signed integer division and unsigned integer division are distinct
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003498 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
3499
Chris Lattner2f2427e2008-01-28 00:36:27 +00003500<p>Division by zero leads to undefined behavior. Overflow also leads to
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003501 undefined behavior; this is a rare case, but can occur, for example, by doing
3502 a 32-bit division of -2147483648 by -1.</p>
3503
Dan Gohman71dfd782009-07-22 00:04:19 +00003504<p>If the <tt>exact</tt> keyword is present, the result value of the
Dan Gohman57255802010-04-23 15:23:32 +00003505 <tt>sdiv</tt> is a <a href="#trapvalues">trap value</a> if the result would
Dan Gohmane501ff72010-07-11 00:08:34 +00003506 be rounded.</p>
Dan Gohmanb07de442009-07-20 22:41:19 +00003507
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003508<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003509<pre>
3510 &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003511</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003512
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003513</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003514
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003515<!-- _______________________________________________________________________ -->
3516<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
Chris Lattner48b383b02003-11-25 01:02:51 +00003517Instruction</a> </div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003518
Misha Brukman76307852003-11-08 01:05:38 +00003519<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003520
Chris Lattner2f7c9632001-06-06 20:29:01 +00003521<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003522<pre>
Gabor Greif0f75ad02008-08-07 21:46:00 +00003523 &lt;result&gt; = fdiv &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner48b383b02003-11-25 01:02:51 +00003524</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003525
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003526<h5>Overview:</h5>
3527<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003528
Chris Lattner48b383b02003-11-25 01:02:51 +00003529<h5>Arguments:</h5>
Jeff Cohen5819f182007-04-22 01:17:39 +00003530<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003531 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3532 floating point values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003533
Chris Lattner48b383b02003-11-25 01:02:51 +00003534<h5>Semantics:</h5>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00003535<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003536
Chris Lattner48b383b02003-11-25 01:02:51 +00003537<h5>Example:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003538<pre>
3539 &lt;result&gt; = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
Chris Lattner48b383b02003-11-25 01:02:51 +00003540</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003541
Chris Lattner48b383b02003-11-25 01:02:51 +00003542</div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003543
Chris Lattner48b383b02003-11-25 01:02:51 +00003544<!-- _______________________________________________________________________ -->
Reid Spencer7eb55b32006-11-02 01:53:59 +00003545<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
3546</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003547
Reid Spencer7eb55b32006-11-02 01:53:59 +00003548<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003549
Reid Spencer7eb55b32006-11-02 01:53:59 +00003550<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003551<pre>
3552 &lt;result&gt; = urem &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer7eb55b32006-11-02 01:53:59 +00003553</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003554
Reid Spencer7eb55b32006-11-02 01:53:59 +00003555<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003556<p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3557 division of its two arguments.</p>
3558
Reid Spencer7eb55b32006-11-02 01:53:59 +00003559<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00003560<p>The two arguments to the '<tt>urem</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003561 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3562 values. Both arguments must have identical types.</p>
3563
Reid Spencer7eb55b32006-11-02 01:53:59 +00003564<h5>Semantics:</h5>
3565<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003566 This instruction always performs an unsigned division to get the
3567 remainder.</p>
3568
Chris Lattner2f2427e2008-01-28 00:36:27 +00003569<p>Note that unsigned integer remainder and signed integer remainder are
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003570 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3571
Chris Lattner2f2427e2008-01-28 00:36:27 +00003572<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003573
Reid Spencer7eb55b32006-11-02 01:53:59 +00003574<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003575<pre>
3576 &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
Reid Spencer7eb55b32006-11-02 01:53:59 +00003577</pre>
3578
3579</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003580
Reid Spencer7eb55b32006-11-02 01:53:59 +00003581<!-- _______________________________________________________________________ -->
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003582<div class="doc_subsubsection">
3583 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
3584</div>
3585
Chris Lattner48b383b02003-11-25 01:02:51 +00003586<div class="doc_text">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003587
Chris Lattner48b383b02003-11-25 01:02:51 +00003588<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003589<pre>
Gabor Greif0f75ad02008-08-07 21:46:00 +00003590 &lt;result&gt; = srem &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner48b383b02003-11-25 01:02:51 +00003591</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003592
Chris Lattner48b383b02003-11-25 01:02:51 +00003593<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003594<p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3595 division of its two operands. This instruction can also take
3596 <a href="#t_vector">vector</a> versions of the values in which case the
3597 elements must be integers.</p>
Chris Lattnerb8f816e2008-01-04 04:33:49 +00003598
Chris Lattner48b383b02003-11-25 01:02:51 +00003599<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00003600<p>The two arguments to the '<tt>srem</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003601 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3602 values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003603
Chris Lattner48b383b02003-11-25 01:02:51 +00003604<h5>Semantics:</h5>
Reid Spencer7eb55b32006-11-02 01:53:59 +00003605<p>This instruction returns the <i>remainder</i> of a division (where the result
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003606 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
3607 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
3608 a value. For more information about the difference,
3609 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3610 Math Forum</a>. For a table of how this is implemented in various languages,
3611 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3612 Wikipedia: modulo operation</a>.</p>
3613
Chris Lattner2f2427e2008-01-28 00:36:27 +00003614<p>Note that signed integer remainder and unsigned integer remainder are
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003615 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3616
Chris Lattner2f2427e2008-01-28 00:36:27 +00003617<p>Taking the remainder of a division by zero leads to undefined behavior.
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003618 Overflow also leads to undefined behavior; this is a rare case, but can
3619 occur, for example, by taking the remainder of a 32-bit division of
3620 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3621 lets srem be implemented using instructions that return both the result of
3622 the division and the remainder.)</p>
3623
Chris Lattner48b383b02003-11-25 01:02:51 +00003624<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003625<pre>
3626 &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
Reid Spencer7eb55b32006-11-02 01:53:59 +00003627</pre>
3628
3629</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003630
Reid Spencer7eb55b32006-11-02 01:53:59 +00003631<!-- _______________________________________________________________________ -->
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003632<div class="doc_subsubsection">
3633 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
3634
Reid Spencer7eb55b32006-11-02 01:53:59 +00003635<div class="doc_text">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003636
Reid Spencer7eb55b32006-11-02 01:53:59 +00003637<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003638<pre>
3639 &lt;result&gt; = frem &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer7eb55b32006-11-02 01:53:59 +00003640</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003641
Reid Spencer7eb55b32006-11-02 01:53:59 +00003642<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003643<p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3644 its two operands.</p>
3645
Reid Spencer7eb55b32006-11-02 01:53:59 +00003646<h5>Arguments:</h5>
3647<p>The two arguments to the '<tt>frem</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003648 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3649 floating point values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003650
Reid Spencer7eb55b32006-11-02 01:53:59 +00003651<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003652<p>This instruction returns the <i>remainder</i> of a division. The remainder
3653 has the same sign as the dividend.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003654
Reid Spencer7eb55b32006-11-02 01:53:59 +00003655<h5>Example:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003656<pre>
3657 &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
Chris Lattner48b383b02003-11-25 01:02:51 +00003658</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003659
Misha Brukman76307852003-11-08 01:05:38 +00003660</div>
Robert Bocchino820bc75b2006-02-17 21:18:08 +00003661
Reid Spencer2ab01932007-02-02 13:57:07 +00003662<!-- ======================================================================= -->
3663<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
3664Operations</a> </div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003665
Reid Spencer2ab01932007-02-02 13:57:07 +00003666<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003667
3668<p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3669 program. They are generally very efficient instructions and can commonly be
3670 strength reduced from other instructions. They require two operands of the
3671 same type, execute an operation on them, and produce a single value. The
3672 resulting value is the same type as its operands.</p>
3673
Reid Spencer2ab01932007-02-02 13:57:07 +00003674</div>
3675
Reid Spencer04e259b2007-01-31 21:39:12 +00003676<!-- _______________________________________________________________________ -->
3677<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
3678Instruction</a> </div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003679
Reid Spencer04e259b2007-01-31 21:39:12 +00003680<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003681
Reid Spencer04e259b2007-01-31 21:39:12 +00003682<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003683<pre>
3684 &lt;result&gt; = shl &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer04e259b2007-01-31 21:39:12 +00003685</pre>
Chris Lattnerf0e50112007-10-03 21:01:14 +00003686
Reid Spencer04e259b2007-01-31 21:39:12 +00003687<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003688<p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3689 a specified number of bits.</p>
Chris Lattnerf0e50112007-10-03 21:01:14 +00003690
Reid Spencer04e259b2007-01-31 21:39:12 +00003691<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003692<p>Both arguments to the '<tt>shl</tt>' instruction must be the
3693 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3694 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Eric Christopher455c5772009-12-05 02:46:03 +00003695
Reid Spencer04e259b2007-01-31 21:39:12 +00003696<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003697<p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3698 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3699 is (statically or dynamically) negative or equal to or larger than the number
3700 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3701 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3702 shift amount in <tt>op2</tt>.</p>
Chris Lattnerf0e50112007-10-03 21:01:14 +00003703
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003704<h5>Example:</h5>
3705<pre>
Reid Spencer04e259b2007-01-31 21:39:12 +00003706 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
3707 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
3708 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerf0e50112007-10-03 21:01:14 +00003709 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Mon P Wang4dd832d2008-12-09 05:46:39 +00003710 &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 Spencer04e259b2007-01-31 21:39:12 +00003711</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003712
Reid Spencer04e259b2007-01-31 21:39:12 +00003713</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003714
Reid Spencer04e259b2007-01-31 21:39:12 +00003715<!-- _______________________________________________________________________ -->
3716<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
3717Instruction</a> </div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003718
Reid Spencer04e259b2007-01-31 21:39:12 +00003719<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003720
Reid Spencer04e259b2007-01-31 21:39:12 +00003721<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003722<pre>
3723 &lt;result&gt; = lshr &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer04e259b2007-01-31 21:39:12 +00003724</pre>
3725
3726<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003727<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
3728 operand shifted to the right a specified number of bits with zero fill.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +00003729
3730<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00003731<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003732 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3733 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +00003734
3735<h5>Semantics:</h5>
3736<p>This instruction always performs a logical shift right operation. The most
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003737 significant bits of the result will be filled with zero bits after the shift.
3738 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
3739 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3740 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3741 shift amount in <tt>op2</tt>.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +00003742
3743<h5>Example:</h5>
3744<pre>
3745 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
3746 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
3747 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
3748 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerf0e50112007-10-03 21:01:14 +00003749 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Mon P Wang4dd832d2008-12-09 05:46:39 +00003750 &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 Spencer04e259b2007-01-31 21:39:12 +00003751</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003752
Reid Spencer04e259b2007-01-31 21:39:12 +00003753</div>
3754
Reid Spencer2ab01932007-02-02 13:57:07 +00003755<!-- _______________________________________________________________________ -->
Reid Spencer04e259b2007-01-31 21:39:12 +00003756<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
3757Instruction</a> </div>
3758<div class="doc_text">
3759
3760<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003761<pre>
3762 &lt;result&gt; = ashr &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer04e259b2007-01-31 21:39:12 +00003763</pre>
3764
3765<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003766<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3767 operand shifted to the right a specified number of bits with sign
3768 extension.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +00003769
3770<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00003771<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003772 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3773 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +00003774
3775<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003776<p>This instruction always performs an arithmetic shift right operation, The
3777 most significant bits of the result will be filled with the sign bit
3778 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3779 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
3780 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
3781 the corresponding shift amount in <tt>op2</tt>.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +00003782
3783<h5>Example:</h5>
3784<pre>
3785 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3786 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3787 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3788 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerf0e50112007-10-03 21:01:14 +00003789 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Mon P Wang4dd832d2008-12-09 05:46:39 +00003790 &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 Spencer04e259b2007-01-31 21:39:12 +00003791</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003792
Reid Spencer04e259b2007-01-31 21:39:12 +00003793</div>
3794
Chris Lattner2f7c9632001-06-06 20:29:01 +00003795<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00003796<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
3797Instruction</a> </div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003798
Misha Brukman76307852003-11-08 01:05:38 +00003799<div class="doc_text">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003800
Chris Lattner2f7c9632001-06-06 20:29:01 +00003801<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003802<pre>
Gabor Greif0f75ad02008-08-07 21:46:00 +00003803 &lt;result&gt; = and &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003804</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003805
Chris Lattner2f7c9632001-06-06 20:29:01 +00003806<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003807<p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
3808 operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003809
Chris Lattner2f7c9632001-06-06 20:29:01 +00003810<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00003811<p>The two arguments to the '<tt>and</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003812 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3813 values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003814
Chris Lattner2f7c9632001-06-06 20:29:01 +00003815<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00003816<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003817
Misha Brukman76307852003-11-08 01:05:38 +00003818<table border="1" cellspacing="0" cellpadding="4">
Chris Lattner48b383b02003-11-25 01:02:51 +00003819 <tbody>
3820 <tr>
3821 <td>In0</td>
3822 <td>In1</td>
3823 <td>Out</td>
3824 </tr>
3825 <tr>
3826 <td>0</td>
3827 <td>0</td>
3828 <td>0</td>
3829 </tr>
3830 <tr>
3831 <td>0</td>
3832 <td>1</td>
3833 <td>0</td>
3834 </tr>
3835 <tr>
3836 <td>1</td>
3837 <td>0</td>
3838 <td>0</td>
3839 </tr>
3840 <tr>
3841 <td>1</td>
3842 <td>1</td>
3843 <td>1</td>
3844 </tr>
3845 </tbody>
3846</table>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003847
Chris Lattner2f7c9632001-06-06 20:29:01 +00003848<h5>Example:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003849<pre>
3850 &lt;result&gt; = and i32 4, %var <i>; yields {i32}:result = 4 &amp; %var</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003851 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
3852 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003853</pre>
Misha Brukman76307852003-11-08 01:05:38 +00003854</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003855<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00003856<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003857
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003858<div class="doc_text">
3859
3860<h5>Syntax:</h5>
3861<pre>
3862 &lt;result&gt; = or &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3863</pre>
3864
3865<h5>Overview:</h5>
3866<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
3867 two operands.</p>
3868
3869<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00003870<p>The two arguments to the '<tt>or</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003871 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3872 values. Both arguments must have identical types.</p>
3873
Chris Lattner2f7c9632001-06-06 20:29:01 +00003874<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00003875<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003876
Chris Lattner48b383b02003-11-25 01:02:51 +00003877<table border="1" cellspacing="0" cellpadding="4">
3878 <tbody>
3879 <tr>
3880 <td>In0</td>
3881 <td>In1</td>
3882 <td>Out</td>
3883 </tr>
3884 <tr>
3885 <td>0</td>
3886 <td>0</td>
3887 <td>0</td>
3888 </tr>
3889 <tr>
3890 <td>0</td>
3891 <td>1</td>
3892 <td>1</td>
3893 </tr>
3894 <tr>
3895 <td>1</td>
3896 <td>0</td>
3897 <td>1</td>
3898 </tr>
3899 <tr>
3900 <td>1</td>
3901 <td>1</td>
3902 <td>1</td>
3903 </tr>
3904 </tbody>
3905</table>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003906
Chris Lattner2f7c9632001-06-06 20:29:01 +00003907<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003908<pre>
3909 &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003910 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
3911 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003912</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003913
Misha Brukman76307852003-11-08 01:05:38 +00003914</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003915
Chris Lattner2f7c9632001-06-06 20:29:01 +00003916<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00003917<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3918Instruction</a> </div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003919
Misha Brukman76307852003-11-08 01:05:38 +00003920<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003921
Chris Lattner2f7c9632001-06-06 20:29:01 +00003922<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003923<pre>
3924 &lt;result&gt; = xor &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003925</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003926
Chris Lattner2f7c9632001-06-06 20:29:01 +00003927<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003928<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
3929 its two operands. The <tt>xor</tt> is used to implement the "one's
3930 complement" operation, which is the "~" operator in C.</p>
3931
Chris Lattner2f7c9632001-06-06 20:29:01 +00003932<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00003933<p>The two arguments to the '<tt>xor</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003934 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3935 values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003936
Chris Lattner2f7c9632001-06-06 20:29:01 +00003937<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00003938<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003939
Chris Lattner48b383b02003-11-25 01:02:51 +00003940<table border="1" cellspacing="0" cellpadding="4">
3941 <tbody>
3942 <tr>
3943 <td>In0</td>
3944 <td>In1</td>
3945 <td>Out</td>
3946 </tr>
3947 <tr>
3948 <td>0</td>
3949 <td>0</td>
3950 <td>0</td>
3951 </tr>
3952 <tr>
3953 <td>0</td>
3954 <td>1</td>
3955 <td>1</td>
3956 </tr>
3957 <tr>
3958 <td>1</td>
3959 <td>0</td>
3960 <td>1</td>
3961 </tr>
3962 <tr>
3963 <td>1</td>
3964 <td>1</td>
3965 <td>0</td>
3966 </tr>
3967 </tbody>
3968</table>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003969
Chris Lattner2f7c9632001-06-06 20:29:01 +00003970<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003971<pre>
3972 &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003973 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3974 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3975 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003976</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003977
Misha Brukman76307852003-11-08 01:05:38 +00003978</div>
Chris Lattner54611b42005-11-06 08:02:57 +00003979
Chris Lattner2f7c9632001-06-06 20:29:01 +00003980<!-- ======================================================================= -->
Eric Christopher455c5772009-12-05 02:46:03 +00003981<div class="doc_subsection">
Chris Lattnerce83bff2006-04-08 23:07:04 +00003982 <a name="vectorops">Vector Operations</a>
3983</div>
3984
3985<div class="doc_text">
3986
3987<p>LLVM supports several instructions to represent vector operations in a
Bill Wendlingd9a66f72009-07-20 02:29:24 +00003988 target-independent manner. These instructions cover the element-access and
3989 vector-specific operations needed to process vectors effectively. While LLVM
3990 does directly support these vector operations, many sophisticated algorithms
3991 will want to use target-specific intrinsics to take full advantage of a
3992 specific target.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +00003993
3994</div>
3995
3996<!-- _______________________________________________________________________ -->
3997<div class="doc_subsubsection">
3998 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3999</div>
4000
4001<div class="doc_text">
4002
4003<h5>Syntax:</h5>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004004<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004005 &lt;result&gt; = extractelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, i32 &lt;idx&gt; <i>; yields &lt;ty&gt;</i>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004006</pre>
4007
4008<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004009<p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
4010 from a vector at a specified index.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004011
4012
4013<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004014<p>The first operand of an '<tt>extractelement</tt>' instruction is a value
4015 of <a href="#t_vector">vector</a> type. The second operand is an index
4016 indicating the position from which to extract the element. The index may be
4017 a variable.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004018
4019<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004020<p>The result is a scalar of the same type as the element type of
4021 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
4022 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4023 results are undefined.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004024
4025<h5>Example:</h5>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004026<pre>
Gabor Greif03ab4dc2009-10-28 13:14:50 +00004027 &lt;result&gt; = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004028</pre>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004029
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004030</div>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004031
4032<!-- _______________________________________________________________________ -->
4033<div class="doc_subsubsection">
4034 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
4035</div>
4036
4037<div class="doc_text">
4038
4039<h5>Syntax:</h5>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004040<pre>
Dan Gohman43ba0672008-05-12 23:38:42 +00004041 &lt;result&gt; = insertelement &lt;n x &lt;ty&gt;&gt; &lt;val&gt;, &lt;ty&gt; &lt;elt&gt;, i32 &lt;idx&gt; <i>; yields &lt;n x &lt;ty&gt;&gt;</i>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004042</pre>
4043
4044<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004045<p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
4046 vector at a specified index.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004047
4048<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004049<p>The first operand of an '<tt>insertelement</tt>' instruction is a value
4050 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
4051 whose type must equal the element type of the first operand. The third
4052 operand is an index indicating the position at which to insert the value.
4053 The index may be a variable.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004054
4055<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004056<p>The result is a vector of the same type as <tt>val</tt>. Its element values
4057 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
4058 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4059 results are undefined.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004060
4061<h5>Example:</h5>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004062<pre>
Gabor Greif03ab4dc2009-10-28 13:14:50 +00004063 &lt;result&gt; = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004064</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004065
Chris Lattnerce83bff2006-04-08 23:07:04 +00004066</div>
4067
4068<!-- _______________________________________________________________________ -->
4069<div class="doc_subsubsection">
4070 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
4071</div>
4072
4073<div class="doc_text">
4074
4075<h5>Syntax:</h5>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004076<pre>
Mon P Wang25f01062008-11-10 04:46:22 +00004077 &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 Lattnerce83bff2006-04-08 23:07:04 +00004078</pre>
4079
4080<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004081<p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
4082 from two input vectors, returning a vector with the same element type as the
4083 input and length that is the same as the shuffle mask.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004084
4085<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004086<p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
4087 with types that match each other. The third argument is a shuffle mask whose
4088 element type is always 'i32'. The result of the instruction is a vector
4089 whose length is the same as the shuffle mask and whose element type is the
4090 same as the element type of the first two operands.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004091
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004092<p>The shuffle mask operand is required to be a constant vector with either
4093 constant integer or undef values.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004094
4095<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004096<p>The elements of the two input vectors are numbered from left to right across
4097 both of the vectors. The shuffle mask operand specifies, for each element of
4098 the result vector, which element of the two input vectors the result element
4099 gets. The element selector may be undef (meaning "don't care") and the
4100 second operand may be undef if performing a shuffle from only one vector.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004101
4102<h5>Example:</h5>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004103<pre>
Eric Christopher455c5772009-12-05 02:46:03 +00004104 &lt;result&gt; = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
Jeff Cohen5819f182007-04-22 01:17:39 +00004105 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
Eric Christopher455c5772009-12-05 02:46:03 +00004106 &lt;result&gt; = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004107 &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 Christopher455c5772009-12-05 02:46:03 +00004108 &lt;result&gt; = shufflevector &lt;8 x i32&gt; %v1, &lt;8 x i32&gt; undef,
Mon P Wang25f01062008-11-10 04:46:22 +00004109 &lt;4 x i32&gt; &lt;i32 0, i32 1, i32 2, i32 3&gt; <i>; yields &lt;4 x i32&gt;</i>
Eric Christopher455c5772009-12-05 02:46:03 +00004110 &lt;result&gt; = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
Mon P Wang25f01062008-11-10 04:46:22 +00004111 &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 Lattnerce83bff2006-04-08 23:07:04 +00004112</pre>
Chris Lattnerce83bff2006-04-08 23:07:04 +00004113
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004114</div>
Tanya Lattnerb138bbe2006-04-14 19:24:33 +00004115
Chris Lattnerce83bff2006-04-08 23:07:04 +00004116<!-- ======================================================================= -->
Eric Christopher455c5772009-12-05 02:46:03 +00004117<div class="doc_subsection">
Dan Gohmanb9d66602008-05-12 23:51:09 +00004118 <a name="aggregateops">Aggregate Operations</a>
4119</div>
4120
4121<div class="doc_text">
4122
Chris Lattner392be582010-02-12 20:49:41 +00004123<p>LLVM supports several instructions for working with
4124 <a href="#t_aggregate">aggregate</a> values.</p>
Dan Gohmanb9d66602008-05-12 23:51:09 +00004125
4126</div>
4127
4128<!-- _______________________________________________________________________ -->
4129<div class="doc_subsubsection">
4130 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
4131</div>
4132
4133<div class="doc_text">
4134
4135<h5>Syntax:</h5>
Dan Gohmanb9d66602008-05-12 23:51:09 +00004136<pre>
4137 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
4138</pre>
4139
4140<h5>Overview:</h5>
Chris Lattner392be582010-02-12 20:49:41 +00004141<p>The '<tt>extractvalue</tt>' instruction extracts the value of a member field
4142 from an <a href="#t_aggregate">aggregate</a> value.</p>
Dan Gohmanb9d66602008-05-12 23:51:09 +00004143
4144<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004145<p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
Chris Lattner392be582010-02-12 20:49:41 +00004146 of <a href="#t_struct">struct</a>, <a href="#t_union">union</a> or
4147 <a href="#t_array">array</a> type. The operands are constant indices to
4148 specify which value to extract in a similar manner as indices in a
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004149 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
Dan Gohmanb9d66602008-05-12 23:51:09 +00004150
4151<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004152<p>The result is the value at the position in the aggregate specified by the
4153 index operands.</p>
Dan Gohmanb9d66602008-05-12 23:51:09 +00004154
4155<h5>Example:</h5>
Dan Gohmanb9d66602008-05-12 23:51:09 +00004156<pre>
Gabor Greif03ab4dc2009-10-28 13:14:50 +00004157 &lt;result&gt; = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
Dan Gohmanb9d66602008-05-12 23:51:09 +00004158</pre>
Dan Gohmanb9d66602008-05-12 23:51:09 +00004159
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004160</div>
Dan Gohmanb9d66602008-05-12 23:51:09 +00004161
4162<!-- _______________________________________________________________________ -->
4163<div class="doc_subsubsection">
4164 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
4165</div>
4166
4167<div class="doc_text">
4168
4169<h5>Syntax:</h5>
Dan Gohmanb9d66602008-05-12 23:51:09 +00004170<pre>
Jeffrey Yasskinf991bbb2010-01-11 19:19:26 +00004171 &lt;result&gt; = insertvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;ty&gt; &lt;elt&gt;, &lt;idx&gt; <i>; yields &lt;aggregate type&gt;</i>
Dan Gohmanb9d66602008-05-12 23:51:09 +00004172</pre>
4173
4174<h5>Overview:</h5>
Chris Lattner392be582010-02-12 20:49:41 +00004175<p>The '<tt>insertvalue</tt>' instruction inserts a value into a member field
4176 in an <a href="#t_aggregate">aggregate</a> value.</p>
Dan Gohmanb9d66602008-05-12 23:51:09 +00004177
4178<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004179<p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
Chris Lattner392be582010-02-12 20:49:41 +00004180 of <a href="#t_struct">struct</a>, <a href="#t_union">union</a> or
4181 <a href="#t_array">array</a> type. The second operand is a first-class
4182 value to insert. The following operands are constant indices indicating
4183 the position at which to insert the value in a similar manner as indices in a
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004184 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction. The
4185 value to insert must have the same type as the value identified by the
4186 indices.</p>
Dan Gohmanb9d66602008-05-12 23:51:09 +00004187
4188<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004189<p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
4190 that of <tt>val</tt> except that the value at the position specified by the
4191 indices is that of <tt>elt</tt>.</p>
Dan Gohmanb9d66602008-05-12 23:51:09 +00004192
4193<h5>Example:</h5>
Dan Gohmanb9d66602008-05-12 23:51:09 +00004194<pre>
Jeffrey Yasskinf991bbb2010-01-11 19:19:26 +00004195 %agg1 = insertvalue {i32, float} undef, i32 1, 0 <i>; yields {i32 1, float undef}</i>
4196 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 <i>; yields {i32 1, float %val}</i>
Dan Gohmanb9d66602008-05-12 23:51:09 +00004197</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004198
Dan Gohmanb9d66602008-05-12 23:51:09 +00004199</div>
4200
4201
4202<!-- ======================================================================= -->
Eric Christopher455c5772009-12-05 02:46:03 +00004203<div class="doc_subsection">
Chris Lattner6ab66722006-08-15 00:45:58 +00004204 <a name="memoryops">Memory Access and Addressing Operations</a>
Chris Lattner54611b42005-11-06 08:02:57 +00004205</div>
4206
Misha Brukman76307852003-11-08 01:05:38 +00004207<div class="doc_text">
Chris Lattner54611b42005-11-06 08:02:57 +00004208
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004209<p>A key design point of an SSA-based representation is how it represents
4210 memory. In LLVM, no memory locations are in SSA form, which makes things
Victor Hernandeza70c6df2009-10-26 23:44:29 +00004211 very simple. This section describes how to read, write, and allocate
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004212 memory in LLVM.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00004213
Misha Brukman76307852003-11-08 01:05:38 +00004214</div>
Chris Lattner54611b42005-11-06 08:02:57 +00004215
Chris Lattner2f7c9632001-06-06 20:29:01 +00004216<!-- _______________________________________________________________________ -->
Chris Lattner54611b42005-11-06 08:02:57 +00004217<div class="doc_subsubsection">
Chris Lattner54611b42005-11-06 08:02:57 +00004218 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
4219</div>
4220
Misha Brukman76307852003-11-08 01:05:38 +00004221<div class="doc_text">
Chris Lattner54611b42005-11-06 08:02:57 +00004222
Chris Lattner2f7c9632001-06-06 20:29:01 +00004223<h5>Syntax:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00004224<pre>
Dan Gohman2140a742010-05-28 01:14:11 +00004225 &lt;result&gt; = alloca &lt;type&gt;[, &lt;ty&gt; &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00004226</pre>
Chris Lattner54611b42005-11-06 08:02:57 +00004227
Chris Lattner2f7c9632001-06-06 20:29:01 +00004228<h5>Overview:</h5>
Jeff Cohen5819f182007-04-22 01:17:39 +00004229<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004230 currently executing function, to be automatically released when this function
4231 returns to its caller. The object is always allocated in the generic address
4232 space (address space zero).</p>
Chris Lattner54611b42005-11-06 08:02:57 +00004233
Chris Lattner2f7c9632001-06-06 20:29:01 +00004234<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004235<p>The '<tt>alloca</tt>' instruction
4236 allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt> bytes of memory on the
4237 runtime stack, returning a pointer of the appropriate type to the program.
4238 If "NumElements" is specified, it is the number of elements allocated,
4239 otherwise "NumElements" is defaulted to be one. If a constant alignment is
4240 specified, the value result of the allocation is guaranteed to be aligned to
4241 at least that boundary. If not specified, or if zero, the target can choose
4242 to align the allocation on any convenient boundary compatible with the
4243 type.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00004244
Misha Brukman76307852003-11-08 01:05:38 +00004245<p>'<tt>type</tt>' may be any sized type.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00004246
Chris Lattner2f7c9632001-06-06 20:29:01 +00004247<h5>Semantics:</h5>
Bill Wendling9ee6a312009-05-08 20:49:29 +00004248<p>Memory is allocated; a pointer is returned. The operation is undefined if
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004249 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
4250 memory is automatically released when the function returns. The
4251 '<tt>alloca</tt>' instruction is commonly used to represent automatic
4252 variables that must have an address available. When the function returns
4253 (either with the <tt><a href="#i_ret">ret</a></tt>
4254 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
4255 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00004256
Chris Lattner2f7c9632001-06-06 20:29:01 +00004257<h5>Example:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00004258<pre>
Dan Gohman7a5acb52009-01-04 23:49:44 +00004259 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
4260 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
4261 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
4262 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00004263</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004264
Misha Brukman76307852003-11-08 01:05:38 +00004265</div>
Chris Lattner54611b42005-11-06 08:02:57 +00004266
Chris Lattner2f7c9632001-06-06 20:29:01 +00004267<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00004268<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
4269Instruction</a> </div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004270
Misha Brukman76307852003-11-08 01:05:38 +00004271<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004272
Chris Lattner095735d2002-05-06 03:03:22 +00004273<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004274<pre>
Bill Wendlingbc4024f2010-02-25 21:23:24 +00004275 &lt;result&gt; = load &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;][, !nontemporal !&lt;index&gt;]
4276 &lt;result&gt; = volatile load &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;][, !nontemporal !&lt;index&gt;]
4277 !&lt;index&gt; = !{ i32 1 }
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004278</pre>
4279
Chris Lattner095735d2002-05-06 03:03:22 +00004280<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00004281<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004282
Chris Lattner095735d2002-05-06 03:03:22 +00004283<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004284<p>The argument to the '<tt>load</tt>' instruction specifies the memory address
4285 from which to load. The pointer must point to
4286 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
4287 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +00004288 number or order of execution of this <tt>load</tt> with other <a
4289 href="#volatile">volatile operations</a>.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004290
Bill Wendlingbc4024f2010-02-25 21:23:24 +00004291<p>The optional constant <tt>align</tt> argument specifies the alignment of the
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004292 operation (that is, the alignment of the memory address). A value of 0 or an
Bill Wendlingbc4024f2010-02-25 21:23:24 +00004293 omitted <tt>align</tt> argument means that the operation has the preferential
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004294 alignment for the target. It is the responsibility of the code emitter to
4295 ensure that the alignment information is correct. Overestimating the
Bill Wendlingbc4024f2010-02-25 21:23:24 +00004296 alignment results in undefined behavior. Underestimating the alignment may
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004297 produce less efficient code. An alignment of 1 is always safe.</p>
4298
Bill Wendlingbc4024f2010-02-25 21:23:24 +00004299<p>The optional <tt>!nontemporal</tt> metadata must reference a single
4300 metatadata name &lt;index&gt; corresponding to a metadata node with
Dan Gohmana269a0a2010-03-01 17:41:39 +00004301 one <tt>i32</tt> entry of value 1. The existence of
Bill Wendlingbc4024f2010-02-25 21:23:24 +00004302 the <tt>!nontemporal</tt> metatadata on the instruction tells the optimizer
4303 and code generator that this load is not expected to be reused in the cache.
4304 The code generator may select special instructions to save cache bandwidth,
Dan Gohmana269a0a2010-03-01 17:41:39 +00004305 such as the <tt>MOVNT</tt> instruction on x86.</p>
David Greene9641d062010-02-16 20:50:18 +00004306
Chris Lattner095735d2002-05-06 03:03:22 +00004307<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004308<p>The location of memory pointed to is loaded. If the value being loaded is of
4309 scalar type then the number of bytes read does not exceed the minimum number
4310 of bytes needed to hold all bits of the type. For example, loading an
4311 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
4312 <tt>i20</tt> with a size that is not an integral number of bytes, the result
4313 is undefined if the value was not originally written using a store of the
4314 same type.</p>
4315
Chris Lattner095735d2002-05-06 03:03:22 +00004316<h5>Examples:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004317<pre>
4318 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4319 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004320 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
Chris Lattner095735d2002-05-06 03:03:22 +00004321</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004322
Misha Brukman76307852003-11-08 01:05:38 +00004323</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004324
Chris Lattner095735d2002-05-06 03:03:22 +00004325<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00004326<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
4327Instruction</a> </div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004328
Reid Spencera89fb182006-11-09 21:18:01 +00004329<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004330
Chris Lattner095735d2002-05-06 03:03:22 +00004331<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004332<pre>
Benjamin Kramer79698be2010-07-13 12:26:09 +00004333 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>
4334 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 Lattner095735d2002-05-06 03:03:22 +00004335</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004336
Chris Lattner095735d2002-05-06 03:03:22 +00004337<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00004338<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004339
Chris Lattner095735d2002-05-06 03:03:22 +00004340<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004341<p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
4342 and an address at which to store it. The type of the
4343 '<tt>&lt;pointer&gt;</tt>' operand must be a pointer to
4344 the <a href="#t_firstclass">first class</a> type of the
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +00004345 '<tt>&lt;value&gt;</tt>' operand. If the <tt>store</tt> is marked as
4346 <tt>volatile</tt>, then the optimizer is not allowed to modify the number or
4347 order of execution of this <tt>store</tt> with other <a
4348 href="#volatile">volatile operations</a>.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004349
4350<p>The optional constant "align" argument specifies the alignment of the
4351 operation (that is, the alignment of the memory address). A value of 0 or an
4352 omitted "align" argument means that the operation has the preferential
4353 alignment for the target. It is the responsibility of the code emitter to
4354 ensure that the alignment information is correct. Overestimating the
4355 alignment results in an undefined behavior. Underestimating the alignment may
4356 produce less efficient code. An alignment of 1 is always safe.</p>
4357
David Greene9641d062010-02-16 20:50:18 +00004358<p>The optional !nontemporal metadata must reference a single metatadata
Benjamin Kramer79698be2010-07-13 12:26:09 +00004359 name &lt;index&gt; corresponding to a metadata node with one i32 entry of
Dan Gohmana269a0a2010-03-01 17:41:39 +00004360 value 1. The existence of the !nontemporal metatadata on the
David Greene9641d062010-02-16 20:50:18 +00004361 instruction tells the optimizer and code generator that this load is
4362 not expected to be reused in the cache. The code generator may
4363 select special instructions to save cache bandwidth, such as the
Dan Gohmana269a0a2010-03-01 17:41:39 +00004364 MOVNT instruction on x86.</p>
David Greene9641d062010-02-16 20:50:18 +00004365
4366
Chris Lattner48b383b02003-11-25 01:02:51 +00004367<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004368<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>' at the
4369 location specified by the '<tt>&lt;pointer&gt;</tt>' operand. If
4370 '<tt>&lt;value&gt;</tt>' is of scalar type then the number of bytes written
4371 does not exceed the minimum number of bytes needed to hold all bits of the
4372 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
4373 writing a value of a type like <tt>i20</tt> with a size that is not an
4374 integral number of bytes, it is unspecified what happens to the extra bits
4375 that do not belong to the type, but they will typically be overwritten.</p>
4376
Chris Lattner095735d2002-05-06 03:03:22 +00004377<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004378<pre>
4379 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling8830ffe2007-10-22 05:10:05 +00004380 store i32 3, i32* %ptr <i>; yields {void}</i>
4381 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
Chris Lattner095735d2002-05-06 03:03:22 +00004382</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004383
Reid Spencer443460a2006-11-09 21:15:49 +00004384</div>
4385
Chris Lattner095735d2002-05-06 03:03:22 +00004386<!-- _______________________________________________________________________ -->
Chris Lattner33fd7022004-04-05 01:30:49 +00004387<div class="doc_subsubsection">
4388 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
4389</div>
4390
Misha Brukman76307852003-11-08 01:05:38 +00004391<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004392
Chris Lattner590645f2002-04-14 06:13:44 +00004393<h5>Syntax:</h5>
Chris Lattner33fd7022004-04-05 01:30:49 +00004394<pre>
Matthijs Kooijman0e268272008-10-13 13:44:15 +00004395 &lt;result&gt; = getelementptr &lt;pty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
Dan Gohman1639c392009-07-27 21:53:46 +00004396 &lt;result&gt; = getelementptr inbounds &lt;pty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
Chris Lattner33fd7022004-04-05 01:30:49 +00004397</pre>
4398
Chris Lattner590645f2002-04-14 06:13:44 +00004399<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004400<p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
Chris Lattner392be582010-02-12 20:49:41 +00004401 subelement of an <a href="#t_aggregate">aggregate</a> data structure.
4402 It performs address calculation only and does not access memory.</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00004403
Chris Lattner590645f2002-04-14 06:13:44 +00004404<h5>Arguments:</h5>
Matthijs Kooijman0e268272008-10-13 13:44:15 +00004405<p>The first argument is always a pointer, and forms the basis of the
Chris Lattnera40b9122009-07-29 06:44:13 +00004406 calculation. The remaining arguments are indices that indicate which of the
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004407 elements of the aggregate object are indexed. The interpretation of each
4408 index is dependent on the type being indexed into. The first index always
4409 indexes the pointer value given as the first argument, the second index
4410 indexes a value of the type pointed to (not necessarily the value directly
4411 pointed to, since the first index can be non-zero), etc. The first type
Chris Lattner392be582010-02-12 20:49:41 +00004412 indexed into must be a pointer value, subsequent types can be arrays,
4413 vectors, structs and unions. Note that subsequent types being indexed into
4414 can never be pointers, since that would require loading the pointer before
4415 continuing calculation.</p>
Matthijs Kooijman0e268272008-10-13 13:44:15 +00004416
4417<p>The type of each index argument depends on the type it is indexing into.
Chris Lattner392be582010-02-12 20:49:41 +00004418 When indexing into a (optionally packed) structure or union, only <tt>i32</tt>
4419 integer <b>constants</b> are allowed. When indexing into an array, pointer
4420 or vector, integers of any width are allowed, and they are not required to be
Chris Lattnera40b9122009-07-29 06:44:13 +00004421 constant.</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00004422
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004423<p>For example, let's consider a C code fragment and how it gets compiled to
4424 LLVM:</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00004425
Benjamin Kramer79698be2010-07-13 12:26:09 +00004426<pre class="doc_code">
Bill Wendling3716c5d2007-05-29 09:04:49 +00004427struct RT {
4428 char A;
Chris Lattnera446f1b2007-05-29 15:43:56 +00004429 int B[10][20];
Bill Wendling3716c5d2007-05-29 09:04:49 +00004430 char C;
4431};
4432struct ST {
Chris Lattnera446f1b2007-05-29 15:43:56 +00004433 int X;
Bill Wendling3716c5d2007-05-29 09:04:49 +00004434 double Y;
4435 struct RT Z;
4436};
Chris Lattner33fd7022004-04-05 01:30:49 +00004437
Chris Lattnera446f1b2007-05-29 15:43:56 +00004438int *foo(struct ST *s) {
Bill Wendling3716c5d2007-05-29 09:04:49 +00004439 return &amp;s[1].Z.B[5][13];
4440}
Chris Lattner33fd7022004-04-05 01:30:49 +00004441</pre>
4442
Misha Brukman76307852003-11-08 01:05:38 +00004443<p>The LLVM code generated by the GCC frontend is:</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00004444
Benjamin Kramer79698be2010-07-13 12:26:09 +00004445<pre class="doc_code">
Chris Lattnerbc088212009-01-11 20:53:49 +00004446%RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
4447%ST = <a href="#namedtypes">type</a> { i32, double, %RT }
Chris Lattner33fd7022004-04-05 01:30:49 +00004448
Dan Gohman6b867702009-07-25 02:23:48 +00004449define i32* @foo(%ST* %s) {
Bill Wendling3716c5d2007-05-29 09:04:49 +00004450entry:
4451 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
4452 ret i32* %reg
4453}
Chris Lattner33fd7022004-04-05 01:30:49 +00004454</pre>
4455
Chris Lattner590645f2002-04-14 06:13:44 +00004456<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00004457<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004458 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
4459 }</tt>' type, a structure. The second index indexes into the third element
4460 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
4461 i8 }</tt>' type, another structure. The third index indexes into the second
4462 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
4463 array. The two dimensions of the array are subscripted into, yielding an
4464 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
4465 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00004466
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004467<p>Note that it is perfectly legal to index partially through a structure,
4468 returning a pointer to an inner element. Because of this, the LLVM code for
4469 the given testcase is equivalent to:</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00004470
4471<pre>
Dan Gohman6b867702009-07-25 02:23:48 +00004472 define i32* @foo(%ST* %s) {
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004473 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
Jeff Cohen5819f182007-04-22 01:17:39 +00004474 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
4475 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004476 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
4477 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
4478 ret i32* %t5
Chris Lattner33fd7022004-04-05 01:30:49 +00004479 }
Chris Lattnera8292f32002-05-06 22:08:29 +00004480</pre>
Chris Lattnerc0ad71e2005-06-24 17:22:57 +00004481
Dan Gohman1639c392009-07-27 21:53:46 +00004482<p>If the <tt>inbounds</tt> keyword is present, the result value of the
Dan Gohman57255802010-04-23 15:23:32 +00004483 <tt>getelementptr</tt> is a <a href="#trapvalues">trap value</a> if the
4484 base pointer is not an <i>in bounds</i> address of an allocated object,
4485 or if any of the addresses that would be formed by successive addition of
4486 the offsets implied by the indices to the base address with infinitely
4487 precise arithmetic are not an <i>in bounds</i> address of that allocated
4488 object. The <i>in bounds</i> addresses for an allocated object are all
4489 the addresses that point into the object, plus the address one byte past
4490 the end.</p>
Dan Gohman1639c392009-07-27 21:53:46 +00004491
4492<p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
4493 the base address with silently-wrapping two's complement arithmetic, and
4494 the result value of the <tt>getelementptr</tt> may be outside the object
4495 pointed to by the base pointer. The result value may not necessarily be
4496 used to access memory though, even if it happens to point into allocated
4497 storage. See the <a href="#pointeraliasing">Pointer Aliasing Rules</a>
4498 section for more information.</p>
4499
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004500<p>The getelementptr instruction is often confusing. For some more insight into
4501 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
Chris Lattner6ab66722006-08-15 00:45:58 +00004502
Chris Lattner590645f2002-04-14 06:13:44 +00004503<h5>Example:</h5>
Chris Lattner33fd7022004-04-05 01:30:49 +00004504<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004505 <i>; yields [12 x i8]*:aptr</i>
Matthijs Kooijman0e268272008-10-13 13:44:15 +00004506 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
4507 <i>; yields i8*:vptr</i>
Dan Gohmanef9462f2008-10-14 16:51:45 +00004508 %vptr = getelementptr {i32, &lt;2 x i8&gt;}* %svptr, i64 0, i32 1, i32 1
Matthijs Kooijman0e268272008-10-13 13:44:15 +00004509 <i>; yields i8*:eptr</i>
4510 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
Sanjiv Gupta0c155e62009-04-25 07:27:44 +00004511 <i>; yields i32*:iptr</i>
Sanjiv Gupta77abea02009-04-24 16:38:13 +00004512 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
Chris Lattner33fd7022004-04-05 01:30:49 +00004513</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004514
Chris Lattner33fd7022004-04-05 01:30:49 +00004515</div>
Reid Spencer443460a2006-11-09 21:15:49 +00004516
Chris Lattner2f7c9632001-06-06 20:29:01 +00004517<!-- ======================================================================= -->
Reid Spencer97c5fa42006-11-08 01:18:52 +00004518<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
Misha Brukman76307852003-11-08 01:05:38 +00004519</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004520
Misha Brukman76307852003-11-08 01:05:38 +00004521<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004522
Reid Spencer97c5fa42006-11-08 01:18:52 +00004523<p>The instructions in this category are the conversion instructions (casting)
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004524 which all take a single operand and a type. They perform various bit
4525 conversions on the operand.</p>
4526
Misha Brukman76307852003-11-08 01:05:38 +00004527</div>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00004528
Chris Lattnera8292f32002-05-06 22:08:29 +00004529<!-- _______________________________________________________________________ -->
Chris Lattnerb53c28d2004-03-12 05:50:16 +00004530<div class="doc_subsubsection">
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004531 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
4532</div>
4533<div class="doc_text">
4534
4535<h5>Syntax:</h5>
4536<pre>
4537 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4538</pre>
4539
4540<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004541<p>The '<tt>trunc</tt>' instruction truncates its operand to the
4542 type <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004543
4544<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004545<p>The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
4546 be an <a href="#t_integer">integer</a> type, and a type that specifies the
4547 size and type of the result, which must be
4548 an <a href="#t_integer">integer</a> type. The bit size of <tt>value</tt> must
4549 be larger than the bit size of <tt>ty2</tt>. Equal sized types are not
4550 allowed.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004551
4552<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004553<p>The '<tt>trunc</tt>' instruction truncates the high order bits
4554 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
4555 source size must be larger than the destination size, <tt>trunc</tt> cannot
4556 be a <i>no-op cast</i>. It will always truncate bits.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004557
4558<h5>Example:</h5>
4559<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004560 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
Reid Spencer36a15422007-01-12 03:35:51 +00004561 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
Gabor Greiff50fd572009-10-28 09:21:30 +00004562 %Z = trunc i32 122 to i1 <i>; yields i1:false</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004563</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004564
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004565</div>
4566
4567<!-- _______________________________________________________________________ -->
4568<div class="doc_subsubsection">
4569 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
4570</div>
4571<div class="doc_text">
4572
4573<h5>Syntax:</h5>
4574<pre>
4575 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4576</pre>
4577
4578<h5>Overview:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00004579<p>The '<tt>zext</tt>' instruction zero extends its operand to type
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004580 <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004581
4582
4583<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00004584<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004585 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4586 also be of <a href="#t_integer">integer</a> type. The bit size of the
Eric Christopher455c5772009-12-05 02:46:03 +00004587 <tt>value</tt> must be smaller than the bit size of the destination type,
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004588 <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004589
4590<h5>Semantics:</h5>
4591<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004592 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004593
Reid Spencer07c9c682007-01-12 15:46:11 +00004594<p>When zero extending from i1, the result will always be either 0 or 1.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004595
4596<h5>Example:</h5>
4597<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004598 %X = zext i32 257 to i64 <i>; yields i64:257</i>
Reid Spencer36a15422007-01-12 03:35:51 +00004599 %Y = zext i1 true to i32 <i>; yields i32:1</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004600</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004601
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004602</div>
4603
4604<!-- _______________________________________________________________________ -->
4605<div class="doc_subsubsection">
4606 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
4607</div>
4608<div class="doc_text">
4609
4610<h5>Syntax:</h5>
4611<pre>
4612 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4613</pre>
4614
4615<h5>Overview:</h5>
4616<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
4617
4618<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00004619<p>The '<tt>sext</tt>' instruction takes a value to cast, which must be of
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004620 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4621 also be of <a href="#t_integer">integer</a> type. The bit size of the
Eric Christopher455c5772009-12-05 02:46:03 +00004622 <tt>value</tt> must be smaller than the bit size of the destination type,
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004623 <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004624
4625<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004626<p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
4627 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
4628 of the type <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004629
Reid Spencer36a15422007-01-12 03:35:51 +00004630<p>When sign extending from i1, the extension always results in -1 or 0.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004631
4632<h5>Example:</h5>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004633<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004634 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
Reid Spencer36a15422007-01-12 03:35:51 +00004635 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004636</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004637
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004638</div>
4639
4640<!-- _______________________________________________________________________ -->
4641<div class="doc_subsubsection">
Reid Spencer2e2740d2006-11-09 21:48:10 +00004642 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
4643</div>
4644
4645<div class="doc_text">
4646
4647<h5>Syntax:</h5>
Reid Spencer2e2740d2006-11-09 21:48:10 +00004648<pre>
4649 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4650</pre>
4651
4652<h5>Overview:</h5>
4653<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004654 <tt>ty2</tt>.</p>
Reid Spencer2e2740d2006-11-09 21:48:10 +00004655
4656<h5>Arguments:</h5>
4657<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004658 point</a> value to cast and a <a href="#t_floating">floating point</a> type
4659 to cast it to. The size of <tt>value</tt> must be larger than the size of
Eric Christopher455c5772009-12-05 02:46:03 +00004660 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004661 <i>no-op cast</i>.</p>
Reid Spencer2e2740d2006-11-09 21:48:10 +00004662
4663<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004664<p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
Eric Christopher455c5772009-12-05 02:46:03 +00004665 <a href="#t_floating">floating point</a> type to a smaller
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004666 <a href="#t_floating">floating point</a> type. If the value cannot fit
4667 within the destination type, <tt>ty2</tt>, then the results are
4668 undefined.</p>
Reid Spencer2e2740d2006-11-09 21:48:10 +00004669
4670<h5>Example:</h5>
4671<pre>
4672 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4673 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4674</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004675
Reid Spencer2e2740d2006-11-09 21:48:10 +00004676</div>
4677
4678<!-- _______________________________________________________________________ -->
4679<div class="doc_subsubsection">
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004680 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
4681</div>
4682<div class="doc_text">
4683
4684<h5>Syntax:</h5>
4685<pre>
4686 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4687</pre>
4688
4689<h5>Overview:</h5>
4690<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004691 floating point value.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004692
4693<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00004694<p>The '<tt>fpext</tt>' instruction takes a
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004695 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
4696 a <a href="#t_floating">floating point</a> type to cast it to. The source
4697 type must be smaller than the destination type.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004698
4699<h5>Semantics:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00004700<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004701 <a href="#t_floating">floating point</a> type to a larger
4702 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4703 used to make a <i>no-op cast</i> because it always changes bits. Use
4704 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004705
4706<h5>Example:</h5>
4707<pre>
4708 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
4709 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
4710</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004711
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004712</div>
4713
4714<!-- _______________________________________________________________________ -->
4715<div class="doc_subsubsection">
Reid Spencer2eadb532007-01-21 00:29:26 +00004716 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004717</div>
4718<div class="doc_text">
4719
4720<h5>Syntax:</h5>
4721<pre>
Reid Spencer753163d2007-07-31 14:40:14 +00004722 &lt;result&gt; = fptoui &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004723</pre>
4724
4725<h5>Overview:</h5>
Reid Spencer753163d2007-07-31 14:40:14 +00004726<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004727 unsigned integer equivalent of type <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004728
4729<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004730<p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
4731 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4732 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4733 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4734 vector integer type with the same number of elements as <tt>ty</tt></p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004735
4736<h5>Semantics:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00004737<p>The '<tt>fptoui</tt>' instruction converts its
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004738 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4739 towards zero) unsigned integer value. If the value cannot fit
4740 in <tt>ty2</tt>, the results are undefined.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004741
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004742<h5>Example:</h5>
4743<pre>
Reid Spencer753163d2007-07-31 14:40:14 +00004744 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner5b95a172007-09-22 03:17:52 +00004745 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Gabor Greiff50fd572009-10-28 09:21:30 +00004746 %Z = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004747</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004748
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004749</div>
4750
4751<!-- _______________________________________________________________________ -->
4752<div class="doc_subsubsection">
Reid Spencer51b07252006-11-09 23:03:26 +00004753 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004754</div>
4755<div class="doc_text">
4756
4757<h5>Syntax:</h5>
4758<pre>
Reid Spencer51b07252006-11-09 23:03:26 +00004759 &lt;result&gt; = fptosi &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004760</pre>
4761
4762<h5>Overview:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00004763<p>The '<tt>fptosi</tt>' instruction converts
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004764 <a href="#t_floating">floating point</a> <tt>value</tt> to
4765 type <tt>ty2</tt>.</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00004766
Chris Lattnera8292f32002-05-06 22:08:29 +00004767<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004768<p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
4769 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4770 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4771 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4772 vector integer type with the same number of elements as <tt>ty</tt></p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00004773
Chris Lattnera8292f32002-05-06 22:08:29 +00004774<h5>Semantics:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00004775<p>The '<tt>fptosi</tt>' instruction converts its
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004776 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4777 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4778 the results are undefined.</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00004779
Chris Lattner70de6632001-07-09 00:26:23 +00004780<h5>Example:</h5>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00004781<pre>
Reid Spencer36a15422007-01-12 03:35:51 +00004782 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner5b95a172007-09-22 03:17:52 +00004783 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Gabor Greiff50fd572009-10-28 09:21:30 +00004784 %Z = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004785</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004786
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004787</div>
4788
4789<!-- _______________________________________________________________________ -->
4790<div class="doc_subsubsection">
Reid Spencer51b07252006-11-09 23:03:26 +00004791 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004792</div>
4793<div class="doc_text">
4794
4795<h5>Syntax:</h5>
4796<pre>
Reid Spencer51b07252006-11-09 23:03:26 +00004797 &lt;result&gt; = uitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004798</pre>
4799
4800<h5>Overview:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00004801<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004802 integer and converts that value to the <tt>ty2</tt> type.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004803
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004804<h5>Arguments:</h5>
Nate Begemand4d45c22007-11-17 03:58:34 +00004805<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004806 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4807 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4808 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4809 floating point type with the same number of elements as <tt>ty</tt></p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004810
4811<h5>Semantics:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00004812<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004813 integer quantity and converts it to the corresponding floating point
4814 value. If the value cannot fit in the floating point value, the results are
4815 undefined.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004816
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004817<h5>Example:</h5>
4818<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004819 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohmanef9462f2008-10-14 16:51:45 +00004820 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004821</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004822
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004823</div>
4824
4825<!-- _______________________________________________________________________ -->
4826<div class="doc_subsubsection">
Reid Spencer51b07252006-11-09 23:03:26 +00004827 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004828</div>
4829<div class="doc_text">
4830
4831<h5>Syntax:</h5>
4832<pre>
Reid Spencer51b07252006-11-09 23:03:26 +00004833 &lt;result&gt; = sitofp &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004834</pre>
4835
4836<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004837<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
4838 and converts that value to the <tt>ty2</tt> type.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004839
4840<h5>Arguments:</h5>
Nate Begemand4d45c22007-11-17 03:58:34 +00004841<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004842 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4843 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4844 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4845 floating point type with the same number of elements as <tt>ty</tt></p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004846
4847<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004848<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
4849 quantity and converts it to the corresponding floating point value. If the
4850 value cannot fit in the floating point value, the results are undefined.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004851
4852<h5>Example:</h5>
4853<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004854 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohmanef9462f2008-10-14 16:51:45 +00004855 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004856</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004857
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004858</div>
4859
4860<!-- _______________________________________________________________________ -->
4861<div class="doc_subsubsection">
Reid Spencerb7344ff2006-11-11 21:00:47 +00004862 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4863</div>
4864<div class="doc_text">
4865
4866<h5>Syntax:</h5>
4867<pre>
4868 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4869</pre>
4870
4871<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004872<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4873 the integer type <tt>ty2</tt>.</p>
Reid Spencerb7344ff2006-11-11 21:00:47 +00004874
4875<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004876<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4877 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
4878 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
Reid Spencerb7344ff2006-11-11 21:00:47 +00004879
4880<h5>Semantics:</h5>
4881<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004882 <tt>ty2</tt> by interpreting the pointer value as an integer and either
4883 truncating or zero extending that value to the size of the integer type. If
4884 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4885 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4886 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4887 change.</p>
Reid Spencerb7344ff2006-11-11 21:00:47 +00004888
4889<h5>Example:</h5>
4890<pre>
Jeff Cohen222a8a42007-04-29 01:07:00 +00004891 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4892 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
Reid Spencerb7344ff2006-11-11 21:00:47 +00004893</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004894
Reid Spencerb7344ff2006-11-11 21:00:47 +00004895</div>
4896
4897<!-- _______________________________________________________________________ -->
4898<div class="doc_subsubsection">
4899 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4900</div>
4901<div class="doc_text">
4902
4903<h5>Syntax:</h5>
4904<pre>
4905 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4906</pre>
4907
4908<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004909<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
4910 pointer type, <tt>ty2</tt>.</p>
Reid Spencerb7344ff2006-11-11 21:00:47 +00004911
4912<h5>Arguments:</h5>
Duncan Sands16f122e2007-03-30 12:22:09 +00004913<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004914 value to cast, and a type to cast it to, which must be a
4915 <a href="#t_pointer">pointer</a> type.</p>
Reid Spencerb7344ff2006-11-11 21:00:47 +00004916
4917<h5>Semantics:</h5>
4918<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004919 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
4920 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4921 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
4922 than the size of a pointer then a zero extension is done. If they are the
4923 same size, nothing is done (<i>no-op cast</i>).</p>
Reid Spencerb7344ff2006-11-11 21:00:47 +00004924
4925<h5>Example:</h5>
4926<pre>
Jeff Cohen222a8a42007-04-29 01:07:00 +00004927 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
Gabor Greiff50fd572009-10-28 09:21:30 +00004928 %Y = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4929 %Z = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
Reid Spencerb7344ff2006-11-11 21:00:47 +00004930</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004931
Reid Spencerb7344ff2006-11-11 21:00:47 +00004932</div>
4933
4934<!-- _______________________________________________________________________ -->
4935<div class="doc_subsubsection">
Reid Spencer5b950642006-11-11 23:08:07 +00004936 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004937</div>
4938<div class="doc_text">
4939
4940<h5>Syntax:</h5>
4941<pre>
Reid Spencer5b950642006-11-11 23:08:07 +00004942 &lt;result&gt; = bitcast &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004943</pre>
4944
4945<h5>Overview:</h5>
Reid Spencer5b950642006-11-11 23:08:07 +00004946<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004947 <tt>ty2</tt> without changing any bits.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004948
4949<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004950<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
4951 non-aggregate first class value, and a type to cast it to, which must also be
4952 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
4953 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
4954 identical. If the source type is a pointer, the destination type must also be
4955 a pointer. This instruction supports bitwise conversion of vectors to
4956 integers and to vectors of other types (as long as they have the same
4957 size).</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004958
4959<h5>Semantics:</h5>
Reid Spencer5b950642006-11-11 23:08:07 +00004960<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004961 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4962 this conversion. The conversion is done as if the <tt>value</tt> had been
4963 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
4964 be converted to other pointer types with this instruction. To convert
4965 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
4966 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004967
4968<h5>Example:</h5>
4969<pre>
Jeff Cohen222a8a42007-04-29 01:07:00 +00004970 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004971 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
Eric Christopher455c5772009-12-05 02:46:03 +00004972 %Z = bitcast &lt;2 x int&gt; %V to i64; <i>; yields i64: %V</i>
Chris Lattner70de6632001-07-09 00:26:23 +00004973</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004974
Misha Brukman76307852003-11-08 01:05:38 +00004975</div>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00004976
Reid Spencer97c5fa42006-11-08 01:18:52 +00004977<!-- ======================================================================= -->
4978<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004979
Reid Spencer97c5fa42006-11-08 01:18:52 +00004980<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004981
4982<p>The instructions in this category are the "miscellaneous" instructions, which
4983 defy better classification.</p>
4984
Reid Spencer97c5fa42006-11-08 01:18:52 +00004985</div>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004986
4987<!-- _______________________________________________________________________ -->
4988<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4989</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004990
Reid Spencerc828a0e2006-11-18 21:50:54 +00004991<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004992
Reid Spencerc828a0e2006-11-18 21:50:54 +00004993<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004994<pre>
4995 &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 Spencerc828a0e2006-11-18 21:50:54 +00004996</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004997
Reid Spencerc828a0e2006-11-18 21:50:54 +00004998<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00004999<p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
5000 boolean values based on comparison of its two integer, integer vector, or
5001 pointer operands.</p>
5002
Reid Spencerc828a0e2006-11-18 21:50:54 +00005003<h5>Arguments:</h5>
5004<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005005 the condition code indicating the kind of comparison to perform. It is not a
5006 value, just a keyword. The possible condition code are:</p>
5007
Reid Spencerc828a0e2006-11-18 21:50:54 +00005008<ol>
5009 <li><tt>eq</tt>: equal</li>
5010 <li><tt>ne</tt>: not equal </li>
5011 <li><tt>ugt</tt>: unsigned greater than</li>
5012 <li><tt>uge</tt>: unsigned greater or equal</li>
5013 <li><tt>ult</tt>: unsigned less than</li>
5014 <li><tt>ule</tt>: unsigned less or equal</li>
5015 <li><tt>sgt</tt>: signed greater than</li>
5016 <li><tt>sge</tt>: signed greater or equal</li>
5017 <li><tt>slt</tt>: signed less than</li>
5018 <li><tt>sle</tt>: signed less or equal</li>
5019</ol>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005020
Chris Lattnerc0f423a2007-01-15 01:54:13 +00005021<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005022 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
5023 typed. They must also be identical types.</p>
5024
Reid Spencerc828a0e2006-11-18 21:50:54 +00005025<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005026<p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
5027 condition code given as <tt>cond</tt>. The comparison performed always yields
Nick Lewycky84a1eeb2009-09-27 00:45:11 +00005028 either an <a href="#t_integer"><tt>i1</tt></a> or vector of <tt>i1</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005029 result, as follows:</p>
5030
Reid Spencerc828a0e2006-11-18 21:50:54 +00005031<ol>
Eric Christopher455c5772009-12-05 02:46:03 +00005032 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005033 <tt>false</tt> otherwise. No sign interpretation is necessary or
5034 performed.</li>
5035
Eric Christopher455c5772009-12-05 02:46:03 +00005036 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005037 <tt>false</tt> otherwise. No sign interpretation is necessary or
5038 performed.</li>
5039
Reid Spencerc828a0e2006-11-18 21:50:54 +00005040 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005041 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5042
Reid Spencerc828a0e2006-11-18 21:50:54 +00005043 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005044 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5045 to <tt>op2</tt>.</li>
5046
Reid Spencerc828a0e2006-11-18 21:50:54 +00005047 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005048 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5049
Reid Spencerc828a0e2006-11-18 21:50:54 +00005050 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005051 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5052
Reid Spencerc828a0e2006-11-18 21:50:54 +00005053 <li><tt>sgt</tt>: interprets the operands as signed values and yields
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005054 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5055
Reid Spencerc828a0e2006-11-18 21:50:54 +00005056 <li><tt>sge</tt>: interprets the operands as signed values and yields
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005057 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5058 to <tt>op2</tt>.</li>
5059
Reid Spencerc828a0e2006-11-18 21:50:54 +00005060 <li><tt>slt</tt>: interprets the operands as signed values and yields
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005061 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5062
Reid Spencerc828a0e2006-11-18 21:50:54 +00005063 <li><tt>sle</tt>: interprets the operands as signed values and yields
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005064 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00005065</ol>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005066
Reid Spencerc828a0e2006-11-18 21:50:54 +00005067<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005068 values are compared as if they were integers.</p>
5069
5070<p>If the operands are integer vectors, then they are compared element by
5071 element. The result is an <tt>i1</tt> vector with the same number of elements
5072 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
Reid Spencerc828a0e2006-11-18 21:50:54 +00005073
5074<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005075<pre>
5076 &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00005077 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
5078 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
5079 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
5080 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
5081 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
Reid Spencerc828a0e2006-11-18 21:50:54 +00005082</pre>
Dan Gohmana5127ab2009-01-22 01:39:38 +00005083
5084<p>Note that the code generator does not yet support vector types with
5085 the <tt>icmp</tt> instruction.</p>
5086
Reid Spencerc828a0e2006-11-18 21:50:54 +00005087</div>
5088
5089<!-- _______________________________________________________________________ -->
5090<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
5091</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005092
Reid Spencerc828a0e2006-11-18 21:50:54 +00005093<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005094
Reid Spencerc828a0e2006-11-18 21:50:54 +00005095<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005096<pre>
5097 &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 Spencerc828a0e2006-11-18 21:50:54 +00005098</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005099
Reid Spencerc828a0e2006-11-18 21:50:54 +00005100<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005101<p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
5102 values based on comparison of its operands.</p>
5103
5104<p>If the operands are floating point scalars, then the result type is a boolean
Nick Lewycky84a1eeb2009-09-27 00:45:11 +00005105(<a href="#t_integer"><tt>i1</tt></a>).</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005106
5107<p>If the operands are floating point vectors, then the result type is a vector
5108 of boolean with the same number of elements as the operands being
5109 compared.</p>
5110
Reid Spencerc828a0e2006-11-18 21:50:54 +00005111<h5>Arguments:</h5>
5112<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005113 the condition code indicating the kind of comparison to perform. It is not a
5114 value, just a keyword. The possible condition code are:</p>
5115
Reid Spencerc828a0e2006-11-18 21:50:54 +00005116<ol>
Reid Spencerf69acf32006-11-19 03:00:14 +00005117 <li><tt>false</tt>: no comparison, always returns false</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00005118 <li><tt>oeq</tt>: ordered and equal</li>
5119 <li><tt>ogt</tt>: ordered and greater than </li>
5120 <li><tt>oge</tt>: ordered and greater than or equal</li>
5121 <li><tt>olt</tt>: ordered and less than </li>
5122 <li><tt>ole</tt>: ordered and less than or equal</li>
5123 <li><tt>one</tt>: ordered and not equal</li>
5124 <li><tt>ord</tt>: ordered (no nans)</li>
5125 <li><tt>ueq</tt>: unordered or equal</li>
5126 <li><tt>ugt</tt>: unordered or greater than </li>
5127 <li><tt>uge</tt>: unordered or greater than or equal</li>
5128 <li><tt>ult</tt>: unordered or less than </li>
5129 <li><tt>ule</tt>: unordered or less than or equal</li>
5130 <li><tt>une</tt>: unordered or not equal</li>
5131 <li><tt>uno</tt>: unordered (either nans)</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00005132 <li><tt>true</tt>: no comparison, always returns true</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00005133</ol>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005134
Jeff Cohen222a8a42007-04-29 01:07:00 +00005135<p><i>Ordered</i> means that neither operand is a QNAN while
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005136 <i>unordered</i> means that either operand may be a QNAN.</p>
5137
5138<p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
5139 a <a href="#t_floating">floating point</a> type or
5140 a <a href="#t_vector">vector</a> of floating point type. They must have
5141 identical types.</p>
5142
Reid Spencerc828a0e2006-11-18 21:50:54 +00005143<h5>Semantics:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +00005144<p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005145 according to the condition code given as <tt>cond</tt>. If the operands are
5146 vectors, then the vectors are compared element by element. Each comparison
Nick Lewycky84a1eeb2009-09-27 00:45:11 +00005147 performed always yields an <a href="#t_integer">i1</a> result, as
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005148 follows:</p>
5149
Reid Spencerc828a0e2006-11-18 21:50:54 +00005150<ol>
5151 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005152
Eric Christopher455c5772009-12-05 02:46:03 +00005153 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005154 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5155
Reid Spencerf69acf32006-11-19 03:00:14 +00005156 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Dan Gohmana269a0a2010-03-01 17:41:39 +00005157 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005158
Eric Christopher455c5772009-12-05 02:46:03 +00005159 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005160 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5161
Eric Christopher455c5772009-12-05 02:46:03 +00005162 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005163 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5164
Eric Christopher455c5772009-12-05 02:46:03 +00005165 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005166 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5167
Eric Christopher455c5772009-12-05 02:46:03 +00005168 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005169 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5170
Reid Spencerf69acf32006-11-19 03:00:14 +00005171 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005172
Eric Christopher455c5772009-12-05 02:46:03 +00005173 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005174 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5175
Eric Christopher455c5772009-12-05 02:46:03 +00005176 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005177 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5178
Eric Christopher455c5772009-12-05 02:46:03 +00005179 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005180 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5181
Eric Christopher455c5772009-12-05 02:46:03 +00005182 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005183 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5184
Eric Christopher455c5772009-12-05 02:46:03 +00005185 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005186 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5187
Eric Christopher455c5772009-12-05 02:46:03 +00005188 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005189 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5190
Reid Spencerf69acf32006-11-19 03:00:14 +00005191 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005192
Reid Spencerc828a0e2006-11-18 21:50:54 +00005193 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
5194</ol>
Reid Spencerc828a0e2006-11-18 21:50:54 +00005195
5196<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005197<pre>
5198 &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
Dan Gohmanc579d972008-09-09 01:02:47 +00005199 &lt;result&gt; = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
5200 &lt;result&gt; = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
5201 &lt;result&gt; = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
Reid Spencerc828a0e2006-11-18 21:50:54 +00005202</pre>
Dan Gohmana5127ab2009-01-22 01:39:38 +00005203
5204<p>Note that the code generator does not yet support vector types with
5205 the <tt>fcmp</tt> instruction.</p>
5206
Reid Spencerc828a0e2006-11-18 21:50:54 +00005207</div>
5208
Reid Spencer97c5fa42006-11-08 01:18:52 +00005209<!-- _______________________________________________________________________ -->
Nate Begemand2195702008-05-12 19:01:56 +00005210<div class="doc_subsubsection">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00005211 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
5212</div>
5213
Reid Spencer97c5fa42006-11-08 01:18:52 +00005214<div class="doc_text">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00005215
Reid Spencer97c5fa42006-11-08 01:18:52 +00005216<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005217<pre>
5218 &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...
5219</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00005220
Reid Spencer97c5fa42006-11-08 01:18:52 +00005221<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005222<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in the
5223 SSA graph representing the function.</p>
5224
Reid Spencer97c5fa42006-11-08 01:18:52 +00005225<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005226<p>The type of the incoming values is specified with the first type field. After
5227 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
5228 one pair for each predecessor basic block of the current block. Only values
5229 of <a href="#t_firstclass">first class</a> type may be used as the value
5230 arguments to the PHI node. Only labels may be used as the label
5231 arguments.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00005232
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005233<p>There must be no non-phi instructions between the start of a basic block and
5234 the PHI instructions: i.e. PHI instructions must be first in a basic
5235 block.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00005236
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005237<p>For the purposes of the SSA form, the use of each incoming value is deemed to
5238 occur on the edge from the corresponding predecessor block to the current
5239 block (but after any definition of an '<tt>invoke</tt>' instruction's return
5240 value on the same edge).</p>
Jay Foad1a4eea52009-06-03 10:20:10 +00005241
Reid Spencer97c5fa42006-11-08 01:18:52 +00005242<h5>Semantics:</h5>
Jeff Cohen222a8a42007-04-29 01:07:00 +00005243<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005244 specified by the pair corresponding to the predecessor basic block that
5245 executed just prior to the current block.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00005246
Reid Spencer97c5fa42006-11-08 01:18:52 +00005247<h5>Example:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00005248<pre>
5249Loop: ; Infinite loop that counts from 0 on up...
5250 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
5251 %nextindvar = add i32 %indvar, 1
5252 br label %Loop
5253</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005254
Reid Spencer97c5fa42006-11-08 01:18:52 +00005255</div>
5256
Chris Lattnerb53c28d2004-03-12 05:50:16 +00005257<!-- _______________________________________________________________________ -->
5258<div class="doc_subsubsection">
5259 <a name="i_select">'<tt>select</tt>' Instruction</a>
5260</div>
5261
5262<div class="doc_text">
5263
5264<h5>Syntax:</h5>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00005265<pre>
Dan Gohmanc579d972008-09-09 01:02:47 +00005266 &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>
5267
Dan Gohmanef9462f2008-10-14 16:51:45 +00005268 <i>selty</i> is either i1 or {&lt;N x i1&gt;}
Chris Lattnerb53c28d2004-03-12 05:50:16 +00005269</pre>
5270
5271<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005272<p>The '<tt>select</tt>' instruction is used to choose one value based on a
5273 condition, without branching.</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00005274
5275
5276<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005277<p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
5278 values indicating the condition, and two values of the
5279 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
5280 vectors and the condition is a scalar, then entire vectors are selected, not
5281 individual elements.</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00005282
5283<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005284<p>If the condition is an i1 and it evaluates to 1, the instruction returns the
5285 first value argument; otherwise, it returns the second value argument.</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00005286
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005287<p>If the condition is a vector of i1, then the value arguments must be vectors
5288 of the same size, and the selection is done element by element.</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00005289
5290<h5>Example:</h5>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00005291<pre>
Reid Spencer36a15422007-01-12 03:35:51 +00005292 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00005293</pre>
Dan Gohmana5127ab2009-01-22 01:39:38 +00005294
5295<p>Note that the code generator does not yet support conditions
5296 with vector type.</p>
5297
Chris Lattnerb53c28d2004-03-12 05:50:16 +00005298</div>
5299
Robert Bocchinof72fdfe2006-01-15 20:48:27 +00005300<!-- _______________________________________________________________________ -->
5301<div class="doc_subsubsection">
Chris Lattnere23c1392005-05-06 05:47:36 +00005302 <a name="i_call">'<tt>call</tt>' Instruction</a>
5303</div>
5304
Misha Brukman76307852003-11-08 01:05:38 +00005305<div class="doc_text">
Chris Lattnere23c1392005-05-06 05:47:36 +00005306
Chris Lattner2f7c9632001-06-06 20:29:01 +00005307<h5>Syntax:</h5>
Chris Lattnere23c1392005-05-06 05:47:36 +00005308<pre>
Devang Patel02256232008-10-07 17:48:33 +00005309 &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 Lattnere23c1392005-05-06 05:47:36 +00005310</pre>
5311
Chris Lattner2f7c9632001-06-06 20:29:01 +00005312<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00005313<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
Chris Lattnere23c1392005-05-06 05:47:36 +00005314
Chris Lattner2f7c9632001-06-06 20:29:01 +00005315<h5>Arguments:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00005316<p>This instruction requires several arguments:</p>
Chris Lattnere23c1392005-05-06 05:47:36 +00005317
Chris Lattnera8292f32002-05-06 22:08:29 +00005318<ol>
Jeffrey Yasskinb8677462010-01-09 19:44:16 +00005319 <li>The optional "tail" marker indicates that the callee function does not
5320 access any allocas or varargs in the caller. Note that calls may be
5321 marked "tail" even if they do not occur before
5322 a <a href="#i_ret"><tt>ret</tt></a> instruction. If the "tail" marker is
5323 present, the function call is eligible for tail call optimization,
5324 but <a href="CodeGenerator.html#tailcallopt">might not in fact be
Evan Cheng59676492010-03-08 21:05:02 +00005325 optimized into a jump</a>. The code generator may optimize calls marked
5326 "tail" with either 1) automatic <a href="CodeGenerator.html#sibcallopt">
5327 sibling call optimization</a> when the caller and callee have
5328 matching signatures, or 2) forced tail call optimization when the
5329 following extra requirements are met:
Jeffrey Yasskinb8677462010-01-09 19:44:16 +00005330 <ul>
5331 <li>Caller and callee both have the calling
5332 convention <tt>fastcc</tt>.</li>
5333 <li>The call is in tail position (ret immediately follows call and ret
5334 uses value of call or is void).</li>
5335 <li>Option <tt>-tailcallopt</tt> is enabled,
Dan Gohman6232f732010-03-02 01:08:11 +00005336 or <code>llvm::GuaranteedTailCallOpt</code> is <code>true</code>.</li>
Jeffrey Yasskinb8677462010-01-09 19:44:16 +00005337 <li><a href="CodeGenerator.html#tailcallopt">Platform specific
5338 constraints are met.</a></li>
5339 </ul>
5340 </li>
Devang Patel7e9b05e2008-10-06 18:50:38 +00005341
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005342 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
5343 convention</a> the call should use. If none is specified, the call
Jeffrey Yasskinb8677462010-01-09 19:44:16 +00005344 defaults to using C calling conventions. The calling convention of the
5345 call must match the calling convention of the target function, or else the
5346 behavior is undefined.</li>
Devang Patel7e9b05e2008-10-06 18:50:38 +00005347
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005348 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
5349 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
5350 '<tt>inreg</tt>' attributes are valid here.</li>
5351
5352 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
5353 type of the return value. Functions that return no value are marked
5354 <tt><a href="#t_void">void</a></tt>.</li>
5355
5356 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
5357 being invoked. The argument types must match the types implied by this
5358 signature. This type can be omitted if the function is not varargs and if
5359 the function type does not return a pointer to a function.</li>
5360
5361 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
5362 be invoked. In most cases, this is a direct function invocation, but
5363 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
5364 to function value.</li>
5365
5366 <li>'<tt>function args</tt>': argument list whose types match the function
Chris Lattner47f2a832010-03-02 06:36:51 +00005367 signature argument types and parameter attributes. All arguments must be
5368 of <a href="#t_firstclass">first class</a> type. If the function
5369 signature indicates the function accepts a variable number of arguments,
5370 the extra arguments can be specified.</li>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005371
5372 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
5373 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
5374 '<tt>readnone</tt>' attributes are valid here.</li>
Chris Lattnera8292f32002-05-06 22:08:29 +00005375</ol>
Chris Lattnere23c1392005-05-06 05:47:36 +00005376
Chris Lattner2f7c9632001-06-06 20:29:01 +00005377<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005378<p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
5379 a specified function, with its incoming arguments bound to the specified
5380 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
5381 function, control flow continues with the instruction after the function
5382 call, and the return value of the function is bound to the result
5383 argument.</p>
Chris Lattnere23c1392005-05-06 05:47:36 +00005384
Chris Lattner2f7c9632001-06-06 20:29:01 +00005385<h5>Example:</h5>
Chris Lattnere23c1392005-05-06 05:47:36 +00005386<pre>
Nick Lewyckya9b13d52007-09-08 13:57:50 +00005387 %retval = call i32 @test(i32 %argc)
Dan Gohmanaabfdb32010-05-28 17:13:49 +00005388 call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42) <i>; yields i32</i>
Chris Lattnerfb7c88d2008-03-21 17:24:17 +00005389 %X = tail call i32 @foo() <i>; yields i32</i>
5390 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
5391 call void %foo(i8 97 signext)
Devang Pateld6cff512008-03-10 20:49:15 +00005392
5393 %struct.A = type { i32, i8 }
Devang Patel7e9b05e2008-10-06 18:50:38 +00005394 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
Dan Gohmancc3132e2008-10-04 19:00:07 +00005395 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
5396 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
Chris Lattner6cbe8e92008-10-08 06:26:11 +00005397 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
Matthijs Kooijmaneefa7df2008-10-07 10:03:45 +00005398 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
Chris Lattnere23c1392005-05-06 05:47:36 +00005399</pre>
5400
Dale Johannesen68f971b2009-09-24 18:38:21 +00005401<p>llvm treats calls to some functions with names and arguments that match the
Dale Johannesen722212d2009-09-25 17:04:42 +00005402standard C99 library as being the C99 library functions, and may perform
5403optimizations or generate code for them under that assumption. This is
5404something we'd like to change in the future to provide better support for
Dan Gohmana269a0a2010-03-01 17:41:39 +00005405freestanding environments and non-C-based languages.</p>
Dale Johannesen68f971b2009-09-24 18:38:21 +00005406
Misha Brukman76307852003-11-08 01:05:38 +00005407</div>
Chris Lattner6a4a0492004-09-27 21:51:25 +00005408
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00005409<!-- _______________________________________________________________________ -->
Chris Lattner6a4a0492004-09-27 21:51:25 +00005410<div class="doc_subsubsection">
Chris Lattner33337472006-01-13 23:26:01 +00005411 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
Chris Lattner6a4a0492004-09-27 21:51:25 +00005412</div>
5413
Misha Brukman76307852003-11-08 01:05:38 +00005414<div class="doc_text">
Chris Lattner6a4a0492004-09-27 21:51:25 +00005415
Chris Lattner26ca62e2003-10-18 05:51:36 +00005416<h5>Syntax:</h5>
Chris Lattner6a4a0492004-09-27 21:51:25 +00005417<pre>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00005418 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
Chris Lattner6a4a0492004-09-27 21:51:25 +00005419</pre>
5420
Chris Lattner26ca62e2003-10-18 05:51:36 +00005421<h5>Overview:</h5>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00005422<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005423 the "variable argument" area of a function call. It is used to implement the
5424 <tt>va_arg</tt> macro in C.</p>
Chris Lattner6a4a0492004-09-27 21:51:25 +00005425
Chris Lattner26ca62e2003-10-18 05:51:36 +00005426<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005427<p>This instruction takes a <tt>va_list*</tt> value and the type of the
5428 argument. It returns a value of the specified argument type and increments
5429 the <tt>va_list</tt> to point to the next argument. The actual type
5430 of <tt>va_list</tt> is target specific.</p>
Chris Lattner6a4a0492004-09-27 21:51:25 +00005431
Chris Lattner26ca62e2003-10-18 05:51:36 +00005432<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005433<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
5434 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
5435 to the next argument. For more information, see the variable argument
5436 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
Chris Lattner6a4a0492004-09-27 21:51:25 +00005437
5438<p>It is legal for this instruction to be called in a function which does not
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005439 take a variable number of arguments, for example, the <tt>vfprintf</tt>
5440 function.</p>
Chris Lattner6a4a0492004-09-27 21:51:25 +00005441
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005442<p><tt>va_arg</tt> is an LLVM instruction instead of
5443 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
5444 argument.</p>
Chris Lattner6a4a0492004-09-27 21:51:25 +00005445
Chris Lattner26ca62e2003-10-18 05:51:36 +00005446<h5>Example:</h5>
Chris Lattner6a4a0492004-09-27 21:51:25 +00005447<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
5448
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005449<p>Note that the code generator does not yet fully support va_arg on many
5450 targets. Also, it does not currently support va_arg with aggregate types on
5451 any target.</p>
Dan Gohman3065b612009-01-12 23:12:39 +00005452
Misha Brukman76307852003-11-08 01:05:38 +00005453</div>
Chris Lattner941515c2004-01-06 05:31:32 +00005454
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00005455<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +00005456<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
5457<!-- *********************************************************************** -->
Chris Lattner941515c2004-01-06 05:31:32 +00005458
Misha Brukman76307852003-11-08 01:05:38 +00005459<div class="doc_text">
Chris Lattnerfee11462004-02-12 17:01:32 +00005460
5461<p>LLVM supports the notion of an "intrinsic function". These functions have
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005462 well known names and semantics and are required to follow certain
5463 restrictions. Overall, these intrinsics represent an extension mechanism for
5464 the LLVM language that does not require changing all of the transformations
5465 in LLVM when adding to the language (or the bitcode reader/writer, the
5466 parser, etc...).</p>
Chris Lattnerfee11462004-02-12 17:01:32 +00005467
John Criswell88190562005-05-16 16:17:45 +00005468<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005469 prefix is reserved in LLVM for intrinsic names; thus, function names may not
5470 begin with this prefix. Intrinsic functions must always be external
5471 functions: you cannot define the body of intrinsic functions. Intrinsic
5472 functions may only be used in call or invoke instructions: it is illegal to
5473 take the address of an intrinsic function. Additionally, because intrinsic
5474 functions are part of the LLVM language, it is required if any are added that
5475 they be documented here.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +00005476
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005477<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
5478 family of functions that perform the same operation but on different data
5479 types. Because LLVM can represent over 8 million different integer types,
5480 overloading is used commonly to allow an intrinsic function to operate on any
5481 integer type. One or more of the argument types or the result type can be
5482 overloaded to accept any integer type. Argument types may also be defined as
5483 exactly matching a previous argument's type or the result type. This allows
5484 an intrinsic function which accepts multiple arguments, but needs all of them
5485 to be of the same type, to only be overloaded with respect to a single
5486 argument or the result.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +00005487
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005488<p>Overloaded intrinsics will have the names of its overloaded argument types
5489 encoded into its function name, each preceded by a period. Only those types
5490 which are overloaded result in a name suffix. Arguments whose type is matched
5491 against another type do not. For example, the <tt>llvm.ctpop</tt> function
5492 can take an integer of any width and returns an integer of exactly the same
5493 integer width. This leads to a family of functions such as
5494 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
5495 %val)</tt>. Only one type, the return type, is overloaded, and only one type
5496 suffix is required. Because the argument's type is matched against the return
5497 type, it does not require its own name suffix.</p>
Reid Spencer4eefaab2007-04-01 08:04:23 +00005498
Eric Christopher455c5772009-12-05 02:46:03 +00005499<p>To learn how to add an intrinsic function, please see the
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005500 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +00005501
Misha Brukman76307852003-11-08 01:05:38 +00005502</div>
Chris Lattner941515c2004-01-06 05:31:32 +00005503
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00005504<!-- ======================================================================= -->
Chris Lattner941515c2004-01-06 05:31:32 +00005505<div class="doc_subsection">
5506 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
5507</div>
5508
Misha Brukman76307852003-11-08 01:05:38 +00005509<div class="doc_text">
Chris Lattner757528b0b2004-05-23 21:06:01 +00005510
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005511<p>Variable argument support is defined in LLVM with
5512 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
5513 intrinsic functions. These functions are related to the similarly named
5514 macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005515
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005516<p>All of these functions operate on arguments that use a target-specific value
5517 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
5518 not define what this type is, so all transformations should be prepared to
5519 handle these functions regardless of the type used.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005520
Chris Lattner30b868d2006-05-15 17:26:46 +00005521<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005522 instruction and the variable argument handling intrinsic functions are
5523 used.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005524
Benjamin Kramer79698be2010-07-13 12:26:09 +00005525<pre class="doc_code">
Anton Korobeynikov640bbe02007-03-21 23:58:04 +00005526define i32 @test(i32 %X, ...) {
Chris Lattnerfee11462004-02-12 17:01:32 +00005527 ; Initialize variable argument processing
Jeff Cohen222a8a42007-04-29 01:07:00 +00005528 %ap = alloca i8*
Chris Lattnerdb0790c2007-01-08 07:55:15 +00005529 %ap2 = bitcast i8** %ap to i8*
Anton Korobeynikov640bbe02007-03-21 23:58:04 +00005530 call void @llvm.va_start(i8* %ap2)
Chris Lattnerfee11462004-02-12 17:01:32 +00005531
5532 ; Read a single integer argument
Jeff Cohen222a8a42007-04-29 01:07:00 +00005533 %tmp = va_arg i8** %ap, i32
Chris Lattnerfee11462004-02-12 17:01:32 +00005534
5535 ; Demonstrate usage of llvm.va_copy and llvm.va_end
Jeff Cohen222a8a42007-04-29 01:07:00 +00005536 %aq = alloca i8*
Chris Lattnerdb0790c2007-01-08 07:55:15 +00005537 %aq2 = bitcast i8** %aq to i8*
Jeff Cohen222a8a42007-04-29 01:07:00 +00005538 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
Anton Korobeynikov640bbe02007-03-21 23:58:04 +00005539 call void @llvm.va_end(i8* %aq2)
Chris Lattnerfee11462004-02-12 17:01:32 +00005540
5541 ; Stop processing of arguments.
Anton Korobeynikov640bbe02007-03-21 23:58:04 +00005542 call void @llvm.va_end(i8* %ap2)
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00005543 ret i32 %tmp
Chris Lattnerfee11462004-02-12 17:01:32 +00005544}
Anton Korobeynikov640bbe02007-03-21 23:58:04 +00005545
5546declare void @llvm.va_start(i8*)
5547declare void @llvm.va_copy(i8*, i8*)
5548declare void @llvm.va_end(i8*)
Chris Lattnerfee11462004-02-12 17:01:32 +00005549</pre>
Chris Lattner941515c2004-01-06 05:31:32 +00005550
Bill Wendling3716c5d2007-05-29 09:04:49 +00005551</div>
5552
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00005553<!-- _______________________________________________________________________ -->
Chris Lattner941515c2004-01-06 05:31:32 +00005554<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005555 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
Chris Lattner941515c2004-01-06 05:31:32 +00005556</div>
5557
5558
Misha Brukman76307852003-11-08 01:05:38 +00005559<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005560
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00005561<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005562<pre>
5563 declare void %llvm.va_start(i8* &lt;arglist&gt;)
5564</pre>
5565
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00005566<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005567<p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*&lt;arglist&gt;</tt>
5568 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00005569
5570<h5>Arguments:</h5>
Dan Gohmanef9462f2008-10-14 16:51:45 +00005571<p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00005572
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00005573<h5>Semantics:</h5>
Dan Gohmanef9462f2008-10-14 16:51:45 +00005574<p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005575 macro available in C. In a target-dependent way, it initializes
5576 the <tt>va_list</tt> element to which the argument points, so that the next
5577 call to <tt>va_arg</tt> will produce the first variable argument passed to
5578 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
5579 need to know the last argument of the function as the compiler can figure
5580 that out.</p>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00005581
Misha Brukman76307852003-11-08 01:05:38 +00005582</div>
Chris Lattner941515c2004-01-06 05:31:32 +00005583
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00005584<!-- _______________________________________________________________________ -->
Chris Lattner941515c2004-01-06 05:31:32 +00005585<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005586 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
Chris Lattner941515c2004-01-06 05:31:32 +00005587</div>
5588
Misha Brukman76307852003-11-08 01:05:38 +00005589<div class="doc_text">
Chris Lattnerdb0790c2007-01-08 07:55:15 +00005590
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005591<h5>Syntax:</h5>
5592<pre>
5593 declare void @llvm.va_end(i8* &lt;arglist&gt;)
5594</pre>
5595
5596<h5>Overview:</h5>
Jeff Cohen222a8a42007-04-29 01:07:00 +00005597<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005598 which has been initialized previously
5599 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
5600 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00005601
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00005602<h5>Arguments:</h5>
Jeff Cohen222a8a42007-04-29 01:07:00 +00005603<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00005604
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00005605<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00005606<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005607 macro available in C. In a target-dependent way, it destroys
5608 the <tt>va_list</tt> element to which the argument points. Calls
5609 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
5610 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
5611 with calls to <tt>llvm.va_end</tt>.</p>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00005612
Misha Brukman76307852003-11-08 01:05:38 +00005613</div>
Chris Lattner941515c2004-01-06 05:31:32 +00005614
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00005615<!-- _______________________________________________________________________ -->
Chris Lattner941515c2004-01-06 05:31:32 +00005616<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005617 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
Chris Lattner941515c2004-01-06 05:31:32 +00005618</div>
5619
Misha Brukman76307852003-11-08 01:05:38 +00005620<div class="doc_text">
Chris Lattner757528b0b2004-05-23 21:06:01 +00005621
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00005622<h5>Syntax:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005623<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00005624 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
Chris Lattner757528b0b2004-05-23 21:06:01 +00005625</pre>
5626
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00005627<h5>Overview:</h5>
Jeff Cohen222a8a42007-04-29 01:07:00 +00005628<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005629 from the source argument list to the destination argument list.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005630
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00005631<h5>Arguments:</h5>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00005632<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005633 The second argument is a pointer to a <tt>va_list</tt> element to copy
5634 from.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005635
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00005636<h5>Semantics:</h5>
Jeff Cohen222a8a42007-04-29 01:07:00 +00005637<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005638 macro available in C. In a target-dependent way, it copies the
5639 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
5640 element. This intrinsic is necessary because
5641 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
5642 arbitrarily complex and require, for example, memory allocation.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005643
Misha Brukman76307852003-11-08 01:05:38 +00005644</div>
Chris Lattner941515c2004-01-06 05:31:32 +00005645
Chris Lattnerfee11462004-02-12 17:01:32 +00005646<!-- ======================================================================= -->
5647<div class="doc_subsection">
Chris Lattner757528b0b2004-05-23 21:06:01 +00005648 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
5649</div>
5650
5651<div class="doc_text">
5652
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005653<p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
Chris Lattner67c37d12008-08-05 18:29:16 +00005654Collection</a> (GC) requires the implementation and generation of these
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005655intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
5656roots on the stack</a>, as well as garbage collector implementations that
5657require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
5658barriers. Front-ends for type-safe garbage collected languages should generate
5659these intrinsics to make use of the LLVM garbage collectors. For more details,
5660see <a href="GarbageCollection.html">Accurate Garbage Collection with
5661LLVM</a>.</p>
Christopher Lamb55c6d4f2007-12-17 01:00:21 +00005662
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005663<p>The garbage collection intrinsics only operate on objects in the generic
5664 address space (address space zero).</p>
Christopher Lamb55c6d4f2007-12-17 01:00:21 +00005665
Chris Lattner757528b0b2004-05-23 21:06:01 +00005666</div>
5667
5668<!-- _______________________________________________________________________ -->
5669<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005670 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005671</div>
5672
5673<div class="doc_text">
5674
5675<h5>Syntax:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005676<pre>
Chris Lattner12477732007-09-21 17:30:40 +00005677 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Chris Lattner757528b0b2004-05-23 21:06:01 +00005678</pre>
5679
5680<h5>Overview:</h5>
John Criswelldfe6a862004-12-10 15:51:16 +00005681<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005682 the code generator, and allows some metadata to be associated with it.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005683
5684<h5>Arguments:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005685<p>The first argument specifies the address of a stack object that contains the
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005686 root pointer. The second pointer (which must be either a constant or a
5687 global value address) contains the meta-data to be associated with the
5688 root.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005689
5690<h5>Semantics:</h5>
Chris Lattner851b7712008-04-24 05:59:56 +00005691<p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005692 location. At compile-time, the code generator generates information to allow
5693 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5694 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5695 algorithm</a>.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005696
5697</div>
5698
Chris Lattner757528b0b2004-05-23 21:06:01 +00005699<!-- _______________________________________________________________________ -->
5700<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005701 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005702</div>
5703
5704<div class="doc_text">
5705
5706<h5>Syntax:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005707<pre>
Chris Lattner12477732007-09-21 17:30:40 +00005708 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Chris Lattner757528b0b2004-05-23 21:06:01 +00005709</pre>
5710
5711<h5>Overview:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005712<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005713 locations, allowing garbage collector implementations that require read
5714 barriers.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005715
5716<h5>Arguments:</h5>
Chris Lattnerf9228072006-03-14 20:02:51 +00005717<p>The second argument is the address to read from, which should be an address
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005718 allocated from the garbage collector. The first object is a pointer to the
5719 start of the referenced object, if needed by the language runtime (otherwise
5720 null).</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005721
5722<h5>Semantics:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005723<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005724 instruction, but may be replaced with substantially more complex code by the
5725 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5726 may only be used in a function which <a href="#gc">specifies a GC
5727 algorithm</a>.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005728
5729</div>
5730
Chris Lattner757528b0b2004-05-23 21:06:01 +00005731<!-- _______________________________________________________________________ -->
5732<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005733 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005734</div>
5735
5736<div class="doc_text">
5737
5738<h5>Syntax:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005739<pre>
Chris Lattner12477732007-09-21 17:30:40 +00005740 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Chris Lattner757528b0b2004-05-23 21:06:01 +00005741</pre>
5742
5743<h5>Overview:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005744<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005745 locations, allowing garbage collector implementations that require write
5746 barriers (such as generational or reference counting collectors).</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005747
5748<h5>Arguments:</h5>
Chris Lattnerf9228072006-03-14 20:02:51 +00005749<p>The first argument is the reference to store, the second is the start of the
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005750 object to store it to, and the third is the address of the field of Obj to
5751 store to. If the runtime does not require a pointer to the object, Obj may
5752 be null.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005753
5754<h5>Semantics:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005755<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005756 instruction, but may be replaced with substantially more complex code by the
5757 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5758 may only be used in a function which <a href="#gc">specifies a GC
5759 algorithm</a>.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005760
5761</div>
5762
Chris Lattner757528b0b2004-05-23 21:06:01 +00005763<!-- ======================================================================= -->
5764<div class="doc_subsection">
Chris Lattner3649c3a2004-02-14 04:08:35 +00005765 <a name="int_codegen">Code Generator Intrinsics</a>
5766</div>
5767
5768<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005769
5770<p>These intrinsics are provided by LLVM to expose special features that may
5771 only be implemented with code generator support.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00005772
5773</div>
5774
5775<!-- _______________________________________________________________________ -->
5776<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005777 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
Chris Lattner3649c3a2004-02-14 04:08:35 +00005778</div>
5779
5780<div class="doc_text">
5781
5782<h5>Syntax:</h5>
5783<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00005784 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
Chris Lattner3649c3a2004-02-14 04:08:35 +00005785</pre>
5786
5787<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005788<p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5789 target-specific value indicating the return address of the current function
5790 or one of its callers.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00005791
5792<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005793<p>The argument to this intrinsic indicates which function to return the address
5794 for. Zero indicates the calling function, one indicates its caller, etc.
5795 The argument is <b>required</b> to be a constant integer value.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00005796
5797<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005798<p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
5799 indicating the return address of the specified call frame, or zero if it
5800 cannot be identified. The value returned by this intrinsic is likely to be
5801 incorrect or 0 for arguments other than zero, so it should only be used for
5802 debugging purposes.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00005803
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005804<p>Note that calling this intrinsic does not prevent function inlining or other
5805 aggressive transformations, so the value returned may not be that of the
5806 obvious source-language caller.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00005807
Chris Lattner3649c3a2004-02-14 04:08:35 +00005808</div>
5809
Chris Lattner3649c3a2004-02-14 04:08:35 +00005810<!-- _______________________________________________________________________ -->
5811<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005812 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
Chris Lattner3649c3a2004-02-14 04:08:35 +00005813</div>
5814
5815<div class="doc_text">
5816
5817<h5>Syntax:</h5>
5818<pre>
Dan Gohmanaabfdb32010-05-28 17:13:49 +00005819 declare i8* @llvm.frameaddress(i32 &lt;level&gt;)
Chris Lattner3649c3a2004-02-14 04:08:35 +00005820</pre>
5821
5822<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005823<p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5824 target-specific frame pointer value for the specified stack frame.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00005825
5826<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005827<p>The argument to this intrinsic indicates which function to return the frame
5828 pointer for. Zero indicates the calling function, one indicates its caller,
5829 etc. The argument is <b>required</b> to be a constant integer value.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00005830
5831<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005832<p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
5833 indicating the frame address of the specified call frame, or zero if it
5834 cannot be identified. The value returned by this intrinsic is likely to be
5835 incorrect or 0 for arguments other than zero, so it should only be used for
5836 debugging purposes.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00005837
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005838<p>Note that calling this intrinsic does not prevent function inlining or other
5839 aggressive transformations, so the value returned may not be that of the
5840 obvious source-language caller.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00005841
Chris Lattner3649c3a2004-02-14 04:08:35 +00005842</div>
5843
Chris Lattnerc8a2c222005-02-28 19:24:19 +00005844<!-- _______________________________________________________________________ -->
5845<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005846 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
Chris Lattner2f0f0012006-01-13 02:03:13 +00005847</div>
5848
5849<div class="doc_text">
5850
5851<h5>Syntax:</h5>
5852<pre>
Dan Gohmanaabfdb32010-05-28 17:13:49 +00005853 declare i8* @llvm.stacksave()
Chris Lattner2f0f0012006-01-13 02:03:13 +00005854</pre>
5855
5856<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005857<p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
5858 of the function stack, for use
5859 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
5860 useful for implementing language features like scoped automatic variable
5861 sized arrays in C99.</p>
Chris Lattner2f0f0012006-01-13 02:03:13 +00005862
5863<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005864<p>This intrinsic returns a opaque pointer value that can be passed
5865 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
5866 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
5867 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
5868 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
5869 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
5870 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
Chris Lattner2f0f0012006-01-13 02:03:13 +00005871
5872</div>
5873
5874<!-- _______________________________________________________________________ -->
5875<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005876 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
Chris Lattner2f0f0012006-01-13 02:03:13 +00005877</div>
5878
5879<div class="doc_text">
5880
5881<h5>Syntax:</h5>
5882<pre>
Dan Gohmanaabfdb32010-05-28 17:13:49 +00005883 declare void @llvm.stackrestore(i8* %ptr)
Chris Lattner2f0f0012006-01-13 02:03:13 +00005884</pre>
5885
5886<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005887<p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5888 the function stack to the state it was in when the
5889 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
5890 executed. This is useful for implementing language features like scoped
5891 automatic variable sized arrays in C99.</p>
Chris Lattner2f0f0012006-01-13 02:03:13 +00005892
5893<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005894<p>See the description
5895 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
Chris Lattner2f0f0012006-01-13 02:03:13 +00005896
5897</div>
5898
Chris Lattner2f0f0012006-01-13 02:03:13 +00005899<!-- _______________________________________________________________________ -->
5900<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005901 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
Chris Lattnerc8a2c222005-02-28 19:24:19 +00005902</div>
5903
5904<div class="doc_text">
5905
5906<h5>Syntax:</h5>
5907<pre>
Chris Lattner12477732007-09-21 17:30:40 +00005908 declare void @llvm.prefetch(i8* &lt;address&gt;, i32 &lt;rw&gt;, i32 &lt;locality&gt;)
Chris Lattnerc8a2c222005-02-28 19:24:19 +00005909</pre>
5910
5911<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005912<p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
5913 insert a prefetch instruction if supported; otherwise, it is a noop.
5914 Prefetches have no effect on the behavior of the program but can change its
5915 performance characteristics.</p>
Chris Lattnerc8a2c222005-02-28 19:24:19 +00005916
5917<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005918<p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
5919 specifier determining if the fetch should be for a read (0) or write (1),
5920 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5921 locality, to (3) - extremely local keep in cache. The <tt>rw</tt>
5922 and <tt>locality</tt> arguments must be constant integers.</p>
Chris Lattnerc8a2c222005-02-28 19:24:19 +00005923
5924<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005925<p>This intrinsic does not modify the behavior of the program. In particular,
5926 prefetches cannot trap and do not produce a value. On targets that support
5927 this intrinsic, the prefetch can provide hints to the processor cache for
5928 better performance.</p>
Chris Lattnerc8a2c222005-02-28 19:24:19 +00005929
5930</div>
5931
Andrew Lenharthb4427912005-03-28 20:05:49 +00005932<!-- _______________________________________________________________________ -->
5933<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005934 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
Andrew Lenharthb4427912005-03-28 20:05:49 +00005935</div>
5936
5937<div class="doc_text">
5938
5939<h5>Syntax:</h5>
5940<pre>
Chris Lattner12477732007-09-21 17:30:40 +00005941 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Andrew Lenharthb4427912005-03-28 20:05:49 +00005942</pre>
5943
5944<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005945<p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
5946 Counter (PC) in a region of code to simulators and other tools. The method
5947 is target specific, but it is expected that the marker will use exported
5948 symbols to transmit the PC of the marker. The marker makes no guarantees
5949 that it will remain with any specific instruction after optimizations. It is
5950 possible that the presence of a marker will inhibit optimizations. The
5951 intended use is to be inserted after optimizations to allow correlations of
5952 simulation runs.</p>
Andrew Lenharthb4427912005-03-28 20:05:49 +00005953
5954<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005955<p><tt>id</tt> is a numerical id identifying the marker.</p>
Andrew Lenharthb4427912005-03-28 20:05:49 +00005956
5957<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005958<p>This intrinsic does not modify the behavior of the program. Backends that do
Dan Gohmana269a0a2010-03-01 17:41:39 +00005959 not support this intrinsic may ignore it.</p>
Andrew Lenharthb4427912005-03-28 20:05:49 +00005960
5961</div>
5962
Andrew Lenharth01aa5632005-11-11 16:47:30 +00005963<!-- _______________________________________________________________________ -->
5964<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005965 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
Andrew Lenharth01aa5632005-11-11 16:47:30 +00005966</div>
5967
5968<div class="doc_text">
5969
5970<h5>Syntax:</h5>
5971<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00005972 declare i64 @llvm.readcyclecounter()
Andrew Lenharth01aa5632005-11-11 16:47:30 +00005973</pre>
5974
5975<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005976<p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5977 counter register (or similar low latency, high accuracy clocks) on those
5978 targets that support it. On X86, it should map to RDTSC. On Alpha, it
5979 should map to RPCC. As the backing counters overflow quickly (on the order
5980 of 9 seconds on alpha), this should only be used for small timings.</p>
Andrew Lenharth01aa5632005-11-11 16:47:30 +00005981
5982<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005983<p>When directly supported, reading the cycle counter should not modify any
5984 memory. Implementations are allowed to either return a application specific
5985 value or a system wide value. On backends without support, this is lowered
5986 to a constant 0.</p>
Andrew Lenharth01aa5632005-11-11 16:47:30 +00005987
5988</div>
5989
Chris Lattner3649c3a2004-02-14 04:08:35 +00005990<!-- ======================================================================= -->
5991<div class="doc_subsection">
Chris Lattnerfee11462004-02-12 17:01:32 +00005992 <a name="int_libc">Standard C Library Intrinsics</a>
5993</div>
5994
5995<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00005996
5997<p>LLVM provides intrinsics for a few important standard C library functions.
5998 These intrinsics allow source-language front-ends to pass information about
5999 the alignment of the pointer arguments to the code generator, providing
6000 opportunity for more efficient code generation.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +00006001
6002</div>
6003
6004<!-- _______________________________________________________________________ -->
6005<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00006006 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
Chris Lattnerfee11462004-02-12 17:01:32 +00006007</div>
6008
6009<div class="doc_text">
6010
6011<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006012<p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
Mon P Wang508127b2010-04-07 06:35:53 +00006013 integer bit width and for different address spaces. Not all targets support
6014 all bit widths however.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006015
Chris Lattnerfee11462004-02-12 17:01:32 +00006016<pre>
Dan Gohmanaabfdb32010-05-28 17:13:49 +00006017 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* &lt;dest&gt;, i8* &lt;src&gt;,
Chris Lattnerbd4ca622010-04-08 00:53:57 +00006018 i32 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Dan Gohmanaabfdb32010-05-28 17:13:49 +00006019 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* &lt;dest&gt;, i8* &lt;src&gt;,
Chris Lattnerbd4ca622010-04-08 00:53:57 +00006020 i64 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Chris Lattnerfee11462004-02-12 17:01:32 +00006021</pre>
6022
6023<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006024<p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6025 source location to the destination location.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +00006026
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006027<p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
Chris Lattnerbd4ca622010-04-08 00:53:57 +00006028 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6029 and the pointers can be in specified address spaces.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +00006030
6031<h5>Arguments:</h5>
Chris Lattnerbd4ca622010-04-08 00:53:57 +00006032
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006033<p>The first argument is a pointer to the destination, the second is a pointer
6034 to the source. The third argument is an integer argument specifying the
Chris Lattnerbd4ca622010-04-08 00:53:57 +00006035 number of bytes to copy, the fourth argument is the alignment of the
6036 source and destination locations, and the fifth is a boolean indicating a
6037 volatile access.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +00006038
Dan Gohmana269a0a2010-03-01 17:41:39 +00006039<p>If the call to this intrinsic has an alignment value that is not 0 or 1,
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006040 then the caller guarantees that both the source and destination pointers are
6041 aligned to that boundary.</p>
Chris Lattner4c67c482004-02-12 21:18:15 +00006042
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +00006043<p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6044 <tt>llvm.memcpy</tt> call is a <a href="#volatile">volatile operation</a>.
6045 The detailed access behavior is not very cleanly specified and it is unwise
6046 to depend on it.</p>
Chris Lattnerbd4ca622010-04-08 00:53:57 +00006047
Chris Lattnerfee11462004-02-12 17:01:32 +00006048<h5>Semantics:</h5>
Chris Lattnerbd4ca622010-04-08 00:53:57 +00006049
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006050<p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6051 source location to the destination location, which are not allowed to
6052 overlap. It copies "len" bytes of memory over. If the argument is known to
6053 be aligned to some boundary, this can be specified as the fourth argument,
6054 otherwise it should be set to 0 or 1.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +00006055
Chris Lattnerfee11462004-02-12 17:01:32 +00006056</div>
6057
Chris Lattnerf30152e2004-02-12 18:10:10 +00006058<!-- _______________________________________________________________________ -->
6059<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00006060 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
Chris Lattnerf30152e2004-02-12 18:10:10 +00006061</div>
6062
6063<div class="doc_text">
6064
6065<h5>Syntax:</h5>
Chris Lattnerdd708342008-11-21 16:42:48 +00006066<p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
Mon P Wang508127b2010-04-07 06:35:53 +00006067 width and for different address space. Not all targets support all bit
6068 widths however.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006069
Chris Lattnerf30152e2004-02-12 18:10:10 +00006070<pre>
Dan Gohmanaabfdb32010-05-28 17:13:49 +00006071 declare void @llvm.memmove.p0i8.p0i8.i32(i8* &lt;dest&gt;, i8* &lt;src&gt;,
Chris Lattnerbd4ca622010-04-08 00:53:57 +00006072 i32 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Dan Gohmanaabfdb32010-05-28 17:13:49 +00006073 declare void @llvm.memmove.p0i8.p0i8.i64(i8* &lt;dest&gt;, i8* &lt;src&gt;,
Chris Lattnerbd4ca622010-04-08 00:53:57 +00006074 i64 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Chris Lattnerf30152e2004-02-12 18:10:10 +00006075</pre>
6076
6077<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006078<p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
6079 source location to the destination location. It is similar to the
6080 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
6081 overlap.</p>
Chris Lattnerf30152e2004-02-12 18:10:10 +00006082
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006083<p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
Chris Lattnerbd4ca622010-04-08 00:53:57 +00006084 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6085 and the pointers can be in specified address spaces.</p>
Chris Lattnerf30152e2004-02-12 18:10:10 +00006086
6087<h5>Arguments:</h5>
Chris Lattnerbd4ca622010-04-08 00:53:57 +00006088
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006089<p>The first argument is a pointer to the destination, the second is a pointer
6090 to the source. The third argument is an integer argument specifying the
Chris Lattnerbd4ca622010-04-08 00:53:57 +00006091 number of bytes to copy, the fourth argument is the alignment of the
6092 source and destination locations, and the fifth is a boolean indicating a
6093 volatile access.</p>
Chris Lattnerf30152e2004-02-12 18:10:10 +00006094
Dan Gohmana269a0a2010-03-01 17:41:39 +00006095<p>If the call to this intrinsic has an alignment value that is not 0 or 1,
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006096 then the caller guarantees that the source and destination pointers are
6097 aligned to that boundary.</p>
Chris Lattner4c67c482004-02-12 21:18:15 +00006098
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +00006099<p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6100 <tt>llvm.memmove</tt> call is a <a href="#volatile">volatile operation</a>.
6101 The detailed access behavior is not very cleanly specified and it is unwise
6102 to depend on it.</p>
Chris Lattnerbd4ca622010-04-08 00:53:57 +00006103
Chris Lattnerf30152e2004-02-12 18:10:10 +00006104<h5>Semantics:</h5>
Chris Lattnerbd4ca622010-04-08 00:53:57 +00006105
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006106<p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
6107 source location to the destination location, which may overlap. It copies
6108 "len" bytes of memory over. If the argument is known to be aligned to some
6109 boundary, this can be specified as the fourth argument, otherwise it should
6110 be set to 0 or 1.</p>
Chris Lattnerf30152e2004-02-12 18:10:10 +00006111
Chris Lattnerf30152e2004-02-12 18:10:10 +00006112</div>
6113
Chris Lattner3649c3a2004-02-14 04:08:35 +00006114<!-- _______________________________________________________________________ -->
6115<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00006116 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
Chris Lattner3649c3a2004-02-14 04:08:35 +00006117</div>
6118
6119<div class="doc_text">
6120
6121<h5>Syntax:</h5>
Chris Lattnerdd708342008-11-21 16:42:48 +00006122<p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
John Criswellad05ae42010-07-30 16:30:28 +00006123 width and for different address spaces. However, not all targets support all
6124 bit widths.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006125
Chris Lattner3649c3a2004-02-14 04:08:35 +00006126<pre>
Dan Gohmanaabfdb32010-05-28 17:13:49 +00006127 declare void @llvm.memset.p0i8.i32(i8* &lt;dest&gt;, i8 &lt;val&gt;,
Chris Lattner685db9d2010-04-08 00:54:34 +00006128 i32 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Dan Gohmanaabfdb32010-05-28 17:13:49 +00006129 declare void @llvm.memset.p0i8.i64(i8* &lt;dest&gt;, i8 &lt;val&gt;,
Chris Lattner685db9d2010-04-08 00:54:34 +00006130 i64 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Chris Lattner3649c3a2004-02-14 04:08:35 +00006131</pre>
6132
6133<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006134<p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
6135 particular byte value.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00006136
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006137<p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
John Criswellad05ae42010-07-30 16:30:28 +00006138 intrinsic does not return a value and takes extra alignment/volatile
6139 arguments. Also, the destination can be in an arbitrary address space.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00006140
6141<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006142<p>The first argument is a pointer to the destination to fill, the second is the
John Criswellad05ae42010-07-30 16:30:28 +00006143 byte value with which to fill it, the third argument is an integer argument
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006144 specifying the number of bytes to fill, and the fourth argument is the known
John Criswellad05ae42010-07-30 16:30:28 +00006145 alignment of the destination location.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00006146
Dan Gohmana269a0a2010-03-01 17:41:39 +00006147<p>If the call to this intrinsic has an alignment value that is not 0 or 1,
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006148 then the caller guarantees that the destination pointer is aligned to that
6149 boundary.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00006150
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +00006151<p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6152 <tt>llvm.memset</tt> call is a <a href="#volatile">volatile operation</a>.
6153 The detailed access behavior is not very cleanly specified and it is unwise
6154 to depend on it.</p>
Chris Lattnerbd4ca622010-04-08 00:53:57 +00006155
Chris Lattner3649c3a2004-02-14 04:08:35 +00006156<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006157<p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
6158 at the destination location. If the argument is known to be aligned to some
6159 boundary, this can be specified as the fourth argument, otherwise it should
6160 be set to 0 or 1.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00006161
Chris Lattner3649c3a2004-02-14 04:08:35 +00006162</div>
6163
Chris Lattner3b4f4372004-06-11 02:28:03 +00006164<!-- _______________________________________________________________________ -->
6165<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00006166 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
Chris Lattner8a8f2e52005-07-21 01:29:16 +00006167</div>
6168
6169<div class="doc_text">
6170
6171<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006172<p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
6173 floating point or vector of floating point type. Not all targets support all
6174 types however.</p>
6175
Chris Lattner8a8f2e52005-07-21 01:29:16 +00006176<pre>
Dale Johannesendd89d272007-10-02 17:47:38 +00006177 declare float @llvm.sqrt.f32(float %Val)
6178 declare double @llvm.sqrt.f64(double %Val)
6179 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
6180 declare fp128 @llvm.sqrt.f128(fp128 %Val)
6181 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Chris Lattner8a8f2e52005-07-21 01:29:16 +00006182</pre>
6183
6184<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006185<p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
6186 returning the same value as the libm '<tt>sqrt</tt>' functions would.
6187 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
6188 behavior for negative numbers other than -0.0 (which allows for better
6189 optimization, because there is no need to worry about errno being
6190 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
Chris Lattner8a8f2e52005-07-21 01:29:16 +00006191
6192<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006193<p>The argument and return value are floating point numbers of the same
6194 type.</p>
Chris Lattner8a8f2e52005-07-21 01:29:16 +00006195
6196<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006197<p>This function returns the sqrt of the specified operand if it is a
6198 nonnegative floating point number.</p>
Chris Lattner8a8f2e52005-07-21 01:29:16 +00006199
Chris Lattner8a8f2e52005-07-21 01:29:16 +00006200</div>
6201
Chris Lattner33b73f92006-09-08 06:34:02 +00006202<!-- _______________________________________________________________________ -->
6203<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00006204 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
Chris Lattner33b73f92006-09-08 06:34:02 +00006205</div>
6206
6207<div class="doc_text">
6208
6209<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006210<p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
6211 floating point or vector of floating point type. Not all targets support all
6212 types however.</p>
6213
Chris Lattner33b73f92006-09-08 06:34:02 +00006214<pre>
Dale Johannesendd89d272007-10-02 17:47:38 +00006215 declare float @llvm.powi.f32(float %Val, i32 %power)
6216 declare double @llvm.powi.f64(double %Val, i32 %power)
6217 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
6218 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
6219 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Chris Lattner33b73f92006-09-08 06:34:02 +00006220</pre>
6221
6222<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006223<p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
6224 specified (positive or negative) power. The order of evaluation of
6225 multiplications is not defined. When a vector of floating point type is
6226 used, the second argument remains a scalar integer value.</p>
Chris Lattner33b73f92006-09-08 06:34:02 +00006227
6228<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006229<p>The second argument is an integer power, and the first is a value to raise to
6230 that power.</p>
Chris Lattner33b73f92006-09-08 06:34:02 +00006231
6232<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006233<p>This function returns the first value raised to the second power with an
6234 unspecified sequence of rounding operations.</p>
Chris Lattner33b73f92006-09-08 06:34:02 +00006235
Chris Lattner33b73f92006-09-08 06:34:02 +00006236</div>
6237
Dan Gohmanb6324c12007-10-15 20:30:11 +00006238<!-- _______________________________________________________________________ -->
6239<div class="doc_subsubsection">
6240 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
6241</div>
6242
6243<div class="doc_text">
6244
6245<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006246<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
6247 floating point or vector of floating point type. Not all targets support all
6248 types however.</p>
6249
Dan Gohmanb6324c12007-10-15 20:30:11 +00006250<pre>
6251 declare float @llvm.sin.f32(float %Val)
6252 declare double @llvm.sin.f64(double %Val)
6253 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
6254 declare fp128 @llvm.sin.f128(fp128 %Val)
6255 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
6256</pre>
6257
6258<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006259<p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +00006260
6261<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006262<p>The argument and return value are floating point numbers of the same
6263 type.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +00006264
6265<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006266<p>This function returns the sine of the specified operand, returning the same
6267 values as the libm <tt>sin</tt> functions would, and handles error conditions
6268 in the same way.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +00006269
Dan Gohmanb6324c12007-10-15 20:30:11 +00006270</div>
6271
6272<!-- _______________________________________________________________________ -->
6273<div class="doc_subsubsection">
6274 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
6275</div>
6276
6277<div class="doc_text">
6278
6279<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006280<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
6281 floating point or vector of floating point type. Not all targets support all
6282 types however.</p>
6283
Dan Gohmanb6324c12007-10-15 20:30:11 +00006284<pre>
6285 declare float @llvm.cos.f32(float %Val)
6286 declare double @llvm.cos.f64(double %Val)
6287 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
6288 declare fp128 @llvm.cos.f128(fp128 %Val)
6289 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
6290</pre>
6291
6292<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006293<p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +00006294
6295<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006296<p>The argument and return value are floating point numbers of the same
6297 type.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +00006298
6299<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006300<p>This function returns the cosine of the specified operand, returning the same
6301 values as the libm <tt>cos</tt> functions would, and handles error conditions
6302 in the same way.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +00006303
Dan Gohmanb6324c12007-10-15 20:30:11 +00006304</div>
6305
6306<!-- _______________________________________________________________________ -->
6307<div class="doc_subsubsection">
6308 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
6309</div>
6310
6311<div class="doc_text">
6312
6313<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006314<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
6315 floating point or vector of floating point type. Not all targets support all
6316 types however.</p>
6317
Dan Gohmanb6324c12007-10-15 20:30:11 +00006318<pre>
6319 declare float @llvm.pow.f32(float %Val, float %Power)
6320 declare double @llvm.pow.f64(double %Val, double %Power)
6321 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
6322 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
6323 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
6324</pre>
6325
6326<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006327<p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
6328 specified (positive or negative) power.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +00006329
6330<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006331<p>The second argument is a floating point power, and the first is a value to
6332 raise to that power.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +00006333
6334<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006335<p>This function returns the first value raised to the second power, returning
6336 the same values as the libm <tt>pow</tt> functions would, and handles error
6337 conditions in the same way.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +00006338
Dan Gohmanb6324c12007-10-15 20:30:11 +00006339</div>
6340
Andrew Lenharth1d463522005-05-03 18:01:48 +00006341<!-- ======================================================================= -->
6342<div class="doc_subsection">
Nate Begeman0f223bb2006-01-13 23:26:38 +00006343 <a name="int_manip">Bit Manipulation Intrinsics</a>
Andrew Lenharth1d463522005-05-03 18:01:48 +00006344</div>
6345
6346<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006347
6348<p>LLVM provides intrinsics for a few important bit manipulation operations.
6349 These allow efficient code generation for some algorithms.</p>
Andrew Lenharth1d463522005-05-03 18:01:48 +00006350
6351</div>
6352
6353<!-- _______________________________________________________________________ -->
6354<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00006355 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
Nate Begeman0f223bb2006-01-13 23:26:38 +00006356</div>
6357
6358<div class="doc_text">
6359
6360<h5>Syntax:</h5>
Reid Spencer4eefaab2007-04-01 08:04:23 +00006361<p>This is an overloaded intrinsic function. You can use bswap on any integer
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006362 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
6363
Nate Begeman0f223bb2006-01-13 23:26:38 +00006364<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +00006365 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
6366 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
6367 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Nate Begeman0f223bb2006-01-13 23:26:38 +00006368</pre>
6369
6370<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006371<p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
6372 values with an even number of bytes (positive multiple of 16 bits). These
6373 are useful for performing operations on data that is not in the target's
6374 native byte order.</p>
Nate Begeman0f223bb2006-01-13 23:26:38 +00006375
6376<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006377<p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
6378 and low byte of the input i16 swapped. Similarly,
6379 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
6380 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
6381 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
6382 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
6383 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
6384 more, respectively).</p>
Nate Begeman0f223bb2006-01-13 23:26:38 +00006385
6386</div>
6387
6388<!-- _______________________________________________________________________ -->
6389<div class="doc_subsubsection">
Reid Spencerb4f9a6f2006-01-16 21:12:35 +00006390 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
Andrew Lenharth1d463522005-05-03 18:01:48 +00006391</div>
6392
6393<div class="doc_text">
6394
6395<h5>Syntax:</h5>
Reid Spencer4eefaab2007-04-01 08:04:23 +00006396<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006397 width. Not all targets support all bit widths however.</p>
6398
Andrew Lenharth1d463522005-05-03 18:01:48 +00006399<pre>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006400 declare i8 @llvm.ctpop.i8(i8 &lt;src&gt;)
Chandler Carruth7132e002007-08-04 01:51:18 +00006401 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00006402 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carruth7132e002007-08-04 01:51:18 +00006403 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
6404 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Andrew Lenharth1d463522005-05-03 18:01:48 +00006405</pre>
6406
6407<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006408<p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
6409 in a value.</p>
Andrew Lenharth1d463522005-05-03 18:01:48 +00006410
6411<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006412<p>The only argument is the value to be counted. The argument may be of any
6413 integer type. The return type must match the argument type.</p>
Andrew Lenharth1d463522005-05-03 18:01:48 +00006414
6415<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006416<p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.</p>
Andrew Lenharth1d463522005-05-03 18:01:48 +00006417
Andrew Lenharth1d463522005-05-03 18:01:48 +00006418</div>
6419
6420<!-- _______________________________________________________________________ -->
6421<div class="doc_subsubsection">
Chris Lattnerb748c672006-01-16 22:34:14 +00006422 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
Andrew Lenharth1d463522005-05-03 18:01:48 +00006423</div>
6424
6425<div class="doc_text">
6426
6427<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006428<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
6429 integer bit width. Not all targets support all bit widths however.</p>
6430
Andrew Lenharth1d463522005-05-03 18:01:48 +00006431<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +00006432 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
6433 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00006434 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carruth7132e002007-08-04 01:51:18 +00006435 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
6436 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Andrew Lenharth1d463522005-05-03 18:01:48 +00006437</pre>
6438
6439<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006440<p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
6441 leading zeros in a variable.</p>
Andrew Lenharth1d463522005-05-03 18:01:48 +00006442
6443<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006444<p>The only argument is the value to be counted. The argument may be of any
6445 integer type. The return type must match the argument type.</p>
Andrew Lenharth1d463522005-05-03 18:01:48 +00006446
6447<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006448<p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
6449 zeros in a variable. If the src == 0 then the result is the size in bits of
6450 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
Andrew Lenharth1d463522005-05-03 18:01:48 +00006451
Andrew Lenharth1d463522005-05-03 18:01:48 +00006452</div>
Chris Lattner3b4f4372004-06-11 02:28:03 +00006453
Chris Lattnerefa20fa2005-05-15 19:39:26 +00006454<!-- _______________________________________________________________________ -->
6455<div class="doc_subsubsection">
Chris Lattnerb748c672006-01-16 22:34:14 +00006456 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
Chris Lattnerefa20fa2005-05-15 19:39:26 +00006457</div>
6458
6459<div class="doc_text">
6460
6461<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006462<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
6463 integer bit width. Not all targets support all bit widths however.</p>
6464
Chris Lattnerefa20fa2005-05-15 19:39:26 +00006465<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +00006466 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
6467 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00006468 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carruth7132e002007-08-04 01:51:18 +00006469 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
6470 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Chris Lattnerefa20fa2005-05-15 19:39:26 +00006471</pre>
6472
6473<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006474<p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
6475 trailing zeros.</p>
Chris Lattnerefa20fa2005-05-15 19:39:26 +00006476
6477<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006478<p>The only argument is the value to be counted. The argument may be of any
6479 integer type. The return type must match the argument type.</p>
Chris Lattnerefa20fa2005-05-15 19:39:26 +00006480
6481<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006482<p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
6483 zeros in a variable. If the src == 0 then the result is the size in bits of
6484 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
Chris Lattnerefa20fa2005-05-15 19:39:26 +00006485
Chris Lattnerefa20fa2005-05-15 19:39:26 +00006486</div>
6487
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006488<!-- ======================================================================= -->
6489<div class="doc_subsection">
6490 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
6491</div>
6492
6493<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006494
6495<p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006496
6497</div>
6498
Bill Wendlingf4d70622009-02-08 01:40:31 +00006499<!-- _______________________________________________________________________ -->
6500<div class="doc_subsubsection">
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006501 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006502</div>
6503
6504<div class="doc_text">
6505
6506<h5>Syntax:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006507<p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006508 on any integer bit width.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006509
6510<pre>
6511 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
6512 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6513 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6514</pre>
6515
6516<h5>Overview:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006517<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006518 a signed addition of the two arguments, and indicate whether an overflow
6519 occurred during the signed summation.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006520
6521<h5>Arguments:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006522<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006523 be of integer types of any bit width, but they must have the same bit
6524 width. The second element of the result structure must be of
6525 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6526 undergo signed addition.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006527
6528<h5>Semantics:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006529<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006530 a signed addition of the two variables. They return a structure &mdash; the
6531 first element of which is the signed summation, and the second element of
6532 which is a bit specifying if the signed summation resulted in an
6533 overflow.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006534
6535<h5>Examples:</h5>
6536<pre>
6537 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6538 %sum = extractvalue {i32, i1} %res, 0
6539 %obit = extractvalue {i32, i1} %res, 1
6540 br i1 %obit, label %overflow, label %normal
6541</pre>
6542
6543</div>
6544
6545<!-- _______________________________________________________________________ -->
6546<div class="doc_subsubsection">
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006547 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006548</div>
6549
6550<div class="doc_text">
6551
6552<h5>Syntax:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006553<p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006554 on any integer bit width.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006555
6556<pre>
6557 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6558 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6559 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6560</pre>
6561
6562<h5>Overview:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006563<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006564 an unsigned addition of the two arguments, and indicate whether a carry
6565 occurred during the unsigned summation.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006566
6567<h5>Arguments:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006568<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006569 be of integer types of any bit width, but they must have the same bit
6570 width. The second element of the result structure must be of
6571 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6572 undergo unsigned addition.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006573
6574<h5>Semantics:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006575<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006576 an unsigned addition of the two arguments. They return a structure &mdash;
6577 the first element of which is the sum, and the second element of which is a
6578 bit specifying if the unsigned summation resulted in a carry.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006579
6580<h5>Examples:</h5>
6581<pre>
6582 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6583 %sum = extractvalue {i32, i1} %res, 0
6584 %obit = extractvalue {i32, i1} %res, 1
6585 br i1 %obit, label %carry, label %normal
6586</pre>
6587
6588</div>
6589
6590<!-- _______________________________________________________________________ -->
6591<div class="doc_subsubsection">
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006592 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006593</div>
6594
6595<div class="doc_text">
6596
6597<h5>Syntax:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006598<p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006599 on any integer bit width.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006600
6601<pre>
6602 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6603 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6604 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6605</pre>
6606
6607<h5>Overview:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006608<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006609 a signed subtraction of the two arguments, and indicate whether an overflow
6610 occurred during the signed subtraction.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006611
6612<h5>Arguments:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006613<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006614 be of integer types of any bit width, but they must have the same bit
6615 width. The second element of the result structure must be of
6616 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6617 undergo signed subtraction.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006618
6619<h5>Semantics:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006620<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006621 a signed subtraction of the two arguments. They return a structure &mdash;
6622 the first element of which is the subtraction, and the second element of
6623 which is a bit specifying if the signed subtraction resulted in an
6624 overflow.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006625
6626<h5>Examples:</h5>
6627<pre>
6628 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6629 %sum = extractvalue {i32, i1} %res, 0
6630 %obit = extractvalue {i32, i1} %res, 1
6631 br i1 %obit, label %overflow, label %normal
6632</pre>
6633
6634</div>
6635
6636<!-- _______________________________________________________________________ -->
6637<div class="doc_subsubsection">
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006638 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006639</div>
6640
6641<div class="doc_text">
6642
6643<h5>Syntax:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006644<p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006645 on any integer bit width.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006646
6647<pre>
6648 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6649 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6650 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6651</pre>
6652
6653<h5>Overview:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006654<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006655 an unsigned subtraction of the two arguments, and indicate whether an
6656 overflow occurred during the unsigned subtraction.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006657
6658<h5>Arguments:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006659<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006660 be of integer types of any bit width, but they must have the same bit
6661 width. The second element of the result structure must be of
6662 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6663 undergo unsigned subtraction.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006664
6665<h5>Semantics:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006666<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006667 an unsigned subtraction of the two arguments. They return a structure &mdash;
6668 the first element of which is the subtraction, and the second element of
6669 which is a bit specifying if the unsigned subtraction resulted in an
6670 overflow.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006671
6672<h5>Examples:</h5>
6673<pre>
6674 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6675 %sum = extractvalue {i32, i1} %res, 0
6676 %obit = extractvalue {i32, i1} %res, 1
6677 br i1 %obit, label %overflow, label %normal
6678</pre>
6679
6680</div>
6681
6682<!-- _______________________________________________________________________ -->
6683<div class="doc_subsubsection">
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006684 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006685</div>
6686
6687<div class="doc_text">
6688
6689<h5>Syntax:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006690<p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006691 on any integer bit width.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006692
6693<pre>
6694 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6695 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6696 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6697</pre>
6698
6699<h5>Overview:</h5>
6700
6701<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006702 a signed multiplication of the two arguments, and indicate whether an
6703 overflow occurred during the signed multiplication.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006704
6705<h5>Arguments:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006706<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006707 be of integer types of any bit width, but they must have the same bit
6708 width. The second element of the result structure must be of
6709 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6710 undergo signed multiplication.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006711
6712<h5>Semantics:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006713<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006714 a signed multiplication of the two arguments. They return a structure &mdash;
6715 the first element of which is the multiplication, and the second element of
6716 which is a bit specifying if the signed multiplication resulted in an
6717 overflow.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006718
6719<h5>Examples:</h5>
6720<pre>
6721 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6722 %sum = extractvalue {i32, i1} %res, 0
6723 %obit = extractvalue {i32, i1} %res, 1
6724 br i1 %obit, label %overflow, label %normal
6725</pre>
6726
Reid Spencer5bf54c82007-04-11 23:23:49 +00006727</div>
6728
Bill Wendlingb9a73272009-02-08 23:00:09 +00006729<!-- _______________________________________________________________________ -->
6730<div class="doc_subsubsection">
6731 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6732</div>
6733
6734<div class="doc_text">
6735
6736<h5>Syntax:</h5>
Bill Wendlingb9a73272009-02-08 23:00:09 +00006737<p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006738 on any integer bit width.</p>
Bill Wendlingb9a73272009-02-08 23:00:09 +00006739
6740<pre>
6741 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6742 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6743 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6744</pre>
6745
6746<h5>Overview:</h5>
Bill Wendlingb9a73272009-02-08 23:00:09 +00006747<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006748 a unsigned multiplication of the two arguments, and indicate whether an
6749 overflow occurred during the unsigned multiplication.</p>
Bill Wendlingb9a73272009-02-08 23:00:09 +00006750
6751<h5>Arguments:</h5>
Bill Wendlingb9a73272009-02-08 23:00:09 +00006752<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006753 be of integer types of any bit width, but they must have the same bit
6754 width. The second element of the result structure must be of
6755 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6756 undergo unsigned multiplication.</p>
Bill Wendlingb9a73272009-02-08 23:00:09 +00006757
6758<h5>Semantics:</h5>
Bill Wendlingb9a73272009-02-08 23:00:09 +00006759<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006760 an unsigned multiplication of the two arguments. They return a structure
6761 &mdash; the first element of which is the multiplication, and the second
6762 element of which is a bit specifying if the unsigned multiplication resulted
6763 in an overflow.</p>
Bill Wendlingb9a73272009-02-08 23:00:09 +00006764
6765<h5>Examples:</h5>
6766<pre>
6767 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6768 %sum = extractvalue {i32, i1} %res, 0
6769 %obit = extractvalue {i32, i1} %res, 1
6770 br i1 %obit, label %overflow, label %normal
6771</pre>
6772
6773</div>
6774
Chris Lattner941515c2004-01-06 05:31:32 +00006775<!-- ======================================================================= -->
6776<div class="doc_subsection">
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +00006777 <a name="int_fp16">Half Precision Floating Point Intrinsics</a>
6778</div>
6779
6780<div class="doc_text">
6781
Chris Lattner022a9fb2010-03-15 04:12:21 +00006782<p>Half precision floating point is a storage-only format. This means that it is
6783 a dense encoding (in memory) but does not support computation in the
6784 format.</p>
Chris Lattnerbbd8bd32010-03-14 23:03:31 +00006785
Chris Lattner022a9fb2010-03-15 04:12:21 +00006786<p>This means that code must first load the half-precision floating point
Chris Lattnerbbd8bd32010-03-14 23:03:31 +00006787 value as an i16, then convert it to float with <a
6788 href="#int_convert_from_fp16"><tt>llvm.convert.from.fp16</tt></a>.
6789 Computation can then be performed on the float value (including extending to
Chris Lattner022a9fb2010-03-15 04:12:21 +00006790 double etc). To store the value back to memory, it is first converted to
6791 float if needed, then converted to i16 with
Chris Lattnerbbd8bd32010-03-14 23:03:31 +00006792 <a href="#int_convert_to_fp16"><tt>llvm.convert.to.fp16</tt></a>, then
6793 storing as an i16 value.</p>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +00006794</div>
6795
6796<!-- _______________________________________________________________________ -->
6797<div class="doc_subsubsection">
Chris Lattnerbbd8bd32010-03-14 23:03:31 +00006798 <a name="int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +00006799</div>
6800
6801<div class="doc_text">
6802
6803<h5>Syntax:</h5>
6804<pre>
6805 declare i16 @llvm.convert.to.fp16(f32 %a)
6806</pre>
6807
6808<h5>Overview:</h5>
6809<p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
6810 a conversion from single precision floating point format to half precision
6811 floating point format.</p>
6812
6813<h5>Arguments:</h5>
6814<p>The intrinsic function contains single argument - the value to be
6815 converted.</p>
6816
6817<h5>Semantics:</h5>
6818<p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
6819 a conversion from single precision floating point format to half precision
Chris Lattner022a9fb2010-03-15 04:12:21 +00006820 floating point format. The return value is an <tt>i16</tt> which
Chris Lattnerbbd8bd32010-03-14 23:03:31 +00006821 contains the converted number.</p>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +00006822
6823<h5>Examples:</h5>
6824<pre>
6825 %res = call i16 @llvm.convert.to.fp16(f32 %a)
6826 store i16 %res, i16* @x, align 2
6827</pre>
6828
6829</div>
6830
6831<!-- _______________________________________________________________________ -->
6832<div class="doc_subsubsection">
Chris Lattnerbbd8bd32010-03-14 23:03:31 +00006833 <a name="int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +00006834</div>
6835
6836<div class="doc_text">
6837
6838<h5>Syntax:</h5>
6839<pre>
6840 declare f32 @llvm.convert.from.fp16(i16 %a)
6841</pre>
6842
6843<h5>Overview:</h5>
6844<p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs
6845 a conversion from half precision floating point format to single precision
6846 floating point format.</p>
6847
6848<h5>Arguments:</h5>
6849<p>The intrinsic function contains single argument - the value to be
6850 converted.</p>
6851
6852<h5>Semantics:</h5>
6853<p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs a
Chris Lattner022a9fb2010-03-15 04:12:21 +00006854 conversion from half single precision floating point format to single
Chris Lattnerbbd8bd32010-03-14 23:03:31 +00006855 precision floating point format. The input half-float value is represented by
6856 an <tt>i16</tt> value.</p>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +00006857
6858<h5>Examples:</h5>
6859<pre>
6860 %a = load i16* @x, align 2
6861 %res = call f32 @llvm.convert.from.fp16(i16 %a)
6862</pre>
6863
6864</div>
6865
6866<!-- ======================================================================= -->
6867<div class="doc_subsection">
Chris Lattner941515c2004-01-06 05:31:32 +00006868 <a name="int_debugger">Debugger Intrinsics</a>
6869</div>
6870
6871<div class="doc_text">
Chris Lattner941515c2004-01-06 05:31:32 +00006872
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006873<p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
6874 prefix), are described in
6875 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
6876 Level Debugging</a> document.</p>
6877
6878</div>
Chris Lattner941515c2004-01-06 05:31:32 +00006879
Jim Laskey2211f492007-03-14 19:31:19 +00006880<!-- ======================================================================= -->
6881<div class="doc_subsection">
6882 <a name="int_eh">Exception Handling Intrinsics</a>
6883</div>
6884
6885<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006886
6887<p>The LLVM exception handling intrinsics (which all start with
6888 <tt>llvm.eh.</tt> prefix), are described in
6889 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6890 Handling</a> document.</p>
6891
Jim Laskey2211f492007-03-14 19:31:19 +00006892</div>
6893
Tanya Lattnercb1b9602007-06-15 20:50:54 +00006894<!-- ======================================================================= -->
6895<div class="doc_subsection">
Duncan Sands86e01192007-09-11 14:10:23 +00006896 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands644f9172007-07-27 12:58:54 +00006897</div>
6898
6899<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006900
6901<p>This intrinsic makes it possible to excise one parameter, marked with
Dan Gohman3770af52010-07-02 23:18:08 +00006902 the <a href="#nest"><tt>nest</tt></a> attribute, from a function.
6903 The result is a callable
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006904 function pointer lacking the nest parameter - the caller does not need to
6905 provide a value for it. Instead, the value to use is stored in advance in a
6906 "trampoline", a block of memory usually allocated on the stack, which also
6907 contains code to splice the nest value into the argument list. This is used
6908 to implement the GCC nested function address extension.</p>
6909
6910<p>For example, if the function is
6911 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
6912 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
6913 follows:</p>
6914
Benjamin Kramer79698be2010-07-13 12:26:09 +00006915<pre class="doc_code">
Duncan Sands86e01192007-09-11 14:10:23 +00006916 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6917 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
Dan Gohmand6a6f612010-05-28 17:07:41 +00006918 %p = call i8* @llvm.init.trampoline(i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval)
Duncan Sands86e01192007-09-11 14:10:23 +00006919 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands644f9172007-07-27 12:58:54 +00006920</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006921
Dan Gohmand6a6f612010-05-28 17:07:41 +00006922<p>The call <tt>%val = call i32 %fp(i32 %x, i32 %y)</tt> is then equivalent
6923 to <tt>%val = call i32 %f(i8* %nval, i32 %x, i32 %y)</tt>.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006924
Duncan Sands644f9172007-07-27 12:58:54 +00006925</div>
6926
6927<!-- _______________________________________________________________________ -->
6928<div class="doc_subsubsection">
6929 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6930</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006931
Duncan Sands644f9172007-07-27 12:58:54 +00006932<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006933
Duncan Sands644f9172007-07-27 12:58:54 +00006934<h5>Syntax:</h5>
6935<pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006936 declare i8* @llvm.init.trampoline(i8* &lt;tramp&gt;, i8* &lt;func&gt;, i8* &lt;nval&gt;)
Duncan Sands644f9172007-07-27 12:58:54 +00006937</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006938
Duncan Sands644f9172007-07-27 12:58:54 +00006939<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006940<p>This fills the memory pointed to by <tt>tramp</tt> with code and returns a
6941 function pointer suitable for executing it.</p>
6942
Duncan Sands644f9172007-07-27 12:58:54 +00006943<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006944<p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6945 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
6946 sufficiently aligned block of memory; this memory is written to by the
6947 intrinsic. Note that the size and the alignment are target-specific - LLVM
6948 currently provides no portable way of determining them, so a front-end that
6949 generates this intrinsic needs to have some target-specific knowledge.
6950 The <tt>func</tt> argument must hold a function bitcast to
6951 an <tt>i8*</tt>.</p>
6952
Duncan Sands644f9172007-07-27 12:58:54 +00006953<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006954<p>The block of memory pointed to by <tt>tramp</tt> is filled with target
6955 dependent code, turning it into a function. A pointer to this function is
6956 returned, but needs to be bitcast to an <a href="#int_trampoline">appropriate
6957 function pointer type</a> before being called. The new function's signature
6958 is the same as that of <tt>func</tt> with any arguments marked with
6959 the <tt>nest</tt> attribute removed. At most one such <tt>nest</tt> argument
6960 is allowed, and it must be of pointer type. Calling the new function is
6961 equivalent to calling <tt>func</tt> with the same argument list, but
6962 with <tt>nval</tt> used for the missing <tt>nest</tt> argument. If, after
6963 calling <tt>llvm.init.trampoline</tt>, the memory pointed to
6964 by <tt>tramp</tt> is modified, then the effect of any later call to the
6965 returned function pointer is undefined.</p>
6966
Duncan Sands644f9172007-07-27 12:58:54 +00006967</div>
6968
6969<!-- ======================================================================= -->
6970<div class="doc_subsection">
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00006971 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6972</div>
6973
6974<div class="doc_text">
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00006975
Bill Wendlingd9a66f72009-07-20 02:29:24 +00006976<p>These intrinsic functions expand the "universal IR" of LLVM to represent
6977 hardware constructs for atomic operations and memory synchronization. This
6978 provides an interface to the hardware, not an interface to the programmer. It
6979 is aimed at a low enough level to allow any programming models or APIs
6980 (Application Programming Interfaces) which need atomic behaviors to map
6981 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
6982 hardware provides a "universal IR" for source languages, it also provides a
6983 starting point for developing a "universal" atomic operation and
6984 synchronization IR.</p>
6985
6986<p>These do <em>not</em> form an API such as high-level threading libraries,
6987 software transaction memory systems, atomic primitives, and intrinsic
6988 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6989 application libraries. The hardware interface provided by LLVM should allow
6990 a clean implementation of all of these APIs and parallel programming models.
6991 No one model or paradigm should be selected above others unless the hardware
6992 itself ubiquitously does so.</p>
6993
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00006994</div>
6995
6996<!-- _______________________________________________________________________ -->
6997<div class="doc_subsubsection">
6998 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6999</div>
7000<div class="doc_text">
7001<h5>Syntax:</h5>
7002<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007003 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 Lenharth9b254ee2008-02-16 01:24:58 +00007004</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007005
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00007006<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007007<p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
7008 specific pairs of memory access types.</p>
7009
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00007010<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007011<p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
7012 The first four arguments enables a specific barrier as listed below. The
Dan Gohmana269a0a2010-03-01 17:41:39 +00007013 fifth argument specifies that the barrier applies to io or device or uncached
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007014 memory.</p>
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00007015
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007016<ul>
7017 <li><tt>ll</tt>: load-load barrier</li>
7018 <li><tt>ls</tt>: load-store barrier</li>
7019 <li><tt>sl</tt>: store-load barrier</li>
7020 <li><tt>ss</tt>: store-store barrier</li>
7021 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
7022</ul>
7023
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00007024<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007025<p>This intrinsic causes the system to enforce some ordering constraints upon
7026 the loads and stores of the program. This barrier does not
7027 indicate <em>when</em> any events will occur, it only enforces
7028 an <em>order</em> in which they occur. For any of the specified pairs of load
7029 and store operations (f.ex. load-load, or store-load), all of the first
7030 operations preceding the barrier will complete before any of the second
7031 operations succeeding the barrier begin. Specifically the semantics for each
7032 pairing is as follows:</p>
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00007033
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007034<ul>
7035 <li><tt>ll</tt>: All loads before the barrier must complete before any load
7036 after the barrier begins.</li>
Eric Christopher455c5772009-12-05 02:46:03 +00007037 <li><tt>ls</tt>: All loads before the barrier must complete before any
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007038 store after the barrier begins.</li>
Eric Christopher455c5772009-12-05 02:46:03 +00007039 <li><tt>ss</tt>: All stores before the barrier must complete before any
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007040 store after the barrier begins.</li>
Eric Christopher455c5772009-12-05 02:46:03 +00007041 <li><tt>sl</tt>: All stores before the barrier must complete before any
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007042 load after the barrier begins.</li>
7043</ul>
7044
7045<p>These semantics are applied with a logical "and" behavior when more than one
7046 is enabled in a single memory barrier intrinsic.</p>
7047
7048<p>Backends may implement stronger barriers than those requested when they do
7049 not support as fine grained a barrier as requested. Some architectures do
7050 not need all types of barriers and on such architectures, these become
7051 noops.</p>
7052
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00007053<h5>Example:</h5>
7054<pre>
Victor Hernandeza70c6df2009-10-26 23:44:29 +00007055%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7056%ptr = bitcast i8* %mallocP to i32*
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00007057 store i32 4, %ptr
7058
7059%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007060 call void @llvm.memory.barrier(i1 false, i1 true, i1 false, i1 false)
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00007061 <i>; guarantee the above finishes</i>
7062 store i32 8, %ptr <i>; before this begins</i>
7063</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007064
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00007065</div>
7066
Andrew Lenharth95528942008-02-21 06:45:13 +00007067<!-- _______________________________________________________________________ -->
7068<div class="doc_subsubsection">
Mon P Wang6a490372008-06-25 08:15:39 +00007069 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
Andrew Lenharth95528942008-02-21 06:45:13 +00007070</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007071
Andrew Lenharth95528942008-02-21 06:45:13 +00007072<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007073
Andrew Lenharth95528942008-02-21 06:45:13 +00007074<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007075<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
7076 any integer bit width and for different address spaces. Not all targets
7077 support all bit widths however.</p>
Andrew Lenharth95528942008-02-21 06:45:13 +00007078
7079<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007080 declare i8 @llvm.atomic.cmp.swap.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt;)
7081 declare i16 @llvm.atomic.cmp.swap.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt;)
7082 declare i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt;)
7083 declare i64 @llvm.atomic.cmp.swap.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;cmp&gt;, i64 &lt;val&gt;)
Andrew Lenharth95528942008-02-21 06:45:13 +00007084</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007085
Andrew Lenharth95528942008-02-21 06:45:13 +00007086<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007087<p>This loads a value in memory and compares it to a given value. If they are
7088 equal, it stores a new value into the memory.</p>
7089
Andrew Lenharth95528942008-02-21 06:45:13 +00007090<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007091<p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
7092 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
7093 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
7094 this integer type. While any bit width integer may be used, targets may only
7095 lower representations they support in hardware.</p>
Andrew Lenharth95528942008-02-21 06:45:13 +00007096
Andrew Lenharth95528942008-02-21 06:45:13 +00007097<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007098<p>This entire intrinsic must be executed atomically. It first loads the value
7099 in memory pointed to by <tt>ptr</tt> and compares it with the
7100 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
7101 memory. The loaded value is yielded in all cases. This provides the
7102 equivalent of an atomic compare-and-swap operation within the SSA
7103 framework.</p>
Andrew Lenharth95528942008-02-21 06:45:13 +00007104
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007105<h5>Examples:</h5>
Andrew Lenharth95528942008-02-21 06:45:13 +00007106<pre>
Victor Hernandeza70c6df2009-10-26 23:44:29 +00007107%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7108%ptr = bitcast i8* %mallocP to i32*
Andrew Lenharth95528942008-02-21 06:45:13 +00007109 store i32 4, %ptr
7110
7111%val1 = add i32 4, 4
Dan Gohmand6a6f612010-05-28 17:07:41 +00007112%result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 4, %val1)
Andrew Lenharth95528942008-02-21 06:45:13 +00007113 <i>; yields {i32}:result1 = 4</i>
7114%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7115%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7116
7117%val2 = add i32 1, 1
Dan Gohmand6a6f612010-05-28 17:07:41 +00007118%result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 5, %val2)
Andrew Lenharth95528942008-02-21 06:45:13 +00007119 <i>; yields {i32}:result2 = 8</i>
7120%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
7121
7122%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
7123</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007124
Andrew Lenharth95528942008-02-21 06:45:13 +00007125</div>
7126
7127<!-- _______________________________________________________________________ -->
7128<div class="doc_subsubsection">
7129 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
7130</div>
7131<div class="doc_text">
7132<h5>Syntax:</h5>
7133
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007134<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
7135 integer bit width. Not all targets support all bit widths however.</p>
7136
Andrew Lenharth95528942008-02-21 06:45:13 +00007137<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007138 declare i8 @llvm.atomic.swap.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;val&gt;)
7139 declare i16 @llvm.atomic.swap.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;val&gt;)
7140 declare i32 @llvm.atomic.swap.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;val&gt;)
7141 declare i64 @llvm.atomic.swap.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;val&gt;)
Andrew Lenharth95528942008-02-21 06:45:13 +00007142</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007143
Andrew Lenharth95528942008-02-21 06:45:13 +00007144<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007145<p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
7146 the value from memory. It then stores the value in <tt>val</tt> in the memory
7147 at <tt>ptr</tt>.</p>
7148
Andrew Lenharth95528942008-02-21 06:45:13 +00007149<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007150<p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
7151 the <tt>val</tt> argument and the result must be integers of the same bit
7152 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
7153 integer type. The targets may only lower integer representations they
7154 support.</p>
Andrew Lenharth95528942008-02-21 06:45:13 +00007155
Andrew Lenharth95528942008-02-21 06:45:13 +00007156<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007157<p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
7158 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
7159 equivalent of an atomic swap operation within the SSA framework.</p>
Andrew Lenharth95528942008-02-21 06:45:13 +00007160
Andrew Lenharth95528942008-02-21 06:45:13 +00007161<h5>Examples:</h5>
7162<pre>
Victor Hernandeza70c6df2009-10-26 23:44:29 +00007163%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7164%ptr = bitcast i8* %mallocP to i32*
Andrew Lenharth95528942008-02-21 06:45:13 +00007165 store i32 4, %ptr
7166
7167%val1 = add i32 4, 4
Dan Gohmand6a6f612010-05-28 17:07:41 +00007168%result1 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val1)
Andrew Lenharth95528942008-02-21 06:45:13 +00007169 <i>; yields {i32}:result1 = 4</i>
7170%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7171%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7172
7173%val2 = add i32 1, 1
Dan Gohmand6a6f612010-05-28 17:07:41 +00007174%result2 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val2)
Andrew Lenharth95528942008-02-21 06:45:13 +00007175 <i>; yields {i32}:result2 = 8</i>
7176
7177%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
7178%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
7179</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007180
Andrew Lenharth95528942008-02-21 06:45:13 +00007181</div>
7182
7183<!-- _______________________________________________________________________ -->
7184<div class="doc_subsubsection">
Mon P Wang6a490372008-06-25 08:15:39 +00007185 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
Andrew Lenharth95528942008-02-21 06:45:13 +00007186
7187</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007188
Andrew Lenharth95528942008-02-21 06:45:13 +00007189<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007190
Andrew Lenharth95528942008-02-21 06:45:13 +00007191<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007192<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
7193 any integer bit width. Not all targets support all bit widths however.</p>
7194
Andrew Lenharth95528942008-02-21 06:45:13 +00007195<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007196 declare i8 @llvm.atomic.load.add.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7197 declare i16 @llvm.atomic.load.add.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7198 declare i32 @llvm.atomic.load.add.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7199 declare i64 @llvm.atomic.load.add.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Andrew Lenharth95528942008-02-21 06:45:13 +00007200</pre>
Andrew Lenharth95528942008-02-21 06:45:13 +00007201
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007202<h5>Overview:</h5>
7203<p>This intrinsic adds <tt>delta</tt> to the value stored in memory
7204 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7205
7206<h5>Arguments:</h5>
7207<p>The intrinsic takes two arguments, the first a pointer to an integer value
7208 and the second an integer value. The result is also an integer value. These
7209 integer types can have any bit width, but they must all have the same bit
7210 width. The targets may only lower integer representations they support.</p>
7211
Andrew Lenharth95528942008-02-21 06:45:13 +00007212<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007213<p>This intrinsic does a series of operations atomically. It first loads the
7214 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
7215 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
Andrew Lenharth95528942008-02-21 06:45:13 +00007216
7217<h5>Examples:</h5>
7218<pre>
Victor Hernandeza70c6df2009-10-26 23:44:29 +00007219%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7220%ptr = bitcast i8* %mallocP to i32*
7221 store i32 4, %ptr
Dan Gohmand6a6f612010-05-28 17:07:41 +00007222%result1 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 4)
Andrew Lenharth95528942008-02-21 06:45:13 +00007223 <i>; yields {i32}:result1 = 4</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007224%result2 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 2)
Andrew Lenharth95528942008-02-21 06:45:13 +00007225 <i>; yields {i32}:result2 = 8</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007226%result3 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 5)
Andrew Lenharth95528942008-02-21 06:45:13 +00007227 <i>; yields {i32}:result3 = 10</i>
Mon P Wang6a490372008-06-25 08:15:39 +00007228%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
Andrew Lenharth95528942008-02-21 06:45:13 +00007229</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007230
Andrew Lenharth95528942008-02-21 06:45:13 +00007231</div>
7232
Mon P Wang6a490372008-06-25 08:15:39 +00007233<!-- _______________________________________________________________________ -->
7234<div class="doc_subsubsection">
7235 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
7236
7237</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007238
Mon P Wang6a490372008-06-25 08:15:39 +00007239<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007240
Mon P Wang6a490372008-06-25 08:15:39 +00007241<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007242<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
7243 any integer bit width and for different address spaces. Not all targets
7244 support all bit widths however.</p>
7245
Mon P Wang6a490372008-06-25 08:15:39 +00007246<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007247 declare i8 @llvm.atomic.load.sub.i8.p0i32(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7248 declare i16 @llvm.atomic.load.sub.i16.p0i32(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7249 declare i32 @llvm.atomic.load.sub.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7250 declare i64 @llvm.atomic.load.sub.i64.p0i32(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +00007251</pre>
Mon P Wang6a490372008-06-25 08:15:39 +00007252
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007253<h5>Overview:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00007254<p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007255 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7256
7257<h5>Arguments:</h5>
7258<p>The intrinsic takes two arguments, the first a pointer to an integer value
7259 and the second an integer value. The result is also an integer value. These
7260 integer types can have any bit width, but they must all have the same bit
7261 width. The targets may only lower integer representations they support.</p>
7262
Mon P Wang6a490372008-06-25 08:15:39 +00007263<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007264<p>This intrinsic does a series of operations atomically. It first loads the
7265 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
7266 result to <tt>ptr</tt>. It yields the original value stored
7267 at <tt>ptr</tt>.</p>
Mon P Wang6a490372008-06-25 08:15:39 +00007268
7269<h5>Examples:</h5>
7270<pre>
Victor Hernandeza70c6df2009-10-26 23:44:29 +00007271%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7272%ptr = bitcast i8* %mallocP to i32*
7273 store i32 8, %ptr
Dan Gohmand6a6f612010-05-28 17:07:41 +00007274%result1 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 4)
Mon P Wang6a490372008-06-25 08:15:39 +00007275 <i>; yields {i32}:result1 = 8</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007276%result2 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 2)
Mon P Wang6a490372008-06-25 08:15:39 +00007277 <i>; yields {i32}:result2 = 4</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007278%result3 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 5)
Mon P Wang6a490372008-06-25 08:15:39 +00007279 <i>; yields {i32}:result3 = 2</i>
7280%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
7281</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007282
Mon P Wang6a490372008-06-25 08:15:39 +00007283</div>
7284
7285<!-- _______________________________________________________________________ -->
7286<div class="doc_subsubsection">
7287 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
7288 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
7289 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
7290 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
Mon P Wang6a490372008-06-25 08:15:39 +00007291</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007292
Mon P Wang6a490372008-06-25 08:15:39 +00007293<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007294
Mon P Wang6a490372008-06-25 08:15:39 +00007295<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007296<p>These are overloaded intrinsics. You can
7297 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
7298 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
7299 bit width and for different address spaces. Not all targets support all bit
7300 widths however.</p>
Mon P Wang6a490372008-06-25 08:15:39 +00007301
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007302<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007303 declare i8 @llvm.atomic.load.and.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7304 declare i16 @llvm.atomic.load.and.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7305 declare i32 @llvm.atomic.load.and.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7306 declare i64 @llvm.atomic.load.and.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +00007307</pre>
7308
7309<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007310 declare i8 @llvm.atomic.load.or.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7311 declare i16 @llvm.atomic.load.or.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7312 declare i32 @llvm.atomic.load.or.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7313 declare i64 @llvm.atomic.load.or.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +00007314</pre>
7315
7316<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007317 declare i8 @llvm.atomic.load.nand.i8.p0i32(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7318 declare i16 @llvm.atomic.load.nand.i16.p0i32(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7319 declare i32 @llvm.atomic.load.nand.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7320 declare i64 @llvm.atomic.load.nand.i64.p0i32(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +00007321</pre>
7322
7323<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007324 declare i8 @llvm.atomic.load.xor.i8.p0i32(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7325 declare i16 @llvm.atomic.load.xor.i16.p0i32(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7326 declare i32 @llvm.atomic.load.xor.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7327 declare i64 @llvm.atomic.load.xor.i64.p0i32(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +00007328</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007329
Mon P Wang6a490372008-06-25 08:15:39 +00007330<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007331<p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
7332 the value stored in memory at <tt>ptr</tt>. It yields the original value
7333 at <tt>ptr</tt>.</p>
Mon P Wang6a490372008-06-25 08:15:39 +00007334
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007335<h5>Arguments:</h5>
7336<p>These intrinsics take two arguments, the first a pointer to an integer value
7337 and the second an integer value. The result is also an integer value. These
7338 integer types can have any bit width, but they must all have the same bit
7339 width. The targets may only lower integer representations they support.</p>
7340
Mon P Wang6a490372008-06-25 08:15:39 +00007341<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007342<p>These intrinsics does a series of operations atomically. They first load the
7343 value stored at <tt>ptr</tt>. They then do the bitwise
7344 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
7345 original value stored at <tt>ptr</tt>.</p>
Mon P Wang6a490372008-06-25 08:15:39 +00007346
7347<h5>Examples:</h5>
7348<pre>
Victor Hernandeza70c6df2009-10-26 23:44:29 +00007349%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7350%ptr = bitcast i8* %mallocP to i32*
7351 store i32 0x0F0F, %ptr
Dan Gohmand6a6f612010-05-28 17:07:41 +00007352%result0 = call i32 @llvm.atomic.load.nand.i32.p0i32(i32* %ptr, i32 0xFF)
Mon P Wang6a490372008-06-25 08:15:39 +00007353 <i>; yields {i32}:result0 = 0x0F0F</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007354%result1 = call i32 @llvm.atomic.load.and.i32.p0i32(i32* %ptr, i32 0xFF)
Mon P Wang6a490372008-06-25 08:15:39 +00007355 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007356%result2 = call i32 @llvm.atomic.load.or.i32.p0i32(i32* %ptr, i32 0F)
Mon P Wang6a490372008-06-25 08:15:39 +00007357 <i>; yields {i32}:result2 = 0xF0</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007358%result3 = call i32 @llvm.atomic.load.xor.i32.p0i32(i32* %ptr, i32 0F)
Mon P Wang6a490372008-06-25 08:15:39 +00007359 <i>; yields {i32}:result3 = FF</i>
7360%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
7361</pre>
Mon P Wang6a490372008-06-25 08:15:39 +00007362
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007363</div>
Mon P Wang6a490372008-06-25 08:15:39 +00007364
7365<!-- _______________________________________________________________________ -->
7366<div class="doc_subsubsection">
7367 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
7368 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
7369 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
7370 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
Mon P Wang6a490372008-06-25 08:15:39 +00007371</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007372
Mon P Wang6a490372008-06-25 08:15:39 +00007373<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007374
Mon P Wang6a490372008-06-25 08:15:39 +00007375<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007376<p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
7377 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
7378 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
7379 address spaces. Not all targets support all bit widths however.</p>
Mon P Wang6a490372008-06-25 08:15:39 +00007380
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007381<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007382 declare i8 @llvm.atomic.load.max.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7383 declare i16 @llvm.atomic.load.max.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7384 declare i32 @llvm.atomic.load.max.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7385 declare i64 @llvm.atomic.load.max.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +00007386</pre>
7387
7388<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007389 declare i8 @llvm.atomic.load.min.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7390 declare i16 @llvm.atomic.load.min.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7391 declare i32 @llvm.atomic.load.min.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7392 declare i64 @llvm.atomic.load.min.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +00007393</pre>
7394
7395<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007396 declare i8 @llvm.atomic.load.umax.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7397 declare i16 @llvm.atomic.load.umax.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7398 declare i32 @llvm.atomic.load.umax.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7399 declare i64 @llvm.atomic.load.umax.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +00007400</pre>
7401
7402<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007403 declare i8 @llvm.atomic.load.umin.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
7404 declare i16 @llvm.atomic.load.umin.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
7405 declare i32 @llvm.atomic.load.umin.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
7406 declare i64 @llvm.atomic.load.umin.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +00007407</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007408
Mon P Wang6a490372008-06-25 08:15:39 +00007409<h5>Overview:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00007410<p>These intrinsics takes the signed or unsigned minimum or maximum of
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007411 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
7412 original value at <tt>ptr</tt>.</p>
Mon P Wang6a490372008-06-25 08:15:39 +00007413
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007414<h5>Arguments:</h5>
7415<p>These intrinsics take two arguments, the first a pointer to an integer value
7416 and the second an integer value. The result is also an integer value. These
7417 integer types can have any bit width, but they must all have the same bit
7418 width. The targets may only lower integer representations they support.</p>
7419
Mon P Wang6a490372008-06-25 08:15:39 +00007420<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007421<p>These intrinsics does a series of operations atomically. They first load the
7422 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
7423 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
7424 yield the original value stored at <tt>ptr</tt>.</p>
Mon P Wang6a490372008-06-25 08:15:39 +00007425
7426<h5>Examples:</h5>
7427<pre>
Victor Hernandeza70c6df2009-10-26 23:44:29 +00007428%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7429%ptr = bitcast i8* %mallocP to i32*
7430 store i32 7, %ptr
Dan Gohmand6a6f612010-05-28 17:07:41 +00007431%result0 = call i32 @llvm.atomic.load.min.i32.p0i32(i32* %ptr, i32 -2)
Mon P Wang6a490372008-06-25 08:15:39 +00007432 <i>; yields {i32}:result0 = 7</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007433%result1 = call i32 @llvm.atomic.load.max.i32.p0i32(i32* %ptr, i32 8)
Mon P Wang6a490372008-06-25 08:15:39 +00007434 <i>; yields {i32}:result1 = -2</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007435%result2 = call i32 @llvm.atomic.load.umin.i32.p0i32(i32* %ptr, i32 10)
Mon P Wang6a490372008-06-25 08:15:39 +00007436 <i>; yields {i32}:result2 = 8</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007437%result3 = call i32 @llvm.atomic.load.umax.i32.p0i32(i32* %ptr, i32 30)
Mon P Wang6a490372008-06-25 08:15:39 +00007438 <i>; yields {i32}:result3 = 8</i>
7439%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
7440</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007441
Mon P Wang6a490372008-06-25 08:15:39 +00007442</div>
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00007443
Nick Lewycky6f7d8342009-10-13 07:03:23 +00007444
7445<!-- ======================================================================= -->
7446<div class="doc_subsection">
7447 <a name="int_memorymarkers">Memory Use Markers</a>
7448</div>
7449
7450<div class="doc_text">
7451
7452<p>This class of intrinsics exists to information about the lifetime of memory
7453 objects and ranges where variables are immutable.</p>
7454
7455</div>
7456
7457<!-- _______________________________________________________________________ -->
7458<div class="doc_subsubsection">
7459 <a name="int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a>
7460</div>
7461
7462<div class="doc_text">
7463
7464<h5>Syntax:</h5>
7465<pre>
7466 declare void @llvm.lifetime.start(i64 &lt;size&gt;, i8* nocapture &lt;ptr&gt;)
7467</pre>
7468
7469<h5>Overview:</h5>
7470<p>The '<tt>llvm.lifetime.start</tt>' intrinsic specifies the start of a memory
7471 object's lifetime.</p>
7472
7473<h5>Arguments:</h5>
Nick Lewycky9bc89042009-10-13 07:57:33 +00007474<p>The first argument is a constant integer representing the size of the
7475 object, or -1 if it is variable sized. The second argument is a pointer to
7476 the object.</p>
Nick Lewycky6f7d8342009-10-13 07:03:23 +00007477
7478<h5>Semantics:</h5>
7479<p>This intrinsic indicates that before this point in the code, the value of the
7480 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
Nick Lewyckyd20fd592009-10-27 16:56:58 +00007481 never be used and has an undefined value. A load from the pointer that
7482 precedes this intrinsic can be replaced with
Nick Lewycky6f7d8342009-10-13 07:03:23 +00007483 <tt>'<a href="#undefvalues">undef</a>'</tt>.</p>
7484
7485</div>
7486
7487<!-- _______________________________________________________________________ -->
7488<div class="doc_subsubsection">
7489 <a name="int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a>
7490</div>
7491
7492<div class="doc_text">
7493
7494<h5>Syntax:</h5>
7495<pre>
7496 declare void @llvm.lifetime.end(i64 &lt;size&gt;, i8* nocapture &lt;ptr&gt;)
7497</pre>
7498
7499<h5>Overview:</h5>
7500<p>The '<tt>llvm.lifetime.end</tt>' intrinsic specifies the end of a memory
7501 object's lifetime.</p>
7502
7503<h5>Arguments:</h5>
Nick Lewycky9bc89042009-10-13 07:57:33 +00007504<p>The first argument is a constant integer representing the size of the
7505 object, or -1 if it is variable sized. The second argument is a pointer to
7506 the object.</p>
Nick Lewycky6f7d8342009-10-13 07:03:23 +00007507
7508<h5>Semantics:</h5>
7509<p>This intrinsic indicates that after this point in the code, the value of the
7510 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7511 never be used and has an undefined value. Any stores into the memory object
7512 following this intrinsic may be removed as dead.
7513
7514</div>
7515
7516<!-- _______________________________________________________________________ -->
7517<div class="doc_subsubsection">
7518 <a name="int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a>
7519</div>
7520
7521<div class="doc_text">
7522
7523<h5>Syntax:</h5>
7524<pre>
7525 declare {}* @llvm.invariant.start(i64 &lt;size&gt;, i8* nocapture &lt;ptr&gt;) readonly
7526</pre>
7527
7528<h5>Overview:</h5>
7529<p>The '<tt>llvm.invariant.start</tt>' intrinsic specifies that the contents of
7530 a memory object will not change.</p>
7531
7532<h5>Arguments:</h5>
Nick Lewycky9bc89042009-10-13 07:57:33 +00007533<p>The first argument is a constant integer representing the size of the
7534 object, or -1 if it is variable sized. The second argument is a pointer to
7535 the object.</p>
Nick Lewycky6f7d8342009-10-13 07:03:23 +00007536
7537<h5>Semantics:</h5>
7538<p>This intrinsic indicates that until an <tt>llvm.invariant.end</tt> that uses
7539 the return value, the referenced memory location is constant and
7540 unchanging.</p>
7541
7542</div>
7543
7544<!-- _______________________________________________________________________ -->
7545<div class="doc_subsubsection">
7546 <a name="int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a>
7547</div>
7548
7549<div class="doc_text">
7550
7551<h5>Syntax:</h5>
7552<pre>
7553 declare void @llvm.invariant.end({}* &lt;start&gt;, i64 &lt;size&gt;, i8* nocapture &lt;ptr&gt;)
7554</pre>
7555
7556<h5>Overview:</h5>
7557<p>The '<tt>llvm.invariant.end</tt>' intrinsic specifies that the contents of
7558 a memory object are mutable.</p>
7559
7560<h5>Arguments:</h5>
7561<p>The first argument is the matching <tt>llvm.invariant.start</tt> intrinsic.
Nick Lewycky9bc89042009-10-13 07:57:33 +00007562 The second argument is a constant integer representing the size of the
7563 object, or -1 if it is variable sized and the third argument is a pointer
7564 to the object.</p>
Nick Lewycky6f7d8342009-10-13 07:03:23 +00007565
7566<h5>Semantics:</h5>
7567<p>This intrinsic indicates that the memory is mutable again.</p>
7568
7569</div>
7570
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00007571<!-- ======================================================================= -->
7572<div class="doc_subsection">
Tanya Lattnercb1b9602007-06-15 20:50:54 +00007573 <a name="int_general">General Intrinsics</a>
7574</div>
7575
7576<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007577
7578<p>This class of intrinsics is designed to be generic and has no specific
7579 purpose.</p>
7580
Tanya Lattnercb1b9602007-06-15 20:50:54 +00007581</div>
7582
7583<!-- _______________________________________________________________________ -->
7584<div class="doc_subsubsection">
7585 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
7586</div>
7587
7588<div class="doc_text">
7589
7590<h5>Syntax:</h5>
7591<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007592 declare void @llvm.var.annotation(i8* &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
Tanya Lattnercb1b9602007-06-15 20:50:54 +00007593</pre>
7594
7595<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007596<p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
Tanya Lattnercb1b9602007-06-15 20:50:54 +00007597
7598<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007599<p>The first argument is a pointer to a value, the second is a pointer to a
7600 global string, the third is a pointer to a global string which is the source
7601 file name, and the last argument is the line number.</p>
Tanya Lattnercb1b9602007-06-15 20:50:54 +00007602
7603<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007604<p>This intrinsic allows annotation of local variables with arbitrary strings.
7605 This can be useful for special purpose optimizations that want to look for
7606 these annotations. These have no other defined use, they are ignored by code
7607 generation and optimization.</p>
Tanya Lattnercb1b9602007-06-15 20:50:54 +00007608
Tanya Lattnercb1b9602007-06-15 20:50:54 +00007609</div>
7610
Tanya Lattner293c0372007-09-21 22:59:12 +00007611<!-- _______________________________________________________________________ -->
7612<div class="doc_subsubsection">
Tanya Lattner0186a652007-09-21 23:57:59 +00007613 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattner293c0372007-09-21 22:59:12 +00007614</div>
7615
7616<div class="doc_text">
7617
7618<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007619<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
7620 any integer bit width.</p>
7621
Tanya Lattner293c0372007-09-21 22:59:12 +00007622<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007623 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
7624 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
7625 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
7626 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
7627 declare i256 @llvm.annotation.i256(i256 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
Tanya Lattner293c0372007-09-21 22:59:12 +00007628</pre>
7629
7630<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007631<p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
Tanya Lattner293c0372007-09-21 22:59:12 +00007632
7633<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007634<p>The first argument is an integer value (result of some expression), the
7635 second is a pointer to a global string, the third is a pointer to a global
7636 string which is the source file name, and the last argument is the line
7637 number. It returns the value of the first argument.</p>
Tanya Lattner293c0372007-09-21 22:59:12 +00007638
7639<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007640<p>This intrinsic allows annotations to be put on arbitrary expressions with
7641 arbitrary strings. This can be useful for special purpose optimizations that
7642 want to look for these annotations. These have no other defined use, they
7643 are ignored by code generation and optimization.</p>
Tanya Lattner293c0372007-09-21 22:59:12 +00007644
Tanya Lattner293c0372007-09-21 22:59:12 +00007645</div>
Jim Laskey2211f492007-03-14 19:31:19 +00007646
Anton Korobeynikov06cbb652008-01-15 22:31:34 +00007647<!-- _______________________________________________________________________ -->
7648<div class="doc_subsubsection">
7649 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
7650</div>
7651
7652<div class="doc_text">
7653
7654<h5>Syntax:</h5>
7655<pre>
7656 declare void @llvm.trap()
7657</pre>
7658
7659<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007660<p>The '<tt>llvm.trap</tt>' intrinsic.</p>
Anton Korobeynikov06cbb652008-01-15 22:31:34 +00007661
7662<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007663<p>None.</p>
Anton Korobeynikov06cbb652008-01-15 22:31:34 +00007664
7665<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007666<p>This intrinsics is lowered to the target dependent trap instruction. If the
7667 target does not have a trap instruction, this intrinsic will be lowered to
7668 the call of the <tt>abort()</tt> function.</p>
Anton Korobeynikov06cbb652008-01-15 22:31:34 +00007669
Anton Korobeynikov06cbb652008-01-15 22:31:34 +00007670</div>
7671
Bill Wendling14313312008-11-19 05:56:17 +00007672<!-- _______________________________________________________________________ -->
7673<div class="doc_subsubsection">
Misha Brukman50de2b22008-11-22 23:55:29 +00007674 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
Bill Wendling14313312008-11-19 05:56:17 +00007675</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007676
Bill Wendling14313312008-11-19 05:56:17 +00007677<div class="doc_text">
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007678
Bill Wendling14313312008-11-19 05:56:17 +00007679<h5>Syntax:</h5>
7680<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007681 declare void @llvm.stackprotector(i8* &lt;guard&gt;, i8** &lt;slot&gt;)
Bill Wendling14313312008-11-19 05:56:17 +00007682</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007683
Bill Wendling14313312008-11-19 05:56:17 +00007684<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007685<p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
7686 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
7687 ensure that it is placed on the stack before local variables.</p>
7688
Bill Wendling14313312008-11-19 05:56:17 +00007689<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007690<p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
7691 arguments. The first argument is the value loaded from the stack
7692 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
7693 that has enough space to hold the value of the guard.</p>
7694
Bill Wendling14313312008-11-19 05:56:17 +00007695<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +00007696<p>This intrinsic causes the prologue/epilogue inserter to force the position of
7697 the <tt>AllocaInst</tt> stack slot to be before local variables on the
7698 stack. This is to ensure that if a local variable on the stack is
7699 overwritten, it will destroy the value of the guard. When the function exits,
7700 the guard on the stack is checked against the original guard. If they're
7701 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
7702 function.</p>
7703
Bill Wendling14313312008-11-19 05:56:17 +00007704</div>
7705
Eric Christopher73484322009-11-30 08:03:53 +00007706<!-- _______________________________________________________________________ -->
7707<div class="doc_subsubsection">
7708 <a name="int_objectsize">'<tt>llvm.objectsize</tt>' Intrinsic</a>
7709</div>
7710
7711<div class="doc_text">
7712
7713<h5>Syntax:</h5>
7714<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +00007715 declare i32 @llvm.objectsize.i32(i8* &lt;object&gt;, i1 &lt;type&gt;)
7716 declare i64 @llvm.objectsize.i64(i8* &lt;object&gt;, i1 &lt;type&gt;)
Eric Christopher73484322009-11-30 08:03:53 +00007717</pre>
7718
7719<h5>Overview:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +00007720<p>The <tt>llvm.objectsize</tt> intrinsic is designed to provide information
Eric Christopher3070e162010-01-08 21:42:39 +00007721 to the optimizers to discover at compile time either a) when an
Eric Christopher455c5772009-12-05 02:46:03 +00007722 operation like memcpy will either overflow a buffer that corresponds to
7723 an object, or b) to determine that a runtime check for overflow isn't
7724 necessary. An object in this context means an allocation of a
Eric Christopher31e39bd2009-12-23 00:29:49 +00007725 specific class, structure, array, or other object.</p>
Eric Christopher73484322009-11-30 08:03:53 +00007726
7727<h5>Arguments:</h5>
7728<p>The <tt>llvm.objectsize</tt> intrinsic takes two arguments. The first
Eric Christopher31e39bd2009-12-23 00:29:49 +00007729 argument is a pointer to or into the <tt>object</tt>. The second argument
7730 is a boolean 0 or 1. This argument determines whether you want the
7731 maximum (0) or minimum (1) bytes remaining. This needs to be a literal 0 or
7732 1, variables are not allowed.</p>
7733
Eric Christopher73484322009-11-30 08:03:53 +00007734<h5>Semantics:</h5>
7735<p>The <tt>llvm.objectsize</tt> intrinsic is lowered to either a constant
Eric Christopher455c5772009-12-05 02:46:03 +00007736 representing the size of the object concerned or <tt>i32/i64 -1 or 0</tt>
7737 (depending on the <tt>type</tt> argument if the size cannot be determined
7738 at compile time.</p>
Eric Christopher73484322009-11-30 08:03:53 +00007739
7740</div>
7741
Chris Lattner2f7c9632001-06-06 20:29:01 +00007742<!-- *********************************************************************** -->
Chris Lattner2f7c9632001-06-06 20:29:01 +00007743<hr>
Misha Brukmanc501f552004-03-01 17:47:27 +00007744<address>
7745 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman86242e12008-12-11 17:34:48 +00007746 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Misha Brukmanc501f552004-03-01 17:47:27 +00007747 <a href="http://validator.w3.org/check/referer"><img
Misha Brukman86242e12008-12-11 17:34:48 +00007748 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
Misha Brukmanc501f552004-03-01 17:47:27 +00007749
7750 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
Reid Spencerca058542006-03-14 05:39:39 +00007751 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
Misha Brukmanc501f552004-03-01 17:47:27 +00007752 Last modified: $Date$
7753</address>
Chris Lattnerb8f816e2008-01-04 04:33:49 +00007754
Misha Brukman76307852003-11-08 01:05:38 +00007755</body>
7756</html>