Gitiles
Code Review Sign In
gerrit-public.fairphone.software / toolchain / llvm-project / ff2c52ce63adfac29b241c2a87da0edfce9289d3 / . / llvm / docs / LangRef.html
blob: 145c19f3e5ee231ce8ff3afc0aadcf95a1974238 [file] [log] [blame]
Misha Brukmanc501f552004-03-01 17:47:27 +0000[diff] [blame]1<!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 +0000[diff] [blame]3<html>
4<head>
5 <title>LLVM Assembly Language Reference Manual</title>
Reid Spencercb84e432004-08-26 20:44:00 +0000[diff] [blame]6 <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 +0000[diff] [blame]8 <meta name="description"
Reid Spencercb84e432004-08-26 20:44:00 +0000[diff] [blame]9 content="LLVM Assembly Language Reference Manual.">
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]10 <link rel="stylesheet" href="llvm.css" type="text/css">
11</head>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]12
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]13<body>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]14
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]15<h1>LLVM Language Reference Manual</h1>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]16<ol>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]17 <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 +0000[diff] [blame]20 <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 +0000[diff] [blame]23 <li><a href="#linkage">Linkage Types</a>
24 <ol>
Bill Wendling8693ef82009-07-20 02:41:50 +0000[diff] [blame]25 <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 +0000[diff] [blame]27 <li><a href="#linkage_linker_private_weak">'<tt>linker_private_weak</tt>' Linkage</a></li>
Bill Wendling578ee402010-08-20 22:05:50 +0000[diff] [blame]28 <li><a href="#linkage_linker_private_weak_def_auto">'<tt>linker_private_weak_def_auto</tt>' Linkage</a></li>
Bill Wendling8693ef82009-07-20 02:41:50 +0000[diff] [blame]29 <li><a href="#linkage_internal">'<tt>internal</tt>' Linkage</a></li>
30 <li><a href="#linkage_available_externally">'<tt>available_externally</tt>' Linkage</a></li>
31 <li><a href="#linkage_linkonce">'<tt>linkonce</tt>' Linkage</a></li>
32 <li><a href="#linkage_common">'<tt>common</tt>' Linkage</a></li>
33 <li><a href="#linkage_weak">'<tt>weak</tt>' Linkage</a></li>
34 <li><a href="#linkage_appending">'<tt>appending</tt>' Linkage</a></li>
35 <li><a href="#linkage_externweak">'<tt>extern_weak</tt>' Linkage</a></li>
Chris Lattner80d73c72009-10-10 18:26:06 +0000[diff] [blame]36 <li><a href="#linkage_linkonce_odr">'<tt>linkonce_odr</tt>' Linkage</a></li>
Bill Wendling8693ef82009-07-20 02:41:50 +0000[diff] [blame]37 <li><a href="#linkage_weak">'<tt>weak_odr</tt>' Linkage</a></li>
Bill Wendlingb4d076e2011-10-11 06:41:28 +0000[diff] [blame]38 <li><a href="#linkage_external">'<tt>external</tt>' Linkage</a></li>
Bill Wendling8693ef82009-07-20 02:41:50 +0000[diff] [blame]39 <li><a href="#linkage_dllimport">'<tt>dllimport</tt>' Linkage</a></li>
40 <li><a href="#linkage_dllexport">'<tt>dllexport</tt>' Linkage</a></li>
Bill Wendlinga3c6f6b2009-07-20 01:03:30 +0000[diff] [blame]41 </ol>
42 </li>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]43 <li><a href="#callingconv">Calling Conventions</a></li>
Chris Lattnerbc088212009-01-11 20:53:49 +0000[diff] [blame]44 <li><a href="#namedtypes">Named Types</a></li>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]45 <li><a href="#globalvars">Global Variables</a></li>
Chris Lattner91c15c42006-01-23 23:23:47 +0000[diff] [blame]46 <li><a href="#functionstructure">Functions</a></li>
Dan Gohmanef9462f2008-10-14 16:51:45 +0000[diff] [blame]47 <li><a href="#aliasstructure">Aliases</a></li>
Devang Pateld1a89692010-01-11 19:35:55 +0000[diff] [blame]48 <li><a href="#namedmetadatastructure">Named Metadata</a></li>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]49 <li><a href="#paramattrs">Parameter Attributes</a></li>
Devang Patel9eb525d2008-09-26 23:51:19 +0000[diff] [blame]50 <li><a href="#fnattrs">Function Attributes</a></li>
Gordon Henriksen71183b62007-12-10 03:18:06 +0000[diff] [blame]51 <li><a href="#gc">Garbage Collector Names</a></li>
Chris Lattner91c15c42006-01-23 23:23:47 +0000[diff] [blame]52 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]53 <li><a href="#datalayout">Data Layout</a></li>
Dan Gohman6154a012009-07-27 18:07:55 +0000[diff] [blame]54 <li><a href="#pointeraliasing">Pointer Aliasing Rules</a></li>
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +0000[diff] [blame]55 <li><a href="#volatile">Volatile Memory Accesses</a></li>
Eli Friedman35b54aa2011-07-20 21:35:53 +0000[diff] [blame]56 <li><a href="#memmodel">Memory Model for Concurrent Operations</a></li>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]57 <li><a href="#ordering">Atomic Memory Ordering Constraints</a></li>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]58 </ol>
59 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]60 <li><a href="#typesystem">Type System</a>
61 <ol>
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]62 <li><a href="#t_classifications">Type Classifications</a></li>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]63 <li><a href="#t_primitive">Primitive Types</a>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]64 <ol>
Nick Lewycky84a1eeb2009-09-27 00:45:11 +0000[diff] [blame]65 <li><a href="#t_integer">Integer Type</a></li>
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]66 <li><a href="#t_floating">Floating Point Types</a></li>
Dale Johannesen33e5c352010-10-01 00:48:59 +0000[diff] [blame]67 <li><a href="#t_x86mmx">X86mmx Type</a></li>
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]68 <li><a href="#t_void">Void Type</a></li>
69 <li><a href="#t_label">Label Type</a></li>
Nick Lewyckyadbc2842009-05-30 05:06:04 +0000[diff] [blame]70 <li><a href="#t_metadata">Metadata Type</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]71 </ol>
72 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]73 <li><a href="#t_derived">Derived Types</a>
74 <ol>
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]75 <li><a href="#t_aggregate">Aggregate Types</a>
76 <ol>
77 <li><a href="#t_array">Array Type</a></li>
78 <li><a href="#t_struct">Structure Type</a></li>
Chris Lattner2a843822011-07-23 19:59:08 +0000[diff] [blame]79 <li><a href="#t_opaque">Opaque Structure Types</a></li>
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]80 <li><a href="#t_vector">Vector Type</a></li>
81 </ol>
82 </li>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]83 <li><a href="#t_function">Function Type</a></li>
84 <li><a href="#t_pointer">Pointer Type</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]85 </ol>
86 </li>
87 </ol>
88 </li>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]89 <li><a href="#constants">Constants</a>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]90 <ol>
Dan Gohmanef9462f2008-10-14 16:51:45 +0000[diff] [blame]91 <li><a href="#simpleconstants">Simple Constants</a></li>
Chris Lattner361bfcd2009-02-28 18:32:25 +0000[diff] [blame]92 <li><a href="#complexconstants">Complex Constants</a></li>
Dan Gohmanef9462f2008-10-14 16:51:45 +0000[diff] [blame]93 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
94 <li><a href="#undefvalues">Undefined Values</a></li>
Dan Gohmanffc9a6b2010-04-22 23:14:21 +0000[diff] [blame]95 <li><a href="#trapvalues">Trap Values</a></li>
Chris Lattner2bfd3202009-10-27 21:19:13 +0000[diff] [blame]96 <li><a href="#blockaddress">Addresses of Basic Blocks</a></li>
Dan Gohmanef9462f2008-10-14 16:51:45 +0000[diff] [blame]97 <li><a href="#constantexprs">Constant Expressions</a></li>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]98 </ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]99 </li>
Chris Lattner98f013c2006-01-25 23:47:57 +0000[diff] [blame]100 <li><a href="#othervalues">Other Values</a>
101 <ol>
Dan Gohmanef9462f2008-10-14 16:51:45 +0000[diff] [blame]102 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
Devang Pateld1a89692010-01-11 19:35:55 +0000[diff] [blame]103 <li><a href="#metadata">Metadata Nodes and Metadata Strings</a></li>
Chris Lattner98f013c2006-01-25 23:47:57 +0000[diff] [blame]104 </ol>
105 </li>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]106 <li><a href="#intrinsic_globals">Intrinsic Global Variables</a>
107 <ol>
108 <li><a href="#intg_used">The '<tt>llvm.used</tt>' Global Variable</a></li>
Chris Lattner58f9bb22009-07-20 06:14:25 +0000[diff] [blame]109 <li><a href="#intg_compiler_used">The '<tt>llvm.compiler.used</tt>'
110 Global Variable</a></li>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]111 <li><a href="#intg_global_ctors">The '<tt>llvm.global_ctors</tt>'
112 Global Variable</a></li>
113 <li><a href="#intg_global_dtors">The '<tt>llvm.global_dtors</tt>'
114 Global Variable</a></li>
115 </ol>
116 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]117 <li><a href="#instref">Instruction Reference</a>
118 <ol>
119 <li><a href="#terminators">Terminator Instructions</a>
120 <ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]121 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
122 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]123 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
Chris Lattnerd04cb6d2009-10-28 00:19:10 +0000[diff] [blame]124 <li><a href="#i_indirectbr">'<tt>indirectbr</tt>' Instruction</a></li>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]125 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]126 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
Bill Wendlingf891bf82011-07-31 06:30:59 +0000[diff] [blame]127 <li><a href="#i_resume">'<tt>resume</tt>' Instruction</a></li>
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]128 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]129 </ol>
130 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]131 <li><a href="#binaryops">Binary Operations</a>
132 <ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]133 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]134 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]135 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]136 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]137 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]138 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]139 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
140 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
141 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
Reid Spencer7eb55b32006-11-02 01:53:59 +0000[diff] [blame]142 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
143 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
144 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]145 </ol>
146 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]147 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
148 <ol>
Reid Spencer2ab01932007-02-02 13:57:07 +0000[diff] [blame]149 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
150 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
151 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]152 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]153 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]154 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]155 </ol>
156 </li>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]157 <li><a href="#vectorops">Vector Operations</a>
158 <ol>
159 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
160 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
161 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]162 </ol>
163 </li>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]164 <li><a href="#aggregateops">Aggregate Operations</a>
165 <ol>
166 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
167 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
168 </ol>
169 </li>
Chris Lattner6ab66722006-08-15 00:45:58 +0000[diff] [blame]170 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]171 <ol>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]172 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
173 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
174 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
175 <li><a href="#i_fence">'<tt>fence</tt>' Instruction</a></li>
176 <li><a href="#i_cmpxchg">'<tt>cmpxchg</tt>' Instruction</a></li>
177 <li><a href="#i_atomicrmw">'<tt>atomicrmw</tt>' Instruction</a></li>
Robert Bocchino820bc75b2006-02-17 21:18:08 +0000[diff] [blame]178 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]179 </ol>
180 </li>
Reid Spencer97c5fa42006-11-08 01:18:52 +0000[diff] [blame]181 <li><a href="#convertops">Conversion Operations</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]182 <ol>
183 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
184 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
185 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
186 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
187 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
Reid Spencer51b07252006-11-09 23:03:26 +0000[diff] [blame]188 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
189 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
190 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
191 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
Reid Spencerb7344ff2006-11-11 21:00:47 +0000[diff] [blame]192 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
193 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
Reid Spencer5b950642006-11-11 23:08:07 +0000[diff] [blame]194 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]195 </ol>
Dan Gohmanef9462f2008-10-14 16:51:45 +0000[diff] [blame]196 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]197 <li><a href="#otherops">Other Operations</a>
198 <ol>
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]199 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
200 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]201 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]202 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]203 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
Chris Lattner33337472006-01-13 23:26:01 +0000[diff] [blame]204 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
Bill Wendlingbbcb7cd2011-08-02 21:52:38 +0000[diff] [blame]205 <li><a href="#i_landingpad">'<tt>landingpad</tt>' Instruction</a></li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]206 </ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]207 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]208 </ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]209 </li>
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]210 <li><a href="#intrinsics">Intrinsic Functions</a>
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]211 <ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]212 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
213 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]214 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
215 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
216 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]217 </ol>
218 </li>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]219 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
220 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]221 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
222 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
223 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]224 </ol>
225 </li>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]226 <li><a href="#int_codegen">Code Generator Intrinsics</a>
227 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]228 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
229 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
230 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
231 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
232 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
233 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
Dan Gohmane58f7b32010-05-26 21:56:15 +0000[diff] [blame]234 <li><a href="#int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
John Criswellaa1c3c12004-04-09 16:43:20 +0000[diff] [blame]235 </ol>
236 </li>
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]237 <li><a href="#int_libc">Standard C Library Intrinsics</a>
238 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]239 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
240 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
241 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
242 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
243 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]244 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
245 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
246 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Dan Gohmane635c522011-05-27 00:36:31 +0000[diff] [blame]247 <li><a href="#int_exp">'<tt>llvm.exp.*</tt>' Intrinsic</a></li>
248 <li><a href="#int_log">'<tt>llvm.log.*</tt>' Intrinsic</a></li>
Cameron Zwarichf03fa182011-07-08 21:39:21 +0000[diff] [blame]249 <li><a href="#int_fma">'<tt>llvm.fma.*</tt>' Intrinsic</a></li>
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]250 </ol>
251 </li>
Nate Begeman0f223bb2006-01-13 23:26:38 +0000[diff] [blame]252 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]253 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]254 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
Chris Lattnerb748c672006-01-16 22:34:14 +0000[diff] [blame]255 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
256 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
257 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]258 </ol>
259 </li>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]260 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
261 <ol>
Bill Wendlingfd2bd722009-02-08 04:04:40 +0000[diff] [blame]262 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
263 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
264 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
265 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
266 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
Bill Wendlingb9a73272009-02-08 23:00:09 +0000[diff] [blame]267 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]268 </ol>
269 </li>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +0000[diff] [blame]270 <li><a href="#int_fp16">Half Precision Floating Point Intrinsics</a>
271 <ol>
Chris Lattnerbbd8bd32010-03-14 23:03:31 +0000[diff] [blame]272 <li><a href="#int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a></li>
273 <li><a href="#int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a></li>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +0000[diff] [blame]274 </ol>
275 </li>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]276 <li><a href="#int_debugger">Debugger intrinsics</a></li>
Jim Laskey2211f492007-03-14 19:31:19 +0000[diff] [blame]277 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sandsa0984362011-09-06 13:37:06 +0000[diff] [blame]278 <li><a href="#int_trampoline">Trampoline Intrinsics</a>
Duncan Sands644f9172007-07-27 12:58:54 +0000[diff] [blame]279 <ol>
280 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sandsa0984362011-09-06 13:37:06 +0000[diff] [blame]281 <li><a href="#int_at">'<tt>llvm.adjust.trampoline</tt>' Intrinsic</a></li>
Duncan Sands644f9172007-07-27 12:58:54 +0000[diff] [blame]282 </ol>
283 </li>
Bill Wendlingf85850f2008-11-18 22:10:53 +0000[diff] [blame]284 <li><a href="#int_atomics">Atomic intrinsics</a>
285 <ol>
286 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
287 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
288 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
289 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
290 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
291 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
292 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
293 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
294 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
295 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
296 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
297 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
298 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
299 </ol>
300 </li>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]301 <li><a href="#int_memorymarkers">Memory Use Markers</a>
302 <ol>
303 <li><a href="#int_lifetime_start"><tt>llvm.lifetime.start</tt></a></li>
304 <li><a href="#int_lifetime_end"><tt>llvm.lifetime.end</tt></a></li>
305 <li><a href="#int_invariant_start"><tt>llvm.invariant.start</tt></a></li>
306 <li><a href="#int_invariant_end"><tt>llvm.invariant.end</tt></a></li>
307 </ol>
308 </li>
Reid Spencer5b2cb0f2007-07-20 19:59:11 +0000[diff] [blame]309 <li><a href="#int_general">General intrinsics</a>
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000[diff] [blame]310 <ol>
Reid Spencer5b2cb0f2007-07-20 19:59:11 +0000[diff] [blame]311 <li><a href="#int_var_annotation">
Bill Wendling14313312008-11-19 05:56:17 +0000[diff] [blame]312 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Tanya Lattner293c0372007-09-21 22:59:12 +0000[diff] [blame]313 <li><a href="#int_annotation">
Bill Wendling14313312008-11-19 05:56:17 +0000[diff] [blame]314 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Anton Korobeynikov06cbb652008-01-15 22:31:34 +0000[diff] [blame]315 <li><a href="#int_trap">
Bill Wendling14313312008-11-19 05:56:17 +0000[diff] [blame]316 '<tt>llvm.trap</tt>' Intrinsic</a></li>
317 <li><a href="#int_stackprotector">
318 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
Eric Christopher73484322009-11-30 08:03:53 +0000[diff] [blame]319 <li><a href="#int_objectsize">
320 '<tt>llvm.objectsize</tt>' Intrinsic</a></li>
Tanya Lattner293c0372007-09-21 22:59:12 +0000[diff] [blame]321 </ol>
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000[diff] [blame]322 </li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]323 </ol>
324 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]325</ol>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]326
327<div class="doc_author">
328 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
329 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]330</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]331
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]332<!-- *********************************************************************** -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]333<h2><a name="abstract">Abstract</a></h2>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]334<!-- *********************************************************************** -->
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]335
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]336<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]337
338<p>This document is a reference manual for the LLVM assembly language. LLVM is
339 a Static Single Assignment (SSA) based representation that provides type
340 safety, low-level operations, flexibility, and the capability of representing
341 'all' high-level languages cleanly. It is the common code representation
342 used throughout all phases of the LLVM compilation strategy.</p>
343
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]344</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]345
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]346<!-- *********************************************************************** -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]347<h2><a name="introduction">Introduction</a></h2>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]348<!-- *********************************************************************** -->
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]349
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]350<div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]351
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]352<p>The LLVM code representation is designed to be used in three different forms:
353 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
354 for fast loading by a Just-In-Time compiler), and as a human readable
355 assembly language representation. This allows LLVM to provide a powerful
356 intermediate representation for efficient compiler transformations and
357 analysis, while providing a natural means to debug and visualize the
358 transformations. The three different forms of LLVM are all equivalent. This
359 document describes the human readable representation and notation.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]360
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]361<p>The LLVM representation aims to be light-weight and low-level while being
362 expressive, typed, and extensible at the same time. It aims to be a
363 "universal IR" of sorts, by being at a low enough level that high-level ideas
364 may be cleanly mapped to it (similar to how microprocessors are "universal
365 IR's", allowing many source languages to be mapped to them). By providing
366 type information, LLVM can be used as the target of optimizations: for
367 example, through pointer analysis, it can be proven that a C automatic
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]368 variable is never accessed outside of the current function, allowing it to
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]369 be promoted to a simple SSA value instead of a memory location.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]370
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]371<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]372<h4>
373 <a name="wellformed">Well-Formedness</a>
374</h4>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]375
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]376<div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]377
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]378<p>It is important to note that this document describes 'well formed' LLVM
379 assembly language. There is a difference between what the parser accepts and
380 what is considered 'well formed'. For example, the following instruction is
381 syntactically okay, but not well formed:</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]382
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]383<pre class="doc_code">
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]384%x = <a href="#i_add">add</a> i32 1, %x
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]385</pre>
386
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]387<p>because the definition of <tt>%x</tt> does not dominate all of its uses. The
388 LLVM infrastructure provides a verification pass that may be used to verify
389 that an LLVM module is well formed. This pass is automatically run by the
390 parser after parsing input assembly and by the optimizer before it outputs
391 bitcode. The violations pointed out by the verifier pass indicate bugs in
392 transformation passes or input to the parser.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]393
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]394</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]395
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]396</div>
397
Chris Lattner87a3dbe2007-10-03 17:34:29 +0000[diff] [blame]398<!-- Describe the typesetting conventions here. -->
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]399
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]400<!-- *********************************************************************** -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]401<h2><a name="identifiers">Identifiers</a></h2>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]402<!-- *********************************************************************** -->
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]403
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]404<div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]405
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]406<p>LLVM identifiers come in two basic types: global and local. Global
407 identifiers (functions, global variables) begin with the <tt>'@'</tt>
408 character. Local identifiers (register names, types) begin with
409 the <tt>'%'</tt> character. Additionally, there are three different formats
410 for identifiers, for different purposes:</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]411
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]412<ol>
Reid Spencerb23b65f2007-08-07 14:34:28 +0000[diff] [blame]413 <li>Named values are represented as a string of characters with their prefix.
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]414 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
415 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
416 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
417 other characters in their names can be surrounded with quotes. Special
418 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
419 ASCII code for the character in hexadecimal. In this way, any character
420 can be used in a name value, even quotes themselves.</li>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]421
Reid Spencerb23b65f2007-08-07 14:34:28 +0000[diff] [blame]422 <li>Unnamed values are represented as an unsigned numeric value with their
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]423 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]424
Reid Spencer8f08d802004-12-09 18:02:53 +0000[diff] [blame]425 <li>Constants, which are described in a <a href="#constants">section about
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]426 constants</a>, below.</li>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]427</ol>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]428
Reid Spencerb23b65f2007-08-07 14:34:28 +0000[diff] [blame]429<p>LLVM requires that values start with a prefix for two reasons: Compilers
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]430 don't need to worry about name clashes with reserved words, and the set of
431 reserved words may be expanded in the future without penalty. Additionally,
432 unnamed identifiers allow a compiler to quickly come up with a temporary
433 variable without having to avoid symbol table conflicts.</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]434
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]435<p>Reserved words in LLVM are very similar to reserved words in other
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]436 languages. There are keywords for different opcodes
437 ('<tt><a href="#i_add">add</a></tt>',
438 '<tt><a href="#i_bitcast">bitcast</a></tt>',
439 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
440 ('<tt><a href="#t_void">void</a></tt>',
441 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
442 reserved words cannot conflict with variable names, because none of them
443 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]444
445<p>Here is an example of LLVM code to multiply the integer variable
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]446 '<tt>%X</tt>' by 8:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]447
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]448<p>The easy way:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]449
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]450<pre class="doc_code">
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]451%result = <a href="#i_mul">mul</a> i32 %X, 8
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]452</pre>
453
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]454<p>After strength reduction:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]455
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]456<pre class="doc_code">
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]457%result = <a href="#i_shl">shl</a> i32 %X, i8 3
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]458</pre>
459
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]460<p>And the hard way:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]461
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]462<pre class="doc_code">
Gabor Greifbd0328f2009-10-28 13:05:07 +0000[diff] [blame]463%0 = <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
464%1 = <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]465%result = <a href="#i_add">add</a> i32 %1, %1
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]466</pre>
467
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]468<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
469 lexical features of LLVM:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]470
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]471<ol>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]472 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]473 line.</li>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]474
475 <li>Unnamed temporaries are created when the result of a computation is not
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]476 assigned to a named value.</li>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]477
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]478 <li>Unnamed temporaries are numbered sequentially</li>
479</ol>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]480
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]481<p>It also shows a convention that we follow in this document. When
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]482 demonstrating instructions, we will follow an instruction with a comment that
483 defines the type and name of value produced. Comments are shown in italic
484 text.</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]485
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]486</div>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]487
488<!-- *********************************************************************** -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]489<h2><a name="highlevel">High Level Structure</a></h2>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]490<!-- *********************************************************************** -->
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]491<div>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]492<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]493<h3>
494 <a name="modulestructure">Module Structure</a>
495</h3>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]496
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]497<div>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]498
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]499<p>LLVM programs are composed of "Module"s, each of which is a translation unit
500 of the input programs. Each module consists of functions, global variables,
501 and symbol table entries. Modules may be combined together with the LLVM
502 linker, which merges function (and global variable) definitions, resolves
503 forward declarations, and merges symbol table entries. Here is an example of
504 the "hello world" module:</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]505
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]506<pre class="doc_code">
Chris Lattner54a7be72010-08-17 17:13:42 +0000[diff] [blame]507<i>; Declare the string constant as a global constant.</i>&nbsp;
Nick Lewyckyfea7ddc2011-01-29 01:09:53 +0000[diff] [blame]508<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a>&nbsp;<a href="#globalvars">constant</a>&nbsp;<a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>&nbsp;
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]509
Chris Lattner54a7be72010-08-17 17:13:42 +0000[diff] [blame]510<i>; External declaration of the puts function</i>&nbsp;
511<a href="#functionstructure">declare</a> i32 @puts(i8*) <i>; i32 (i8*)* </i>&nbsp;
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]512
513<i>; Definition of main function</i>
Chris Lattner54a7be72010-08-17 17:13:42 +0000[diff] [blame]514define i32 @main() { <i>; i32()* </i>&nbsp;
515 <i>; Convert [13 x i8]* to i8 *...</i>&nbsp;
516 %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 +0000[diff] [blame]517
Chris Lattner54a7be72010-08-17 17:13:42 +0000[diff] [blame]518 <i>; Call puts function to write out the string to stdout.</i>&nbsp;
519 <a href="#i_call">call</a> i32 @puts(i8* %cast210) <i>; i32</i>&nbsp;
520 <a href="#i_ret">ret</a> i32 0&nbsp;
521}
Devang Pateld1a89692010-01-11 19:35:55 +0000[diff] [blame]522
523<i>; Named metadata</i>
524!1 = metadata !{i32 41}
525!foo = !{!1, null}
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]526</pre>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]527
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]528<p>This example is made up of a <a href="#globalvars">global variable</a> named
Devang Pateld1a89692010-01-11 19:35:55 +0000[diff] [blame]529 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function,
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]530 a <a href="#functionstructure">function definition</a> for
Devang Pateld1a89692010-01-11 19:35:55 +0000[diff] [blame]531 "<tt>main</tt>" and <a href="#namedmetadatastructure">named metadata</a>
532 "<tt>foo"</tt>.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]533
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]534<p>In general, a module is made up of a list of global values, where both
535 functions and global variables are global values. Global values are
536 represented by a pointer to a memory location (in this case, a pointer to an
537 array of char, and a pointer to a function), and have one of the
538 following <a href="#linkage">linkage types</a>.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]539
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]540</div>
541
542<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]543<h3>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]544 <a name="linkage">Linkage Types</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]545</h3>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]546
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]547<div>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]548
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]549<p>All Global Variables and Functions have one of the following types of
550 linkage:</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]551
552<dl>
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]553 <dt><tt><b><a name="linkage_private">private</a></b></tt></dt>
Bill Wendling03bcd6e2010-07-01 21:55:59 +0000[diff] [blame]554 <dd>Global values with "<tt>private</tt>" linkage are only directly accessible
555 by objects in the current module. In particular, linking code into a
556 module with an private global value may cause the private to be renamed as
557 necessary to avoid collisions. Because the symbol is private to the
558 module, all references can be updated. This doesn't show up in any symbol
559 table in the object file.</dd>
Rafael Espindola6de96a12009-01-15 20:18:42 +0000[diff] [blame]560
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]561 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt></dt>
Bill Wendling03bcd6e2010-07-01 21:55:59 +0000[diff] [blame]562 <dd>Similar to <tt>private</tt>, but the symbol is passed through the
563 assembler and evaluated by the linker. Unlike normal strong symbols, they
564 are removed by the linker from the final linked image (executable or
565 dynamic library).</dd>
566
567 <dt><tt><b><a name="linkage_linker_private_weak">linker_private_weak</a></b></tt></dt>
568 <dd>Similar to "<tt>linker_private</tt>", but the symbol is weak. Note that
569 <tt>linker_private_weak</tt> symbols are subject to coalescing by the
570 linker. The symbols are removed by the linker from the final linked image
571 (executable or dynamic library).</dd>
Bill Wendlinga3c6f6b2009-07-20 01:03:30 +0000[diff] [blame]572
Bill Wendling578ee402010-08-20 22:05:50 +0000[diff] [blame]573 <dt><tt><b><a name="linkage_linker_private_weak_def_auto">linker_private_weak_def_auto</a></b></tt></dt>
574 <dd>Similar to "<tt>linker_private_weak</tt>", but it's known that the address
575 of the object is not taken. For instance, functions that had an inline
576 definition, but the compiler decided not to inline it. Note,
577 unlike <tt>linker_private</tt> and <tt>linker_private_weak</tt>,
578 <tt>linker_private_weak_def_auto</tt> may have only <tt>default</tt>
579 visibility. The symbols are removed by the linker from the final linked
580 image (executable or dynamic library).</dd>
581
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]582 <dt><tt><b><a name="linkage_internal">internal</a></b></tt></dt>
Bill Wendling36321712010-06-29 22:34:52 +0000[diff] [blame]583 <dd>Similar to private, but the value shows as a local symbol
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]584 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
585 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]586
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]587 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt></dt>
Chris Lattner184f1be2009-04-13 05:44:34 +0000[diff] [blame]588 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]589 into the object file corresponding to the LLVM module. They exist to
590 allow inlining and other optimizations to take place given knowledge of
591 the definition of the global, which is known to be somewhere outside the
592 module. Globals with <tt>available_externally</tt> linkage are allowed to
593 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
594 This linkage type is only allowed on definitions, not declarations.</dd>
Chris Lattner184f1be2009-04-13 05:44:34 +0000[diff] [blame]595
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]596 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt></dt>
Chris Lattnere20b4702007-01-14 06:51:48 +0000[diff] [blame]597 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
Chris Lattner0de4caa2010-01-09 19:15:14 +0000[diff] [blame]598 the same name when linkage occurs. This can be used to implement
599 some forms of inline functions, templates, or other code which must be
600 generated in each translation unit that uses it, but where the body may
601 be overridden with a more definitive definition later. Unreferenced
602 <tt>linkonce</tt> globals are allowed to be discarded. Note that
603 <tt>linkonce</tt> linkage does not actually allow the optimizer to
604 inline the body of this function into callers because it doesn't know if
605 this definition of the function is the definitive definition within the
606 program or whether it will be overridden by a stronger definition.
607 To enable inlining and other optimizations, use "<tt>linkonce_odr</tt>"
608 linkage.</dd>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]609
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]610 <dt><tt><b><a name="linkage_weak">weak</a></b></tt></dt>
Chris Lattnerd0554882009-08-05 05:21:07 +0000[diff] [blame]611 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
612 <tt>linkonce</tt> linkage, except that unreferenced globals with
613 <tt>weak</tt> linkage may not be discarded. This is used for globals that
614 are declared "weak" in C source code.</dd>
615
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]616 <dt><tt><b><a name="linkage_common">common</a></b></tt></dt>
Chris Lattnerd0554882009-08-05 05:21:07 +0000[diff] [blame]617 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
618 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
619 global scope.
620 Symbols with "<tt>common</tt>" linkage are merged in the same way as
621 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
Chris Lattner0aff0b22009-08-05 05:41:44 +0000[diff] [blame]622 <tt>common</tt> symbols may not have an explicit section,
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]623 must have a zero initializer, and may not be marked '<a
Chris Lattner0aff0b22009-08-05 05:41:44 +0000[diff] [blame]624 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
625 have common linkage.</dd>
Chris Lattnerd0554882009-08-05 05:21:07 +0000[diff] [blame]626
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]627
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]628 <dt><tt><b><a name="linkage_appending">appending</a></b></tt></dt>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]629 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]630 pointer to array type. When two global variables with appending linkage
631 are linked together, the two global arrays are appended together. This is
632 the LLVM, typesafe, equivalent of having the system linker append together
633 "sections" with identical names when .o files are linked.</dd>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]634
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]635 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]636 <dd>The semantics of this linkage follow the ELF object file model: the symbol
637 is weak until linked, if not linked, the symbol becomes null instead of
638 being an undefined reference.</dd>
Anton Korobeynikova0554d92007-01-12 19:20:47 +0000[diff] [blame]639
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]640 <dt><tt><b><a name="linkage_linkonce_odr">linkonce_odr</a></b></tt></dt>
641 <dt><tt><b><a name="linkage_weak_odr">weak_odr</a></b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]642 <dd>Some languages allow differing globals to be merged, such as two functions
643 with different semantics. Other languages, such as <tt>C++</tt>, ensure
Bill Wendling03bcd6e2010-07-01 21:55:59 +0000[diff] [blame]644 that only equivalent globals are ever merged (the "one definition rule"
645 &mdash; "ODR"). Such languages can use the <tt>linkonce_odr</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]646 and <tt>weak_odr</tt> linkage types to indicate that the global will only
647 be merged with equivalent globals. These linkage types are otherwise the
648 same as their non-<tt>odr</tt> versions.</dd>
Duncan Sands12da8ce2009-03-07 15:45:40 +0000[diff] [blame]649
Bill Wendlingb4d076e2011-10-11 06:41:28 +0000[diff] [blame]650 <dt><tt><b><a name="linkage_external">external</a></b></tt>:</dt>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000[diff] [blame]651 <dd>If none of the above identifiers are used, the global is externally
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]652 visible, meaning that it participates in linkage and can be used to
653 resolve external symbol references.</dd>
Reid Spencer7972c472007-04-11 23:49:50 +0000[diff] [blame]654</dl>
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000[diff] [blame]655
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]656<p>The next two types of linkage are targeted for Microsoft Windows platform
657 only. They are designed to support importing (exporting) symbols from (to)
658 DLLs (Dynamic Link Libraries).</p>
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000[diff] [blame]659
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]660<dl>
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]661 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt></dt>
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000[diff] [blame]662 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]663 or variable via a global pointer to a pointer that is set up by the DLL
664 exporting the symbol. On Microsoft Windows targets, the pointer name is
665 formed by combining <code>__imp_</code> and the function or variable
666 name.</dd>
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000[diff] [blame]667
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]668 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt></dt>
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000[diff] [blame]669 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]670 pointer to a pointer in a DLL, so that it can be referenced with the
671 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
672 name is formed by combining <code>__imp_</code> and the function or
673 variable name.</dd>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]674</dl>
675
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]676<p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
677 another module defined a "<tt>.LC0</tt>" variable and was linked with this
678 one, one of the two would be renamed, preventing a collision. Since
679 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
680 declarations), they are accessible outside of the current module.</p>
681
682<p>It is illegal for a function <i>declaration</i> to have any linkage type
Bill Wendlingb4d076e2011-10-11 06:41:28 +0000[diff] [blame]683 other than <tt>external</tt>, <tt>dllimport</tt>
684 or <tt>extern_weak</tt>.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]685
Duncan Sands12da8ce2009-03-07 15:45:40 +0000[diff] [blame]686<p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]687 or <tt>weak_odr</tt> linkages.</p>
688
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]689</div>
690
691<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]692<h3>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]693 <a name="callingconv">Calling Conventions</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]694</h3>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]695
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]696<div>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]697
698<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]699 and <a href="#i_invoke">invokes</a> can all have an optional calling
700 convention specified for the call. The calling convention of any pair of
701 dynamic caller/callee must match, or the behavior of the program is
702 undefined. The following calling conventions are supported by LLVM, and more
703 may be added in the future:</p>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]704
705<dl>
706 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]707 <dd>This calling convention (the default if no other calling convention is
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]708 specified) matches the target C calling conventions. This calling
709 convention supports varargs function calls and tolerates some mismatch in
710 the declared prototype and implemented declaration of the function (as
711 does normal C).</dd>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]712
713 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]714 <dd>This calling convention attempts to make calls as fast as possible
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]715 (e.g. by passing things in registers). This calling convention allows the
716 target to use whatever tricks it wants to produce fast code for the
717 target, without having to conform to an externally specified ABI
Jeffrey Yasskinb8677462010-01-09 19:44:16 +0000[diff] [blame]718 (Application Binary Interface).
719 <a href="CodeGenerator.html#tailcallopt">Tail calls can only be optimized
Chris Lattnera179e4d2010-03-11 00:22:57 +0000[diff] [blame]720 when this or the GHC convention is used.</a> This calling convention
721 does not support varargs and requires the prototype of all callees to
722 exactly match the prototype of the function definition.</dd>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]723
724 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]725 <dd>This calling convention attempts to make code in the caller as efficient
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]726 as possible under the assumption that the call is not commonly executed.
727 As such, these calls often preserve all registers so that the call does
728 not break any live ranges in the caller side. This calling convention
729 does not support varargs and requires the prototype of all callees to
730 exactly match the prototype of the function definition.</dd>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]731
Chris Lattnera179e4d2010-03-11 00:22:57 +0000[diff] [blame]732 <dt><b>"<tt>cc <em>10</em></tt>" - GHC convention</b>:</dt>
733 <dd>This calling convention has been implemented specifically for use by the
734 <a href="http://www.haskell.org/ghc">Glasgow Haskell Compiler (GHC)</a>.
735 It passes everything in registers, going to extremes to achieve this by
736 disabling callee save registers. This calling convention should not be
737 used lightly but only for specific situations such as an alternative to
738 the <em>register pinning</em> performance technique often used when
739 implementing functional programming languages.At the moment only X86
740 supports this convention and it has the following limitations:
741 <ul>
742 <li>On <em>X86-32</em> only supports up to 4 bit type parameters. No
743 floating point types are supported.</li>
744 <li>On <em>X86-64</em> only supports up to 10 bit type parameters and
745 6 floating point parameters.</li>
746 </ul>
747 This calling convention supports
748 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> but
749 requires both the caller and callee are using it.
750 </dd>
751
Chris Lattner573f64e2005-05-07 01:46:40 +0000[diff] [blame]752 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]753 <dd>Any calling convention may be specified by number, allowing
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]754 target-specific calling conventions to be used. Target specific calling
755 conventions start at 64.</dd>
Chris Lattner573f64e2005-05-07 01:46:40 +0000[diff] [blame]756</dl>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]757
758<p>More calling conventions can be added/defined on an as-needed basis, to
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]759 support Pascal conventions or any other well-known target-independent
760 convention.</p>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]761
762</div>
763
764<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]765<h3>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000[diff] [blame]766 <a name="visibility">Visibility Styles</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]767</h3>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000[diff] [blame]768
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]769<div>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000[diff] [blame]770
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]771<p>All Global Variables and Functions have one of the following visibility
772 styles:</p>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000[diff] [blame]773
774<dl>
775 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
Chris Lattner67c37d12008-08-05 18:29:16 +0000[diff] [blame]776 <dd>On targets that use the ELF object file format, default visibility means
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]777 that the declaration is visible to other modules and, in shared libraries,
778 means that the declared entity may be overridden. On Darwin, default
779 visibility means that the declaration is visible to other modules. Default
780 visibility corresponds to "external linkage" in the language.</dd>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000[diff] [blame]781
782 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000[diff] [blame]783 <dd>Two declarations of an object with hidden visibility refer to the same
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]784 object if they are in the same shared object. Usually, hidden visibility
785 indicates that the symbol will not be placed into the dynamic symbol
786 table, so no other module (executable or shared library) can reference it
787 directly.</dd>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000[diff] [blame]788
Anton Korobeynikov39f3cff2007-04-29 18:35:00 +0000[diff] [blame]789 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
Anton Korobeynikov39f3cff2007-04-29 18:35:00 +0000[diff] [blame]790 <dd>On ELF, protected visibility indicates that the symbol will be placed in
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]791 the dynamic symbol table, but that references within the defining module
792 will bind to the local symbol. That is, the symbol cannot be overridden by
793 another module.</dd>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000[diff] [blame]794</dl>
795
796</div>
797
798<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]799<h3>
Chris Lattnerbc088212009-01-11 20:53:49 +0000[diff] [blame]800 <a name="namedtypes">Named Types</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]801</h3>
Chris Lattnerbc088212009-01-11 20:53:49 +0000[diff] [blame]802
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]803<div>
Chris Lattnerbc088212009-01-11 20:53:49 +0000[diff] [blame]804
805<p>LLVM IR allows you to specify name aliases for certain types. This can make
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]806 it easier to read the IR and make the IR more condensed (particularly when
807 recursive types are involved). An example of a name specification is:</p>
Chris Lattnerbc088212009-01-11 20:53:49 +0000[diff] [blame]808
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]809<pre class="doc_code">
Chris Lattnerbc088212009-01-11 20:53:49 +0000[diff] [blame]810%mytype = type { %mytype*, i32 }
811</pre>
Chris Lattnerbc088212009-01-11 20:53:49 +0000[diff] [blame]812
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]813<p>You may give a name to any <a href="#typesystem">type</a> except
Chris Lattner249b9762010-08-17 23:26:04 +0000[diff] [blame]814 "<a href="#t_void">void</a>". Type name aliases may be used anywhere a type
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]815 is expected with the syntax "%mytype".</p>
Chris Lattnerbc088212009-01-11 20:53:49 +0000[diff] [blame]816
817<p>Note that type names are aliases for the structural type that they indicate,
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]818 and that you can therefore specify multiple names for the same type. This
819 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
820 uses structural typing, the name is not part of the type. When printing out
821 LLVM IR, the printer will pick <em>one name</em> to render all types of a
822 particular shape. This means that if you have code where two different
823 source types end up having the same LLVM type, that the dumper will sometimes
824 print the "wrong" or unexpected type. This is an important design point and
825 isn't going to change.</p>
Chris Lattnerbc088212009-01-11 20:53:49 +0000[diff] [blame]826
827</div>
828
Chris Lattnerbc088212009-01-11 20:53:49 +0000[diff] [blame]829<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]830<h3>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]831 <a name="globalvars">Global Variables</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]832</h3>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]833
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]834<div>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]835
Chris Lattner5d5aede2005-02-12 19:30:21 +0000[diff] [blame]836<p>Global variables define regions of memory allocated at compilation time
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]837 instead of run-time. Global variables may optionally be initialized, may
838 have an explicit section to be placed in, and may have an optional explicit
839 alignment specified. A variable may be defined as "thread_local", which
840 means that it will not be shared by threads (each thread will have a
841 separated copy of the variable). A variable may be defined as a global
842 "constant," which indicates that the contents of the variable
843 will <b>never</b> be modified (enabling better optimization, allowing the
844 global data to be placed in the read-only section of an executable, etc).
845 Note that variables that need runtime initialization cannot be marked
846 "constant" as there is a store to the variable.</p>
Chris Lattner5d5aede2005-02-12 19:30:21 +0000[diff] [blame]847
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]848<p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
849 constant, even if the final definition of the global is not. This capability
850 can be used to enable slightly better optimization of the program, but
851 requires the language definition to guarantee that optimizations based on the
852 'constantness' are valid for the translation units that do not include the
853 definition.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]854
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]855<p>As SSA values, global variables define pointer values that are in scope
856 (i.e. they dominate) all basic blocks in the program. Global variables
857 always define a pointer to their "content" type because they describe a
858 region of memory, and all memory objects in LLVM are accessed through
859 pointers.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]860
Rafael Espindola45e6c192011-01-08 16:42:36 +0000[diff] [blame]861<p>Global variables can be marked with <tt>unnamed_addr</tt> which indicates
862 that the address is not significant, only the content. Constants marked
Rafael Espindolaf1ed7812011-01-15 08:20:57 +0000[diff] [blame]863 like this can be merged with other constants if they have the same
864 initializer. Note that a constant with significant address <em>can</em>
865 be merged with a <tt>unnamed_addr</tt> constant, the result being a
866 constant whose address is significant.</p>
Rafael Espindola45e6c192011-01-08 16:42:36 +0000[diff] [blame]867
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]868<p>A global variable may be declared to reside in a target-specific numbered
869 address space. For targets that support them, address spaces may affect how
870 optimizations are performed and/or what target instructions are used to
871 access the variable. The default address space is zero. The address space
872 qualifier must precede any other attributes.</p>
Christopher Lamb308121c2007-12-11 09:31:00 +0000[diff] [blame]873
Chris Lattner662c8722005-11-12 00:45:07 +0000[diff] [blame]874<p>LLVM allows an explicit section to be specified for globals. If the target
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]875 supports it, it will emit globals to the section specified.</p>
Chris Lattner662c8722005-11-12 00:45:07 +0000[diff] [blame]876
Chris Lattner78e00bc2010-04-28 00:13:42 +0000[diff] [blame]877<p>An explicit alignment may be specified for a global, which must be a power
878 of 2. If not present, or if the alignment is set to zero, the alignment of
879 the global is set by the target to whatever it feels convenient. If an
880 explicit alignment is specified, the global is forced to have exactly that
Chris Lattner4bd85e42010-04-28 00:31:12 +0000[diff] [blame]881 alignment. Targets and optimizers are not allowed to over-align the global
882 if the global has an assigned section. In this case, the extra alignment
883 could be observable: for example, code could assume that the globals are
884 densely packed in their section and try to iterate over them as an array,
885 alignment padding would break this iteration.</p>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]886
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]887<p>For example, the following defines a global in a numbered address space with
888 an initializer, section, and alignment:</p>
Chris Lattner5760c502007-01-14 00:27:09 +0000[diff] [blame]889
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]890<pre class="doc_code">
Dan Gohmanaaa679b2009-01-11 00:40:00 +0000[diff] [blame]891@G = addrspace(5) constant float 1.0, section "foo", align 4
Chris Lattner5760c502007-01-14 00:27:09 +0000[diff] [blame]892</pre>
893
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]894</div>
895
896
897<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]898<h3>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]899 <a name="functionstructure">Functions</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]900</h3>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]901
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]902<div>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]903
Dan Gohmana269a0a2010-03-01 17:41:39 +0000[diff] [blame]904<p>LLVM function definitions consist of the "<tt>define</tt>" keyword, an
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]905 optional <a href="#linkage">linkage type</a>, an optional
906 <a href="#visibility">visibility style</a>, an optional
Rafael Espindola45e6c192011-01-08 16:42:36 +0000[diff] [blame]907 <a href="#callingconv">calling convention</a>,
908 an optional <tt>unnamed_addr</tt> attribute, a return type, an optional
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]909 <a href="#paramattrs">parameter attribute</a> for the return type, a function
910 name, a (possibly empty) argument list (each with optional
911 <a href="#paramattrs">parameter attributes</a>), optional
912 <a href="#fnattrs">function attributes</a>, an optional section, an optional
913 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
914 curly brace, a list of basic blocks, and a closing curly brace.</p>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000[diff] [blame]915
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]916<p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
917 optional <a href="#linkage">linkage type</a>, an optional
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]918 <a href="#visibility">visibility style</a>, an optional
Rafael Espindola45e6c192011-01-08 16:42:36 +0000[diff] [blame]919 <a href="#callingconv">calling convention</a>,
920 an optional <tt>unnamed_addr</tt> attribute, a return type, an optional
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]921 <a href="#paramattrs">parameter attribute</a> for the return type, a function
922 name, a possibly empty list of arguments, an optional alignment, and an
923 optional <a href="#gc">garbage collector name</a>.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]924
Chris Lattner67c37d12008-08-05 18:29:16 +0000[diff] [blame]925<p>A function definition contains a list of basic blocks, forming the CFG
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]926 (Control Flow Graph) for the function. Each basic block may optionally start
927 with a label (giving the basic block a symbol table entry), contains a list
928 of instructions, and ends with a <a href="#terminators">terminator</a>
929 instruction (such as a branch or function return).</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]930
Chris Lattnera59fb102007-06-08 16:52:14 +0000[diff] [blame]931<p>The first basic block in a function is special in two ways: it is immediately
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]932 executed on entrance to the function, and it is not allowed to have
933 predecessor basic blocks (i.e. there can not be any branches to the entry
934 block of a function). Because the block can have no predecessors, it also
935 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]936
Chris Lattner662c8722005-11-12 00:45:07 +0000[diff] [blame]937<p>LLVM allows an explicit section to be specified for functions. If the target
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]938 supports it, it will emit functions to the section specified.</p>
Chris Lattner662c8722005-11-12 00:45:07 +0000[diff] [blame]939
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]940<p>An explicit alignment may be specified for a function. If not present, or if
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]941 the alignment is set to zero, the alignment of the function is set by the
942 target to whatever it feels convenient. If an explicit alignment is
943 specified, the function is forced to have at least that much alignment. All
944 alignments must be a power of 2.</p>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]945
Rafael Espindola45e6c192011-01-08 16:42:36 +0000[diff] [blame]946<p>If the <tt>unnamed_addr</tt> attribute is given, the address is know to not
947 be significant and two identical functions can be merged</p>.
948
Bill Wendling30235112009-07-20 02:39:26 +0000[diff] [blame]949<h5>Syntax:</h5>
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]950<pre class="doc_code">
Chris Lattner0ae02092008-10-13 16:55:18 +0000[diff] [blame]951define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]952 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
953 &lt;ResultType&gt; @&lt;FunctionName&gt; ([argument list])
954 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
955 [<a href="#gc">gc</a>] { ... }
956</pre>
Devang Patel02256232008-10-07 17:48:33 +0000[diff] [blame]957
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]958</div>
959
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000[diff] [blame]960<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]961<h3>
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000[diff] [blame]962 <a name="aliasstructure">Aliases</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]963</h3>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]964
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]965<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]966
967<p>Aliases act as "second name" for the aliasee value (which can be either
968 function, global variable, another alias or bitcast of global value). Aliases
969 may have an optional <a href="#linkage">linkage type</a>, and an
970 optional <a href="#visibility">visibility style</a>.</p>
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000[diff] [blame]971
Bill Wendling30235112009-07-20 02:39:26 +0000[diff] [blame]972<h5>Syntax:</h5>
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]973<pre class="doc_code">
Duncan Sands7e99a942008-09-12 20:48:21 +0000[diff] [blame]974@&lt;Name&gt; = alias [Linkage] [Visibility] &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
Bill Wendling2d8b9a82007-05-29 09:42:13 +0000[diff] [blame]975</pre>
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000[diff] [blame]976
977</div>
978
Chris Lattner91c15c42006-01-23 23:23:47 +0000[diff] [blame]979<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]980<h3>
Devang Pateld1a89692010-01-11 19:35:55 +0000[diff] [blame]981 <a name="namedmetadatastructure">Named Metadata</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]982</h3>
Devang Pateld1a89692010-01-11 19:35:55 +0000[diff] [blame]983
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]984<div>
Devang Pateld1a89692010-01-11 19:35:55 +0000[diff] [blame]985
Chris Lattnerc2f8f162010-01-15 21:50:19 +0000[diff] [blame]986<p>Named metadata is a collection of metadata. <a href="#metadata">Metadata
Dan Gohman093cb792010-07-21 18:54:18 +0000[diff] [blame]987 nodes</a> (but not metadata strings) are the only valid operands for
Chris Lattnerc2f8f162010-01-15 21:50:19 +0000[diff] [blame]988 a named metadata.</p>
Devang Pateld1a89692010-01-11 19:35:55 +0000[diff] [blame]989
990<h5>Syntax:</h5>
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]991<pre class="doc_code">
Dan Gohman093cb792010-07-21 18:54:18 +0000[diff] [blame]992; Some unnamed metadata nodes, which are referenced by the named metadata.
993!0 = metadata !{metadata !"zero"}
Devang Pateld1a89692010-01-11 19:35:55 +0000[diff] [blame]994!1 = metadata !{metadata !"one"}
Dan Gohman093cb792010-07-21 18:54:18 +0000[diff] [blame]995!2 = metadata !{metadata !"two"}
Dan Gohman58cd65f2010-07-13 19:48:13 +0000[diff] [blame]996; A named metadata.
Dan Gohman093cb792010-07-21 18:54:18 +0000[diff] [blame]997!name = !{!0, !1, !2}
Devang Pateld1a89692010-01-11 19:35:55 +0000[diff] [blame]998</pre>
Devang Pateld1a89692010-01-11 19:35:55 +0000[diff] [blame]999
1000</div>
1001
1002<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1003<h3>
1004 <a name="paramattrs">Parameter Attributes</a>
1005</h3>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]1006
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1007<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1008
1009<p>The return type and each parameter of a function type may have a set of
1010 <i>parameter attributes</i> associated with them. Parameter attributes are
1011 used to communicate additional information about the result or parameters of
1012 a function. Parameter attributes are considered to be part of the function,
1013 not of the function type, so functions with different parameter attributes
1014 can have the same function type.</p>
1015
1016<p>Parameter attributes are simple keywords that follow the type specified. If
1017 multiple parameter attributes are needed, they are space separated. For
1018 example:</p>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]1019
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]1020<pre class="doc_code">
Nick Lewyckydac78d82009-02-15 23:06:14 +0000[diff] [blame]1021declare i32 @printf(i8* noalias nocapture, ...)
Chris Lattnerd2597d72008-10-04 18:33:34 +0000[diff] [blame]1022declare i32 @atoi(i8 zeroext)
1023declare signext i8 @returns_signed_char()
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]1024</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]1025
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1026<p>Note that any attributes for the function result (<tt>nounwind</tt>,
1027 <tt>readonly</tt>) come immediately after the argument list.</p>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]1028
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1029<p>Currently, only the following parameter attributes are defined:</p>
Chris Lattner5cee13f2008-01-11 06:20:47 +0000[diff] [blame]1030
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1031<dl>
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]1032 <dt><tt><b>zeroext</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1033 <dd>This indicates to the code generator that the parameter or return value
Cameron Zwarichac106272011-03-16 22:20:18 +0000[diff] [blame]1034 should be zero-extended to the extent required by the target's ABI (which
1035 is usually 32-bits, but is 8-bits for a i1 on x86-64) by the caller (for a
1036 parameter) or the callee (for a return value).</dd>
Chris Lattner5cee13f2008-01-11 06:20:47 +0000[diff] [blame]1037
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]1038 <dt><tt><b>signext</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1039 <dd>This indicates to the code generator that the parameter or return value
Cameron Zwarich341c36d2011-03-17 14:21:58 +0000[diff] [blame]1040 should be sign-extended to the extent required by the target's ABI (which
1041 is usually 32-bits) by the caller (for a parameter) or the callee (for a
1042 return value).</dd>
Chris Lattner5cee13f2008-01-11 06:20:47 +0000[diff] [blame]1043
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]1044 <dt><tt><b>inreg</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1045 <dd>This indicates that this parameter or return value should be treated in a
1046 special target-dependent fashion during while emitting code for a function
1047 call or return (usually, by putting it in a register as opposed to memory,
1048 though some targets use it to distinguish between two different kinds of
1049 registers). Use of this attribute is target-specific.</dd>
Chris Lattner5cee13f2008-01-11 06:20:47 +0000[diff] [blame]1050
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]1051 <dt><tt><b><a name="byval">byval</a></b></tt></dt>
Chris Lattnerd78dbee2010-11-20 23:49:06 +0000[diff] [blame]1052 <dd><p>This indicates that the pointer parameter should really be passed by
1053 value to the function. The attribute implies that a hidden copy of the
1054 pointee
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1055 is made between the caller and the callee, so the callee is unable to
1056 modify the value in the callee. This attribute is only valid on LLVM
1057 pointer arguments. It is generally used to pass structs and arrays by
1058 value, but is also valid on pointers to scalars. The copy is considered
1059 to belong to the caller not the callee (for example,
1060 <tt><a href="#readonly">readonly</a></tt> functions should not write to
1061 <tt>byval</tt> parameters). This is not a valid attribute for return
Chris Lattnerd78dbee2010-11-20 23:49:06 +0000[diff] [blame]1062 values.</p>
1063
1064 <p>The byval attribute also supports specifying an alignment with
1065 the align attribute. It indicates the alignment of the stack slot to
1066 form and the known alignment of the pointer specified to the call site. If
1067 the alignment is not specified, then the code generator makes a
1068 target-specific assumption.</p></dd>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1069
Dan Gohman3770af52010-07-02 23:18:08 +0000[diff] [blame]1070 <dt><tt><b><a name="sret">sret</a></b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1071 <dd>This indicates that the pointer parameter specifies the address of a
1072 structure that is the return value of the function in the source program.
1073 This pointer must be guaranteed by the caller to be valid: loads and
1074 stores to the structure may be assumed by the callee to not to trap. This
1075 may only be applied to the first parameter. This is not a valid attribute
1076 for return values. </dd>
1077
Dan Gohman3770af52010-07-02 23:18:08 +0000[diff] [blame]1078 <dt><tt><b><a name="noalias">noalias</a></b></tt></dt>
Dan Gohmandf12d082010-07-02 18:41:32 +0000[diff] [blame]1079 <dd>This indicates that pointer values
1080 <a href="#pointeraliasing"><i>based</i></a> on the argument or return
Dan Gohmande256292010-07-02 23:46:54 +0000[diff] [blame]1081 value do not alias pointer values which are not <i>based</i> on it,
1082 ignoring certain "irrelevant" dependencies.
1083 For a call to the parent function, dependencies between memory
1084 references from before or after the call and from those during the call
1085 are "irrelevant" to the <tt>noalias</tt> keyword for the arguments and
1086 return value used in that call.
Dan Gohmandf12d082010-07-02 18:41:32 +0000[diff] [blame]1087 The caller shares the responsibility with the callee for ensuring that
1088 these requirements are met.
1089 For further details, please see the discussion of the NoAlias response in
Dan Gohman6c858db2010-07-06 15:26:33 +0000[diff] [blame]1090 <a href="AliasAnalysis.html#MustMayNo">alias analysis</a>.<br>
1091<br>
John McCall72ed8902010-07-06 21:07:14 +0000[diff] [blame]1092 Note that this definition of <tt>noalias</tt> is intentionally
1093 similar to the definition of <tt>restrict</tt> in C99 for function
Chris Lattner5eff9ca2010-07-06 20:51:35 +0000[diff] [blame]1094 arguments, though it is slightly weaker.
Dan Gohman6c858db2010-07-06 15:26:33 +0000[diff] [blame]1095<br>
1096 For function return values, C99's <tt>restrict</tt> is not meaningful,
1097 while LLVM's <tt>noalias</tt> is.
1098 </dd>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1099
Dan Gohman3770af52010-07-02 23:18:08 +0000[diff] [blame]1100 <dt><tt><b><a name="nocapture">nocapture</a></b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1101 <dd>This indicates that the callee does not make any copies of the pointer
1102 that outlive the callee itself. This is not a valid attribute for return
1103 values.</dd>
1104
Dan Gohman3770af52010-07-02 23:18:08 +0000[diff] [blame]1105 <dt><tt><b><a name="nest">nest</a></b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1106 <dd>This indicates that the pointer parameter can be excised using the
1107 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
1108 attribute for return values.</dd>
1109</dl>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]1110
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]1111</div>
1112
1113<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1114<h3>
Gordon Henriksen71183b62007-12-10 03:18:06 +0000[diff] [blame]1115 <a name="gc">Garbage Collector Names</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1116</h3>
Gordon Henriksen71183b62007-12-10 03:18:06 +0000[diff] [blame]1117
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1118<div>
Gordon Henriksen71183b62007-12-10 03:18:06 +0000[diff] [blame]1119
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1120<p>Each function may specify a garbage collector name, which is simply a
1121 string:</p>
1122
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]1123<pre class="doc_code">
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]1124define void @f() gc "name" { ... }
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1125</pre>
Gordon Henriksen71183b62007-12-10 03:18:06 +0000[diff] [blame]1126
1127<p>The compiler declares the supported values of <i>name</i>. Specifying a
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1128 collector which will cause the compiler to alter its output in order to
1129 support the named garbage collection algorithm.</p>
1130
Gordon Henriksen71183b62007-12-10 03:18:06 +0000[diff] [blame]1131</div>
1132
1133<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1134<h3>
Devang Patel9eb525d2008-09-26 23:51:19 +0000[diff] [blame]1135 <a name="fnattrs">Function Attributes</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1136</h3>
Devang Patelcaacdba2008-09-04 23:05:13 +0000[diff] [blame]1137
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1138<div>
Devang Patel9eb525d2008-09-26 23:51:19 +0000[diff] [blame]1139
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1140<p>Function attributes are set to communicate additional information about a
1141 function. Function attributes are considered to be part of the function, not
1142 of the function type, so functions with different parameter attributes can
1143 have the same function type.</p>
Devang Patel9eb525d2008-09-26 23:51:19 +0000[diff] [blame]1144
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1145<p>Function attributes are simple keywords that follow the type specified. If
1146 multiple attributes are needed, they are space separated. For example:</p>
Devang Patelcaacdba2008-09-04 23:05:13 +0000[diff] [blame]1147
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]1148<pre class="doc_code">
Devang Patel9eb525d2008-09-26 23:51:19 +0000[diff] [blame]1149define void @f() noinline { ... }
1150define void @f() alwaysinline { ... }
1151define void @f() alwaysinline optsize { ... }
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]1152define void @f() optsize { ... }
Bill Wendlingb175fa42008-09-07 10:26:33 +0000[diff] [blame]1153</pre>
Devang Patelcaacdba2008-09-04 23:05:13 +0000[diff] [blame]1154
Bill Wendlingb175fa42008-09-07 10:26:33 +0000[diff] [blame]1155<dl>
Charles Davisbe5557e2010-02-12 00:31:15 +0000[diff] [blame]1156 <dt><tt><b>alignstack(&lt;<em>n</em>&gt;)</b></tt></dt>
1157 <dd>This attribute indicates that, when emitting the prologue and epilogue,
1158 the backend should forcibly align the stack pointer. Specify the
1159 desired alignment, which must be a power of two, in parentheses.
1160
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]1161 <dt><tt><b>alwaysinline</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1162 <dd>This attribute indicates that the inliner should attempt to inline this
1163 function into callers whenever possible, ignoring any active inlining size
1164 threshold for this caller.</dd>
Bill Wendlingb175fa42008-09-07 10:26:33 +0000[diff] [blame]1165
Dan Gohman8bd11f12011-06-16 16:03:13 +0000[diff] [blame]1166 <dt><tt><b>nonlazybind</b></tt></dt>
1167 <dd>This attribute suppresses lazy symbol binding for the function. This
1168 may make calls to the function faster, at the cost of extra program
1169 startup time if the function is not called during program startup.</dd>
1170
Jakob Stoklund Olesen74bb06c2010-02-06 01:16:28 +0000[diff] [blame]1171 <dt><tt><b>inlinehint</b></tt></dt>
1172 <dd>This attribute indicates that the source code contained a hint that inlining
1173 this function is desirable (such as the "inline" keyword in C/C++). It
1174 is just a hint; it imposes no requirements on the inliner.</dd>
1175
Nick Lewycky14b58da2010-07-06 18:24:09 +0000[diff] [blame]1176 <dt><tt><b>naked</b></tt></dt>
1177 <dd>This attribute disables prologue / epilogue emission for the function.
1178 This can have very system-specific consequences.</dd>
1179
1180 <dt><tt><b>noimplicitfloat</b></tt></dt>
1181 <dd>This attributes disables implicit floating point instructions.</dd>
1182
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]1183 <dt><tt><b>noinline</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1184 <dd>This attribute indicates that the inliner should never inline this
1185 function in any situation. This attribute may not be used together with
1186 the <tt>alwaysinline</tt> attribute.</dd>
Devang Patel9eb525d2008-09-26 23:51:19 +0000[diff] [blame]1187
Nick Lewycky14b58da2010-07-06 18:24:09 +0000[diff] [blame]1188 <dt><tt><b>noredzone</b></tt></dt>
1189 <dd>This attribute indicates that the code generator should not use a red
1190 zone, even if the target-specific ABI normally permits it.</dd>
Devang Patel9eb525d2008-09-26 23:51:19 +0000[diff] [blame]1191
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]1192 <dt><tt><b>noreturn</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1193 <dd>This function attribute indicates that the function never returns
1194 normally. This produces undefined behavior at runtime if the function
1195 ever does dynamically return.</dd>
Bill Wendlinga8130172008-11-13 01:02:51 +0000[diff] [blame]1196
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]1197 <dt><tt><b>nounwind</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1198 <dd>This function attribute indicates that the function never returns with an
1199 unwind or exceptional control flow. If the function does unwind, its
1200 runtime behavior is undefined.</dd>
Bill Wendling0f5541e2008-11-26 19:07:40 +0000[diff] [blame]1201
Nick Lewycky14b58da2010-07-06 18:24:09 +0000[diff] [blame]1202 <dt><tt><b>optsize</b></tt></dt>
1203 <dd>This attribute suggests that optimization passes and code generator passes
1204 make choices that keep the code size of this function low, and otherwise
1205 do optimizations specifically to reduce code size.</dd>
1206
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]1207 <dt><tt><b>readnone</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1208 <dd>This attribute indicates that the function computes its result (or decides
1209 to unwind an exception) based strictly on its arguments, without
1210 dereferencing any pointer arguments or otherwise accessing any mutable
1211 state (e.g. memory, control registers, etc) visible to caller functions.
1212 It does not write through any pointer arguments
1213 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1214 changes any state visible to callers. This means that it cannot unwind
1215 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1216 could use the <tt>unwind</tt> instruction.</dd>
Devang Patel310fd4a2009-06-12 19:45:19 +0000[diff] [blame]1217
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]1218 <dt><tt><b><a name="readonly">readonly</a></b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1219 <dd>This attribute indicates that the function does not write through any
1220 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1221 arguments) or otherwise modify any state (e.g. memory, control registers,
1222 etc) visible to caller functions. It may dereference pointer arguments
1223 and read state that may be set in the caller. A readonly function always
1224 returns the same value (or unwinds an exception identically) when called
1225 with the same set of arguments and global state. It cannot unwind an
1226 exception by calling the <tt>C++</tt> exception throwing methods, but may
1227 use the <tt>unwind</tt> instruction.</dd>
Anton Korobeynikovc8ce7b082009-07-17 18:07:26 +0000[diff] [blame]1228
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]1229 <dt><tt><b><a name="ssp">ssp</a></b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1230 <dd>This attribute indicates that the function should emit a stack smashing
1231 protector. It is in the form of a "canary"&mdash;a random value placed on
1232 the stack before the local variables that's checked upon return from the
1233 function to see if it has been overwritten. A heuristic is used to
1234 determine if a function needs stack protectors or not.<br>
1235<br>
1236 If a function that has an <tt>ssp</tt> attribute is inlined into a
1237 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1238 function will have an <tt>ssp</tt> attribute.</dd>
1239
Bill Wendling7f4a3362009-11-02 00:24:16 +0000[diff] [blame]1240 <dt><tt><b>sspreq</b></tt></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1241 <dd>This attribute indicates that the function should <em>always</em> emit a
1242 stack smashing protector. This overrides
Bill Wendling30235112009-07-20 02:39:26 +0000[diff] [blame]1243 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1244<br>
1245 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1246 function that doesn't have an <tt>sspreq</tt> attribute or which has
1247 an <tt>ssp</tt> attribute, then the resulting function will have
1248 an <tt>sspreq</tt> attribute.</dd>
Rafael Espindola163d6752011-07-25 15:27:59 +0000[diff] [blame]1249
1250 <dt><tt><b><a name="uwtable">uwtable</a></b></tt></dt>
1251 <dd>This attribute indicates that the ABI being targeted requires that
1252 an unwind table entry be produce for this function even if we can
1253 show that no exceptions passes by it. This is normally the case for
1254 the ELF x86-64 abi, but it can be disabled for some compilation
1255 units.</dd>
1256
Rafael Espindolacc349c82011-10-03 14:45:37 +0000[diff] [blame]1257 <dt><tt><b><a name="returns_twice">returns_twice</a></b></tt></dt>
1258 <dd>This attribute indicates that this function can return
1259 twice. The C <code>setjmp</code> is an example of such a function.
1260 The compiler disables some optimizations (like tail calls) in the caller of
1261 these functions.</dd>
Bill Wendlingb175fa42008-09-07 10:26:33 +0000[diff] [blame]1262</dl>
1263
Devang Patelcaacdba2008-09-04 23:05:13 +0000[diff] [blame]1264</div>
1265
1266<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1267<h3>
Chris Lattner93564892006-04-08 04:40:53 +0000[diff] [blame]1268 <a name="moduleasm">Module-Level Inline Assembly</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1269</h3>
Chris Lattner91c15c42006-01-23 23:23:47 +0000[diff] [blame]1270
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1271<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1272
1273<p>Modules may contain "module-level inline asm" blocks, which corresponds to
1274 the GCC "file scope inline asm" blocks. These blocks are internally
1275 concatenated by LLVM and treated as a single unit, but may be separated in
1276 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
Chris Lattner91c15c42006-01-23 23:23:47 +0000[diff] [blame]1277
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]1278<pre class="doc_code">
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]1279module asm "inline asm code goes here"
1280module asm "more can go here"
1281</pre>
Chris Lattner91c15c42006-01-23 23:23:47 +0000[diff] [blame]1282
1283<p>The strings can contain any character by escaping non-printable characters.
1284 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1285 for the number.</p>
Chris Lattner91c15c42006-01-23 23:23:47 +0000[diff] [blame]1286
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1287<p>The inline asm code is simply printed to the machine code .s file when
1288 assembly code is generated.</p>
1289
Chris Lattner91c15c42006-01-23 23:23:47 +0000[diff] [blame]1290</div>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]1291
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1292<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1293<h3>
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1294 <a name="datalayout">Data Layout</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1295</h3>
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1296
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1297<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1298
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1299<p>A module may specify a target specific data layout string that specifies how
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1300 data is to be laid out in memory. The syntax for the data layout is
1301 simply:</p>
1302
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]1303<pre class="doc_code">
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1304target datalayout = "<i>layout specification</i>"
1305</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1306
1307<p>The <i>layout specification</i> consists of a list of specifications
1308 separated by the minus sign character ('-'). Each specification starts with
1309 a letter and may include other information after the letter to define some
1310 aspect of the data layout. The specifications accepted are as follows:</p>
1311
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1312<dl>
1313 <dt><tt>E</tt></dt>
1314 <dd>Specifies that the target lays out data in big-endian form. That is, the
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1315 bits with the most significance have the lowest address location.</dd>
1316
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1317 <dt><tt>e</tt></dt>
Chris Lattner67c37d12008-08-05 18:29:16 +0000[diff] [blame]1318 <dd>Specifies that the target lays out data in little-endian form. That is,
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1319 the bits with the least significance have the lowest address
1320 location.</dd>
1321
Lang Hamesde7ab802011-10-10 23:42:08 +0000[diff] [blame]1322 <dt><tt>S<i>size</i></tt></dt>
1323 <dd>Specifies the natural alignment of the stack in bits. Alignment promotion
1324 of stack variables is limited to the natural stack alignment to avoid
1325 dynamic stack realignment. The stack alignment must be a multiple of
Lang Hamesff2c52c2011-10-11 17:50:14 +0000[diff] [blame^]1326 8-bits. If omitted, the natural stack alignment defaults to "unspecified",
1327 which does not prevent any alignment promotions.</dd>
Lang Hamesde7ab802011-10-10 23:42:08 +0000[diff] [blame]1328
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1329 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]1330 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1331 <i>preferred</i> alignments. All sizes are in bits. Specifying
1332 the <i>pref</i> alignment is optional. If omitted, the
1333 preceding <tt>:</tt> should be omitted too.</dd>
1334
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1335 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1336 <dd>This specifies the alignment for an integer type of a given bit
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1337 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1338
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1339 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]1340 <dd>This specifies the alignment for a vector type of a given bit
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1341 <i>size</i>.</dd>
1342
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1343 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]1344 <dd>This specifies the alignment for a floating point type of a given bit
Dale Johannesence522852010-05-28 18:54:47 +0000[diff] [blame]1345 <i>size</i>. Only values of <i>size</i> that are supported by the target
1346 will work. 32 (float) and 64 (double) are supported on all targets;
1347 80 or 128 (different flavors of long double) are also supported on some
1348 targets.
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1349
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1350 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1351 <dd>This specifies the alignment for an aggregate type of a given bit
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1352 <i>size</i>.</dd>
1353
Daniel Dunbar7921a592009-06-08 22:17:53 +0000[diff] [blame]1354 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1355 <dd>This specifies the alignment for a stack object of a given bit
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1356 <i>size</i>.</dd>
Chris Lattnera381eff2009-11-07 09:35:34 +0000[diff] [blame]1357
1358 <dt><tt>n<i>size1</i>:<i>size2</i>:<i>size3</i>...</tt></dt>
1359 <dd>This specifies a set of native integer widths for the target CPU
1360 in bits. For example, it might contain "n32" for 32-bit PowerPC,
1361 "n32:64" for PowerPC 64, or "n8:16:32:64" for X86-64. Elements of
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]1362 this set are considered to support most general arithmetic
Chris Lattnera381eff2009-11-07 09:35:34 +0000[diff] [blame]1363 operations efficiently.</dd>
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1364</dl>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1365
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1366<p>When constructing the data layout for a given target, LLVM starts with a
Dan Gohman61110ae2010-04-28 00:36:01 +0000[diff] [blame]1367 default set of specifications which are then (possibly) overridden by the
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1368 specifications in the <tt>datalayout</tt> keyword. The default specifications
1369 are given in this list:</p>
1370
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1371<ul>
1372 <li><tt>E</tt> - big endian</li>
Dan Gohman8ad777d2010-02-23 02:44:03 +0000[diff] [blame]1373 <li><tt>p:64:64:64</tt> - 64-bit pointers with 64-bit alignment</li>
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1374 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1375 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1376 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1377 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
Chris Lattner67c37d12008-08-05 18:29:16 +0000[diff] [blame]1378 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1379 alignment of 64-bits</li>
1380 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1381 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1382 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1383 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1384 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
Daniel Dunbar7921a592009-06-08 22:17:53 +0000[diff] [blame]1385 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1386</ul>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1387
1388<p>When LLVM is determining the alignment for a given type, it uses the
1389 following rules:</p>
1390
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1391<ol>
1392 <li>If the type sought is an exact match for one of the specifications, that
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1393 specification is used.</li>
1394
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1395 <li>If no match is found, and the type sought is an integer type, then the
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1396 smallest integer type that is larger than the bitwidth of the sought type
1397 is used. If none of the specifications are larger than the bitwidth then
1398 the the largest integer type is used. For example, given the default
1399 specifications above, the i7 type will use the alignment of i8 (next
1400 largest) while both i65 and i256 will use the alignment of i64 (largest
1401 specified).</li>
1402
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1403 <li>If no match is found, and the type sought is a vector type, then the
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1404 largest vector type that is smaller than the sought vector type will be
1405 used as a fall back. This happens because &lt;128 x double&gt; can be
1406 implemented in terms of 64 &lt;2 x double&gt;, for example.</li>
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1407</ol>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1408
Reid Spencer50c723a2007-02-19 23:54:10 +0000[diff] [blame]1409</div>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]1410
Dan Gohman6154a012009-07-27 18:07:55 +0000[diff] [blame]1411<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1412<h3>
Dan Gohman6154a012009-07-27 18:07:55 +0000[diff] [blame]1413 <a name="pointeraliasing">Pointer Aliasing Rules</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1414</h3>
Dan Gohman6154a012009-07-27 18:07:55 +0000[diff] [blame]1415
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1416<div>
Dan Gohman6154a012009-07-27 18:07:55 +0000[diff] [blame]1417
Andreas Bolka8ae4e242009-07-29 00:02:05 +0000[diff] [blame]1418<p>Any memory access must be done through a pointer value associated
Andreas Bolkae39f0332009-07-27 20:37:10 +0000[diff] [blame]1419with an address range of the memory access, otherwise the behavior
Dan Gohman6154a012009-07-27 18:07:55 +0000[diff] [blame]1420is undefined. Pointer values are associated with address ranges
1421according to the following rules:</p>
1422
1423<ul>
Dan Gohmandf12d082010-07-02 18:41:32 +0000[diff] [blame]1424 <li>A pointer value is associated with the addresses associated with
1425 any value it is <i>based</i> on.
Andreas Bolka8ae4e242009-07-29 00:02:05 +0000[diff] [blame]1426 <li>An address of a global variable is associated with the address
Dan Gohman6154a012009-07-27 18:07:55 +0000[diff] [blame]1427 range of the variable's storage.</li>
1428 <li>The result value of an allocation instruction is associated with
1429 the address range of the allocated storage.</li>
1430 <li>A null pointer in the default address-space is associated with
Andreas Bolka8ae4e242009-07-29 00:02:05 +0000[diff] [blame]1431 no address.</li>
Dan Gohman6154a012009-07-27 18:07:55 +0000[diff] [blame]1432 <li>An integer constant other than zero or a pointer value returned
1433 from a function not defined within LLVM may be associated with address
1434 ranges allocated through mechanisms other than those provided by
Andreas Bolka8ae4e242009-07-29 00:02:05 +0000[diff] [blame]1435 LLVM. Such ranges shall not overlap with any ranges of addresses
Dan Gohman6154a012009-07-27 18:07:55 +0000[diff] [blame]1436 allocated by mechanisms provided by LLVM.</li>
Dan Gohmandf12d082010-07-02 18:41:32 +0000[diff] [blame]1437</ul>
1438
1439<p>A pointer value is <i>based</i> on another pointer value according
1440 to the following rules:</p>
1441
1442<ul>
1443 <li>A pointer value formed from a
1444 <tt><a href="#i_getelementptr">getelementptr</a></tt> operation
1445 is <i>based</i> on the first operand of the <tt>getelementptr</tt>.</li>
1446 <li>The result value of a
1447 <tt><a href="#i_bitcast">bitcast</a></tt> is <i>based</i> on the operand
1448 of the <tt>bitcast</tt>.</li>
1449 <li>A pointer value formed by an
1450 <tt><a href="#i_inttoptr">inttoptr</a></tt> is <i>based</i> on all
1451 pointer values that contribute (directly or indirectly) to the
1452 computation of the pointer's value.</li>
1453 <li>The "<i>based</i> on" relationship is transitive.</li>
1454</ul>
1455
1456<p>Note that this definition of <i>"based"</i> is intentionally
1457 similar to the definition of <i>"based"</i> in C99, though it is
1458 slightly weaker.</p>
Dan Gohman6154a012009-07-27 18:07:55 +0000[diff] [blame]1459
1460<p>LLVM IR does not associate types with memory. The result type of a
Andreas Bolka8ae4e242009-07-29 00:02:05 +0000[diff] [blame]1461<tt><a href="#i_load">load</a></tt> merely indicates the size and
1462alignment of the memory from which to load, as well as the
Dan Gohman4eb47192010-06-17 19:23:50 +0000[diff] [blame]1463interpretation of the value. The first operand type of a
Andreas Bolka8ae4e242009-07-29 00:02:05 +0000[diff] [blame]1464<tt><a href="#i_store">store</a></tt> similarly only indicates the size
1465and alignment of the store.</p>
Dan Gohman6154a012009-07-27 18:07:55 +0000[diff] [blame]1466
1467<p>Consequently, type-based alias analysis, aka TBAA, aka
1468<tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1469LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1470additional information which specialized optimization passes may use
1471to implement type-based alias analysis.</p>
1472
1473</div>
1474
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +0000[diff] [blame]1475<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1476<h3>
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +0000[diff] [blame]1477 <a name="volatile">Volatile Memory Accesses</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1478</h3>
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +0000[diff] [blame]1479
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1480<div>
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +0000[diff] [blame]1481
1482<p>Certain memory accesses, such as <a href="#i_load"><tt>load</tt></a>s, <a
1483href="#i_store"><tt>store</tt></a>s, and <a
1484href="#int_memcpy"><tt>llvm.memcpy</tt></a>s may be marked <tt>volatile</tt>.
1485The optimizers must not change the number of volatile operations or change their
1486order of execution relative to other volatile operations. The optimizers
1487<i>may</i> change the order of volatile operations relative to non-volatile
1488operations. This is not Java's "volatile" and has no cross-thread
1489synchronization behavior.</p>
1490
1491</div>
1492
Eli Friedman35b54aa2011-07-20 21:35:53 +0000[diff] [blame]1493<!-- ======================================================================= -->
1494<h3>
1495 <a name="memmodel">Memory Model for Concurrent Operations</a>
1496</h3>
1497
1498<div>
1499
1500<p>The LLVM IR does not define any way to start parallel threads of execution
1501or to register signal handlers. Nonetheless, there are platform-specific
1502ways to create them, and we define LLVM IR's behavior in their presence. This
1503model is inspired by the C++0x memory model.</p>
1504
Eli Friedman95f69a42011-08-22 21:35:27 +0000[diff] [blame]1505<p>For a more informal introduction to this model, see the
1506<a href="Atomics.html">LLVM Atomic Instructions and Concurrency Guide</a>.
1507
Eli Friedman35b54aa2011-07-20 21:35:53 +0000[diff] [blame]1508<p>We define a <i>happens-before</i> partial order as the least partial order
1509that</p>
1510<ul>
1511 <li>Is a superset of single-thread program order, and</li>
1512 <li>When a <i>synchronizes-with</i> <tt>b</tt>, includes an edge from
1513 <tt>a</tt> to <tt>b</tt>. <i>Synchronizes-with</i> pairs are introduced
1514 by platform-specific techniques, like pthread locks, thread
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]1515 creation, thread joining, etc., and by atomic instructions.
1516 (See also <a href="#ordering">Atomic Memory Ordering Constraints</a>).
1517 </li>
Eli Friedman35b54aa2011-07-20 21:35:53 +0000[diff] [blame]1518</ul>
1519
1520<p>Note that program order does not introduce <i>happens-before</i> edges
1521between a thread and signals executing inside that thread.</p>
1522
1523<p>Every (defined) read operation (load instructions, memcpy, atomic
1524loads/read-modify-writes, etc.) <var>R</var> reads a series of bytes written by
1525(defined) write operations (store instructions, atomic
Eli Friedmanf12e4e92011-07-22 03:04:45 +0000[diff] [blame]1526stores/read-modify-writes, memcpy, etc.). For the purposes of this section,
1527initialized globals are considered to have a write of the initializer which is
1528atomic and happens before any other read or write of the memory in question.
1529For each byte of a read <var>R</var>, <var>R<sub>byte</sub></var> may see
1530any write to the same byte, except:</p>
Eli Friedman35b54aa2011-07-20 21:35:53 +0000[diff] [blame]1531
1532<ul>
1533 <li>If <var>write<sub>1</sub></var> happens before
1534 <var>write<sub>2</sub></var>, and <var>write<sub>2</sub></var> happens
1535 before <var>R<sub>byte</sub></var>, then <var>R<sub>byte</sub></var>
Eli Friedmanf12e4e92011-07-22 03:04:45 +0000[diff] [blame]1536 does not see <var>write<sub>1</sub></var>.
Bill Wendling537603b2011-07-31 06:45:03 +0000[diff] [blame]1537 <li>If <var>R<sub>byte</sub></var> happens before
1538 <var>write<sub>3</sub></var>, then <var>R<sub>byte</sub></var> does not
1539 see <var>write<sub>3</sub></var>.
Eli Friedman35b54aa2011-07-20 21:35:53 +0000[diff] [blame]1540</ul>
1541
1542<p>Given that definition, <var>R<sub>byte</sub></var> is defined as follows:
1543<ul>
Eli Friedman95f69a42011-08-22 21:35:27 +0000[diff] [blame]1544 <li>If <var>R</var> is volatile, the result is target-dependent. (Volatile
1545 is supposed to give guarantees which can support
1546 <code>sig_atomic_t</code> in C/C++, and may be used for accesses to
1547 addresses which do not behave like normal memory. It does not generally
1548 provide cross-thread synchronization.)
1549 <li>Otherwise, if there is no write to the same byte that happens before
Eli Friedman35b54aa2011-07-20 21:35:53 +0000[diff] [blame]1550 <var>R<sub>byte</sub></var>, <var>R<sub>byte</sub></var> returns
1551 <tt>undef</tt> for that byte.
Eli Friedmanf12e4e92011-07-22 03:04:45 +0000[diff] [blame]1552 <li>Otherwise, if <var>R<sub>byte</sub></var> may see exactly one write,
Eli Friedman35b54aa2011-07-20 21:35:53 +0000[diff] [blame]1553 <var>R<sub>byte</sub></var> returns the value written by that
1554 write.</li>
Eli Friedmanf12e4e92011-07-22 03:04:45 +0000[diff] [blame]1555 <li>Otherwise, if <var>R</var> is atomic, and all the writes
1556 <var>R<sub>byte</sub></var> may see are atomic, it chooses one of the
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]1557 values written. See the <a href="#ordering">Atomic Memory Ordering
1558 Constraints</a> section for additional constraints on how the choice
1559 is made.
Eli Friedman35b54aa2011-07-20 21:35:53 +0000[diff] [blame]1560 <li>Otherwise <var>R<sub>byte</sub></var> returns <tt>undef</tt>.</li>
1561</ul>
1562
1563<p><var>R</var> returns the value composed of the series of bytes it read.
1564This implies that some bytes within the value may be <tt>undef</tt>
1565<b>without</b> the entire value being <tt>undef</tt>. Note that this only
1566defines the semantics of the operation; it doesn't mean that targets will
1567emit more than one instruction to read the series of bytes.</p>
1568
1569<p>Note that in cases where none of the atomic intrinsics are used, this model
1570places only one restriction on IR transformations on top of what is required
1571for single-threaded execution: introducing a store to a byte which might not
Eli Friedman4bc9f3c2011-08-02 01:15:34 +0000[diff] [blame]1572otherwise be stored is not allowed in general. (Specifically, in the case
1573where another thread might write to and read from an address, introducing a
1574store can change a load that may see exactly one write into a load that may
1575see multiple writes.)</p>
Eli Friedman35b54aa2011-07-20 21:35:53 +0000[diff] [blame]1576
1577<!-- FIXME: This model assumes all targets where concurrency is relevant have
1578a byte-size store which doesn't affect adjacent bytes. As far as I can tell,
1579none of the backends currently in the tree fall into this category; however,
1580there might be targets which care. If there are, we want a paragraph
1581like the following:
1582
1583Targets may specify that stores narrower than a certain width are not
1584available; on such a target, for the purposes of this model, treat any
1585non-atomic write with an alignment or width less than the minimum width
1586as if it writes to the relevant surrounding bytes.
1587-->
1588
1589</div>
1590
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]1591<!-- ======================================================================= -->
NAKAMURA Takumi0300d882011-08-12 06:17:17 +0000[diff] [blame]1592<h3>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]1593 <a name="ordering">Atomic Memory Ordering Constraints</a>
NAKAMURA Takumi0300d882011-08-12 06:17:17 +0000[diff] [blame]1594</h3>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]1595
NAKAMURA Takumi0300d882011-08-12 06:17:17 +0000[diff] [blame]1596<div>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]1597
1598<p>Atomic instructions (<a href="#i_cmpxchg"><code>cmpxchg</code></a>,
Eli Friedman59b66882011-08-09 23:02:53 +0000[diff] [blame]1599<a href="#i_atomicrmw"><code>atomicrmw</code></a>,
1600<a href="#i_fence"><code>fence</code></a>,
1601<a href="#i_load"><code>atomic load</code></a>, and
Eli Friedman75362532011-08-09 23:26:12 +0000[diff] [blame]1602<a href="#i_store"><code>atomic store</code></a>) take an ordering parameter
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]1603that determines which other atomic instructions on the same address they
1604<i>synchronize with</i>. These semantics are borrowed from Java and C++0x,
1605but are somewhat more colloquial. If these descriptions aren't precise enough,
Eli Friedman95f69a42011-08-22 21:35:27 +0000[diff] [blame]1606check those specs (see spec references in the
1607<a href="Atomic.html#introduction">atomics guide</a>).
1608<a href="#i_fence"><code>fence</code></a> instructions
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]1609treat these orderings somewhat differently since they don't take an address.
1610See that instruction's documentation for details.</p>
1611
Eli Friedman95f69a42011-08-22 21:35:27 +0000[diff] [blame]1612<p>For a simpler introduction to the ordering constraints, see the
1613<a href="Atomics.html">LLVM Atomic Instructions and Concurrency Guide</a>.</p>
1614
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]1615<dl>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]1616<dt><code>unordered</code></dt>
1617<dd>The set of values that can be read is governed by the happens-before
1618partial order. A value cannot be read unless some operation wrote it.
1619This is intended to provide a guarantee strong enough to model Java's
1620non-volatile shared variables. This ordering cannot be specified for
1621read-modify-write operations; it is not strong enough to make them atomic
1622in any interesting way.</dd>
1623<dt><code>monotonic</code></dt>
1624<dd>In addition to the guarantees of <code>unordered</code>, there is a single
1625total order for modifications by <code>monotonic</code> operations on each
1626address. All modification orders must be compatible with the happens-before
1627order. There is no guarantee that the modification orders can be combined to
1628a global total order for the whole program (and this often will not be
1629possible). The read in an atomic read-modify-write operation
1630(<a href="#i_cmpxchg"><code>cmpxchg</code></a> and
1631<a href="#i_atomicrmw"><code>atomicrmw</code></a>)
1632reads the value in the modification order immediately before the value it
1633writes. If one atomic read happens before another atomic read of the same
1634address, the later read must see the same value or a later value in the
1635address's modification order. This disallows reordering of
1636<code>monotonic</code> (or stronger) operations on the same address. If an
1637address is written <code>monotonic</code>ally by one thread, and other threads
1638<code>monotonic</code>ally read that address repeatedly, the other threads must
Eli Friedman95f69a42011-08-22 21:35:27 +0000[diff] [blame]1639eventually see the write. This corresponds to the C++0x/C1x
1640<code>memory_order_relaxed</code>.</dd>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]1641<dt><code>acquire</code></dt>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]1642<dd>In addition to the guarantees of <code>monotonic</code>,
Eli Friedman0cb3b562011-08-24 20:28:39 +0000[diff] [blame]1643a <i>synchronizes-with</i> edge may be formed with a <code>release</code>
1644operation. This is intended to model C++'s <code>memory_order_acquire</code>.</dd>
1645<dt><code>release</code></dt>
1646<dd>In addition to the guarantees of <code>monotonic</code>, if this operation
1647writes a value which is subsequently read by an <code>acquire</code> operation,
1648it <i>synchronizes-with</i> that operation. (This isn't a complete
1649description; see the C++0x definition of a release sequence.) This corresponds
1650to the C++0x/C1x <code>memory_order_release</code>.</dd>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]1651<dt><code>acq_rel</code> (acquire+release)</dt><dd>Acts as both an
Eli Friedman95f69a42011-08-22 21:35:27 +0000[diff] [blame]1652<code>acquire</code> and <code>release</code> operation on its address.
1653This corresponds to the C++0x/C1x <code>memory_order_acq_rel</code>.</dd>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]1654<dt><code>seq_cst</code> (sequentially consistent)</dt><dd>
1655<dd>In addition to the guarantees of <code>acq_rel</code>
1656(<code>acquire</code> for an operation which only reads, <code>release</code>
1657for an operation which only writes), there is a global total order on all
1658sequentially-consistent operations on all addresses, which is consistent with
1659the <i>happens-before</i> partial order and with the modification orders of
1660all the affected addresses. Each sequentially-consistent read sees the last
Eli Friedman95f69a42011-08-22 21:35:27 +0000[diff] [blame]1661preceding write to the same address in this global order. This corresponds
1662to the C++0x/C1x <code>memory_order_seq_cst</code> and Java volatile.</dd>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]1663</dl>
1664
1665<p id="singlethread">If an atomic operation is marked <code>singlethread</code>,
1666it only <i>synchronizes with</i> or participates in modification and seq_cst
1667total orderings with other operations running in the same thread (for example,
1668in signal handlers).</p>
1669
1670</div>
1671
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1672</div>
1673
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]1674<!-- *********************************************************************** -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1675<h2><a name="typesystem">Type System</a></h2>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]1676<!-- *********************************************************************** -->
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]1677
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1678<div>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]1679
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]1680<p>The LLVM type system is one of the most important features of the
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1681 intermediate representation. Being typed enables a number of optimizations
1682 to be performed on the intermediate representation directly, without having
1683 to do extra analyses on the side before the transformation. A strong type
1684 system makes it easier to read the generated code and enables novel analyses
1685 and transformations that are not feasible to perform on normal three address
1686 code representations.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000[diff] [blame]1687
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]1688<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1689<h3>
1690 <a name="t_classifications">Type Classifications</a>
1691</h3>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1692
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1693<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1694
1695<p>The types fall into a few useful classifications:</p>
Misha Brukmanc501f552004-03-01 17:47:27 +0000[diff] [blame]1696
1697<table border="1" cellspacing="0" cellpadding="4">
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]1698 <tbody>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]1699 <tr><th>Classification</th><th>Types</th></tr>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]1700 <tr>
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1701 <td><a href="#t_integer">integer</a></td>
Reid Spencer138249b2007-05-16 18:44:01 +0000[diff] [blame]1702 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]1703 </tr>
1704 <tr>
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1705 <td><a href="#t_floating">floating point</a></td>
1706 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]1707 </tr>
1708 <tr>
1709 <td><a name="t_firstclass">first class</a></td>
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1710 <td><a href="#t_integer">integer</a>,
1711 <a href="#t_floating">floating point</a>,
1712 <a href="#t_pointer">pointer</a>,
Dan Gohman08783a882008-06-18 18:42:13 +0000[diff] [blame]1713 <a href="#t_vector">vector</a>,
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]1714 <a href="#t_struct">structure</a>,
1715 <a href="#t_array">array</a>,
Nick Lewyckyadbc2842009-05-30 05:06:04 +0000[diff] [blame]1716 <a href="#t_label">label</a>,
1717 <a href="#t_metadata">metadata</a>.
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]1718 </td>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]1719 </tr>
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1720 <tr>
1721 <td><a href="#t_primitive">primitive</a></td>
1722 <td><a href="#t_label">label</a>,
1723 <a href="#t_void">void</a>,
Tobias Grosser4c8c95b2010-12-28 20:29:31 +0000[diff] [blame]1724 <a href="#t_integer">integer</a>,
Nick Lewyckyadbc2842009-05-30 05:06:04 +0000[diff] [blame]1725 <a href="#t_floating">floating point</a>,
Dale Johannesen33e5c352010-10-01 00:48:59 +0000[diff] [blame]1726 <a href="#t_x86mmx">x86mmx</a>,
Nick Lewyckyadbc2842009-05-30 05:06:04 +0000[diff] [blame]1727 <a href="#t_metadata">metadata</a>.</td>
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1728 </tr>
1729 <tr>
1730 <td><a href="#t_derived">derived</a></td>
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]1731 <td><a href="#t_array">array</a>,
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1732 <a href="#t_function">function</a>,
1733 <a href="#t_pointer">pointer</a>,
1734 <a href="#t_struct">structure</a>,
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1735 <a href="#t_vector">vector</a>,
1736 <a href="#t_opaque">opaque</a>.
Dan Gohman93bf60d2008-10-14 16:32:04 +0000[diff] [blame]1737 </td>
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1738 </tr>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]1739 </tbody>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]1740</table>
Misha Brukmanc501f552004-03-01 17:47:27 +0000[diff] [blame]1741
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1742<p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1743 important. Values of these types are the only ones which can be produced by
Nick Lewycky84a1eeb2009-09-27 00:45:11 +0000[diff] [blame]1744 instructions.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1745
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]1746</div>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]1747
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]1748<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1749<h3>
1750 <a name="t_primitive">Primitive Types</a>
1751</h3>
Chris Lattner43542b32008-01-04 04:34:14 +0000[diff] [blame]1752
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1753<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1754
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1755<p>The primitive types are the fundamental building blocks of the LLVM
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1756 system.</p>
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1757
1758<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1759<h4>
1760 <a name="t_integer">Integer Type</a>
1761</h4>
Nick Lewycky84a1eeb2009-09-27 00:45:11 +0000[diff] [blame]1762
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1763<div>
Nick Lewycky84a1eeb2009-09-27 00:45:11 +0000[diff] [blame]1764
1765<h5>Overview:</h5>
1766<p>The integer type is a very simple type that simply specifies an arbitrary
1767 bit width for the integer type desired. Any bit width from 1 bit to
1768 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
1769
1770<h5>Syntax:</h5>
1771<pre>
1772 iN
1773</pre>
1774
1775<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1776 value.</p>
1777
1778<h5>Examples:</h5>
1779<table class="layout">
1780 <tr class="layout">
1781 <td class="left"><tt>i1</tt></td>
1782 <td class="left">a single-bit integer.</td>
1783 </tr>
1784 <tr class="layout">
1785 <td class="left"><tt>i32</tt></td>
1786 <td class="left">a 32-bit integer.</td>
1787 </tr>
1788 <tr class="layout">
1789 <td class="left"><tt>i1942652</tt></td>
1790 <td class="left">a really big integer of over 1 million bits.</td>
1791 </tr>
1792</table>
1793
Nick Lewycky84a1eeb2009-09-27 00:45:11 +0000[diff] [blame]1794</div>
1795
1796<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1797<h4>
1798 <a name="t_floating">Floating Point Types</a>
1799</h4>
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1800
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1801<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1802
1803<table>
1804 <tbody>
1805 <tr><th>Type</th><th>Description</th></tr>
1806 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1807 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1808 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1809 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1810 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1811 </tbody>
1812</table>
1813
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1814</div>
1815
1816<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1817<h4>
1818 <a name="t_x86mmx">X86mmx Type</a>
1819</h4>
Dale Johannesen33e5c352010-10-01 00:48:59 +0000[diff] [blame]1820
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1821<div>
Dale Johannesen33e5c352010-10-01 00:48:59 +0000[diff] [blame]1822
1823<h5>Overview:</h5>
1824<p>The x86mmx type represents a value held in an MMX register on an x86 machine. The operations allowed on it are quite limited: parameters and return values, load and store, and bitcast. User-specified MMX instructions are represented as intrinsic or asm calls with arguments and/or results of this type. There are no arrays, vectors or constants of this type.</p>
1825
1826<h5>Syntax:</h5>
1827<pre>
Dale Johannesenb1f0ff12010-10-01 01:07:02 +0000[diff] [blame]1828 x86mmx
Dale Johannesen33e5c352010-10-01 00:48:59 +0000[diff] [blame]1829</pre>
1830
1831</div>
1832
1833<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1834<h4>
1835 <a name="t_void">Void Type</a>
1836</h4>
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1837
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1838<div>
Bill Wendling30235112009-07-20 02:39:26 +0000[diff] [blame]1839
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1840<h5>Overview:</h5>
1841<p>The void type does not represent any value and has no size.</p>
1842
1843<h5>Syntax:</h5>
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1844<pre>
1845 void
1846</pre>
Bill Wendling30235112009-07-20 02:39:26 +0000[diff] [blame]1847
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1848</div>
1849
1850<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1851<h4>
1852 <a name="t_label">Label Type</a>
1853</h4>
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1854
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1855<div>
Bill Wendling30235112009-07-20 02:39:26 +0000[diff] [blame]1856
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1857<h5>Overview:</h5>
1858<p>The label type represents code labels.</p>
1859
1860<h5>Syntax:</h5>
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1861<pre>
1862 label
1863</pre>
Bill Wendling30235112009-07-20 02:39:26 +0000[diff] [blame]1864
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1865</div>
1866
Nick Lewyckyadbc2842009-05-30 05:06:04 +0000[diff] [blame]1867<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1868<h4>
1869 <a name="t_metadata">Metadata Type</a>
1870</h4>
Nick Lewyckyadbc2842009-05-30 05:06:04 +0000[diff] [blame]1871
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1872<div>
Bill Wendling30235112009-07-20 02:39:26 +0000[diff] [blame]1873
Nick Lewyckyadbc2842009-05-30 05:06:04 +0000[diff] [blame]1874<h5>Overview:</h5>
Nick Lewycky93e06a52009-09-27 23:27:42 +0000[diff] [blame]1875<p>The metadata type represents embedded metadata. No derived types may be
1876 created from metadata except for <a href="#t_function">function</a>
1877 arguments.
Nick Lewyckyadbc2842009-05-30 05:06:04 +0000[diff] [blame]1878
1879<h5>Syntax:</h5>
Nick Lewyckyadbc2842009-05-30 05:06:04 +0000[diff] [blame]1880<pre>
1881 metadata
1882</pre>
Bill Wendling30235112009-07-20 02:39:26 +0000[diff] [blame]1883
Nick Lewyckyadbc2842009-05-30 05:06:04 +0000[diff] [blame]1884</div>
1885
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1886</div>
Chris Lattner7824d182008-01-04 04:32:38 +0000[diff] [blame]1887
1888<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1889<h3>
1890 <a name="t_derived">Derived Types</a>
1891</h3>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]1892
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1893<div>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]1894
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1895<p>The real power in LLVM comes from the derived types in the system. This is
1896 what allows a programmer to represent arrays, functions, pointers, and other
Nick Lewycky84a1eeb2009-09-27 00:45:11 +0000[diff] [blame]1897 useful types. Each of these types contain one or more element types which
1898 may be a primitive type, or another derived type. For example, it is
1899 possible to have a two dimensional array, using an array as the element type
1900 of another array.</p>
Dan Gohman142ccc02009-01-24 15:58:40 +0000[diff] [blame]1901
Chris Lattnerb1ed91f2011-07-09 17:41:24 +0000[diff] [blame]1902</div>
1903
1904
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]1905<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1906<h4>
1907 <a name="t_aggregate">Aggregate Types</a>
1908</h4>
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]1909
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1910<div>
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]1911
1912<p>Aggregate Types are a subset of derived types that can contain multiple
1913 member types. <a href="#t_array">Arrays</a>,
Chris Lattner13ee7952010-08-28 04:09:24 +0000[diff] [blame]1914 <a href="#t_struct">structs</a>, and <a href="#t_vector">vectors</a> are
1915 aggregate types.</p>
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]1916
1917</div>
1918
Reid Spencer138249b2007-05-16 18:44:01 +0000[diff] [blame]1919<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1920<h4>
1921 <a name="t_array">Array Type</a>
1922</h4>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]1923
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1924<div>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]1925
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]1926<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]1927<p>The array type is a very simple derived type that arranges elements
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1928 sequentially in memory. The array type requires a size (number of elements)
1929 and an underlying data type.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]1930
Chris Lattner590645f2002-04-14 06:13:44 +0000[diff] [blame]1931<h5>Syntax:</h5>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]1932<pre>
1933 [&lt;# elements&gt; x &lt;elementtype&gt;]
1934</pre>
1935
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1936<p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1937 be any type with a size.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]1938
Chris Lattner590645f2002-04-14 06:13:44 +0000[diff] [blame]1939<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]1940<table class="layout">
1941 <tr class="layout">
Chris Lattner747359f2007-12-19 05:04:11 +0000[diff] [blame]1942 <td class="left"><tt>[40 x i32]</tt></td>
1943 <td class="left">Array of 40 32-bit integer values.</td>
1944 </tr>
1945 <tr class="layout">
1946 <td class="left"><tt>[41 x i32]</tt></td>
1947 <td class="left">Array of 41 32-bit integer values.</td>
1948 </tr>
1949 <tr class="layout">
1950 <td class="left"><tt>[4 x i8]</tt></td>
1951 <td class="left">Array of 4 8-bit integer values.</td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]1952 </tr>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]1953</table>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]1954<p>Here are some examples of multidimensional arrays:</p>
1955<table class="layout">
1956 <tr class="layout">
Chris Lattner747359f2007-12-19 05:04:11 +0000[diff] [blame]1957 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1958 <td class="left">3x4 array of 32-bit integer values.</td>
1959 </tr>
1960 <tr class="layout">
1961 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1962 <td class="left">12x10 array of single precision floating point values.</td>
1963 </tr>
1964 <tr class="layout">
1965 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1966 <td class="left">2x3x4 array of 16-bit integer values.</td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]1967 </tr>
1968</table>
Chris Lattnerc0ad71e2005-06-24 17:22:57 +0000[diff] [blame]1969
Dan Gohmanc74bc282009-11-09 19:01:53 +0000[diff] [blame]1970<p>There is no restriction on indexing beyond the end of the array implied by
1971 a static type (though there are restrictions on indexing beyond the bounds
1972 of an allocated object in some cases). This means that single-dimension
1973 'variable sized array' addressing can be implemented in LLVM with a zero
1974 length array type. An implementation of 'pascal style arrays' in LLVM could
1975 use the type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
Chris Lattnerc0ad71e2005-06-24 17:22:57 +0000[diff] [blame]1976
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]1977</div>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]1978
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]1979<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]1980<h4>
1981 <a name="t_function">Function Type</a>
1982</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1983
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]1984<div>
Chris Lattnerda508ac2008-04-23 04:59:35 +0000[diff] [blame]1985
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]1986<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1987<p>The function type can be thought of as a function signature. It consists of
1988 a return type and a list of formal parameter types. The return type of a
Chris Lattner13ee7952010-08-28 04:09:24 +0000[diff] [blame]1989 function type is a first class type or a void type.</p>
Devang Pateld6cff512008-03-10 20:49:15 +0000[diff] [blame]1990
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]1991<h5>Syntax:</h5>
Chris Lattnerda508ac2008-04-23 04:59:35 +0000[diff] [blame]1992<pre>
Nick Lewycky14d1ccc2009-09-27 07:55:32 +0000[diff] [blame]1993 &lt;returntype&gt; (&lt;parameter list&gt;)
Chris Lattnerda508ac2008-04-23 04:59:35 +0000[diff] [blame]1994</pre>
1995
John Criswell4c0cf7f2005-10-24 16:17:18 +0000[diff] [blame]1996<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]1997 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1998 which indicates that the function takes a variable number of arguments.
1999 Variable argument functions can access their arguments with
2000 the <a href="#int_varargs">variable argument handling intrinsic</a>
Chris Lattner47f2a832010-03-02 06:36:51 +0000[diff] [blame]2001 functions. '<tt>&lt;returntype&gt;</tt>' is any type except
Nick Lewycky93e06a52009-09-27 23:27:42 +0000[diff] [blame]2002 <a href="#t_label">label</a>.</p>
Chris Lattnerda508ac2008-04-23 04:59:35 +0000[diff] [blame]2003
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]2004<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]2005<table class="layout">
2006 <tr class="layout">
Reid Spencer58c08712006-12-31 07:18:34 +0000[diff] [blame]2007 <td class="left"><tt>i32 (i32)</tt></td>
2008 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]2009 </td>
Reid Spencer58c08712006-12-31 07:18:34 +0000[diff] [blame]2010 </tr><tr class="layout">
Chris Lattner47f2a832010-03-02 06:36:51 +0000[diff] [blame]2011 <td class="left"><tt>float&nbsp;(i16,&nbsp;i32&nbsp;*)&nbsp;*
Reid Spencer655dcc62006-12-31 07:20:23 +0000[diff] [blame]2012 </tt></td>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]2013 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
Chris Lattner47f2a832010-03-02 06:36:51 +0000[diff] [blame]2014 an <tt>i16</tt> and a <a href="#t_pointer">pointer</a> to <tt>i32</tt>,
2015 returning <tt>float</tt>.
Reid Spencer58c08712006-12-31 07:18:34 +0000[diff] [blame]2016 </td>
2017 </tr><tr class="layout">
2018 <td class="left"><tt>i32 (i8*, ...)</tt></td>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]2019 <td class="left">A vararg function that takes at least one
2020 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
2021 which returns an integer. This is the signature for <tt>printf</tt> in
Reid Spencer58c08712006-12-31 07:18:34 +0000[diff] [blame]2022 LLVM.
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]2023 </td>
Devang Patele3dfc1c2008-03-24 05:35:41 +0000[diff] [blame]2024 </tr><tr class="layout">
2025 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Nick Lewycky14d1ccc2009-09-27 07:55:32 +0000[diff] [blame]2026 <td class="left">A function taking an <tt>i32</tt>, returning a
2027 <a href="#t_struct">structure</a> containing two <tt>i32</tt> values
Devang Patele3dfc1c2008-03-24 05:35:41 +0000[diff] [blame]2028 </td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]2029 </tr>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]2030</table>
Misha Brukmanc501f552004-03-01 17:47:27 +0000[diff] [blame]2031
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]2032</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2033
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]2034<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2035<h4>
2036 <a name="t_struct">Structure Type</a>
2037</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2038
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2039<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2040
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]2041<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2042<p>The structure type is used to represent a collection of data members together
Chris Lattnerb1ed91f2011-07-09 17:41:24 +0000[diff] [blame]2043 in memory. The elements of a structure may be any type that has a size.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2044
Jeffrey Yasskinf991bbb2010-01-11 19:19:26 +0000[diff] [blame]2045<p>Structures in memory are accessed using '<tt><a href="#i_load">load</a></tt>'
2046 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field
2047 with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
2048 Structures in registers are accessed using the
2049 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' and
2050 '<tt><a href="#i_insertvalue">insertvalue</a></tt>' instructions.</p>
Chris Lattnerb1ed91f2011-07-09 17:41:24 +0000[diff] [blame]2051
2052<p>Structures may optionally be "packed" structures, which indicate that the
2053 alignment of the struct is one byte, and that there is no padding between
Chris Lattner190552d2011-08-12 17:31:02 +0000[diff] [blame]2054 the elements. In non-packed structs, padding between field types is inserted
2055 as defined by the TargetData string in the module, which is required to match
2056 what the underlying processor expects.</p>
Chris Lattnerb1ed91f2011-07-09 17:41:24 +0000[diff] [blame]2057
Chris Lattner190552d2011-08-12 17:31:02 +0000[diff] [blame]2058<p>Structures can either be "literal" or "identified". A literal structure is
2059 defined inline with other types (e.g. <tt>{i32, i32}*</tt>) whereas identified
2060 types are always defined at the top level with a name. Literal types are
2061 uniqued by their contents and can never be recursive or opaque since there is
Chris Lattner32531732011-08-12 18:12:40 +0000[diff] [blame]2062 no way to write one. Identified types can be recursive, can be opaqued, and are
Chris Lattner190552d2011-08-12 17:31:02 +0000[diff] [blame]2063 never uniqued.
Chris Lattnerb1ed91f2011-07-09 17:41:24 +0000[diff] [blame]2064</p>
2065
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]2066<h5>Syntax:</h5>
Bill Wendling30235112009-07-20 02:39:26 +0000[diff] [blame]2067<pre>
Chris Lattner190552d2011-08-12 17:31:02 +0000[diff] [blame]2068 %T1 = type { &lt;type list&gt; } <i>; Identified normal struct type</i>
2069 %T2 = type &lt;{ &lt;type list&gt; }&gt; <i>; Identified packed struct type</i>
Bill Wendling30235112009-07-20 02:39:26 +0000[diff] [blame]2070</pre>
Chris Lattnerb1ed91f2011-07-09 17:41:24 +0000[diff] [blame]2071
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]2072<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]2073<table class="layout">
2074 <tr class="layout">
Jeff Cohen5819f182007-04-22 01:17:39 +0000[diff] [blame]2075 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
2076 <td class="left">A triple of three <tt>i32</tt> values</td>
Chris Lattnerb1ed91f2011-07-09 17:41:24 +0000[diff] [blame]2077 </tr>
2078 <tr class="layout">
Jeff Cohen5819f182007-04-22 01:17:39 +0000[diff] [blame]2079 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
2080 <td class="left">A pair, where the first element is a <tt>float</tt> and the
2081 second element is a <a href="#t_pointer">pointer</a> to a
2082 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
2083 an <tt>i32</tt>.</td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]2084 </tr>
Chris Lattnerb1ed91f2011-07-09 17:41:24 +0000[diff] [blame]2085 <tr class="layout">
2086 <td class="left"><tt>&lt;{ i8, i32 }&gt;</tt></td>
2087 <td class="left">A packed struct known to be 5 bytes in size.</td>
2088 </tr>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]2089</table>
Dan Gohman142ccc02009-01-24 15:58:40 +0000[diff] [blame]2090
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]2091</div>
Chris Lattnerb1ed91f2011-07-09 17:41:24 +0000[diff] [blame]2092
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]2093<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2094<h4>
Chris Lattner2a843822011-07-23 19:59:08 +0000[diff] [blame]2095 <a name="t_opaque">Opaque Structure Types</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2096</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2097
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2098<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2099
Andrew Lenharth8df88e22006-12-08 17:13:00 +0000[diff] [blame]2100<h5>Overview:</h5>
Chris Lattner2a843822011-07-23 19:59:08 +0000[diff] [blame]2101<p>Opaque structure types are used to represent named structure types that do
2102 not have a body specified. This corresponds (for example) to the C notion of
2103 a forward declared structure.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2104
Andrew Lenharth8df88e22006-12-08 17:13:00 +0000[diff] [blame]2105<h5>Syntax:</h5>
Bill Wendling30235112009-07-20 02:39:26 +0000[diff] [blame]2106<pre>
Chris Lattnerb1ed91f2011-07-09 17:41:24 +0000[diff] [blame]2107 %X = type opaque
2108 %52 = type opaque
Bill Wendling30235112009-07-20 02:39:26 +0000[diff] [blame]2109</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2110
Andrew Lenharth8df88e22006-12-08 17:13:00 +0000[diff] [blame]2111<h5>Examples:</h5>
2112<table class="layout">
2113 <tr class="layout">
Chris Lattnerb1ed91f2011-07-09 17:41:24 +0000[diff] [blame]2114 <td class="left"><tt>opaque</tt></td>
2115 <td class="left">An opaque type.</td>
Andrew Lenharth8df88e22006-12-08 17:13:00 +0000[diff] [blame]2116 </tr>
2117</table>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2118
Andrew Lenharth8df88e22006-12-08 17:13:00 +0000[diff] [blame]2119</div>
2120
Chris Lattnerb1ed91f2011-07-09 17:41:24 +0000[diff] [blame]2121
2122
Andrew Lenharth8df88e22006-12-08 17:13:00 +0000[diff] [blame]2123<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2124<h4>
2125 <a name="t_pointer">Pointer Type</a>
2126</h4>
Chris Lattner4a67c912009-02-08 19:53:29 +0000[diff] [blame]2127
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2128<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2129
2130<h5>Overview:</h5>
Dan Gohman88481112010-02-25 16:50:07 +0000[diff] [blame]2131<p>The pointer type is used to specify memory locations.
2132 Pointers are commonly used to reference objects in memory.</p>
2133
2134<p>Pointer types may have an optional address space attribute defining the
2135 numbered address space where the pointed-to object resides. The default
2136 address space is number zero. The semantics of non-zero address
2137 spaces are target-specific.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2138
2139<p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
2140 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
Chris Lattner4a67c912009-02-08 19:53:29 +0000[diff] [blame]2141
Chris Lattner590645f2002-04-14 06:13:44 +0000[diff] [blame]2142<h5>Syntax:</h5>
Bill Wendling30235112009-07-20 02:39:26 +0000[diff] [blame]2143<pre>
2144 &lt;type&gt; *
2145</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2146
Chris Lattner590645f2002-04-14 06:13:44 +0000[diff] [blame]2147<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]2148<table class="layout">
2149 <tr class="layout">
Dan Gohman623806e2009-01-04 23:44:43 +0000[diff] [blame]2150 <td class="left"><tt>[4 x i32]*</tt></td>
Chris Lattner747359f2007-12-19 05:04:11 +0000[diff] [blame]2151 <td class="left">A <a href="#t_pointer">pointer</a> to <a
2152 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
2153 </tr>
2154 <tr class="layout">
Dan Gohmanaabfdb32010-05-28 17:13:49 +0000[diff] [blame]2155 <td class="left"><tt>i32 (i32*) *</tt></td>
Chris Lattner747359f2007-12-19 05:04:11 +0000[diff] [blame]2156 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]2157 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner747359f2007-12-19 05:04:11 +0000[diff] [blame]2158 <tt>i32</tt>.</td>
2159 </tr>
2160 <tr class="layout">
2161 <td class="left"><tt>i32 addrspace(5)*</tt></td>
2162 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
2163 that resides in address space #5.</td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]2164 </tr>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]2165</table>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2166
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]2167</div>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]2168
Chris Lattnerc8cb6952004-08-12 19:12:28 +0000[diff] [blame]2169<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2170<h4>
2171 <a name="t_vector">Vector Type</a>
2172</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2173
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2174<div>
Chris Lattner37b6b092005-04-25 17:34:15 +0000[diff] [blame]2175
Chris Lattnerc8cb6952004-08-12 19:12:28 +0000[diff] [blame]2176<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2177<p>A vector type is a simple derived type that represents a vector of elements.
2178 Vector types are used when multiple primitive data are operated in parallel
2179 using a single instruction (SIMD). A vector type requires a size (number of
Duncan Sands31c0e0e2009-11-27 13:38:03 +0000[diff] [blame]2180 elements) and an underlying primitive data type. Vector types are considered
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2181 <a href="#t_firstclass">first class</a>.</p>
Chris Lattner37b6b092005-04-25 17:34:15 +0000[diff] [blame]2182
Chris Lattnerc8cb6952004-08-12 19:12:28 +0000[diff] [blame]2183<h5>Syntax:</h5>
Chris Lattner37b6b092005-04-25 17:34:15 +0000[diff] [blame]2184<pre>
2185 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
2186</pre>
2187
Chris Lattnerf11031a2010-10-10 18:20:35 +0000[diff] [blame]2188<p>The number of elements is a constant integer value larger than 0; elementtype
2189 may be any integer or floating point type. Vectors of size zero are not
2190 allowed, and pointers are not allowed as the element type.</p>
Chris Lattner37b6b092005-04-25 17:34:15 +0000[diff] [blame]2191
Chris Lattnerc8cb6952004-08-12 19:12:28 +0000[diff] [blame]2192<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]2193<table class="layout">
2194 <tr class="layout">
Chris Lattner747359f2007-12-19 05:04:11 +0000[diff] [blame]2195 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
2196 <td class="left">Vector of 4 32-bit integer values.</td>
2197 </tr>
2198 <tr class="layout">
2199 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
2200 <td class="left">Vector of 8 32-bit floating-point values.</td>
2201 </tr>
2202 <tr class="layout">
2203 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
2204 <td class="left">Vector of 2 64-bit integer values.</td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +0000[diff] [blame]2205 </tr>
2206</table>
Dan Gohman142ccc02009-01-24 15:58:40 +0000[diff] [blame]2207
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]2208</div>
2209
Bill Wendlingae8b5ea2011-07-31 06:47:33 +0000[diff] [blame]2210</div>
2211
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2212<!-- *********************************************************************** -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2213<h2><a name="constants">Constants</a></h2>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2214<!-- *********************************************************************** -->
2215
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2216<div>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2217
2218<p>LLVM has several different basic types of constants. This section describes
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2219 them all and their syntax.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2220
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2221<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2222<h3>
2223 <a name="simpleconstants">Simple Constants</a>
2224</h3>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2225
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2226<div>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2227
2228<dl>
2229 <dt><b>Boolean constants</b></dt>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2230 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
Nick Lewycky84a1eeb2009-09-27 00:45:11 +0000[diff] [blame]2231 constants of the <tt><a href="#t_integer">i1</a></tt> type.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2232
2233 <dt><b>Integer constants</b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2234 <dd>Standard integers (such as '4') are constants of
2235 the <a href="#t_integer">integer</a> type. Negative numbers may be used
2236 with integer types.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2237
2238 <dt><b>Floating point constants</b></dt>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2239 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2240 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
2241 notation (see below). The assembler requires the exact decimal value of a
2242 floating-point constant. For example, the assembler accepts 1.25 but
2243 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
2244 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2245
2246 <dt><b>Null pointer constants</b></dt>
John Criswelldfe6a862004-12-10 15:51:16 +0000[diff] [blame]2247 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2248 and must be of <a href="#t_pointer">pointer type</a>.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2249</dl>
2250
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2251<p>The one non-intuitive notation for constants is the hexadecimal form of
2252 floating point constants. For example, the form '<tt>double
2253 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
2254 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
2255 constants are required (and the only time that they are generated by the
2256 disassembler) is when a floating point constant must be emitted but it cannot
2257 be represented as a decimal floating point number in a reasonable number of
2258 digits. For example, NaN's, infinities, and other special values are
2259 represented in their IEEE hexadecimal format so that assembly and disassembly
2260 do not cause any bits to change in the constants.</p>
2261
Dale Johannesencd4a3012009-02-11 22:14:51 +0000[diff] [blame]2262<p>When using the hexadecimal form, constants of types float and double are
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2263 represented using the 16-digit form shown above (which matches the IEEE754
2264 representation for double); float values must, however, be exactly
2265 representable as IEE754 single precision. Hexadecimal format is always used
2266 for long double, and there are three forms of long double. The 80-bit format
2267 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
2268 The 128-bit format used by PowerPC (two adjacent doubles) is represented
2269 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
2270 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
2271 currently supported target uses this format. Long doubles will only work if
2272 they match the long double format on your target. All hexadecimal formats
2273 are big-endian (sign bit at the left).</p>
2274
Dale Johannesen33e5c352010-10-01 00:48:59 +0000[diff] [blame]2275<p>There are no constants of type x86mmx.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2276</div>
2277
2278<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2279<h3>
Bill Wendling972b7202009-07-20 02:32:41 +0000[diff] [blame]2280<a name="aggregateconstants"></a> <!-- old anchor -->
2281<a name="complexconstants">Complex Constants</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2282</h3>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2283
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2284<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2285
Chris Lattner361bfcd2009-02-28 18:32:25 +0000[diff] [blame]2286<p>Complex constants are a (potentially recursive) combination of simple
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2287 constants and smaller complex constants.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2288
2289<dl>
2290 <dt><b>Structure constants</b></dt>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2291 <dd>Structure constants are represented with notation similar to structure
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2292 type definitions (a comma separated list of elements, surrounded by braces
2293 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
2294 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
2295 Structure constants must have <a href="#t_struct">structure type</a>, and
2296 the number and types of elements must match those specified by the
2297 type.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2298
2299 <dt><b>Array constants</b></dt>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2300 <dd>Array constants are represented with notation similar to array type
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2301 definitions (a comma separated list of elements, surrounded by square
2302 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
2303 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
2304 the number and types of elements must match those specified by the
2305 type.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2306
Reid Spencer404a3252007-02-15 03:07:05 +0000[diff] [blame]2307 <dt><b>Vector constants</b></dt>
Reid Spencer404a3252007-02-15 03:07:05 +0000[diff] [blame]2308 <dd>Vector constants are represented with notation similar to vector type
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2309 definitions (a comma separated list of elements, surrounded by
2310 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32
2311 42, i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must
2312 have <a href="#t_vector">vector type</a>, and the number and types of
2313 elements must match those specified by the type.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2314
2315 <dt><b>Zero initialization</b></dt>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2316 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]2317 value to zero of <em>any</em> type, including scalar and
2318 <a href="#t_aggregate">aggregate</a> types.
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2319 This is often used to avoid having to print large zero initializers
2320 (e.g. for large arrays) and is always exactly equivalent to using explicit
2321 zero initializers.</dd>
Nick Lewycky49f89192009-04-04 07:22:01 +0000[diff] [blame]2322
2323 <dt><b>Metadata node</b></dt>
Nick Lewycky8e2c4f42009-05-30 16:08:30 +0000[diff] [blame]2324 <dd>A metadata node is a structure-like constant with
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2325 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
2326 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
2327 be interpreted as part of the instruction stream, metadata is a place to
2328 attach additional information such as debug info.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2329</dl>
2330
2331</div>
2332
2333<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2334<h3>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2335 <a name="globalconstants">Global Variable and Function Addresses</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2336</h3>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2337
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2338<div>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2339
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2340<p>The addresses of <a href="#globalvars">global variables</a>
2341 and <a href="#functionstructure">functions</a> are always implicitly valid
2342 (link-time) constants. These constants are explicitly referenced when
2343 the <a href="#identifiers">identifier for the global</a> is used and always
2344 have <a href="#t_pointer">pointer</a> type. For example, the following is a
2345 legal LLVM file:</p>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2346
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]2347<pre class="doc_code">
Chris Lattner00538a12007-06-06 18:28:13 +0000[diff] [blame]2348@X = global i32 17
2349@Y = global i32 42
2350@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2351</pre>
2352
2353</div>
2354
2355<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2356<h3>
2357 <a name="undefvalues">Undefined Values</a>
2358</h3>
2359
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2360<div>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2361
Chris Lattnerec72b9b2009-09-07 22:52:39 +0000[diff] [blame]2362<p>The string '<tt>undef</tt>' can be used anywhere a constant is expected, and
Benjamin Kramer0f420382009-10-12 14:46:08 +0000[diff] [blame]2363 indicates that the user of the value may receive an unspecified bit-pattern.
Bill Wendling6bbe0912010-10-27 01:07:41 +0000[diff] [blame]2364 Undefined values may be of any type (other than '<tt>label</tt>'
2365 or '<tt>void</tt>') and be used anywhere a constant is permitted.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2366
Chris Lattner92ada5d2009-09-11 01:49:31 +0000[diff] [blame]2367<p>Undefined values are useful because they indicate to the compiler that the
Chris Lattnerec72b9b2009-09-07 22:52:39 +0000[diff] [blame]2368 program is well defined no matter what value is used. This gives the
2369 compiler more freedom to optimize. Here are some examples of (potentially
2370 surprising) transformations that are valid (in pseudo IR):</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2371
Chris Lattnerec72b9b2009-09-07 22:52:39 +0000[diff] [blame]2372
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]2373<pre class="doc_code">
Chris Lattnerec72b9b2009-09-07 22:52:39 +0000[diff] [blame]2374 %A = add %X, undef
2375 %B = sub %X, undef
2376 %C = xor %X, undef
2377Safe:
2378 %A = undef
2379 %B = undef
2380 %C = undef
2381</pre>
Chris Lattnerec72b9b2009-09-07 22:52:39 +0000[diff] [blame]2382
2383<p>This is safe because all of the output bits are affected by the undef bits.
Bill Wendling6bbe0912010-10-27 01:07:41 +0000[diff] [blame]2384 Any output bit can have a zero or one depending on the input bits.</p>
Chris Lattnerec72b9b2009-09-07 22:52:39 +0000[diff] [blame]2385
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]2386<pre class="doc_code">
Chris Lattnerec72b9b2009-09-07 22:52:39 +0000[diff] [blame]2387 %A = or %X, undef
2388 %B = and %X, undef
2389Safe:
2390 %A = -1
2391 %B = 0
2392Unsafe:
2393 %A = undef
2394 %B = undef
2395</pre>
Chris Lattnerec72b9b2009-09-07 22:52:39 +0000[diff] [blame]2396
2397<p>These logical operations have bits that are not always affected by the input.
Bill Wendling6bbe0912010-10-27 01:07:41 +0000[diff] [blame]2398 For example, if <tt>%X</tt> has a zero bit, then the output of the
2399 '<tt>and</tt>' operation will always be a zero for that bit, no matter what
2400 the corresponding bit from the '<tt>undef</tt>' is. As such, it is unsafe to
2401 optimize or assume that the result of the '<tt>and</tt>' is '<tt>undef</tt>'.
2402 However, it is safe to assume that all bits of the '<tt>undef</tt>' could be
2403 0, and optimize the '<tt>and</tt>' to 0. Likewise, it is safe to assume that
2404 all the bits of the '<tt>undef</tt>' operand to the '<tt>or</tt>' could be
2405 set, allowing the '<tt>or</tt>' to be folded to -1.</p>
Chris Lattnerec72b9b2009-09-07 22:52:39 +0000[diff] [blame]2406
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]2407<pre class="doc_code">
Chris Lattnerec72b9b2009-09-07 22:52:39 +0000[diff] [blame]2408 %A = select undef, %X, %Y
2409 %B = select undef, 42, %Y
2410 %C = select %X, %Y, undef
2411Safe:
2412 %A = %X (or %Y)
2413 %B = 42 (or %Y)
2414 %C = %Y
2415Unsafe:
2416 %A = undef
2417 %B = undef
2418 %C = undef
2419</pre>
Chris Lattnerec72b9b2009-09-07 22:52:39 +0000[diff] [blame]2420
Bill Wendling6bbe0912010-10-27 01:07:41 +0000[diff] [blame]2421<p>This set of examples shows that undefined '<tt>select</tt>' (and conditional
2422 branch) conditions can go <em>either way</em>, but they have to come from one
2423 of the two operands. In the <tt>%A</tt> example, if <tt>%X</tt> and
2424 <tt>%Y</tt> were both known to have a clear low bit, then <tt>%A</tt> would
2425 have to have a cleared low bit. However, in the <tt>%C</tt> example, the
2426 optimizer is allowed to assume that the '<tt>undef</tt>' operand could be the
2427 same as <tt>%Y</tt>, allowing the whole '<tt>select</tt>' to be
2428 eliminated.</p>
Chris Lattnerec72b9b2009-09-07 22:52:39 +0000[diff] [blame]2429
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]2430<pre class="doc_code">
Chris Lattnerec72b9b2009-09-07 22:52:39 +0000[diff] [blame]2431 %A = xor undef, undef
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]2432
Chris Lattnerec72b9b2009-09-07 22:52:39 +0000[diff] [blame]2433 %B = undef
2434 %C = xor %B, %B
2435
2436 %D = undef
2437 %E = icmp lt %D, 4
2438 %F = icmp gte %D, 4
2439
2440Safe:
2441 %A = undef
2442 %B = undef
2443 %C = undef
2444 %D = undef
2445 %E = undef
2446 %F = undef
2447</pre>
Chris Lattnerec72b9b2009-09-07 22:52:39 +0000[diff] [blame]2448
Bill Wendling6bbe0912010-10-27 01:07:41 +0000[diff] [blame]2449<p>This example points out that two '<tt>undef</tt>' operands are not
2450 necessarily the same. This can be surprising to people (and also matches C
2451 semantics) where they assume that "<tt>X^X</tt>" is always zero, even
2452 if <tt>X</tt> is undefined. This isn't true for a number of reasons, but the
2453 short answer is that an '<tt>undef</tt>' "variable" can arbitrarily change
2454 its value over its "live range". This is true because the variable doesn't
2455 actually <em>have a live range</em>. Instead, the value is logically read
2456 from arbitrary registers that happen to be around when needed, so the value
2457 is not necessarily consistent over time. In fact, <tt>%A</tt> and <tt>%C</tt>
2458 need to have the same semantics or the core LLVM "replace all uses with"
2459 concept would not hold.</p>
Chris Lattnera34a7182009-09-07 23:33:52 +0000[diff] [blame]2460
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]2461<pre class="doc_code">
Chris Lattnera34a7182009-09-07 23:33:52 +0000[diff] [blame]2462 %A = fdiv undef, %X
2463 %B = fdiv %X, undef
2464Safe:
2465 %A = undef
2466b: unreachable
2467</pre>
Chris Lattnera34a7182009-09-07 23:33:52 +0000[diff] [blame]2468
2469<p>These examples show the crucial difference between an <em>undefined
Bill Wendling6bbe0912010-10-27 01:07:41 +0000[diff] [blame]2470 value</em> and <em>undefined behavior</em>. An undefined value (like
2471 '<tt>undef</tt>') is allowed to have an arbitrary bit-pattern. This means that
2472 the <tt>%A</tt> operation can be constant folded to '<tt>undef</tt>', because
2473 the '<tt>undef</tt>' could be an SNaN, and <tt>fdiv</tt> is not (currently)
2474 defined on SNaN's. However, in the second example, we can make a more
2475 aggressive assumption: because the <tt>undef</tt> is allowed to be an
2476 arbitrary value, we are allowed to assume that it could be zero. Since a
2477 divide by zero has <em>undefined behavior</em>, we are allowed to assume that
2478 the operation does not execute at all. This allows us to delete the divide and
2479 all code after it. Because the undefined operation "can't happen", the
2480 optimizer can assume that it occurs in dead code.</p>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]2481
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]2482<pre class="doc_code">
Chris Lattnera34a7182009-09-07 23:33:52 +0000[diff] [blame]2483a: store undef -> %X
2484b: store %X -> undef
2485Safe:
2486a: &lt;deleted&gt;
2487b: unreachable
2488</pre>
Chris Lattnera34a7182009-09-07 23:33:52 +0000[diff] [blame]2489
Bill Wendling6bbe0912010-10-27 01:07:41 +0000[diff] [blame]2490<p>These examples reiterate the <tt>fdiv</tt> example: a store <em>of</em> an
2491 undefined value can be assumed to not have any effect; we can assume that the
2492 value is overwritten with bits that happen to match what was already there.
2493 However, a store <em>to</em> an undefined location could clobber arbitrary
2494 memory, therefore, it has undefined behavior.</p>
Chris Lattnera34a7182009-09-07 23:33:52 +0000[diff] [blame]2495
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2496</div>
2497
2498<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2499<h3>
2500 <a name="trapvalues">Trap Values</a>
2501</h3>
2502
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2503<div>
Dan Gohmanffc9a6b2010-04-22 23:14:21 +0000[diff] [blame]2504
Dan Gohmanb2a709b2010-04-26 20:21:21 +0000[diff] [blame]2505<p>Trap values are similar to <a href="#undefvalues">undef values</a>, however
Dan Gohmanffc9a6b2010-04-22 23:14:21 +0000[diff] [blame]2506 instead of representing an unspecified bit pattern, they represent the
2507 fact that an instruction or constant expression which cannot evoke side
2508 effects has nevertheless detected a condition which results in undefined
Dan Gohmanb2a709b2010-04-26 20:21:21 +0000[diff] [blame]2509 behavior.</p>
Dan Gohmanffc9a6b2010-04-22 23:14:21 +0000[diff] [blame]2510
Dan Gohman2f1ae062010-04-28 00:49:41 +0000[diff] [blame]2511<p>There is currently no way of representing a trap value in the IR; they
Dan Gohmanac355aa2010-05-03 14:51:43 +0000[diff] [blame]2512 only exist when produced by operations such as
Dan Gohman2f1ae062010-04-28 00:49:41 +0000[diff] [blame]2513 <a href="#i_add"><tt>add</tt></a> with the <tt>nsw</tt> flag.</p>
Dan Gohmanb8b85c12010-04-26 23:36:52 +0000[diff] [blame]2514
Dan Gohman2f1ae062010-04-28 00:49:41 +0000[diff] [blame]2515<p>Trap value behavior is defined in terms of value <i>dependence</i>:</p>
Dan Gohmanb8b85c12010-04-26 23:36:52 +0000[diff] [blame]2516
Dan Gohman2f1ae062010-04-28 00:49:41 +0000[diff] [blame]2517<ul>
2518<li>Values other than <a href="#i_phi"><tt>phi</tt></a> nodes depend on
2519 their operands.</li>
2520
2521<li><a href="#i_phi"><tt>Phi</tt></a> nodes depend on the operand corresponding
2522 to their dynamic predecessor basic block.</li>
2523
2524<li>Function arguments depend on the corresponding actual argument values in
2525 the dynamic callers of their functions.</li>
2526
2527<li><a href="#i_call"><tt>Call</tt></a> instructions depend on the
2528 <a href="#i_ret"><tt>ret</tt></a> instructions that dynamically transfer
2529 control back to them.</li>
2530
Dan Gohman7292a752010-05-03 14:55:22 +0000[diff] [blame]2531<li><a href="#i_invoke"><tt>Invoke</tt></a> instructions depend on the
2532 <a href="#i_ret"><tt>ret</tt></a>, <a href="#i_unwind"><tt>unwind</tt></a>,
2533 or exception-throwing call instructions that dynamically transfer control
2534 back to them.</li>
2535
Dan Gohman2f1ae062010-04-28 00:49:41 +0000[diff] [blame]2536<li>Non-volatile loads and stores depend on the most recent stores to all of the
2537 referenced memory addresses, following the order in the IR
2538 (including loads and stores implied by intrinsics such as
2539 <a href="#int_memcpy"><tt>@llvm.memcpy</tt></a>.)</li>
2540
Dan Gohman3513ea52010-05-03 14:59:34 +0000[diff] [blame]2541<!-- TODO: In the case of multiple threads, this only applies if the store
2542 "happens-before" the load or store. -->
Dan Gohmanb8b85c12010-04-26 23:36:52 +0000[diff] [blame]2543
Dan Gohman2f1ae062010-04-28 00:49:41 +0000[diff] [blame]2544<!-- TODO: floating-point exception state -->
Dan Gohmanb8b85c12010-04-26 23:36:52 +0000[diff] [blame]2545
Dan Gohman2f1ae062010-04-28 00:49:41 +0000[diff] [blame]2546<li>An instruction with externally visible side effects depends on the most
2547 recent preceding instruction with externally visible side effects, following
Dan Gohman6c858db2010-07-06 15:26:33 +0000[diff] [blame]2548 the order in the IR. (This includes
2549 <a href="#volatile">volatile operations</a>.)</li>
Dan Gohmanffc9a6b2010-04-22 23:14:21 +0000[diff] [blame]2550
Dan Gohman7292a752010-05-03 14:55:22 +0000[diff] [blame]2551<li>An instruction <i>control-depends</i> on a
2552 <a href="#terminators">terminator instruction</a>
2553 if the terminator instruction has multiple successors and the instruction
2554 is always executed when control transfers to one of the successors, and
Chris Lattner0ab5e2c2011-04-15 05:18:47 +0000[diff] [blame]2555 may not be executed when control is transferred to another.</li>
Dan Gohman2f1ae062010-04-28 00:49:41 +0000[diff] [blame]2556
Dan Gohmanc8454ee2011-04-12 23:05:59 +0000[diff] [blame]2557<li>Additionally, an instruction also <i>control-depends</i> on a terminator
2558 instruction if the set of instructions it otherwise depends on would be
Chris Lattner0ab5e2c2011-04-15 05:18:47 +0000[diff] [blame]2559 different if the terminator had transferred control to a different
Dan Gohmanc8454ee2011-04-12 23:05:59 +0000[diff] [blame]2560 successor.</li>
2561
Dan Gohman2f1ae062010-04-28 00:49:41 +0000[diff] [blame]2562<li>Dependence is transitive.</li>
2563
2564</ul>
Dan Gohman2f1ae062010-04-28 00:49:41 +0000[diff] [blame]2565
2566<p>Whenever a trap value is generated, all values which depend on it evaluate
2567 to trap. If they have side effects, the evoke their side effects as if each
2568 operand with a trap value were undef. If they have externally-visible side
2569 effects, the behavior is undefined.</p>
2570
2571<p>Here are some examples:</p>
Dan Gohman48a25882010-04-26 20:54:53 +0000[diff] [blame]2572
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]2573<pre class="doc_code">
Dan Gohmanb8b85c12010-04-26 23:36:52 +0000[diff] [blame]2574entry:
2575 %trap = sub nuw i32 0, 1 ; Results in a trap value.
Dan Gohman2f1ae062010-04-28 00:49:41 +0000[diff] [blame]2576 %still_trap = and i32 %trap, 0 ; Whereas (and i32 undef, 0) would return 0.
2577 %trap_yet_again = getelementptr i32* @h, i32 %still_trap
2578 store i32 0, i32* %trap_yet_again ; undefined behavior
2579
2580 store i32 %trap, i32* @g ; Trap value conceptually stored to memory.
2581 %trap2 = load i32* @g ; Returns a trap value, not just undef.
2582
2583 volatile store i32 %trap, i32* @g ; External observation; undefined behavior.
2584
2585 %narrowaddr = bitcast i32* @g to i16*
2586 %wideaddr = bitcast i32* @g to i64*
Nick Lewycky9c876bf2011-05-16 19:29:30 +0000[diff] [blame]2587 %trap3 = load i16* %narrowaddr ; Returns a trap value.
2588 %trap4 = load i64* %wideaddr ; Returns a trap value.
Dan Gohman2f1ae062010-04-28 00:49:41 +0000[diff] [blame]2589
Nick Lewycky9c876bf2011-05-16 19:29:30 +0000[diff] [blame]2590 %cmp = icmp slt i32 %trap, 0 ; Returns a trap value.
2591 br i1 %cmp, label %true, label %end ; Branch to either destination.
Dan Gohmanb8b85c12010-04-26 23:36:52 +0000[diff] [blame]2592
2593true:
Dan Gohman2f1ae062010-04-28 00:49:41 +0000[diff] [blame]2594 volatile store i32 0, i32* @g ; This is control-dependent on %cmp, so
2595 ; it has undefined behavior.
Dan Gohmanb8b85c12010-04-26 23:36:52 +0000[diff] [blame]2596 br label %end
2597
2598end:
2599 %p = phi i32 [ 0, %entry ], [ 1, %true ]
2600 ; Both edges into this PHI are
2601 ; control-dependent on %cmp, so this
Dan Gohman2f1ae062010-04-28 00:49:41 +0000[diff] [blame]2602 ; always results in a trap value.
Dan Gohmanb8b85c12010-04-26 23:36:52 +0000[diff] [blame]2603
Dan Gohmanc8454ee2011-04-12 23:05:59 +0000[diff] [blame]2604 volatile store i32 0, i32* @g ; This would depend on the store in %true
2605 ; if %cmp is true, or the store in %entry
2606 ; otherwise, so this is undefined behavior.
2607
Nick Lewycky9c876bf2011-05-16 19:29:30 +0000[diff] [blame]2608 br i1 %cmp, label %second_true, label %second_end
Dan Gohmanc8454ee2011-04-12 23:05:59 +0000[diff] [blame]2609 ; The same branch again, but this time the
2610 ; true block doesn't have side effects.
2611
2612second_true:
2613 ; No side effects!
Nick Lewycky9c876bf2011-05-16 19:29:30 +0000[diff] [blame]2614 ret void
Dan Gohmanc8454ee2011-04-12 23:05:59 +0000[diff] [blame]2615
2616second_end:
2617 volatile store i32 0, i32* @g ; This time, the instruction always depends
2618 ; on the store in %end. Also, it is
2619 ; control-equivalent to %end, so this is
Nick Lewycky9c876bf2011-05-16 19:29:30 +0000[diff] [blame]2620 ; well-defined (again, ignoring earlier
Dan Gohmanb8b85c12010-04-26 23:36:52 +0000[diff] [blame]2621 ; undefined behavior in this example).
Dan Gohmanb8b85c12010-04-26 23:36:52 +0000[diff] [blame]2622</pre>
Dan Gohmanffc9a6b2010-04-22 23:14:21 +0000[diff] [blame]2623
Dan Gohmanffc9a6b2010-04-22 23:14:21 +0000[diff] [blame]2624</div>
2625
2626<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2627<h3>
2628 <a name="blockaddress">Addresses of Basic Blocks</a>
2629</h3>
2630
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2631<div>
Chris Lattnere4801f72009-10-27 21:01:34 +0000[diff] [blame]2632
Chris Lattneraa99c942009-11-01 01:27:45 +0000[diff] [blame]2633<p><b><tt>blockaddress(@function, %block)</tt></b></p>
Chris Lattnere4801f72009-10-27 21:01:34 +0000[diff] [blame]2634
2635<p>The '<tt>blockaddress</tt>' constant computes the address of the specified
Chris Lattner5c5f0ac2009-10-27 21:49:40 +0000[diff] [blame]2636 basic block in the specified function, and always has an i8* type. Taking
Chris Lattneraa99c942009-11-01 01:27:45 +0000[diff] [blame]2637 the address of the entry block is illegal.</p>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]2638
Chris Lattnere4801f72009-10-27 21:01:34 +0000[diff] [blame]2639<p>This value only has defined behavior when used as an operand to the
Bill Wendling6bbe0912010-10-27 01:07:41 +0000[diff] [blame]2640 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>' instruction, or for
2641 comparisons against null. Pointer equality tests between labels addresses
2642 results in undefined behavior &mdash; though, again, comparison against null
2643 is ok, and no label is equal to the null pointer. This may be passed around
2644 as an opaque pointer sized value as long as the bits are not inspected. This
2645 allows <tt>ptrtoint</tt> and arithmetic to be performed on these values so
2646 long as the original value is reconstituted before the <tt>indirectbr</tt>
2647 instruction.</p>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]2648
Bill Wendling6bbe0912010-10-27 01:07:41 +0000[diff] [blame]2649<p>Finally, some targets may provide defined semantics when using the value as
2650 the operand to an inline assembly, but that is target specific.</p>
Chris Lattnere4801f72009-10-27 21:01:34 +0000[diff] [blame]2651
2652</div>
2653
2654
2655<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2656<h3>
2657 <a name="constantexprs">Constant Expressions</a>
2658</h3>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2659
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2660<div>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2661
2662<p>Constant expressions are used to allow expressions involving other constants
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2663 to be used as constants. Constant expressions may be of
2664 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2665 operation that does not have side effects (e.g. load and call are not
Bill Wendling6bbe0912010-10-27 01:07:41 +0000[diff] [blame]2666 supported). The following is the syntax for constant expressions:</p>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2667
2668<dl>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2669 <dt><b><tt>trunc (CST to TYPE)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2670 <dd>Truncate a constant to another type. The bit size of CST must be larger
2671 than the bit size of TYPE. Both types must be integers.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2672
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2673 <dt><b><tt>zext (CST to TYPE)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2674 <dd>Zero extend a constant to another type. The bit size of CST must be
Duncan Sandsa522e562010-07-13 12:06:14 +0000[diff] [blame]2675 smaller than the bit size of TYPE. Both types must be integers.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]2676
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2677 <dt><b><tt>sext (CST to TYPE)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2678 <dd>Sign extend a constant to another type. The bit size of CST must be
Duncan Sandsa522e562010-07-13 12:06:14 +0000[diff] [blame]2679 smaller than the bit size of TYPE. Both types must be integers.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]2680
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2681 <dt><b><tt>fptrunc (CST to TYPE)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2682 <dd>Truncate a floating point constant to another floating point type. The
2683 size of CST must be larger than the size of TYPE. Both types must be
2684 floating point.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]2685
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2686 <dt><b><tt>fpext (CST to TYPE)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2687 <dd>Floating point extend a constant to another type. The size of CST must be
2688 smaller or equal to the size of TYPE. Both types must be floating
2689 point.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]2690
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2691 <dt><b><tt>fptoui (CST to TYPE)</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]2692 <dd>Convert a floating point constant to the corresponding unsigned integer
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2693 constant. TYPE must be a scalar or vector integer type. CST must be of
2694 scalar or vector floating point type. Both CST and TYPE must be scalars,
2695 or vectors of the same number of elements. If the value won't fit in the
2696 integer type, the results are undefined.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]2697
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2698 <dt><b><tt>fptosi (CST to TYPE)</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]2699 <dd>Convert a floating point constant to the corresponding signed integer
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2700 constant. TYPE must be a scalar or vector integer type. CST must be of
2701 scalar or vector floating point type. Both CST and TYPE must be scalars,
2702 or vectors of the same number of elements. If the value won't fit in the
2703 integer type, the results are undefined.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]2704
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2705 <dt><b><tt>uitofp (CST to TYPE)</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]2706 <dd>Convert an unsigned integer constant to the corresponding floating point
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2707 constant. TYPE must be a scalar or vector floating point type. CST must be
2708 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2709 vectors of the same number of elements. If the value won't fit in the
2710 floating point type, the results are undefined.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]2711
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2712 <dt><b><tt>sitofp (CST to TYPE)</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]2713 <dd>Convert a signed integer constant to the corresponding floating point
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2714 constant. TYPE must be a scalar or vector floating point type. CST must be
2715 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2716 vectors of the same number of elements. If the value won't fit in the
2717 floating point type, the results are undefined.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]2718
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2719 <dt><b><tt>ptrtoint (CST to TYPE)</tt></b></dt>
Reid Spencer5b950642006-11-11 23:08:07 +0000[diff] [blame]2720 <dd>Convert a pointer typed constant to the corresponding integer constant
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2721 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2722 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2723 make it fit in <tt>TYPE</tt>.</dd>
Reid Spencer5b950642006-11-11 23:08:07 +0000[diff] [blame]2724
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2725 <dt><b><tt>inttoptr (CST to TYPE)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2726 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2727 type. CST must be of integer type. The CST value is zero extended,
2728 truncated, or unchanged to make it fit in a pointer size. This one is
2729 <i>really</i> dangerous!</dd>
Reid Spencer5b950642006-11-11 23:08:07 +0000[diff] [blame]2730
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2731 <dt><b><tt>bitcast (CST to TYPE)</tt></b></dt>
Chris Lattner789dee32009-02-28 18:27:03 +0000[diff] [blame]2732 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2733 are the same as those for the <a href="#i_bitcast">bitcast
2734 instruction</a>.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2735
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2736 <dt><b><tt>getelementptr (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2737 <dt><b><tt>getelementptr inbounds (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2738 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2739 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2740 instruction, the index list may have zero or more indexes, which are
2741 required to make sense for the type of "CSTPTR".</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2742
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2743 <dt><b><tt>select (COND, VAL1, VAL2)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2744 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
Reid Spencer9965ee72006-12-04 19:23:19 +0000[diff] [blame]2745
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2746 <dt><b><tt>icmp COND (VAL1, VAL2)</tt></b></dt>
Reid Spencer9965ee72006-12-04 19:23:19 +0000[diff] [blame]2747 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2748
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2749 <dt><b><tt>fcmp COND (VAL1, VAL2)</tt></b></dt>
Reid Spencer9965ee72006-12-04 19:23:19 +0000[diff] [blame]2750 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
Robert Bocchino7e97a6d2006-01-10 19:31:34 +0000[diff] [blame]2751
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2752 <dt><b><tt>extractelement (VAL, IDX)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2753 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2754 constants.</dd>
Robert Bocchino7e97a6d2006-01-10 19:31:34 +0000[diff] [blame]2755
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2756 <dt><b><tt>insertelement (VAL, ELT, IDX)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2757 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2758 constants.</dd>
Chris Lattner016a0e52006-04-08 00:13:41 +0000[diff] [blame]2759
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2760 <dt><b><tt>shufflevector (VEC1, VEC2, IDXMASK)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2761 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2762 constants.</dd>
Chris Lattner016a0e52006-04-08 00:13:41 +0000[diff] [blame]2763
Nick Lewycky9ab9a7f2010-05-29 06:44:15 +0000[diff] [blame]2764 <dt><b><tt>extractvalue (VAL, IDX0, IDX1, ...)</tt></b></dt>
2765 <dd>Perform the <a href="#i_extractvalue">extractvalue operation</a> on
2766 constants. The index list is interpreted in a similar manner as indices in
2767 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2768 index value must be specified.</dd>
2769
2770 <dt><b><tt>insertvalue (VAL, ELT, IDX0, IDX1, ...)</tt></b></dt>
2771 <dd>Perform the <a href="#i_insertvalue">insertvalue operation</a> on
2772 constants. The index list is interpreted in a similar manner as indices in
2773 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2774 index value must be specified.</dd>
2775
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]2776 <dt><b><tt>OPCODE (LHS, RHS)</tt></b></dt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2777 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2778 be any of the <a href="#binaryops">binary</a>
2779 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2780 on operands are the same as those for the corresponding instruction
2781 (e.g. no bitwise operations on floating point values are allowed).</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2782</dl>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2783
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]2784</div>
Chris Lattnerb1652612004-03-08 16:49:10 +0000[diff] [blame]2785
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2786</div>
2787
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]2788<!-- *********************************************************************** -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2789<h2><a name="othervalues">Other Values</a></h2>
Chris Lattner98f013c2006-01-25 23:47:57 +0000[diff] [blame]2790<!-- *********************************************************************** -->
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2791<div>
Chris Lattner98f013c2006-01-25 23:47:57 +0000[diff] [blame]2792<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2793<h3>
Chris Lattner98f013c2006-01-25 23:47:57 +0000[diff] [blame]2794<a name="inlineasm">Inline Assembler Expressions</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2795</h3>
Chris Lattner98f013c2006-01-25 23:47:57 +0000[diff] [blame]2796
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2797<div>
Chris Lattner98f013c2006-01-25 23:47:57 +0000[diff] [blame]2798
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2799<p>LLVM supports inline assembler expressions (as opposed
2800 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2801 a special value. This value represents the inline assembler as a string
2802 (containing the instructions to emit), a list of operand constraints (stored
Dale Johannesen63c94fe2009-10-13 21:56:55 +0000[diff] [blame]2803 as a string), a flag that indicates whether or not the inline asm
Dale Johannesen1cfb9582009-10-21 23:28:00 +0000[diff] [blame]2804 expression has side effects, and a flag indicating whether the function
2805 containing the asm needs to align its stack conservatively. An example
2806 inline assembler expression is:</p>
Chris Lattner98f013c2006-01-25 23:47:57 +0000[diff] [blame]2807
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]2808<pre class="doc_code">
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]2809i32 (i32) asm "bswap $0", "=r,r"
Chris Lattner98f013c2006-01-25 23:47:57 +0000[diff] [blame]2810</pre>
2811
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2812<p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2813 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2814 have:</p>
Chris Lattner98f013c2006-01-25 23:47:57 +0000[diff] [blame]2815
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]2816<pre class="doc_code">
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]2817%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
Chris Lattner98f013c2006-01-25 23:47:57 +0000[diff] [blame]2818</pre>
2819
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2820<p>Inline asms with side effects not visible in the constraint list must be
2821 marked as having side effects. This is done through the use of the
2822 '<tt>sideeffect</tt>' keyword, like so:</p>
Chris Lattner98f013c2006-01-25 23:47:57 +0000[diff] [blame]2823
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]2824<pre class="doc_code">
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]2825call void asm sideeffect "eieio", ""()
Chris Lattner98f013c2006-01-25 23:47:57 +0000[diff] [blame]2826</pre>
2827
Dale Johannesen1cfb9582009-10-21 23:28:00 +0000[diff] [blame]2828<p>In some cases inline asms will contain code that will not work unless the
2829 stack is aligned in some way, such as calls or SSE instructions on x86,
2830 yet will not contain code that does that alignment within the asm.
2831 The compiler should make conservative assumptions about what the asm might
2832 contain and should generate its usual stack alignment code in the prologue
2833 if the '<tt>alignstack</tt>' keyword is present:</p>
Dale Johannesen63c94fe2009-10-13 21:56:55 +0000[diff] [blame]2834
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]2835<pre class="doc_code">
Dale Johannesen1cfb9582009-10-21 23:28:00 +0000[diff] [blame]2836call void asm alignstack "eieio", ""()
Dale Johannesen63c94fe2009-10-13 21:56:55 +0000[diff] [blame]2837</pre>
Dale Johannesen63c94fe2009-10-13 21:56:55 +0000[diff] [blame]2838
2839<p>If both keywords appear the '<tt>sideeffect</tt>' keyword must come
2840 first.</p>
2841
Chris Lattner98f013c2006-01-25 23:47:57 +0000[diff] [blame]2842<p>TODO: The format of the asm and constraints string still need to be
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]2843 documented here. Constraints on what can be done (e.g. duplication, moving,
2844 etc need to be documented). This is probably best done by reference to
2845 another document that covers inline asm from a holistic perspective.</p>
Chris Lattner51065562010-04-07 05:38:05 +0000[diff] [blame]2846
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2847<h4>
Chris Lattner51065562010-04-07 05:38:05 +0000[diff] [blame]2848<a name="inlineasm_md">Inline Asm Metadata</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2849</h4>
Chris Lattner51065562010-04-07 05:38:05 +0000[diff] [blame]2850
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2851<div>
Chris Lattner51065562010-04-07 05:38:05 +0000[diff] [blame]2852
2853<p>The call instructions that wrap inline asm nodes may have a "!srcloc" MDNode
Chris Lattner79ffdc72010-11-17 08:20:42 +0000[diff] [blame]2854 attached to it that contains a list of constant integers. If present, the
2855 code generator will use the integer as the location cookie value when report
Chris Lattner51065562010-04-07 05:38:05 +0000[diff] [blame]2856 errors through the LLVMContext error reporting mechanisms. This allows a
Dan Gohman61110ae2010-04-28 00:36:01 +0000[diff] [blame]2857 front-end to correlate backend errors that occur with inline asm back to the
Chris Lattner51065562010-04-07 05:38:05 +0000[diff] [blame]2858 source code that produced it. For example:</p>
2859
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]2860<pre class="doc_code">
Chris Lattner51065562010-04-07 05:38:05 +0000[diff] [blame]2861call void asm sideeffect "something bad", ""()<b>, !srcloc !42</b>
2862...
2863!42 = !{ i32 1234567 }
2864</pre>
Chris Lattner51065562010-04-07 05:38:05 +0000[diff] [blame]2865
2866<p>It is up to the front-end to make sense of the magic numbers it places in the
Chris Lattner79ffdc72010-11-17 08:20:42 +0000[diff] [blame]2867 IR. If the MDNode contains multiple constants, the code generator will use
2868 the one that corresponds to the line of the asm that the error occurs on.</p>
Chris Lattner98f013c2006-01-25 23:47:57 +0000[diff] [blame]2869
2870</div>
2871
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2872</div>
2873
Chris Lattnerc2f8f162010-01-15 21:50:19 +0000[diff] [blame]2874<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2875<h3>
2876 <a name="metadata">Metadata Nodes and Metadata Strings</a>
2877</h3>
Chris Lattnerc2f8f162010-01-15 21:50:19 +0000[diff] [blame]2878
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2879<div>
Chris Lattnerc2f8f162010-01-15 21:50:19 +0000[diff] [blame]2880
2881<p>LLVM IR allows metadata to be attached to instructions in the program that
2882 can convey extra information about the code to the optimizers and code
2883 generator. One example application of metadata is source-level debug
2884 information. There are two metadata primitives: strings and nodes. All
2885 metadata has the <tt>metadata</tt> type and is identified in syntax by a
2886 preceding exclamation point ('<tt>!</tt>').</p>
2887
2888<p>A metadata string is a string surrounded by double quotes. It can contain
2889 any character by escaping non-printable characters with "\xx" where "xx" is
2890 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
2891
2892<p>Metadata nodes are represented with notation similar to structure constants
2893 (a comma separated list of elements, surrounded by braces and preceded by an
2894 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2895 10}</tt>". Metadata nodes can have any values as their operand.</p>
2896
2897<p>A <a href="#namedmetadatastructure">named metadata</a> is a collection of
2898 metadata nodes, which can be looked up in the module symbol table. For
2899 example: "<tt>!foo = metadata !{!4, !3}</tt>".
2900
Devang Patel9984bd62010-03-04 23:44:48 +0000[diff] [blame]2901<p>Metadata can be used as function arguments. Here <tt>llvm.dbg.value</tt>
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]2902 function is using two metadata arguments.</p>
Devang Patel9984bd62010-03-04 23:44:48 +0000[diff] [blame]2903
Bill Wendlingc0e10672011-03-02 02:17:11 +0000[diff] [blame]2904<div class="doc_code">
2905<pre>
2906call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
2907</pre>
2908</div>
Devang Patel9984bd62010-03-04 23:44:48 +0000[diff] [blame]2909
2910<p>Metadata can be attached with an instruction. Here metadata <tt>!21</tt> is
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]2911 attached with <tt>add</tt> instruction using <tt>!dbg</tt> identifier.</p>
Devang Patel9984bd62010-03-04 23:44:48 +0000[diff] [blame]2912
Bill Wendlingc0e10672011-03-02 02:17:11 +0000[diff] [blame]2913<div class="doc_code">
2914<pre>
2915%indvar.next = add i64 %indvar, 1, !dbg !21
2916</pre>
2917</div>
2918
Chris Lattnerc2f8f162010-01-15 21:50:19 +0000[diff] [blame]2919</div>
2920
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2921</div>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]2922
2923<!-- *********************************************************************** -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2924<h2>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]2925 <a name="intrinsic_globals">Intrinsic Global Variables</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2926</h2>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]2927<!-- *********************************************************************** -->
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2928<div>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]2929<p>LLVM has a number of "magic" global variables that contain data that affect
2930code generation or other IR semantics. These are documented here. All globals
Chris Lattner58f9bb22009-07-20 06:14:25 +0000[diff] [blame]2931of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2932section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2933by LLVM.</p>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]2934
2935<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2936<h3>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]2937<a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2938</h3>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]2939
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2940<div>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]2941
2942<p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2943href="#linkage_appending">appending linkage</a>. This array contains a list of
2944pointers to global variables and functions which may optionally have a pointer
2945cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2946
2947<pre>
2948 @X = global i8 4
2949 @Y = global i32 123
2950
2951 @llvm.used = appending global [2 x i8*] [
2952 i8* @X,
2953 i8* bitcast (i32* @Y to i8*)
2954 ], section "llvm.metadata"
2955</pre>
2956
2957<p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2958compiler, assembler, and linker are required to treat the symbol as if there is
2959a reference to the global that it cannot see. For example, if a variable has
2960internal linkage and no references other than that from the <tt>@llvm.used</tt>
2961list, it cannot be deleted. This is commonly used to represent references from
2962inline asms and other things the compiler cannot "see", and corresponds to
2963"attribute((used))" in GNU C.</p>
2964
2965<p>On some targets, the code generator must emit a directive to the assembler or
2966object file to prevent the assembler and linker from molesting the symbol.</p>
2967
2968</div>
2969
2970<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2971<h3>
2972 <a name="intg_compiler_used">
2973 The '<tt>llvm.compiler.used</tt>' Global Variable
2974 </a>
2975</h3>
Chris Lattner58f9bb22009-07-20 06:14:25 +0000[diff] [blame]2976
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2977<div>
Chris Lattner58f9bb22009-07-20 06:14:25 +0000[diff] [blame]2978
2979<p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2980<tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2981touching the symbol. On targets that support it, this allows an intelligent
2982linker to optimize references to the symbol without being impeded as it would be
2983by <tt>@llvm.used</tt>.</p>
2984
2985<p>This is a rare construct that should only be used in rare circumstances, and
2986should not be exposed to source languages.</p>
2987
2988</div>
2989
2990<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2991<h3>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]2992<a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]2993</h3>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]2994
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]2995<div>
David Chisnalla9d4a6f2010-04-30 19:23:49 +0000[diff] [blame]2996<pre>
2997%0 = type { i32, void ()* }
David Chisnallb492b812010-04-30 19:27:35 +0000[diff] [blame]2998@llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor }]
David Chisnalla9d4a6f2010-04-30 19:23:49 +0000[diff] [blame]2999</pre>
3000<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.
3001</p>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]3002
3003</div>
3004
3005<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3006<h3>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]3007<a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3008</h3>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]3009
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3010<div>
David Chisnalla9d4a6f2010-04-30 19:23:49 +0000[diff] [blame]3011<pre>
3012%0 = type { i32, void ()* }
David Chisnallb492b812010-04-30 19:27:35 +0000[diff] [blame]3013@llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor }]
David Chisnalla9d4a6f2010-04-30 19:23:49 +0000[diff] [blame]3014</pre>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]3015
David Chisnalla9d4a6f2010-04-30 19:23:49 +0000[diff] [blame]3016<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.
3017</p>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]3018
3019</div>
3020
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3021</div>
Chris Lattnerae76db52009-07-20 05:55:19 +0000[diff] [blame]3022
Chris Lattner98f013c2006-01-25 23:47:57 +0000[diff] [blame]3023<!-- *********************************************************************** -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3024<h2><a name="instref">Instruction Reference</a></h2>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]3025<!-- *********************************************************************** -->
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]3026
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3027<div>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]3028
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3029<p>The LLVM instruction set consists of several different classifications of
3030 instructions: <a href="#terminators">terminator
3031 instructions</a>, <a href="#binaryops">binary instructions</a>,
3032 <a href="#bitwiseops">bitwise binary instructions</a>,
3033 <a href="#memoryops">memory instructions</a>, and
3034 <a href="#otherops">other instructions</a>.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]3035
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3036<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3037<h3>
3038 <a name="terminators">Terminator Instructions</a>
3039</h3>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]3040
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3041<div>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]3042
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3043<p>As mentioned <a href="#functionstructure">previously</a>, every basic block
3044 in a program ends with a "Terminator" instruction, which indicates which
3045 block should be executed after the current block is finished. These
3046 terminator instructions typically yield a '<tt>void</tt>' value: they produce
3047 control flow, not values (the one exception being the
3048 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
3049
Chris Lattnerd3d65ab2011-08-02 20:29:13 +0000[diff] [blame]3050<p>The terminator instructions are:
3051 '<a href="#i_ret"><tt>ret</tt></a>',
3052 '<a href="#i_br"><tt>br</tt></a>',
3053 '<a href="#i_switch"><tt>switch</tt></a>',
3054 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>',
3055 '<a href="#i_invoke"><tt>invoke</tt></a>',
3056 '<a href="#i_unwind"><tt>unwind</tt></a>',
3057 '<a href="#i_resume"><tt>resume</tt></a>', and
3058 '<a href="#i_unreachable"><tt>unreachable</tt></a>'.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +0000[diff] [blame]3059
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3060<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3061<h4>
3062 <a name="i_ret">'<tt>ret</tt>' Instruction</a>
3063</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3064
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3065<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3066
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3067<h5>Syntax:</h5>
Dan Gohmancc3132e2008-10-04 19:00:07 +0000[diff] [blame]3068<pre>
3069 ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
Chris Lattner590645f2002-04-14 06:13:44 +0000[diff] [blame]3070 ret void <i>; Return from void function</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3071</pre>
Chris Lattnerda508ac2008-04-23 04:59:35 +0000[diff] [blame]3072
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3073<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3074<p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
3075 a value) from a function back to the caller.</p>
3076
3077<p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
3078 value and then causes control flow, and one that just causes control flow to
3079 occur.</p>
Chris Lattnerda508ac2008-04-23 04:59:35 +0000[diff] [blame]3080
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3081<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3082<p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
3083 return value. The type of the return value must be a
3084 '<a href="#t_firstclass">first class</a>' type.</p>
Dan Gohmancc3132e2008-10-04 19:00:07 +0000[diff] [blame]3085
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3086<p>A function is not <a href="#wellformed">well formed</a> if it it has a
3087 non-void return type and contains a '<tt>ret</tt>' instruction with no return
3088 value or a return value with a type that does not match its type, or if it
3089 has a void return type and contains a '<tt>ret</tt>' instruction with a
3090 return value.</p>
Chris Lattnerda508ac2008-04-23 04:59:35 +0000[diff] [blame]3091
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3092<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3093<p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
3094 the calling function's context. If the caller is a
3095 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
3096 instruction after the call. If the caller was an
3097 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
3098 the beginning of the "normal" destination block. If the instruction returns
3099 a value, that value shall set the call or invoke instruction's return
3100 value.</p>
Chris Lattnerda508ac2008-04-23 04:59:35 +0000[diff] [blame]3101
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3102<h5>Example:</h5>
Chris Lattnerda508ac2008-04-23 04:59:35 +0000[diff] [blame]3103<pre>
3104 ret i32 5 <i>; Return an integer value of 5</i>
Chris Lattner590645f2002-04-14 06:13:44 +0000[diff] [blame]3105 ret void <i>; Return from a void function</i>
Bill Wendling050ee8f2009-02-28 22:12:54 +0000[diff] [blame]3106 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3107</pre>
Dan Gohman3065b612009-01-12 23:12:39 +0000[diff] [blame]3108
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]3109</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3110<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3111<h4>
3112 <a name="i_br">'<tt>br</tt>' Instruction</a>
3113</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3114
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3115<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3116
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3117<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3118<pre>
Bill Wendling16b86742011-07-26 10:41:15 +0000[diff] [blame]3119 br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;
3120 br label &lt;dest&gt; <i>; Unconditional branch</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3121</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3122
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3123<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3124<p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
3125 different basic block in the current function. There are two forms of this
3126 instruction, corresponding to a conditional branch and an unconditional
3127 branch.</p>
3128
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3129<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3130<p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
3131 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
3132 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
3133 target.</p>
3134
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3135<h5>Semantics:</h5>
Reid Spencer36a15422007-01-12 03:35:51 +0000[diff] [blame]3136<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3137 argument is evaluated. If the value is <tt>true</tt>, control flows to the
3138 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
3139 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
3140
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3141<h5>Example:</h5>
Bill Wendling30235112009-07-20 02:39:26 +0000[diff] [blame]3142<pre>
3143Test:
3144 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
3145 br i1 %cond, label %IfEqual, label %IfUnequal
3146IfEqual:
3147 <a href="#i_ret">ret</a> i32 1
3148IfUnequal:
3149 <a href="#i_ret">ret</a> i32 0
3150</pre>
3151
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]3152</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3153
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3154<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3155<h4>
Chris Lattnercf96c6c2004-02-24 04:54:45 +0000[diff] [blame]3156 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3157</h4>
Chris Lattnercf96c6c2004-02-24 04:54:45 +0000[diff] [blame]3158
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3159<div>
Chris Lattnercf96c6c2004-02-24 04:54:45 +0000[diff] [blame]3160
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3161<h5>Syntax:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +0000[diff] [blame]3162<pre>
3163 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
3164</pre>
3165
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3166<h5>Overview:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +0000[diff] [blame]3167<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3168 several different places. It is a generalization of the '<tt>br</tt>'
3169 instruction, allowing a branch to occur to one of many possible
3170 destinations.</p>
Chris Lattnercf96c6c2004-02-24 04:54:45 +0000[diff] [blame]3171
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3172<h5>Arguments:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +0000[diff] [blame]3173<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3174 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
3175 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
3176 The table is not allowed to contain duplicate constant entries.</p>
Chris Lattnercf96c6c2004-02-24 04:54:45 +0000[diff] [blame]3177
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3178<h5>Semantics:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]3179<p>The <tt>switch</tt> instruction specifies a table of values and
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3180 destinations. When the '<tt>switch</tt>' instruction is executed, this table
3181 is searched for the given value. If the value is found, control flow is
Benjamin Kramer0f420382009-10-12 14:46:08 +0000[diff] [blame]3182 transferred to the corresponding destination; otherwise, control flow is
3183 transferred to the default destination.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3184
Chris Lattnercf96c6c2004-02-24 04:54:45 +0000[diff] [blame]3185<h5>Implementation:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +0000[diff] [blame]3186<p>Depending on properties of the target machine and the particular
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3187 <tt>switch</tt> instruction, this instruction may be code generated in
3188 different ways. For example, it could be generated as a series of chained
3189 conditional branches or with a lookup table.</p>
Chris Lattnercf96c6c2004-02-24 04:54:45 +0000[diff] [blame]3190
3191<h5>Example:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +0000[diff] [blame]3192<pre>
3193 <i>; Emulate a conditional br instruction</i>
Reid Spencer36a15422007-01-12 03:35:51 +0000[diff] [blame]3194 %Val = <a href="#i_zext">zext</a> i1 %value to i32
Dan Gohman623806e2009-01-04 23:44:43 +0000[diff] [blame]3195 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
Chris Lattnercf96c6c2004-02-24 04:54:45 +0000[diff] [blame]3196
3197 <i>; Emulate an unconditional br instruction</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]3198 switch i32 0, label %dest [ ]
Chris Lattnercf96c6c2004-02-24 04:54:45 +0000[diff] [blame]3199
3200 <i>; Implement a jump table:</i>
Dan Gohman623806e2009-01-04 23:44:43 +0000[diff] [blame]3201 switch i32 %val, label %otherwise [ i32 0, label %onzero
3202 i32 1, label %onone
3203 i32 2, label %ontwo ]
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3204</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3205
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]3206</div>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3207
Chris Lattner3ed871f2009-10-27 19:13:16 +0000[diff] [blame]3208
3209<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3210<h4>
Chris Lattnerd04cb6d2009-10-28 00:19:10 +0000[diff] [blame]3211 <a name="i_indirectbr">'<tt>indirectbr</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3212</h4>
Chris Lattner3ed871f2009-10-27 19:13:16 +0000[diff] [blame]3213
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3214<div>
Chris Lattner3ed871f2009-10-27 19:13:16 +0000[diff] [blame]3215
3216<h5>Syntax:</h5>
3217<pre>
Chris Lattnerd04cb6d2009-10-28 00:19:10 +0000[diff] [blame]3218 indirectbr &lt;somety&gt;* &lt;address&gt;, [ label &lt;dest1&gt;, label &lt;dest2&gt;, ... ]
Chris Lattner3ed871f2009-10-27 19:13:16 +0000[diff] [blame]3219</pre>
3220
3221<h5>Overview:</h5>
3222
Chris Lattnerd04cb6d2009-10-28 00:19:10 +0000[diff] [blame]3223<p>The '<tt>indirectbr</tt>' instruction implements an indirect branch to a label
Chris Lattner3ed871f2009-10-27 19:13:16 +0000[diff] [blame]3224 within the current function, whose address is specified by
Chris Lattnere4801f72009-10-27 21:01:34 +0000[diff] [blame]3225 "<tt>address</tt>". Address must be derived from a <a
3226 href="#blockaddress">blockaddress</a> constant.</p>
Chris Lattner3ed871f2009-10-27 19:13:16 +0000[diff] [blame]3227
3228<h5>Arguments:</h5>
3229
3230<p>The '<tt>address</tt>' argument is the address of the label to jump to. The
3231 rest of the arguments indicate the full set of possible destinations that the
3232 address may point to. Blocks are allowed to occur multiple times in the
3233 destination list, though this isn't particularly useful.</p>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]3234
Chris Lattner3ed871f2009-10-27 19:13:16 +0000[diff] [blame]3235<p>This destination list is required so that dataflow analysis has an accurate
3236 understanding of the CFG.</p>
3237
3238<h5>Semantics:</h5>
3239
3240<p>Control transfers to the block specified in the address argument. All
3241 possible destination blocks must be listed in the label list, otherwise this
3242 instruction has undefined behavior. This implies that jumps to labels
3243 defined in other functions have undefined behavior as well.</p>
3244
3245<h5>Implementation:</h5>
3246
3247<p>This is typically implemented with a jump through a register.</p>
3248
3249<h5>Example:</h5>
3250<pre>
Chris Lattnerd04cb6d2009-10-28 00:19:10 +0000[diff] [blame]3251 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
Chris Lattner3ed871f2009-10-27 19:13:16 +0000[diff] [blame]3252</pre>
3253
3254</div>
3255
3256
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3257<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3258<h4>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3259 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3260</h4>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3261
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3262<div>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3263
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3264<h5>Syntax:</h5>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3265<pre>
Devang Patel02256232008-10-07 17:48:33 +0000[diff] [blame]3266 &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 +0000[diff] [blame]3267 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3268</pre>
3269
Chris Lattnera8292f32002-05-06 22:08:29 +0000[diff] [blame]3270<h5>Overview:</h5>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3271<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3272 function, with the possibility of control flow transfer to either the
3273 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
3274 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
3275 control flow will return to the "normal" label. If the callee (or any
3276 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
3277 instruction, control is interrupted and continued at the dynamically nearest
3278 "exception" label.</p>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3279
Bill Wendlingbbcb7cd2011-08-02 21:52:38 +0000[diff] [blame]3280<p>The '<tt>exception</tt>' label is a
3281 <i><a href="ExceptionHandling.html#overview">landing pad</a></i> for the
3282 exception. As such, '<tt>exception</tt>' label is required to have the
3283 "<a href="#i_landingpad"><tt>landingpad</tt></a>" instruction, which contains
3284 the information about about the behavior of the program after unwinding
3285 happens, as its first non-PHI instruction. The restrictions on the
3286 "<tt>landingpad</tt>" instruction's tightly couples it to the
3287 "<tt>invoke</tt>" instruction, so that the important information contained
3288 within the "<tt>landingpad</tt>" instruction can't be lost through normal
3289 code motion.</p>
3290
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3291<h5>Arguments:</h5>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]3292<p>This instruction requires several arguments:</p>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3293
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3294<ol>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3295 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
3296 convention</a> the call should use. If none is specified, the call
3297 defaults to using C calling conventions.</li>
Devang Patel7e9b05e2008-10-06 18:50:38 +0000[diff] [blame]3298
3299 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3300 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
3301 '<tt>inreg</tt>' attributes are valid here.</li>
Devang Patel7e9b05e2008-10-06 18:50:38 +0000[diff] [blame]3302
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3303 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3304 function value being invoked. In most cases, this is a direct function
3305 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
3306 off an arbitrary pointer to function value.</li>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3307
3308 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3309 function to be invoked. </li>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3310
3311 <li>'<tt>function args</tt>': argument list whose types match the function
Chris Lattner47f2a832010-03-02 06:36:51 +0000[diff] [blame]3312 signature argument types and parameter attributes. All arguments must be
3313 of <a href="#t_firstclass">first class</a> type. If the function
3314 signature indicates the function accepts a variable number of arguments,
3315 the extra arguments can be specified.</li>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3316
3317 <li>'<tt>normal label</tt>': the label reached when the called function
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3318 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3319
3320 <li>'<tt>exception label</tt>': the label reached when a callee returns with
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3321 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3322
Devang Patel02256232008-10-07 17:48:33 +0000[diff] [blame]3323 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3324 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
3325 '<tt>readnone</tt>' attributes are valid here.</li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3326</ol>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3327
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3328<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3329<p>This instruction is designed to operate as a standard
3330 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
3331 primary difference is that it establishes an association with a label, which
3332 is used by the runtime library to unwind the stack.</p>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3333
3334<p>This instruction is used in languages with destructors to ensure that proper
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3335 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
3336 exception. Additionally, this is important for implementation of
3337 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3338
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3339<p>For the purposes of the SSA form, the definition of the value returned by the
3340 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
3341 block to the "normal" label. If the callee unwinds then no return value is
3342 available.</p>
Dan Gohman9069d892009-05-22 21:47:08 +0000[diff] [blame]3343
Chris Lattner97257f82010-01-15 18:08:37 +0000[diff] [blame]3344<p>Note that the code generator does not yet completely support unwind, and
3345that the invoke/unwind semantics are likely to change in future versions.</p>
3346
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3347<h5>Example:</h5>
Chris Lattner0132aff2005-05-06 22:57:40 +0000[diff] [blame]3348<pre>
Nick Lewycky084ab472008-03-16 07:18:12 +0000[diff] [blame]3349 %retval = invoke i32 @Test(i32 15) to label %Continue
Jeff Cohen5819f182007-04-22 01:17:39 +0000[diff] [blame]3350 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewycky084ab472008-03-16 07:18:12 +0000[diff] [blame]3351 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Jeff Cohen5819f182007-04-22 01:17:39 +0000[diff] [blame]3352 unwind label %TestCleanup <i>; {i32}:retval set</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3353</pre>
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]3354
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3355</div>
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]3356
Chris Lattner5ed60612003-09-03 00:41:47 +0000[diff] [blame]3357<!-- _______________________________________________________________________ -->
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]3358
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3359<h4>
3360 <a name="i_unwind">'<tt>unwind</tt>' Instruction</a>
3361</h4>
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]3362
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3363<div>
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]3364
Chris Lattner5ed60612003-09-03 00:41:47 +0000[diff] [blame]3365<h5>Syntax:</h5>
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]3366<pre>
3367 unwind
3368</pre>
3369
Chris Lattner5ed60612003-09-03 00:41:47 +0000[diff] [blame]3370<h5>Overview:</h5>
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]3371<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3372 at the first callee in the dynamic call stack which used
3373 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
3374 This is primarily used to implement exception handling.</p>
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]3375
Chris Lattner5ed60612003-09-03 00:41:47 +0000[diff] [blame]3376<h5>Semantics:</h5>
Chris Lattnerfe8519c2008-04-19 21:01:16 +0000[diff] [blame]3377<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3378 immediately halt. The dynamic call stack is then searched for the
3379 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
3380 Once found, execution continues at the "exceptional" destination block
3381 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
3382 instruction in the dynamic call chain, undefined behavior results.</p>
3383
Chris Lattner97257f82010-01-15 18:08:37 +0000[diff] [blame]3384<p>Note that the code generator does not yet completely support unwind, and
3385that the invoke/unwind semantics are likely to change in future versions.</p>
3386
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]3387</div>
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]3388
Bill Wendlingf891bf82011-07-31 06:30:59 +0000[diff] [blame]3389 <!-- _______________________________________________________________________ -->
3390
3391<h4>
3392 <a name="i_resume">'<tt>resume</tt>' Instruction</a>
3393</h4>
3394
3395<div>
3396
3397<h5>Syntax:</h5>
3398<pre>
3399 resume &lt;type&gt; &lt;value&gt;
3400</pre>
3401
3402<h5>Overview:</h5>
3403<p>The '<tt>resume</tt>' instruction is a terminator instruction that has no
3404 successors.</p>
3405
3406<h5>Arguments:</h5>
Bill Wendlingbbcb7cd2011-08-02 21:52:38 +0000[diff] [blame]3407<p>The '<tt>resume</tt>' instruction requires one argument, which must have the
Bill Wendlingc5a13612011-08-03 18:37:32 +0000[diff] [blame]3408 same type as the result of any '<tt>landingpad</tt>' instruction in the same
3409 function.</p>
Bill Wendlingf891bf82011-07-31 06:30:59 +0000[diff] [blame]3410
3411<h5>Semantics:</h5>
3412<p>The '<tt>resume</tt>' instruction resumes propagation of an existing
3413 (in-flight) exception whose unwinding was interrupted with
Bill Wendlingbbcb7cd2011-08-02 21:52:38 +0000[diff] [blame]3414 a <a href="#i_landingpad"><tt>landingpad</tt></a> instruction.</p>
Bill Wendlingf891bf82011-07-31 06:30:59 +0000[diff] [blame]3415
3416<h5>Example:</h5>
3417<pre>
Bill Wendlingbbcb7cd2011-08-02 21:52:38 +0000[diff] [blame]3418 resume { i8*, i32 } %exn
Bill Wendlingf891bf82011-07-31 06:30:59 +0000[diff] [blame]3419</pre>
3420
3421</div>
3422
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]3423<!-- _______________________________________________________________________ -->
3424
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3425<h4>
3426 <a name="i_unreachable">'<tt>unreachable</tt>' Instruction</a>
3427</h4>
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]3428
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3429<div>
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]3430
3431<h5>Syntax:</h5>
3432<pre>
3433 unreachable
3434</pre>
3435
3436<h5>Overview:</h5>
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]3437<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3438 instruction is used to inform the optimizer that a particular portion of the
3439 code is not reachable. This can be used to indicate that the code after a
3440 no-return function cannot be reached, and other facts.</p>
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]3441
3442<h5>Semantics:</h5>
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]3443<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3444
Chris Lattner08b7d5b2004-10-16 18:04:13 +0000[diff] [blame]3445</div>
3446
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3447</div>
3448
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3449<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3450<h3>
3451 <a name="binaryops">Binary Operations</a>
3452</h3>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3453
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3454<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3455
3456<p>Binary operators are used to do most of the computation in a program. They
3457 require two operands of the same type, execute an operation on them, and
3458 produce a single value. The operands might represent multiple data, as is
3459 the case with the <a href="#t_vector">vector</a> data type. The result value
3460 has the same type as its operands.</p>
3461
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]3462<p>There are several different binary operators:</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3463
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3464<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3465<h4>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3466 <a name="i_add">'<tt>add</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3467</h4>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3468
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3469<div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3470
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3471<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3472<pre>
Dan Gohmanb07de442009-07-20 22:41:19 +0000[diff] [blame]3473 &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 +0000[diff] [blame]3474 &lt;result&gt; = add nuw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3475 &lt;result&gt; = add nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3476 &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 +0000[diff] [blame]3477</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3478
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3479<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]3480<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3481
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3482<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3483<p>The two arguments to the '<tt>add</tt>' instruction must
3484 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3485 integer values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3486
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3487<h5>Semantics:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3488<p>The value produced is the integer sum of the two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3489
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3490<p>If the sum has unsigned overflow, the result returned is the mathematical
3491 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3492
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3493<p>Because LLVM integers use a two's complement representation, this instruction
3494 is appropriate for both signed and unsigned integers.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3495
Dan Gohman902dfff2009-07-22 22:44:56 +0000[diff] [blame]3496<p><tt>nuw</tt> and <tt>nsw</tt> stand for &quot;No Unsigned Wrap&quot;
3497 and &quot;No Signed Wrap&quot;, respectively. If the <tt>nuw</tt> and/or
3498 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
Dan Gohmanffc9a6b2010-04-22 23:14:21 +0000[diff] [blame]3499 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3500 respectively, occurs.</p>
Dan Gohmanb07de442009-07-20 22:41:19 +0000[diff] [blame]3501
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3502<h5>Example:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3503<pre>
3504 &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3505</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3506
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]3507</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3508
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3509<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3510<h4>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3511 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3512</h4>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3513
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3514<div>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3515
3516<h5>Syntax:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3517<pre>
3518 &lt;result&gt; = fadd &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3519</pre>
3520
3521<h5>Overview:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3522<p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
3523
3524<h5>Arguments:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3525<p>The two arguments to the '<tt>fadd</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3526 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3527 floating point values. Both arguments must have identical types.</p>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3528
3529<h5>Semantics:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3530<p>The value produced is the floating point sum of the two operands.</p>
3531
3532<h5>Example:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3533<pre>
3534 &lt;result&gt; = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
3535</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3536
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3537</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3538
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3539<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3540<h4>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3541 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3542</h4>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3543
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3544<div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3545
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3546<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3547<pre>
Dan Gohman902dfff2009-07-22 22:44:56 +0000[diff] [blame]3548 &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 +0000[diff] [blame]3549 &lt;result&gt; = sub nuw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3550 &lt;result&gt; = sub nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3551 &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 +0000[diff] [blame]3552</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3553
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3554<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]3555<p>The '<tt>sub</tt>' instruction returns the difference of its two
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3556 operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3557
3558<p>Note that the '<tt>sub</tt>' instruction is used to represent the
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3559 '<tt>neg</tt>' instruction present in most other intermediate
3560 representations.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3561
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3562<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3563<p>The two arguments to the '<tt>sub</tt>' instruction must
3564 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3565 integer values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3566
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3567<h5>Semantics:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3568<p>The value produced is the integer difference of the two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3569
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3570<p>If the difference has unsigned overflow, the result returned is the
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3571 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
3572 result.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3573
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3574<p>Because LLVM integers use a two's complement representation, this instruction
3575 is appropriate for both signed and unsigned integers.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3576
Dan Gohman902dfff2009-07-22 22:44:56 +0000[diff] [blame]3577<p><tt>nuw</tt> and <tt>nsw</tt> stand for &quot;No Unsigned Wrap&quot;
3578 and &quot;No Signed Wrap&quot;, respectively. If the <tt>nuw</tt> and/or
3579 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
Dan Gohmanffc9a6b2010-04-22 23:14:21 +0000[diff] [blame]3580 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3581 respectively, occurs.</p>
Dan Gohmanb07de442009-07-20 22:41:19 +0000[diff] [blame]3582
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3583<h5>Example:</h5>
Bill Wendling2d8b9a82007-05-29 09:42:13 +0000[diff] [blame]3584<pre>
3585 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]3586 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3587</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3588
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]3589</div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3590
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3591<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3592<h4>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3593 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3594</h4>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3595
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3596<div>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3597
3598<h5>Syntax:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3599<pre>
3600 &lt;result&gt; = fsub &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3601</pre>
3602
3603<h5>Overview:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3604<p>The '<tt>fsub</tt>' instruction returns the difference of its two
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3605 operands.</p>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3606
3607<p>Note that the '<tt>fsub</tt>' instruction is used to represent the
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3608 '<tt>fneg</tt>' instruction present in most other intermediate
3609 representations.</p>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3610
3611<h5>Arguments:</h5>
Bill Wendling972b7202009-07-20 02:32:41 +0000[diff] [blame]3612<p>The two arguments to the '<tt>fsub</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3613 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3614 floating point values. Both arguments must have identical types.</p>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3615
3616<h5>Semantics:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3617<p>The value produced is the floating point difference of the two operands.</p>
3618
3619<h5>Example:</h5>
3620<pre>
3621 &lt;result&gt; = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
3622 &lt;result&gt; = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
3623</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3624
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3625</div>
3626
3627<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3628<h4>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3629 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3630</h4>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3631
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3632<div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3633
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3634<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3635<pre>
Dan Gohman902dfff2009-07-22 22:44:56 +0000[diff] [blame]3636 &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 +0000[diff] [blame]3637 &lt;result&gt; = mul nuw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3638 &lt;result&gt; = mul nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3639 &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 +0000[diff] [blame]3640</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3641
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3642<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3643<p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3644
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3645<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3646<p>The two arguments to the '<tt>mul</tt>' instruction must
3647 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3648 integer values. Both arguments must have identical types.</p>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]3649
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3650<h5>Semantics:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3651<p>The value produced is the integer product of the two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3652
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3653<p>If the result of the multiplication has unsigned overflow, the result
3654 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
3655 width of the result.</p>
3656
3657<p>Because LLVM integers use a two's complement representation, and the result
3658 is the same width as the operands, this instruction returns the correct
3659 result for both signed and unsigned integers. If a full product
3660 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
3661 be sign-extended or zero-extended as appropriate to the width of the full
3662 product.</p>
3663
Dan Gohman902dfff2009-07-22 22:44:56 +0000[diff] [blame]3664<p><tt>nuw</tt> and <tt>nsw</tt> stand for &quot;No Unsigned Wrap&quot;
3665 and &quot;No Signed Wrap&quot;, respectively. If the <tt>nuw</tt> and/or
3666 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
Dan Gohmanffc9a6b2010-04-22 23:14:21 +0000[diff] [blame]3667 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3668 respectively, occurs.</p>
Dan Gohmanb07de442009-07-20 22:41:19 +0000[diff] [blame]3669
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3670<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3671<pre>
3672 &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3673</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3674
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]3675</div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3676
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3677<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3678<h4>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3679 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3680</h4>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3681
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3682<div>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3683
3684<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3685<pre>
3686 &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 +0000[diff] [blame]3687</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3688
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3689<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3690<p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3691
3692<h5>Arguments:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3693<p>The two arguments to the '<tt>fmul</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3694 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3695 floating point values. Both arguments must have identical types.</p>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3696
3697<h5>Semantics:</h5>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3698<p>The value produced is the floating point product of the two operands.</p>
3699
3700<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3701<pre>
3702 &lt;result&gt; = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3703</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3704
Dan Gohmana5b96452009-06-04 22:49:04 +0000[diff] [blame]3705</div>
3706
3707<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3708<h4>
3709 <a name="i_udiv">'<tt>udiv</tt>' Instruction</a>
3710</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3711
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3712<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3713
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3714<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3715<pre>
Chris Lattner35315d02011-02-06 21:44:57 +0000[diff] [blame]3716 &lt;result&gt; = udiv &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3717 &lt;result&gt; = udiv exact &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3718</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3719
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3720<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3721<p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3722
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3723<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]3724<p>The two arguments to the '<tt>udiv</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3725 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3726 values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3727
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3728<h5>Semantics:</h5>
Chris Lattner2f2427e2008-01-28 00:36:27 +0000[diff] [blame]3729<p>The value produced is the unsigned integer quotient of the two operands.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3730
Chris Lattner2f2427e2008-01-28 00:36:27 +0000[diff] [blame]3731<p>Note that unsigned integer division and signed integer division are distinct
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3732 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
3733
Chris Lattner2f2427e2008-01-28 00:36:27 +0000[diff] [blame]3734<p>Division by zero leads to undefined behavior.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3735
Chris Lattner35315d02011-02-06 21:44:57 +0000[diff] [blame]3736<p>If the <tt>exact</tt> keyword is present, the result value of the
3737 <tt>udiv</tt> is a <a href="#trapvalues">trap value</a> if %op1 is not a
3738 multiple of %op2 (as such, "((a udiv exact b) mul b) == a").</p>
3739
3740
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3741<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3742<pre>
3743 &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3744</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3745
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3746</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3747
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3748<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3749<h4>
3750 <a name="i_sdiv">'<tt>sdiv</tt>' Instruction</a>
3751</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3752
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3753<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3754
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3755<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3756<pre>
Dan Gohmanb07de442009-07-20 22:41:19 +0000[diff] [blame]3757 &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 +0000[diff] [blame]3758 &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 +0000[diff] [blame]3759</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3760
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3761<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3762<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3763
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3764<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]3765<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3766 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3767 values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3768
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3769<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3770<p>The value produced is the signed integer quotient of the two operands rounded
3771 towards zero.</p>
3772
Chris Lattner2f2427e2008-01-28 00:36:27 +0000[diff] [blame]3773<p>Note that signed integer division and unsigned integer division are distinct
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3774 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
3775
Chris Lattner2f2427e2008-01-28 00:36:27 +0000[diff] [blame]3776<p>Division by zero leads to undefined behavior. Overflow also leads to
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3777 undefined behavior; this is a rare case, but can occur, for example, by doing
3778 a 32-bit division of -2147483648 by -1.</p>
3779
Dan Gohman71dfd782009-07-22 00:04:19 +0000[diff] [blame]3780<p>If the <tt>exact</tt> keyword is present, the result value of the
Dan Gohman57255802010-04-23 15:23:32 +0000[diff] [blame]3781 <tt>sdiv</tt> is a <a href="#trapvalues">trap value</a> if the result would
Dan Gohmane501ff72010-07-11 00:08:34 +0000[diff] [blame]3782 be rounded.</p>
Dan Gohmanb07de442009-07-20 22:41:19 +0000[diff] [blame]3783
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3784<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3785<pre>
3786 &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3787</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3788
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3789</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3790
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3791<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3792<h4>
3793 <a name="i_fdiv">'<tt>fdiv</tt>' Instruction</a>
3794</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3795
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3796<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3797
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]3798<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3799<pre>
Gabor Greif0f75ad02008-08-07 21:46:00 +0000[diff] [blame]3800 &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 +0000[diff] [blame]3801</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3802
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3803<h5>Overview:</h5>
3804<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3805
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]3806<h5>Arguments:</h5>
Jeff Cohen5819f182007-04-22 01:17:39 +0000[diff] [blame]3807<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3808 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3809 floating point values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3810
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]3811<h5>Semantics:</h5>
Reid Spencer7e80b0b2006-10-26 06:15:43 +0000[diff] [blame]3812<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3813
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]3814<h5>Example:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3815<pre>
3816 &lt;result&gt; = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]3817</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3818
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]3819</div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3820
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]3821<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3822<h4>
3823 <a name="i_urem">'<tt>urem</tt>' Instruction</a>
3824</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3825
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3826<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3827
Reid Spencer7eb55b32006-11-02 01:53:59 +0000[diff] [blame]3828<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3829<pre>
3830 &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 +0000[diff] [blame]3831</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3832
Reid Spencer7eb55b32006-11-02 01:53:59 +0000[diff] [blame]3833<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3834<p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3835 division of its two arguments.</p>
3836
Reid Spencer7eb55b32006-11-02 01:53:59 +0000[diff] [blame]3837<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]3838<p>The two arguments to the '<tt>urem</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3839 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3840 values. Both arguments must have identical types.</p>
3841
Reid Spencer7eb55b32006-11-02 01:53:59 +0000[diff] [blame]3842<h5>Semantics:</h5>
3843<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3844 This instruction always performs an unsigned division to get the
3845 remainder.</p>
3846
Chris Lattner2f2427e2008-01-28 00:36:27 +0000[diff] [blame]3847<p>Note that unsigned integer remainder and signed integer remainder are
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3848 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3849
Chris Lattner2f2427e2008-01-28 00:36:27 +0000[diff] [blame]3850<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3851
Reid Spencer7eb55b32006-11-02 01:53:59 +0000[diff] [blame]3852<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3853<pre>
3854 &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
Reid Spencer7eb55b32006-11-02 01:53:59 +0000[diff] [blame]3855</pre>
3856
3857</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3858
Reid Spencer7eb55b32006-11-02 01:53:59 +0000[diff] [blame]3859<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3860<h4>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3861 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3862</h4>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3863
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3864<div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3865
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]3866<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3867<pre>
Gabor Greif0f75ad02008-08-07 21:46:00 +0000[diff] [blame]3868 &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 +0000[diff] [blame]3869</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3870
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]3871<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3872<p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3873 division of its two operands. This instruction can also take
3874 <a href="#t_vector">vector</a> versions of the values in which case the
3875 elements must be integers.</p>
Chris Lattnerb8f816e2008-01-04 04:33:49 +0000[diff] [blame]3876
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]3877<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]3878<p>The two arguments to the '<tt>srem</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3879 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3880 values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3881
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]3882<h5>Semantics:</h5>
Reid Spencer7eb55b32006-11-02 01:53:59 +0000[diff] [blame]3883<p>This instruction returns the <i>remainder</i> of a division (where the result
Duncan Sands2769c6e2011-03-07 09:12:24 +0000[diff] [blame]3884 is either zero or has the same sign as the dividend, <tt>op1</tt>), not the
3885 <i>modulo</i> operator (where the result is either zero or has the same sign
3886 as the divisor, <tt>op2</tt>) of a value.
3887 For more information about the difference,
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3888 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3889 Math Forum</a>. For a table of how this is implemented in various languages,
3890 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3891 Wikipedia: modulo operation</a>.</p>
3892
Chris Lattner2f2427e2008-01-28 00:36:27 +0000[diff] [blame]3893<p>Note that signed integer remainder and unsigned integer remainder are
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3894 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3895
Chris Lattner2f2427e2008-01-28 00:36:27 +0000[diff] [blame]3896<p>Taking the remainder of a division by zero leads to undefined behavior.
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3897 Overflow also leads to undefined behavior; this is a rare case, but can
3898 occur, for example, by taking the remainder of a 32-bit division of
3899 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3900 lets srem be implemented using instructions that return both the result of
3901 the division and the remainder.)</p>
3902
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]3903<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3904<pre>
3905 &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
Reid Spencer7eb55b32006-11-02 01:53:59 +0000[diff] [blame]3906</pre>
3907
3908</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3909
Reid Spencer7eb55b32006-11-02 01:53:59 +0000[diff] [blame]3910<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3911<h4>
3912 <a name="i_frem">'<tt>frem</tt>' Instruction</a>
3913</h4>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3914
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3915<div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3916
Reid Spencer7eb55b32006-11-02 01:53:59 +0000[diff] [blame]3917<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3918<pre>
3919 &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 +0000[diff] [blame]3920</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3921
Reid Spencer7eb55b32006-11-02 01:53:59 +0000[diff] [blame]3922<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3923<p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3924 its two operands.</p>
3925
Reid Spencer7eb55b32006-11-02 01:53:59 +0000[diff] [blame]3926<h5>Arguments:</h5>
3927<p>The two arguments to the '<tt>frem</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3928 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3929 floating point values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3930
Reid Spencer7eb55b32006-11-02 01:53:59 +0000[diff] [blame]3931<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3932<p>This instruction returns the <i>remainder</i> of a division. The remainder
3933 has the same sign as the dividend.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3934
Reid Spencer7eb55b32006-11-02 01:53:59 +0000[diff] [blame]3935<h5>Example:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]3936<pre>
3937 &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]3938</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3939
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]3940</div>
Robert Bocchino820bc75b2006-02-17 21:18:08 +0000[diff] [blame]3941
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3942</div>
3943
Reid Spencer2ab01932007-02-02 13:57:07 +0000[diff] [blame]3944<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3945<h3>
3946 <a name="bitwiseops">Bitwise Binary Operations</a>
3947</h3>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3948
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3949<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3950
3951<p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3952 program. They are generally very efficient instructions and can commonly be
3953 strength reduced from other instructions. They require two operands of the
3954 same type, execute an operation on them, and produce a single value. The
3955 resulting value is the same type as its operands.</p>
3956
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]3957<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]3958<h4>
3959 <a name="i_shl">'<tt>shl</tt>' Instruction</a>
3960</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3961
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]3962<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3963
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]3964<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3965<pre>
Chris Lattnera676c0f2011-02-07 16:40:21 +0000[diff] [blame]3966 &lt;result&gt; = shl &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3967 &lt;result&gt; = shl nuw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3968 &lt;result&gt; = shl nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
3969 &lt;result&gt; = shl nuw nsw &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]3970</pre>
Chris Lattnerf0e50112007-10-03 21:01:14 +0000[diff] [blame]3971
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]3972<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3973<p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3974 a specified number of bits.</p>
Chris Lattnerf0e50112007-10-03 21:01:14 +0000[diff] [blame]3975
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]3976<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3977<p>Both arguments to the '<tt>shl</tt>' instruction must be the
3978 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3979 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]3980
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]3981<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3982<p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3983 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3984 is (statically or dynamically) negative or equal to or larger than the number
3985 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3986 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3987 shift amount in <tt>op2</tt>.</p>
Chris Lattnerf0e50112007-10-03 21:01:14 +0000[diff] [blame]3988
Chris Lattnera676c0f2011-02-07 16:40:21 +0000[diff] [blame]3989<p>If the <tt>nuw</tt> keyword is present, then the shift produces a
3990 <a href="#trapvalues">trap value</a> if it shifts out any non-zero bits. If
Chris Lattnerf10dfdc2011-02-09 16:44:44 +0000[diff] [blame]3991 the <tt>nsw</tt> keyword is present, then the shift produces a
Chris Lattnera676c0f2011-02-07 16:40:21 +0000[diff] [blame]3992 <a href="#trapvalues">trap value</a> if it shifts out any bits that disagree
3993 with the resultant sign bit. As such, NUW/NSW have the same semantics as
3994 they would if the shift were expressed as a mul instruction with the same
3995 nsw/nuw bits in (mul %op1, (shl 1, %op2)).</p>
3996
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]3997<h5>Example:</h5>
3998<pre>
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]3999 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
4000 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
4001 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerf0e50112007-10-03 21:01:14 +0000[diff] [blame]4002 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Mon P Wang4dd832d2008-12-09 05:46:39 +0000[diff] [blame]4003 &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 +0000[diff] [blame]4004</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4005
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]4006</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4007
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]4008<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4009<h4>
4010 <a name="i_lshr">'<tt>lshr</tt>' Instruction</a>
4011</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4012
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4013<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4014
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]4015<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4016<pre>
Chris Lattnera676c0f2011-02-07 16:40:21 +0000[diff] [blame]4017 &lt;result&gt; = lshr &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
4018 &lt;result&gt; = lshr exact &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]4019</pre>
4020
4021<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4022<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
4023 operand shifted to the right a specified number of bits with zero fill.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]4024
4025<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]4026<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4027 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4028 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]4029
4030<h5>Semantics:</h5>
4031<p>This instruction always performs a logical shift right operation. The most
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4032 significant bits of the result will be filled with zero bits after the shift.
4033 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
4034 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
4035 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
4036 shift amount in <tt>op2</tt>.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]4037
Chris Lattnera676c0f2011-02-07 16:40:21 +0000[diff] [blame]4038<p>If the <tt>exact</tt> keyword is present, the result value of the
4039 <tt>lshr</tt> is a <a href="#trapvalues">trap value</a> if any of the bits
4040 shifted out are non-zero.</p>
4041
4042
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]4043<h5>Example:</h5>
4044<pre>
4045 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
4046 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
4047 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
4048 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerf0e50112007-10-03 21:01:14 +0000[diff] [blame]4049 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Mon P Wang4dd832d2008-12-09 05:46:39 +0000[diff] [blame]4050 &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 +0000[diff] [blame]4051</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4052
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]4053</div>
4054
Reid Spencer2ab01932007-02-02 13:57:07 +0000[diff] [blame]4055<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4056<h4>
4057 <a name="i_ashr">'<tt>ashr</tt>' Instruction</a>
4058</h4>
4059
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4060<div>
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]4061
4062<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4063<pre>
Chris Lattnera676c0f2011-02-07 16:40:21 +0000[diff] [blame]4064 &lt;result&gt; = ashr &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
4065 &lt;result&gt; = ashr exact &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]4066</pre>
4067
4068<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4069<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
4070 operand shifted to the right a specified number of bits with sign
4071 extension.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]4072
4073<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]4074<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4075 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4076 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]4077
4078<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4079<p>This instruction always performs an arithmetic shift right operation, The
4080 most significant bits of the result will be filled with the sign bit
4081 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
4082 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
4083 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
4084 the corresponding shift amount in <tt>op2</tt>.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]4085
Chris Lattnera676c0f2011-02-07 16:40:21 +0000[diff] [blame]4086<p>If the <tt>exact</tt> keyword is present, the result value of the
4087 <tt>ashr</tt> is a <a href="#trapvalues">trap value</a> if any of the bits
4088 shifted out are non-zero.</p>
4089
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]4090<h5>Example:</h5>
4091<pre>
4092 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
4093 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
4094 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
4095 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerf0e50112007-10-03 21:01:14 +0000[diff] [blame]4096 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Mon P Wang4dd832d2008-12-09 05:46:39 +0000[diff] [blame]4097 &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 +0000[diff] [blame]4098</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4099
Reid Spencer04e259b2007-01-31 21:39:12 +0000[diff] [blame]4100</div>
4101
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4102<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4103<h4>
4104 <a name="i_and">'<tt>and</tt>' Instruction</a>
4105</h4>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]4106
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4107<div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]4108
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4109<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]4110<pre>
Gabor Greif0f75ad02008-08-07 21:46:00 +0000[diff] [blame]4111 &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 +0000[diff] [blame]4112</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]4113
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4114<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4115<p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
4116 operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]4117
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4118<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]4119<p>The two arguments to the '<tt>and</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4120 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4121 values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]4122
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4123<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]4124<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4125
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]4126<table border="1" cellspacing="0" cellpadding="4">
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]4127 <tbody>
4128 <tr>
4129 <td>In0</td>
4130 <td>In1</td>
4131 <td>Out</td>
4132 </tr>
4133 <tr>
4134 <td>0</td>
4135 <td>0</td>
4136 <td>0</td>
4137 </tr>
4138 <tr>
4139 <td>0</td>
4140 <td>1</td>
4141 <td>0</td>
4142 </tr>
4143 <tr>
4144 <td>1</td>
4145 <td>0</td>
4146 <td>0</td>
4147 </tr>
4148 <tr>
4149 <td>1</td>
4150 <td>1</td>
4151 <td>1</td>
4152 </tr>
4153 </tbody>
4154</table>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4155
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4156<h5>Example:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]4157<pre>
4158 &lt;result&gt; = and i32 4, %var <i>; yields {i32}:result = 4 &amp; %var</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]4159 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
4160 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4161</pre>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]4162</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4163<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4164<h4>
4165 <a name="i_or">'<tt>or</tt>' Instruction</a>
4166</h4>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]4167
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4168<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4169
4170<h5>Syntax:</h5>
4171<pre>
4172 &lt;result&gt; = or &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
4173</pre>
4174
4175<h5>Overview:</h5>
4176<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
4177 two operands.</p>
4178
4179<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]4180<p>The two arguments to the '<tt>or</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4181 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4182 values. Both arguments must have identical types.</p>
4183
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4184<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]4185<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4186
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]4187<table border="1" cellspacing="0" cellpadding="4">
4188 <tbody>
4189 <tr>
4190 <td>In0</td>
4191 <td>In1</td>
4192 <td>Out</td>
4193 </tr>
4194 <tr>
4195 <td>0</td>
4196 <td>0</td>
4197 <td>0</td>
4198 </tr>
4199 <tr>
4200 <td>0</td>
4201 <td>1</td>
4202 <td>1</td>
4203 </tr>
4204 <tr>
4205 <td>1</td>
4206 <td>0</td>
4207 <td>1</td>
4208 </tr>
4209 <tr>
4210 <td>1</td>
4211 <td>1</td>
4212 <td>1</td>
4213 </tr>
4214 </tbody>
4215</table>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4216
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4217<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4218<pre>
4219 &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]4220 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
4221 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4222</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4223
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]4224</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4225
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4226<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4227<h4>
4228 <a name="i_xor">'<tt>xor</tt>' Instruction</a>
4229</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4230
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4231<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4232
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4233<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4234<pre>
4235 &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 +0000[diff] [blame]4236</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4237
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4238<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4239<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
4240 its two operands. The <tt>xor</tt> is used to implement the "one's
4241 complement" operation, which is the "~" operator in C.</p>
4242
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4243<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]4244<p>The two arguments to the '<tt>xor</tt>' instruction must be
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4245 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4246 values. Both arguments must have identical types.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]4247
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4248<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]4249<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4250
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]4251<table border="1" cellspacing="0" cellpadding="4">
4252 <tbody>
4253 <tr>
4254 <td>In0</td>
4255 <td>In1</td>
4256 <td>Out</td>
4257 </tr>
4258 <tr>
4259 <td>0</td>
4260 <td>0</td>
4261 <td>0</td>
4262 </tr>
4263 <tr>
4264 <td>0</td>
4265 <td>1</td>
4266 <td>1</td>
4267 </tr>
4268 <tr>
4269 <td>1</td>
4270 <td>0</td>
4271 <td>1</td>
4272 </tr>
4273 <tr>
4274 <td>1</td>
4275 <td>1</td>
4276 <td>0</td>
4277 </tr>
4278 </tbody>
4279</table>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4280
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4281<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4282<pre>
4283 &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]4284 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
4285 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
4286 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4287</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4288
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]4289</div>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]4290
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4291</div>
4292
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4293<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4294<h3>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4295 <a name="vectorops">Vector Operations</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4296</h3>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4297
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4298<div>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4299
4300<p>LLVM supports several instructions to represent vector operations in a
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4301 target-independent manner. These instructions cover the element-access and
4302 vector-specific operations needed to process vectors effectively. While LLVM
4303 does directly support these vector operations, many sophisticated algorithms
4304 will want to use target-specific intrinsics to take full advantage of a
4305 specific target.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4306
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4307<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4308<h4>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4309 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4310</h4>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4311
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4312<div>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4313
4314<h5>Syntax:</h5>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4315<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]4316 &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 +0000[diff] [blame]4317</pre>
4318
4319<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4320<p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
4321 from a vector at a specified index.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4322
4323
4324<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4325<p>The first operand of an '<tt>extractelement</tt>' instruction is a value
4326 of <a href="#t_vector">vector</a> type. The second operand is an index
4327 indicating the position from which to extract the element. The index may be
4328 a variable.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4329
4330<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4331<p>The result is a scalar of the same type as the element type of
4332 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
4333 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4334 results are undefined.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4335
4336<h5>Example:</h5>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4337<pre>
Gabor Greif03ab4dc2009-10-28 13:14:50 +0000[diff] [blame]4338 &lt;result&gt; = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4339</pre>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4340
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4341</div>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4342
4343<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4344<h4>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4345 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4346</h4>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4347
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4348<div>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4349
4350<h5>Syntax:</h5>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4351<pre>
Dan Gohman43ba0672008-05-12 23:38:42 +0000[diff] [blame]4352 &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 +0000[diff] [blame]4353</pre>
4354
4355<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4356<p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
4357 vector at a specified index.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4358
4359<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4360<p>The first operand of an '<tt>insertelement</tt>' instruction is a value
4361 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
4362 whose type must equal the element type of the first operand. The third
4363 operand is an index indicating the position at which to insert the value.
4364 The index may be a variable.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4365
4366<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4367<p>The result is a vector of the same type as <tt>val</tt>. Its element values
4368 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
4369 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4370 results are undefined.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4371
4372<h5>Example:</h5>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4373<pre>
Gabor Greif03ab4dc2009-10-28 13:14:50 +0000[diff] [blame]4374 &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 +0000[diff] [blame]4375</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4376
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4377</div>
4378
4379<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4380<h4>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4381 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4382</h4>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4383
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4384<div>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4385
4386<h5>Syntax:</h5>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4387<pre>
Mon P Wang25f01062008-11-10 04:46:22 +0000[diff] [blame]4388 &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 +0000[diff] [blame]4389</pre>
4390
4391<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4392<p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
4393 from two input vectors, returning a vector with the same element type as the
4394 input and length that is the same as the shuffle mask.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4395
4396<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4397<p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
4398 with types that match each other. The third argument is a shuffle mask whose
4399 element type is always 'i32'. The result of the instruction is a vector
4400 whose length is the same as the shuffle mask and whose element type is the
4401 same as the element type of the first two operands.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4402
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4403<p>The shuffle mask operand is required to be a constant vector with either
4404 constant integer or undef values.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4405
4406<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4407<p>The elements of the two input vectors are numbered from left to right across
4408 both of the vectors. The shuffle mask operand specifies, for each element of
4409 the result vector, which element of the two input vectors the result element
4410 gets. The element selector may be undef (meaning "don't care") and the
4411 second operand may be undef if performing a shuffle from only one vector.</p>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4412
4413<h5>Example:</h5>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4414<pre>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]4415 &lt;result&gt; = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
Jeff Cohen5819f182007-04-22 01:17:39 +0000[diff] [blame]4416 &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 +0000[diff] [blame]4417 &lt;result&gt; = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]4418 &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 +0000[diff] [blame]4419 &lt;result&gt; = shufflevector &lt;8 x i32&gt; %v1, &lt;8 x i32&gt; undef,
Mon P Wang25f01062008-11-10 04:46:22 +0000[diff] [blame]4420 &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 +0000[diff] [blame]4421 &lt;result&gt; = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
Mon P Wang25f01062008-11-10 04:46:22 +0000[diff] [blame]4422 &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 +0000[diff] [blame]4423</pre>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4424
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4425</div>
Tanya Lattnerb138bbe2006-04-14 19:24:33 +0000[diff] [blame]4426
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4427</div>
4428
Chris Lattnerce83bff2006-04-08 23:07:04 +0000[diff] [blame]4429<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4430<h3>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4431 <a name="aggregateops">Aggregate Operations</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4432</h3>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4433
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4434<div>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4435
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]4436<p>LLVM supports several instructions for working with
4437 <a href="#t_aggregate">aggregate</a> values.</p>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4438
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4439<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4440<h4>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4441 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4442</h4>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4443
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4444<div>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4445
4446<h5>Syntax:</h5>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4447<pre>
4448 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
4449</pre>
4450
4451<h5>Overview:</h5>
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]4452<p>The '<tt>extractvalue</tt>' instruction extracts the value of a member field
4453 from an <a href="#t_aggregate">aggregate</a> value.</p>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4454
4455<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4456<p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
Chris Lattner13ee7952010-08-28 04:09:24 +0000[diff] [blame]4457 of <a href="#t_struct">struct</a> or
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]4458 <a href="#t_array">array</a> type. The operands are constant indices to
4459 specify which value to extract in a similar manner as indices in a
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4460 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
Frits van Bommel7cf63ac2010-12-05 20:54:38 +0000[diff] [blame]4461 <p>The major differences to <tt>getelementptr</tt> indexing are:</p>
4462 <ul>
4463 <li>Since the value being indexed is not a pointer, the first index is
4464 omitted and assumed to be zero.</li>
4465 <li>At least one index must be specified.</li>
4466 <li>Not only struct indices but also array indices must be in
4467 bounds.</li>
4468 </ul>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4469
4470<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4471<p>The result is the value at the position in the aggregate specified by the
4472 index operands.</p>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4473
4474<h5>Example:</h5>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4475<pre>
Gabor Greif03ab4dc2009-10-28 13:14:50 +0000[diff] [blame]4476 &lt;result&gt; = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4477</pre>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4478
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4479</div>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4480
4481<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4482<h4>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4483 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4484</h4>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4485
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4486<div>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4487
4488<h5>Syntax:</h5>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4489<pre>
Bill Wendlingf6a91cf2011-07-26 20:42:28 +0000[diff] [blame]4490 &lt;result&gt; = insertvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;ty&gt; &lt;elt&gt;, &lt;idx&gt;{, &lt;idx&gt;}* <i>; yields &lt;aggregate type&gt;</i>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4491</pre>
4492
4493<h5>Overview:</h5>
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]4494<p>The '<tt>insertvalue</tt>' instruction inserts a value into a member field
4495 in an <a href="#t_aggregate">aggregate</a> value.</p>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4496
4497<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4498<p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
Chris Lattner13ee7952010-08-28 04:09:24 +0000[diff] [blame]4499 of <a href="#t_struct">struct</a> or
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]4500 <a href="#t_array">array</a> type. The second operand is a first-class
4501 value to insert. The following operands are constant indices indicating
4502 the position at which to insert the value in a similar manner as indices in a
Frits van Bommel7cf63ac2010-12-05 20:54:38 +0000[diff] [blame]4503 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' instruction. The
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4504 value to insert must have the same type as the value identified by the
4505 indices.</p>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4506
4507<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4508<p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
4509 that of <tt>val</tt> except that the value at the position specified by the
4510 indices is that of <tt>elt</tt>.</p>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4511
4512<h5>Example:</h5>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4513<pre>
Chris Lattnerc2e85402011-05-22 07:18:08 +0000[diff] [blame]4514 %agg1 = insertvalue {i32, float} undef, i32 1, 0 <i>; yields {i32 1, float undef}</i>
4515 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 <i>; yields {i32 1, float %val}</i>
4516 %agg3 = insertvalue {i32, {float}} %agg1, float %val, 1, 0 <i>; yields {i32 1, float %val}</i>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4517</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4518
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4519</div>
4520
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4521</div>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000[diff] [blame]4522
4523<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4524<h3>
Chris Lattner6ab66722006-08-15 00:45:58 +0000[diff] [blame]4525 <a name="memoryops">Memory Access and Addressing Operations</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4526</h3>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]4527
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4528<div>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]4529
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4530<p>A key design point of an SSA-based representation is how it represents
4531 memory. In LLVM, no memory locations are in SSA form, which makes things
Victor Hernandeza70c6df2009-10-26 23:44:29 +0000[diff] [blame]4532 very simple. This section describes how to read, write, and allocate
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4533 memory in LLVM.</p>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]4534
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4535<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4536<h4>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]4537 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4538</h4>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]4539
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4540<div>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]4541
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4542<h5>Syntax:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]4543<pre>
Dan Gohman2140a742010-05-28 01:14:11 +0000[diff] [blame]4544 &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 +0000[diff] [blame]4545</pre>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]4546
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4547<h5>Overview:</h5>
Jeff Cohen5819f182007-04-22 01:17:39 +0000[diff] [blame]4548<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4549 currently executing function, to be automatically released when this function
4550 returns to its caller. The object is always allocated in the generic address
4551 space (address space zero).</p>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]4552
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4553<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4554<p>The '<tt>alloca</tt>' instruction
4555 allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt> bytes of memory on the
4556 runtime stack, returning a pointer of the appropriate type to the program.
4557 If "NumElements" is specified, it is the number of elements allocated,
4558 otherwise "NumElements" is defaulted to be one. If a constant alignment is
4559 specified, the value result of the allocation is guaranteed to be aligned to
4560 at least that boundary. If not specified, or if zero, the target can choose
4561 to align the allocation on any convenient boundary compatible with the
4562 type.</p>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]4563
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]4564<p>'<tt>type</tt>' may be any sized type.</p>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]4565
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4566<h5>Semantics:</h5>
Bill Wendling9ee6a312009-05-08 20:49:29 +0000[diff] [blame]4567<p>Memory is allocated; a pointer is returned. The operation is undefined if
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4568 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
4569 memory is automatically released when the function returns. The
4570 '<tt>alloca</tt>' instruction is commonly used to represent automatic
4571 variables that must have an address available. When the function returns
4572 (either with the <tt><a href="#i_ret">ret</a></tt>
4573 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
4574 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]4575
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4576<h5>Example:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]4577<pre>
Dan Gohman7a5acb52009-01-04 23:49:44 +0000[diff] [blame]4578 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
4579 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
4580 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
4581 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4582</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4583
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]4584</div>
Chris Lattner54611b42005-11-06 08:02:57 +0000[diff] [blame]4585
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]4586<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4587<h4>
4588 <a name="i_load">'<tt>load</tt>' Instruction</a>
4589</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4590
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4591<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4592
Chris Lattner095735d2002-05-06 03:03:22 +0000[diff] [blame]4593<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4594<pre>
Eli Friedman02e737b2011-08-12 22:50:01 +0000[diff] [blame]4595 &lt;result&gt; = load [volatile] &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;][, !nontemporal !&lt;index&gt;]
4596 &lt;result&gt; = load atomic [volatile] &lt;ty&gt;* &lt;pointer&gt; [singlethread] &lt;ordering&gt;, align &lt;alignment&gt;
Bill Wendlingbc4024f2010-02-25 21:23:24 +0000[diff] [blame]4597 !&lt;index&gt; = !{ i32 1 }
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4598</pre>
4599
Chris Lattner095735d2002-05-06 03:03:22 +0000[diff] [blame]4600<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]4601<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4602
Chris Lattner095735d2002-05-06 03:03:22 +0000[diff] [blame]4603<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4604<p>The argument to the '<tt>load</tt>' instruction specifies the memory address
4605 from which to load. The pointer must point to
4606 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
4607 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +0000[diff] [blame]4608 number or order of execution of this <tt>load</tt> with other <a
4609 href="#volatile">volatile operations</a>.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4610
Eli Friedman59b66882011-08-09 23:02:53 +0000[diff] [blame]4611<p>If the <code>load</code> is marked as <code>atomic</code>, it takes an extra
4612 <a href="#ordering">ordering</a> and optional <code>singlethread</code>
4613 argument. The <code>release</code> and <code>acq_rel</code> orderings are
4614 not valid on <code>load</code> instructions. Atomic loads produce <a
4615 href="#memorymodel">defined</a> results when they may see multiple atomic
4616 stores. The type of the pointee must be an integer type whose bit width
4617 is a power of two greater than or equal to eight and less than or equal
4618 to a target-specific size limit. <code>align</code> must be explicitly
4619 specified on atomic loads, and the load has undefined behavior if the
4620 alignment is not set to a value which is at least the size in bytes of
4621 the pointee. <code>!nontemporal</code> does not have any defined semantics
4622 for atomic loads.</p>
4623
Bill Wendlingbc4024f2010-02-25 21:23:24 +0000[diff] [blame]4624<p>The optional constant <tt>align</tt> argument specifies the alignment of the
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4625 operation (that is, the alignment of the memory address). A value of 0 or an
Bill Wendlingbc4024f2010-02-25 21:23:24 +0000[diff] [blame]4626 omitted <tt>align</tt> argument means that the operation has the preferential
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4627 alignment for the target. It is the responsibility of the code emitter to
4628 ensure that the alignment information is correct. Overestimating the
Bill Wendlingbc4024f2010-02-25 21:23:24 +0000[diff] [blame]4629 alignment results in undefined behavior. Underestimating the alignment may
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4630 produce less efficient code. An alignment of 1 is always safe.</p>
4631
Bill Wendlingbc4024f2010-02-25 21:23:24 +0000[diff] [blame]4632<p>The optional <tt>!nontemporal</tt> metadata must reference a single
4633 metatadata name &lt;index&gt; corresponding to a metadata node with
Dan Gohmana269a0a2010-03-01 17:41:39 +0000[diff] [blame]4634 one <tt>i32</tt> entry of value 1. The existence of
Bill Wendlingbc4024f2010-02-25 21:23:24 +0000[diff] [blame]4635 the <tt>!nontemporal</tt> metatadata on the instruction tells the optimizer
4636 and code generator that this load is not expected to be reused in the cache.
4637 The code generator may select special instructions to save cache bandwidth,
Dan Gohmana269a0a2010-03-01 17:41:39 +0000[diff] [blame]4638 such as the <tt>MOVNT</tt> instruction on x86.</p>
David Greene9641d062010-02-16 20:50:18 +0000[diff] [blame]4639
Chris Lattner095735d2002-05-06 03:03:22 +0000[diff] [blame]4640<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4641<p>The location of memory pointed to is loaded. If the value being loaded is of
4642 scalar type then the number of bytes read does not exceed the minimum number
4643 of bytes needed to hold all bits of the type. For example, loading an
4644 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
4645 <tt>i20</tt> with a size that is not an integral number of bytes, the result
4646 is undefined if the value was not originally written using a store of the
4647 same type.</p>
4648
Chris Lattner095735d2002-05-06 03:03:22 +0000[diff] [blame]4649<h5>Examples:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4650<pre>
4651 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4652 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]4653 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
Chris Lattner095735d2002-05-06 03:03:22 +0000[diff] [blame]4654</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4655
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]4656</div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4657
Chris Lattner095735d2002-05-06 03:03:22 +0000[diff] [blame]4658<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4659<h4>
4660 <a name="i_store">'<tt>store</tt>' Instruction</a>
4661</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4662
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4663<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4664
Chris Lattner095735d2002-05-06 03:03:22 +0000[diff] [blame]4665<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4666<pre>
Eli Friedman02e737b2011-08-12 22:50:01 +0000[diff] [blame]4667 store [volatile] &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;][, !nontemporal !&lt;index&gt;] <i>; yields {void}</i>
4668 store atomic [volatile] &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt; [singlethread] &lt;ordering&gt;, align &lt;alignment&gt; <i>; yields {void}</i>
Chris Lattner095735d2002-05-06 03:03:22 +0000[diff] [blame]4669</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4670
Chris Lattner095735d2002-05-06 03:03:22 +0000[diff] [blame]4671<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]4672<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4673
Chris Lattner095735d2002-05-06 03:03:22 +0000[diff] [blame]4674<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4675<p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
4676 and an address at which to store it. The type of the
4677 '<tt>&lt;pointer&gt;</tt>' operand must be a pointer to
4678 the <a href="#t_firstclass">first class</a> type of the
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +0000[diff] [blame]4679 '<tt>&lt;value&gt;</tt>' operand. If the <tt>store</tt> is marked as
4680 <tt>volatile</tt>, then the optimizer is not allowed to modify the number or
4681 order of execution of this <tt>store</tt> with other <a
4682 href="#volatile">volatile operations</a>.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4683
Eli Friedman59b66882011-08-09 23:02:53 +0000[diff] [blame]4684<p>If the <code>store</code> is marked as <code>atomic</code>, it takes an extra
4685 <a href="#ordering">ordering</a> and optional <code>singlethread</code>
4686 argument. The <code>acquire</code> and <code>acq_rel</code> orderings aren't
4687 valid on <code>store</code> instructions. Atomic loads produce <a
4688 href="#memorymodel">defined</a> results when they may see multiple atomic
4689 stores. The type of the pointee must be an integer type whose bit width
4690 is a power of two greater than or equal to eight and less than or equal
4691 to a target-specific size limit. <code>align</code> must be explicitly
4692 specified on atomic stores, and the store has undefined behavior if the
4693 alignment is not set to a value which is at least the size in bytes of
4694 the pointee. <code>!nontemporal</code> does not have any defined semantics
4695 for atomic stores.</p>
4696
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4697<p>The optional constant "align" argument specifies the alignment of the
4698 operation (that is, the alignment of the memory address). A value of 0 or an
4699 omitted "align" argument means that the operation has the preferential
4700 alignment for the target. It is the responsibility of the code emitter to
4701 ensure that the alignment information is correct. Overestimating the
4702 alignment results in an undefined behavior. Underestimating the alignment may
4703 produce less efficient code. An alignment of 1 is always safe.</p>
4704
David Greene9641d062010-02-16 20:50:18 +0000[diff] [blame]4705<p>The optional !nontemporal metadata must reference a single metatadata
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]4706 name &lt;index&gt; corresponding to a metadata node with one i32 entry of
Dan Gohmana269a0a2010-03-01 17:41:39 +0000[diff] [blame]4707 value 1. The existence of the !nontemporal metatadata on the
David Greene9641d062010-02-16 20:50:18 +0000[diff] [blame]4708 instruction tells the optimizer and code generator that this load is
4709 not expected to be reused in the cache. The code generator may
4710 select special instructions to save cache bandwidth, such as the
Dan Gohmana269a0a2010-03-01 17:41:39 +0000[diff] [blame]4711 MOVNT instruction on x86.</p>
David Greene9641d062010-02-16 20:50:18 +0000[diff] [blame]4712
4713
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]4714<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4715<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>' at the
4716 location specified by the '<tt>&lt;pointer&gt;</tt>' operand. If
4717 '<tt>&lt;value&gt;</tt>' is of scalar type then the number of bytes written
4718 does not exceed the minimum number of bytes needed to hold all bits of the
4719 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
4720 writing a value of a type like <tt>i20</tt> with a size that is not an
4721 integral number of bytes, it is unspecified what happens to the extra bits
4722 that do not belong to the type, but they will typically be overwritten.</p>
4723
Chris Lattner095735d2002-05-06 03:03:22 +0000[diff] [blame]4724<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4725<pre>
4726 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling8830ffe2007-10-22 05:10:05 +0000[diff] [blame]4727 store i32 3, i32* %ptr <i>; yields {void}</i>
4728 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
Chris Lattner095735d2002-05-06 03:03:22 +0000[diff] [blame]4729</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4730
Reid Spencer443460a2006-11-09 21:15:49 +0000[diff] [blame]4731</div>
4732
Chris Lattner095735d2002-05-06 03:03:22 +0000[diff] [blame]4733<!-- _______________________________________________________________________ -->
NAKAMURA Takumi0300d882011-08-12 06:17:17 +0000[diff] [blame]4734<h4>
4735<a name="i_fence">'<tt>fence</tt>' Instruction</a>
4736</h4>
Eli Friedmanfee02c62011-07-25 23:16:38 +0000[diff] [blame]4737
NAKAMURA Takumi0300d882011-08-12 06:17:17 +0000[diff] [blame]4738<div>
Eli Friedmanfee02c62011-07-25 23:16:38 +0000[diff] [blame]4739
4740<h5>Syntax:</h5>
4741<pre>
4742 fence [singlethread] &lt;ordering&gt; <i>; yields {void}</i>
4743</pre>
4744
4745<h5>Overview:</h5>
4746<p>The '<tt>fence</tt>' instruction is used to introduce happens-before edges
4747between operations.</p>
4748
4749<h5>Arguments:</h5> <p>'<code>fence</code>' instructions take an <a
4750href="#ordering">ordering</a> argument which defines what
4751<i>synchronizes-with</i> edges they add. They can only be given
4752<code>acquire</code>, <code>release</code>, <code>acq_rel</code>, and
4753<code>seq_cst</code> orderings.</p>
4754
4755<h5>Semantics:</h5>
4756<p>A fence <var>A</var> which has (at least) <code>release</code> ordering
4757semantics <i>synchronizes with</i> a fence <var>B</var> with (at least)
4758<code>acquire</code> ordering semantics if and only if there exist atomic
4759operations <var>X</var> and <var>Y</var>, both operating on some atomic object
4760<var>M</var>, such that <var>A</var> is sequenced before <var>X</var>,
4761<var>X</var> modifies <var>M</var> (either directly or through some side effect
4762of a sequence headed by <var>X</var>), <var>Y</var> is sequenced before
4763<var>B</var>, and <var>Y</var> observes <var>M</var>. This provides a
4764<i>happens-before</i> dependency between <var>A</var> and <var>B</var>. Rather
4765than an explicit <code>fence</code>, one (but not both) of the atomic operations
4766<var>X</var> or <var>Y</var> might provide a <code>release</code> or
4767<code>acquire</code> (resp.) ordering constraint and still
4768<i>synchronize-with</i> the explicit <code>fence</code> and establish the
4769<i>happens-before</i> edge.</p>
4770
4771<p>A <code>fence</code> which has <code>seq_cst</code> ordering, in addition to
4772having both <code>acquire</code> and <code>release</code> semantics specified
4773above, participates in the global program order of other <code>seq_cst</code>
4774operations and/or fences.</p>
4775
4776<p>The optional "<a href="#singlethread"><code>singlethread</code></a>" argument
4777specifies that the fence only synchronizes with other fences in the same
4778thread. (This is useful for interacting with signal handlers.)</p>
4779
Eli Friedmanfee02c62011-07-25 23:16:38 +0000[diff] [blame]4780<h5>Example:</h5>
4781<pre>
4782 fence acquire <i>; yields {void}</i>
4783 fence singlethread seq_cst <i>; yields {void}</i>
4784</pre>
4785
4786</div>
4787
4788<!-- _______________________________________________________________________ -->
NAKAMURA Takumi0300d882011-08-12 06:17:17 +0000[diff] [blame]4789<h4>
4790<a name="i_cmpxchg">'<tt>cmpxchg</tt>' Instruction</a>
4791</h4>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]4792
NAKAMURA Takumi0300d882011-08-12 06:17:17 +0000[diff] [blame]4793<div>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]4794
4795<h5>Syntax:</h5>
4796<pre>
Eli Friedman02e737b2011-08-12 22:50:01 +0000[diff] [blame]4797 cmpxchg [volatile] &lt;ty&gt;* &lt;pointer&gt;, &lt;ty&gt; &lt;cmp&gt;, &lt;ty&gt; &lt;new&gt; [singlethread] &lt;ordering&gt; <i>; yields {ty}</i>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]4798</pre>
4799
4800<h5>Overview:</h5>
4801<p>The '<tt>cmpxchg</tt>' instruction is used to atomically modify memory.
4802It loads a value in memory and compares it to a given value. If they are
4803equal, it stores a new value into the memory.</p>
4804
4805<h5>Arguments:</h5>
4806<p>There are three arguments to the '<code>cmpxchg</code>' instruction: an
4807address to operate on, a value to compare to the value currently be at that
4808address, and a new value to place at that address if the compared values are
4809equal. The type of '<var>&lt;cmp&gt;</var>' must be an integer type whose
4810bit width is a power of two greater than or equal to eight and less than
4811or equal to a target-specific size limit. '<var>&lt;cmp&gt;</var>' and
4812'<var>&lt;new&gt;</var>' must have the same type, and the type of
4813'<var>&lt;pointer&gt;</var>' must be a pointer to that type. If the
4814<code>cmpxchg</code> is marked as <code>volatile</code>, then the
4815optimizer is not allowed to modify the number or order of execution
4816of this <code>cmpxchg</code> with other <a href="#volatile">volatile
4817operations</a>.</p>
4818
4819<!-- FIXME: Extend allowed types. -->
4820
4821<p>The <a href="#ordering"><var>ordering</var></a> argument specifies how this
4822<code>cmpxchg</code> synchronizes with other atomic operations.</p>
4823
4824<p>The optional "<code>singlethread</code>" argument declares that the
4825<code>cmpxchg</code> is only atomic with respect to code (usually signal
4826handlers) running in the same thread as the <code>cmpxchg</code>. Otherwise the
4827cmpxchg is atomic with respect to all other code in the system.</p>
4828
4829<p>The pointer passed into cmpxchg must have alignment greater than or equal to
4830the size in memory of the operand.
4831
4832<h5>Semantics:</h5>
4833<p>The contents of memory at the location specified by the
4834'<tt>&lt;pointer&gt;</tt>' operand is read and compared to
4835'<tt>&lt;cmp&gt;</tt>'; if the read value is the equal,
4836'<tt>&lt;new&gt;</tt>' is written. The original value at the location
4837is returned.
4838
4839<p>A successful <code>cmpxchg</code> is a read-modify-write instruction for the
4840purpose of identifying <a href="#release_sequence">release sequences</a>. A
4841failed <code>cmpxchg</code> is equivalent to an atomic load with an ordering
4842parameter determined by dropping any <code>release</code> part of the
4843<code>cmpxchg</code>'s ordering.</p>
4844
4845<!--
4846FIXME: Is compare_exchange_weak() necessary? (Consider after we've done
4847optimization work on ARM.)
4848
4849FIXME: Is a weaker ordering constraint on failure helpful in practice?
4850-->
4851
4852<h5>Example:</h5>
4853<pre>
4854entry:
4855 %orig = atomic <a href="#i_load">load</a> i32* %ptr unordered <i>; yields {i32}</i>
4856 <a href="#i_br">br</a> label %loop
4857
4858loop:
4859 %cmp = <a href="#i_phi">phi</a> i32 [ %orig, %entry ], [%old, %loop]
4860 %squared = <a href="#i_mul">mul</a> i32 %cmp, %cmp
4861 %old = cmpxchg i32* %ptr, i32 %cmp, i32 %squared <i>; yields {i32}</i>
4862 %success = <a href="#i_icmp">icmp</a> eq i32 %cmp, %old
4863 <a href="#i_br">br</a> i1 %success, label %done, label %loop
4864
4865done:
4866 ...
4867</pre>
4868
4869</div>
4870
4871<!-- _______________________________________________________________________ -->
NAKAMURA Takumi0300d882011-08-12 06:17:17 +0000[diff] [blame]4872<h4>
4873<a name="i_atomicrmw">'<tt>atomicrmw</tt>' Instruction</a>
4874</h4>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]4875
NAKAMURA Takumi0300d882011-08-12 06:17:17 +0000[diff] [blame]4876<div>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]4877
4878<h5>Syntax:</h5>
4879<pre>
Eli Friedman02e737b2011-08-12 22:50:01 +0000[diff] [blame]4880 atomicrmw [volatile] &lt;operation&gt; &lt;ty&gt;* &lt;pointer&gt;, &lt;ty&gt; &lt;value&gt; [singlethread] &lt;ordering&gt; <i>; yields {ty}</i>
Eli Friedmanc9a551e2011-07-28 21:48:00 +0000[diff] [blame]4881</pre>
4882
4883<h5>Overview:</h5>
4884<p>The '<tt>atomicrmw</tt>' instruction is used to atomically modify memory.</p>
4885
4886<h5>Arguments:</h5>
4887<p>There are three arguments to the '<code>atomicrmw</code>' instruction: an
4888operation to apply, an address whose value to modify, an argument to the
4889operation. The operation must be one of the following keywords:</p>
4890<ul>
4891 <li>xchg</li>
4892 <li>add</li>
4893 <li>sub</li>
4894 <li>and</li>
4895 <li>nand</li>
4896 <li>or</li>
4897 <li>xor</li>
4898 <li>max</li>
4899 <li>min</li>
4900 <li>umax</li>
4901 <li>umin</li>
4902</ul>
4903
4904<p>The type of '<var>&lt;value&gt;</var>' must be an integer type whose
4905bit width is a power of two greater than or equal to eight and less than
4906or equal to a target-specific size limit. The type of the
4907'<code>&lt;pointer&gt;</code>' operand must be a pointer to that type.
4908If the <code>atomicrmw</code> is marked as <code>volatile</code>, then the
4909optimizer is not allowed to modify the number or order of execution of this
4910<code>atomicrmw</code> with other <a href="#volatile">volatile
4911 operations</a>.</p>
4912
4913<!-- FIXME: Extend allowed types. -->
4914
4915<h5>Semantics:</h5>
4916<p>The contents of memory at the location specified by the
4917'<tt>&lt;pointer&gt;</tt>' operand are atomically read, modified, and written
4918back. The original value at the location is returned. The modification is
4919specified by the <var>operation</var> argument:</p>
4920
4921<ul>
4922 <li>xchg: <code>*ptr = val</code></li>
4923 <li>add: <code>*ptr = *ptr + val</code></li>
4924 <li>sub: <code>*ptr = *ptr - val</code></li>
4925 <li>and: <code>*ptr = *ptr &amp; val</code></li>
4926 <li>nand: <code>*ptr = ~(*ptr &amp; val)</code></li>
4927 <li>or: <code>*ptr = *ptr | val</code></li>
4928 <li>xor: <code>*ptr = *ptr ^ val</code></li>
4929 <li>max: <code>*ptr = *ptr &gt; val ? *ptr : val</code> (using a signed comparison)</li>
4930 <li>min: <code>*ptr = *ptr &lt; val ? *ptr : val</code> (using a signed comparison)</li>
4931 <li>umax: <code>*ptr = *ptr &gt; val ? *ptr : val</code> (using an unsigned comparison)</li>
4932 <li>umin: <code>*ptr = *ptr &lt; val ? *ptr : val</code> (using an unsigned comparison)</li>
4933</ul>
4934
4935<h5>Example:</h5>
4936<pre>
4937 %old = atomicrmw add i32* %ptr, i32 1 acquire <i>; yields {i32}</i>
4938</pre>
4939
4940</div>
4941
4942<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4943<h4>
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]4944 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]4945</h4>
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]4946
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]4947<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4948
Chris Lattner590645f2002-04-14 06:13:44 +0000[diff] [blame]4949<h5>Syntax:</h5>
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]4950<pre>
Matthijs Kooijman0e268272008-10-13 13:44:15 +0000[diff] [blame]4951 &lt;result&gt; = getelementptr &lt;pty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
Dan Gohman1639c392009-07-27 21:53:46 +0000[diff] [blame]4952 &lt;result&gt; = getelementptr inbounds &lt;pty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]4953</pre>
4954
Chris Lattner590645f2002-04-14 06:13:44 +0000[diff] [blame]4955<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4956<p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]4957 subelement of an <a href="#t_aggregate">aggregate</a> data structure.
4958 It performs address calculation only and does not access memory.</p>
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]4959
Chris Lattner590645f2002-04-14 06:13:44 +0000[diff] [blame]4960<h5>Arguments:</h5>
Matthijs Kooijman0e268272008-10-13 13:44:15 +0000[diff] [blame]4961<p>The first argument is always a pointer, and forms the basis of the
Chris Lattnera40b9122009-07-29 06:44:13 +0000[diff] [blame]4962 calculation. The remaining arguments are indices that indicate which of the
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4963 elements of the aggregate object are indexed. The interpretation of each
4964 index is dependent on the type being indexed into. The first index always
4965 indexes the pointer value given as the first argument, the second index
4966 indexes a value of the type pointed to (not necessarily the value directly
4967 pointed to, since the first index can be non-zero), etc. The first type
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]4968 indexed into must be a pointer value, subsequent types can be arrays,
Chris Lattner13ee7952010-08-28 04:09:24 +0000[diff] [blame]4969 vectors, and structs. Note that subsequent types being indexed into
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]4970 can never be pointers, since that would require loading the pointer before
4971 continuing calculation.</p>
Matthijs Kooijman0e268272008-10-13 13:44:15 +0000[diff] [blame]4972
4973<p>The type of each index argument depends on the type it is indexing into.
Chris Lattner13ee7952010-08-28 04:09:24 +0000[diff] [blame]4974 When indexing into a (optionally packed) structure, only <tt>i32</tt>
Chris Lattner392be582010-02-12 20:49:41 +0000[diff] [blame]4975 integer <b>constants</b> are allowed. When indexing into an array, pointer
4976 or vector, integers of any width are allowed, and they are not required to be
Eli Friedmand8874dc2011-08-12 23:37:55 +0000[diff] [blame]4977 constant. These integers are treated as signed values where relevant.</p>
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]4978
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]4979<p>For example, let's consider a C code fragment and how it gets compiled to
4980 LLVM:</p>
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]4981
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]4982<pre class="doc_code">
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]4983struct RT {
4984 char A;
Chris Lattnera446f1b2007-05-29 15:43:56 +0000[diff] [blame]4985 int B[10][20];
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]4986 char C;
4987};
4988struct ST {
Chris Lattnera446f1b2007-05-29 15:43:56 +0000[diff] [blame]4989 int X;
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]4990 double Y;
4991 struct RT Z;
4992};
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]4993
Chris Lattnera446f1b2007-05-29 15:43:56 +0000[diff] [blame]4994int *foo(struct ST *s) {
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]4995 return &amp;s[1].Z.B[5][13];
4996}
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]4997</pre>
4998
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]4999<p>The LLVM code generated by the GCC frontend is:</p>
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]5000
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]5001<pre class="doc_code">
Chris Lattnerbc088212009-01-11 20:53:49 +0000[diff] [blame]5002%RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
5003%ST = <a href="#namedtypes">type</a> { i32, double, %RT }
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]5004
Dan Gohman6b867702009-07-25 02:23:48 +0000[diff] [blame]5005define i32* @foo(%ST* %s) {
Bill Wendling3716c5d2007-05-29 09:04:49 +0000[diff] [blame]5006entry:
5007 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
5008 ret i32* %reg
5009}
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]5010</pre>
5011
Chris Lattner590645f2002-04-14 06:13:44 +0000[diff] [blame]5012<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]5013<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5014 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
5015 }</tt>' type, a structure. The second index indexes into the third element
5016 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
5017 i8 }</tt>' type, another structure. The third index indexes into the second
5018 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
5019 array. The two dimensions of the array are subscripted into, yielding an
5020 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
5021 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]5022
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5023<p>Note that it is perfectly legal to index partially through a structure,
5024 returning a pointer to an inner element. Because of this, the LLVM code for
5025 the given testcase is equivalent to:</p>
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]5026
5027<pre>
Dan Gohman6b867702009-07-25 02:23:48 +0000[diff] [blame]5028 define i32* @foo(%ST* %s) {
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]5029 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
Jeff Cohen5819f182007-04-22 01:17:39 +0000[diff] [blame]5030 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
5031 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]5032 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
5033 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
5034 ret i32* %t5
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]5035 }
Chris Lattnera8292f32002-05-06 22:08:29 +0000[diff] [blame]5036</pre>
Chris Lattnerc0ad71e2005-06-24 17:22:57 +0000[diff] [blame]5037
Dan Gohman1639c392009-07-27 21:53:46 +0000[diff] [blame]5038<p>If the <tt>inbounds</tt> keyword is present, the result value of the
Dan Gohman57255802010-04-23 15:23:32 +0000[diff] [blame]5039 <tt>getelementptr</tt> is a <a href="#trapvalues">trap value</a> if the
5040 base pointer is not an <i>in bounds</i> address of an allocated object,
5041 or if any of the addresses that would be formed by successive addition of
5042 the offsets implied by the indices to the base address with infinitely
Eli Friedmand8874dc2011-08-12 23:37:55 +0000[diff] [blame]5043 precise signed arithmetic are not an <i>in bounds</i> address of that
5044 allocated object. The <i>in bounds</i> addresses for an allocated object
5045 are all the addresses that point into the object, plus the address one
5046 byte past the end.</p>
Dan Gohman1639c392009-07-27 21:53:46 +0000[diff] [blame]5047
5048<p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
Eli Friedmand8874dc2011-08-12 23:37:55 +0000[diff] [blame]5049 the base address with silently-wrapping two's complement arithmetic. If the
5050 offsets have a different width from the pointer, they are sign-extended or
5051 truncated to the width of the pointer. The result value of the
5052 <tt>getelementptr</tt> may be outside the object pointed to by the base
5053 pointer. The result value may not necessarily be used to access memory
5054 though, even if it happens to point into allocated storage. See the
5055 <a href="#pointeraliasing">Pointer Aliasing Rules</a> section for more
5056 information.</p>
Dan Gohman1639c392009-07-27 21:53:46 +0000[diff] [blame]5057
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5058<p>The getelementptr instruction is often confusing. For some more insight into
5059 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
Chris Lattner6ab66722006-08-15 00:45:58 +0000[diff] [blame]5060
Chris Lattner590645f2002-04-14 06:13:44 +0000[diff] [blame]5061<h5>Example:</h5>
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]5062<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]5063 <i>; yields [12 x i8]*:aptr</i>
Matthijs Kooijman0e268272008-10-13 13:44:15 +0000[diff] [blame]5064 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
5065 <i>; yields i8*:vptr</i>
Dan Gohmanef9462f2008-10-14 16:51:45 +0000[diff] [blame]5066 %vptr = getelementptr {i32, &lt;2 x i8&gt;}* %svptr, i64 0, i32 1, i32 1
Matthijs Kooijman0e268272008-10-13 13:44:15 +0000[diff] [blame]5067 <i>; yields i8*:eptr</i>
5068 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
Sanjiv Gupta0c155e62009-04-25 07:27:44 +0000[diff] [blame]5069 <i>; yields i32*:iptr</i>
Sanjiv Gupta77abea02009-04-24 16:38:13 +0000[diff] [blame]5070 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]5071</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5072
Chris Lattner33fd7022004-04-05 01:30:49 +0000[diff] [blame]5073</div>
Reid Spencer443460a2006-11-09 21:15:49 +0000[diff] [blame]5074
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5075</div>
5076
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]5077<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5078<h3>
5079 <a name="convertops">Conversion Operations</a>
5080</h3>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5081
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5082<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5083
Reid Spencer97c5fa42006-11-08 01:18:52 +0000[diff] [blame]5084<p>The instructions in this category are the conversion instructions (casting)
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5085 which all take a single operand and a type. They perform various bit
5086 conversions on the operand.</p>
5087
Chris Lattnera8292f32002-05-06 22:08:29 +0000[diff] [blame]5088<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5089<h4>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5090 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5091</h4>
5092
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5093<div>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5094
5095<h5>Syntax:</h5>
5096<pre>
5097 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
5098</pre>
5099
5100<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5101<p>The '<tt>trunc</tt>' instruction truncates its operand to the
5102 type <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5103
5104<h5>Arguments:</h5>
Nadav Rotem502f1b92011-02-24 21:01:34 +0000[diff] [blame]5105<p>The '<tt>trunc</tt>' instruction takes a value to trunc, and a type to trunc it to.
5106 Both types must be of <a href="#t_integer">integer</a> types, or vectors
5107 of the same number of integers.
5108 The bit size of the <tt>value</tt> must be larger than
5109 the bit size of the destination type, <tt>ty2</tt>.
5110 Equal sized types are not allowed.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5111
5112<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5113<p>The '<tt>trunc</tt>' instruction truncates the high order bits
5114 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
5115 source size must be larger than the destination size, <tt>trunc</tt> cannot
5116 be a <i>no-op cast</i>. It will always truncate bits.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5117
5118<h5>Example:</h5>
5119<pre>
Nadav Rotem502f1b92011-02-24 21:01:34 +0000[diff] [blame]5120 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
5121 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
5122 %Z = trunc i32 122 to i1 <i>; yields i1:false</i>
5123 %W = trunc &lt;2 x i16&gt; &lt;i16 8, i16 7&gt; to &lt;2 x i8&gt; <i>; yields &lt;i8 8, i8 7&gt;</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5124</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5125
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5126</div>
5127
5128<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5129<h4>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5130 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5131</h4>
5132
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5133<div>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5134
5135<h5>Syntax:</h5>
5136<pre>
5137 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
5138</pre>
5139
5140<h5>Overview:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5141<p>The '<tt>zext</tt>' instruction zero extends its operand to type
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5142 <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5143
5144
5145<h5>Arguments:</h5>
Nadav Rotem25f2ac92011-02-20 12:37:50 +0000[diff] [blame]5146<p>The '<tt>zext</tt>' instruction takes a value to cast, and a type to cast it to.
5147 Both types must be of <a href="#t_integer">integer</a> types, or vectors
5148 of the same number of integers.
5149 The bit size of the <tt>value</tt> must be smaller than
5150 the bit size of the destination type,
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5151 <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5152
5153<h5>Semantics:</h5>
5154<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5155 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5156
Reid Spencer07c9c682007-01-12 15:46:11 +0000[diff] [blame]5157<p>When zero extending from i1, the result will always be either 0 or 1.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5158
5159<h5>Example:</h5>
5160<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]5161 %X = zext i32 257 to i64 <i>; yields i64:257</i>
Reid Spencer36a15422007-01-12 03:35:51 +0000[diff] [blame]5162 %Y = zext i1 true to i32 <i>; yields i32:1</i>
Nadav Rotem25f2ac92011-02-20 12:37:50 +0000[diff] [blame]5163 %Z = zext &lt;2 x i16&gt; &lt;i16 8, i16 7&gt; to &lt;2 x i32&gt; <i>; yields &lt;i32 8, i32 7&gt;</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5164</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5165
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5166</div>
5167
5168<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5169<h4>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5170 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5171</h4>
5172
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5173<div>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5174
5175<h5>Syntax:</h5>
5176<pre>
5177 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
5178</pre>
5179
5180<h5>Overview:</h5>
5181<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
5182
5183<h5>Arguments:</h5>
Nadav Rotem502f1b92011-02-24 21:01:34 +0000[diff] [blame]5184<p>The '<tt>sext</tt>' instruction takes a value to cast, and a type to cast it to.
5185 Both types must be of <a href="#t_integer">integer</a> types, or vectors
5186 of the same number of integers.
5187 The bit size of the <tt>value</tt> must be smaller than
5188 the bit size of the destination type,
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5189 <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5190
5191<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5192<p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
5193 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
5194 of the type <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5195
Reid Spencer36a15422007-01-12 03:35:51 +0000[diff] [blame]5196<p>When sign extending from i1, the extension always results in -1 or 0.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5197
5198<h5>Example:</h5>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5199<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]5200 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
Reid Spencer36a15422007-01-12 03:35:51 +0000[diff] [blame]5201 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
Nadav Rotem502f1b92011-02-24 21:01:34 +0000[diff] [blame]5202 %Z = sext &lt;2 x i16&gt; &lt;i16 8, i16 7&gt; to &lt;2 x i32&gt; <i>; yields &lt;i32 8, i32 7&gt;</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5203</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5204
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5205</div>
5206
5207<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5208<h4>
Reid Spencer2e2740d2006-11-09 21:48:10 +0000[diff] [blame]5209 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5210</h4>
Reid Spencer2e2740d2006-11-09 21:48:10 +0000[diff] [blame]5211
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5212<div>
Reid Spencer2e2740d2006-11-09 21:48:10 +0000[diff] [blame]5213
5214<h5>Syntax:</h5>
Reid Spencer2e2740d2006-11-09 21:48:10 +0000[diff] [blame]5215<pre>
5216 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
5217</pre>
5218
5219<h5>Overview:</h5>
5220<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5221 <tt>ty2</tt>.</p>
Reid Spencer2e2740d2006-11-09 21:48:10 +0000[diff] [blame]5222
5223<h5>Arguments:</h5>
5224<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5225 point</a> value to cast and a <a href="#t_floating">floating point</a> type
5226 to cast it to. The size of <tt>value</tt> must be larger than the size of
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5227 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5228 <i>no-op cast</i>.</p>
Reid Spencer2e2740d2006-11-09 21:48:10 +0000[diff] [blame]5229
5230<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5231<p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5232 <a href="#t_floating">floating point</a> type to a smaller
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5233 <a href="#t_floating">floating point</a> type. If the value cannot fit
5234 within the destination type, <tt>ty2</tt>, then the results are
5235 undefined.</p>
Reid Spencer2e2740d2006-11-09 21:48:10 +0000[diff] [blame]5236
5237<h5>Example:</h5>
5238<pre>
5239 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
5240 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
5241</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5242
Reid Spencer2e2740d2006-11-09 21:48:10 +0000[diff] [blame]5243</div>
5244
5245<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5246<h4>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5247 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5248</h4>
5249
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5250<div>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5251
5252<h5>Syntax:</h5>
5253<pre>
5254 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
5255</pre>
5256
5257<h5>Overview:</h5>
5258<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5259 floating point value.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5260
5261<h5>Arguments:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5262<p>The '<tt>fpext</tt>' instruction takes a
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5263 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
5264 a <a href="#t_floating">floating point</a> type to cast it to. The source
5265 type must be smaller than the destination type.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5266
5267<h5>Semantics:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +0000[diff] [blame]5268<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5269 <a href="#t_floating">floating point</a> type to a larger
5270 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
5271 used to make a <i>no-op cast</i> because it always changes bits. Use
5272 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5273
5274<h5>Example:</h5>
5275<pre>
Nick Lewycky9feca672011-03-31 18:20:19 +0000[diff] [blame]5276 %X = fpext float 3.125 to double <i>; yields double:3.125000e+00</i>
5277 %Y = fpext double %X to fp128 <i>; yields fp128:0xL00000000000000004000900000000000</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5278</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5279
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5280</div>
5281
5282<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5283<h4>
Reid Spencer2eadb532007-01-21 00:29:26 +0000[diff] [blame]5284 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5285</h4>
5286
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5287<div>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5288
5289<h5>Syntax:</h5>
5290<pre>
Reid Spencer753163d2007-07-31 14:40:14 +0000[diff] [blame]5291 &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 +0000[diff] [blame]5292</pre>
5293
5294<h5>Overview:</h5>
Reid Spencer753163d2007-07-31 14:40:14 +0000[diff] [blame]5295<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5296 unsigned integer equivalent of type <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5297
5298<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5299<p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
5300 scalar or vector <a href="#t_floating">floating point</a> value, and a type
5301 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
5302 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
5303 vector integer type with the same number of elements as <tt>ty</tt></p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5304
5305<h5>Semantics:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5306<p>The '<tt>fptoui</tt>' instruction converts its
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5307 <a href="#t_floating">floating point</a> operand into the nearest (rounding
5308 towards zero) unsigned integer value. If the value cannot fit
5309 in <tt>ty2</tt>, the results are undefined.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5310
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5311<h5>Example:</h5>
5312<pre>
Reid Spencer753163d2007-07-31 14:40:14 +0000[diff] [blame]5313 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner5b95a172007-09-22 03:17:52 +0000[diff] [blame]5314 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Gabor Greiff50fd572009-10-28 09:21:30 +0000[diff] [blame]5315 %Z = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5316</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5317
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5318</div>
5319
5320<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5321<h4>
Reid Spencer51b07252006-11-09 23:03:26 +0000[diff] [blame]5322 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5323</h4>
5324
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5325<div>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5326
5327<h5>Syntax:</h5>
5328<pre>
Reid Spencer51b07252006-11-09 23:03:26 +0000[diff] [blame]5329 &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 +0000[diff] [blame]5330</pre>
5331
5332<h5>Overview:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5333<p>The '<tt>fptosi</tt>' instruction converts
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5334 <a href="#t_floating">floating point</a> <tt>value</tt> to
5335 type <tt>ty2</tt>.</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5336
Chris Lattnera8292f32002-05-06 22:08:29 +0000[diff] [blame]5337<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5338<p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
5339 scalar or vector <a href="#t_floating">floating point</a> value, and a type
5340 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
5341 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
5342 vector integer type with the same number of elements as <tt>ty</tt></p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5343
Chris Lattnera8292f32002-05-06 22:08:29 +0000[diff] [blame]5344<h5>Semantics:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5345<p>The '<tt>fptosi</tt>' instruction converts its
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5346 <a href="#t_floating">floating point</a> operand into the nearest (rounding
5347 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
5348 the results are undefined.</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5349
Chris Lattner70de6632001-07-09 00:26:23 +0000[diff] [blame]5350<h5>Example:</h5>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5351<pre>
Reid Spencer36a15422007-01-12 03:35:51 +0000[diff] [blame]5352 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner5b95a172007-09-22 03:17:52 +0000[diff] [blame]5353 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Gabor Greiff50fd572009-10-28 09:21:30 +0000[diff] [blame]5354 %Z = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5355</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5356
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5357</div>
5358
5359<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5360<h4>
Reid Spencer51b07252006-11-09 23:03:26 +0000[diff] [blame]5361 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5362</h4>
5363
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5364<div>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5365
5366<h5>Syntax:</h5>
5367<pre>
Reid Spencer51b07252006-11-09 23:03:26 +0000[diff] [blame]5368 &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 +0000[diff] [blame]5369</pre>
5370
5371<h5>Overview:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +0000[diff] [blame]5372<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5373 integer and converts that value to the <tt>ty2</tt> type.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5374
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5375<h5>Arguments:</h5>
Nate Begemand4d45c22007-11-17 03:58:34 +0000[diff] [blame]5376<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5377 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
5378 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
5379 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
5380 floating point type with the same number of elements as <tt>ty</tt></p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5381
5382<h5>Semantics:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +0000[diff] [blame]5383<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5384 integer quantity and converts it to the corresponding floating point
5385 value. If the value cannot fit in the floating point value, the results are
5386 undefined.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5387
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5388<h5>Example:</h5>
5389<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]5390 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohmanef9462f2008-10-14 16:51:45 +0000[diff] [blame]5391 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5392</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5393
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5394</div>
5395
5396<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5397<h4>
Reid Spencer51b07252006-11-09 23:03:26 +0000[diff] [blame]5398 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5399</h4>
5400
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5401<div>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5402
5403<h5>Syntax:</h5>
5404<pre>
Reid Spencer51b07252006-11-09 23:03:26 +0000[diff] [blame]5405 &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 +0000[diff] [blame]5406</pre>
5407
5408<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5409<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
5410 and converts that value to the <tt>ty2</tt> type.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5411
5412<h5>Arguments:</h5>
Nate Begemand4d45c22007-11-17 03:58:34 +0000[diff] [blame]5413<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5414 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
5415 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
5416 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
5417 floating point type with the same number of elements as <tt>ty</tt></p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5418
5419<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5420<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
5421 quantity and converts it to the corresponding floating point value. If the
5422 value cannot fit in the floating point value, the results are undefined.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5423
5424<h5>Example:</h5>
5425<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]5426 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohmanef9462f2008-10-14 16:51:45 +0000[diff] [blame]5427 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5428</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5429
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5430</div>
5431
5432<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5433<h4>
Reid Spencerb7344ff2006-11-11 21:00:47 +0000[diff] [blame]5434 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5435</h4>
5436
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5437<div>
Reid Spencerb7344ff2006-11-11 21:00:47 +0000[diff] [blame]5438
5439<h5>Syntax:</h5>
5440<pre>
5441 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
5442</pre>
5443
5444<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5445<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
5446 the integer type <tt>ty2</tt>.</p>
Reid Spencerb7344ff2006-11-11 21:00:47 +0000[diff] [blame]5447
5448<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5449<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
5450 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
5451 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
Reid Spencerb7344ff2006-11-11 21:00:47 +0000[diff] [blame]5452
5453<h5>Semantics:</h5>
5454<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5455 <tt>ty2</tt> by interpreting the pointer value as an integer and either
5456 truncating or zero extending that value to the size of the integer type. If
5457 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
5458 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
5459 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
5460 change.</p>
Reid Spencerb7344ff2006-11-11 21:00:47 +0000[diff] [blame]5461
5462<h5>Example:</h5>
5463<pre>
Jeff Cohen222a8a42007-04-29 01:07:00 +0000[diff] [blame]5464 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
5465 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
Reid Spencerb7344ff2006-11-11 21:00:47 +0000[diff] [blame]5466</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5467
Reid Spencerb7344ff2006-11-11 21:00:47 +0000[diff] [blame]5468</div>
5469
5470<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5471<h4>
Reid Spencerb7344ff2006-11-11 21:00:47 +0000[diff] [blame]5472 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5473</h4>
5474
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5475<div>
Reid Spencerb7344ff2006-11-11 21:00:47 +0000[diff] [blame]5476
5477<h5>Syntax:</h5>
5478<pre>
5479 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
5480</pre>
5481
5482<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5483<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
5484 pointer type, <tt>ty2</tt>.</p>
Reid Spencerb7344ff2006-11-11 21:00:47 +0000[diff] [blame]5485
5486<h5>Arguments:</h5>
Duncan Sands16f122e2007-03-30 12:22:09 +0000[diff] [blame]5487<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5488 value to cast, and a type to cast it to, which must be a
5489 <a href="#t_pointer">pointer</a> type.</p>
Reid Spencerb7344ff2006-11-11 21:00:47 +0000[diff] [blame]5490
5491<h5>Semantics:</h5>
5492<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5493 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
5494 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
5495 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
5496 than the size of a pointer then a zero extension is done. If they are the
5497 same size, nothing is done (<i>no-op cast</i>).</p>
Reid Spencerb7344ff2006-11-11 21:00:47 +0000[diff] [blame]5498
5499<h5>Example:</h5>
5500<pre>
Jeff Cohen222a8a42007-04-29 01:07:00 +0000[diff] [blame]5501 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
Gabor Greiff50fd572009-10-28 09:21:30 +0000[diff] [blame]5502 %Y = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
5503 %Z = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
Reid Spencerb7344ff2006-11-11 21:00:47 +0000[diff] [blame]5504</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5505
Reid Spencerb7344ff2006-11-11 21:00:47 +0000[diff] [blame]5506</div>
5507
5508<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5509<h4>
Reid Spencer5b950642006-11-11 23:08:07 +0000[diff] [blame]5510 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5511</h4>
5512
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5513<div>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5514
5515<h5>Syntax:</h5>
5516<pre>
Reid Spencer5b950642006-11-11 23:08:07 +0000[diff] [blame]5517 &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 +0000[diff] [blame]5518</pre>
5519
5520<h5>Overview:</h5>
Reid Spencer5b950642006-11-11 23:08:07 +0000[diff] [blame]5521<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5522 <tt>ty2</tt> without changing any bits.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5523
5524<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5525<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
5526 non-aggregate first class value, and a type to cast it to, which must also be
5527 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
5528 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
5529 identical. If the source type is a pointer, the destination type must also be
5530 a pointer. This instruction supports bitwise conversion of vectors to
5531 integers and to vectors of other types (as long as they have the same
5532 size).</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5533
5534<h5>Semantics:</h5>
Reid Spencer5b950642006-11-11 23:08:07 +0000[diff] [blame]5535<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5536 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
5537 this conversion. The conversion is done as if the <tt>value</tt> had been
5538 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
5539 be converted to other pointer types with this instruction. To convert
5540 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
5541 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000[diff] [blame]5542
5543<h5>Example:</h5>
5544<pre>
Jeff Cohen222a8a42007-04-29 01:07:00 +0000[diff] [blame]5545 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]5546 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5547 %Z = bitcast &lt;2 x int&gt; %V to i64; <i>; yields i64: %V</i>
Chris Lattner70de6632001-07-09 00:26:23 +0000[diff] [blame]5548</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5549
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]5550</div>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5551
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5552</div>
5553
Reid Spencer97c5fa42006-11-08 01:18:52 +0000[diff] [blame]5554<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5555<h3>
5556 <a name="otherops">Other Operations</a>
5557</h3>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5558
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5559<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5560
5561<p>The instructions in this category are the "miscellaneous" instructions, which
5562 defy better classification.</p>
5563
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5564<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5565<h4>
5566 <a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
5567</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5568
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5569<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5570
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5571<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5572<pre>
5573 &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 +0000[diff] [blame]5574</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5575
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5576<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5577<p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
5578 boolean values based on comparison of its two integer, integer vector, or
5579 pointer operands.</p>
5580
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5581<h5>Arguments:</h5>
5582<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5583 the condition code indicating the kind of comparison to perform. It is not a
5584 value, just a keyword. The possible condition code are:</p>
5585
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5586<ol>
5587 <li><tt>eq</tt>: equal</li>
5588 <li><tt>ne</tt>: not equal </li>
5589 <li><tt>ugt</tt>: unsigned greater than</li>
5590 <li><tt>uge</tt>: unsigned greater or equal</li>
5591 <li><tt>ult</tt>: unsigned less than</li>
5592 <li><tt>ule</tt>: unsigned less or equal</li>
5593 <li><tt>sgt</tt>: signed greater than</li>
5594 <li><tt>sge</tt>: signed greater or equal</li>
5595 <li><tt>slt</tt>: signed less than</li>
5596 <li><tt>sle</tt>: signed less or equal</li>
5597</ol>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5598
Chris Lattnerc0f423a2007-01-15 01:54:13 +0000[diff] [blame]5599<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5600 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
5601 typed. They must also be identical types.</p>
5602
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5603<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5604<p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
5605 condition code given as <tt>cond</tt>. The comparison performed always yields
Nick Lewycky84a1eeb2009-09-27 00:45:11 +0000[diff] [blame]5606 either an <a href="#t_integer"><tt>i1</tt></a> or vector of <tt>i1</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5607 result, as follows:</p>
5608
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5609<ol>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5610 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5611 <tt>false</tt> otherwise. No sign interpretation is necessary or
5612 performed.</li>
5613
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5614 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5615 <tt>false</tt> otherwise. No sign interpretation is necessary or
5616 performed.</li>
5617
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5618 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5619 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5620
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5621 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5622 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5623 to <tt>op2</tt>.</li>
5624
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5625 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5626 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5627
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5628 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5629 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5630
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5631 <li><tt>sgt</tt>: interprets the operands as signed values and yields
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5632 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5633
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5634 <li><tt>sge</tt>: interprets the operands as signed values and yields
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5635 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5636 to <tt>op2</tt>.</li>
5637
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5638 <li><tt>slt</tt>: interprets the operands as signed values and yields
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5639 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5640
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5641 <li><tt>sle</tt>: interprets the operands as signed values and yields
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5642 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5643</ol>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5644
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5645<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5646 values are compared as if they were integers.</p>
5647
5648<p>If the operands are integer vectors, then they are compared element by
5649 element. The result is an <tt>i1</tt> vector with the same number of elements
5650 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5651
5652<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5653<pre>
5654 &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]5655 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
5656 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
5657 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
5658 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
5659 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5660</pre>
Dan Gohmana5127ab2009-01-22 01:39:38 +0000[diff] [blame]5661
5662<p>Note that the code generator does not yet support vector types with
5663 the <tt>icmp</tt> instruction.</p>
5664
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5665</div>
5666
5667<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5668<h4>
5669 <a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
5670</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5671
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5672<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5673
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5674<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5675<pre>
5676 &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 +0000[diff] [blame]5677</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5678
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5679<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5680<p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
5681 values based on comparison of its operands.</p>
5682
5683<p>If the operands are floating point scalars, then the result type is a boolean
Nick Lewycky84a1eeb2009-09-27 00:45:11 +0000[diff] [blame]5684(<a href="#t_integer"><tt>i1</tt></a>).</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5685
5686<p>If the operands are floating point vectors, then the result type is a vector
5687 of boolean with the same number of elements as the operands being
5688 compared.</p>
5689
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5690<h5>Arguments:</h5>
5691<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5692 the condition code indicating the kind of comparison to perform. It is not a
5693 value, just a keyword. The possible condition code are:</p>
5694
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5695<ol>
Reid Spencerf69acf32006-11-19 03:00:14 +0000[diff] [blame]5696 <li><tt>false</tt>: no comparison, always returns false</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5697 <li><tt>oeq</tt>: ordered and equal</li>
5698 <li><tt>ogt</tt>: ordered and greater than </li>
5699 <li><tt>oge</tt>: ordered and greater than or equal</li>
5700 <li><tt>olt</tt>: ordered and less than </li>
5701 <li><tt>ole</tt>: ordered and less than or equal</li>
5702 <li><tt>one</tt>: ordered and not equal</li>
5703 <li><tt>ord</tt>: ordered (no nans)</li>
5704 <li><tt>ueq</tt>: unordered or equal</li>
5705 <li><tt>ugt</tt>: unordered or greater than </li>
5706 <li><tt>uge</tt>: unordered or greater than or equal</li>
5707 <li><tt>ult</tt>: unordered or less than </li>
5708 <li><tt>ule</tt>: unordered or less than or equal</li>
5709 <li><tt>une</tt>: unordered or not equal</li>
5710 <li><tt>uno</tt>: unordered (either nans)</li>
Reid Spencerf69acf32006-11-19 03:00:14 +0000[diff] [blame]5711 <li><tt>true</tt>: no comparison, always returns true</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5712</ol>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5713
Jeff Cohen222a8a42007-04-29 01:07:00 +0000[diff] [blame]5714<p><i>Ordered</i> means that neither operand is a QNAN while
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5715 <i>unordered</i> means that either operand may be a QNAN.</p>
5716
5717<p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
5718 a <a href="#t_floating">floating point</a> type or
5719 a <a href="#t_vector">vector</a> of floating point type. They must have
5720 identical types.</p>
5721
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5722<h5>Semantics:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +0000[diff] [blame]5723<p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5724 according to the condition code given as <tt>cond</tt>. If the operands are
5725 vectors, then the vectors are compared element by element. Each comparison
Nick Lewycky84a1eeb2009-09-27 00:45:11 +0000[diff] [blame]5726 performed always yields an <a href="#t_integer">i1</a> result, as
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5727 follows:</p>
5728
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5729<ol>
5730 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5731
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5732 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5733 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5734
Reid Spencerf69acf32006-11-19 03:00:14 +0000[diff] [blame]5735 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Dan Gohmana269a0a2010-03-01 17:41:39 +0000[diff] [blame]5736 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5737
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5738 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5739 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5740
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5741 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5742 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5743
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5744 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5745 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5746
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5747 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5748 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5749
Reid Spencerf69acf32006-11-19 03:00:14 +0000[diff] [blame]5750 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5751
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5752 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5753 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5754
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5755 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5756 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5757
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5758 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5759 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5760
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5761 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5762 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5763
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5764 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5765 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5766
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]5767 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5768 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5769
Reid Spencerf69acf32006-11-19 03:00:14 +0000[diff] [blame]5770 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5771
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5772 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
5773</ol>
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5774
5775<h5>Example:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5776<pre>
5777 &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
Dan Gohmanc579d972008-09-09 01:02:47 +0000[diff] [blame]5778 &lt;result&gt; = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
5779 &lt;result&gt; = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
5780 &lt;result&gt; = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5781</pre>
Dan Gohmana5127ab2009-01-22 01:39:38 +0000[diff] [blame]5782
5783<p>Note that the code generator does not yet support vector types with
5784 the <tt>fcmp</tt> instruction.</p>
5785
Reid Spencerc828a0e2006-11-18 21:50:54 +0000[diff] [blame]5786</div>
5787
Reid Spencer97c5fa42006-11-08 01:18:52 +0000[diff] [blame]5788<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5789<h4>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]5790 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5791</h4>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]5792
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5793<div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]5794
Reid Spencer97c5fa42006-11-08 01:18:52 +0000[diff] [blame]5795<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5796<pre>
5797 &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...
5798</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]5799
Reid Spencer97c5fa42006-11-08 01:18:52 +0000[diff] [blame]5800<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5801<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in the
5802 SSA graph representing the function.</p>
5803
Reid Spencer97c5fa42006-11-08 01:18:52 +0000[diff] [blame]5804<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5805<p>The type of the incoming values is specified with the first type field. After
5806 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
5807 one pair for each predecessor basic block of the current block. Only values
5808 of <a href="#t_firstclass">first class</a> type may be used as the value
5809 arguments to the PHI node. Only labels may be used as the label
5810 arguments.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]5811
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5812<p>There must be no non-phi instructions between the start of a basic block and
5813 the PHI instructions: i.e. PHI instructions must be first in a basic
5814 block.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]5815
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5816<p>For the purposes of the SSA form, the use of each incoming value is deemed to
5817 occur on the edge from the corresponding predecessor block to the current
5818 block (but after any definition of an '<tt>invoke</tt>' instruction's return
5819 value on the same edge).</p>
Jay Foad1a4eea52009-06-03 10:20:10 +0000[diff] [blame]5820
Reid Spencer97c5fa42006-11-08 01:18:52 +0000[diff] [blame]5821<h5>Semantics:</h5>
Jeff Cohen222a8a42007-04-29 01:07:00 +0000[diff] [blame]5822<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5823 specified by the pair corresponding to the predecessor basic block that
5824 executed just prior to the current block.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]5825
Reid Spencer97c5fa42006-11-08 01:18:52 +0000[diff] [blame]5826<h5>Example:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +0000[diff] [blame]5827<pre>
5828Loop: ; Infinite loop that counts from 0 on up...
5829 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
5830 %nextindvar = add i32 %indvar, 1
5831 br label %Loop
5832</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5833
Reid Spencer97c5fa42006-11-08 01:18:52 +0000[diff] [blame]5834</div>
5835
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5836<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5837<h4>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5838 <a name="i_select">'<tt>select</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5839</h4>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5840
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5841<div>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5842
5843<h5>Syntax:</h5>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5844<pre>
Dan Gohmanc579d972008-09-09 01:02:47 +0000[diff] [blame]5845 &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>
5846
Dan Gohmanef9462f2008-10-14 16:51:45 +0000[diff] [blame]5847 <i>selty</i> is either i1 or {&lt;N x i1&gt;}
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5848</pre>
5849
5850<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5851<p>The '<tt>select</tt>' instruction is used to choose one value based on a
5852 condition, without branching.</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5853
5854
5855<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5856<p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
5857 values indicating the condition, and two values of the
5858 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
5859 vectors and the condition is a scalar, then entire vectors are selected, not
5860 individual elements.</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5861
5862<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5863<p>If the condition is an i1 and it evaluates to 1, the instruction returns the
5864 first value argument; otherwise, it returns the second value argument.</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5865
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5866<p>If the condition is a vector of i1, then the value arguments must be vectors
5867 of the same size, and the selection is done element by element.</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5868
5869<h5>Example:</h5>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5870<pre>
Reid Spencer36a15422007-01-12 03:35:51 +0000[diff] [blame]5871 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5872</pre>
Dan Gohmana5127ab2009-01-22 01:39:38 +0000[diff] [blame]5873
5874<p>Note that the code generator does not yet support conditions
5875 with vector type.</p>
5876
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000[diff] [blame]5877</div>
5878
Robert Bocchinof72fdfe2006-01-15 20:48:27 +0000[diff] [blame]5879<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5880<h4>
Chris Lattnere23c1392005-05-06 05:47:36 +0000[diff] [blame]5881 <a name="i_call">'<tt>call</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5882</h4>
Chris Lattnere23c1392005-05-06 05:47:36 +0000[diff] [blame]5883
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5884<div>
Chris Lattnere23c1392005-05-06 05:47:36 +0000[diff] [blame]5885
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]5886<h5>Syntax:</h5>
Chris Lattnere23c1392005-05-06 05:47:36 +0000[diff] [blame]5887<pre>
Devang Patel02256232008-10-07 17:48:33 +0000[diff] [blame]5888 &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 +0000[diff] [blame]5889</pre>
5890
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]5891<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]5892<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
Chris Lattnere23c1392005-05-06 05:47:36 +0000[diff] [blame]5893
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]5894<h5>Arguments:</h5>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]5895<p>This instruction requires several arguments:</p>
Chris Lattnere23c1392005-05-06 05:47:36 +0000[diff] [blame]5896
Chris Lattnera8292f32002-05-06 22:08:29 +0000[diff] [blame]5897<ol>
Jeffrey Yasskinb8677462010-01-09 19:44:16 +0000[diff] [blame]5898 <li>The optional "tail" marker indicates that the callee function does not
5899 access any allocas or varargs in the caller. Note that calls may be
5900 marked "tail" even if they do not occur before
5901 a <a href="#i_ret"><tt>ret</tt></a> instruction. If the "tail" marker is
5902 present, the function call is eligible for tail call optimization,
5903 but <a href="CodeGenerator.html#tailcallopt">might not in fact be
Evan Cheng59676492010-03-08 21:05:02 +0000[diff] [blame]5904 optimized into a jump</a>. The code generator may optimize calls marked
5905 "tail" with either 1) automatic <a href="CodeGenerator.html#sibcallopt">
5906 sibling call optimization</a> when the caller and callee have
5907 matching signatures, or 2) forced tail call optimization when the
5908 following extra requirements are met:
Jeffrey Yasskinb8677462010-01-09 19:44:16 +0000[diff] [blame]5909 <ul>
5910 <li>Caller and callee both have the calling
5911 convention <tt>fastcc</tt>.</li>
5912 <li>The call is in tail position (ret immediately follows call and ret
5913 uses value of call or is void).</li>
5914 <li>Option <tt>-tailcallopt</tt> is enabled,
Dan Gohman6232f732010-03-02 01:08:11 +0000[diff] [blame]5915 or <code>llvm::GuaranteedTailCallOpt</code> is <code>true</code>.</li>
Jeffrey Yasskinb8677462010-01-09 19:44:16 +0000[diff] [blame]5916 <li><a href="CodeGenerator.html#tailcallopt">Platform specific
5917 constraints are met.</a></li>
5918 </ul>
5919 </li>
Devang Patel7e9b05e2008-10-06 18:50:38 +0000[diff] [blame]5920
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5921 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
5922 convention</a> the call should use. If none is specified, the call
Jeffrey Yasskinb8677462010-01-09 19:44:16 +0000[diff] [blame]5923 defaults to using C calling conventions. The calling convention of the
5924 call must match the calling convention of the target function, or else the
5925 behavior is undefined.</li>
Devang Patel7e9b05e2008-10-06 18:50:38 +0000[diff] [blame]5926
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5927 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
5928 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
5929 '<tt>inreg</tt>' attributes are valid here.</li>
5930
5931 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
5932 type of the return value. Functions that return no value are marked
5933 <tt><a href="#t_void">void</a></tt>.</li>
5934
5935 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
5936 being invoked. The argument types must match the types implied by this
5937 signature. This type can be omitted if the function is not varargs and if
5938 the function type does not return a pointer to a function.</li>
5939
5940 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
5941 be invoked. In most cases, this is a direct function invocation, but
5942 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
5943 to function value.</li>
5944
5945 <li>'<tt>function args</tt>': argument list whose types match the function
Chris Lattner47f2a832010-03-02 06:36:51 +0000[diff] [blame]5946 signature argument types and parameter attributes. All arguments must be
5947 of <a href="#t_firstclass">first class</a> type. If the function
5948 signature indicates the function accepts a variable number of arguments,
5949 the extra arguments can be specified.</li>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5950
5951 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
5952 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
5953 '<tt>readnone</tt>' attributes are valid here.</li>
Chris Lattnera8292f32002-05-06 22:08:29 +0000[diff] [blame]5954</ol>
Chris Lattnere23c1392005-05-06 05:47:36 +0000[diff] [blame]5955
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]5956<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]5957<p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
5958 a specified function, with its incoming arguments bound to the specified
5959 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
5960 function, control flow continues with the instruction after the function
5961 call, and the return value of the function is bound to the result
5962 argument.</p>
Chris Lattnere23c1392005-05-06 05:47:36 +0000[diff] [blame]5963
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]5964<h5>Example:</h5>
Chris Lattnere23c1392005-05-06 05:47:36 +0000[diff] [blame]5965<pre>
Nick Lewyckya9b13d52007-09-08 13:57:50 +0000[diff] [blame]5966 %retval = call i32 @test(i32 %argc)
Dan Gohmanaabfdb32010-05-28 17:13:49 +0000[diff] [blame]5967 call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42) <i>; yields i32</i>
Chris Lattnerfb7c88d2008-03-21 17:24:17 +0000[diff] [blame]5968 %X = tail call i32 @foo() <i>; yields i32</i>
5969 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
5970 call void %foo(i8 97 signext)
Devang Pateld6cff512008-03-10 20:49:15 +0000[diff] [blame]5971
5972 %struct.A = type { i32, i8 }
Devang Patel7e9b05e2008-10-06 18:50:38 +0000[diff] [blame]5973 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
Dan Gohmancc3132e2008-10-04 19:00:07 +0000[diff] [blame]5974 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
5975 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
Chris Lattner6cbe8e92008-10-08 06:26:11 +0000[diff] [blame]5976 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
Matthijs Kooijmaneefa7df2008-10-07 10:03:45 +0000[diff] [blame]5977 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
Chris Lattnere23c1392005-05-06 05:47:36 +0000[diff] [blame]5978</pre>
5979
Dale Johannesen68f971b2009-09-24 18:38:21 +0000[diff] [blame]5980<p>llvm treats calls to some functions with names and arguments that match the
Dale Johannesen722212d2009-09-25 17:04:42 +0000[diff] [blame]5981standard C99 library as being the C99 library functions, and may perform
5982optimizations or generate code for them under that assumption. This is
5983something we'd like to change in the future to provide better support for
Dan Gohmana269a0a2010-03-01 17:41:39 +0000[diff] [blame]5984freestanding environments and non-C-based languages.</p>
Dale Johannesen68f971b2009-09-24 18:38:21 +0000[diff] [blame]5985
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]5986</div>
Chris Lattner6a4a0492004-09-27 21:51:25 +0000[diff] [blame]5987
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]5988<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5989<h4>
Chris Lattner33337472006-01-13 23:26:01 +0000[diff] [blame]5990 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]5991</h4>
Chris Lattner6a4a0492004-09-27 21:51:25 +0000[diff] [blame]5992
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]5993<div>
Chris Lattner6a4a0492004-09-27 21:51:25 +0000[diff] [blame]5994
Chris Lattner26ca62e2003-10-18 05:51:36 +0000[diff] [blame]5995<h5>Syntax:</h5>
Chris Lattner6a4a0492004-09-27 21:51:25 +0000[diff] [blame]5996<pre>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +0000[diff] [blame]5997 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
Chris Lattner6a4a0492004-09-27 21:51:25 +0000[diff] [blame]5998</pre>
5999
Chris Lattner26ca62e2003-10-18 05:51:36 +0000[diff] [blame]6000<h5>Overview:</h5>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +0000[diff] [blame]6001<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6002 the "variable argument" area of a function call. It is used to implement the
6003 <tt>va_arg</tt> macro in C.</p>
Chris Lattner6a4a0492004-09-27 21:51:25 +0000[diff] [blame]6004
Chris Lattner26ca62e2003-10-18 05:51:36 +0000[diff] [blame]6005<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6006<p>This instruction takes a <tt>va_list*</tt> value and the type of the
6007 argument. It returns a value of the specified argument type and increments
6008 the <tt>va_list</tt> to point to the next argument. The actual type
6009 of <tt>va_list</tt> is target specific.</p>
Chris Lattner6a4a0492004-09-27 21:51:25 +0000[diff] [blame]6010
Chris Lattner26ca62e2003-10-18 05:51:36 +0000[diff] [blame]6011<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6012<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
6013 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
6014 to the next argument. For more information, see the variable argument
6015 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
Chris Lattner6a4a0492004-09-27 21:51:25 +0000[diff] [blame]6016
6017<p>It is legal for this instruction to be called in a function which does not
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6018 take a variable number of arguments, for example, the <tt>vfprintf</tt>
6019 function.</p>
Chris Lattner6a4a0492004-09-27 21:51:25 +0000[diff] [blame]6020
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6021<p><tt>va_arg</tt> is an LLVM instruction instead of
6022 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
6023 argument.</p>
Chris Lattner6a4a0492004-09-27 21:51:25 +0000[diff] [blame]6024
Chris Lattner26ca62e2003-10-18 05:51:36 +0000[diff] [blame]6025<h5>Example:</h5>
Chris Lattner6a4a0492004-09-27 21:51:25 +0000[diff] [blame]6026<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
6027
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6028<p>Note that the code generator does not yet fully support va_arg on many
6029 targets. Also, it does not currently support va_arg with aggregate types on
6030 any target.</p>
Dan Gohman3065b612009-01-12 23:12:39 +0000[diff] [blame]6031
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]6032</div>
Chris Lattner941515c2004-01-06 05:31:32 +0000[diff] [blame]6033
Bill Wendlingbbcb7cd2011-08-02 21:52:38 +0000[diff] [blame]6034<!-- _______________________________________________________________________ -->
6035<h4>
6036 <a name="i_landingpad">'<tt>landingpad</tt>' Instruction</a>
6037</h4>
6038
6039<div>
6040
6041<h5>Syntax:</h5>
6042<pre>
Bill Wendling49bfb122011-08-08 08:06:05 +0000[diff] [blame]6043 &lt;resultval&gt; = landingpad &lt;somety&gt; personality &lt;type&gt; &lt;pers_fn&gt; &lt;clause&gt;+
6044 &lt;resultval&gt; = landingpad &lt;somety&gt; personality &lt;type&gt; &lt;pers_fn&gt; cleanup &lt;clause&gt;*
6045
Bill Wendlingbbcb7cd2011-08-02 21:52:38 +0000[diff] [blame]6046 &lt;clause&gt; := catch &lt;type&gt; &lt;value&gt;
Bill Wendlingfae14752011-08-12 20:24:12 +0000[diff] [blame]6047 &lt;clause&gt; := filter &lt;array constant type&gt; &lt;array constant&gt;
Bill Wendlingbbcb7cd2011-08-02 21:52:38 +0000[diff] [blame]6048</pre>
6049
6050<h5>Overview:</h5>
6051<p>The '<tt>landingpad</tt>' instruction is used by
6052 <a href="ExceptionHandling.html#overview">LLVM's exception handling
6053 system</a> to specify that a basic block is a landing pad &mdash; one where
6054 the exception lands, and corresponds to the code found in the
6055 <i><tt>catch</tt></i> portion of a <i><tt>try/catch</tt></i> sequence. It
6056 defines values supplied by the personality function (<tt>pers_fn</tt>) upon
6057 re-entry to the function. The <tt>resultval</tt> has the
6058 type <tt>somety</tt>.</p>
6059
6060<h5>Arguments:</h5>
6061<p>This instruction takes a <tt>pers_fn</tt> value. This is the personality
6062 function associated with the unwinding mechanism. The optional
6063 <tt>cleanup</tt> flag indicates that the landing pad block is a cleanup.</p>
6064
6065<p>A <tt>clause</tt> begins with the clause type &mdash; <tt>catch</tt>
Bill Wendlingfae14752011-08-12 20:24:12 +0000[diff] [blame]6066 or <tt>filter</tt> &mdash; and contains the global variable representing the
6067 "type" that may be caught or filtered respectively. Unlike the
6068 <tt>catch</tt> clause, the <tt>filter</tt> clause takes an array constant as
6069 its argument. Use "<tt>[0 x i8**] undef</tt>" for a filter which cannot
6070 throw. The '<tt>landingpad</tt>' instruction must contain <em>at least</em>
Bill Wendlingbbcb7cd2011-08-02 21:52:38 +0000[diff] [blame]6071 one <tt>clause</tt> or the <tt>cleanup</tt> flag.</p>
6072
6073<h5>Semantics:</h5>
6074<p>The '<tt>landingpad</tt>' instruction defines the values which are set by the
6075 personality function (<tt>pers_fn</tt>) upon re-entry to the function, and
6076 therefore the "result type" of the <tt>landingpad</tt> instruction. As with
6077 calling conventions, how the personality function results are represented in
6078 LLVM IR is target specific.</p>
6079
Bill Wendling0524b8d2011-08-03 17:17:06 +0000[diff] [blame]6080<p>The clauses are applied in order from top to bottom. If two
6081 <tt>landingpad</tt> instructions are merged together through inlining, the
Bill Wendlinga503fc02011-08-08 07:58:58 +0000[diff] [blame]6082 clauses from the calling function are appended to the list of clauses.</p>
Bill Wendling0524b8d2011-08-03 17:17:06 +0000[diff] [blame]6083
Bill Wendlingbbcb7cd2011-08-02 21:52:38 +0000[diff] [blame]6084<p>The <tt>landingpad</tt> instruction has several restrictions:</p>
6085
6086<ul>
6087 <li>A landing pad block is a basic block which is the unwind destination of an
6088 '<tt>invoke</tt>' instruction.</li>
6089 <li>A landing pad block must have a '<tt>landingpad</tt>' instruction as its
6090 first non-PHI instruction.</li>
6091 <li>There can be only one '<tt>landingpad</tt>' instruction within the landing
6092 pad block.</li>
6093 <li>A basic block that is not a landing pad block may not include a
6094 '<tt>landingpad</tt>' instruction.</li>
6095 <li>All '<tt>landingpad</tt>' instructions in a function must have the same
6096 personality function.</li>
6097</ul>
6098
6099<h5>Example:</h5>
6100<pre>
6101 ;; A landing pad which can catch an integer.
6102 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6103 catch i8** @_ZTIi
6104 ;; A landing pad that is a cleanup.
6105 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
Bill Wendlingfae14752011-08-12 20:24:12 +0000[diff] [blame]6106 cleanup
Bill Wendlingbbcb7cd2011-08-02 21:52:38 +0000[diff] [blame]6107 ;; A landing pad which can catch an integer and can only throw a double.
6108 %res = landingpad { i8*, i32 } personality i32 (...)* @__gxx_personality_v0
6109 catch i8** @_ZTIi
Bill Wendlingfae14752011-08-12 20:24:12 +0000[diff] [blame]6110 filter [1 x i8**] [@_ZTId]
Bill Wendlingbbcb7cd2011-08-02 21:52:38 +0000[diff] [blame]6111</pre>
6112
6113</div>
6114
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6115</div>
6116
6117</div>
6118
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]6119<!-- *********************************************************************** -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6120<h2><a name="intrinsics">Intrinsic Functions</a></h2>
Chris Lattner48b383b02003-11-25 01:02:51 +0000[diff] [blame]6121<!-- *********************************************************************** -->
Chris Lattner941515c2004-01-06 05:31:32 +0000[diff] [blame]6122
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6123<div>
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6124
6125<p>LLVM supports the notion of an "intrinsic function". These functions have
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6126 well known names and semantics and are required to follow certain
6127 restrictions. Overall, these intrinsics represent an extension mechanism for
6128 the LLVM language that does not require changing all of the transformations
6129 in LLVM when adding to the language (or the bitcode reader/writer, the
6130 parser, etc...).</p>
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6131
John Criswell88190562005-05-16 16:17:45 +0000[diff] [blame]6132<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6133 prefix is reserved in LLVM for intrinsic names; thus, function names may not
6134 begin with this prefix. Intrinsic functions must always be external
6135 functions: you cannot define the body of intrinsic functions. Intrinsic
6136 functions may only be used in call or invoke instructions: it is illegal to
6137 take the address of an intrinsic function. Additionally, because intrinsic
6138 functions are part of the LLVM language, it is required if any are added that
6139 they be documented here.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6140
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6141<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
6142 family of functions that perform the same operation but on different data
6143 types. Because LLVM can represent over 8 million different integer types,
6144 overloading is used commonly to allow an intrinsic function to operate on any
6145 integer type. One or more of the argument types or the result type can be
6146 overloaded to accept any integer type. Argument types may also be defined as
6147 exactly matching a previous argument's type or the result type. This allows
6148 an intrinsic function which accepts multiple arguments, but needs all of them
6149 to be of the same type, to only be overloaded with respect to a single
6150 argument or the result.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6151
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6152<p>Overloaded intrinsics will have the names of its overloaded argument types
6153 encoded into its function name, each preceded by a period. Only those types
6154 which are overloaded result in a name suffix. Arguments whose type is matched
6155 against another type do not. For example, the <tt>llvm.ctpop</tt> function
6156 can take an integer of any width and returns an integer of exactly the same
6157 integer width. This leads to a family of functions such as
6158 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
6159 %val)</tt>. Only one type, the return type, is overloaded, and only one type
6160 suffix is required. Because the argument's type is matched against the return
6161 type, it does not require its own name suffix.</p>
Reid Spencer4eefaab2007-04-01 08:04:23 +0000[diff] [blame]6162
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]6163<p>To learn how to add an intrinsic function, please see the
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6164 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6165
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]6166<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6167<h3>
Chris Lattner941515c2004-01-06 05:31:32 +0000[diff] [blame]6168 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6169</h3>
Chris Lattner941515c2004-01-06 05:31:32 +0000[diff] [blame]6170
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6171<div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6172
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6173<p>Variable argument support is defined in LLVM with
6174 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
6175 intrinsic functions. These functions are related to the similarly named
6176 macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6177
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6178<p>All of these functions operate on arguments that use a target-specific value
6179 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
6180 not define what this type is, so all transformations should be prepared to
6181 handle these functions regardless of the type used.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6182
Chris Lattner30b868d2006-05-15 17:26:46 +0000[diff] [blame]6183<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6184 instruction and the variable argument handling intrinsic functions are
6185 used.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6186
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]6187<pre class="doc_code">
Anton Korobeynikov640bbe02007-03-21 23:58:04 +0000[diff] [blame]6188define i32 @test(i32 %X, ...) {
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6189 ; Initialize variable argument processing
Jeff Cohen222a8a42007-04-29 01:07:00 +0000[diff] [blame]6190 %ap = alloca i8*
Chris Lattnerdb0790c2007-01-08 07:55:15 +0000[diff] [blame]6191 %ap2 = bitcast i8** %ap to i8*
Anton Korobeynikov640bbe02007-03-21 23:58:04 +0000[diff] [blame]6192 call void @llvm.va_start(i8* %ap2)
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6193
6194 ; Read a single integer argument
Jeff Cohen222a8a42007-04-29 01:07:00 +0000[diff] [blame]6195 %tmp = va_arg i8** %ap, i32
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6196
6197 ; Demonstrate usage of llvm.va_copy and llvm.va_end
Jeff Cohen222a8a42007-04-29 01:07:00 +0000[diff] [blame]6198 %aq = alloca i8*
Chris Lattnerdb0790c2007-01-08 07:55:15 +0000[diff] [blame]6199 %aq2 = bitcast i8** %aq to i8*
Jeff Cohen222a8a42007-04-29 01:07:00 +0000[diff] [blame]6200 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
Anton Korobeynikov640bbe02007-03-21 23:58:04 +0000[diff] [blame]6201 call void @llvm.va_end(i8* %aq2)
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6202
6203 ; Stop processing of arguments.
Anton Korobeynikov640bbe02007-03-21 23:58:04 +0000[diff] [blame]6204 call void @llvm.va_end(i8* %ap2)
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000[diff] [blame]6205 ret i32 %tmp
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6206}
Anton Korobeynikov640bbe02007-03-21 23:58:04 +0000[diff] [blame]6207
6208declare void @llvm.va_start(i8*)
6209declare void @llvm.va_copy(i8*, i8*)
6210declare void @llvm.va_end(i8*)
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6211</pre>
Chris Lattner941515c2004-01-06 05:31:32 +0000[diff] [blame]6212
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]6213<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6214<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6215 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6216</h4>
Chris Lattner941515c2004-01-06 05:31:32 +0000[diff] [blame]6217
6218
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6219<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6220
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]6221<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6222<pre>
6223 declare void %llvm.va_start(i8* &lt;arglist&gt;)
6224</pre>
6225
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]6226<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6227<p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*&lt;arglist&gt;</tt>
6228 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +0000[diff] [blame]6229
6230<h5>Arguments:</h5>
Dan Gohmanef9462f2008-10-14 16:51:45 +0000[diff] [blame]6231<p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +0000[diff] [blame]6232
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]6233<h5>Semantics:</h5>
Dan Gohmanef9462f2008-10-14 16:51:45 +0000[diff] [blame]6234<p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6235 macro available in C. In a target-dependent way, it initializes
6236 the <tt>va_list</tt> element to which the argument points, so that the next
6237 call to <tt>va_arg</tt> will produce the first variable argument passed to
6238 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
6239 need to know the last argument of the function as the compiler can figure
6240 that out.</p>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +0000[diff] [blame]6241
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]6242</div>
Chris Lattner941515c2004-01-06 05:31:32 +0000[diff] [blame]6243
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]6244<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6245<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6246 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6247</h4>
Chris Lattner941515c2004-01-06 05:31:32 +0000[diff] [blame]6248
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6249<div>
Chris Lattnerdb0790c2007-01-08 07:55:15 +0000[diff] [blame]6250
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6251<h5>Syntax:</h5>
6252<pre>
6253 declare void @llvm.va_end(i8* &lt;arglist&gt;)
6254</pre>
6255
6256<h5>Overview:</h5>
Jeff Cohen222a8a42007-04-29 01:07:00 +0000[diff] [blame]6257<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6258 which has been initialized previously
6259 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
6260 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
Chris Lattnerdb0790c2007-01-08 07:55:15 +0000[diff] [blame]6261
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]6262<h5>Arguments:</h5>
Jeff Cohen222a8a42007-04-29 01:07:00 +0000[diff] [blame]6263<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
Chris Lattnerdb0790c2007-01-08 07:55:15 +0000[diff] [blame]6264
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]6265<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]6266<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6267 macro available in C. In a target-dependent way, it destroys
6268 the <tt>va_list</tt> element to which the argument points. Calls
6269 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
6270 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
6271 with calls to <tt>llvm.va_end</tt>.</p>
Chris Lattnerdb0790c2007-01-08 07:55:15 +0000[diff] [blame]6272
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]6273</div>
Chris Lattner941515c2004-01-06 05:31:32 +0000[diff] [blame]6274
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]6275<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6276<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6277 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6278</h4>
Chris Lattner941515c2004-01-06 05:31:32 +0000[diff] [blame]6279
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6280<div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6281
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]6282<h5>Syntax:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6283<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +0000[diff] [blame]6284 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6285</pre>
6286
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]6287<h5>Overview:</h5>
Jeff Cohen222a8a42007-04-29 01:07:00 +0000[diff] [blame]6288<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6289 from the source argument list to the destination argument list.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6290
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]6291<h5>Arguments:</h5>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +0000[diff] [blame]6292<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6293 The second argument is a pointer to a <tt>va_list</tt> element to copy
6294 from.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6295
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000[diff] [blame]6296<h5>Semantics:</h5>
Jeff Cohen222a8a42007-04-29 01:07:00 +0000[diff] [blame]6297<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6298 macro available in C. In a target-dependent way, it copies the
6299 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
6300 element. This intrinsic is necessary because
6301 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
6302 arbitrarily complex and require, for example, memory allocation.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6303
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]6304</div>
Chris Lattner941515c2004-01-06 05:31:32 +0000[diff] [blame]6305
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6306</div>
6307
Bill Wendling537603b2011-07-31 06:45:03 +0000[diff] [blame]6308</div>
6309
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6310<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6311<h3>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6312 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6313</h3>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6314
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6315<div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6316
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6317<p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
Chris Lattner67c37d12008-08-05 18:29:16 +0000[diff] [blame]6318Collection</a> (GC) requires the implementation and generation of these
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6319intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
6320roots on the stack</a>, as well as garbage collector implementations that
6321require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
6322barriers. Front-ends for type-safe garbage collected languages should generate
6323these intrinsics to make use of the LLVM garbage collectors. For more details,
6324see <a href="GarbageCollection.html">Accurate Garbage Collection with
6325LLVM</a>.</p>
Christopher Lamb55c6d4f2007-12-17 01:00:21 +0000[diff] [blame]6326
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6327<p>The garbage collection intrinsics only operate on objects in the generic
6328 address space (address space zero).</p>
Christopher Lamb55c6d4f2007-12-17 01:00:21 +0000[diff] [blame]6329
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6330<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6331<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6332 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6333</h4>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6334
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6335<div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6336
6337<h5>Syntax:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6338<pre>
Chris Lattner12477732007-09-21 17:30:40 +0000[diff] [blame]6339 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6340</pre>
6341
6342<h5>Overview:</h5>
John Criswelldfe6a862004-12-10 15:51:16 +0000[diff] [blame]6343<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6344 the code generator, and allows some metadata to be associated with it.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6345
6346<h5>Arguments:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6347<p>The first argument specifies the address of a stack object that contains the
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6348 root pointer. The second pointer (which must be either a constant or a
6349 global value address) contains the meta-data to be associated with the
6350 root.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6351
6352<h5>Semantics:</h5>
Chris Lattner851b7712008-04-24 05:59:56 +0000[diff] [blame]6353<p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6354 location. At compile-time, the code generator generates information to allow
6355 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
6356 intrinsic may only be used in a function which <a href="#gc">specifies a GC
6357 algorithm</a>.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6358
6359</div>
6360
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6361<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6362<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6363 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6364</h4>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6365
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6366<div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6367
6368<h5>Syntax:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6369<pre>
Chris Lattner12477732007-09-21 17:30:40 +0000[diff] [blame]6370 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6371</pre>
6372
6373<h5>Overview:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6374<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6375 locations, allowing garbage collector implementations that require read
6376 barriers.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6377
6378<h5>Arguments:</h5>
Chris Lattnerf9228072006-03-14 20:02:51 +0000[diff] [blame]6379<p>The second argument is the address to read from, which should be an address
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6380 allocated from the garbage collector. The first object is a pointer to the
6381 start of the referenced object, if needed by the language runtime (otherwise
6382 null).</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6383
6384<h5>Semantics:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6385<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6386 instruction, but may be replaced with substantially more complex code by the
6387 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
6388 may only be used in a function which <a href="#gc">specifies a GC
6389 algorithm</a>.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6390
6391</div>
6392
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6393<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6394<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6395 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6396</h4>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6397
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6398<div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6399
6400<h5>Syntax:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6401<pre>
Chris Lattner12477732007-09-21 17:30:40 +0000[diff] [blame]6402 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6403</pre>
6404
6405<h5>Overview:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6406<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6407 locations, allowing garbage collector implementations that require write
6408 barriers (such as generational or reference counting collectors).</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6409
6410<h5>Arguments:</h5>
Chris Lattnerf9228072006-03-14 20:02:51 +0000[diff] [blame]6411<p>The first argument is the reference to store, the second is the start of the
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6412 object to store it to, and the third is the address of the field of Obj to
6413 store to. If the runtime does not require a pointer to the object, Obj may
6414 be null.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6415
6416<h5>Semantics:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6417<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6418 instruction, but may be replaced with substantially more complex code by the
6419 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
6420 may only be used in a function which <a href="#gc">specifies a GC
6421 algorithm</a>.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6422
6423</div>
6424
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6425</div>
6426
Chris Lattner757528b0b2004-05-23 21:06:01 +0000[diff] [blame]6427<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6428<h3>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6429 <a name="int_codegen">Code Generator Intrinsics</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6430</h3>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6431
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6432<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6433
6434<p>These intrinsics are provided by LLVM to expose special features that may
6435 only be implemented with code generator support.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6436
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6437<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6438<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6439 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6440</h4>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6441
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6442<div>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6443
6444<h5>Syntax:</h5>
6445<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +0000[diff] [blame]6446 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6447</pre>
6448
6449<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6450<p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
6451 target-specific value indicating the return address of the current function
6452 or one of its callers.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6453
6454<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6455<p>The argument to this intrinsic indicates which function to return the address
6456 for. Zero indicates the calling function, one indicates its caller, etc.
6457 The argument is <b>required</b> to be a constant integer value.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6458
6459<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6460<p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
6461 indicating the return address of the specified call frame, or zero if it
6462 cannot be identified. The value returned by this intrinsic is likely to be
6463 incorrect or 0 for arguments other than zero, so it should only be used for
6464 debugging purposes.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6465
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6466<p>Note that calling this intrinsic does not prevent function inlining or other
6467 aggressive transformations, so the value returned may not be that of the
6468 obvious source-language caller.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6469
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6470</div>
6471
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6472<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6473<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6474 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6475</h4>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6476
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6477<div>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6478
6479<h5>Syntax:</h5>
6480<pre>
Dan Gohmanaabfdb32010-05-28 17:13:49 +0000[diff] [blame]6481 declare i8* @llvm.frameaddress(i32 &lt;level&gt;)
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6482</pre>
6483
6484<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6485<p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
6486 target-specific frame pointer value for the specified stack frame.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6487
6488<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6489<p>The argument to this intrinsic indicates which function to return the frame
6490 pointer for. Zero indicates the calling function, one indicates its caller,
6491 etc. The argument is <b>required</b> to be a constant integer value.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6492
6493<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6494<p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
6495 indicating the frame address of the specified call frame, or zero if it
6496 cannot be identified. The value returned by this intrinsic is likely to be
6497 incorrect or 0 for arguments other than zero, so it should only be used for
6498 debugging purposes.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6499
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6500<p>Note that calling this intrinsic does not prevent function inlining or other
6501 aggressive transformations, so the value returned may not be that of the
6502 obvious source-language caller.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6503
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6504</div>
6505
Chris Lattnerc8a2c222005-02-28 19:24:19 +0000[diff] [blame]6506<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6507<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6508 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6509</h4>
Chris Lattner2f0f0012006-01-13 02:03:13 +0000[diff] [blame]6510
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6511<div>
Chris Lattner2f0f0012006-01-13 02:03:13 +0000[diff] [blame]6512
6513<h5>Syntax:</h5>
6514<pre>
Dan Gohmanaabfdb32010-05-28 17:13:49 +0000[diff] [blame]6515 declare i8* @llvm.stacksave()
Chris Lattner2f0f0012006-01-13 02:03:13 +0000[diff] [blame]6516</pre>
6517
6518<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6519<p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
6520 of the function stack, for use
6521 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
6522 useful for implementing language features like scoped automatic variable
6523 sized arrays in C99.</p>
Chris Lattner2f0f0012006-01-13 02:03:13 +0000[diff] [blame]6524
6525<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6526<p>This intrinsic returns a opaque pointer value that can be passed
6527 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
6528 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
6529 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
6530 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
6531 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
6532 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
Chris Lattner2f0f0012006-01-13 02:03:13 +0000[diff] [blame]6533
6534</div>
6535
6536<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6537<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6538 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6539</h4>
Chris Lattner2f0f0012006-01-13 02:03:13 +0000[diff] [blame]6540
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6541<div>
Chris Lattner2f0f0012006-01-13 02:03:13 +0000[diff] [blame]6542
6543<h5>Syntax:</h5>
6544<pre>
Dan Gohmanaabfdb32010-05-28 17:13:49 +0000[diff] [blame]6545 declare void @llvm.stackrestore(i8* %ptr)
Chris Lattner2f0f0012006-01-13 02:03:13 +0000[diff] [blame]6546</pre>
6547
6548<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6549<p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
6550 the function stack to the state it was in when the
6551 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
6552 executed. This is useful for implementing language features like scoped
6553 automatic variable sized arrays in C99.</p>
Chris Lattner2f0f0012006-01-13 02:03:13 +0000[diff] [blame]6554
6555<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6556<p>See the description
6557 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
Chris Lattner2f0f0012006-01-13 02:03:13 +0000[diff] [blame]6558
6559</div>
6560
Chris Lattner2f0f0012006-01-13 02:03:13 +0000[diff] [blame]6561<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6562<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6563 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6564</h4>
Chris Lattnerc8a2c222005-02-28 19:24:19 +0000[diff] [blame]6565
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6566<div>
Chris Lattnerc8a2c222005-02-28 19:24:19 +0000[diff] [blame]6567
6568<h5>Syntax:</h5>
6569<pre>
Bruno Cardoso Lopesdc9ff3a2011-06-14 04:58:37 +0000[diff] [blame]6570 declare void @llvm.prefetch(i8* &lt;address&gt;, i32 &lt;rw&gt;, i32 &lt;locality&gt;, i32 &lt;cache type&gt;)
Chris Lattnerc8a2c222005-02-28 19:24:19 +0000[diff] [blame]6571</pre>
6572
6573<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6574<p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
6575 insert a prefetch instruction if supported; otherwise, it is a noop.
6576 Prefetches have no effect on the behavior of the program but can change its
6577 performance characteristics.</p>
Chris Lattnerc8a2c222005-02-28 19:24:19 +0000[diff] [blame]6578
6579<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6580<p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
6581 specifier determining if the fetch should be for a read (0) or write (1),
6582 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
Bruno Cardoso Lopesdc9ff3a2011-06-14 04:58:37 +0000[diff] [blame]6583 locality, to (3) - extremely local keep in cache. The <tt>cache type</tt>
6584 specifies whether the prefetch is performed on the data (1) or instruction (0)
6585 cache. The <tt>rw</tt>, <tt>locality</tt> and <tt>cache type</tt> arguments
6586 must be constant integers.</p>
Chris Lattnerc8a2c222005-02-28 19:24:19 +0000[diff] [blame]6587
6588<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6589<p>This intrinsic does not modify the behavior of the program. In particular,
6590 prefetches cannot trap and do not produce a value. On targets that support
6591 this intrinsic, the prefetch can provide hints to the processor cache for
6592 better performance.</p>
Chris Lattnerc8a2c222005-02-28 19:24:19 +0000[diff] [blame]6593
6594</div>
6595
Andrew Lenharthb4427912005-03-28 20:05:49 +0000[diff] [blame]6596<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6597<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6598 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6599</h4>
Andrew Lenharthb4427912005-03-28 20:05:49 +0000[diff] [blame]6600
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6601<div>
Andrew Lenharthb4427912005-03-28 20:05:49 +0000[diff] [blame]6602
6603<h5>Syntax:</h5>
6604<pre>
Chris Lattner12477732007-09-21 17:30:40 +0000[diff] [blame]6605 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Andrew Lenharthb4427912005-03-28 20:05:49 +0000[diff] [blame]6606</pre>
6607
6608<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6609<p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
6610 Counter (PC) in a region of code to simulators and other tools. The method
6611 is target specific, but it is expected that the marker will use exported
6612 symbols to transmit the PC of the marker. The marker makes no guarantees
6613 that it will remain with any specific instruction after optimizations. It is
6614 possible that the presence of a marker will inhibit optimizations. The
6615 intended use is to be inserted after optimizations to allow correlations of
6616 simulation runs.</p>
Andrew Lenharthb4427912005-03-28 20:05:49 +0000[diff] [blame]6617
6618<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6619<p><tt>id</tt> is a numerical id identifying the marker.</p>
Andrew Lenharthb4427912005-03-28 20:05:49 +0000[diff] [blame]6620
6621<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6622<p>This intrinsic does not modify the behavior of the program. Backends that do
Dan Gohmana269a0a2010-03-01 17:41:39 +0000[diff] [blame]6623 not support this intrinsic may ignore it.</p>
Andrew Lenharthb4427912005-03-28 20:05:49 +0000[diff] [blame]6624
6625</div>
6626
Andrew Lenharth01aa5632005-11-11 16:47:30 +0000[diff] [blame]6627<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6628<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6629 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6630</h4>
Andrew Lenharth01aa5632005-11-11 16:47:30 +0000[diff] [blame]6631
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6632<div>
Andrew Lenharth01aa5632005-11-11 16:47:30 +0000[diff] [blame]6633
6634<h5>Syntax:</h5>
6635<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]6636 declare i64 @llvm.readcyclecounter()
Andrew Lenharth01aa5632005-11-11 16:47:30 +0000[diff] [blame]6637</pre>
6638
6639<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6640<p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
6641 counter register (or similar low latency, high accuracy clocks) on those
6642 targets that support it. On X86, it should map to RDTSC. On Alpha, it
6643 should map to RPCC. As the backing counters overflow quickly (on the order
6644 of 9 seconds on alpha), this should only be used for small timings.</p>
Andrew Lenharth01aa5632005-11-11 16:47:30 +0000[diff] [blame]6645
6646<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6647<p>When directly supported, reading the cycle counter should not modify any
6648 memory. Implementations are allowed to either return a application specific
6649 value or a system wide value. On backends without support, this is lowered
6650 to a constant 0.</p>
Andrew Lenharth01aa5632005-11-11 16:47:30 +0000[diff] [blame]6651
6652</div>
6653
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6654</div>
6655
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6656<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6657<h3>
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6658 <a name="int_libc">Standard C Library Intrinsics</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6659</h3>
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6660
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6661<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6662
6663<p>LLVM provides intrinsics for a few important standard C library functions.
6664 These intrinsics allow source-language front-ends to pass information about
6665 the alignment of the pointer arguments to the code generator, providing
6666 opportunity for more efficient code generation.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6667
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6668<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6669<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6670 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6671</h4>
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6672
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6673<div>
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6674
6675<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6676<p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
Mon P Wang508127b2010-04-07 06:35:53 +0000[diff] [blame]6677 integer bit width and for different address spaces. Not all targets support
6678 all bit widths however.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6679
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6680<pre>
Dan Gohmanaabfdb32010-05-28 17:13:49 +0000[diff] [blame]6681 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* &lt;dest&gt;, i8* &lt;src&gt;,
Chris Lattnerbd4ca622010-04-08 00:53:57 +0000[diff] [blame]6682 i32 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Dan Gohmanaabfdb32010-05-28 17:13:49 +0000[diff] [blame]6683 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* &lt;dest&gt;, i8* &lt;src&gt;,
Chris Lattnerbd4ca622010-04-08 00:53:57 +0000[diff] [blame]6684 i64 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6685</pre>
6686
6687<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6688<p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6689 source location to the destination location.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6690
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6691<p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
Chris Lattnerbd4ca622010-04-08 00:53:57 +0000[diff] [blame]6692 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6693 and the pointers can be in specified address spaces.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6694
6695<h5>Arguments:</h5>
Chris Lattnerbd4ca622010-04-08 00:53:57 +0000[diff] [blame]6696
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6697<p>The first argument is a pointer to the destination, the second is a pointer
6698 to the source. The third argument is an integer argument specifying the
Chris Lattnerbd4ca622010-04-08 00:53:57 +0000[diff] [blame]6699 number of bytes to copy, the fourth argument is the alignment of the
6700 source and destination locations, and the fifth is a boolean indicating a
6701 volatile access.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6702
Dan Gohmana269a0a2010-03-01 17:41:39 +0000[diff] [blame]6703<p>If the call to this intrinsic has an alignment value that is not 0 or 1,
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6704 then the caller guarantees that both the source and destination pointers are
6705 aligned to that boundary.</p>
Chris Lattner4c67c482004-02-12 21:18:15 +0000[diff] [blame]6706
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +0000[diff] [blame]6707<p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6708 <tt>llvm.memcpy</tt> call is a <a href="#volatile">volatile operation</a>.
6709 The detailed access behavior is not very cleanly specified and it is unwise
6710 to depend on it.</p>
Chris Lattnerbd4ca622010-04-08 00:53:57 +0000[diff] [blame]6711
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6712<h5>Semantics:</h5>
Chris Lattnerbd4ca622010-04-08 00:53:57 +0000[diff] [blame]6713
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6714<p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6715 source location to the destination location, which are not allowed to
6716 overlap. It copies "len" bytes of memory over. If the argument is known to
6717 be aligned to some boundary, this can be specified as the fourth argument,
6718 otherwise it should be set to 0 or 1.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6719
Chris Lattnerfee11462004-02-12 17:01:32 +0000[diff] [blame]6720</div>
6721
Chris Lattnerf30152e2004-02-12 18:10:10 +0000[diff] [blame]6722<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6723<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6724 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6725</h4>
Chris Lattnerf30152e2004-02-12 18:10:10 +0000[diff] [blame]6726
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6727<div>
Chris Lattnerf30152e2004-02-12 18:10:10 +0000[diff] [blame]6728
6729<h5>Syntax:</h5>
Chris Lattnerdd708342008-11-21 16:42:48 +0000[diff] [blame]6730<p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
Mon P Wang508127b2010-04-07 06:35:53 +0000[diff] [blame]6731 width and for different address space. Not all targets support all bit
6732 widths however.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6733
Chris Lattnerf30152e2004-02-12 18:10:10 +0000[diff] [blame]6734<pre>
Dan Gohmanaabfdb32010-05-28 17:13:49 +0000[diff] [blame]6735 declare void @llvm.memmove.p0i8.p0i8.i32(i8* &lt;dest&gt;, i8* &lt;src&gt;,
Chris Lattnerbd4ca622010-04-08 00:53:57 +0000[diff] [blame]6736 i32 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Dan Gohmanaabfdb32010-05-28 17:13:49 +0000[diff] [blame]6737 declare void @llvm.memmove.p0i8.p0i8.i64(i8* &lt;dest&gt;, i8* &lt;src&gt;,
Chris Lattnerbd4ca622010-04-08 00:53:57 +0000[diff] [blame]6738 i64 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Chris Lattnerf30152e2004-02-12 18:10:10 +0000[diff] [blame]6739</pre>
6740
6741<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6742<p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
6743 source location to the destination location. It is similar to the
6744 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
6745 overlap.</p>
Chris Lattnerf30152e2004-02-12 18:10:10 +0000[diff] [blame]6746
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6747<p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
Chris Lattnerbd4ca622010-04-08 00:53:57 +0000[diff] [blame]6748 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6749 and the pointers can be in specified address spaces.</p>
Chris Lattnerf30152e2004-02-12 18:10:10 +0000[diff] [blame]6750
6751<h5>Arguments:</h5>
Chris Lattnerbd4ca622010-04-08 00:53:57 +0000[diff] [blame]6752
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6753<p>The first argument is a pointer to the destination, the second is a pointer
6754 to the source. The third argument is an integer argument specifying the
Chris Lattnerbd4ca622010-04-08 00:53:57 +0000[diff] [blame]6755 number of bytes to copy, the fourth argument is the alignment of the
6756 source and destination locations, and the fifth is a boolean indicating a
6757 volatile access.</p>
Chris Lattnerf30152e2004-02-12 18:10:10 +0000[diff] [blame]6758
Dan Gohmana269a0a2010-03-01 17:41:39 +0000[diff] [blame]6759<p>If the call to this intrinsic has an alignment value that is not 0 or 1,
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6760 then the caller guarantees that the source and destination pointers are
6761 aligned to that boundary.</p>
Chris Lattner4c67c482004-02-12 21:18:15 +0000[diff] [blame]6762
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +0000[diff] [blame]6763<p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6764 <tt>llvm.memmove</tt> call is a <a href="#volatile">volatile operation</a>.
6765 The detailed access behavior is not very cleanly specified and it is unwise
6766 to depend on it.</p>
Chris Lattnerbd4ca622010-04-08 00:53:57 +0000[diff] [blame]6767
Chris Lattnerf30152e2004-02-12 18:10:10 +0000[diff] [blame]6768<h5>Semantics:</h5>
Chris Lattnerbd4ca622010-04-08 00:53:57 +0000[diff] [blame]6769
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6770<p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
6771 source location to the destination location, which may overlap. It copies
6772 "len" bytes of memory over. If the argument is known to be aligned to some
6773 boundary, this can be specified as the fourth argument, otherwise it should
6774 be set to 0 or 1.</p>
Chris Lattnerf30152e2004-02-12 18:10:10 +0000[diff] [blame]6775
Chris Lattnerf30152e2004-02-12 18:10:10 +0000[diff] [blame]6776</div>
6777
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6778<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6779<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6780 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6781</h4>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6782
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6783<div>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6784
6785<h5>Syntax:</h5>
Chris Lattnerdd708342008-11-21 16:42:48 +0000[diff] [blame]6786<p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
John Criswellad05ae42010-07-30 16:30:28 +0000[diff] [blame]6787 width and for different address spaces. However, not all targets support all
6788 bit widths.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6789
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6790<pre>
Dan Gohmanaabfdb32010-05-28 17:13:49 +0000[diff] [blame]6791 declare void @llvm.memset.p0i8.i32(i8* &lt;dest&gt;, i8 &lt;val&gt;,
Chris Lattner685db9d2010-04-08 00:54:34 +0000[diff] [blame]6792 i32 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Dan Gohmanaabfdb32010-05-28 17:13:49 +0000[diff] [blame]6793 declare void @llvm.memset.p0i8.i64(i8* &lt;dest&gt;, i8 &lt;val&gt;,
Chris Lattner685db9d2010-04-08 00:54:34 +0000[diff] [blame]6794 i64 &lt;len&gt;, i32 &lt;align&gt;, i1 &lt;isvolatile&gt;)
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6795</pre>
6796
6797<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6798<p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
6799 particular byte value.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6800
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6801<p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
John Criswellad05ae42010-07-30 16:30:28 +0000[diff] [blame]6802 intrinsic does not return a value and takes extra alignment/volatile
6803 arguments. Also, the destination can be in an arbitrary address space.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6804
6805<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6806<p>The first argument is a pointer to the destination to fill, the second is the
John Criswellad05ae42010-07-30 16:30:28 +0000[diff] [blame]6807 byte value with which to fill it, the third argument is an integer argument
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6808 specifying the number of bytes to fill, and the fourth argument is the known
John Criswellad05ae42010-07-30 16:30:28 +0000[diff] [blame]6809 alignment of the destination location.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6810
Dan Gohmana269a0a2010-03-01 17:41:39 +0000[diff] [blame]6811<p>If the call to this intrinsic has an alignment value that is not 0 or 1,
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6812 then the caller guarantees that the destination pointer is aligned to that
6813 boundary.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6814
Jeffrey Yasskin5d284ae2010-04-26 21:21:24 +0000[diff] [blame]6815<p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6816 <tt>llvm.memset</tt> call is a <a href="#volatile">volatile operation</a>.
6817 The detailed access behavior is not very cleanly specified and it is unwise
6818 to depend on it.</p>
Chris Lattnerbd4ca622010-04-08 00:53:57 +0000[diff] [blame]6819
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6820<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6821<p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
6822 at the destination location. If the argument is known to be aligned to some
6823 boundary, this can be specified as the fourth argument, otherwise it should
6824 be set to 0 or 1.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6825
Chris Lattner3649c3a2004-02-14 04:08:35 +0000[diff] [blame]6826</div>
6827
Chris Lattner3b4f4372004-06-11 02:28:03 +0000[diff] [blame]6828<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6829<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6830 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6831</h4>
Chris Lattner8a8f2e52005-07-21 01:29:16 +0000[diff] [blame]6832
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6833<div>
Chris Lattner8a8f2e52005-07-21 01:29:16 +0000[diff] [blame]6834
6835<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6836<p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
6837 floating point or vector of floating point type. Not all targets support all
6838 types however.</p>
6839
Chris Lattner8a8f2e52005-07-21 01:29:16 +0000[diff] [blame]6840<pre>
Dale Johannesendd89d272007-10-02 17:47:38 +0000[diff] [blame]6841 declare float @llvm.sqrt.f32(float %Val)
6842 declare double @llvm.sqrt.f64(double %Val)
6843 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
6844 declare fp128 @llvm.sqrt.f128(fp128 %Val)
6845 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Chris Lattner8a8f2e52005-07-21 01:29:16 +0000[diff] [blame]6846</pre>
6847
6848<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6849<p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
6850 returning the same value as the libm '<tt>sqrt</tt>' functions would.
6851 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
6852 behavior for negative numbers other than -0.0 (which allows for better
6853 optimization, because there is no need to worry about errno being
6854 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
Chris Lattner8a8f2e52005-07-21 01:29:16 +0000[diff] [blame]6855
6856<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6857<p>The argument and return value are floating point numbers of the same
6858 type.</p>
Chris Lattner8a8f2e52005-07-21 01:29:16 +0000[diff] [blame]6859
6860<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6861<p>This function returns the sqrt of the specified operand if it is a
6862 nonnegative floating point number.</p>
Chris Lattner8a8f2e52005-07-21 01:29:16 +0000[diff] [blame]6863
Chris Lattner8a8f2e52005-07-21 01:29:16 +0000[diff] [blame]6864</div>
6865
Chris Lattner33b73f92006-09-08 06:34:02 +0000[diff] [blame]6866<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6867<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]6868 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6869</h4>
Chris Lattner33b73f92006-09-08 06:34:02 +0000[diff] [blame]6870
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6871<div>
Chris Lattner33b73f92006-09-08 06:34:02 +0000[diff] [blame]6872
6873<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6874<p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
6875 floating point or vector of floating point type. Not all targets support all
6876 types however.</p>
6877
Chris Lattner33b73f92006-09-08 06:34:02 +0000[diff] [blame]6878<pre>
Dale Johannesendd89d272007-10-02 17:47:38 +0000[diff] [blame]6879 declare float @llvm.powi.f32(float %Val, i32 %power)
6880 declare double @llvm.powi.f64(double %Val, i32 %power)
6881 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
6882 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
6883 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Chris Lattner33b73f92006-09-08 06:34:02 +0000[diff] [blame]6884</pre>
6885
6886<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6887<p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
6888 specified (positive or negative) power. The order of evaluation of
6889 multiplications is not defined. When a vector of floating point type is
6890 used, the second argument remains a scalar integer value.</p>
Chris Lattner33b73f92006-09-08 06:34:02 +0000[diff] [blame]6891
6892<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6893<p>The second argument is an integer power, and the first is a value to raise to
6894 that power.</p>
Chris Lattner33b73f92006-09-08 06:34:02 +0000[diff] [blame]6895
6896<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6897<p>This function returns the first value raised to the second power with an
6898 unspecified sequence of rounding operations.</p>
Chris Lattner33b73f92006-09-08 06:34:02 +0000[diff] [blame]6899
Chris Lattner33b73f92006-09-08 06:34:02 +0000[diff] [blame]6900</div>
6901
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6902<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6903<h4>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6904 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6905</h4>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6906
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6907<div>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6908
6909<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6910<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
6911 floating point or vector of floating point type. Not all targets support all
6912 types however.</p>
6913
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6914<pre>
6915 declare float @llvm.sin.f32(float %Val)
6916 declare double @llvm.sin.f64(double %Val)
6917 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
6918 declare fp128 @llvm.sin.f128(fp128 %Val)
6919 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
6920</pre>
6921
6922<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6923<p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6924
6925<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6926<p>The argument and return value are floating point numbers of the same
6927 type.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6928
6929<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6930<p>This function returns the sine of the specified operand, returning the same
6931 values as the libm <tt>sin</tt> functions would, and handles error conditions
6932 in the same way.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6933
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6934</div>
6935
6936<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6937<h4>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6938 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6939</h4>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6940
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6941<div>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6942
6943<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6944<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
6945 floating point or vector of floating point type. Not all targets support all
6946 types however.</p>
6947
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6948<pre>
6949 declare float @llvm.cos.f32(float %Val)
6950 declare double @llvm.cos.f64(double %Val)
6951 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
6952 declare fp128 @llvm.cos.f128(fp128 %Val)
6953 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
6954</pre>
6955
6956<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6957<p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6958
6959<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6960<p>The argument and return value are floating point numbers of the same
6961 type.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6962
6963<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6964<p>This function returns the cosine of the specified operand, returning the same
6965 values as the libm <tt>cos</tt> functions would, and handles error conditions
6966 in the same way.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6967
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6968</div>
6969
6970<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6971<h4>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6972 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]6973</h4>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6974
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]6975<div>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6976
6977<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6978<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
6979 floating point or vector of floating point type. Not all targets support all
6980 types however.</p>
6981
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6982<pre>
6983 declare float @llvm.pow.f32(float %Val, float %Power)
6984 declare double @llvm.pow.f64(double %Val, double %Power)
6985 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
6986 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
6987 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
6988</pre>
6989
6990<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6991<p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
6992 specified (positive or negative) power.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6993
6994<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6995<p>The second argument is a floating point power, and the first is a value to
6996 raise to that power.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]6997
6998<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]6999<p>This function returns the first value raised to the second power, returning
7000 the same values as the libm <tt>pow</tt> functions would, and handles error
7001 conditions in the same way.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]7002
Dan Gohmanb6324c12007-10-15 20:30:11 +0000[diff] [blame]7003</div>
7004
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7005</div>
7006
Dan Gohman911fa902011-05-23 21:13:03 +0000[diff] [blame]7007<!-- _______________________________________________________________________ -->
7008<h4>
7009 <a name="int_exp">'<tt>llvm.exp.*</tt>' Intrinsic</a>
7010</h4>
7011
7012<div>
7013
7014<h5>Syntax:</h5>
7015<p>This is an overloaded intrinsic. You can use <tt>llvm.exp</tt> on any
7016 floating point or vector of floating point type. Not all targets support all
7017 types however.</p>
7018
7019<pre>
7020 declare float @llvm.exp.f32(float %Val)
7021 declare double @llvm.exp.f64(double %Val)
7022 declare x86_fp80 @llvm.exp.f80(x86_fp80 %Val)
7023 declare fp128 @llvm.exp.f128(fp128 %Val)
7024 declare ppc_fp128 @llvm.exp.ppcf128(ppc_fp128 %Val)
7025</pre>
7026
7027<h5>Overview:</h5>
7028<p>The '<tt>llvm.exp.*</tt>' intrinsics perform the exp function.</p>
7029
7030<h5>Arguments:</h5>
7031<p>The argument and return value are floating point numbers of the same
7032 type.</p>
7033
7034<h5>Semantics:</h5>
7035<p>This function returns the same values as the libm <tt>exp</tt> functions
7036 would, and handles error conditions in the same way.</p>
7037
7038</div>
7039
7040<!-- _______________________________________________________________________ -->
7041<h4>
7042 <a name="int_log">'<tt>llvm.log.*</tt>' Intrinsic</a>
7043</h4>
7044
7045<div>
7046
7047<h5>Syntax:</h5>
7048<p>This is an overloaded intrinsic. You can use <tt>llvm.log</tt> on any
7049 floating point or vector of floating point type. Not all targets support all
7050 types however.</p>
7051
7052<pre>
7053 declare float @llvm.log.f32(float %Val)
7054 declare double @llvm.log.f64(double %Val)
7055 declare x86_fp80 @llvm.log.f80(x86_fp80 %Val)
7056 declare fp128 @llvm.log.f128(fp128 %Val)
7057 declare ppc_fp128 @llvm.log.ppcf128(ppc_fp128 %Val)
7058</pre>
7059
7060<h5>Overview:</h5>
7061<p>The '<tt>llvm.log.*</tt>' intrinsics perform the log function.</p>
7062
7063<h5>Arguments:</h5>
7064<p>The argument and return value are floating point numbers of the same
7065 type.</p>
7066
7067<h5>Semantics:</h5>
7068<p>This function returns the same values as the libm <tt>log</tt> functions
7069 would, and handles error conditions in the same way.</p>
7070
Cameron Zwarichf03fa182011-07-08 21:39:21 +0000[diff] [blame]7071<h4>
7072 <a name="int_fma">'<tt>llvm.fma.*</tt>' Intrinsic</a>
7073</h4>
7074
7075<div>
7076
7077<h5>Syntax:</h5>
7078<p>This is an overloaded intrinsic. You can use <tt>llvm.fma</tt> on any
7079 floating point or vector of floating point type. Not all targets support all
7080 types however.</p>
7081
7082<pre>
7083 declare float @llvm.fma.f32(float %a, float %b, float %c)
7084 declare double @llvm.fma.f64(double %a, double %b, double %c)
7085 declare x86_fp80 @llvm.fma.f80(x86_fp80 %a, x86_fp80 %b, x86_fp80 %c)
7086 declare fp128 @llvm.fma.f128(fp128 %a, fp128 %b, fp128 %c)
7087 declare ppc_fp128 @llvm.fma.ppcf128(ppc_fp128 %a, ppc_fp128 %b, ppc_fp128 %c)
7088</pre>
7089
7090<h5>Overview:</h5>
Cameron Zwaricha32fd212011-07-08 22:13:55 +0000[diff] [blame]7091<p>The '<tt>llvm.fma.*</tt>' intrinsics perform the fused multiply-add
Cameron Zwarichf03fa182011-07-08 21:39:21 +0000[diff] [blame]7092 operation.</p>
7093
7094<h5>Arguments:</h5>
7095<p>The argument and return value are floating point numbers of the same
7096 type.</p>
7097
7098<h5>Semantics:</h5>
7099<p>This function returns the same values as the libm <tt>fma</tt> functions
7100 would.</p>
7101
Dan Gohman911fa902011-05-23 21:13:03 +0000[diff] [blame]7102</div>
7103
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7104<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7105<h3>
Nate Begeman0f223bb2006-01-13 23:26:38 +0000[diff] [blame]7106 <a name="int_manip">Bit Manipulation Intrinsics</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7107</h3>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7108
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7109<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7110
7111<p>LLVM provides intrinsics for a few important bit manipulation operations.
7112 These allow efficient code generation for some algorithms.</p>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7113
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7114<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7115<h4>
Reid Spencer96a5f022007-04-04 02:42:35 +0000[diff] [blame]7116 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7117</h4>
Nate Begeman0f223bb2006-01-13 23:26:38 +0000[diff] [blame]7118
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7119<div>
Nate Begeman0f223bb2006-01-13 23:26:38 +0000[diff] [blame]7120
7121<h5>Syntax:</h5>
Reid Spencer4eefaab2007-04-01 08:04:23 +0000[diff] [blame]7122<p>This is an overloaded intrinsic function. You can use bswap on any integer
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7123 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
7124
Nate Begeman0f223bb2006-01-13 23:26:38 +0000[diff] [blame]7125<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +0000[diff] [blame]7126 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
7127 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
7128 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Nate Begeman0f223bb2006-01-13 23:26:38 +0000[diff] [blame]7129</pre>
7130
7131<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7132<p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
7133 values with an even number of bytes (positive multiple of 16 bits). These
7134 are useful for performing operations on data that is not in the target's
7135 native byte order.</p>
Nate Begeman0f223bb2006-01-13 23:26:38 +0000[diff] [blame]7136
7137<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7138<p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
7139 and low byte of the input i16 swapped. Similarly,
7140 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
7141 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
7142 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
7143 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
7144 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
7145 more, respectively).</p>
Nate Begeman0f223bb2006-01-13 23:26:38 +0000[diff] [blame]7146
7147</div>
7148
7149<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7150<h4>
Reid Spencerb4f9a6f2006-01-16 21:12:35 +0000[diff] [blame]7151 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7152</h4>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7153
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7154<div>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7155
7156<h5>Syntax:</h5>
Reid Spencer4eefaab2007-04-01 08:04:23 +0000[diff] [blame]7157<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
Owen Anderson2f37bdc2011-07-01 21:52:38 +0000[diff] [blame]7158 width, or on any vector with integer elements. Not all targets support all
7159 bit widths or vector types, however.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7160
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7161<pre>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7162 declare i8 @llvm.ctpop.i8(i8 &lt;src&gt;)
Chandler Carruth7132e002007-08-04 01:51:18 +0000[diff] [blame]7163 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +0000[diff] [blame]7164 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carruth7132e002007-08-04 01:51:18 +0000[diff] [blame]7165 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
7166 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Owen Anderson2f37bdc2011-07-01 21:52:38 +0000[diff] [blame]7167 declare &lt;2 x i32&gt; @llvm.ctpop.v2i32(&lt;2 x i32&gt; &lt;src&gt;)
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7168</pre>
7169
7170<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7171<p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
7172 in a value.</p>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7173
7174<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7175<p>The only argument is the value to be counted. The argument may be of any
Owen Anderson2f37bdc2011-07-01 21:52:38 +0000[diff] [blame]7176 integer type, or a vector with integer elements.
7177 The return type must match the argument type.</p>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7178
7179<h5>Semantics:</h5>
Owen Anderson2f37bdc2011-07-01 21:52:38 +0000[diff] [blame]7180<p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable, or within each
7181 element of a vector.</p>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7182
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7183</div>
7184
7185<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7186<h4>
Chris Lattnerb748c672006-01-16 22:34:14 +0000[diff] [blame]7187 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7188</h4>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7189
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7190<div>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7191
7192<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7193<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
Owen Anderson2f37bdc2011-07-01 21:52:38 +0000[diff] [blame]7194 integer bit width, or any vector whose elements are integers. Not all
7195 targets support all bit widths or vector types, however.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7196
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7197<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +0000[diff] [blame]7198 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
7199 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +0000[diff] [blame]7200 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carruth7132e002007-08-04 01:51:18 +0000[diff] [blame]7201 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
7202 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Owen Anderson2f37bdc2011-07-01 21:52:38 +0000[diff] [blame]7203 declare &lt;2 x i32&gt; @llvm.ctlz.v2i32(&lt;2 x i32&gt; &lt;src;gt)
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7204</pre>
7205
7206<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7207<p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
7208 leading zeros in a variable.</p>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7209
7210<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7211<p>The only argument is the value to be counted. The argument may be of any
Owen Anderson2f37bdc2011-07-01 21:52:38 +0000[diff] [blame]7212 integer type, or any vector type with integer element type.
7213 The return type must match the argument type.</p>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7214
7215<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7216<p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
Owen Anderson2f37bdc2011-07-01 21:52:38 +0000[diff] [blame]7217 zeros in a variable, or within each element of the vector if the operation
7218 is of vector type. If the src == 0 then the result is the size in bits of
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7219 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7220
Andrew Lenharth1d463522005-05-03 18:01:48 +0000[diff] [blame]7221</div>
Chris Lattner3b4f4372004-06-11 02:28:03 +0000[diff] [blame]7222
Chris Lattnerefa20fa2005-05-15 19:39:26 +0000[diff] [blame]7223<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7224<h4>
Chris Lattnerb748c672006-01-16 22:34:14 +0000[diff] [blame]7225 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7226</h4>
Chris Lattnerefa20fa2005-05-15 19:39:26 +0000[diff] [blame]7227
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7228<div>
Chris Lattnerefa20fa2005-05-15 19:39:26 +0000[diff] [blame]7229
7230<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7231<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
Owen Anderson2f37bdc2011-07-01 21:52:38 +0000[diff] [blame]7232 integer bit width, or any vector of integer elements. Not all targets
7233 support all bit widths or vector types, however.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7234
Chris Lattnerefa20fa2005-05-15 19:39:26 +0000[diff] [blame]7235<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +0000[diff] [blame]7236 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
7237 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +0000[diff] [blame]7238 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carruth7132e002007-08-04 01:51:18 +0000[diff] [blame]7239 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
7240 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Owen Anderson2f37bdc2011-07-01 21:52:38 +0000[diff] [blame]7241 declase &lt;2 x i32&gt; @llvm.cttz.v2i32(&lt;2 x i32&gt; &lt;src&gt;)
Chris Lattnerefa20fa2005-05-15 19:39:26 +0000[diff] [blame]7242</pre>
7243
7244<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7245<p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
7246 trailing zeros.</p>
Chris Lattnerefa20fa2005-05-15 19:39:26 +0000[diff] [blame]7247
7248<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7249<p>The only argument is the value to be counted. The argument may be of any
Owen Anderson2f37bdc2011-07-01 21:52:38 +0000[diff] [blame]7250 integer type, or a vectory with integer element type.. The return type
7251 must match the argument type.</p>
Chris Lattnerefa20fa2005-05-15 19:39:26 +0000[diff] [blame]7252
7253<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7254<p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
Owen Anderson2f37bdc2011-07-01 21:52:38 +0000[diff] [blame]7255 zeros in a variable, or within each element of a vector.
7256 If the src == 0 then the result is the size in bits of
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7257 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
Chris Lattnerefa20fa2005-05-15 19:39:26 +0000[diff] [blame]7258
Chris Lattnerefa20fa2005-05-15 19:39:26 +0000[diff] [blame]7259</div>
7260
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7261</div>
7262
Bill Wendlingfd2bd722009-02-08 04:04:40 +0000[diff] [blame]7263<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7264<h3>
Bill Wendlingfd2bd722009-02-08 04:04:40 +0000[diff] [blame]7265 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7266</h3>
Bill Wendlingfd2bd722009-02-08 04:04:40 +0000[diff] [blame]7267
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7268<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7269
7270<p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
Bill Wendlingfd2bd722009-02-08 04:04:40 +0000[diff] [blame]7271
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7272<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7273<h4>
7274 <a name="int_sadd_overflow">
7275 '<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics
7276 </a>
7277</h4>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7278
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7279<div>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7280
7281<h5>Syntax:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7282<p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7283 on any integer bit width.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7284
7285<pre>
7286 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
7287 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
7288 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
7289</pre>
7290
7291<h5>Overview:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7292<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7293 a signed addition of the two arguments, and indicate whether an overflow
7294 occurred during the signed summation.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7295
7296<h5>Arguments:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7297<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7298 be of integer types of any bit width, but they must have the same bit
7299 width. The second element of the result structure must be of
7300 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7301 undergo signed addition.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7302
7303<h5>Semantics:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7304<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7305 a signed addition of the two variables. They return a structure &mdash; the
7306 first element of which is the signed summation, and the second element of
7307 which is a bit specifying if the signed summation resulted in an
7308 overflow.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7309
7310<h5>Examples:</h5>
7311<pre>
7312 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
7313 %sum = extractvalue {i32, i1} %res, 0
7314 %obit = extractvalue {i32, i1} %res, 1
7315 br i1 %obit, label %overflow, label %normal
7316</pre>
7317
7318</div>
7319
7320<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7321<h4>
7322 <a name="int_uadd_overflow">
7323 '<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics
7324 </a>
7325</h4>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7326
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7327<div>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7328
7329<h5>Syntax:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7330<p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7331 on any integer bit width.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7332
7333<pre>
7334 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
7335 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
7336 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
7337</pre>
7338
7339<h5>Overview:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7340<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7341 an unsigned addition of the two arguments, and indicate whether a carry
7342 occurred during the unsigned summation.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7343
7344<h5>Arguments:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7345<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7346 be of integer types of any bit width, but they must have the same bit
7347 width. The second element of the result structure must be of
7348 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7349 undergo unsigned addition.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7350
7351<h5>Semantics:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7352<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7353 an unsigned addition of the two arguments. They return a structure &mdash;
7354 the first element of which is the sum, and the second element of which is a
7355 bit specifying if the unsigned summation resulted in a carry.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7356
7357<h5>Examples:</h5>
7358<pre>
7359 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
7360 %sum = extractvalue {i32, i1} %res, 0
7361 %obit = extractvalue {i32, i1} %res, 1
7362 br i1 %obit, label %carry, label %normal
7363</pre>
7364
7365</div>
7366
7367<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7368<h4>
7369 <a name="int_ssub_overflow">
7370 '<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics
7371 </a>
7372</h4>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7373
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7374<div>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7375
7376<h5>Syntax:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7377<p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7378 on any integer bit width.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7379
7380<pre>
7381 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
7382 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
7383 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
7384</pre>
7385
7386<h5>Overview:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7387<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7388 a signed subtraction of the two arguments, and indicate whether an overflow
7389 occurred during the signed subtraction.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7390
7391<h5>Arguments:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7392<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7393 be of integer types of any bit width, but they must have the same bit
7394 width. The second element of the result structure must be of
7395 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7396 undergo signed subtraction.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7397
7398<h5>Semantics:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7399<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7400 a signed subtraction of the two arguments. They return a structure &mdash;
7401 the first element of which is the subtraction, and the second element of
7402 which is a bit specifying if the signed subtraction resulted in an
7403 overflow.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7404
7405<h5>Examples:</h5>
7406<pre>
7407 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
7408 %sum = extractvalue {i32, i1} %res, 0
7409 %obit = extractvalue {i32, i1} %res, 1
7410 br i1 %obit, label %overflow, label %normal
7411</pre>
7412
7413</div>
7414
7415<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7416<h4>
7417 <a name="int_usub_overflow">
7418 '<tt>llvm.usub.with.overflow.*</tt>' Intrinsics
7419 </a>
7420</h4>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7421
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7422<div>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7423
7424<h5>Syntax:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7425<p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7426 on any integer bit width.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7427
7428<pre>
7429 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
7430 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
7431 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
7432</pre>
7433
7434<h5>Overview:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7435<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7436 an unsigned subtraction of the two arguments, and indicate whether an
7437 overflow occurred during the unsigned subtraction.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7438
7439<h5>Arguments:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7440<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7441 be of integer types of any bit width, but they must have the same bit
7442 width. The second element of the result structure must be of
7443 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7444 undergo unsigned subtraction.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7445
7446<h5>Semantics:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7447<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7448 an unsigned subtraction of the two arguments. They return a structure &mdash;
7449 the first element of which is the subtraction, and the second element of
7450 which is a bit specifying if the unsigned subtraction resulted in an
7451 overflow.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7452
7453<h5>Examples:</h5>
7454<pre>
7455 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
7456 %sum = extractvalue {i32, i1} %res, 0
7457 %obit = extractvalue {i32, i1} %res, 1
7458 br i1 %obit, label %overflow, label %normal
7459</pre>
7460
7461</div>
7462
7463<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7464<h4>
7465 <a name="int_smul_overflow">
7466 '<tt>llvm.smul.with.overflow.*</tt>' Intrinsics
7467 </a>
7468</h4>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7469
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7470<div>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7471
7472<h5>Syntax:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7473<p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7474 on any integer bit width.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7475
7476<pre>
7477 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
7478 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
7479 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
7480</pre>
7481
7482<h5>Overview:</h5>
7483
7484<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7485 a signed multiplication of the two arguments, and indicate whether an
7486 overflow occurred during the signed multiplication.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7487
7488<h5>Arguments:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7489<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7490 be of integer types of any bit width, but they must have the same bit
7491 width. The second element of the result structure must be of
7492 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7493 undergo signed multiplication.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7494
7495<h5>Semantics:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7496<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7497 a signed multiplication of the two arguments. They return a structure &mdash;
7498 the first element of which is the multiplication, and the second element of
7499 which is a bit specifying if the signed multiplication resulted in an
7500 overflow.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000[diff] [blame]7501
7502<h5>Examples:</h5>
7503<pre>
7504 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
7505 %sum = extractvalue {i32, i1} %res, 0
7506 %obit = extractvalue {i32, i1} %res, 1
7507 br i1 %obit, label %overflow, label %normal
7508</pre>
7509
Reid Spencer5bf54c82007-04-11 23:23:49 +0000[diff] [blame]7510</div>
7511
Bill Wendlingb9a73272009-02-08 23:00:09 +0000[diff] [blame]7512<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7513<h4>
7514 <a name="int_umul_overflow">
7515 '<tt>llvm.umul.with.overflow.*</tt>' Intrinsics
7516 </a>
7517</h4>
Bill Wendlingb9a73272009-02-08 23:00:09 +0000[diff] [blame]7518
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7519<div>
Bill Wendlingb9a73272009-02-08 23:00:09 +0000[diff] [blame]7520
7521<h5>Syntax:</h5>
Bill Wendlingb9a73272009-02-08 23:00:09 +0000[diff] [blame]7522<p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7523 on any integer bit width.</p>
Bill Wendlingb9a73272009-02-08 23:00:09 +0000[diff] [blame]7524
7525<pre>
7526 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
7527 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
7528 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
7529</pre>
7530
7531<h5>Overview:</h5>
Bill Wendlingb9a73272009-02-08 23:00:09 +0000[diff] [blame]7532<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7533 a unsigned multiplication of the two arguments, and indicate whether an
7534 overflow occurred during the unsigned multiplication.</p>
Bill Wendlingb9a73272009-02-08 23:00:09 +0000[diff] [blame]7535
7536<h5>Arguments:</h5>
Bill Wendlingb9a73272009-02-08 23:00:09 +0000[diff] [blame]7537<p>The arguments (%a and %b) and the first element of the result structure may
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7538 be of integer types of any bit width, but they must have the same bit
7539 width. The second element of the result structure must be of
7540 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7541 undergo unsigned multiplication.</p>
Bill Wendlingb9a73272009-02-08 23:00:09 +0000[diff] [blame]7542
7543<h5>Semantics:</h5>
Bill Wendlingb9a73272009-02-08 23:00:09 +0000[diff] [blame]7544<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7545 an unsigned multiplication of the two arguments. They return a structure
7546 &mdash; the first element of which is the multiplication, and the second
7547 element of which is a bit specifying if the unsigned multiplication resulted
7548 in an overflow.</p>
Bill Wendlingb9a73272009-02-08 23:00:09 +0000[diff] [blame]7549
7550<h5>Examples:</h5>
7551<pre>
7552 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
7553 %sum = extractvalue {i32, i1} %res, 0
7554 %obit = extractvalue {i32, i1} %res, 1
7555 br i1 %obit, label %overflow, label %normal
7556</pre>
7557
7558</div>
7559
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7560</div>
7561
Chris Lattner941515c2004-01-06 05:31:32 +0000[diff] [blame]7562<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7563<h3>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +0000[diff] [blame]7564 <a name="int_fp16">Half Precision Floating Point Intrinsics</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7565</h3>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +0000[diff] [blame]7566
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7567<div>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +0000[diff] [blame]7568
Chris Lattner022a9fb2010-03-15 04:12:21 +0000[diff] [blame]7569<p>Half precision floating point is a storage-only format. This means that it is
7570 a dense encoding (in memory) but does not support computation in the
7571 format.</p>
Chris Lattnerbbd8bd32010-03-14 23:03:31 +0000[diff] [blame]7572
Chris Lattner022a9fb2010-03-15 04:12:21 +0000[diff] [blame]7573<p>This means that code must first load the half-precision floating point
Chris Lattnerbbd8bd32010-03-14 23:03:31 +0000[diff] [blame]7574 value as an i16, then convert it to float with <a
7575 href="#int_convert_from_fp16"><tt>llvm.convert.from.fp16</tt></a>.
7576 Computation can then be performed on the float value (including extending to
Chris Lattner022a9fb2010-03-15 04:12:21 +0000[diff] [blame]7577 double etc). To store the value back to memory, it is first converted to
7578 float if needed, then converted to i16 with
Chris Lattnerbbd8bd32010-03-14 23:03:31 +0000[diff] [blame]7579 <a href="#int_convert_to_fp16"><tt>llvm.convert.to.fp16</tt></a>, then
7580 storing as an i16 value.</p>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +0000[diff] [blame]7581
7582<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7583<h4>
7584 <a name="int_convert_to_fp16">
7585 '<tt>llvm.convert.to.fp16</tt>' Intrinsic
7586 </a>
7587</h4>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +0000[diff] [blame]7588
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7589<div>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +0000[diff] [blame]7590
7591<h5>Syntax:</h5>
7592<pre>
7593 declare i16 @llvm.convert.to.fp16(f32 %a)
7594</pre>
7595
7596<h5>Overview:</h5>
7597<p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
7598 a conversion from single precision floating point format to half precision
7599 floating point format.</p>
7600
7601<h5>Arguments:</h5>
7602<p>The intrinsic function contains single argument - the value to be
7603 converted.</p>
7604
7605<h5>Semantics:</h5>
7606<p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
7607 a conversion from single precision floating point format to half precision
Chris Lattner022a9fb2010-03-15 04:12:21 +0000[diff] [blame]7608 floating point format. The return value is an <tt>i16</tt> which
Chris Lattnerbbd8bd32010-03-14 23:03:31 +0000[diff] [blame]7609 contains the converted number.</p>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +0000[diff] [blame]7610
7611<h5>Examples:</h5>
7612<pre>
7613 %res = call i16 @llvm.convert.to.fp16(f32 %a)
7614 store i16 %res, i16* @x, align 2
7615</pre>
7616
7617</div>
7618
7619<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7620<h4>
7621 <a name="int_convert_from_fp16">
7622 '<tt>llvm.convert.from.fp16</tt>' Intrinsic
7623 </a>
7624</h4>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +0000[diff] [blame]7625
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7626<div>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +0000[diff] [blame]7627
7628<h5>Syntax:</h5>
7629<pre>
7630 declare f32 @llvm.convert.from.fp16(i16 %a)
7631</pre>
7632
7633<h5>Overview:</h5>
7634<p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs
7635 a conversion from half precision floating point format to single precision
7636 floating point format.</p>
7637
7638<h5>Arguments:</h5>
7639<p>The intrinsic function contains single argument - the value to be
7640 converted.</p>
7641
7642<h5>Semantics:</h5>
7643<p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs a
Chris Lattner022a9fb2010-03-15 04:12:21 +0000[diff] [blame]7644 conversion from half single precision floating point format to single
Chris Lattnerbbd8bd32010-03-14 23:03:31 +0000[diff] [blame]7645 precision floating point format. The input half-float value is represented by
7646 an <tt>i16</tt> value.</p>
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +0000[diff] [blame]7647
7648<h5>Examples:</h5>
7649<pre>
7650 %a = load i16* @x, align 2
7651 %res = call f32 @llvm.convert.from.fp16(i16 %a)
7652</pre>
7653
7654</div>
7655
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7656</div>
7657
Anton Korobeynikovcd4dd9c2010-03-14 18:42:47 +0000[diff] [blame]7658<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7659<h3>
Chris Lattner941515c2004-01-06 05:31:32 +0000[diff] [blame]7660 <a name="int_debugger">Debugger Intrinsics</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7661</h3>
Chris Lattner941515c2004-01-06 05:31:32 +0000[diff] [blame]7662
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7663<div>
Chris Lattner941515c2004-01-06 05:31:32 +0000[diff] [blame]7664
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7665<p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
7666 prefix), are described in
7667 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
7668 Level Debugging</a> document.</p>
7669
7670</div>
Chris Lattner941515c2004-01-06 05:31:32 +0000[diff] [blame]7671
Jim Laskey2211f492007-03-14 19:31:19 +0000[diff] [blame]7672<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7673<h3>
Jim Laskey2211f492007-03-14 19:31:19 +0000[diff] [blame]7674 <a name="int_eh">Exception Handling Intrinsics</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7675</h3>
Jim Laskey2211f492007-03-14 19:31:19 +0000[diff] [blame]7676
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7677<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7678
7679<p>The LLVM exception handling intrinsics (which all start with
7680 <tt>llvm.eh.</tt> prefix), are described in
7681 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
7682 Handling</a> document.</p>
7683
Jim Laskey2211f492007-03-14 19:31:19 +0000[diff] [blame]7684</div>
7685
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000[diff] [blame]7686<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7687<h3>
Duncan Sandsa0984362011-09-06 13:37:06 +0000[diff] [blame]7688 <a name="int_trampoline">Trampoline Intrinsics</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7689</h3>
Duncan Sands644f9172007-07-27 12:58:54 +0000[diff] [blame]7690
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7691<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7692
Duncan Sandsa0984362011-09-06 13:37:06 +0000[diff] [blame]7693<p>These intrinsics make it possible to excise one parameter, marked with
Dan Gohman3770af52010-07-02 23:18:08 +0000[diff] [blame]7694 the <a href="#nest"><tt>nest</tt></a> attribute, from a function.
7695 The result is a callable
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7696 function pointer lacking the nest parameter - the caller does not need to
7697 provide a value for it. Instead, the value to use is stored in advance in a
7698 "trampoline", a block of memory usually allocated on the stack, which also
7699 contains code to splice the nest value into the argument list. This is used
7700 to implement the GCC nested function address extension.</p>
7701
7702<p>For example, if the function is
7703 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
7704 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
7705 follows:</p>
7706
Benjamin Kramer79698be2010-07-13 12:26:09 +0000[diff] [blame]7707<pre class="doc_code">
Duncan Sands86e01192007-09-11 14:10:23 +0000[diff] [blame]7708 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
7709 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
Duncan Sandsa0984362011-09-06 13:37:06 +0000[diff] [blame]7710 call i8* @llvm.init.trampoline(i8* %tramp1, i8* bitcast (i32 (i8*, i32, i32)* @f to i8*), i8* %nval)
7711 %p = call i8* @llvm.adjust.trampoline(i8* %tramp1)
Duncan Sands86e01192007-09-11 14:10:23 +0000[diff] [blame]7712 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands644f9172007-07-27 12:58:54 +0000[diff] [blame]7713</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7714
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]7715<p>The call <tt>%val = call i32 %fp(i32 %x, i32 %y)</tt> is then equivalent
7716 to <tt>%val = call i32 %f(i8* %nval, i32 %x, i32 %y)</tt>.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7717
Duncan Sands644f9172007-07-27 12:58:54 +0000[diff] [blame]7718<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7719<h4>
7720 <a name="int_it">
7721 '<tt>llvm.init.trampoline</tt>' Intrinsic
7722 </a>
7723</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7724
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7725<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7726
Duncan Sands644f9172007-07-27 12:58:54 +0000[diff] [blame]7727<h5>Syntax:</h5>
7728<pre>
Duncan Sandsa0984362011-09-06 13:37:06 +0000[diff] [blame]7729 declare void @llvm.init.trampoline(i8* &lt;tramp&gt;, i8* &lt;func&gt;, i8* &lt;nval&gt;)
Duncan Sands644f9172007-07-27 12:58:54 +0000[diff] [blame]7730</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7731
Duncan Sands644f9172007-07-27 12:58:54 +0000[diff] [blame]7732<h5>Overview:</h5>
Duncan Sandsa0984362011-09-06 13:37:06 +0000[diff] [blame]7733<p>This fills the memory pointed to by <tt>tramp</tt> with executable code,
7734 turning it into a trampoline.</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7735
Duncan Sands644f9172007-07-27 12:58:54 +0000[diff] [blame]7736<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7737<p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
7738 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
7739 sufficiently aligned block of memory; this memory is written to by the
7740 intrinsic. Note that the size and the alignment are target-specific - LLVM
7741 currently provides no portable way of determining them, so a front-end that
7742 generates this intrinsic needs to have some target-specific knowledge.
7743 The <tt>func</tt> argument must hold a function bitcast to
7744 an <tt>i8*</tt>.</p>
7745
Duncan Sands644f9172007-07-27 12:58:54 +0000[diff] [blame]7746<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7747<p>The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sandsa0984362011-09-06 13:37:06 +0000[diff] [blame]7748 dependent code, turning it into a function. Then <tt>tramp</tt> needs to be
7749 passed to <a href="#int_at">llvm.adjust.trampoline</a> to get a pointer
7750 which can be <a href="#int_trampoline">bitcast (to a new function) and
7751 called</a>. The new function's signature is the same as that of
7752 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
7753 removed. At most one such <tt>nest</tt> argument is allowed, and it must be of
7754 pointer type. Calling the new function is equivalent to calling <tt>func</tt>
7755 with the same argument list, but with <tt>nval</tt> used for the missing
7756 <tt>nest</tt> argument. If, after calling <tt>llvm.init.trampoline</tt>, the
7757 memory pointed to by <tt>tramp</tt> is modified, then the effect of any later call
7758 to the returned function pointer is undefined.</p>
7759</div>
7760
7761<!-- _______________________________________________________________________ -->
7762<h4>
7763 <a name="int_at">
7764 '<tt>llvm.adjust.trampoline</tt>' Intrinsic
7765 </a>
7766</h4>
7767
7768<div>
7769
7770<h5>Syntax:</h5>
7771<pre>
7772 declare i8* @llvm.adjust.trampoline(i8* &lt;tramp&gt;)
7773</pre>
7774
7775<h5>Overview:</h5>
7776<p>This performs any required machine-specific adjustment to the address of a
7777 trampoline (passed as <tt>tramp</tt>).</p>
7778
7779<h5>Arguments:</h5>
7780<p><tt>tramp</tt> must point to a block of memory which already has trampoline code
7781 filled in by a previous call to <a href="#int_it"><tt>llvm.init.trampoline</tt>
7782 </a>.</p>
7783
7784<h5>Semantics:</h5>
7785<p>On some architectures the address of the code to be executed needs to be
7786 different to the address where the trampoline is actually stored. This
7787 intrinsic returns the executable address corresponding to <tt>tramp</tt>
7788 after performing the required machine specific adjustments.
7789 The pointer returned can then be <a href="#int_trampoline"> bitcast and
7790 executed</a>.
7791</p>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7792
Duncan Sands644f9172007-07-27 12:58:54 +0000[diff] [blame]7793</div>
7794
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7795</div>
7796
Duncan Sands644f9172007-07-27 12:58:54 +0000[diff] [blame]7797<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7798<h3>
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]7799 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7800</h3>
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]7801
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7802<div>
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]7803
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7804<p>These intrinsic functions expand the "universal IR" of LLVM to represent
7805 hardware constructs for atomic operations and memory synchronization. This
7806 provides an interface to the hardware, not an interface to the programmer. It
7807 is aimed at a low enough level to allow any programming models or APIs
7808 (Application Programming Interfaces) which need atomic behaviors to map
7809 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
7810 hardware provides a "universal IR" for source languages, it also provides a
7811 starting point for developing a "universal" atomic operation and
7812 synchronization IR.</p>
7813
7814<p>These do <em>not</em> form an API such as high-level threading libraries,
7815 software transaction memory systems, atomic primitives, and intrinsic
7816 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
7817 application libraries. The hardware interface provided by LLVM should allow
7818 a clean implementation of all of these APIs and parallel programming models.
7819 No one model or paradigm should be selected above others unless the hardware
7820 itself ubiquitously does so.</p>
7821
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]7822<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7823<h4>
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]7824 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7825</h4>
7826
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7827<div>
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]7828<h5>Syntax:</h5>
7829<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]7830 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 +0000[diff] [blame]7831</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7832
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]7833<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7834<p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
7835 specific pairs of memory access types.</p>
7836
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]7837<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7838<p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
7839 The first four arguments enables a specific barrier as listed below. The
Dan Gohmana269a0a2010-03-01 17:41:39 +0000[diff] [blame]7840 fifth argument specifies that the barrier applies to io or device or uncached
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7841 memory.</p>
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]7842
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7843<ul>
7844 <li><tt>ll</tt>: load-load barrier</li>
7845 <li><tt>ls</tt>: load-store barrier</li>
7846 <li><tt>sl</tt>: store-load barrier</li>
7847 <li><tt>ss</tt>: store-store barrier</li>
7848 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
7849</ul>
7850
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]7851<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7852<p>This intrinsic causes the system to enforce some ordering constraints upon
7853 the loads and stores of the program. This barrier does not
7854 indicate <em>when</em> any events will occur, it only enforces
7855 an <em>order</em> in which they occur. For any of the specified pairs of load
7856 and store operations (f.ex. load-load, or store-load), all of the first
7857 operations preceding the barrier will complete before any of the second
7858 operations succeeding the barrier begin. Specifically the semantics for each
7859 pairing is as follows:</p>
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]7860
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7861<ul>
7862 <li><tt>ll</tt>: All loads before the barrier must complete before any load
7863 after the barrier begins.</li>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]7864 <li><tt>ls</tt>: All loads before the barrier must complete before any
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7865 store after the barrier begins.</li>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]7866 <li><tt>ss</tt>: All stores before the barrier must complete before any
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7867 store after the barrier begins.</li>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]7868 <li><tt>sl</tt>: All stores before the barrier must complete before any
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7869 load after the barrier begins.</li>
7870</ul>
7871
7872<p>These semantics are applied with a logical "and" behavior when more than one
7873 is enabled in a single memory barrier intrinsic.</p>
7874
7875<p>Backends may implement stronger barriers than those requested when they do
7876 not support as fine grained a barrier as requested. Some architectures do
7877 not need all types of barriers and on such architectures, these become
7878 noops.</p>
7879
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]7880<h5>Example:</h5>
7881<pre>
Victor Hernandeza70c6df2009-10-26 23:44:29 +0000[diff] [blame]7882%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7883%ptr = bitcast i8* %mallocP to i32*
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]7884 store i32 4, %ptr
7885
7886%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
Evan Cheng0ac49c62011-06-29 17:14:00 +0000[diff] [blame]7887 call void @llvm.memory.barrier(i1 false, i1 true, i1 false, i1 false, i1 true)
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]7888 <i>; guarantee the above finishes</i>
7889 store i32 8, %ptr <i>; before this begins</i>
7890</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7891
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]7892</div>
7893
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7894<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7895<h4>
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]7896 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7897</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7898
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7899<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7900
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7901<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7902<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
7903 any integer bit width and for different address spaces. Not all targets
7904 support all bit widths however.</p>
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7905
7906<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]7907 declare i8 @llvm.atomic.cmp.swap.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt;)
7908 declare i16 @llvm.atomic.cmp.swap.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt;)
7909 declare i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt;)
7910 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 +0000[diff] [blame]7911</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7912
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7913<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7914<p>This loads a value in memory and compares it to a given value. If they are
7915 equal, it stores a new value into the memory.</p>
7916
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7917<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7918<p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
7919 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
7920 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
7921 this integer type. While any bit width integer may be used, targets may only
7922 lower representations they support in hardware.</p>
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7923
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7924<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7925<p>This entire intrinsic must be executed atomically. It first loads the value
7926 in memory pointed to by <tt>ptr</tt> and compares it with the
7927 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
7928 memory. The loaded value is yielded in all cases. This provides the
7929 equivalent of an atomic compare-and-swap operation within the SSA
7930 framework.</p>
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7931
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7932<h5>Examples:</h5>
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7933<pre>
Victor Hernandeza70c6df2009-10-26 23:44:29 +0000[diff] [blame]7934%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7935%ptr = bitcast i8* %mallocP to i32*
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7936 store i32 4, %ptr
7937
7938%val1 = add i32 4, 4
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]7939%result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 4, %val1)
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7940 <i>; yields {i32}:result1 = 4</i>
7941%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7942%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7943
7944%val2 = add i32 1, 1
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]7945%result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 5, %val2)
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7946 <i>; yields {i32}:result2 = 8</i>
7947%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
7948
7949%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
7950</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7951
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7952</div>
7953
7954<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7955<h4>
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7956 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]7957</h4>
7958
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]7959<div>
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7960<h5>Syntax:</h5>
7961
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7962<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
7963 integer bit width. Not all targets support all bit widths however.</p>
7964
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7965<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]7966 declare i8 @llvm.atomic.swap.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;val&gt;)
7967 declare i16 @llvm.atomic.swap.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;val&gt;)
7968 declare i32 @llvm.atomic.swap.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;val&gt;)
7969 declare i64 @llvm.atomic.swap.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;val&gt;)
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7970</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7971
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7972<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7973<p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
7974 the value from memory. It then stores the value in <tt>val</tt> in the memory
7975 at <tt>ptr</tt>.</p>
7976
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7977<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7978<p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
7979 the <tt>val</tt> argument and the result must be integers of the same bit
7980 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
7981 integer type. The targets may only lower integer representations they
7982 support.</p>
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7983
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7984<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]7985<p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
7986 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
7987 equivalent of an atomic swap operation within the SSA framework.</p>
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7988
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7989<h5>Examples:</h5>
7990<pre>
Victor Hernandeza70c6df2009-10-26 23:44:29 +0000[diff] [blame]7991%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7992%ptr = bitcast i8* %mallocP to i32*
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7993 store i32 4, %ptr
7994
7995%val1 = add i32 4, 4
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]7996%result1 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val1)
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]7997 <i>; yields {i32}:result1 = 4</i>
7998%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7999%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
8000
8001%val2 = add i32 1, 1
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8002%result2 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val2)
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]8003 <i>; yields {i32}:result2 = 8</i>
8004
8005%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
8006%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
8007</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8008
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]8009</div>
8010
8011<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8012<h4>
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8013 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8014</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8015
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8016<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8017
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]8018<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8019<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
8020 any integer bit width. Not all targets support all bit widths however.</p>
8021
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]8022<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8023 declare i8 @llvm.atomic.load.add.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
8024 declare i16 @llvm.atomic.load.add.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
8025 declare i32 @llvm.atomic.load.add.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
8026 declare i64 @llvm.atomic.load.add.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]8027</pre>
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]8028
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8029<h5>Overview:</h5>
8030<p>This intrinsic adds <tt>delta</tt> to the value stored in memory
8031 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
8032
8033<h5>Arguments:</h5>
8034<p>The intrinsic takes two arguments, the first a pointer to an integer value
8035 and the second an integer value. The result is also an integer value. These
8036 integer types can have any bit width, but they must all have the same bit
8037 width. The targets may only lower integer representations they support.</p>
8038
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]8039<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8040<p>This intrinsic does a series of operations atomically. It first loads the
8041 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
8042 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]8043
8044<h5>Examples:</h5>
8045<pre>
Victor Hernandeza70c6df2009-10-26 23:44:29 +0000[diff] [blame]8046%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
8047%ptr = bitcast i8* %mallocP to i32*
8048 store i32 4, %ptr
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8049%result1 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 4)
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]8050 <i>; yields {i32}:result1 = 4</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8051%result2 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 2)
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]8052 <i>; yields {i32}:result2 = 8</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8053%result3 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 5)
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]8054 <i>; yields {i32}:result3 = 10</i>
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8055%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]8056</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8057
Andrew Lenharth95528942008-02-21 06:45:13 +0000[diff] [blame]8058</div>
8059
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8060<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8061<h4>
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8062 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8063</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8064
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8065<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8066
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8067<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8068<p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
8069 any integer bit width and for different address spaces. Not all targets
8070 support all bit widths however.</p>
8071
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8072<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8073 declare i8 @llvm.atomic.load.sub.i8.p0i32(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
8074 declare i16 @llvm.atomic.load.sub.i16.p0i32(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
8075 declare i32 @llvm.atomic.load.sub.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
8076 declare i64 @llvm.atomic.load.sub.i64.p0i32(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8077</pre>
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8078
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8079<h5>Overview:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]8080<p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8081 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
8082
8083<h5>Arguments:</h5>
8084<p>The intrinsic takes two arguments, the first a pointer to an integer value
8085 and the second an integer value. The result is also an integer value. These
8086 integer types can have any bit width, but they must all have the same bit
8087 width. The targets may only lower integer representations they support.</p>
8088
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8089<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8090<p>This intrinsic does a series of operations atomically. It first loads the
8091 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
8092 result to <tt>ptr</tt>. It yields the original value stored
8093 at <tt>ptr</tt>.</p>
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8094
8095<h5>Examples:</h5>
8096<pre>
Victor Hernandeza70c6df2009-10-26 23:44:29 +0000[diff] [blame]8097%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
8098%ptr = bitcast i8* %mallocP to i32*
8099 store i32 8, %ptr
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8100%result1 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 4)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8101 <i>; yields {i32}:result1 = 8</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8102%result2 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 2)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8103 <i>; yields {i32}:result2 = 4</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8104%result3 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 5)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8105 <i>; yields {i32}:result3 = 2</i>
8106%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
8107</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8108
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8109</div>
8110
8111<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8112<h4>
8113 <a name="int_atomic_load_and">
8114 '<tt>llvm.atomic.load.and.*</tt>' Intrinsic
8115 </a>
8116 <br>
8117 <a name="int_atomic_load_nand">
8118 '<tt>llvm.atomic.load.nand.*</tt>' Intrinsic
8119 </a>
8120 <br>
8121 <a name="int_atomic_load_or">
8122 '<tt>llvm.atomic.load.or.*</tt>' Intrinsic
8123 </a>
8124 <br>
8125 <a name="int_atomic_load_xor">
8126 '<tt>llvm.atomic.load.xor.*</tt>' Intrinsic
8127 </a>
8128</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8129
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8130<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8131
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8132<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8133<p>These are overloaded intrinsics. You can
8134 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
8135 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
8136 bit width and for different address spaces. Not all targets support all bit
8137 widths however.</p>
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8138
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8139<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8140 declare i8 @llvm.atomic.load.and.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
8141 declare i16 @llvm.atomic.load.and.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
8142 declare i32 @llvm.atomic.load.and.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
8143 declare i64 @llvm.atomic.load.and.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8144</pre>
8145
8146<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8147 declare i8 @llvm.atomic.load.or.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
8148 declare i16 @llvm.atomic.load.or.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
8149 declare i32 @llvm.atomic.load.or.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
8150 declare i64 @llvm.atomic.load.or.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8151</pre>
8152
8153<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8154 declare i8 @llvm.atomic.load.nand.i8.p0i32(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
8155 declare i16 @llvm.atomic.load.nand.i16.p0i32(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
8156 declare i32 @llvm.atomic.load.nand.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
8157 declare i64 @llvm.atomic.load.nand.i64.p0i32(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8158</pre>
8159
8160<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8161 declare i8 @llvm.atomic.load.xor.i8.p0i32(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
8162 declare i16 @llvm.atomic.load.xor.i16.p0i32(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
8163 declare i32 @llvm.atomic.load.xor.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
8164 declare i64 @llvm.atomic.load.xor.i64.p0i32(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8165</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8166
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8167<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8168<p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
8169 the value stored in memory at <tt>ptr</tt>. It yields the original value
8170 at <tt>ptr</tt>.</p>
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8171
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8172<h5>Arguments:</h5>
8173<p>These intrinsics take two arguments, the first a pointer to an integer value
8174 and the second an integer value. The result is also an integer value. These
8175 integer types can have any bit width, but they must all have the same bit
8176 width. The targets may only lower integer representations they support.</p>
8177
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8178<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8179<p>These intrinsics does a series of operations atomically. They first load the
8180 value stored at <tt>ptr</tt>. They then do the bitwise
8181 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
8182 original value stored at <tt>ptr</tt>.</p>
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8183
8184<h5>Examples:</h5>
8185<pre>
Victor Hernandeza70c6df2009-10-26 23:44:29 +0000[diff] [blame]8186%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
8187%ptr = bitcast i8* %mallocP to i32*
8188 store i32 0x0F0F, %ptr
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8189%result0 = call i32 @llvm.atomic.load.nand.i32.p0i32(i32* %ptr, i32 0xFF)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8190 <i>; yields {i32}:result0 = 0x0F0F</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8191%result1 = call i32 @llvm.atomic.load.and.i32.p0i32(i32* %ptr, i32 0xFF)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8192 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8193%result2 = call i32 @llvm.atomic.load.or.i32.p0i32(i32* %ptr, i32 0F)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8194 <i>; yields {i32}:result2 = 0xF0</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8195%result3 = call i32 @llvm.atomic.load.xor.i32.p0i32(i32* %ptr, i32 0F)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8196 <i>; yields {i32}:result3 = FF</i>
8197%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
8198</pre>
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8199
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8200</div>
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8201
8202<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8203<h4>
8204 <a name="int_atomic_load_max">
8205 '<tt>llvm.atomic.load.max.*</tt>' Intrinsic
8206 </a>
8207 <br>
8208 <a name="int_atomic_load_min">
8209 '<tt>llvm.atomic.load.min.*</tt>' Intrinsic
8210 </a>
8211 <br>
8212 <a name="int_atomic_load_umax">
8213 '<tt>llvm.atomic.load.umax.*</tt>' Intrinsic
8214 </a>
8215 <br>
8216 <a name="int_atomic_load_umin">
8217 '<tt>llvm.atomic.load.umin.*</tt>' Intrinsic
8218 </a>
8219</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8220
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8221<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8222
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8223<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8224<p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
8225 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
8226 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
8227 address spaces. Not all targets support all bit widths however.</p>
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8228
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8229<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8230 declare i8 @llvm.atomic.load.max.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
8231 declare i16 @llvm.atomic.load.max.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
8232 declare i32 @llvm.atomic.load.max.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
8233 declare i64 @llvm.atomic.load.max.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8234</pre>
8235
8236<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8237 declare i8 @llvm.atomic.load.min.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
8238 declare i16 @llvm.atomic.load.min.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
8239 declare i32 @llvm.atomic.load.min.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
8240 declare i64 @llvm.atomic.load.min.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8241</pre>
8242
8243<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8244 declare i8 @llvm.atomic.load.umax.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
8245 declare i16 @llvm.atomic.load.umax.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
8246 declare i32 @llvm.atomic.load.umax.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
8247 declare i64 @llvm.atomic.load.umax.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8248</pre>
8249
8250<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8251 declare i8 @llvm.atomic.load.umin.i8.p0i8(i8* &lt;ptr&gt;, i8 &lt;delta&gt;)
8252 declare i16 @llvm.atomic.load.umin.i16.p0i16(i16* &lt;ptr&gt;, i16 &lt;delta&gt;)
8253 declare i32 @llvm.atomic.load.umin.i32.p0i32(i32* &lt;ptr&gt;, i32 &lt;delta&gt;)
8254 declare i64 @llvm.atomic.load.umin.i64.p0i64(i64* &lt;ptr&gt;, i64 &lt;delta&gt;)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8255</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8256
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8257<h5>Overview:</h5>
Eric Christopher455c5772009-12-05 02:46:03 +0000[diff] [blame]8258<p>These intrinsics takes the signed or unsigned minimum or maximum of
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8259 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
8260 original value at <tt>ptr</tt>.</p>
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8261
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8262<h5>Arguments:</h5>
8263<p>These intrinsics take two arguments, the first a pointer to an integer value
8264 and the second an integer value. The result is also an integer value. These
8265 integer types can have any bit width, but they must all have the same bit
8266 width. The targets may only lower integer representations they support.</p>
8267
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8268<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8269<p>These intrinsics does a series of operations atomically. They first load the
8270 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
8271 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
8272 yield the original value stored at <tt>ptr</tt>.</p>
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8273
8274<h5>Examples:</h5>
8275<pre>
Victor Hernandeza70c6df2009-10-26 23:44:29 +0000[diff] [blame]8276%mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
8277%ptr = bitcast i8* %mallocP to i32*
8278 store i32 7, %ptr
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8279%result0 = call i32 @llvm.atomic.load.min.i32.p0i32(i32* %ptr, i32 -2)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8280 <i>; yields {i32}:result0 = 7</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8281%result1 = call i32 @llvm.atomic.load.max.i32.p0i32(i32* %ptr, i32 8)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8282 <i>; yields {i32}:result1 = -2</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8283%result2 = call i32 @llvm.atomic.load.umin.i32.p0i32(i32* %ptr, i32 10)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8284 <i>; yields {i32}:result2 = 8</i>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8285%result3 = call i32 @llvm.atomic.load.umax.i32.p0i32(i32* %ptr, i32 30)
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8286 <i>; yields {i32}:result3 = 8</i>
8287%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
8288</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8289
Mon P Wang6a490372008-06-25 08:15:39 +0000[diff] [blame]8290</div>
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]8291
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8292</div>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8293
8294<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8295<h3>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8296 <a name="int_memorymarkers">Memory Use Markers</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8297</h3>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8298
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8299<div>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8300
8301<p>This class of intrinsics exists to information about the lifetime of memory
8302 objects and ranges where variables are immutable.</p>
8303
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8304<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8305<h4>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8306 <a name="int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8307</h4>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8308
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8309<div>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8310
8311<h5>Syntax:</h5>
8312<pre>
8313 declare void @llvm.lifetime.start(i64 &lt;size&gt;, i8* nocapture &lt;ptr&gt;)
8314</pre>
8315
8316<h5>Overview:</h5>
8317<p>The '<tt>llvm.lifetime.start</tt>' intrinsic specifies the start of a memory
8318 object's lifetime.</p>
8319
8320<h5>Arguments:</h5>
Nick Lewycky9bc89042009-10-13 07:57:33 +0000[diff] [blame]8321<p>The first argument is a constant integer representing the size of the
8322 object, or -1 if it is variable sized. The second argument is a pointer to
8323 the object.</p>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8324
8325<h5>Semantics:</h5>
8326<p>This intrinsic indicates that before this point in the code, the value of the
8327 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
Nick Lewyckyd20fd592009-10-27 16:56:58 +0000[diff] [blame]8328 never be used and has an undefined value. A load from the pointer that
8329 precedes this intrinsic can be replaced with
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8330 <tt>'<a href="#undefvalues">undef</a>'</tt>.</p>
8331
8332</div>
8333
8334<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8335<h4>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8336 <a name="int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8337</h4>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8338
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8339<div>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8340
8341<h5>Syntax:</h5>
8342<pre>
8343 declare void @llvm.lifetime.end(i64 &lt;size&gt;, i8* nocapture &lt;ptr&gt;)
8344</pre>
8345
8346<h5>Overview:</h5>
8347<p>The '<tt>llvm.lifetime.end</tt>' intrinsic specifies the end of a memory
8348 object's lifetime.</p>
8349
8350<h5>Arguments:</h5>
Nick Lewycky9bc89042009-10-13 07:57:33 +0000[diff] [blame]8351<p>The first argument is a constant integer representing the size of the
8352 object, or -1 if it is variable sized. The second argument is a pointer to
8353 the object.</p>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8354
8355<h5>Semantics:</h5>
8356<p>This intrinsic indicates that after this point in the code, the value of the
8357 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
8358 never be used and has an undefined value. Any stores into the memory object
8359 following this intrinsic may be removed as dead.
8360
8361</div>
8362
8363<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8364<h4>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8365 <a name="int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8366</h4>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8367
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8368<div>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8369
8370<h5>Syntax:</h5>
8371<pre>
Nick Lewycky2965d3e2010-11-30 04:13:41 +0000[diff] [blame]8372 declare {}* @llvm.invariant.start(i64 &lt;size&gt;, i8* nocapture &lt;ptr&gt;)
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8373</pre>
8374
8375<h5>Overview:</h5>
8376<p>The '<tt>llvm.invariant.start</tt>' intrinsic specifies that the contents of
8377 a memory object will not change.</p>
8378
8379<h5>Arguments:</h5>
Nick Lewycky9bc89042009-10-13 07:57:33 +0000[diff] [blame]8380<p>The first argument is a constant integer representing the size of the
8381 object, or -1 if it is variable sized. The second argument is a pointer to
8382 the object.</p>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8383
8384<h5>Semantics:</h5>
8385<p>This intrinsic indicates that until an <tt>llvm.invariant.end</tt> that uses
8386 the return value, the referenced memory location is constant and
8387 unchanging.</p>
8388
8389</div>
8390
8391<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8392<h4>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8393 <a name="int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8394</h4>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8395
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8396<div>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8397
8398<h5>Syntax:</h5>
8399<pre>
8400 declare void @llvm.invariant.end({}* &lt;start&gt;, i64 &lt;size&gt;, i8* nocapture &lt;ptr&gt;)
8401</pre>
8402
8403<h5>Overview:</h5>
8404<p>The '<tt>llvm.invariant.end</tt>' intrinsic specifies that the contents of
8405 a memory object are mutable.</p>
8406
8407<h5>Arguments:</h5>
8408<p>The first argument is the matching <tt>llvm.invariant.start</tt> intrinsic.
Nick Lewycky9bc89042009-10-13 07:57:33 +0000[diff] [blame]8409 The second argument is a constant integer representing the size of the
8410 object, or -1 if it is variable sized and the third argument is a pointer
8411 to the object.</p>
Nick Lewycky6f7d8342009-10-13 07:03:23 +0000[diff] [blame]8412
8413<h5>Semantics:</h5>
8414<p>This intrinsic indicates that the memory is mutable again.</p>
8415
8416</div>
8417
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8418</div>
8419
Andrew Lenharth9b254ee2008-02-16 01:24:58 +0000[diff] [blame]8420<!-- ======================================================================= -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8421<h3>
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000[diff] [blame]8422 <a name="int_general">General Intrinsics</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8423</h3>
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000[diff] [blame]8424
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8425<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8426
8427<p>This class of intrinsics is designed to be generic and has no specific
8428 purpose.</p>
8429
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000[diff] [blame]8430<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8431<h4>
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000[diff] [blame]8432 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8433</h4>
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000[diff] [blame]8434
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8435<div>
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000[diff] [blame]8436
8437<h5>Syntax:</h5>
8438<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8439 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 +0000[diff] [blame]8440</pre>
8441
8442<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8443<p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000[diff] [blame]8444
8445<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8446<p>The first argument is a pointer to a value, the second is a pointer to a
8447 global string, the third is a pointer to a global string which is the source
8448 file name, and the last argument is the line number.</p>
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000[diff] [blame]8449
8450<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8451<p>This intrinsic allows annotation of local variables with arbitrary strings.
8452 This can be useful for special purpose optimizations that want to look for
John Criswellf0d536a2011-08-19 16:57:55 +0000[diff] [blame]8453 these annotations. These have no other defined use; they are ignored by code
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8454 generation and optimization.</p>
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000[diff] [blame]8455
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000[diff] [blame]8456</div>
8457
Tanya Lattner293c0372007-09-21 22:59:12 +0000[diff] [blame]8458<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8459<h4>
Tanya Lattner0186a652007-09-21 23:57:59 +0000[diff] [blame]8460 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8461</h4>
Tanya Lattner293c0372007-09-21 22:59:12 +0000[diff] [blame]8462
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8463<div>
Tanya Lattner293c0372007-09-21 22:59:12 +0000[diff] [blame]8464
8465<h5>Syntax:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8466<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
8467 any integer bit width.</p>
8468
Tanya Lattner293c0372007-09-21 22:59:12 +0000[diff] [blame]8469<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8470 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
8471 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
8472 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
8473 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt;)
8474 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 +0000[diff] [blame]8475</pre>
8476
8477<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8478<p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
Tanya Lattner293c0372007-09-21 22:59:12 +0000[diff] [blame]8479
8480<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8481<p>The first argument is an integer value (result of some expression), the
8482 second is a pointer to a global string, the third is a pointer to a global
8483 string which is the source file name, and the last argument is the line
8484 number. It returns the value of the first argument.</p>
Tanya Lattner293c0372007-09-21 22:59:12 +0000[diff] [blame]8485
8486<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8487<p>This intrinsic allows annotations to be put on arbitrary expressions with
8488 arbitrary strings. This can be useful for special purpose optimizations that
John Criswellf0d536a2011-08-19 16:57:55 +0000[diff] [blame]8489 want to look for these annotations. These have no other defined use; they
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8490 are ignored by code generation and optimization.</p>
Tanya Lattner293c0372007-09-21 22:59:12 +0000[diff] [blame]8491
Tanya Lattner293c0372007-09-21 22:59:12 +0000[diff] [blame]8492</div>
Jim Laskey2211f492007-03-14 19:31:19 +0000[diff] [blame]8493
Anton Korobeynikov06cbb652008-01-15 22:31:34 +0000[diff] [blame]8494<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8495<h4>
Anton Korobeynikov06cbb652008-01-15 22:31:34 +0000[diff] [blame]8496 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8497</h4>
Anton Korobeynikov06cbb652008-01-15 22:31:34 +0000[diff] [blame]8498
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8499<div>
Anton Korobeynikov06cbb652008-01-15 22:31:34 +0000[diff] [blame]8500
8501<h5>Syntax:</h5>
8502<pre>
8503 declare void @llvm.trap()
8504</pre>
8505
8506<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8507<p>The '<tt>llvm.trap</tt>' intrinsic.</p>
Anton Korobeynikov06cbb652008-01-15 22:31:34 +0000[diff] [blame]8508
8509<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8510<p>None.</p>
Anton Korobeynikov06cbb652008-01-15 22:31:34 +0000[diff] [blame]8511
8512<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8513<p>This intrinsics is lowered to the target dependent trap instruction. If the
8514 target does not have a trap instruction, this intrinsic will be lowered to
8515 the call of the <tt>abort()</tt> function.</p>
Anton Korobeynikov06cbb652008-01-15 22:31:34 +0000[diff] [blame]8516
Anton Korobeynikov06cbb652008-01-15 22:31:34 +0000[diff] [blame]8517</div>
8518
Bill Wendling14313312008-11-19 05:56:17 +0000[diff] [blame]8519<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8520<h4>
Misha Brukman50de2b22008-11-22 23:55:29 +0000[diff] [blame]8521 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8522</h4>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8523
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8524<div>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8525
Bill Wendling14313312008-11-19 05:56:17 +0000[diff] [blame]8526<h5>Syntax:</h5>
8527<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8528 declare void @llvm.stackprotector(i8* &lt;guard&gt;, i8** &lt;slot&gt;)
Bill Wendling14313312008-11-19 05:56:17 +0000[diff] [blame]8529</pre>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8530
Bill Wendling14313312008-11-19 05:56:17 +0000[diff] [blame]8531<h5>Overview:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8532<p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
8533 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
8534 ensure that it is placed on the stack before local variables.</p>
8535
Bill Wendling14313312008-11-19 05:56:17 +0000[diff] [blame]8536<h5>Arguments:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8537<p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
8538 arguments. The first argument is the value loaded from the stack
8539 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
8540 that has enough space to hold the value of the guard.</p>
8541
Bill Wendling14313312008-11-19 05:56:17 +0000[diff] [blame]8542<h5>Semantics:</h5>
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8543<p>This intrinsic causes the prologue/epilogue inserter to force the position of
8544 the <tt>AllocaInst</tt> stack slot to be before local variables on the
8545 stack. This is to ensure that if a local variable on the stack is
8546 overwritten, it will destroy the value of the guard. When the function exits,
Bill Wendling6bbe0912010-10-27 01:07:41 +0000[diff] [blame]8547 the guard on the stack is checked against the original guard. If they are
Bill Wendlingd9a66f72009-07-20 02:29:24 +0000[diff] [blame]8548 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
8549 function.</p>
8550
Bill Wendling14313312008-11-19 05:56:17 +0000[diff] [blame]8551</div>
8552
Eric Christopher73484322009-11-30 08:03:53 +0000[diff] [blame]8553<!-- _______________________________________________________________________ -->
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8554<h4>
Eric Christopher73484322009-11-30 08:03:53 +0000[diff] [blame]8555 <a name="int_objectsize">'<tt>llvm.objectsize</tt>' Intrinsic</a>
NAKAMURA Takumifc8d9302011-04-18 23:59:50 +0000[diff] [blame]8556</h4>
Eric Christopher73484322009-11-30 08:03:53 +0000[diff] [blame]8557
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8558<div>
Eric Christopher73484322009-11-30 08:03:53 +0000[diff] [blame]8559
8560<h5>Syntax:</h5>
8561<pre>
Dan Gohmand6a6f612010-05-28 17:07:41 +0000[diff] [blame]8562 declare i32 @llvm.objectsize.i32(i8* &lt;object&gt;, i1 &lt;type&gt;)
8563 declare i64 @llvm.objectsize.i64(i8* &lt;object&gt;, i1 &lt;type&gt;)
Eric Christopher73484322009-11-30 08:03:53 +0000[diff] [blame]8564</pre>
8565
8566<h5>Overview:</h5>
Bill Wendling6bbe0912010-10-27 01:07:41 +0000[diff] [blame]8567<p>The <tt>llvm.objectsize</tt> intrinsic is designed to provide information to
8568 the optimizers to determine at compile time whether a) an operation (like
8569 memcpy) will overflow a buffer that corresponds to an object, or b) that a
8570 runtime check for overflow isn't necessary. An object in this context means
8571 an allocation of a specific class, structure, array, or other object.</p>
Eric Christopher73484322009-11-30 08:03:53 +0000[diff] [blame]8572
8573<h5>Arguments:</h5>
Bill Wendling6bbe0912010-10-27 01:07:41 +0000[diff] [blame]8574<p>The <tt>llvm.objectsize</tt> intrinsic takes two arguments. The first
Eric Christopher31e39bd2009-12-23 00:29:49 +0000[diff] [blame]8575 argument is a pointer to or into the <tt>object</tt>. The second argument
Bill Wendling6bbe0912010-10-27 01:07:41 +0000[diff] [blame]8576 is a boolean 0 or 1. This argument determines whether you want the
8577 maximum (0) or minimum (1) bytes remaining. This needs to be a literal 0 or
Eric Christopher31e39bd2009-12-23 00:29:49 +0000[diff] [blame]8578 1, variables are not allowed.</p>
8579
Eric Christopher73484322009-11-30 08:03:53 +0000[diff] [blame]8580<h5>Semantics:</h5>
8581<p>The <tt>llvm.objectsize</tt> intrinsic is lowered to either a constant
Bill Wendling6bbe0912010-10-27 01:07:41 +0000[diff] [blame]8582 representing the size of the object concerned, or <tt>i32/i64 -1 or 0</tt>,
8583 depending on the <tt>type</tt> argument, if the size cannot be determined at
8584 compile time.</p>
Eric Christopher73484322009-11-30 08:03:53 +0000[diff] [blame]8585
8586</div>
8587
NAKAMURA Takumiaa3d6242011-04-23 00:30:22 +0000[diff] [blame]8588</div>
8589
8590</div>
8591
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]8592<!-- *********************************************************************** -->
Chris Lattner2f7c9632001-06-06 20:29:01 +0000[diff] [blame]8593<hr>
Misha Brukmanc501f552004-03-01 17:47:27 +0000[diff] [blame]8594<address>
8595 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman86242e12008-12-11 17:34:48 +0000[diff] [blame]8596 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Misha Brukmanc501f552004-03-01 17:47:27 +0000[diff] [blame]8597 <a href="http://validator.w3.org/check/referer"><img
Misha Brukman86242e12008-12-11 17:34:48 +0000[diff] [blame]8598 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
Misha Brukmanc501f552004-03-01 17:47:27 +0000[diff] [blame]8599
8600 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
NAKAMURA Takumica46f5a2011-04-09 02:13:37 +0000[diff] [blame]8601 <a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br>
Misha Brukmanc501f552004-03-01 17:47:27 +0000[diff] [blame]8602 Last modified: $Date$
8603</address>
Chris Lattnerb8f816e2008-01-04 04:33:49 +0000[diff] [blame]8604
Misha Brukman76307852003-11-08 01:05:38 +0000[diff] [blame]8605</body>
8606</html>