blob: 9fc14b4fea560010a12b552ff9286a06c83e5a05 [file] [log] [blame]
Misha Brukmanc501f552004-03-01 17:47:27 +00001<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
Misha Brukman76307852003-11-08 01:05:38 +00003<html>
4<head>
5 <title>LLVM Assembly Language Reference Manual</title>
Reid Spencercb84e432004-08-26 20:44:00 +00006 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
8 <meta name="description"
9 content="LLVM Assembly Language Reference Manual.">
Misha Brukman76307852003-11-08 01:05:38 +000010 <link rel="stylesheet" href="llvm.css" type="text/css">
11</head>
Chris Lattner757528b0b2004-05-23 21:06:01 +000012
Misha Brukman76307852003-11-08 01:05:38 +000013<body>
Chris Lattner757528b0b2004-05-23 21:06:01 +000014
Chris Lattner48b383b02003-11-25 01:02:51 +000015<div class="doc_title"> LLVM Language Reference Manual </div>
Chris Lattner2f7c9632001-06-06 20:29:01 +000016<ol>
Misha Brukman76307852003-11-08 01:05:38 +000017 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
Chris Lattner6af02f32004-12-09 16:11:40 +000020 <li><a href="#highlevel">High Level Structure</a>
21 <ol>
22 <li><a href="#modulestructure">Module Structure</a></li>
Chris Lattnerd79749a2004-12-09 16:36:40 +000023 <li><a href="#linkage">Linkage Types</a></li>
Chris Lattner0132aff2005-05-06 22:57:40 +000024 <li><a href="#callingconv">Calling Conventions</a></li>
Chris Lattnerbc088212009-01-11 20:53:49 +000025 <li><a href="#namedtypes">Named Types</a></li>
Chris Lattner6af02f32004-12-09 16:11:40 +000026 <li><a href="#globalvars">Global Variables</a></li>
Chris Lattner91c15c42006-01-23 23:23:47 +000027 <li><a href="#functionstructure">Functions</a></li>
Dan Gohmanef9462f2008-10-14 16:51:45 +000028 <li><a href="#aliasstructure">Aliases</a></li>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +000029 <li><a href="#paramattrs">Parameter Attributes</a></li>
Devang Patel9eb525d2008-09-26 23:51:19 +000030 <li><a href="#fnattrs">Function Attributes</a></li>
Gordon Henriksen71183b62007-12-10 03:18:06 +000031 <li><a href="#gc">Garbage Collector Names</a></li>
Chris Lattner91c15c42006-01-23 23:23:47 +000032 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
Reid Spencer50c723a2007-02-19 23:54:10 +000033 <li><a href="#datalayout">Data Layout</a></li>
Chris Lattner6af02f32004-12-09 16:11:40 +000034 </ol>
35 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +000036 <li><a href="#typesystem">Type System</a>
37 <ol>
Chris Lattner7824d182008-01-04 04:32:38 +000038 <li><a href="#t_classifications">Type Classifications</a></li>
Robert Bocchino820bc75b2006-02-17 21:18:08 +000039 <li><a href="#t_primitive">Primitive Types</a>
Chris Lattner48b383b02003-11-25 01:02:51 +000040 <ol>
Chris Lattner7824d182008-01-04 04:32:38 +000041 <li><a href="#t_floating">Floating Point Types</a></li>
42 <li><a href="#t_void">Void Type</a></li>
43 <li><a href="#t_label">Label Type</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000044 </ol>
45 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +000046 <li><a href="#t_derived">Derived Types</a>
47 <ol>
Chris Lattner9a2e3cb2007-12-18 06:18:21 +000048 <li><a href="#t_integer">Integer Type</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000049 <li><a href="#t_array">Array Type</a></li>
Misha Brukman76307852003-11-08 01:05:38 +000050 <li><a href="#t_function">Function Type</a></li>
51 <li><a href="#t_pointer">Pointer Type</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000052 <li><a href="#t_struct">Structure Type</a></li>
Andrew Lenharth8df88e22006-12-08 17:13:00 +000053 <li><a href="#t_pstruct">Packed Structure Type</a></li>
Reid Spencer404a3252007-02-15 03:07:05 +000054 <li><a href="#t_vector">Vector Type</a></li>
Chris Lattner37b6b092005-04-25 17:34:15 +000055 <li><a href="#t_opaque">Opaque Type</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000056 </ol>
57 </li>
Chris Lattnercf7a5842009-02-02 07:32:36 +000058 <li><a href="#t_uprefs">Type Up-references</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000059 </ol>
60 </li>
Chris Lattner6af02f32004-12-09 16:11:40 +000061 <li><a href="#constants">Constants</a>
Chris Lattner74d3f822004-12-09 17:30:23 +000062 <ol>
Dan Gohmanef9462f2008-10-14 16:51:45 +000063 <li><a href="#simpleconstants">Simple Constants</a></li>
Chris Lattner361bfcd2009-02-28 18:32:25 +000064 <li><a href="#complexconstants">Complex Constants</a></li>
Dan Gohmanef9462f2008-10-14 16:51:45 +000065 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
66 <li><a href="#undefvalues">Undefined Values</a></li>
67 <li><a href="#constantexprs">Constant Expressions</a></li>
Nick Lewycky49f89192009-04-04 07:22:01 +000068 <li><a href="#metadata">Embedded Metadata</a></li>
Chris Lattner74d3f822004-12-09 17:30:23 +000069 </ol>
Chris Lattner48b383b02003-11-25 01:02:51 +000070 </li>
Chris Lattner98f013c2006-01-25 23:47:57 +000071 <li><a href="#othervalues">Other Values</a>
72 <ol>
Dan Gohmanef9462f2008-10-14 16:51:45 +000073 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
Chris Lattner98f013c2006-01-25 23:47:57 +000074 </ol>
75 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +000076 <li><a href="#instref">Instruction Reference</a>
77 <ol>
78 <li><a href="#terminators">Terminator Instructions</a>
79 <ol>
Chris Lattner48b383b02003-11-25 01:02:51 +000080 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
81 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
Misha Brukman76307852003-11-08 01:05:38 +000082 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
83 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000084 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
Chris Lattner08b7d5b2004-10-16 18:04:13 +000085 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000086 </ol>
87 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +000088 <li><a href="#binaryops">Binary Operations</a>
89 <ol>
Chris Lattner48b383b02003-11-25 01:02:51 +000090 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
91 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
92 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
Reid Spencer7e80b0b2006-10-26 06:15:43 +000093 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
94 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
95 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
Reid Spencer7eb55b32006-11-02 01:53:59 +000096 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
97 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
98 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +000099 </ol>
100 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000101 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
102 <ol>
Reid Spencer2ab01932007-02-02 13:57:07 +0000103 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
104 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
105 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
Misha Brukman76307852003-11-08 01:05:38 +0000106 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000107 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
Misha Brukman76307852003-11-08 01:05:38 +0000108 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000109 </ol>
110 </li>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000111 <li><a href="#vectorops">Vector Operations</a>
112 <ol>
113 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
114 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
115 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
Chris Lattnerce83bff2006-04-08 23:07:04 +0000116 </ol>
117 </li>
Dan Gohmanb9d66602008-05-12 23:51:09 +0000118 <li><a href="#aggregateops">Aggregate Operations</a>
119 <ol>
120 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
121 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
122 </ol>
123 </li>
Chris Lattner6ab66722006-08-15 00:45:58 +0000124 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000125 <ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000126 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
127 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
128 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
Robert Bocchino820bc75b2006-02-17 21:18:08 +0000129 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
130 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
131 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000132 </ol>
133 </li>
Reid Spencer97c5fa42006-11-08 01:18:52 +0000134 <li><a href="#convertops">Conversion Operations</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000135 <ol>
136 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
137 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
138 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
139 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
140 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
Reid Spencer51b07252006-11-09 23:03:26 +0000141 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
142 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
143 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
144 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
Reid Spencerb7344ff2006-11-11 21:00:47 +0000145 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
146 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
Reid Spencer5b950642006-11-11 23:08:07 +0000147 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
Reid Spencer59b6b7d2006-11-08 01:11:31 +0000148 </ol>
Dan Gohmanef9462f2008-10-14 16:51:45 +0000149 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000150 <li><a href="#otherops">Other Operations</a>
151 <ol>
Reid Spencerc828a0e2006-11-18 21:50:54 +0000152 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
153 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
Nate Begemand2195702008-05-12 19:01:56 +0000154 <li><a href="#i_vicmp">'<tt>vicmp</tt>' Instruction</a></li>
155 <li><a href="#i_vfcmp">'<tt>vfcmp</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000156 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
Chris Lattnerb53c28d2004-03-12 05:50:16 +0000157 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000158 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
Chris Lattner33337472006-01-13 23:26:01 +0000159 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000160 </ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000161 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000162 </ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000163 </li>
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000164 <li><a href="#intrinsics">Intrinsic Functions</a>
Chris Lattnerbd64b4e2003-05-08 04:57:36 +0000165 <ol>
Chris Lattner48b383b02003-11-25 01:02:51 +0000166 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
167 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000168 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
169 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
170 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000171 </ol>
172 </li>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000173 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
174 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000175 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
176 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
177 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000178 </ol>
179 </li>
Chris Lattner3649c3a2004-02-14 04:08:35 +0000180 <li><a href="#int_codegen">Code Generator Intrinsics</a>
181 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000182 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
183 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
184 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
185 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
186 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
187 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
188 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
John Criswellaa1c3c12004-04-09 16:43:20 +0000189 </ol>
190 </li>
Chris Lattnerfee11462004-02-12 17:01:32 +0000191 <li><a href="#int_libc">Standard C Library Intrinsics</a>
192 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000193 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
194 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
195 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
196 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
197 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
Dan Gohmanb6324c12007-10-15 20:30:11 +0000198 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
199 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
200 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
Chris Lattnerfee11462004-02-12 17:01:32 +0000201 </ol>
202 </li>
Nate Begeman0f223bb2006-01-13 23:26:38 +0000203 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000204 <ol>
Reid Spencer96a5f022007-04-04 02:42:35 +0000205 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
Chris Lattnerb748c672006-01-16 22:34:14 +0000206 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
207 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
208 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
Reid Spencer5bf54c82007-04-11 23:23:49 +0000209 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
210 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
Andrew Lenharth1d463522005-05-03 18:01:48 +0000211 </ol>
212 </li>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000213 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
214 <ol>
Bill Wendlingfd2bd722009-02-08 04:04:40 +0000215 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
216 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
217 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
218 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
219 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
Bill Wendlingb9a73272009-02-08 23:00:09 +0000220 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
Bill Wendlingf4d70622009-02-08 01:40:31 +0000221 </ol>
222 </li>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000223 <li><a href="#int_debugger">Debugger intrinsics</a></li>
Jim Laskey2211f492007-03-14 19:31:19 +0000224 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
Duncan Sands86e01192007-09-11 14:10:23 +0000225 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
Duncan Sands644f9172007-07-27 12:58:54 +0000226 <ol>
227 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
Duncan Sands644f9172007-07-27 12:58:54 +0000228 </ol>
229 </li>
Bill Wendlingf85850f2008-11-18 22:10:53 +0000230 <li><a href="#int_atomics">Atomic intrinsics</a>
231 <ol>
232 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
233 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
234 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
235 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
236 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
237 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
238 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
239 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
240 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
241 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
242 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
243 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
244 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
245 </ol>
246 </li>
Reid Spencer5b2cb0f2007-07-20 19:59:11 +0000247 <li><a href="#int_general">General intrinsics</a>
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000248 <ol>
Reid Spencer5b2cb0f2007-07-20 19:59:11 +0000249 <li><a href="#int_var_annotation">
Bill Wendling14313312008-11-19 05:56:17 +0000250 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
Tanya Lattner293c0372007-09-21 22:59:12 +0000251 <li><a href="#int_annotation">
Bill Wendling14313312008-11-19 05:56:17 +0000252 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
Anton Korobeynikov06cbb652008-01-15 22:31:34 +0000253 <li><a href="#int_trap">
Bill Wendling14313312008-11-19 05:56:17 +0000254 '<tt>llvm.trap</tt>' Intrinsic</a></li>
255 <li><a href="#int_stackprotector">
256 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
Tanya Lattner293c0372007-09-21 22:59:12 +0000257 </ol>
Tanya Lattnercb1b9602007-06-15 20:50:54 +0000258 </li>
Chris Lattner48b383b02003-11-25 01:02:51 +0000259 </ol>
260 </li>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000261</ol>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000262
263<div class="doc_author">
264 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
265 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
Misha Brukman76307852003-11-08 01:05:38 +0000266</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000267
Chris Lattner2f7c9632001-06-06 20:29:01 +0000268<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +0000269<div class="doc_section"> <a name="abstract">Abstract </a></div>
270<!-- *********************************************************************** -->
Chris Lattner757528b0b2004-05-23 21:06:01 +0000271
Misha Brukman76307852003-11-08 01:05:38 +0000272<div class="doc_text">
Chris Lattner48b383b02003-11-25 01:02:51 +0000273<p>This document is a reference manual for the LLVM assembly language.
Bill Wendling6e03f9a2008-08-05 22:29:16 +0000274LLVM is a Static Single Assignment (SSA) based representation that provides
Chris Lattner67c37d12008-08-05 18:29:16 +0000275type safety, low-level operations, flexibility, and the capability of
276representing 'all' high-level languages cleanly. It is the common code
Chris Lattner48b383b02003-11-25 01:02:51 +0000277representation used throughout all phases of the LLVM compilation
278strategy.</p>
Misha Brukman76307852003-11-08 01:05:38 +0000279</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000280
Chris Lattner2f7c9632001-06-06 20:29:01 +0000281<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +0000282<div class="doc_section"> <a name="introduction">Introduction</a> </div>
283<!-- *********************************************************************** -->
Chris Lattner757528b0b2004-05-23 21:06:01 +0000284
Misha Brukman76307852003-11-08 01:05:38 +0000285<div class="doc_text">
Chris Lattner757528b0b2004-05-23 21:06:01 +0000286
Chris Lattner48b383b02003-11-25 01:02:51 +0000287<p>The LLVM code representation is designed to be used in three
Gabor Greifa54634a2007-07-06 22:07:22 +0000288different forms: as an in-memory compiler IR, as an on-disk bitcode
Chris Lattner48b383b02003-11-25 01:02:51 +0000289representation (suitable for fast loading by a Just-In-Time compiler),
290and as a human readable assembly language representation. This allows
291LLVM to provide a powerful intermediate representation for efficient
292compiler transformations and analysis, while providing a natural means
293to debug and visualize the transformations. The three different forms
294of LLVM are all equivalent. This document describes the human readable
295representation and notation.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000296
John Criswell4a3327e2005-05-13 22:25:59 +0000297<p>The LLVM representation aims to be light-weight and low-level
Chris Lattner48b383b02003-11-25 01:02:51 +0000298while being expressive, typed, and extensible at the same time. It
299aims to be a "universal IR" of sorts, by being at a low enough level
300that high-level ideas may be cleanly mapped to it (similar to how
301microprocessors are "universal IR's", allowing many source languages to
302be mapped to them). By providing type information, LLVM can be used as
303the target of optimizations: for example, through pointer analysis, it
304can be proven that a C automatic variable is never accessed outside of
305the current function... allowing it to be promoted to a simple SSA
306value instead of a memory location.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000307
Misha Brukman76307852003-11-08 01:05:38 +0000308</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000309
Chris Lattner2f7c9632001-06-06 20:29:01 +0000310<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +0000311<div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000312
Misha Brukman76307852003-11-08 01:05:38 +0000313<div class="doc_text">
Chris Lattner757528b0b2004-05-23 21:06:01 +0000314
Chris Lattner48b383b02003-11-25 01:02:51 +0000315<p>It is important to note that this document describes 'well formed'
316LLVM assembly language. There is a difference between what the parser
317accepts and what is considered 'well formed'. For example, the
318following instruction is syntactically okay, but not well formed:</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000319
Bill Wendling3716c5d2007-05-29 09:04:49 +0000320<div class="doc_code">
Chris Lattner757528b0b2004-05-23 21:06:01 +0000321<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000322%x = <a href="#i_add">add</a> i32 1, %x
Chris Lattner757528b0b2004-05-23 21:06:01 +0000323</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000324</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000325
Chris Lattner48b383b02003-11-25 01:02:51 +0000326<p>...because the definition of <tt>%x</tt> does not dominate all of
327its uses. The LLVM infrastructure provides a verification pass that may
328be used to verify that an LLVM module is well formed. This pass is
John Criswell4a3327e2005-05-13 22:25:59 +0000329automatically run by the parser after parsing input assembly and by
Gabor Greifa54634a2007-07-06 22:07:22 +0000330the optimizer before it outputs bitcode. The violations pointed out
Chris Lattner48b383b02003-11-25 01:02:51 +0000331by the verifier pass indicate bugs in transformation passes or input to
332the parser.</p>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000333</div>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000334
Chris Lattner87a3dbe2007-10-03 17:34:29 +0000335<!-- Describe the typesetting conventions here. -->
Chris Lattner757528b0b2004-05-23 21:06:01 +0000336
Chris Lattner2f7c9632001-06-06 20:29:01 +0000337<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +0000338<div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
Chris Lattner2f7c9632001-06-06 20:29:01 +0000339<!-- *********************************************************************** -->
Chris Lattner757528b0b2004-05-23 21:06:01 +0000340
Misha Brukman76307852003-11-08 01:05:38 +0000341<div class="doc_text">
Chris Lattner757528b0b2004-05-23 21:06:01 +0000342
Reid Spencerb23b65f2007-08-07 14:34:28 +0000343 <p>LLVM identifiers come in two basic types: global and local. Global
344 identifiers (functions, global variables) begin with the @ character. Local
345 identifiers (register names, types) begin with the % character. Additionally,
Dan Gohmanef9462f2008-10-14 16:51:45 +0000346 there are three different formats for identifiers, for different purposes:</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +0000347
Chris Lattner2f7c9632001-06-06 20:29:01 +0000348<ol>
Reid Spencerb23b65f2007-08-07 14:34:28 +0000349 <li>Named values are represented as a string of characters with their prefix.
350 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
351 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
Chris Lattnerd79749a2004-12-09 16:36:40 +0000352 Identifiers which require other characters in their names can be surrounded
Daniel Dunbar0f8155a2008-10-14 23:51:43 +0000353 with quotes. Special characters may be escaped using "\xx" where xx is the
354 ASCII code for the character in hexadecimal. In this way, any character can
355 be used in a name value, even quotes themselves.
Chris Lattnerd79749a2004-12-09 16:36:40 +0000356
Reid Spencerb23b65f2007-08-07 14:34:28 +0000357 <li>Unnamed values are represented as an unsigned numeric value with their
358 prefix. For example, %12, @2, %44.</li>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000359
Reid Spencer8f08d802004-12-09 18:02:53 +0000360 <li>Constants, which are described in a <a href="#constants">section about
361 constants</a>, below.</li>
Misha Brukman76307852003-11-08 01:05:38 +0000362</ol>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000363
Reid Spencerb23b65f2007-08-07 14:34:28 +0000364<p>LLVM requires that values start with a prefix for two reasons: Compilers
Chris Lattnerd79749a2004-12-09 16:36:40 +0000365don't need to worry about name clashes with reserved words, and the set of
366reserved words may be expanded in the future without penalty. Additionally,
367unnamed identifiers allow a compiler to quickly come up with a temporary
368variable without having to avoid symbol table conflicts.</p>
369
Chris Lattner48b383b02003-11-25 01:02:51 +0000370<p>Reserved words in LLVM are very similar to reserved words in other
Reid Spencer5b950642006-11-11 23:08:07 +0000371languages. There are keywords for different opcodes
372('<tt><a href="#i_add">add</a></tt>',
373 '<tt><a href="#i_bitcast">bitcast</a></tt>',
374 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000375href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
Chris Lattnerd79749a2004-12-09 16:36:40 +0000376and others. These reserved words cannot conflict with variable names, because
Reid Spencerb23b65f2007-08-07 14:34:28 +0000377none of them start with a prefix character ('%' or '@').</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000378
379<p>Here is an example of LLVM code to multiply the integer variable
380'<tt>%X</tt>' by 8:</p>
381
Misha Brukman76307852003-11-08 01:05:38 +0000382<p>The easy way:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000383
Bill Wendling3716c5d2007-05-29 09:04:49 +0000384<div class="doc_code">
Chris Lattnerd79749a2004-12-09 16:36:40 +0000385<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000386%result = <a href="#i_mul">mul</a> i32 %X, 8
Chris Lattnerd79749a2004-12-09 16:36:40 +0000387</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000388</div>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000389
Misha Brukman76307852003-11-08 01:05:38 +0000390<p>After strength reduction:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000391
Bill Wendling3716c5d2007-05-29 09:04:49 +0000392<div class="doc_code">
Chris Lattnerd79749a2004-12-09 16:36:40 +0000393<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000394%result = <a href="#i_shl">shl</a> i32 %X, i8 3
Chris Lattnerd79749a2004-12-09 16:36:40 +0000395</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000396</div>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000397
Misha Brukman76307852003-11-08 01:05:38 +0000398<p>And the hard way:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000399
Bill Wendling3716c5d2007-05-29 09:04:49 +0000400<div class="doc_code">
Chris Lattnerd79749a2004-12-09 16:36:40 +0000401<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000402<a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
403<a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
404%result = <a href="#i_add">add</a> i32 %1, %1
Chris Lattnerd79749a2004-12-09 16:36:40 +0000405</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000406</div>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000407
Chris Lattner48b383b02003-11-25 01:02:51 +0000408<p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
409important lexical features of LLVM:</p>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000410
Chris Lattner2f7c9632001-06-06 20:29:01 +0000411<ol>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000412
413 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
414 line.</li>
415
416 <li>Unnamed temporaries are created when the result of a computation is not
417 assigned to a named value.</li>
418
Misha Brukman76307852003-11-08 01:05:38 +0000419 <li>Unnamed temporaries are numbered sequentially</li>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000420
Misha Brukman76307852003-11-08 01:05:38 +0000421</ol>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000422
John Criswell02fdc6f2005-05-12 16:52:32 +0000423<p>...and it also shows a convention that we follow in this document. When
Chris Lattnerd79749a2004-12-09 16:36:40 +0000424demonstrating instructions, we will follow an instruction with a comment that
425defines the type and name of value produced. Comments are shown in italic
426text.</p>
427
Misha Brukman76307852003-11-08 01:05:38 +0000428</div>
Chris Lattner6af02f32004-12-09 16:11:40 +0000429
430<!-- *********************************************************************** -->
431<div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
432<!-- *********************************************************************** -->
433
434<!-- ======================================================================= -->
435<div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
436</div>
437
438<div class="doc_text">
439
440<p>LLVM programs are composed of "Module"s, each of which is a
441translation unit of the input programs. Each module consists of
442functions, global variables, and symbol table entries. Modules may be
443combined together with the LLVM linker, which merges function (and
444global variable) definitions, resolves forward declarations, and merges
445symbol table entries. Here is an example of the "hello world" module:</p>
446
Bill Wendling3716c5d2007-05-29 09:04:49 +0000447<div class="doc_code">
Chris Lattner6af02f32004-12-09 16:11:40 +0000448<pre><i>; Declare the string constant as a global constant...</i>
Chris Lattner2b0bf4f2007-06-12 17:00:26 +0000449<a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
450 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
Chris Lattner6af02f32004-12-09 16:11:40 +0000451
452<i>; External declaration of the puts function</i>
Chris Lattner2b0bf4f2007-06-12 17:00:26 +0000453<a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
Chris Lattner6af02f32004-12-09 16:11:40 +0000454
455<i>; Definition of main function</i>
Chris Lattner2b0bf4f2007-06-12 17:00:26 +0000456define i32 @main() { <i>; i32()* </i>
Dan Gohman623806e2009-01-04 23:44:43 +0000457 <i>; Convert [13 x i8]* to i8 *...</i>
Chris Lattner6af02f32004-12-09 16:11:40 +0000458 %cast210 = <a
Dan Gohman623806e2009-01-04 23:44:43 +0000459 href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
Chris Lattner6af02f32004-12-09 16:11:40 +0000460
461 <i>; Call puts function to write out the string to stdout...</i>
462 <a
Chris Lattner2b0bf4f2007-06-12 17:00:26 +0000463 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
Chris Lattner6af02f32004-12-09 16:11:40 +0000464 <a
Bill Wendling3716c5d2007-05-29 09:04:49 +0000465 href="#i_ret">ret</a> i32 0<br>}<br>
466</pre>
467</div>
Chris Lattner6af02f32004-12-09 16:11:40 +0000468
469<p>This example is made up of a <a href="#globalvars">global variable</a>
470named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
471function, and a <a href="#functionstructure">function definition</a>
472for "<tt>main</tt>".</p>
473
Chris Lattnerd79749a2004-12-09 16:36:40 +0000474<p>In general, a module is made up of a list of global values,
475where both functions and global variables are global values. Global values are
476represented by a pointer to a memory location (in this case, a pointer to an
477array of char, and a pointer to a function), and have one of the following <a
478href="#linkage">linkage types</a>.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000479
Chris Lattnerd79749a2004-12-09 16:36:40 +0000480</div>
481
482<!-- ======================================================================= -->
483<div class="doc_subsection">
484 <a name="linkage">Linkage Types</a>
485</div>
486
487<div class="doc_text">
488
489<p>
490All Global Variables and Functions have one of the following types of linkage:
491</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000492
493<dl>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000494
Rafael Espindola6de96a12009-01-15 20:18:42 +0000495 <dt><tt><b><a name="linkage_private">private</a></b></tt>: </dt>
496
497 <dd>Global values with private linkage are only directly accessible by
498 objects in the current module. In particular, linking code into a module with
499 an private global value may cause the private to be renamed as necessary to
500 avoid collisions. Because the symbol is private to the module, all
501 references can be updated. This doesn't show up in any symbol table in the
502 object file.
503 </dd>
504
Dale Johannesen4188aad2008-05-23 23:13:41 +0000505 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000506
Duncan Sands35e43c12009-01-16 09:29:46 +0000507 <dd> Similar to private, but the value shows as a local symbol (STB_LOCAL in
Rafael Espindola6de96a12009-01-15 20:18:42 +0000508 the case of ELF) in the object file. This corresponds to the notion of the
Chris Lattnere20b4702007-01-14 06:51:48 +0000509 '<tt>static</tt>' keyword in C.
Chris Lattner6af02f32004-12-09 16:11:40 +0000510 </dd>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000511
Chris Lattner184f1be2009-04-13 05:44:34 +0000512 <dt><tt><b><a name="available_externally">available_externally</a></b></tt>:
513 </dt>
514
515 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
516 into the object file corresponding to the LLVM module. They exist to
517 allow inlining and other optimizations to take place given knowledge of the
518 definition of the global, which is known to be somewhere outside the module.
519 Globals with <tt>available_externally</tt> linkage are allowed to be discarded
520 at will, and are otherwise the same as <tt>linkonce_odr</tt>. This linkage
521 type is only allowed on definitions, not declarations.</dd>
522
Chris Lattner6af02f32004-12-09 16:11:40 +0000523 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000524
Chris Lattnere20b4702007-01-14 06:51:48 +0000525 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
526 the same name when linkage occurs. This is typically used to implement
527 inline functions, templates, or other code which must be generated in each
528 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
529 allowed to be discarded.
Chris Lattner6af02f32004-12-09 16:11:40 +0000530 </dd>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000531
Dale Johannesen4188aad2008-05-23 23:13:41 +0000532 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
533
534 <dd>"<tt>common</tt>" linkage is exactly the same as <tt>linkonce</tt>
535 linkage, except that unreferenced <tt>common</tt> globals may not be
536 discarded. This is used for globals that may be emitted in multiple
537 translation units, but that are not guaranteed to be emitted into every
538 translation unit that uses them. One example of this is tentative
539 definitions in C, such as "<tt>int X;</tt>" at global scope.
540 </dd>
541
Chris Lattner6af02f32004-12-09 16:11:40 +0000542 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000543
Dale Johannesen4188aad2008-05-23 23:13:41 +0000544 <dd>"<tt>weak</tt>" linkage is the same as <tt>common</tt> linkage, except
545 that some targets may choose to emit different assembly sequences for them
546 for target-dependent reasons. This is used for globals that are declared
547 "weak" in C source code.
Chris Lattner6af02f32004-12-09 16:11:40 +0000548 </dd>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000549
Chris Lattner6af02f32004-12-09 16:11:40 +0000550 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000551
552 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
553 pointer to array type. When two global variables with appending linkage are
554 linked together, the two global arrays are appended together. This is the
555 LLVM, typesafe, equivalent of having the system linker append together
556 "sections" with identical names when .o files are linked.
Chris Lattner6af02f32004-12-09 16:11:40 +0000557 </dd>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000558
Anton Korobeynikova0554d92007-01-12 19:20:47 +0000559 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
Duncan Sands12da8ce2009-03-07 15:45:40 +0000560
Chris Lattner67c37d12008-08-05 18:29:16 +0000561 <dd>The semantics of this linkage follow the ELF object file model: the
562 symbol is weak until linked, if not linked, the symbol becomes null instead
563 of being an undefined reference.
Anton Korobeynikova0554d92007-01-12 19:20:47 +0000564 </dd>
Anton Korobeynikova0554d92007-01-12 19:20:47 +0000565
Duncan Sands12da8ce2009-03-07 15:45:40 +0000566 <dt><tt><b><a name="linkage_linkonce">linkonce_odr</a></b></tt>: </dt>
Duncan Sands12da8ce2009-03-07 15:45:40 +0000567 <dt><tt><b><a name="linkage_weak">weak_odr</a></b></tt>: </dt>
Chris Lattner184f1be2009-04-13 05:44:34 +0000568 <dd>Some languages allow differing globals to be merged, such as two
Duncan Sands12da8ce2009-03-07 15:45:40 +0000569 functions with different semantics. Other languages, such as <tt>C++</tt>,
570 ensure that only equivalent globals are ever merged (the "one definition
Chris Lattner184f1be2009-04-13 05:44:34 +0000571 rule" - "ODR"). Such languages can use the <tt>linkonce_odr</tt>
Duncan Sands4581beb2009-03-11 20:14:15 +0000572 and <tt>weak_odr</tt> linkage types to indicate that the global will only
573 be merged with equivalent globals. These linkage types are otherwise the
574 same as their non-<tt>odr</tt> versions.
Duncan Sands12da8ce2009-03-07 15:45:40 +0000575 </dd>
576
Chris Lattner6af02f32004-12-09 16:11:40 +0000577 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
Chris Lattnerd79749a2004-12-09 16:36:40 +0000578
579 <dd>If none of the above identifiers are used, the global is externally
580 visible, meaning that it participates in linkage and can be used to resolve
581 external symbol references.
Chris Lattner6af02f32004-12-09 16:11:40 +0000582 </dd>
Reid Spencer7972c472007-04-11 23:49:50 +0000583</dl>
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000584
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000585 <p>
586 The next two types of linkage are targeted for Microsoft Windows platform
587 only. They are designed to support importing (exporting) symbols from (to)
Chris Lattner67c37d12008-08-05 18:29:16 +0000588 DLLs (Dynamic Link Libraries).
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000589 </p>
590
Anton Korobeynikova0554d92007-01-12 19:20:47 +0000591 <dl>
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000592 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
593
594 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
595 or variable via a global pointer to a pointer that is set up by the DLL
596 exporting the symbol. On Microsoft Windows targets, the pointer name is
Dan Gohman33a9cef2009-01-12 21:35:55 +0000597 formed by combining <code>__imp_</code> and the function or variable name.
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000598 </dd>
599
600 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
601
602 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
603 pointer to a pointer in a DLL, so that it can be referenced with the
604 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
Dan Gohman33a9cef2009-01-12 21:35:55 +0000605 name is formed by combining <code>__imp_</code> and the function or variable
Anton Korobeynikovd61d39e2006-09-14 18:23:27 +0000606 name.
607 </dd>
608
Chris Lattner6af02f32004-12-09 16:11:40 +0000609</dl>
610
Dan Gohman8ef44982008-11-24 17:18:39 +0000611<p>For example, since the "<tt>.LC0</tt>"
Chris Lattner6af02f32004-12-09 16:11:40 +0000612variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
613variable and was linked with this one, one of the two would be renamed,
614preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
615external (i.e., lacking any linkage declarations), they are accessible
Reid Spencer92c671e2007-01-05 00:59:10 +0000616outside of the current module.</p>
617<p>It is illegal for a function <i>declaration</i>
Duncan Sandse2881052009-03-11 08:08:06 +0000618to have any linkage type other than "externally visible", <tt>dllimport</tt>
619or <tt>extern_weak</tt>.</p>
Duncan Sands12da8ce2009-03-07 15:45:40 +0000620<p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
621or <tt>weak_odr</tt> linkages.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000622</div>
623
624<!-- ======================================================================= -->
625<div class="doc_subsection">
Chris Lattner0132aff2005-05-06 22:57:40 +0000626 <a name="callingconv">Calling Conventions</a>
627</div>
628
629<div class="doc_text">
630
631<p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
632and <a href="#i_invoke">invokes</a> can all have an optional calling convention
633specified for the call. The calling convention of any pair of dynamic
634caller/callee must match, or the behavior of the program is undefined. The
635following calling conventions are supported by LLVM, and more may be added in
636the future:</p>
637
638<dl>
639 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
640
641 <dd>This calling convention (the default if no other calling convention is
642 specified) matches the target C calling conventions. This calling convention
John Criswell02fdc6f2005-05-12 16:52:32 +0000643 supports varargs function calls and tolerates some mismatch in the declared
Reid Spencer72ba4992006-12-31 21:30:18 +0000644 prototype and implemented declaration of the function (as does normal C).
Chris Lattner0132aff2005-05-06 22:57:40 +0000645 </dd>
646
647 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
648
649 <dd>This calling convention attempts to make calls as fast as possible
650 (e.g. by passing things in registers). This calling convention allows the
651 target to use whatever tricks it wants to produce fast code for the target,
Chris Lattner67c37d12008-08-05 18:29:16 +0000652 without having to conform to an externally specified ABI (Application Binary
653 Interface). Implementations of this convention should allow arbitrary
Arnold Schwaighofer2c6b8882008-05-14 09:17:12 +0000654 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> to be
655 supported. This calling convention does not support varargs and requires the
656 prototype of all callees to exactly match the prototype of the function
657 definition.
Chris Lattner0132aff2005-05-06 22:57:40 +0000658 </dd>
659
660 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
661
662 <dd>This calling convention attempts to make code in the caller as efficient
663 as possible under the assumption that the call is not commonly executed. As
664 such, these calls often preserve all registers so that the call does not break
665 any live ranges in the caller side. This calling convention does not support
666 varargs and requires the prototype of all callees to exactly match the
667 prototype of the function definition.
668 </dd>
669
Chris Lattner573f64e2005-05-07 01:46:40 +0000670 <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
Chris Lattner0132aff2005-05-06 22:57:40 +0000671
672 <dd>Any calling convention may be specified by number, allowing
673 target-specific calling conventions to be used. Target specific calling
674 conventions start at 64.
675 </dd>
Chris Lattner573f64e2005-05-07 01:46:40 +0000676</dl>
Chris Lattner0132aff2005-05-06 22:57:40 +0000677
678<p>More calling conventions can be added/defined on an as-needed basis, to
679support pascal conventions or any other well-known target-independent
680convention.</p>
681
682</div>
683
684<!-- ======================================================================= -->
685<div class="doc_subsection">
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000686 <a name="visibility">Visibility Styles</a>
687</div>
688
689<div class="doc_text">
690
691<p>
692All Global Variables and Functions have one of the following visibility styles:
693</p>
694
695<dl>
696 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
697
Chris Lattner67c37d12008-08-05 18:29:16 +0000698 <dd>On targets that use the ELF object file format, default visibility means
699 that the declaration is visible to other
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000700 modules and, in shared libraries, means that the declared entity may be
701 overridden. On Darwin, default visibility means that the declaration is
702 visible to other modules. Default visibility corresponds to "external
703 linkage" in the language.
704 </dd>
705
706 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
707
708 <dd>Two declarations of an object with hidden visibility refer to the same
709 object if they are in the same shared object. Usually, hidden visibility
710 indicates that the symbol will not be placed into the dynamic symbol table,
711 so no other module (executable or shared library) can reference it
712 directly.
713 </dd>
714
Anton Korobeynikov39f3cff2007-04-29 18:35:00 +0000715 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
716
717 <dd>On ELF, protected visibility indicates that the symbol will be placed in
718 the dynamic symbol table, but that references within the defining module will
719 bind to the local symbol. That is, the symbol cannot be overridden by another
720 module.
721 </dd>
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000722</dl>
723
724</div>
725
726<!-- ======================================================================= -->
727<div class="doc_subsection">
Chris Lattnerbc088212009-01-11 20:53:49 +0000728 <a name="namedtypes">Named Types</a>
729</div>
730
731<div class="doc_text">
732
733<p>LLVM IR allows you to specify name aliases for certain types. This can make
734it easier to read the IR and make the IR more condensed (particularly when
735recursive types are involved). An example of a name specification is:
736</p>
737
738<div class="doc_code">
739<pre>
740%mytype = type { %mytype*, i32 }
741</pre>
742</div>
743
744<p>You may give a name to any <a href="#typesystem">type</a> except "<a
745href="t_void">void</a>". Type name aliases may be used anywhere a type is
746expected with the syntax "%mytype".</p>
747
748<p>Note that type names are aliases for the structural type that they indicate,
749and that you can therefore specify multiple names for the same type. This often
750leads to confusing behavior when dumping out a .ll file. Since LLVM IR uses
751structural typing, the name is not part of the type. When printing out LLVM IR,
752the printer will pick <em>one name</em> to render all types of a particular
753shape. This means that if you have code where two different source types end up
754having the same LLVM type, that the dumper will sometimes print the "wrong" or
755unexpected type. This is an important design point and isn't going to
756change.</p>
757
758</div>
759
Chris Lattnerbc088212009-01-11 20:53:49 +0000760<!-- ======================================================================= -->
761<div class="doc_subsection">
Chris Lattner6af02f32004-12-09 16:11:40 +0000762 <a name="globalvars">Global Variables</a>
763</div>
764
765<div class="doc_text">
766
Chris Lattner5d5aede2005-02-12 19:30:21 +0000767<p>Global variables define regions of memory allocated at compilation time
Chris Lattner662c8722005-11-12 00:45:07 +0000768instead of run-time. Global variables may optionally be initialized, may have
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000769an explicit section to be placed in, and may have an optional explicit alignment
770specified. A variable may be defined as "thread_local", which means that it
771will not be shared by threads (each thread will have a separated copy of the
772variable). A variable may be defined as a global "constant," which indicates
773that the contents of the variable will <b>never</b> be modified (enabling better
Chris Lattner5d5aede2005-02-12 19:30:21 +0000774optimization, allowing the global data to be placed in the read-only section of
775an executable, etc). Note that variables that need runtime initialization
John Criswell4c0cf7f2005-10-24 16:17:18 +0000776cannot be marked "constant" as there is a store to the variable.</p>
Chris Lattner5d5aede2005-02-12 19:30:21 +0000777
778<p>
779LLVM explicitly allows <em>declarations</em> of global variables to be marked
780constant, even if the final definition of the global is not. This capability
781can be used to enable slightly better optimization of the program, but requires
782the language definition to guarantee that optimizations based on the
783'constantness' are valid for the translation units that do not include the
784definition.
785</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000786
787<p>As SSA values, global variables define pointer values that are in
788scope (i.e. they dominate) all basic blocks in the program. Global
789variables always define a pointer to their "content" type because they
790describe a region of memory, and all memory objects in LLVM are
791accessed through pointers.</p>
792
Christopher Lamb308121c2007-12-11 09:31:00 +0000793<p>A global variable may be declared to reside in a target-specifc numbered
794address space. For targets that support them, address spaces may affect how
795optimizations are performed and/or what target instructions are used to access
Christopher Lamb25f50762007-12-12 08:44:39 +0000796the variable. The default address space is zero. The address space qualifier
797must precede any other attributes.</p>
Christopher Lamb308121c2007-12-11 09:31:00 +0000798
Chris Lattner662c8722005-11-12 00:45:07 +0000799<p>LLVM allows an explicit section to be specified for globals. If the target
800supports it, it will emit globals to the section specified.</p>
801
Chris Lattner54611b42005-11-06 08:02:57 +0000802<p>An explicit alignment may be specified for a global. If not present, or if
803the alignment is set to zero, the alignment of the global is set by the target
804to whatever it feels convenient. If an explicit alignment is specified, the
805global is forced to have at least that much alignment. All alignments must be
806a power of 2.</p>
807
Christopher Lamb308121c2007-12-11 09:31:00 +0000808<p>For example, the following defines a global in a numbered address space with
809an initializer, section, and alignment:</p>
Chris Lattner5760c502007-01-14 00:27:09 +0000810
Bill Wendling3716c5d2007-05-29 09:04:49 +0000811<div class="doc_code">
Chris Lattner5760c502007-01-14 00:27:09 +0000812<pre>
Dan Gohmanaaa679b2009-01-11 00:40:00 +0000813@G = addrspace(5) constant float 1.0, section "foo", align 4
Chris Lattner5760c502007-01-14 00:27:09 +0000814</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000815</div>
Chris Lattner5760c502007-01-14 00:27:09 +0000816
Chris Lattner6af02f32004-12-09 16:11:40 +0000817</div>
818
819
820<!-- ======================================================================= -->
821<div class="doc_subsection">
822 <a name="functionstructure">Functions</a>
823</div>
824
825<div class="doc_text">
826
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000827<p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
828an optional <a href="#linkage">linkage type</a>, an optional
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000829<a href="#visibility">visibility style</a>, an optional
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000830<a href="#callingconv">calling convention</a>, a return type, an optional
831<a href="#paramattrs">parameter attribute</a> for the return type, a function
832name, a (possibly empty) argument list (each with optional
Devang Patel7e9b05e2008-10-06 18:50:38 +0000833<a href="#paramattrs">parameter attributes</a>), optional
834<a href="#fnattrs">function attributes</a>, an optional section,
835an optional alignment, an optional <a href="#gc">garbage collector name</a>,
Chris Lattnercbc4d2a2008-10-04 18:10:21 +0000836an opening curly brace, a list of basic blocks, and a closing curly brace.
Anton Korobeynikovc7f9f3d2007-01-23 12:35:46 +0000837
838LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
839optional <a href="#linkage">linkage type</a>, an optional
840<a href="#visibility">visibility style</a>, an optional
841<a href="#callingconv">calling convention</a>, a return type, an optional
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000842<a href="#paramattrs">parameter attribute</a> for the return type, a function
Gordon Henriksen71183b62007-12-10 03:18:06 +0000843name, a possibly empty list of arguments, an optional alignment, and an optional
Gordon Henriksendc5cafb2007-12-10 03:30:21 +0000844<a href="#gc">garbage collector name</a>.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +0000845
Chris Lattner67c37d12008-08-05 18:29:16 +0000846<p>A function definition contains a list of basic blocks, forming the CFG
847(Control Flow Graph) for
Chris Lattner6af02f32004-12-09 16:11:40 +0000848the function. Each basic block may optionally start with a label (giving the
849basic block a symbol table entry), contains a list of instructions, and ends
850with a <a href="#terminators">terminator</a> instruction (such as a branch or
851function return).</p>
852
Chris Lattnera59fb102007-06-08 16:52:14 +0000853<p>The first basic block in a function is special in two ways: it is immediately
Chris Lattner6af02f32004-12-09 16:11:40 +0000854executed on entrance to the function, and it is not allowed to have predecessor
855basic blocks (i.e. there can not be any branches to the entry block of a
856function). Because the block can have no predecessors, it also cannot have any
857<a href="#i_phi">PHI nodes</a>.</p>
858
Chris Lattner662c8722005-11-12 00:45:07 +0000859<p>LLVM allows an explicit section to be specified for functions. If the target
860supports it, it will emit functions to the section specified.</p>
861
Chris Lattner54611b42005-11-06 08:02:57 +0000862<p>An explicit alignment may be specified for a function. If not present, or if
863the alignment is set to zero, the alignment of the function is set by the target
864to whatever it feels convenient. If an explicit alignment is specified, the
865function is forced to have at least that much alignment. All alignments must be
866a power of 2.</p>
867
Devang Patel02256232008-10-07 17:48:33 +0000868 <h5>Syntax:</h5>
869
870<div class="doc_code">
Chris Lattner0ae02092008-10-13 16:55:18 +0000871<tt>
872define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
873 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
874 &lt;ResultType&gt; @&lt;FunctionName&gt; ([argument list])
875 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
876 [<a href="#gc">gc</a>] { ... }
877</tt>
Devang Patel02256232008-10-07 17:48:33 +0000878</div>
879
Chris Lattner6af02f32004-12-09 16:11:40 +0000880</div>
881
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000882
883<!-- ======================================================================= -->
884<div class="doc_subsection">
885 <a name="aliasstructure">Aliases</a>
886</div>
887<div class="doc_text">
888 <p>Aliases act as "second name" for the aliasee value (which can be either
Anton Korobeynikov25b2e822008-03-22 08:36:14 +0000889 function, global variable, another alias or bitcast of global value). Aliases
890 may have an optional <a href="#linkage">linkage type</a>, and an
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000891 optional <a href="#visibility">visibility style</a>.</p>
892
893 <h5>Syntax:</h5>
894
Bill Wendling3716c5d2007-05-29 09:04:49 +0000895<div class="doc_code">
Bill Wendling2d8b9a82007-05-29 09:42:13 +0000896<pre>
Duncan Sands7e99a942008-09-12 20:48:21 +0000897@&lt;Name&gt; = alias [Linkage] [Visibility] &lt;AliaseeTy&gt; @&lt;Aliasee&gt;
Bill Wendling2d8b9a82007-05-29 09:42:13 +0000898</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +0000899</div>
Anton Korobeynikova97b6942007-04-25 14:27:10 +0000900
901</div>
902
903
904
Chris Lattner91c15c42006-01-23 23:23:47 +0000905<!-- ======================================================================= -->
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000906<div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
907<div class="doc_text">
908 <p>The return type and each parameter of a function type may have a set of
909 <i>parameter attributes</i> associated with them. Parameter attributes are
910 used to communicate additional information about the result or parameters of
Duncan Sandsad0ea2d2007-11-27 13:23:08 +0000911 a function. Parameter attributes are considered to be part of the function,
912 not of the function type, so functions with different parameter attributes
913 can have the same function type.</p>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000914
Reid Spencercf7ebf52007-01-15 18:27:39 +0000915 <p>Parameter attributes are simple keywords that follow the type specified. If
916 multiple parameter attributes are needed, they are space separated. For
Bill Wendling3716c5d2007-05-29 09:04:49 +0000917 example:</p>
918
919<div class="doc_code">
920<pre>
Nick Lewyckydac78d82009-02-15 23:06:14 +0000921declare i32 @printf(i8* noalias nocapture, ...)
Chris Lattnerd2597d72008-10-04 18:33:34 +0000922declare i32 @atoi(i8 zeroext)
923declare signext i8 @returns_signed_char()
Bill Wendling3716c5d2007-05-29 09:04:49 +0000924</pre>
925</div>
926
Duncan Sandsad0ea2d2007-11-27 13:23:08 +0000927 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
928 <tt>readonly</tt>) come immediately after the argument list.</p>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000929
Anton Korobeynikova0554d92007-01-12 19:20:47 +0000930 <p>Currently, only the following parameter attributes are defined:</p>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000931 <dl>
Reid Spencer314e1cb2007-07-19 23:13:04 +0000932 <dt><tt>zeroext</tt></dt>
Chris Lattnerd2597d72008-10-04 18:33:34 +0000933 <dd>This indicates to the code generator that the parameter or return value
934 should be zero-extended to a 32-bit value by the caller (for a parameter)
935 or the callee (for a return value).</dd>
Chris Lattner5cee13f2008-01-11 06:20:47 +0000936
Reid Spencer314e1cb2007-07-19 23:13:04 +0000937 <dt><tt>signext</tt></dt>
Chris Lattnerd2597d72008-10-04 18:33:34 +0000938 <dd>This indicates to the code generator that the parameter or return value
939 should be sign-extended to a 32-bit value by the caller (for a parameter)
940 or the callee (for a return value).</dd>
Chris Lattner5cee13f2008-01-11 06:20:47 +0000941
Anton Korobeynikove8166852007-01-28 14:30:45 +0000942 <dt><tt>inreg</tt></dt>
Dale Johannesenc50ada22008-09-25 20:47:45 +0000943 <dd>This indicates that this parameter or return value should be treated
944 in a special target-dependent fashion during while emitting code for a
945 function call or return (usually, by putting it in a register as opposed
Chris Lattnerd2597d72008-10-04 18:33:34 +0000946 to memory, though some targets use it to distinguish between two different
947 kinds of registers). Use of this attribute is target-specific.</dd>
Chris Lattner5cee13f2008-01-11 06:20:47 +0000948
Duncan Sands2a1d8ba2008-10-06 08:14:18 +0000949 <dt><tt><a name="byval">byval</a></tt></dt>
Chris Lattner352ab9b2008-01-15 04:34:22 +0000950 <dd>This indicates that the pointer parameter should really be passed by
951 value to the function. The attribute implies that a hidden copy of the
952 pointee is made between the caller and the callee, so the callee is unable
Chris Lattner1ca5c642008-08-05 18:21:08 +0000953 to modify the value in the callee. This attribute is only valid on LLVM
Chris Lattner352ab9b2008-01-15 04:34:22 +0000954 pointer arguments. It is generally used to pass structs and arrays by
Duncan Sands2a1d8ba2008-10-06 08:14:18 +0000955 value, but is also valid on pointers to scalars. The copy is considered to
956 belong to the caller not the callee (for example,
957 <tt><a href="#readonly">readonly</a></tt> functions should not write to
Devang Patel7e9b05e2008-10-06 18:50:38 +0000958 <tt>byval</tt> parameters). This is not a valid attribute for return
Chris Lattner08aa9062009-02-05 05:42:28 +0000959 values. The byval attribute also supports specifying an alignment with the
960 align attribute. This has a target-specific effect on the code generator
961 that usually indicates a desired alignment for the synthesized stack
962 slot.</dd>
Chris Lattner5cee13f2008-01-11 06:20:47 +0000963
Anton Korobeynikove8166852007-01-28 14:30:45 +0000964 <dt><tt>sret</tt></dt>
Duncan Sandsfa4b6732008-02-18 04:19:38 +0000965 <dd>This indicates that the pointer parameter specifies the address of a
966 structure that is the return value of the function in the source program.
Chris Lattnerd2597d72008-10-04 18:33:34 +0000967 This pointer must be guaranteed by the caller to be valid: loads and stores
968 to the structure may be assumed by the callee to not to trap. This may only
Devang Patel7e9b05e2008-10-06 18:50:38 +0000969 be applied to the first parameter. This is not a valid attribute for
970 return values. </dd>
Chris Lattner5cee13f2008-01-11 06:20:47 +0000971
Zhou Sheng2444a9a2007-06-05 05:28:26 +0000972 <dt><tt>noalias</tt></dt>
Nick Lewyckyf5ffcbc2008-11-24 03:41:24 +0000973 <dd>This indicates that the pointer does not alias any global or any other
974 parameter. The caller is responsible for ensuring that this is the
Nick Lewyckyd59572c2008-11-24 05:00:44 +0000975 case. On a function return value, <tt>noalias</tt> additionally indicates
976 that the pointer does not alias any other pointers visible to the
Nick Lewycky2abb1082008-12-19 06:39:12 +0000977 caller. For further details, please see the discussion of the NoAlias
978 response in
979 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
980 analysis</a>.</dd>
981
982 <dt><tt>nocapture</tt></dt>
983 <dd>This indicates that the callee does not make any copies of the pointer
984 that outlive the callee itself. This is not a valid attribute for return
985 values.</dd>
Chris Lattner5cee13f2008-01-11 06:20:47 +0000986
Duncan Sands27e91592007-07-27 19:57:41 +0000987 <dt><tt>nest</tt></dt>
Duncan Sands825bde42008-07-08 09:27:25 +0000988 <dd>This indicates that the pointer parameter can be excised using the
Devang Patel7e9b05e2008-10-06 18:50:38 +0000989 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
990 attribute for return values.</dd>
Anton Korobeynikova0554d92007-01-12 19:20:47 +0000991 </dl>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000992
Reid Spencerb5ebf3d2006-12-31 07:07:53 +0000993</div>
994
995<!-- ======================================================================= -->
Chris Lattner91c15c42006-01-23 23:23:47 +0000996<div class="doc_subsection">
Gordon Henriksen71183b62007-12-10 03:18:06 +0000997 <a name="gc">Garbage Collector Names</a>
998</div>
999
1000<div class="doc_text">
1001<p>Each function may specify a garbage collector name, which is simply a
1002string.</p>
1003
1004<div class="doc_code"><pre
1005>define void @f() gc "name" { ...</pre></div>
1006
1007<p>The compiler declares the supported values of <i>name</i>. Specifying a
1008collector which will cause the compiler to alter its output in order to support
1009the named garbage collection algorithm.</p>
1010</div>
1011
1012<!-- ======================================================================= -->
1013<div class="doc_subsection">
Devang Patel9eb525d2008-09-26 23:51:19 +00001014 <a name="fnattrs">Function Attributes</a>
Devang Patelcaacdba2008-09-04 23:05:13 +00001015</div>
1016
1017<div class="doc_text">
Devang Patel9eb525d2008-09-26 23:51:19 +00001018
1019<p>Function attributes are set to communicate additional information about
1020 a function. Function attributes are considered to be part of the function,
1021 not of the function type, so functions with different parameter attributes
1022 can have the same function type.</p>
1023
1024 <p>Function attributes are simple keywords that follow the type specified. If
1025 multiple attributes are needed, they are space separated. For
1026 example:</p>
Devang Patelcaacdba2008-09-04 23:05:13 +00001027
1028<div class="doc_code">
Bill Wendlingb175fa42008-09-07 10:26:33 +00001029<pre>
Devang Patel9eb525d2008-09-26 23:51:19 +00001030define void @f() noinline { ... }
1031define void @f() alwaysinline { ... }
1032define void @f() alwaysinline optsize { ... }
1033define void @f() optsize
Bill Wendlingb175fa42008-09-07 10:26:33 +00001034</pre>
Devang Patelcaacdba2008-09-04 23:05:13 +00001035</div>
1036
Bill Wendlingb175fa42008-09-07 10:26:33 +00001037<dl>
Devang Patel9eb525d2008-09-26 23:51:19 +00001038<dt><tt>alwaysinline</tt></dt>
Chris Lattnerfbf60a42008-10-04 18:23:17 +00001039<dd>This attribute indicates that the inliner should attempt to inline this
1040function into callers whenever possible, ignoring any active inlining size
1041threshold for this caller.</dd>
Bill Wendlingb175fa42008-09-07 10:26:33 +00001042
Devang Patel9eb525d2008-09-26 23:51:19 +00001043<dt><tt>noinline</tt></dt>
Chris Lattnerfbf60a42008-10-04 18:23:17 +00001044<dd>This attribute indicates that the inliner should never inline this function
Chris Lattner0625c282008-10-05 17:14:59 +00001045in any situation. This attribute may not be used together with the
Chris Lattnerfbf60a42008-10-04 18:23:17 +00001046<tt>alwaysinline</tt> attribute.</dd>
Bill Wendlingb175fa42008-09-07 10:26:33 +00001047
Devang Patel9eb525d2008-09-26 23:51:19 +00001048<dt><tt>optsize</tt></dt>
Devang Patele9743902008-09-29 18:34:44 +00001049<dd>This attribute suggests that optimization passes and code generator passes
Chris Lattnerfbf60a42008-10-04 18:23:17 +00001050make choices that keep the code size of this function low, and otherwise do
1051optimizations specifically to reduce code size.</dd>
Bill Wendlingb175fa42008-09-07 10:26:33 +00001052
Devang Patel9eb525d2008-09-26 23:51:19 +00001053<dt><tt>noreturn</tt></dt>
Chris Lattnerfbf60a42008-10-04 18:23:17 +00001054<dd>This function attribute indicates that the function never returns normally.
1055This produces undefined behavior at runtime if the function ever does
1056dynamically return.</dd>
Devang Patel9eb525d2008-09-26 23:51:19 +00001057
1058<dt><tt>nounwind</tt></dt>
Chris Lattnerfbf60a42008-10-04 18:23:17 +00001059<dd>This function attribute indicates that the function never returns with an
1060unwind or exceptional control flow. If the function does unwind, its runtime
1061behavior is undefined.</dd>
1062
1063<dt><tt>readnone</tt></dt>
Duncan Sands1efabaa2009-05-06 06:49:50 +00001064<dd>This attribute indicates that the function computes its result (or decides to
1065unwind an exception) based strictly on its arguments, without dereferencing any
Duncan Sands2a1d8ba2008-10-06 08:14:18 +00001066pointer arguments or otherwise accessing any mutable state (e.g. memory, control
1067registers, etc) visible to caller functions. It does not write through any
1068pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments) and
Duncan Sands1efabaa2009-05-06 06:49:50 +00001069never changes any state visible to callers. This means that it cannot unwind
1070exceptions by calling the <tt>C++</tt> exception throwing methods, but could
1071use the <tt>unwind</tt> instruction.</dd>
Devang Patel9eb525d2008-09-26 23:51:19 +00001072
Duncan Sands2a1d8ba2008-10-06 08:14:18 +00001073<dt><tt><a name="readonly">readonly</a></tt></dt>
1074<dd>This attribute indicates that the function does not write through any
1075pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments)
1076or otherwise modify any state (e.g. memory, control registers, etc) visible to
1077caller functions. It may dereference pointer arguments and read state that may
Duncan Sands1efabaa2009-05-06 06:49:50 +00001078be set in the caller. A readonly function always returns the same value (or
1079unwinds an exception identically) when called with the same set of arguments
1080and global state. It cannot unwind an exception by calling the <tt>C++</tt>
1081exception throwing methods, but may use the <tt>unwind</tt> instruction.</dd>
Bill Wendlinga8130172008-11-13 01:02:51 +00001082
1083<dt><tt><a name="ssp">ssp</a></tt></dt>
Bill Wendling6e41add2008-11-26 19:19:05 +00001084<dd>This attribute indicates that the function should emit a stack smashing
Bill Wendlinga8130172008-11-13 01:02:51 +00001085protector. It is in the form of a "canary"&mdash;a random value placed on the
1086stack before the local variables that's checked upon return from the function to
1087see if it has been overwritten. A heuristic is used to determine if a function
Bill Wendling6e41add2008-11-26 19:19:05 +00001088needs stack protectors or not.
Bill Wendlinga8130172008-11-13 01:02:51 +00001089
Bill Wendling0f5541e2008-11-26 19:07:40 +00001090<p>If a function that has an <tt>ssp</tt> attribute is inlined into a function
1091that doesn't have an <tt>ssp</tt> attribute, then the resulting function will
1092have an <tt>ssp</tt> attribute.</p></dd>
1093
1094<dt><tt>sspreq</tt></dt>
Bill Wendling6e41add2008-11-26 19:19:05 +00001095<dd>This attribute indicates that the function should <em>always</em> emit a
Bill Wendlinga8130172008-11-13 01:02:51 +00001096stack smashing protector. This overrides the <tt><a href="#ssp">ssp</a></tt>
Bill Wendling6e41add2008-11-26 19:19:05 +00001097function attribute.
Bill Wendling0f5541e2008-11-26 19:07:40 +00001098
1099<p>If a function that has an <tt>sspreq</tt> attribute is inlined into a
1100function that doesn't have an <tt>sspreq</tt> attribute or which has
1101an <tt>ssp</tt> attribute, then the resulting function will have
1102an <tt>sspreq</tt> attribute.</p></dd>
Bill Wendlingb175fa42008-09-07 10:26:33 +00001103</dl>
1104
Devang Patelcaacdba2008-09-04 23:05:13 +00001105</div>
1106
1107<!-- ======================================================================= -->
1108<div class="doc_subsection">
Chris Lattner93564892006-04-08 04:40:53 +00001109 <a name="moduleasm">Module-Level Inline Assembly</a>
Chris Lattner91c15c42006-01-23 23:23:47 +00001110</div>
1111
1112<div class="doc_text">
1113<p>
1114Modules may contain "module-level inline asm" blocks, which corresponds to the
1115GCC "file scope inline asm" blocks. These blocks are internally concatenated by
1116LLVM and treated as a single unit, but may be separated in the .ll file if
1117desired. The syntax is very simple:
1118</p>
1119
Bill Wendling3716c5d2007-05-29 09:04:49 +00001120<div class="doc_code">
1121<pre>
1122module asm "inline asm code goes here"
1123module asm "more can go here"
1124</pre>
1125</div>
Chris Lattner91c15c42006-01-23 23:23:47 +00001126
1127<p>The strings can contain any character by escaping non-printable characters.
1128 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1129 for the number.
1130</p>
1131
1132<p>
1133 The inline asm code is simply printed to the machine code .s file when
1134 assembly code is generated.
1135</p>
1136</div>
Chris Lattner6af02f32004-12-09 16:11:40 +00001137
Reid Spencer50c723a2007-02-19 23:54:10 +00001138<!-- ======================================================================= -->
1139<div class="doc_subsection">
1140 <a name="datalayout">Data Layout</a>
1141</div>
1142
1143<div class="doc_text">
1144<p>A module may specify a target specific data layout string that specifies how
Reid Spencer7972c472007-04-11 23:49:50 +00001145data is to be laid out in memory. The syntax for the data layout is simply:</p>
1146<pre> target datalayout = "<i>layout specification</i>"</pre>
1147<p>The <i>layout specification</i> consists of a list of specifications
1148separated by the minus sign character ('-'). Each specification starts with a
1149letter and may include other information after the letter to define some
1150aspect of the data layout. The specifications accepted are as follows: </p>
Reid Spencer50c723a2007-02-19 23:54:10 +00001151<dl>
1152 <dt><tt>E</tt></dt>
1153 <dd>Specifies that the target lays out data in big-endian form. That is, the
1154 bits with the most significance have the lowest address location.</dd>
1155 <dt><tt>e</tt></dt>
Chris Lattner67c37d12008-08-05 18:29:16 +00001156 <dd>Specifies that the target lays out data in little-endian form. That is,
Reid Spencer50c723a2007-02-19 23:54:10 +00001157 the bits with the least significance have the lowest address location.</dd>
1158 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1159 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1160 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1161 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1162 too.</dd>
1163 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1164 <dd>This specifies the alignment for an integer type of a given bit
1165 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1166 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1167 <dd>This specifies the alignment for a vector type of a given bit
1168 <i>size</i>.</dd>
1169 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1170 <dd>This specifies the alignment for a floating point type of a given bit
1171 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1172 (double).</dd>
1173 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1174 <dd>This specifies the alignment for an aggregate type of a given bit
1175 <i>size</i>.</dd>
1176</dl>
1177<p>When constructing the data layout for a given target, LLVM starts with a
1178default set of specifications which are then (possibly) overriden by the
1179specifications in the <tt>datalayout</tt> keyword. The default specifications
1180are given in this list:</p>
1181<ul>
1182 <li><tt>E</tt> - big endian</li>
1183 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1184 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1185 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1186 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1187 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
Chris Lattner67c37d12008-08-05 18:29:16 +00001188 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
Reid Spencer50c723a2007-02-19 23:54:10 +00001189 alignment of 64-bits</li>
1190 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1191 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1192 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1193 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1194 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1195</ul>
Chris Lattner1ca5c642008-08-05 18:21:08 +00001196<p>When LLVM is determining the alignment for a given type, it uses the
Dan Gohmanef9462f2008-10-14 16:51:45 +00001197following rules:</p>
Reid Spencer50c723a2007-02-19 23:54:10 +00001198<ol>
1199 <li>If the type sought is an exact match for one of the specifications, that
1200 specification is used.</li>
1201 <li>If no match is found, and the type sought is an integer type, then the
1202 smallest integer type that is larger than the bitwidth of the sought type is
1203 used. If none of the specifications are larger than the bitwidth then the the
1204 largest integer type is used. For example, given the default specifications
1205 above, the i7 type will use the alignment of i8 (next largest) while both
1206 i65 and i256 will use the alignment of i64 (largest specified).</li>
1207 <li>If no match is found, and the type sought is a vector type, then the
1208 largest vector type that is smaller than the sought vector type will be used
Dan Gohmanef9462f2008-10-14 16:51:45 +00001209 as a fall back. This happens because &lt;128 x double&gt; can be implemented
1210 in terms of 64 &lt;2 x double&gt;, for example.</li>
Reid Spencer50c723a2007-02-19 23:54:10 +00001211</ol>
1212</div>
Chris Lattner6af02f32004-12-09 16:11:40 +00001213
Chris Lattner2f7c9632001-06-06 20:29:01 +00001214<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001215<div class="doc_section"> <a name="typesystem">Type System</a> </div>
1216<!-- *********************************************************************** -->
Chris Lattner6af02f32004-12-09 16:11:40 +00001217
Misha Brukman76307852003-11-08 01:05:38 +00001218<div class="doc_text">
Chris Lattner6af02f32004-12-09 16:11:40 +00001219
Misha Brukman76307852003-11-08 01:05:38 +00001220<p>The LLVM type system is one of the most important features of the
Chris Lattner48b383b02003-11-25 01:02:51 +00001221intermediate representation. Being typed enables a number of
Chris Lattner67c37d12008-08-05 18:29:16 +00001222optimizations to be performed on the intermediate representation directly,
1223without having to do
Chris Lattner48b383b02003-11-25 01:02:51 +00001224extra analyses on the side before the transformation. A strong type
1225system makes it easier to read the generated code and enables novel
1226analyses and transformations that are not feasible to perform on normal
1227three address code representations.</p>
Chris Lattner6af02f32004-12-09 16:11:40 +00001228
1229</div>
1230
Chris Lattner2f7c9632001-06-06 20:29:01 +00001231<!-- ======================================================================= -->
Chris Lattner7824d182008-01-04 04:32:38 +00001232<div class="doc_subsection"> <a name="t_classifications">Type
Chris Lattner48b383b02003-11-25 01:02:51 +00001233Classifications</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00001234<div class="doc_text">
Chris Lattner7824d182008-01-04 04:32:38 +00001235<p>The types fall into a few useful
Chris Lattner48b383b02003-11-25 01:02:51 +00001236classifications:</p>
Misha Brukmanc501f552004-03-01 17:47:27 +00001237
1238<table border="1" cellspacing="0" cellpadding="4">
Chris Lattner48b383b02003-11-25 01:02:51 +00001239 <tbody>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001240 <tr><th>Classification</th><th>Types</th></tr>
Chris Lattner48b383b02003-11-25 01:02:51 +00001241 <tr>
Chris Lattner7824d182008-01-04 04:32:38 +00001242 <td><a href="#t_integer">integer</a></td>
Reid Spencer138249b2007-05-16 18:44:01 +00001243 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
Chris Lattner48b383b02003-11-25 01:02:51 +00001244 </tr>
1245 <tr>
Chris Lattner7824d182008-01-04 04:32:38 +00001246 <td><a href="#t_floating">floating point</a></td>
1247 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
Chris Lattner48b383b02003-11-25 01:02:51 +00001248 </tr>
1249 <tr>
1250 <td><a name="t_firstclass">first class</a></td>
Chris Lattner7824d182008-01-04 04:32:38 +00001251 <td><a href="#t_integer">integer</a>,
1252 <a href="#t_floating">floating point</a>,
1253 <a href="#t_pointer">pointer</a>,
Dan Gohman08783a882008-06-18 18:42:13 +00001254 <a href="#t_vector">vector</a>,
Dan Gohmanb9d66602008-05-12 23:51:09 +00001255 <a href="#t_struct">structure</a>,
1256 <a href="#t_array">array</a>,
Dan Gohmanda52d212008-05-23 22:50:26 +00001257 <a href="#t_label">label</a>.
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001258 </td>
Chris Lattner48b383b02003-11-25 01:02:51 +00001259 </tr>
Chris Lattner7824d182008-01-04 04:32:38 +00001260 <tr>
1261 <td><a href="#t_primitive">primitive</a></td>
1262 <td><a href="#t_label">label</a>,
1263 <a href="#t_void">void</a>,
Chris Lattner7824d182008-01-04 04:32:38 +00001264 <a href="#t_floating">floating point</a>.</td>
1265 </tr>
1266 <tr>
1267 <td><a href="#t_derived">derived</a></td>
1268 <td><a href="#t_integer">integer</a>,
1269 <a href="#t_array">array</a>,
1270 <a href="#t_function">function</a>,
1271 <a href="#t_pointer">pointer</a>,
1272 <a href="#t_struct">structure</a>,
1273 <a href="#t_pstruct">packed structure</a>,
1274 <a href="#t_vector">vector</a>,
1275 <a href="#t_opaque">opaque</a>.
Dan Gohman93bf60d2008-10-14 16:32:04 +00001276 </td>
Chris Lattner7824d182008-01-04 04:32:38 +00001277 </tr>
Chris Lattner48b383b02003-11-25 01:02:51 +00001278 </tbody>
Misha Brukman76307852003-11-08 01:05:38 +00001279</table>
Misha Brukmanc501f552004-03-01 17:47:27 +00001280
Chris Lattner48b383b02003-11-25 01:02:51 +00001281<p>The <a href="#t_firstclass">first class</a> types are perhaps the
1282most important. Values of these types are the only ones which can be
1283produced by instructions, passed as arguments, or used as operands to
Dan Gohman34d1c0d2008-05-23 21:53:15 +00001284instructions.</p>
Misha Brukman76307852003-11-08 01:05:38 +00001285</div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001286
Chris Lattner2f7c9632001-06-06 20:29:01 +00001287<!-- ======================================================================= -->
Chris Lattner7824d182008-01-04 04:32:38 +00001288<div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
Chris Lattner43542b32008-01-04 04:34:14 +00001289
Chris Lattner7824d182008-01-04 04:32:38 +00001290<div class="doc_text">
1291<p>The primitive types are the fundamental building blocks of the LLVM
1292system.</p>
1293
Chris Lattner43542b32008-01-04 04:34:14 +00001294</div>
1295
Chris Lattner7824d182008-01-04 04:32:38 +00001296<!-- _______________________________________________________________________ -->
1297<div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1298
1299<div class="doc_text">
1300 <table>
1301 <tbody>
1302 <tr><th>Type</th><th>Description</th></tr>
1303 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1304 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1305 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1306 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1307 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1308 </tbody>
1309 </table>
1310</div>
1311
1312<!-- _______________________________________________________________________ -->
1313<div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1314
1315<div class="doc_text">
1316<h5>Overview:</h5>
1317<p>The void type does not represent any value and has no size.</p>
1318
1319<h5>Syntax:</h5>
1320
1321<pre>
1322 void
1323</pre>
1324</div>
1325
1326<!-- _______________________________________________________________________ -->
1327<div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1328
1329<div class="doc_text">
1330<h5>Overview:</h5>
1331<p>The label type represents code labels.</p>
1332
1333<h5>Syntax:</h5>
1334
1335<pre>
1336 label
1337</pre>
1338</div>
1339
1340
1341<!-- ======================================================================= -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001342<div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001343
Misha Brukman76307852003-11-08 01:05:38 +00001344<div class="doc_text">
Chris Lattner74d3f822004-12-09 17:30:23 +00001345
Chris Lattner48b383b02003-11-25 01:02:51 +00001346<p>The real power in LLVM comes from the derived types in the system.
1347This is what allows a programmer to represent arrays, functions,
1348pointers, and other useful types. Note that these derived types may be
1349recursive: For example, it is possible to have a two dimensional array.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001350
Misha Brukman76307852003-11-08 01:05:38 +00001351</div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001352
Chris Lattner2f7c9632001-06-06 20:29:01 +00001353<!-- _______________________________________________________________________ -->
Reid Spencer138249b2007-05-16 18:44:01 +00001354<div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1355
1356<div class="doc_text">
1357
1358<h5>Overview:</h5>
1359<p>The integer type is a very simple derived type that simply specifies an
1360arbitrary bit width for the integer type desired. Any bit width from 1 bit to
13612^23-1 (about 8 million) can be specified.</p>
1362
1363<h5>Syntax:</h5>
1364
1365<pre>
1366 iN
1367</pre>
1368
1369<p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1370value.</p>
1371
1372<h5>Examples:</h5>
1373<table class="layout">
Nick Lewyckyaab930a2009-05-24 02:46:06 +00001374 <tr class="layout">
1375 <td class="left"><tt>i1</tt></td>
1376 <td class="left">a single-bit integer.</td>
Reid Spencer138249b2007-05-16 18:44:01 +00001377 </tr>
Nick Lewyckyaab930a2009-05-24 02:46:06 +00001378 <tr class="layout">
1379 <td class="left"><tt>i32</tt></td>
1380 <td class="left">a 32-bit integer.</td>
1381 </tr>
1382 <tr class="layout">
1383 <td class="left"><tt>i1942652</tt></td>
1384 <td class="left">a really big integer of over 1 million bits.</td>
1385 </tr>
Reid Spencer138249b2007-05-16 18:44:01 +00001386</table>
Dan Gohman142ccc02009-01-24 15:58:40 +00001387
1388<p>Note that the code generator does not yet support large integer types
1389to be used as function return types. The specific limit on how large a
1390return type the code generator can currently handle is target-dependent;
1391currently it's often 64 bits for 32-bit targets and 128 bits for 64-bit
1392targets.</p>
1393
Bill Wendling3716c5d2007-05-29 09:04:49 +00001394</div>
Reid Spencer138249b2007-05-16 18:44:01 +00001395
1396<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001397<div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001398
Misha Brukman76307852003-11-08 01:05:38 +00001399<div class="doc_text">
Chris Lattner74d3f822004-12-09 17:30:23 +00001400
Chris Lattner2f7c9632001-06-06 20:29:01 +00001401<h5>Overview:</h5>
Chris Lattner74d3f822004-12-09 17:30:23 +00001402
Misha Brukman76307852003-11-08 01:05:38 +00001403<p>The array type is a very simple derived type that arranges elements
Chris Lattner48b383b02003-11-25 01:02:51 +00001404sequentially in memory. The array type requires a size (number of
1405elements) and an underlying data type.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001406
Chris Lattner590645f2002-04-14 06:13:44 +00001407<h5>Syntax:</h5>
Chris Lattner74d3f822004-12-09 17:30:23 +00001408
1409<pre>
1410 [&lt;# elements&gt; x &lt;elementtype&gt;]
1411</pre>
1412
John Criswell02fdc6f2005-05-12 16:52:32 +00001413<p>The number of elements is a constant integer value; elementtype may
Chris Lattner48b383b02003-11-25 01:02:51 +00001414be any type with a size.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001415
Chris Lattner590645f2002-04-14 06:13:44 +00001416<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001417<table class="layout">
1418 <tr class="layout">
Chris Lattner747359f2007-12-19 05:04:11 +00001419 <td class="left"><tt>[40 x i32]</tt></td>
1420 <td class="left">Array of 40 32-bit integer values.</td>
1421 </tr>
1422 <tr class="layout">
1423 <td class="left"><tt>[41 x i32]</tt></td>
1424 <td class="left">Array of 41 32-bit integer values.</td>
1425 </tr>
1426 <tr class="layout">
1427 <td class="left"><tt>[4 x i8]</tt></td>
1428 <td class="left">Array of 4 8-bit integer values.</td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001429 </tr>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001430</table>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001431<p>Here are some examples of multidimensional arrays:</p>
1432<table class="layout">
1433 <tr class="layout">
Chris Lattner747359f2007-12-19 05:04:11 +00001434 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1435 <td class="left">3x4 array of 32-bit integer values.</td>
1436 </tr>
1437 <tr class="layout">
1438 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1439 <td class="left">12x10 array of single precision floating point values.</td>
1440 </tr>
1441 <tr class="layout">
1442 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1443 <td class="left">2x3x4 array of 16-bit integer values.</td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001444 </tr>
1445</table>
Chris Lattnerc0ad71e2005-06-24 17:22:57 +00001446
John Criswell4c0cf7f2005-10-24 16:17:18 +00001447<p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1448length array. Normally, accesses past the end of an array are undefined in
Chris Lattnerc0ad71e2005-06-24 17:22:57 +00001449LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1450As a special case, however, zero length arrays are recognized to be variable
1451length. This allows implementation of 'pascal style arrays' with the LLVM
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001452type "{ i32, [0 x float]}", for example.</p>
Chris Lattnerc0ad71e2005-06-24 17:22:57 +00001453
Dan Gohman142ccc02009-01-24 15:58:40 +00001454<p>Note that the code generator does not yet support large aggregate types
1455to be used as function return types. The specific limit on how large an
1456aggregate return type the code generator can currently handle is
1457target-dependent, and also dependent on the aggregate element types.</p>
1458
Misha Brukman76307852003-11-08 01:05:38 +00001459</div>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001460
Chris Lattner2f7c9632001-06-06 20:29:01 +00001461<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001462<div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00001463<div class="doc_text">
Chris Lattnerda508ac2008-04-23 04:59:35 +00001464
Chris Lattner2f7c9632001-06-06 20:29:01 +00001465<h5>Overview:</h5>
Chris Lattnerda508ac2008-04-23 04:59:35 +00001466
Chris Lattner48b383b02003-11-25 01:02:51 +00001467<p>The function type can be thought of as a function signature. It
Devang Patele3dfc1c2008-03-24 05:35:41 +00001468consists of a return type and a list of formal parameter types. The
Chris Lattnerda508ac2008-04-23 04:59:35 +00001469return type of a function type is a scalar type, a void type, or a struct type.
Devang Patel9c1f8b12008-03-24 20:52:42 +00001470If the return type is a struct type then all struct elements must be of first
Chris Lattnerda508ac2008-04-23 04:59:35 +00001471class types, and the struct must have at least one element.</p>
Devang Pateld6cff512008-03-10 20:49:15 +00001472
Chris Lattner2f7c9632001-06-06 20:29:01 +00001473<h5>Syntax:</h5>
Chris Lattnerda508ac2008-04-23 04:59:35 +00001474
1475<pre>
1476 &lt;returntype list&gt; (&lt;parameter list&gt;)
1477</pre>
1478
John Criswell4c0cf7f2005-10-24 16:17:18 +00001479<p>...where '<tt>&lt;parameter list&gt;</tt>' is a comma-separated list of type
Misha Brukman20f9a622004-08-12 20:16:08 +00001480specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
Chris Lattner5ed60612003-09-03 00:41:47 +00001481which indicates that the function takes a variable number of arguments.
1482Variable argument functions can access their arguments with the <a
Devang Pateld6cff512008-03-10 20:49:15 +00001483 href="#int_varargs">variable argument handling intrinsic</a> functions.
1484'<tt>&lt;returntype list&gt;</tt>' is a comma-separated list of
1485<a href="#t_firstclass">first class</a> type specifiers.</p>
Chris Lattnerda508ac2008-04-23 04:59:35 +00001486
Chris Lattner2f7c9632001-06-06 20:29:01 +00001487<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001488<table class="layout">
1489 <tr class="layout">
Reid Spencer58c08712006-12-31 07:18:34 +00001490 <td class="left"><tt>i32 (i32)</tt></td>
1491 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001492 </td>
Reid Spencer58c08712006-12-31 07:18:34 +00001493 </tr><tr class="layout">
Reid Spencer314e1cb2007-07-19 23:13:04 +00001494 <td class="left"><tt>float&nbsp;(i16&nbsp;signext,&nbsp;i32&nbsp;*)&nbsp;*
Reid Spencer655dcc62006-12-31 07:20:23 +00001495 </tt></td>
Reid Spencer58c08712006-12-31 07:18:34 +00001496 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1497 an <tt>i16</tt> that should be sign extended and a
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001498 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
Reid Spencer58c08712006-12-31 07:18:34 +00001499 <tt>float</tt>.
1500 </td>
1501 </tr><tr class="layout">
1502 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1503 <td class="left">A vararg function that takes at least one
Reid Spencer3e628eb92007-01-04 16:43:23 +00001504 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
Reid Spencer58c08712006-12-31 07:18:34 +00001505 which returns an integer. This is the signature for <tt>printf</tt> in
1506 LLVM.
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001507 </td>
Devang Patele3dfc1c2008-03-24 05:35:41 +00001508 </tr><tr class="layout">
1509 <td class="left"><tt>{i32, i32} (i32)</tt></td>
Misha Brukmanc9813bd2008-11-27 06:41:20 +00001510 <td class="left">A function taking an <tt>i32</tt>, returning two
1511 <tt>i32</tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
Devang Patele3dfc1c2008-03-24 05:35:41 +00001512 </td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001513 </tr>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001514</table>
Misha Brukmanc501f552004-03-01 17:47:27 +00001515
Misha Brukman76307852003-11-08 01:05:38 +00001516</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001517<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001518<div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00001519<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00001520<h5>Overview:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001521<p>The structure type is used to represent a collection of data members
1522together in memory. The packing of the field types is defined to match
1523the ABI of the underlying processor. The elements of a structure may
1524be any type that has a size.</p>
1525<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1526and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1527field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1528instruction.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001529<h5>Syntax:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001530<pre> { &lt;type list&gt; }<br></pre>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001531<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001532<table class="layout">
1533 <tr class="layout">
Jeff Cohen5819f182007-04-22 01:17:39 +00001534 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1535 <td class="left">A triple of three <tt>i32</tt> values</td>
1536 </tr><tr class="layout">
1537 <td class="left"><tt>{&nbsp;float,&nbsp;i32&nbsp;(i32)&nbsp;*&nbsp;}</tt></td>
1538 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1539 second element is a <a href="#t_pointer">pointer</a> to a
1540 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1541 an <tt>i32</tt>.</td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001542 </tr>
Chris Lattner2f7c9632001-06-06 20:29:01 +00001543</table>
Dan Gohman142ccc02009-01-24 15:58:40 +00001544
1545<p>Note that the code generator does not yet support large aggregate types
1546to be used as function return types. The specific limit on how large an
1547aggregate return type the code generator can currently handle is
1548target-dependent, and also dependent on the aggregate element types.</p>
1549
Misha Brukman76307852003-11-08 01:05:38 +00001550</div>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001551
Chris Lattner2f7c9632001-06-06 20:29:01 +00001552<!-- _______________________________________________________________________ -->
Andrew Lenharth8df88e22006-12-08 17:13:00 +00001553<div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1554</div>
1555<div class="doc_text">
1556<h5>Overview:</h5>
1557<p>The packed structure type is used to represent a collection of data members
1558together in memory. There is no padding between fields. Further, the alignment
1559of a packed structure is 1 byte. The elements of a packed structure may
1560be any type that has a size.</p>
1561<p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1562and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1563field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1564instruction.</p>
1565<h5>Syntax:</h5>
1566<pre> &lt; { &lt;type list&gt; } &gt; <br></pre>
1567<h5>Examples:</h5>
1568<table class="layout">
1569 <tr class="layout">
Jeff Cohen5819f182007-04-22 01:17:39 +00001570 <td class="left"><tt>&lt; { i32, i32, i32 } &gt;</tt></td>
1571 <td class="left">A triple of three <tt>i32</tt> values</td>
1572 </tr><tr class="layout">
Bill Wendlingb175fa42008-09-07 10:26:33 +00001573 <td class="left">
1574<tt>&lt;&nbsp;{&nbsp;float,&nbsp;i32&nbsp;(i32)*&nbsp;}&nbsp;&gt;</tt></td>
Jeff Cohen5819f182007-04-22 01:17:39 +00001575 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1576 second element is a <a href="#t_pointer">pointer</a> to a
1577 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1578 an <tt>i32</tt>.</td>
Andrew Lenharth8df88e22006-12-08 17:13:00 +00001579 </tr>
1580</table>
1581</div>
1582
1583<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00001584<div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00001585<div class="doc_text">
Chris Lattner590645f2002-04-14 06:13:44 +00001586<h5>Overview:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001587<p>As in many languages, the pointer type represents a pointer or
Christopher Lamb308121c2007-12-11 09:31:00 +00001588reference to another object, which must live in memory. Pointer types may have
1589an optional address space attribute defining the target-specific numbered
1590address space where the pointed-to object resides. The default address space is
1591zero.</p>
Chris Lattner4a67c912009-02-08 19:53:29 +00001592
1593<p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does
Chris Lattnerd1d4cff2009-02-08 22:21:28 +00001594it permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
Chris Lattner4a67c912009-02-08 19:53:29 +00001595
Chris Lattner590645f2002-04-14 06:13:44 +00001596<h5>Syntax:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00001597<pre> &lt;type&gt; *<br></pre>
Chris Lattner590645f2002-04-14 06:13:44 +00001598<h5>Examples:</h5>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001599<table class="layout">
1600 <tr class="layout">
Dan Gohman623806e2009-01-04 23:44:43 +00001601 <td class="left"><tt>[4 x i32]*</tt></td>
Chris Lattner747359f2007-12-19 05:04:11 +00001602 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1603 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1604 </tr>
1605 <tr class="layout">
1606 <td class="left"><tt>i32 (i32 *) *</tt></td>
1607 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001608 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
Chris Lattner747359f2007-12-19 05:04:11 +00001609 <tt>i32</tt>.</td>
1610 </tr>
1611 <tr class="layout">
1612 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1613 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1614 that resides in address space #5.</td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001615 </tr>
Misha Brukman76307852003-11-08 01:05:38 +00001616</table>
Misha Brukman76307852003-11-08 01:05:38 +00001617</div>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001618
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001619<!-- _______________________________________________________________________ -->
Reid Spencer404a3252007-02-15 03:07:05 +00001620<div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00001621<div class="doc_text">
Chris Lattner37b6b092005-04-25 17:34:15 +00001622
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001623<h5>Overview:</h5>
Chris Lattner37b6b092005-04-25 17:34:15 +00001624
Reid Spencer404a3252007-02-15 03:07:05 +00001625<p>A vector type is a simple derived type that represents a vector
1626of elements. Vector types are used when multiple primitive data
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001627are operated in parallel using a single instruction (SIMD).
Reid Spencer404a3252007-02-15 03:07:05 +00001628A vector type requires a size (number of
Chris Lattner330ce692005-11-10 01:44:22 +00001629elements) and an underlying primitive data type. Vectors must have a power
Reid Spencer404a3252007-02-15 03:07:05 +00001630of two length (1, 2, 4, 8, 16 ...). Vector types are
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001631considered <a href="#t_firstclass">first class</a>.</p>
Chris Lattner37b6b092005-04-25 17:34:15 +00001632
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001633<h5>Syntax:</h5>
Chris Lattner37b6b092005-04-25 17:34:15 +00001634
1635<pre>
1636 &lt; &lt;# elements&gt; x &lt;elementtype&gt; &gt;
1637</pre>
1638
John Criswell4a3327e2005-05-13 22:25:59 +00001639<p>The number of elements is a constant integer value; elementtype may
Chris Lattnerc0f423a2007-01-15 01:54:13 +00001640be any integer or floating point type.</p>
Chris Lattner37b6b092005-04-25 17:34:15 +00001641
Chris Lattnerc8cb6952004-08-12 19:12:28 +00001642<h5>Examples:</h5>
Chris Lattner37b6b092005-04-25 17:34:15 +00001643
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001644<table class="layout">
1645 <tr class="layout">
Chris Lattner747359f2007-12-19 05:04:11 +00001646 <td class="left"><tt>&lt;4 x i32&gt;</tt></td>
1647 <td class="left">Vector of 4 32-bit integer values.</td>
1648 </tr>
1649 <tr class="layout">
1650 <td class="left"><tt>&lt;8 x float&gt;</tt></td>
1651 <td class="left">Vector of 8 32-bit floating-point values.</td>
1652 </tr>
1653 <tr class="layout">
1654 <td class="left"><tt>&lt;2 x i64&gt;</tt></td>
1655 <td class="left">Vector of 2 64-bit integer values.</td>
Reid Spencerc3c4c4f2004-11-01 08:19:36 +00001656 </tr>
1657</table>
Dan Gohman142ccc02009-01-24 15:58:40 +00001658
1659<p>Note that the code generator does not yet support large vector types
1660to be used as function return types. The specific limit on how large a
1661vector return type codegen can currently handle is target-dependent;
1662currently it's often a few times longer than a hardware vector register.</p>
1663
Misha Brukman76307852003-11-08 01:05:38 +00001664</div>
1665
Chris Lattner37b6b092005-04-25 17:34:15 +00001666<!-- _______________________________________________________________________ -->
1667<div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1668<div class="doc_text">
1669
1670<h5>Overview:</h5>
1671
1672<p>Opaque types are used to represent unknown types in the system. This
Gordon Henriksena699c4d2007-10-14 00:34:53 +00001673corresponds (for example) to the C notion of a forward declared structure type.
Chris Lattner37b6b092005-04-25 17:34:15 +00001674In LLVM, opaque types can eventually be resolved to any type (not just a
1675structure type).</p>
1676
1677<h5>Syntax:</h5>
1678
1679<pre>
1680 opaque
1681</pre>
1682
1683<h5>Examples:</h5>
1684
1685<table class="layout">
1686 <tr class="layout">
Chris Lattner747359f2007-12-19 05:04:11 +00001687 <td class="left"><tt>opaque</tt></td>
1688 <td class="left">An opaque type.</td>
Chris Lattner37b6b092005-04-25 17:34:15 +00001689 </tr>
1690</table>
1691</div>
1692
Chris Lattnercf7a5842009-02-02 07:32:36 +00001693<!-- ======================================================================= -->
1694<div class="doc_subsection">
1695 <a name="t_uprefs">Type Up-references</a>
1696</div>
1697
1698<div class="doc_text">
1699<h5>Overview:</h5>
1700<p>
1701An "up reference" allows you to refer to a lexically enclosing type without
1702requiring it to have a name. For instance, a structure declaration may contain a
1703pointer to any of the types it is lexically a member of. Example of up
1704references (with their equivalent as named type declarations) include:</p>
1705
1706<pre>
Chris Lattnerbf1d5452009-02-09 10:00:56 +00001707 { \2 * } %x = type { %x* }
Chris Lattnercf7a5842009-02-02 07:32:36 +00001708 { \2 }* %y = type { %y }*
1709 \1* %z = type %z*
1710</pre>
1711
1712<p>
1713An up reference is needed by the asmprinter for printing out cyclic types when
1714there is no declared name for a type in the cycle. Because the asmprinter does
1715not want to print out an infinite type string, it needs a syntax to handle
1716recursive types that have no names (all names are optional in llvm IR).
1717</p>
1718
1719<h5>Syntax:</h5>
1720<pre>
1721 \&lt;level&gt;
1722</pre>
1723
1724<p>
1725The level is the count of the lexical type that is being referred to.
1726</p>
1727
1728<h5>Examples:</h5>
1729
1730<table class="layout">
1731 <tr class="layout">
1732 <td class="left"><tt>\1*</tt></td>
1733 <td class="left">Self-referential pointer.</td>
1734 </tr>
1735 <tr class="layout">
1736 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
1737 <td class="left">Recursive structure where the upref refers to the out-most
1738 structure.</td>
1739 </tr>
1740</table>
1741</div>
1742
Chris Lattner37b6b092005-04-25 17:34:15 +00001743
Chris Lattner74d3f822004-12-09 17:30:23 +00001744<!-- *********************************************************************** -->
1745<div class="doc_section"> <a name="constants">Constants</a> </div>
1746<!-- *********************************************************************** -->
1747
1748<div class="doc_text">
1749
1750<p>LLVM has several different basic types of constants. This section describes
1751them all and their syntax.</p>
1752
1753</div>
1754
1755<!-- ======================================================================= -->
Reid Spencer8f08d802004-12-09 18:02:53 +00001756<div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001757
1758<div class="doc_text">
1759
1760<dl>
1761 <dt><b>Boolean constants</b></dt>
1762
1763 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
Reid Spencer36a15422007-01-12 03:35:51 +00001764 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
Chris Lattner74d3f822004-12-09 17:30:23 +00001765 </dd>
1766
1767 <dt><b>Integer constants</b></dt>
1768
Reid Spencer8f08d802004-12-09 18:02:53 +00001769 <dd>Standard integers (such as '4') are constants of the <a
Reid Spencer3e628eb92007-01-04 16:43:23 +00001770 href="#t_integer">integer</a> type. Negative numbers may be used with
Chris Lattner74d3f822004-12-09 17:30:23 +00001771 integer types.
1772 </dd>
1773
1774 <dt><b>Floating point constants</b></dt>
1775
1776 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1777 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
Chris Lattner1429e6f2008-04-01 18:45:27 +00001778 notation (see below). The assembler requires the exact decimal value of
1779 a floating-point constant. For example, the assembler accepts 1.25 but
1780 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1781 constants must have a <a href="#t_floating">floating point</a> type. </dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00001782
1783 <dt><b>Null pointer constants</b></dt>
1784
John Criswelldfe6a862004-12-10 15:51:16 +00001785 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
Chris Lattner74d3f822004-12-09 17:30:23 +00001786 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1787
1788</dl>
1789
Dale Johannesencd4a3012009-02-11 22:14:51 +00001790<p>The one non-intuitive notation for constants is the hexadecimal form
Chris Lattner74d3f822004-12-09 17:30:23 +00001791of floating point constants. For example, the form '<tt>double
17920x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
17934.5e+15</tt>'. The only time hexadecimal floating point constants are required
Reid Spencer8f08d802004-12-09 18:02:53 +00001794(and the only time that they are generated by the disassembler) is when a
1795floating point constant must be emitted but it cannot be represented as a
Dale Johannesencd4a3012009-02-11 22:14:51 +00001796decimal floating point number in a reasonable number of digits. For example,
1797NaN's, infinities, and other
Reid Spencer8f08d802004-12-09 18:02:53 +00001798special values are represented in their IEEE hexadecimal format so that
1799assembly and disassembly do not cause any bits to change in the constants.</p>
Dale Johannesencd4a3012009-02-11 22:14:51 +00001800<p>When using the hexadecimal form, constants of types float and double are
1801represented using the 16-digit form shown above (which matches the IEEE754
1802representation for double); float values must, however, be exactly representable
1803as IEE754 single precision.
1804Hexadecimal format is always used for long
1805double, and there are three forms of long double. The 80-bit
1806format used by x86 is represented as <tt>0xK</tt>
1807followed by 20 hexadecimal digits.
1808The 128-bit format used by PowerPC (two adjacent doubles) is represented
1809by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit
1810format is represented
1811by <tt>0xL</tt> followed by 32 hexadecimal digits; no currently supported
1812target uses this format. Long doubles will only work if they match
1813the long double format on your target. All hexadecimal formats are big-endian
1814(sign bit at the left).</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001815</div>
1816
1817<!-- ======================================================================= -->
Chris Lattner361bfcd2009-02-28 18:32:25 +00001818<div class="doc_subsection">
1819<a name="aggregateconstants"> <!-- old anchor -->
1820<a name="complexconstants">Complex Constants</a></a>
Chris Lattner74d3f822004-12-09 17:30:23 +00001821</div>
1822
1823<div class="doc_text">
Chris Lattner361bfcd2009-02-28 18:32:25 +00001824<p>Complex constants are a (potentially recursive) combination of simple
1825constants and smaller complex constants.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001826
1827<dl>
1828 <dt><b>Structure constants</b></dt>
1829
1830 <dd>Structure constants are represented with notation similar to structure
1831 type definitions (a comma separated list of elements, surrounded by braces
Chris Lattnerbea11172007-12-25 20:34:52 +00001832 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1833 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
Chris Lattner455fc8c2005-03-07 22:13:59 +00001834 must have <a href="#t_struct">structure type</a>, and the number and
Chris Lattner74d3f822004-12-09 17:30:23 +00001835 types of elements must match those specified by the type.
1836 </dd>
1837
1838 <dt><b>Array constants</b></dt>
1839
1840 <dd>Array constants are represented with notation similar to array type
1841 definitions (a comma separated list of elements, surrounded by square brackets
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001842 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
Chris Lattner74d3f822004-12-09 17:30:23 +00001843 constants must have <a href="#t_array">array type</a>, and the number and
1844 types of elements must match those specified by the type.
1845 </dd>
1846
Reid Spencer404a3252007-02-15 03:07:05 +00001847 <dt><b>Vector constants</b></dt>
Chris Lattner74d3f822004-12-09 17:30:23 +00001848
Reid Spencer404a3252007-02-15 03:07:05 +00001849 <dd>Vector constants are represented with notation similar to vector type
Chris Lattner74d3f822004-12-09 17:30:23 +00001850 definitions (a comma separated list of elements, surrounded by
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00001851 less-than/greater-than's (<tt>&lt;&gt;</tt>)). For example: "<tt>&lt; i32 42,
Jeff Cohen5819f182007-04-22 01:17:39 +00001852 i32 11, i32 74, i32 100 &gt;</tt>". Vector constants must have <a
Reid Spencer404a3252007-02-15 03:07:05 +00001853 href="#t_vector">vector type</a>, and the number and types of elements must
Chris Lattner74d3f822004-12-09 17:30:23 +00001854 match those specified by the type.
1855 </dd>
1856
1857 <dt><b>Zero initialization</b></dt>
1858
1859 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1860 value to zero of <em>any</em> type, including scalar and aggregate types.
1861 This is often used to avoid having to print large zero initializers (e.g. for
John Criswell4c0cf7f2005-10-24 16:17:18 +00001862 large arrays) and is always exactly equivalent to using explicit zero
Chris Lattner74d3f822004-12-09 17:30:23 +00001863 initializers.
1864 </dd>
Nick Lewycky49f89192009-04-04 07:22:01 +00001865
1866 <dt><b>Metadata node</b></dt>
1867
1868 <dd>A metadata node is a structure-like constant with the type of an empty
1869 struct. For example: "<tt>{ } !{ i32 0, { } !"test" }</tt>". Unlike other
1870 constants that are meant to be interpreted as part of the instruction stream,
1871 metadata is a place to attach additional information such as debug info.
1872 </dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00001873</dl>
1874
1875</div>
1876
1877<!-- ======================================================================= -->
1878<div class="doc_subsection">
1879 <a name="globalconstants">Global Variable and Function Addresses</a>
1880</div>
1881
1882<div class="doc_text">
1883
1884<p>The addresses of <a href="#globalvars">global variables</a> and <a
1885href="#functionstructure">functions</a> are always implicitly valid (link-time)
John Criswelldfe6a862004-12-10 15:51:16 +00001886constants. These constants are explicitly referenced when the <a
1887href="#identifiers">identifier for the global</a> is used and always have <a
Chris Lattner74d3f822004-12-09 17:30:23 +00001888href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1889file:</p>
1890
Bill Wendling3716c5d2007-05-29 09:04:49 +00001891<div class="doc_code">
Chris Lattner74d3f822004-12-09 17:30:23 +00001892<pre>
Chris Lattner00538a12007-06-06 18:28:13 +00001893@X = global i32 17
1894@Y = global i32 42
1895@Z = global [2 x i32*] [ i32* @X, i32* @Y ]
Chris Lattner74d3f822004-12-09 17:30:23 +00001896</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00001897</div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001898
1899</div>
1900
1901<!-- ======================================================================= -->
Reid Spencer641f5c92004-12-09 18:13:12 +00001902<div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
Chris Lattner74d3f822004-12-09 17:30:23 +00001903<div class="doc_text">
Reid Spencer641f5c92004-12-09 18:13:12 +00001904 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
John Criswell4a3327e2005-05-13 22:25:59 +00001905 no specific value. Undefined values may be of any type and be used anywhere
Reid Spencer641f5c92004-12-09 18:13:12 +00001906 a constant is permitted.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001907
Reid Spencer641f5c92004-12-09 18:13:12 +00001908 <p>Undefined values indicate to the compiler that the program is well defined
1909 no matter what value is used, giving the compiler more freedom to optimize.
1910 </p>
Chris Lattner74d3f822004-12-09 17:30:23 +00001911</div>
1912
1913<!-- ======================================================================= -->
1914<div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1915</div>
1916
1917<div class="doc_text">
1918
1919<p>Constant expressions are used to allow expressions involving other constants
1920to be used as constants. Constant expressions may be of any <a
John Criswell4a3327e2005-05-13 22:25:59 +00001921href="#t_firstclass">first class</a> type and may involve any LLVM operation
Chris Lattner74d3f822004-12-09 17:30:23 +00001922that does not have side effects (e.g. load and call are not supported). The
1923following is the syntax for constant expressions:</p>
1924
1925<dl>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001926 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1927 <dd>Truncate a constant to another type. The bit size of CST must be larger
Chris Lattnerc0f423a2007-01-15 01:54:13 +00001928 than the bit size of TYPE. Both types must be integers.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00001929
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001930 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1931 <dd>Zero extend a constant to another type. The bit size of CST must be
Chris Lattnerc0f423a2007-01-15 01:54:13 +00001932 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001933
1934 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1935 <dd>Sign extend a constant to another type. The bit size of CST must be
Chris Lattnerc0f423a2007-01-15 01:54:13 +00001936 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001937
1938 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1939 <dd>Truncate a floating point constant to another floating point type. The
1940 size of CST must be larger than the size of TYPE. Both types must be
1941 floating point.</dd>
1942
1943 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1944 <dd>Floating point extend a constant to another type. The size of CST must be
1945 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1946
Reid Spencer753163d2007-07-31 14:40:14 +00001947 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001948 <dd>Convert a floating point constant to the corresponding unsigned integer
Nate Begemand4d45c22007-11-17 03:58:34 +00001949 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1950 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1951 of the same number of elements. If the value won't fit in the integer type,
1952 the results are undefined.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001953
Reid Spencer51b07252006-11-09 23:03:26 +00001954 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001955 <dd>Convert a floating point constant to the corresponding signed integer
Nate Begemand4d45c22007-11-17 03:58:34 +00001956 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1957 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1958 of the same number of elements. If the value won't fit in the integer type,
1959 the results are undefined.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001960
Reid Spencer51b07252006-11-09 23:03:26 +00001961 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001962 <dd>Convert an unsigned integer constant to the corresponding floating point
Nate Begemand4d45c22007-11-17 03:58:34 +00001963 constant. TYPE must be a scalar or vector floating point type. CST must be of
1964 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1965 of the same number of elements. If the value won't fit in the floating point
1966 type, the results are undefined.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001967
Reid Spencer51b07252006-11-09 23:03:26 +00001968 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001969 <dd>Convert a signed integer constant to the corresponding floating point
Nate Begemand4d45c22007-11-17 03:58:34 +00001970 constant. TYPE must be a scalar or vector floating point type. CST must be of
1971 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1972 of the same number of elements. If the value won't fit in the floating point
1973 type, the results are undefined.</dd>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00001974
Reid Spencer5b950642006-11-11 23:08:07 +00001975 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1976 <dd>Convert a pointer typed constant to the corresponding integer constant
1977 TYPE must be an integer type. CST must be of pointer type. The CST value is
1978 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1979
1980 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1981 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1982 pointer type. CST must be of integer type. The CST value is zero extended,
1983 truncated, or unchanged to make it fit in a pointer size. This one is
1984 <i>really</i> dangerous!</dd>
1985
1986 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
Chris Lattner789dee32009-02-28 18:27:03 +00001987 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
1988 are the same as those for the <a href="#i_bitcast">bitcast
1989 instruction</a>.</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00001990
1991 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1992
1993 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1994 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1995 instruction, the index list may have zero or more indexes, which are required
1996 to make sense for the type of "CSTPTR".</dd>
1997
Robert Bocchino7e97a6d2006-01-10 19:31:34 +00001998 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1999
2000 <dd>Perform the <a href="#i_select">select operation</a> on
Reid Spencer9965ee72006-12-04 19:23:19 +00002001 constants.</dd>
2002
2003 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
2004 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2005
2006 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
2007 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
Robert Bocchino7e97a6d2006-01-10 19:31:34 +00002008
Nate Begemand2195702008-05-12 19:01:56 +00002009 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
2010 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
2011
2012 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
2013 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
2014
Robert Bocchino7e97a6d2006-01-10 19:31:34 +00002015 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
2016
2017 <dd>Perform the <a href="#i_extractelement">extractelement
Dan Gohmanef9462f2008-10-14 16:51:45 +00002018 operation</a> on constants.</dd>
Robert Bocchino7e97a6d2006-01-10 19:31:34 +00002019
Robert Bocchinof72fdfe2006-01-15 20:48:27 +00002020 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
2021
2022 <dd>Perform the <a href="#i_insertelement">insertelement
Reid Spencer9965ee72006-12-04 19:23:19 +00002023 operation</a> on constants.</dd>
Robert Bocchinof72fdfe2006-01-15 20:48:27 +00002024
Chris Lattner016a0e52006-04-08 00:13:41 +00002025
2026 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
2027
2028 <dd>Perform the <a href="#i_shufflevector">shufflevector
Reid Spencer9965ee72006-12-04 19:23:19 +00002029 operation</a> on constants.</dd>
Chris Lattner016a0e52006-04-08 00:13:41 +00002030
Chris Lattner74d3f822004-12-09 17:30:23 +00002031 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
2032
Reid Spencer641f5c92004-12-09 18:13:12 +00002033 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2034 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
Chris Lattner74d3f822004-12-09 17:30:23 +00002035 binary</a> operations. The constraints on operands are the same as those for
2036 the corresponding instruction (e.g. no bitwise operations on floating point
John Criswell02fdc6f2005-05-12 16:52:32 +00002037 values are allowed).</dd>
Chris Lattner74d3f822004-12-09 17:30:23 +00002038</dl>
Chris Lattner74d3f822004-12-09 17:30:23 +00002039</div>
Chris Lattnerb1652612004-03-08 16:49:10 +00002040
Nick Lewycky49f89192009-04-04 07:22:01 +00002041<!-- ======================================================================= -->
2042<div class="doc_subsection"><a name="metadata">Embedded Metadata</a>
2043</div>
2044
2045<div class="doc_text">
2046
2047<p>Embedded metadata provides a way to attach arbitrary data to the
2048instruction stream without affecting the behaviour of the program. There are
2049two metadata primitives, strings and nodes. All metadata has the type of an
2050empty struct and is identified in syntax by a preceding exclamation point
2051('<tt>!</tt>').
2052</p>
2053
2054<p>A metadata string is a string surrounded by double quotes. It can contain
2055any character by escaping non-printable characters with "\xx" where "xx" is
2056the two digit hex code. For example: "<tt>!"test\00"</tt>".
2057</p>
2058
2059<p>Metadata nodes are represented with notation similar to structure constants
2060(a comma separated list of elements, surrounded by braces and preceeded by an
2061exclamation point). For example: "<tt>!{ { } !"test\00", i32 10}</tt>".
2062</p>
2063
Nick Lewyckyb8f9b7a2009-05-10 20:57:05 +00002064<p>A metadata node will attempt to track changes to the values it holds. In
2065the event that a value is deleted, it will be replaced with a typeless
2066"<tt>null</tt>", such as "<tt>{ } !{null, i32 0}</tt>".</p>
2067
Nick Lewycky49f89192009-04-04 07:22:01 +00002068<p>Optimizations may rely on metadata to provide additional information about
2069the program that isn't available in the instructions, or that isn't easily
2070computable. Similarly, the code generator may expect a certain metadata format
2071to be used to express debugging information.</p>
2072</div>
2073
Chris Lattner2f7c9632001-06-06 20:29:01 +00002074<!-- *********************************************************************** -->
Chris Lattner98f013c2006-01-25 23:47:57 +00002075<div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2076<!-- *********************************************************************** -->
2077
2078<!-- ======================================================================= -->
2079<div class="doc_subsection">
2080<a name="inlineasm">Inline Assembler Expressions</a>
2081</div>
2082
2083<div class="doc_text">
2084
2085<p>
2086LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
2087Module-Level Inline Assembly</a>) through the use of a special value. This
2088value represents the inline assembler as a string (containing the instructions
2089to emit), a list of operand constraints (stored as a string), and a flag that
2090indicates whether or not the inline asm expression has side effects. An example
2091inline assembler expression is:
2092</p>
2093
Bill Wendling3716c5d2007-05-29 09:04:49 +00002094<div class="doc_code">
Chris Lattner98f013c2006-01-25 23:47:57 +00002095<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00002096i32 (i32) asm "bswap $0", "=r,r"
Chris Lattner98f013c2006-01-25 23:47:57 +00002097</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00002098</div>
Chris Lattner98f013c2006-01-25 23:47:57 +00002099
2100<p>
2101Inline assembler expressions may <b>only</b> be used as the callee operand of
2102a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
2103</p>
2104
Bill Wendling3716c5d2007-05-29 09:04:49 +00002105<div class="doc_code">
Chris Lattner98f013c2006-01-25 23:47:57 +00002106<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00002107%X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
Chris Lattner98f013c2006-01-25 23:47:57 +00002108</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00002109</div>
Chris Lattner98f013c2006-01-25 23:47:57 +00002110
2111<p>
2112Inline asms with side effects not visible in the constraint list must be marked
2113as having side effects. This is done through the use of the
2114'<tt>sideeffect</tt>' keyword, like so:
2115</p>
2116
Bill Wendling3716c5d2007-05-29 09:04:49 +00002117<div class="doc_code">
Chris Lattner98f013c2006-01-25 23:47:57 +00002118<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00002119call void asm sideeffect "eieio", ""()
Chris Lattner98f013c2006-01-25 23:47:57 +00002120</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00002121</div>
Chris Lattner98f013c2006-01-25 23:47:57 +00002122
2123<p>TODO: The format of the asm and constraints string still need to be
2124documented here. Constraints on what can be done (e.g. duplication, moving, etc
Chris Lattnerd5528262008-10-04 18:36:02 +00002125need to be documented). This is probably best done by reference to another
2126document that covers inline asm from a holistic perspective.
Chris Lattner98f013c2006-01-25 23:47:57 +00002127</p>
2128
2129</div>
2130
2131<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +00002132<div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2133<!-- *********************************************************************** -->
Chris Lattner74d3f822004-12-09 17:30:23 +00002134
Misha Brukman76307852003-11-08 01:05:38 +00002135<div class="doc_text">
Chris Lattner74d3f822004-12-09 17:30:23 +00002136
Chris Lattner48b383b02003-11-25 01:02:51 +00002137<p>The LLVM instruction set consists of several different
2138classifications of instructions: <a href="#terminators">terminator
John Criswell4a3327e2005-05-13 22:25:59 +00002139instructions</a>, <a href="#binaryops">binary instructions</a>,
2140<a href="#bitwiseops">bitwise binary instructions</a>, <a
Chris Lattner48b383b02003-11-25 01:02:51 +00002141 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
2142instructions</a>.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00002143
Misha Brukman76307852003-11-08 01:05:38 +00002144</div>
Chris Lattner74d3f822004-12-09 17:30:23 +00002145
Chris Lattner2f7c9632001-06-06 20:29:01 +00002146<!-- ======================================================================= -->
Chris Lattner48b383b02003-11-25 01:02:51 +00002147<div class="doc_subsection"> <a name="terminators">Terminator
2148Instructions</a> </div>
Chris Lattner74d3f822004-12-09 17:30:23 +00002149
Misha Brukman76307852003-11-08 01:05:38 +00002150<div class="doc_text">
Chris Lattner74d3f822004-12-09 17:30:23 +00002151
Chris Lattner48b383b02003-11-25 01:02:51 +00002152<p>As mentioned <a href="#functionstructure">previously</a>, every
2153basic block in a program ends with a "Terminator" instruction, which
2154indicates which block should be executed after the current block is
2155finished. These terminator instructions typically yield a '<tt>void</tt>'
2156value: they produce control flow, not values (the one exception being
2157the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
John Criswelldfe6a862004-12-10 15:51:16 +00002158<p>There are six different terminator instructions: the '<a
Chris Lattner48b383b02003-11-25 01:02:51 +00002159 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
2160instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
Chris Lattner08b7d5b2004-10-16 18:04:13 +00002161the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
2162 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
2163 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
Chris Lattner74d3f822004-12-09 17:30:23 +00002164
Misha Brukman76307852003-11-08 01:05:38 +00002165</div>
Chris Lattner74d3f822004-12-09 17:30:23 +00002166
Chris Lattner2f7c9632001-06-06 20:29:01 +00002167<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00002168<div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2169Instruction</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00002170<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00002171<h5>Syntax:</h5>
Dan Gohmancc3132e2008-10-04 19:00:07 +00002172<pre>
2173 ret &lt;type&gt; &lt;value&gt; <i>; Return a value from a non-void function</i>
Chris Lattner590645f2002-04-14 06:13:44 +00002174 ret void <i>; Return from void function</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002175</pre>
Chris Lattnerda508ac2008-04-23 04:59:35 +00002176
Chris Lattner2f7c9632001-06-06 20:29:01 +00002177<h5>Overview:</h5>
Chris Lattnerda508ac2008-04-23 04:59:35 +00002178
Dan Gohmancc3132e2008-10-04 19:00:07 +00002179<p>The '<tt>ret</tt>' instruction is used to return control flow (and
2180optionally a value) from a function back to the caller.</p>
John Criswell417228d2004-04-09 16:48:45 +00002181<p>There are two forms of the '<tt>ret</tt>' instruction: one that
Dan Gohmancc3132e2008-10-04 19:00:07 +00002182returns a value and then causes control flow, and one that just causes
Chris Lattner48b383b02003-11-25 01:02:51 +00002183control flow to occur.</p>
Chris Lattnerda508ac2008-04-23 04:59:35 +00002184
Chris Lattner2f7c9632001-06-06 20:29:01 +00002185<h5>Arguments:</h5>
Chris Lattnerda508ac2008-04-23 04:59:35 +00002186
Dan Gohmancc3132e2008-10-04 19:00:07 +00002187<p>The '<tt>ret</tt>' instruction optionally accepts a single argument,
2188the return value. The type of the return value must be a
2189'<a href="#t_firstclass">first class</a>' type.</p>
2190
2191<p>A function is not <a href="#wellformed">well formed</a> if
2192it it has a non-void return type and contains a '<tt>ret</tt>'
2193instruction with no return value or a return value with a type that
2194does not match its type, or if it has a void return type and contains
2195a '<tt>ret</tt>' instruction with a return value.</p>
Chris Lattnerda508ac2008-04-23 04:59:35 +00002196
Chris Lattner2f7c9632001-06-06 20:29:01 +00002197<h5>Semantics:</h5>
Chris Lattnerda508ac2008-04-23 04:59:35 +00002198
Chris Lattner48b383b02003-11-25 01:02:51 +00002199<p>When the '<tt>ret</tt>' instruction is executed, control flow
2200returns back to the calling function's context. If the caller is a "<a
John Criswell40db33f2004-06-25 15:16:57 +00002201 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
Chris Lattner48b383b02003-11-25 01:02:51 +00002202the instruction after the call. If the caller was an "<a
2203 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
John Criswell02fdc6f2005-05-12 16:52:32 +00002204at the beginning of the "normal" destination block. If the instruction
Chris Lattner48b383b02003-11-25 01:02:51 +00002205returns a value, that value shall set the call or invoke instruction's
Dan Gohmanef9462f2008-10-14 16:51:45 +00002206return value.</p>
Chris Lattnerda508ac2008-04-23 04:59:35 +00002207
Chris Lattner2f7c9632001-06-06 20:29:01 +00002208<h5>Example:</h5>
Chris Lattnerda508ac2008-04-23 04:59:35 +00002209
2210<pre>
2211 ret i32 5 <i>; Return an integer value of 5</i>
Chris Lattner590645f2002-04-14 06:13:44 +00002212 ret void <i>; Return from a void function</i>
Bill Wendling050ee8f2009-02-28 22:12:54 +00002213 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002214</pre>
Dan Gohman3065b612009-01-12 23:12:39 +00002215
Dan Gohman142ccc02009-01-24 15:58:40 +00002216<p>Note that the code generator does not yet fully support large
2217 return values. The specific sizes that are currently supported are
2218 dependent on the target. For integers, on 32-bit targets the limit
2219 is often 64 bits, and on 64-bit targets the limit is often 128 bits.
2220 For aggregate types, the current limits are dependent on the element
2221 types; for example targets are often limited to 2 total integer
2222 elements and 2 total floating-point elements.</p>
Dan Gohman3065b612009-01-12 23:12:39 +00002223
Misha Brukman76307852003-11-08 01:05:38 +00002224</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002225<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00002226<div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00002227<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00002228<h5>Syntax:</h5>
Reid Spencer36a15422007-01-12 03:35:51 +00002229<pre> br i1 &lt;cond&gt;, label &lt;iftrue&gt;, label &lt;iffalse&gt;<br> br label &lt;dest&gt; <i>; Unconditional branch</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002230</pre>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002231<h5>Overview:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00002232<p>The '<tt>br</tt>' instruction is used to cause control flow to
2233transfer to a different basic block in the current function. There are
2234two forms of this instruction, corresponding to a conditional branch
2235and an unconditional branch.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002236<h5>Arguments:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00002237<p>The conditional branch form of the '<tt>br</tt>' instruction takes a
Reid Spencer36a15422007-01-12 03:35:51 +00002238single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
Reid Spencer50c723a2007-02-19 23:54:10 +00002239unconditional form of the '<tt>br</tt>' instruction takes a single
2240'<tt>label</tt>' value as a target.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002241<h5>Semantics:</h5>
Reid Spencer36a15422007-01-12 03:35:51 +00002242<p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
Chris Lattner48b383b02003-11-25 01:02:51 +00002243argument is evaluated. If the value is <tt>true</tt>, control flows
2244to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2245control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002246<h5>Example:</h5>
Chris Lattnere648c282009-05-09 18:11:50 +00002247<pre>Test:<br> %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b<br> br i1 %cond, label %IfEqual, label %IfUnequal<br>IfEqual:<br> <a
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002248 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
Misha Brukman76307852003-11-08 01:05:38 +00002249</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002250<!-- _______________________________________________________________________ -->
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002251<div class="doc_subsubsection">
2252 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2253</div>
2254
Misha Brukman76307852003-11-08 01:05:38 +00002255<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00002256<h5>Syntax:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002257
2258<pre>
2259 switch &lt;intty&gt; &lt;value&gt;, label &lt;defaultdest&gt; [ &lt;intty&gt; &lt;val&gt;, label &lt;dest&gt; ... ]
2260</pre>
2261
Chris Lattner2f7c9632001-06-06 20:29:01 +00002262<h5>Overview:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002263
2264<p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2265several different places. It is a generalization of the '<tt>br</tt>'
Misha Brukman76307852003-11-08 01:05:38 +00002266instruction, allowing a branch to occur to one of many possible
2267destinations.</p>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002268
2269
Chris Lattner2f7c9632001-06-06 20:29:01 +00002270<h5>Arguments:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002271
2272<p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2273comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
2274an array of pairs of comparison value constants and '<tt>label</tt>'s. The
2275table is not allowed to contain duplicate constant entries.</p>
2276
Chris Lattner2f7c9632001-06-06 20:29:01 +00002277<h5>Semantics:</h5>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002278
Chris Lattner48b383b02003-11-25 01:02:51 +00002279<p>The <tt>switch</tt> instruction specifies a table of values and
2280destinations. When the '<tt>switch</tt>' instruction is executed, this
John Criswellbcbb18c2004-06-25 16:05:06 +00002281table is searched for the given value. If the value is found, control flow is
2282transfered to the corresponding destination; otherwise, control flow is
2283transfered to the default destination.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002284
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002285<h5>Implementation:</h5>
2286
2287<p>Depending on properties of the target machine and the particular
2288<tt>switch</tt> instruction, this instruction may be code generated in different
John Criswellbcbb18c2004-06-25 16:05:06 +00002289ways. For example, it could be generated as a series of chained conditional
2290branches or with a lookup table.</p>
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002291
2292<h5>Example:</h5>
2293
2294<pre>
2295 <i>; Emulate a conditional br instruction</i>
Reid Spencer36a15422007-01-12 03:35:51 +00002296 %Val = <a href="#i_zext">zext</a> i1 %value to i32
Dan Gohman623806e2009-01-04 23:44:43 +00002297 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002298
2299 <i>; Emulate an unconditional br instruction</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002300 switch i32 0, label %dest [ ]
Chris Lattnercf96c6c2004-02-24 04:54:45 +00002301
2302 <i>; Implement a jump table:</i>
Dan Gohman623806e2009-01-04 23:44:43 +00002303 switch i32 %val, label %otherwise [ i32 0, label %onzero
2304 i32 1, label %onone
2305 i32 2, label %ontwo ]
Chris Lattner2f7c9632001-06-06 20:29:01 +00002306</pre>
Misha Brukman76307852003-11-08 01:05:38 +00002307</div>
Chris Lattner0132aff2005-05-06 22:57:40 +00002308
Chris Lattner2f7c9632001-06-06 20:29:01 +00002309<!-- _______________________________________________________________________ -->
Chris Lattner0132aff2005-05-06 22:57:40 +00002310<div class="doc_subsubsection">
2311 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2312</div>
2313
Misha Brukman76307852003-11-08 01:05:38 +00002314<div class="doc_text">
Chris Lattner0132aff2005-05-06 22:57:40 +00002315
Chris Lattner2f7c9632001-06-06 20:29:01 +00002316<h5>Syntax:</h5>
Chris Lattner0132aff2005-05-06 22:57:40 +00002317
2318<pre>
Devang Patel02256232008-10-07 17:48:33 +00002319 &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 +00002320 to label &lt;normal label&gt; unwind label &lt;exception label&gt;
Chris Lattner0132aff2005-05-06 22:57:40 +00002321</pre>
2322
Chris Lattnera8292f32002-05-06 22:08:29 +00002323<h5>Overview:</h5>
Chris Lattner0132aff2005-05-06 22:57:40 +00002324
2325<p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2326function, with the possibility of control flow transfer to either the
John Criswell02fdc6f2005-05-12 16:52:32 +00002327'<tt>normal</tt>' label or the
2328'<tt>exception</tt>' label. If the callee function returns with the
Chris Lattner0132aff2005-05-06 22:57:40 +00002329"<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2330"normal" label. If the callee (or any indirect callees) returns with the "<a
John Criswell02fdc6f2005-05-12 16:52:32 +00002331href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
Dan Gohmanef9462f2008-10-14 16:51:45 +00002332continued at the dynamically nearest "exception" label.</p>
Chris Lattner0132aff2005-05-06 22:57:40 +00002333
Chris Lattner2f7c9632001-06-06 20:29:01 +00002334<h5>Arguments:</h5>
Chris Lattner0132aff2005-05-06 22:57:40 +00002335
Misha Brukman76307852003-11-08 01:05:38 +00002336<p>This instruction requires several arguments:</p>
Chris Lattner0132aff2005-05-06 22:57:40 +00002337
Chris Lattner2f7c9632001-06-06 20:29:01 +00002338<ol>
Chris Lattner0132aff2005-05-06 22:57:40 +00002339 <li>
Duncan Sands16f122e2007-03-30 12:22:09 +00002340 The optional "cconv" marker indicates which <a href="#callingconv">calling
Chris Lattner0132aff2005-05-06 22:57:40 +00002341 convention</a> the call should use. If none is specified, the call defaults
2342 to using C calling conventions.
2343 </li>
Devang Patel7e9b05e2008-10-06 18:50:38 +00002344
2345 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2346 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
2347 and '<tt>inreg</tt>' attributes are valid here.</li>
2348
Chris Lattner0132aff2005-05-06 22:57:40 +00002349 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2350 function value being invoked. In most cases, this is a direct function
2351 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2352 an arbitrary pointer to function value.
2353 </li>
2354
2355 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2356 function to be invoked. </li>
2357
2358 <li>'<tt>function args</tt>': argument list whose types match the function
2359 signature argument types. If the function signature indicates the function
2360 accepts a variable number of arguments, the extra arguments can be
2361 specified. </li>
2362
2363 <li>'<tt>normal label</tt>': the label reached when the called function
2364 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2365
2366 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2367 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2368
Devang Patel02256232008-10-07 17:48:33 +00002369 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
Devang Patel7e9b05e2008-10-06 18:50:38 +00002370 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2371 '<tt>readnone</tt>' attributes are valid here.</li>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002372</ol>
Chris Lattner0132aff2005-05-06 22:57:40 +00002373
Chris Lattner2f7c9632001-06-06 20:29:01 +00002374<h5>Semantics:</h5>
Chris Lattner0132aff2005-05-06 22:57:40 +00002375
Misha Brukman76307852003-11-08 01:05:38 +00002376<p>This instruction is designed to operate as a standard '<tt><a
Chris Lattner0132aff2005-05-06 22:57:40 +00002377href="#i_call">call</a></tt>' instruction in most regards. The primary
2378difference is that it establishes an association with a label, which is used by
2379the runtime library to unwind the stack.</p>
2380
2381<p>This instruction is used in languages with destructors to ensure that proper
2382cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2383exception. Additionally, this is important for implementation of
2384'<tt>catch</tt>' clauses in high-level languages that support them.</p>
2385
Dan Gohman9069d892009-05-22 21:47:08 +00002386<p>It is not valid to reference the return value of an invoke call from
2387anywhere not dominated by the normal label, since an unwind does not
2388provide a return value.</p>
2389
Chris Lattner2f7c9632001-06-06 20:29:01 +00002390<h5>Example:</h5>
Chris Lattner0132aff2005-05-06 22:57:40 +00002391<pre>
Nick Lewycky084ab472008-03-16 07:18:12 +00002392 %retval = invoke i32 @Test(i32 15) to label %Continue
Jeff Cohen5819f182007-04-22 01:17:39 +00002393 unwind label %TestCleanup <i>; {i32}:retval set</i>
Nick Lewycky084ab472008-03-16 07:18:12 +00002394 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
Jeff Cohen5819f182007-04-22 01:17:39 +00002395 unwind label %TestCleanup <i>; {i32}:retval set</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002396</pre>
Misha Brukman76307852003-11-08 01:05:38 +00002397</div>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00002398
2399
Chris Lattner5ed60612003-09-03 00:41:47 +00002400<!-- _______________________________________________________________________ -->
Chris Lattner08b7d5b2004-10-16 18:04:13 +00002401
Chris Lattner48b383b02003-11-25 01:02:51 +00002402<div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2403Instruction</a> </div>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00002404
Misha Brukman76307852003-11-08 01:05:38 +00002405<div class="doc_text">
Chris Lattner08b7d5b2004-10-16 18:04:13 +00002406
Chris Lattner5ed60612003-09-03 00:41:47 +00002407<h5>Syntax:</h5>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00002408<pre>
2409 unwind
2410</pre>
2411
Chris Lattner5ed60612003-09-03 00:41:47 +00002412<h5>Overview:</h5>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00002413
2414<p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2415at the first callee in the dynamic call stack which used an <a
2416href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2417primarily used to implement exception handling.</p>
2418
Chris Lattner5ed60612003-09-03 00:41:47 +00002419<h5>Semantics:</h5>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00002420
Chris Lattnerfe8519c2008-04-19 21:01:16 +00002421<p>The '<tt>unwind</tt>' instruction causes execution of the current function to
Chris Lattner08b7d5b2004-10-16 18:04:13 +00002422immediately halt. The dynamic call stack is then searched for the first <a
2423href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2424execution continues at the "exceptional" destination block specified by the
2425<tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2426dynamic call chain, undefined behavior results.</p>
Misha Brukman76307852003-11-08 01:05:38 +00002427</div>
Chris Lattner08b7d5b2004-10-16 18:04:13 +00002428
2429<!-- _______________________________________________________________________ -->
2430
2431<div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2432Instruction</a> </div>
2433
2434<div class="doc_text">
2435
2436<h5>Syntax:</h5>
2437<pre>
2438 unreachable
2439</pre>
2440
2441<h5>Overview:</h5>
2442
2443<p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2444instruction is used to inform the optimizer that a particular portion of the
2445code is not reachable. This can be used to indicate that the code after a
2446no-return function cannot be reached, and other facts.</p>
2447
2448<h5>Semantics:</h5>
2449
2450<p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2451</div>
2452
2453
2454
Chris Lattner2f7c9632001-06-06 20:29:01 +00002455<!-- ======================================================================= -->
Chris Lattner48b383b02003-11-25 01:02:51 +00002456<div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00002457<div class="doc_text">
Chris Lattner48b383b02003-11-25 01:02:51 +00002458<p>Binary operators are used to do most of the computation in a
Chris Lattner81f92972008-04-01 18:47:32 +00002459program. They require two operands of the same type, execute an operation on them, and
John Criswelldfe6a862004-12-10 15:51:16 +00002460produce a single value. The operands might represent
Reid Spencer404a3252007-02-15 03:07:05 +00002461multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
Chris Lattner81f92972008-04-01 18:47:32 +00002462The result value has the same type as its operands.</p>
Misha Brukman76307852003-11-08 01:05:38 +00002463<p>There are several different binary operators:</p>
Misha Brukman76307852003-11-08 01:05:38 +00002464</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002465<!-- _______________________________________________________________________ -->
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002466<div class="doc_subsubsection">
2467 <a name="i_add">'<tt>add</tt>' Instruction</a>
2468</div>
2469
Misha Brukman76307852003-11-08 01:05:38 +00002470<div class="doc_text">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002471
Chris Lattner2f7c9632001-06-06 20:29:01 +00002472<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002473
2474<pre>
Gabor Greif0f75ad02008-08-07 21:46:00 +00002475 &lt;result&gt; = add &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002476</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002477
Chris Lattner2f7c9632001-06-06 20:29:01 +00002478<h5>Overview:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002479
Misha Brukman76307852003-11-08 01:05:38 +00002480<p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002481
Chris Lattner2f7c9632001-06-06 20:29:01 +00002482<h5>Arguments:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002483
2484<p>The two arguments to the '<tt>add</tt>' instruction must be <a
2485 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>, or
2486 <a href="#t_vector">vector</a> values. Both arguments must have identical
2487 types.</p>
2488
Chris Lattner2f7c9632001-06-06 20:29:01 +00002489<h5>Semantics:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002490
Misha Brukman76307852003-11-08 01:05:38 +00002491<p>The value produced is the integer or floating point sum of the two
2492operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002493
Chris Lattner2f2427e2008-01-28 00:36:27 +00002494<p>If an integer sum has unsigned overflow, the result returned is the
2495mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2496the result.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002497
Chris Lattner2f2427e2008-01-28 00:36:27 +00002498<p>Because LLVM integers use a two's complement representation, this
2499instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002500
Chris Lattner2f7c9632001-06-06 20:29:01 +00002501<h5>Example:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002502
2503<pre>
2504 &lt;result&gt; = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002505</pre>
Misha Brukman76307852003-11-08 01:05:38 +00002506</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002507<!-- _______________________________________________________________________ -->
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002508<div class="doc_subsubsection">
2509 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2510</div>
2511
Misha Brukman76307852003-11-08 01:05:38 +00002512<div class="doc_text">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002513
Chris Lattner2f7c9632001-06-06 20:29:01 +00002514<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002515
2516<pre>
Gabor Greif0f75ad02008-08-07 21:46:00 +00002517 &lt;result&gt; = sub &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002518</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002519
Chris Lattner2f7c9632001-06-06 20:29:01 +00002520<h5>Overview:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002521
Misha Brukman76307852003-11-08 01:05:38 +00002522<p>The '<tt>sub</tt>' instruction returns the difference of its two
2523operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002524
2525<p>Note that the '<tt>sub</tt>' instruction is used to represent the
2526'<tt>neg</tt>' instruction present in most other intermediate
2527representations.</p>
2528
Chris Lattner2f7c9632001-06-06 20:29:01 +00002529<h5>Arguments:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002530
2531<p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2532 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2533 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2534 types.</p>
2535
Chris Lattner2f7c9632001-06-06 20:29:01 +00002536<h5>Semantics:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002537
Chris Lattner48b383b02003-11-25 01:02:51 +00002538<p>The value produced is the integer or floating point difference of
2539the two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002540
Chris Lattner2f2427e2008-01-28 00:36:27 +00002541<p>If an integer difference has unsigned overflow, the result returned is the
2542mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2543the result.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002544
Chris Lattner2f2427e2008-01-28 00:36:27 +00002545<p>Because LLVM integers use a two's complement representation, this
2546instruction is appropriate for both signed and unsigned integers.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002547
Chris Lattner2f7c9632001-06-06 20:29:01 +00002548<h5>Example:</h5>
Bill Wendling2d8b9a82007-05-29 09:42:13 +00002549<pre>
2550 &lt;result&gt; = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002551 &lt;result&gt; = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002552</pre>
Misha Brukman76307852003-11-08 01:05:38 +00002553</div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002554
Chris Lattner2f7c9632001-06-06 20:29:01 +00002555<!-- _______________________________________________________________________ -->
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002556<div class="doc_subsubsection">
2557 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2558</div>
2559
Misha Brukman76307852003-11-08 01:05:38 +00002560<div class="doc_text">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002561
Chris Lattner2f7c9632001-06-06 20:29:01 +00002562<h5>Syntax:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +00002563<pre> &lt;result&gt; = mul &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002564</pre>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002565<h5>Overview:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00002566<p>The '<tt>mul</tt>' instruction returns the product of its two
2567operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002568
Chris Lattner2f7c9632001-06-06 20:29:01 +00002569<h5>Arguments:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002570
2571<p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2572href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2573or <a href="#t_vector">vector</a> values. Both arguments must have identical
2574types.</p>
2575
Chris Lattner2f7c9632001-06-06 20:29:01 +00002576<h5>Semantics:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002577
Chris Lattner48b383b02003-11-25 01:02:51 +00002578<p>The value produced is the integer or floating point product of the
Misha Brukman76307852003-11-08 01:05:38 +00002579two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002580
Chris Lattner2f2427e2008-01-28 00:36:27 +00002581<p>If the result of an integer multiplication has unsigned overflow,
2582the result returned is the mathematical result modulo
25832<sup>n</sup>, where n is the bit width of the result.</p>
2584<p>Because LLVM integers use a two's complement representation, and the
2585result is the same width as the operands, this instruction returns the
2586correct result for both signed and unsigned integers. If a full product
2587(e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2588should be sign-extended or zero-extended as appropriate to the
2589width of the full product.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002590<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002591<pre> &lt;result&gt; = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002592</pre>
Misha Brukman76307852003-11-08 01:05:38 +00002593</div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002594
Chris Lattner2f7c9632001-06-06 20:29:01 +00002595<!-- _______________________________________________________________________ -->
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002596<div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2597</a></div>
2598<div class="doc_text">
2599<h5>Syntax:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +00002600<pre> &lt;result&gt; = udiv &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002601</pre>
2602<h5>Overview:</h5>
2603<p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2604operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002605
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002606<h5>Arguments:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002607
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002608<p>The two arguments to the '<tt>udiv</tt>' instruction must be
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002609<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2610values. Both arguments must have identical types.</p>
2611
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002612<h5>Semantics:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002613
Chris Lattner2f2427e2008-01-28 00:36:27 +00002614<p>The value produced is the unsigned integer quotient of the two operands.</p>
2615<p>Note that unsigned integer division and signed integer division are distinct
2616operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2617<p>Division by zero leads to undefined behavior.</p>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002618<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002619<pre> &lt;result&gt; = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002620</pre>
2621</div>
2622<!-- _______________________________________________________________________ -->
2623<div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2624</a> </div>
2625<div class="doc_text">
2626<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002627<pre>
Gabor Greif0f75ad02008-08-07 21:46:00 +00002628 &lt;result&gt; = sdiv &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002629</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002630
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002631<h5>Overview:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002632
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002633<p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2634operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002635
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002636<h5>Arguments:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002637
2638<p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2639<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2640values. Both arguments must have identical types.</p>
2641
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002642<h5>Semantics:</h5>
Chris Lattner1429e6f2008-04-01 18:45:27 +00002643<p>The value produced is the signed integer quotient of the two operands rounded towards zero.</p>
Chris Lattner2f2427e2008-01-28 00:36:27 +00002644<p>Note that signed integer division and unsigned integer division are distinct
2645operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2646<p>Division by zero leads to undefined behavior. Overflow also leads to
2647undefined behavior; this is a rare case, but can occur, for example,
2648by doing a 32-bit division of -2147483648 by -1.</p>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002649<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002650<pre> &lt;result&gt; = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002651</pre>
2652</div>
2653<!-- _______________________________________________________________________ -->
2654<div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
Chris Lattner48b383b02003-11-25 01:02:51 +00002655Instruction</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00002656<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00002657<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002658<pre>
Gabor Greif0f75ad02008-08-07 21:46:00 +00002659 &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 +00002660</pre>
2661<h5>Overview:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002662
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002663<p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
Chris Lattner48b383b02003-11-25 01:02:51 +00002664operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002665
Chris Lattner48b383b02003-11-25 01:02:51 +00002666<h5>Arguments:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002667
Jeff Cohen5819f182007-04-22 01:17:39 +00002668<p>The two arguments to the '<tt>fdiv</tt>' instruction must be
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002669<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2670of floating point values. Both arguments must have identical types.</p>
2671
Chris Lattner48b383b02003-11-25 01:02:51 +00002672<h5>Semantics:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002673
Reid Spencer7e80b0b2006-10-26 06:15:43 +00002674<p>The value produced is the floating point quotient of the two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002675
Chris Lattner48b383b02003-11-25 01:02:51 +00002676<h5>Example:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002677
2678<pre>
2679 &lt;result&gt; = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
Chris Lattner48b383b02003-11-25 01:02:51 +00002680</pre>
2681</div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002682
Chris Lattner48b383b02003-11-25 01:02:51 +00002683<!-- _______________________________________________________________________ -->
Reid Spencer7eb55b32006-11-02 01:53:59 +00002684<div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2685</div>
2686<div class="doc_text">
2687<h5>Syntax:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +00002688<pre> &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 +00002689</pre>
2690<h5>Overview:</h5>
2691<p>The '<tt>urem</tt>' instruction returns the remainder from the
2692unsigned division of its two arguments.</p>
2693<h5>Arguments:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002694<p>The two arguments to the '<tt>urem</tt>' instruction must be
2695<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2696values. Both arguments must have identical types.</p>
Reid Spencer7eb55b32006-11-02 01:53:59 +00002697<h5>Semantics:</h5>
2698<p>This instruction returns the unsigned integer <i>remainder</i> of a division.
Chris Lattner1429e6f2008-04-01 18:45:27 +00002699This instruction always performs an unsigned division to get the remainder.</p>
Chris Lattner2f2427e2008-01-28 00:36:27 +00002700<p>Note that unsigned integer remainder and signed integer remainder are
2701distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2702<p>Taking the remainder of a division by zero leads to undefined behavior.</p>
Reid Spencer7eb55b32006-11-02 01:53:59 +00002703<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002704<pre> &lt;result&gt; = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
Reid Spencer7eb55b32006-11-02 01:53:59 +00002705</pre>
2706
2707</div>
2708<!-- _______________________________________________________________________ -->
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002709<div class="doc_subsubsection">
2710 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2711</div>
2712
Chris Lattner48b383b02003-11-25 01:02:51 +00002713<div class="doc_text">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002714
Chris Lattner48b383b02003-11-25 01:02:51 +00002715<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002716
2717<pre>
Gabor Greif0f75ad02008-08-07 21:46:00 +00002718 &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 +00002719</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002720
Chris Lattner48b383b02003-11-25 01:02:51 +00002721<h5>Overview:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002722
Reid Spencer7eb55b32006-11-02 01:53:59 +00002723<p>The '<tt>srem</tt>' instruction returns the remainder from the
Dan Gohman08143e32007-11-05 23:35:22 +00002724signed division of its two operands. This instruction can also take
2725<a href="#t_vector">vector</a> versions of the values in which case
2726the elements must be integers.</p>
Chris Lattnerb8f816e2008-01-04 04:33:49 +00002727
Chris Lattner48b383b02003-11-25 01:02:51 +00002728<h5>Arguments:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002729
Reid Spencer7eb55b32006-11-02 01:53:59 +00002730<p>The two arguments to the '<tt>srem</tt>' instruction must be
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002731<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2732values. Both arguments must have identical types.</p>
2733
Chris Lattner48b383b02003-11-25 01:02:51 +00002734<h5>Semantics:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002735
Reid Spencer7eb55b32006-11-02 01:53:59 +00002736<p>This instruction returns the <i>remainder</i> of a division (where the result
Gabor Greif0f75ad02008-08-07 21:46:00 +00002737has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2738operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
Reid Spencer806ad6a2007-03-24 22:23:39 +00002739a value. For more information about the difference, see <a
Chris Lattner48b383b02003-11-25 01:02:51 +00002740 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
Reid Spencer806ad6a2007-03-24 22:23:39 +00002741Math Forum</a>. For a table of how this is implemented in various languages,
Reid Spencerdb3b93b2007-03-24 22:40:44 +00002742please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
Reid Spencer806ad6a2007-03-24 22:23:39 +00002743Wikipedia: modulo operation</a>.</p>
Chris Lattner2f2427e2008-01-28 00:36:27 +00002744<p>Note that signed integer remainder and unsigned integer remainder are
2745distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2746<p>Taking the remainder of a division by zero leads to undefined behavior.
2747Overflow also leads to undefined behavior; this is a rare case, but can occur,
2748for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2749(The remainder doesn't actually overflow, but this rule lets srem be
2750implemented using instructions that return both the result of the division
2751and the remainder.)</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00002752<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002753<pre> &lt;result&gt; = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
Reid Spencer7eb55b32006-11-02 01:53:59 +00002754</pre>
2755
2756</div>
2757<!-- _______________________________________________________________________ -->
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002758<div class="doc_subsubsection">
2759 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2760
Reid Spencer7eb55b32006-11-02 01:53:59 +00002761<div class="doc_text">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002762
Reid Spencer7eb55b32006-11-02 01:53:59 +00002763<h5>Syntax:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +00002764<pre> &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 +00002765</pre>
2766<h5>Overview:</h5>
2767<p>The '<tt>frem</tt>' instruction returns the remainder from the
2768division of its two operands.</p>
2769<h5>Arguments:</h5>
2770<p>The two arguments to the '<tt>frem</tt>' instruction must be
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002771<a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2772of floating point values. Both arguments must have identical types.</p>
2773
Reid Spencer7eb55b32006-11-02 01:53:59 +00002774<h5>Semantics:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002775
Chris Lattner1429e6f2008-04-01 18:45:27 +00002776<p>This instruction returns the <i>remainder</i> of a division.
2777The remainder has the same sign as the dividend.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002778
Reid Spencer7eb55b32006-11-02 01:53:59 +00002779<h5>Example:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002780
2781<pre>
2782 &lt;result&gt; = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
Chris Lattner48b383b02003-11-25 01:02:51 +00002783</pre>
Misha Brukman76307852003-11-08 01:05:38 +00002784</div>
Robert Bocchino820bc75b2006-02-17 21:18:08 +00002785
Reid Spencer2ab01932007-02-02 13:57:07 +00002786<!-- ======================================================================= -->
2787<div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2788Operations</a> </div>
2789<div class="doc_text">
2790<p>Bitwise binary operators are used to do various forms of
2791bit-twiddling in a program. They are generally very efficient
2792instructions and can commonly be strength reduced from other
Chris Lattner1429e6f2008-04-01 18:45:27 +00002793instructions. They require two operands of the same type, execute an operation on them,
2794and produce a single value. The resulting value is the same type as its operands.</p>
Reid Spencer2ab01932007-02-02 13:57:07 +00002795</div>
2796
Reid Spencer04e259b2007-01-31 21:39:12 +00002797<!-- _______________________________________________________________________ -->
2798<div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2799Instruction</a> </div>
2800<div class="doc_text">
2801<h5>Syntax:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +00002802<pre> &lt;result&gt; = shl &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer04e259b2007-01-31 21:39:12 +00002803</pre>
Chris Lattnerf0e50112007-10-03 21:01:14 +00002804
Reid Spencer04e259b2007-01-31 21:39:12 +00002805<h5>Overview:</h5>
Chris Lattnerf0e50112007-10-03 21:01:14 +00002806
Reid Spencer04e259b2007-01-31 21:39:12 +00002807<p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2808the left a specified number of bits.</p>
Chris Lattnerf0e50112007-10-03 21:01:14 +00002809
Reid Spencer04e259b2007-01-31 21:39:12 +00002810<h5>Arguments:</h5>
Chris Lattnerf0e50112007-10-03 21:01:14 +00002811
Reid Spencer04e259b2007-01-31 21:39:12 +00002812<p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
Nate Begemanfecbc8c2008-07-29 15:49:41 +00002813 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
Gabor Greif0f75ad02008-08-07 21:46:00 +00002814type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Chris Lattnerf0e50112007-10-03 21:01:14 +00002815
Reid Spencer04e259b2007-01-31 21:39:12 +00002816<h5>Semantics:</h5>
Chris Lattnerf0e50112007-10-03 21:01:14 +00002817
Gabor Greif0f75ad02008-08-07 21:46:00 +00002818<p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2819where n is the width of the result. If <tt>op2</tt> is (statically or dynamically) negative or
Mon P Wang68d4eee2008-12-10 08:55:09 +00002820equal to or larger than the number of bits in <tt>op1</tt>, the result is undefined.
2821If the arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2822corresponding shift amount in <tt>op2</tt>.</p>
Chris Lattnerf0e50112007-10-03 21:01:14 +00002823
Reid Spencer04e259b2007-01-31 21:39:12 +00002824<h5>Example:</h5><pre>
2825 &lt;result&gt; = shl i32 4, %var <i>; yields {i32}: 4 &lt;&lt; %var</i>
2826 &lt;result&gt; = shl i32 4, 2 <i>; yields {i32}: 16</i>
2827 &lt;result&gt; = shl i32 1, 10 <i>; yields {i32}: 1024</i>
Chris Lattnerf0e50112007-10-03 21:01:14 +00002828 &lt;result&gt; = shl i32 1, 32 <i>; undefined</i>
Mon P Wang4dd832d2008-12-09 05:46:39 +00002829 &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 +00002830</pre>
2831</div>
2832<!-- _______________________________________________________________________ -->
2833<div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2834Instruction</a> </div>
2835<div class="doc_text">
2836<h5>Syntax:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +00002837<pre> &lt;result&gt; = lshr &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer04e259b2007-01-31 21:39:12 +00002838</pre>
2839
2840<h5>Overview:</h5>
2841<p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
Jeff Cohen5819f182007-04-22 01:17:39 +00002842operand shifted to the right a specified number of bits with zero fill.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +00002843
2844<h5>Arguments:</h5>
2845<p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
Nate Begemanfecbc8c2008-07-29 15:49:41 +00002846<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
Gabor Greif0f75ad02008-08-07 21:46:00 +00002847type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +00002848
2849<h5>Semantics:</h5>
Chris Lattnerf0e50112007-10-03 21:01:14 +00002850
Reid Spencer04e259b2007-01-31 21:39:12 +00002851<p>This instruction always performs a logical shift right operation. The most
2852significant bits of the result will be filled with zero bits after the
Gabor Greif0f75ad02008-08-07 21:46:00 +00002853shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
Mon P Wang68d4eee2008-12-10 08:55:09 +00002854the number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
2855vectors, each vector element of <tt>op1</tt> is shifted by the corresponding shift
2856amount in <tt>op2</tt>.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +00002857
2858<h5>Example:</h5>
2859<pre>
2860 &lt;result&gt; = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2861 &lt;result&gt; = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2862 &lt;result&gt; = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2863 &lt;result&gt; = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
Chris Lattnerf0e50112007-10-03 21:01:14 +00002864 &lt;result&gt; = lshr i32 1, 32 <i>; undefined</i>
Mon P Wang4dd832d2008-12-09 05:46:39 +00002865 &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 +00002866</pre>
2867</div>
2868
Reid Spencer2ab01932007-02-02 13:57:07 +00002869<!-- _______________________________________________________________________ -->
Reid Spencer04e259b2007-01-31 21:39:12 +00002870<div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2871Instruction</a> </div>
2872<div class="doc_text">
2873
2874<h5>Syntax:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +00002875<pre> &lt;result&gt; = ashr &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Reid Spencer04e259b2007-01-31 21:39:12 +00002876</pre>
2877
2878<h5>Overview:</h5>
2879<p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
Jeff Cohen5819f182007-04-22 01:17:39 +00002880operand shifted to the right a specified number of bits with sign extension.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +00002881
2882<h5>Arguments:</h5>
2883<p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
Nate Begemanfecbc8c2008-07-29 15:49:41 +00002884<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
Gabor Greif0f75ad02008-08-07 21:46:00 +00002885type. '<tt>op2</tt>' is treated as an unsigned value.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +00002886
2887<h5>Semantics:</h5>
2888<p>This instruction always performs an arithmetic shift right operation,
2889The most significant bits of the result will be filled with the sign bit
Gabor Greif0f75ad02008-08-07 21:46:00 +00002890of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
Mon P Wang68d4eee2008-12-10 08:55:09 +00002891larger than the number of bits in <tt>op1</tt>, the result is undefined. If the
2892arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2893corresponding shift amount in <tt>op2</tt>.</p>
Reid Spencer04e259b2007-01-31 21:39:12 +00002894
2895<h5>Example:</h5>
2896<pre>
2897 &lt;result&gt; = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2898 &lt;result&gt; = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2899 &lt;result&gt; = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2900 &lt;result&gt; = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
Chris Lattnerf0e50112007-10-03 21:01:14 +00002901 &lt;result&gt; = ashr i32 1, 32 <i>; undefined</i>
Mon P Wang4dd832d2008-12-09 05:46:39 +00002902 &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 +00002903</pre>
2904</div>
2905
Chris Lattner2f7c9632001-06-06 20:29:01 +00002906<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00002907<div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2908Instruction</a> </div>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002909
Misha Brukman76307852003-11-08 01:05:38 +00002910<div class="doc_text">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002911
Chris Lattner2f7c9632001-06-06 20:29:01 +00002912<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002913
2914<pre>
Gabor Greif0f75ad02008-08-07 21:46:00 +00002915 &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 +00002916</pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002917
Chris Lattner2f7c9632001-06-06 20:29:01 +00002918<h5>Overview:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002919
Chris Lattner48b383b02003-11-25 01:02:51 +00002920<p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2921its two operands.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002922
Chris Lattner2f7c9632001-06-06 20:29:01 +00002923<h5>Arguments:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002924
2925<p>The two arguments to the '<tt>and</tt>' instruction must be
2926<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2927values. Both arguments must have identical types.</p>
2928
Chris Lattner2f7c9632001-06-06 20:29:01 +00002929<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00002930<p>The truth table used for the '<tt>and</tt>' instruction is:</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00002931<p> </p>
Bill Wendling5703c6e2008-09-07 10:29:20 +00002932<div>
Misha Brukman76307852003-11-08 01:05:38 +00002933<table border="1" cellspacing="0" cellpadding="4">
Chris Lattner48b383b02003-11-25 01:02:51 +00002934 <tbody>
2935 <tr>
2936 <td>In0</td>
2937 <td>In1</td>
2938 <td>Out</td>
2939 </tr>
2940 <tr>
2941 <td>0</td>
2942 <td>0</td>
2943 <td>0</td>
2944 </tr>
2945 <tr>
2946 <td>0</td>
2947 <td>1</td>
2948 <td>0</td>
2949 </tr>
2950 <tr>
2951 <td>1</td>
2952 <td>0</td>
2953 <td>0</td>
2954 </tr>
2955 <tr>
2956 <td>1</td>
2957 <td>1</td>
2958 <td>1</td>
2959 </tr>
2960 </tbody>
2961</table>
Misha Brukmanc501f552004-03-01 17:47:27 +00002962</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002963<h5>Example:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002964<pre>
2965 &lt;result&gt; = and i32 4, %var <i>; yields {i32}:result = 4 &amp; %var</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00002966 &lt;result&gt; = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2967 &lt;result&gt; = and i32 4, 8 <i>; yields {i32}:result = 0</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002968</pre>
Misha Brukman76307852003-11-08 01:05:38 +00002969</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002970<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00002971<div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00002972<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00002973<h5>Syntax:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +00002974<pre> &lt;result&gt; = or &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002975</pre>
Chris Lattner48b383b02003-11-25 01:02:51 +00002976<h5>Overview:</h5>
2977<p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2978or of its two operands.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002979<h5>Arguments:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00002980
2981<p>The two arguments to the '<tt>or</tt>' instruction must be
2982<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2983values. Both arguments must have identical types.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00002984<h5>Semantics:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00002985<p>The truth table used for the '<tt>or</tt>' instruction is:</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00002986<p> </p>
Bill Wendling5703c6e2008-09-07 10:29:20 +00002987<div>
Chris Lattner48b383b02003-11-25 01:02:51 +00002988<table border="1" cellspacing="0" cellpadding="4">
2989 <tbody>
2990 <tr>
2991 <td>In0</td>
2992 <td>In1</td>
2993 <td>Out</td>
2994 </tr>
2995 <tr>
2996 <td>0</td>
2997 <td>0</td>
2998 <td>0</td>
2999 </tr>
3000 <tr>
3001 <td>0</td>
3002 <td>1</td>
3003 <td>1</td>
3004 </tr>
3005 <tr>
3006 <td>1</td>
3007 <td>0</td>
3008 <td>1</td>
3009 </tr>
3010 <tr>
3011 <td>1</td>
3012 <td>1</td>
3013 <td>1</td>
3014 </tr>
3015 </tbody>
3016</table>
Misha Brukmanc501f552004-03-01 17:47:27 +00003017</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003018<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003019<pre> &lt;result&gt; = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
3020 &lt;result&gt; = or i32 15, 40 <i>; yields {i32}:result = 47</i>
3021 &lt;result&gt; = or i32 4, 8 <i>; yields {i32}:result = 12</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003022</pre>
Misha Brukman76307852003-11-08 01:05:38 +00003023</div>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003024<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00003025<div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3026Instruction</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00003027<div class="doc_text">
Chris Lattner2f7c9632001-06-06 20:29:01 +00003028<h5>Syntax:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +00003029<pre> &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 +00003030</pre>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003031<h5>Overview:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00003032<p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
3033or of its two operands. The <tt>xor</tt> is used to implement the
3034"one's complement" operation, which is the "~" operator in C.</p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003035<h5>Arguments:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003036<p>The two arguments to the '<tt>xor</tt>' instruction must be
3037<a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3038values. Both arguments must have identical types.</p>
3039
Chris Lattner2f7c9632001-06-06 20:29:01 +00003040<h5>Semantics:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00003041
Misha Brukman76307852003-11-08 01:05:38 +00003042<p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00003043<p> </p>
Bill Wendling5703c6e2008-09-07 10:29:20 +00003044<div>
Chris Lattner48b383b02003-11-25 01:02:51 +00003045<table border="1" cellspacing="0" cellpadding="4">
3046 <tbody>
3047 <tr>
3048 <td>In0</td>
3049 <td>In1</td>
3050 <td>Out</td>
3051 </tr>
3052 <tr>
3053 <td>0</td>
3054 <td>0</td>
3055 <td>0</td>
3056 </tr>
3057 <tr>
3058 <td>0</td>
3059 <td>1</td>
3060 <td>1</td>
3061 </tr>
3062 <tr>
3063 <td>1</td>
3064 <td>0</td>
3065 <td>1</td>
3066 </tr>
3067 <tr>
3068 <td>1</td>
3069 <td>1</td>
3070 <td>0</td>
3071 </tr>
3072 </tbody>
3073</table>
Misha Brukmanc501f552004-03-01 17:47:27 +00003074</div>
Chris Lattner48b383b02003-11-25 01:02:51 +00003075<p> </p>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003076<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003077<pre> &lt;result&gt; = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
3078 &lt;result&gt; = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3079 &lt;result&gt; = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3080 &lt;result&gt; = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003081</pre>
Misha Brukman76307852003-11-08 01:05:38 +00003082</div>
Chris Lattner54611b42005-11-06 08:02:57 +00003083
Chris Lattner2f7c9632001-06-06 20:29:01 +00003084<!-- ======================================================================= -->
Chris Lattner54611b42005-11-06 08:02:57 +00003085<div class="doc_subsection">
Chris Lattnerce83bff2006-04-08 23:07:04 +00003086 <a name="vectorops">Vector Operations</a>
3087</div>
3088
3089<div class="doc_text">
3090
3091<p>LLVM supports several instructions to represent vector operations in a
Jeff Cohen5819f182007-04-22 01:17:39 +00003092target-independent manner. These instructions cover the element-access and
Chris Lattnerce83bff2006-04-08 23:07:04 +00003093vector-specific operations needed to process vectors effectively. While LLVM
3094does directly support these vector operations, many sophisticated algorithms
3095will want to use target-specific intrinsics to take full advantage of a specific
3096target.</p>
3097
3098</div>
3099
3100<!-- _______________________________________________________________________ -->
3101<div class="doc_subsubsection">
3102 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3103</div>
3104
3105<div class="doc_text">
3106
3107<h5>Syntax:</h5>
3108
3109<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003110 &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 +00003111</pre>
3112
3113<h5>Overview:</h5>
3114
3115<p>
3116The '<tt>extractelement</tt>' instruction extracts a single scalar
Reid Spencer404a3252007-02-15 03:07:05 +00003117element from a vector at a specified index.
Chris Lattnerce83bff2006-04-08 23:07:04 +00003118</p>
3119
3120
3121<h5>Arguments:</h5>
3122
3123<p>
3124The first operand of an '<tt>extractelement</tt>' instruction is a
Reid Spencer404a3252007-02-15 03:07:05 +00003125value of <a href="#t_vector">vector</a> type. The second operand is
Chris Lattnerce83bff2006-04-08 23:07:04 +00003126an index indicating the position from which to extract the element.
3127The index may be a variable.</p>
3128
3129<h5>Semantics:</h5>
3130
3131<p>
3132The result is a scalar of the same type as the element type of
3133<tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3134<tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3135results are undefined.
3136</p>
3137
3138<h5>Example:</h5>
3139
3140<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003141 %result = extractelement &lt;4 x i32&gt; %vec, i32 0 <i>; yields i32</i>
Chris Lattnerce83bff2006-04-08 23:07:04 +00003142</pre>
3143</div>
3144
3145
3146<!-- _______________________________________________________________________ -->
3147<div class="doc_subsubsection">
3148 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
3149</div>
3150
3151<div class="doc_text">
3152
3153<h5>Syntax:</h5>
3154
3155<pre>
Dan Gohman43ba0672008-05-12 23:38:42 +00003156 &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 +00003157</pre>
3158
3159<h5>Overview:</h5>
3160
3161<p>
3162The '<tt>insertelement</tt>' instruction inserts a scalar
Reid Spencer404a3252007-02-15 03:07:05 +00003163element into a vector at a specified index.
Chris Lattnerce83bff2006-04-08 23:07:04 +00003164</p>
3165
3166
3167<h5>Arguments:</h5>
3168
3169<p>
3170The first operand of an '<tt>insertelement</tt>' instruction is a
Reid Spencer404a3252007-02-15 03:07:05 +00003171value of <a href="#t_vector">vector</a> type. The second operand is a
Chris Lattnerce83bff2006-04-08 23:07:04 +00003172scalar value whose type must equal the element type of the first
3173operand. The third operand is an index indicating the position at
3174which to insert the value. The index may be a variable.</p>
3175
3176<h5>Semantics:</h5>
3177
3178<p>
Reid Spencer404a3252007-02-15 03:07:05 +00003179The result is a vector of the same type as <tt>val</tt>. Its
Chris Lattnerce83bff2006-04-08 23:07:04 +00003180element values are those of <tt>val</tt> except at position
3181<tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
3182exceeds the length of <tt>val</tt>, the results are undefined.
3183</p>
3184
3185<h5>Example:</h5>
3186
3187<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003188 %result = insertelement &lt;4 x i32&gt; %vec, i32 1, i32 0 <i>; yields &lt;4 x i32&gt;</i>
Chris Lattnerce83bff2006-04-08 23:07:04 +00003189</pre>
3190</div>
3191
3192<!-- _______________________________________________________________________ -->
3193<div class="doc_subsubsection">
3194 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3195</div>
3196
3197<div class="doc_text">
3198
3199<h5>Syntax:</h5>
3200
3201<pre>
Mon P Wang25f01062008-11-10 04:46:22 +00003202 &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 +00003203</pre>
3204
3205<h5>Overview:</h5>
3206
3207<p>
3208The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
Mon P Wang25f01062008-11-10 04:46:22 +00003209from two input vectors, returning a vector with the same element type as
3210the input and length that is the same as the shuffle mask.
Chris Lattnerce83bff2006-04-08 23:07:04 +00003211</p>
3212
3213<h5>Arguments:</h5>
3214
3215<p>
Mon P Wang25f01062008-11-10 04:46:22 +00003216The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3217with types that match each other. The third argument is a shuffle mask whose
3218element type is always 'i32'. The result of the instruction is a vector whose
3219length is the same as the shuffle mask and whose element type is the same as
3220the element type of the first two operands.
Chris Lattnerce83bff2006-04-08 23:07:04 +00003221</p>
3222
3223<p>
3224The shuffle mask operand is required to be a constant vector with either
3225constant integer or undef values.
3226</p>
3227
3228<h5>Semantics:</h5>
3229
3230<p>
3231The elements of the two input vectors are numbered from left to right across
3232both of the vectors. The shuffle mask operand specifies, for each element of
Mon P Wang25f01062008-11-10 04:46:22 +00003233the result vector, which element of the two input vectors the result element
Chris Lattnerce83bff2006-04-08 23:07:04 +00003234gets. The element selector may be undef (meaning "don't care") and the second
3235operand may be undef if performing a shuffle from only one vector.
3236</p>
3237
3238<h5>Example:</h5>
3239
3240<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003241 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
Jeff Cohen5819f182007-04-22 01:17:39 +00003242 &lt;4 x i32&gt; &lt;i32 0, i32 4, i32 1, i32 5&gt; <i>; yields &lt;4 x i32&gt;</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003243 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; undef,
3244 &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.
Mon P Wang25f01062008-11-10 04:46:22 +00003245 %result = shufflevector &lt;8 x i32&gt; %v1, &lt;8 x i32&gt; undef,
3246 &lt;4 x i32&gt; &lt;i32 0, i32 1, i32 2, i32 3&gt; <i>; yields &lt;4 x i32&gt;</i>
3247 %result = shufflevector &lt;4 x i32&gt; %v1, &lt;4 x i32&gt; %v2,
3248 &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 +00003249</pre>
3250</div>
3251
Tanya Lattnerb138bbe2006-04-14 19:24:33 +00003252
Chris Lattnerce83bff2006-04-08 23:07:04 +00003253<!-- ======================================================================= -->
3254<div class="doc_subsection">
Dan Gohmanb9d66602008-05-12 23:51:09 +00003255 <a name="aggregateops">Aggregate Operations</a>
3256</div>
3257
3258<div class="doc_text">
3259
3260<p>LLVM supports several instructions for working with aggregate values.
3261</p>
3262
3263</div>
3264
3265<!-- _______________________________________________________________________ -->
3266<div class="doc_subsubsection">
3267 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3268</div>
3269
3270<div class="doc_text">
3271
3272<h5>Syntax:</h5>
3273
3274<pre>
3275 &lt;result&gt; = extractvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;idx&gt;{, &lt;idx&gt;}*
3276</pre>
3277
3278<h5>Overview:</h5>
3279
3280<p>
Dan Gohman35a835c2008-05-13 18:16:06 +00003281The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3282or array element from an aggregate value.
Dan Gohmanb9d66602008-05-12 23:51:09 +00003283</p>
3284
3285
3286<h5>Arguments:</h5>
3287
3288<p>
3289The first operand of an '<tt>extractvalue</tt>' instruction is a
3290value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
Dan Gohman35a835c2008-05-13 18:16:06 +00003291type. The operands are constant indices to specify which value to extract
Dan Gohman1ecaf452008-05-31 00:58:22 +00003292in a similar manner as indices in a
Dan Gohmanb9d66602008-05-12 23:51:09 +00003293'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3294</p>
3295
3296<h5>Semantics:</h5>
3297
3298<p>
3299The result is the value at the position in the aggregate specified by
3300the index operands.
3301</p>
3302
3303<h5>Example:</h5>
3304
3305<pre>
Dan Gohman1ecaf452008-05-31 00:58:22 +00003306 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
Dan Gohmanb9d66602008-05-12 23:51:09 +00003307</pre>
3308</div>
3309
3310
3311<!-- _______________________________________________________________________ -->
3312<div class="doc_subsubsection">
3313 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3314</div>
3315
3316<div class="doc_text">
3317
3318<h5>Syntax:</h5>
3319
3320<pre>
Dan Gohman1ecaf452008-05-31 00:58:22 +00003321 &lt;result&gt; = insertvalue &lt;aggregate type&gt; &lt;val&gt;, &lt;ty&gt; &lt;val&gt;, &lt;idx&gt; <i>; yields &lt;n x &lt;ty&gt;&gt;</i>
Dan Gohmanb9d66602008-05-12 23:51:09 +00003322</pre>
3323
3324<h5>Overview:</h5>
3325
3326<p>
3327The '<tt>insertvalue</tt>' instruction inserts a value
Dan Gohman35a835c2008-05-13 18:16:06 +00003328into a struct field or array element in an aggregate.
Dan Gohmanb9d66602008-05-12 23:51:09 +00003329</p>
3330
3331
3332<h5>Arguments:</h5>
3333
3334<p>
3335The first operand of an '<tt>insertvalue</tt>' instruction is a
3336value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3337The second operand is a first-class value to insert.
Dan Gohman34d1c0d2008-05-23 21:53:15 +00003338The following operands are constant indices
Dan Gohman1ecaf452008-05-31 00:58:22 +00003339indicating the position at which to insert the value in a similar manner as
Dan Gohman35a835c2008-05-13 18:16:06 +00003340indices in a
Dan Gohmanb9d66602008-05-12 23:51:09 +00003341'<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3342The value to insert must have the same type as the value identified
Dan Gohman35a835c2008-05-13 18:16:06 +00003343by the indices.
Dan Gohmanef9462f2008-10-14 16:51:45 +00003344</p>
Dan Gohmanb9d66602008-05-12 23:51:09 +00003345
3346<h5>Semantics:</h5>
3347
3348<p>
3349The result is an aggregate of the same type as <tt>val</tt>. Its
3350value is that of <tt>val</tt> except that the value at the position
Dan Gohman35a835c2008-05-13 18:16:06 +00003351specified by the indices is that of <tt>elt</tt>.
Dan Gohmanb9d66602008-05-12 23:51:09 +00003352</p>
3353
3354<h5>Example:</h5>
3355
3356<pre>
Dan Gohman88ce1a52008-06-23 15:26:37 +00003357 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
Dan Gohmanb9d66602008-05-12 23:51:09 +00003358</pre>
3359</div>
3360
3361
3362<!-- ======================================================================= -->
3363<div class="doc_subsection">
Chris Lattner6ab66722006-08-15 00:45:58 +00003364 <a name="memoryops">Memory Access and Addressing Operations</a>
Chris Lattner54611b42005-11-06 08:02:57 +00003365</div>
3366
Misha Brukman76307852003-11-08 01:05:38 +00003367<div class="doc_text">
Chris Lattner54611b42005-11-06 08:02:57 +00003368
Chris Lattner48b383b02003-11-25 01:02:51 +00003369<p>A key design point of an SSA-based representation is how it
3370represents memory. In LLVM, no memory locations are in SSA form, which
3371makes things very simple. This section describes how to read, write,
John Criswelldfe6a862004-12-10 15:51:16 +00003372allocate, and free memory in LLVM.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00003373
Misha Brukman76307852003-11-08 01:05:38 +00003374</div>
Chris Lattner54611b42005-11-06 08:02:57 +00003375
Chris Lattner2f7c9632001-06-06 20:29:01 +00003376<!-- _______________________________________________________________________ -->
Chris Lattner54611b42005-11-06 08:02:57 +00003377<div class="doc_subsubsection">
3378 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3379</div>
3380
Misha Brukman76307852003-11-08 01:05:38 +00003381<div class="doc_text">
Chris Lattner54611b42005-11-06 08:02:57 +00003382
Chris Lattner2f7c9632001-06-06 20:29:01 +00003383<h5>Syntax:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00003384
3385<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003386 &lt;result&gt; = malloc &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003387</pre>
Chris Lattner54611b42005-11-06 08:02:57 +00003388
Chris Lattner2f7c9632001-06-06 20:29:01 +00003389<h5>Overview:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00003390
Chris Lattner48b383b02003-11-25 01:02:51 +00003391<p>The '<tt>malloc</tt>' instruction allocates memory from the system
Christopher Lamb55c6d4f2007-12-17 01:00:21 +00003392heap and returns a pointer to it. The object is always allocated in the generic
3393address space (address space zero).</p>
Chris Lattner54611b42005-11-06 08:02:57 +00003394
Chris Lattner2f7c9632001-06-06 20:29:01 +00003395<h5>Arguments:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00003396
3397<p>The '<tt>malloc</tt>' instruction allocates
3398<tt>sizeof(&lt;type&gt;)*NumElements</tt>
John Criswella92e5862004-02-24 16:13:56 +00003399bytes of memory from the operating system and returns a pointer of the
Chris Lattner54611b42005-11-06 08:02:57 +00003400appropriate type to the program. If "NumElements" is specified, it is the
Gabor Greifdd1fc982008-02-09 22:24:34 +00003401number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner1f17cce2008-04-02 00:38:26 +00003402If a constant alignment is specified, the value result of the allocation is guaranteed to
Gabor Greifdd1fc982008-02-09 22:24:34 +00003403be aligned to at least that boundary. If not specified, or if zero, the target can
3404choose to align the allocation on any convenient boundary.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00003405
Misha Brukman76307852003-11-08 01:05:38 +00003406<p>'<tt>type</tt>' must be a sized type.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00003407
Chris Lattner2f7c9632001-06-06 20:29:01 +00003408<h5>Semantics:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00003409
Chris Lattner48b383b02003-11-25 01:02:51 +00003410<p>Memory is allocated using the system "<tt>malloc</tt>" function, and
Nick Lewyckyf5ffcbc2008-11-24 03:41:24 +00003411a pointer is returned. The result of a zero byte allocation is undefined. The
Chris Lattnerfe8519c2008-04-19 21:01:16 +00003412result is null if there is insufficient memory available.</p>
Misha Brukman76307852003-11-08 01:05:38 +00003413
Chris Lattner54611b42005-11-06 08:02:57 +00003414<h5>Example:</h5>
3415
3416<pre>
Dan Gohman7a5acb52009-01-04 23:49:44 +00003417 %array = malloc [4 x i8] <i>; yields {[%4 x i8]*}:array</i>
Chris Lattner54611b42005-11-06 08:02:57 +00003418
Bill Wendling2d8b9a82007-05-29 09:42:13 +00003419 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3420 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3421 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3422 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3423 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003424</pre>
Dan Gohman3065b612009-01-12 23:12:39 +00003425
3426<p>Note that the code generator does not yet respect the
3427 alignment value.</p>
3428
Misha Brukman76307852003-11-08 01:05:38 +00003429</div>
Chris Lattner54611b42005-11-06 08:02:57 +00003430
Chris Lattner2f7c9632001-06-06 20:29:01 +00003431<!-- _______________________________________________________________________ -->
Chris Lattner54611b42005-11-06 08:02:57 +00003432<div class="doc_subsubsection">
3433 <a name="i_free">'<tt>free</tt>' Instruction</a>
3434</div>
3435
Misha Brukman76307852003-11-08 01:05:38 +00003436<div class="doc_text">
Chris Lattner54611b42005-11-06 08:02:57 +00003437
Chris Lattner2f7c9632001-06-06 20:29:01 +00003438<h5>Syntax:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00003439
3440<pre>
Dan Gohman7a5acb52009-01-04 23:49:44 +00003441 free &lt;type&gt; &lt;value&gt; <i>; yields {void}</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003442</pre>
Chris Lattner54611b42005-11-06 08:02:57 +00003443
Chris Lattner2f7c9632001-06-06 20:29:01 +00003444<h5>Overview:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00003445
Chris Lattner48b383b02003-11-25 01:02:51 +00003446<p>The '<tt>free</tt>' instruction returns memory back to the unused
John Criswell4a3327e2005-05-13 22:25:59 +00003447memory heap to be reallocated in the future.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00003448
Chris Lattner2f7c9632001-06-06 20:29:01 +00003449<h5>Arguments:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00003450
Chris Lattner48b383b02003-11-25 01:02:51 +00003451<p>'<tt>value</tt>' shall be a pointer value that points to a value
3452that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3453instruction.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00003454
Chris Lattner2f7c9632001-06-06 20:29:01 +00003455<h5>Semantics:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00003456
John Criswelldfe6a862004-12-10 15:51:16 +00003457<p>Access to the memory pointed to by the pointer is no longer defined
Chris Lattner0f103e12008-04-19 22:41:32 +00003458after this instruction executes. If the pointer is null, the operation
3459is a noop.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00003460
Chris Lattner2f7c9632001-06-06 20:29:01 +00003461<h5>Example:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00003462
3463<pre>
Dan Gohman7a5acb52009-01-04 23:49:44 +00003464 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003465 free [4 x i8]* %array
Chris Lattner2f7c9632001-06-06 20:29:01 +00003466</pre>
Misha Brukman76307852003-11-08 01:05:38 +00003467</div>
Chris Lattner54611b42005-11-06 08:02:57 +00003468
Chris Lattner2f7c9632001-06-06 20:29:01 +00003469<!-- _______________________________________________________________________ -->
Chris Lattner54611b42005-11-06 08:02:57 +00003470<div class="doc_subsubsection">
3471 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3472</div>
3473
Misha Brukman76307852003-11-08 01:05:38 +00003474<div class="doc_text">
Chris Lattner54611b42005-11-06 08:02:57 +00003475
Chris Lattner2f7c9632001-06-06 20:29:01 +00003476<h5>Syntax:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00003477
3478<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003479 &lt;result&gt; = alloca &lt;type&gt;[, i32 &lt;NumElements&gt;][, align &lt;alignment&gt;] <i>; yields {type*}:result</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003480</pre>
Chris Lattner54611b42005-11-06 08:02:57 +00003481
Chris Lattner2f7c9632001-06-06 20:29:01 +00003482<h5>Overview:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00003483
Jeff Cohen5819f182007-04-22 01:17:39 +00003484<p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3485currently executing function, to be automatically released when this function
Christopher Lamb55c6d4f2007-12-17 01:00:21 +00003486returns to its caller. The object is always allocated in the generic address
3487space (address space zero).</p>
Chris Lattner54611b42005-11-06 08:02:57 +00003488
Chris Lattner2f7c9632001-06-06 20:29:01 +00003489<h5>Arguments:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00003490
John Criswelldfe6a862004-12-10 15:51:16 +00003491<p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(&lt;type&gt;)*NumElements</tt>
Chris Lattner48b383b02003-11-25 01:02:51 +00003492bytes of memory on the runtime stack, returning a pointer of the
Gabor Greifdd1fc982008-02-09 22:24:34 +00003493appropriate type to the program. If "NumElements" is specified, it is the
3494number of elements allocated, otherwise "NumElements" is defaulted to be one.
Chris Lattner1f17cce2008-04-02 00:38:26 +00003495If a constant alignment is specified, the value result of the allocation is guaranteed
Gabor Greifdd1fc982008-02-09 22:24:34 +00003496to be aligned to at least that boundary. If not specified, or if zero, the target
3497can choose to align the allocation on any convenient boundary.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00003498
Misha Brukman76307852003-11-08 01:05:38 +00003499<p>'<tt>type</tt>' may be any sized type.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00003500
Chris Lattner2f7c9632001-06-06 20:29:01 +00003501<h5>Semantics:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00003502
Bill Wendling9ee6a312009-05-08 20:49:29 +00003503<p>Memory is allocated; a pointer is returned. The operation is undefined if
Chris Lattnerfe8519c2008-04-19 21:01:16 +00003504there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
Chris Lattner48b383b02003-11-25 01:02:51 +00003505memory is automatically released when the function returns. The '<tt>alloca</tt>'
3506instruction is commonly used to represent automatic variables that must
3507have an address available. When the function returns (either with the <tt><a
John Criswellc932bef2005-05-12 16:55:34 +00003508 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
Chris Lattner1f17cce2008-04-02 00:38:26 +00003509instructions), the memory is reclaimed. Allocating zero bytes
3510is legal, but the result is undefined.</p>
Chris Lattner54611b42005-11-06 08:02:57 +00003511
Chris Lattner2f7c9632001-06-06 20:29:01 +00003512<h5>Example:</h5>
Chris Lattner54611b42005-11-06 08:02:57 +00003513
3514<pre>
Dan Gohman7a5acb52009-01-04 23:49:44 +00003515 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3516 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3517 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3518 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
Chris Lattner2f7c9632001-06-06 20:29:01 +00003519</pre>
Misha Brukman76307852003-11-08 01:05:38 +00003520</div>
Chris Lattner54611b42005-11-06 08:02:57 +00003521
Chris Lattner2f7c9632001-06-06 20:29:01 +00003522<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00003523<div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3524Instruction</a> </div>
Misha Brukman76307852003-11-08 01:05:38 +00003525<div class="doc_text">
Chris Lattner095735d2002-05-06 03:03:22 +00003526<h5>Syntax:</h5>
Christopher Lambbff50202007-04-21 08:16:25 +00003527<pre> &lt;result&gt; = load &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;]<br> &lt;result&gt; = volatile load &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;]<br></pre>
Chris Lattner095735d2002-05-06 03:03:22 +00003528<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00003529<p>The '<tt>load</tt>' instruction is used to read from memory.</p>
Chris Lattner095735d2002-05-06 03:03:22 +00003530<h5>Arguments:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00003531<p>The argument to the '<tt>load</tt>' instruction specifies the memory
John Criswell4c0cf7f2005-10-24 16:17:18 +00003532address from which to load. The pointer must point to a <a
Chris Lattner10ee9652004-06-03 22:57:15 +00003533 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
John Criswell4c0cf7f2005-10-24 16:17:18 +00003534marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
Chris Lattner48b383b02003-11-25 01:02:51 +00003535the number or order of execution of this <tt>load</tt> with other
3536volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3537instructions. </p>
Chris Lattner2a1993f2008-01-06 21:04:43 +00003538<p>
Chris Lattner1f17cce2008-04-02 00:38:26 +00003539The optional constant "align" argument specifies the alignment of the operation
Chris Lattner2a1993f2008-01-06 21:04:43 +00003540(that is, the alignment of the memory address). A value of 0 or an
3541omitted "align" argument means that the operation has the preferential
3542alignment for the target. It is the responsibility of the code emitter
3543to ensure that the alignment information is correct. Overestimating
3544the alignment results in an undefined behavior. Underestimating the
3545alignment may produce less efficient code. An alignment of 1 is always
3546safe.
3547</p>
Chris Lattner095735d2002-05-06 03:03:22 +00003548<h5>Semantics:</h5>
Duncan Sandsb1656c12009-03-22 11:33:16 +00003549<p>The location of memory pointed to is loaded. If the value being loaded
3550is of scalar type then the number of bytes read does not exceed the minimum
3551number of bytes needed to hold all bits of the type. For example, loading an
3552<tt>i24</tt> reads at most three bytes. When loading a value of a type like
3553<tt>i20</tt> with a size that is not an integral number of bytes, the result
3554is undefined if the value was not originally written using a store of the
3555same type.</p>
Chris Lattner095735d2002-05-06 03:03:22 +00003556<h5>Examples:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003557<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Chris Lattner48b383b02003-11-25 01:02:51 +00003558 <a
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003559 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3560 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
Chris Lattner095735d2002-05-06 03:03:22 +00003561</pre>
Misha Brukman76307852003-11-08 01:05:38 +00003562</div>
Chris Lattner095735d2002-05-06 03:03:22 +00003563<!-- _______________________________________________________________________ -->
Chris Lattner48b383b02003-11-25 01:02:51 +00003564<div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3565Instruction</a> </div>
Reid Spencera89fb182006-11-09 21:18:01 +00003566<div class="doc_text">
Chris Lattner095735d2002-05-06 03:03:22 +00003567<h5>Syntax:</h5>
Christopher Lambbff50202007-04-21 08:16:25 +00003568<pre> store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
3569 volatile store &lt;ty&gt; &lt;value&gt;, &lt;ty&gt;* &lt;pointer&gt;[, align &lt;alignment&gt;] <i>; yields {void}</i>
Chris Lattner095735d2002-05-06 03:03:22 +00003570</pre>
Chris Lattner095735d2002-05-06 03:03:22 +00003571<h5>Overview:</h5>
Misha Brukman76307852003-11-08 01:05:38 +00003572<p>The '<tt>store</tt>' instruction is used to write to memory.</p>
Chris Lattner095735d2002-05-06 03:03:22 +00003573<h5>Arguments:</h5>
Chris Lattner48b383b02003-11-25 01:02:51 +00003574<p>There are two arguments to the '<tt>store</tt>' instruction: a value
Jeff Cohen5819f182007-04-22 01:17:39 +00003575to store and an address at which to store it. The type of the '<tt>&lt;pointer&gt;</tt>'
Chris Lattner1f17cce2008-04-02 00:38:26 +00003576operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3577of the '<tt>&lt;value&gt;</tt>'
John Criswell4a3327e2005-05-13 22:25:59 +00003578operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
Chris Lattner48b383b02003-11-25 01:02:51 +00003579optimizer is not allowed to modify the number or order of execution of
3580this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3581 href="#i_store">store</a></tt> instructions.</p>
Chris Lattner2a1993f2008-01-06 21:04:43 +00003582<p>
Chris Lattner1f17cce2008-04-02 00:38:26 +00003583The optional constant "align" argument specifies the alignment of the operation
Chris Lattner2a1993f2008-01-06 21:04:43 +00003584(that is, the alignment of the memory address). A value of 0 or an
3585omitted "align" argument means that the operation has the preferential
3586alignment for the target. It is the responsibility of the code emitter
3587to ensure that the alignment information is correct. Overestimating
3588the alignment results in an undefined behavior. Underestimating the
3589alignment may produce less efficient code. An alignment of 1 is always
3590safe.
3591</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00003592<h5>Semantics:</h5>
3593<p>The contents of memory are updated to contain '<tt>&lt;value&gt;</tt>'
Duncan Sandsb1656c12009-03-22 11:33:16 +00003594at the location specified by the '<tt>&lt;pointer&gt;</tt>' operand.
3595If '<tt>&lt;value&gt;</tt>' is of scalar type then the number of bytes
3596written does not exceed the minimum number of bytes needed to hold all
3597bits of the type. For example, storing an <tt>i24</tt> writes at most
3598three bytes. When writing a value of a type like <tt>i20</tt> with a
3599size that is not an integral number of bytes, it is unspecified what
3600happens to the extra bits that do not belong to the type, but they will
3601typically be overwritten.</p>
Chris Lattner095735d2002-05-06 03:03:22 +00003602<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003603<pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
Bill Wendling8830ffe2007-10-22 05:10:05 +00003604 store i32 3, i32* %ptr <i>; yields {void}</i>
3605 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
Chris Lattner095735d2002-05-06 03:03:22 +00003606</pre>
Reid Spencer443460a2006-11-09 21:15:49 +00003607</div>
3608
Chris Lattner095735d2002-05-06 03:03:22 +00003609<!-- _______________________________________________________________________ -->
Chris Lattner33fd7022004-04-05 01:30:49 +00003610<div class="doc_subsubsection">
3611 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3612</div>
3613
Misha Brukman76307852003-11-08 01:05:38 +00003614<div class="doc_text">
Chris Lattner590645f2002-04-14 06:13:44 +00003615<h5>Syntax:</h5>
Chris Lattner33fd7022004-04-05 01:30:49 +00003616<pre>
Matthijs Kooijman0e268272008-10-13 13:44:15 +00003617 &lt;result&gt; = getelementptr &lt;pty&gt;* &lt;ptrval&gt;{, &lt;ty&gt; &lt;idx&gt;}*
Chris Lattner33fd7022004-04-05 01:30:49 +00003618</pre>
3619
Chris Lattner590645f2002-04-14 06:13:44 +00003620<h5>Overview:</h5>
Chris Lattner33fd7022004-04-05 01:30:49 +00003621
3622<p>
3623The '<tt>getelementptr</tt>' instruction is used to get the address of a
Matthijs Kooijman0e268272008-10-13 13:44:15 +00003624subelement of an aggregate data structure. It performs address calculation only
3625and does not access memory.</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00003626
Chris Lattner590645f2002-04-14 06:13:44 +00003627<h5>Arguments:</h5>
Chris Lattner33fd7022004-04-05 01:30:49 +00003628
Matthijs Kooijman0e268272008-10-13 13:44:15 +00003629<p>The first argument is always a pointer, and forms the basis of the
3630calculation. The remaining arguments are indices, that indicate which of the
3631elements of the aggregate object are indexed. The interpretation of each index
3632is dependent on the type being indexed into. The first index always indexes the
3633pointer value given as the first argument, the second index indexes a value of
3634the type pointed to (not necessarily the value directly pointed to, since the
3635first index can be non-zero), etc. The first type indexed into must be a pointer
3636value, subsequent types can be arrays, vectors and structs. Note that subsequent
3637types being indexed into can never be pointers, since that would require loading
3638the pointer before continuing calculation.</p>
3639
3640<p>The type of each index argument depends on the type it is indexing into.
3641When indexing into a (packed) structure, only <tt>i32</tt> integer
3642<b>constants</b> are allowed. When indexing into an array, pointer or vector,
Sanjiv Gupta1f8555a2009-04-27 03:21:00 +00003643integers of any width are allowed (also non-constants).</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00003644
Chris Lattner48b383b02003-11-25 01:02:51 +00003645<p>For example, let's consider a C code fragment and how it gets
3646compiled to LLVM:</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00003647
Bill Wendling3716c5d2007-05-29 09:04:49 +00003648<div class="doc_code">
Chris Lattner33fd7022004-04-05 01:30:49 +00003649<pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00003650struct RT {
3651 char A;
Chris Lattnera446f1b2007-05-29 15:43:56 +00003652 int B[10][20];
Bill Wendling3716c5d2007-05-29 09:04:49 +00003653 char C;
3654};
3655struct ST {
Chris Lattnera446f1b2007-05-29 15:43:56 +00003656 int X;
Bill Wendling3716c5d2007-05-29 09:04:49 +00003657 double Y;
3658 struct RT Z;
3659};
Chris Lattner33fd7022004-04-05 01:30:49 +00003660
Chris Lattnera446f1b2007-05-29 15:43:56 +00003661int *foo(struct ST *s) {
Bill Wendling3716c5d2007-05-29 09:04:49 +00003662 return &amp;s[1].Z.B[5][13];
3663}
Chris Lattner33fd7022004-04-05 01:30:49 +00003664</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00003665</div>
Chris Lattner33fd7022004-04-05 01:30:49 +00003666
Misha Brukman76307852003-11-08 01:05:38 +00003667<p>The LLVM code generated by the GCC frontend is:</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00003668
Bill Wendling3716c5d2007-05-29 09:04:49 +00003669<div class="doc_code">
Chris Lattner33fd7022004-04-05 01:30:49 +00003670<pre>
Chris Lattnerbc088212009-01-11 20:53:49 +00003671%RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
3672%ST = <a href="#namedtypes">type</a> { i32, double, %RT }
Chris Lattner33fd7022004-04-05 01:30:49 +00003673
Bill Wendling3716c5d2007-05-29 09:04:49 +00003674define i32* %foo(%ST* %s) {
3675entry:
3676 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3677 ret i32* %reg
3678}
Chris Lattner33fd7022004-04-05 01:30:49 +00003679</pre>
Bill Wendling3716c5d2007-05-29 09:04:49 +00003680</div>
Chris Lattner33fd7022004-04-05 01:30:49 +00003681
Chris Lattner590645f2002-04-14 06:13:44 +00003682<h5>Semantics:</h5>
Chris Lattner33fd7022004-04-05 01:30:49 +00003683
Misha Brukman76307852003-11-08 01:05:38 +00003684<p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003685type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
Chris Lattner33fd7022004-04-05 01:30:49 +00003686}</tt>' type, a structure. The second index indexes into the third element of
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003687the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3688i8 }</tt>' type, another structure. The third index indexes into the second
3689element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
Chris Lattner33fd7022004-04-05 01:30:49 +00003690array. The two dimensions of the array are subscripted into, yielding an
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003691'<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3692to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00003693
Chris Lattner48b383b02003-11-25 01:02:51 +00003694<p>Note that it is perfectly legal to index partially through a
3695structure, returning a pointer to an inner element. Because of this,
3696the LLVM code for the given testcase is equivalent to:</p>
Chris Lattner33fd7022004-04-05 01:30:49 +00003697
3698<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003699 define i32* %foo(%ST* %s) {
3700 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
Jeff Cohen5819f182007-04-22 01:17:39 +00003701 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3702 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003703 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3704 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3705 ret i32* %t5
Chris Lattner33fd7022004-04-05 01:30:49 +00003706 }
Chris Lattnera8292f32002-05-06 22:08:29 +00003707</pre>
Chris Lattnerc0ad71e2005-06-24 17:22:57 +00003708
Chris Lattnerdd282822009-03-09 20:55:18 +00003709<p>Note that it is undefined to access an array out of bounds: array
3710and pointer indexes must always be within the defined bounds of the
3711array type when accessed with an instruction that dereferences the
3712pointer (e.g. a load or store instruction). The one exception for
3713this rule is zero length arrays. These arrays are defined to be
3714accessible as variable length arrays, which requires access beyond the
3715zero'th element.</p>
Chris Lattnerc0ad71e2005-06-24 17:22:57 +00003716
Chris Lattner6ab66722006-08-15 00:45:58 +00003717<p>The getelementptr instruction is often confusing. For some more insight
3718into how it works, see <a href="GetElementPtr.html">the getelementptr
3719FAQ</a>.</p>
3720
Chris Lattner590645f2002-04-14 06:13:44 +00003721<h5>Example:</h5>
Chris Lattnerc0ad71e2005-06-24 17:22:57 +00003722
Chris Lattner33fd7022004-04-05 01:30:49 +00003723<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003724 <i>; yields [12 x i8]*:aptr</i>
Matthijs Kooijman0e268272008-10-13 13:44:15 +00003725 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
3726 <i>; yields i8*:vptr</i>
Dan Gohmanef9462f2008-10-14 16:51:45 +00003727 %vptr = getelementptr {i32, &lt;2 x i8&gt;}* %svptr, i64 0, i32 1, i32 1
Matthijs Kooijman0e268272008-10-13 13:44:15 +00003728 <i>; yields i8*:eptr</i>
3729 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
Sanjiv Gupta0c155e62009-04-25 07:27:44 +00003730 <i>; yields i32*:iptr</i>
Sanjiv Gupta77abea02009-04-24 16:38:13 +00003731 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
Chris Lattner33fd7022004-04-05 01:30:49 +00003732</pre>
Chris Lattner33fd7022004-04-05 01:30:49 +00003733</div>
Reid Spencer443460a2006-11-09 21:15:49 +00003734
Chris Lattner2f7c9632001-06-06 20:29:01 +00003735<!-- ======================================================================= -->
Reid Spencer97c5fa42006-11-08 01:18:52 +00003736<div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
Misha Brukman76307852003-11-08 01:05:38 +00003737</div>
Misha Brukman76307852003-11-08 01:05:38 +00003738<div class="doc_text">
Reid Spencer97c5fa42006-11-08 01:18:52 +00003739<p>The instructions in this category are the conversion instructions (casting)
3740which all take a single operand and a type. They perform various bit conversions
3741on the operand.</p>
Misha Brukman76307852003-11-08 01:05:38 +00003742</div>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00003743
Chris Lattnera8292f32002-05-06 22:08:29 +00003744<!-- _______________________________________________________________________ -->
Chris Lattnerb53c28d2004-03-12 05:50:16 +00003745<div class="doc_subsubsection">
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003746 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3747</div>
3748<div class="doc_text">
3749
3750<h5>Syntax:</h5>
3751<pre>
3752 &lt;result&gt; = trunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3753</pre>
3754
3755<h5>Overview:</h5>
3756<p>
3757The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3758</p>
3759
3760<h5>Arguments:</h5>
3761<p>
3762The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3763be an <a href="#t_integer">integer</a> type, and a type that specifies the size
Chris Lattnerc0f423a2007-01-15 01:54:13 +00003764and type of the result, which must be an <a href="#t_integer">integer</a>
Reid Spencer51b07252006-11-09 23:03:26 +00003765type. The bit size of <tt>value</tt> must be larger than the bit size of
3766<tt>ty2</tt>. Equal sized types are not allowed.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003767
3768<h5>Semantics:</h5>
3769<p>
3770The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
Reid Spencer51b07252006-11-09 23:03:26 +00003771and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3772larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3773It will always truncate bits.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003774
3775<h5>Example:</h5>
3776<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003777 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
Reid Spencer36a15422007-01-12 03:35:51 +00003778 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3779 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003780</pre>
3781</div>
3782
3783<!-- _______________________________________________________________________ -->
3784<div class="doc_subsubsection">
3785 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3786</div>
3787<div class="doc_text">
3788
3789<h5>Syntax:</h5>
3790<pre>
3791 &lt;result&gt; = zext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3792</pre>
3793
3794<h5>Overview:</h5>
3795<p>The '<tt>zext</tt>' instruction zero extends its operand to type
3796<tt>ty2</tt>.</p>
3797
3798
3799<h5>Arguments:</h5>
3800<p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
Chris Lattnerc0f423a2007-01-15 01:54:13 +00003801<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3802also be of <a href="#t_integer">integer</a> type. The bit size of the
Reid Spencer51b07252006-11-09 23:03:26 +00003803<tt>value</tt> must be smaller than the bit size of the destination type,
3804<tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003805
3806<h5>Semantics:</h5>
3807<p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
Chris Lattnerc87f3df2007-05-24 19:13:27 +00003808bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003809
Reid Spencer07c9c682007-01-12 15:46:11 +00003810<p>When zero extending from i1, the result will always be either 0 or 1.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003811
3812<h5>Example:</h5>
3813<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003814 %X = zext i32 257 to i64 <i>; yields i64:257</i>
Reid Spencer36a15422007-01-12 03:35:51 +00003815 %Y = zext i1 true to i32 <i>; yields i32:1</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003816</pre>
3817</div>
3818
3819<!-- _______________________________________________________________________ -->
3820<div class="doc_subsubsection">
3821 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3822</div>
3823<div class="doc_text">
3824
3825<h5>Syntax:</h5>
3826<pre>
3827 &lt;result&gt; = sext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3828</pre>
3829
3830<h5>Overview:</h5>
3831<p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3832
3833<h5>Arguments:</h5>
3834<p>
3835The '<tt>sext</tt>' instruction takes a value to cast, which must be of
Chris Lattnerc0f423a2007-01-15 01:54:13 +00003836<a href="#t_integer">integer</a> type, and a type to cast it to, which must
3837also be of <a href="#t_integer">integer</a> type. The bit size of the
Reid Spencer51b07252006-11-09 23:03:26 +00003838<tt>value</tt> must be smaller than the bit size of the destination type,
3839<tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003840
3841<h5>Semantics:</h5>
3842<p>
3843The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3844bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
Chris Lattnerc87f3df2007-05-24 19:13:27 +00003845the type <tt>ty2</tt>.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003846
Reid Spencer36a15422007-01-12 03:35:51 +00003847<p>When sign extending from i1, the extension always results in -1 or 0.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003848
3849<h5>Example:</h5>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003850<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003851 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
Reid Spencer36a15422007-01-12 03:35:51 +00003852 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003853</pre>
3854</div>
3855
3856<!-- _______________________________________________________________________ -->
3857<div class="doc_subsubsection">
Reid Spencer2e2740d2006-11-09 21:48:10 +00003858 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3859</div>
3860
3861<div class="doc_text">
3862
3863<h5>Syntax:</h5>
3864
3865<pre>
3866 &lt;result&gt; = fptrunc &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3867</pre>
3868
3869<h5>Overview:</h5>
3870<p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3871<tt>ty2</tt>.</p>
3872
3873
3874<h5>Arguments:</h5>
3875<p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3876 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3877cast it to. The size of <tt>value</tt> must be larger than the size of
3878<tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3879<i>no-op cast</i>.</p>
3880
3881<h5>Semantics:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00003882<p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3883<a href="#t_floating">floating point</a> type to a smaller
3884<a href="#t_floating">floating point</a> type. If the value cannot fit within
3885the destination type, <tt>ty2</tt>, then the results are undefined.</p>
Reid Spencer2e2740d2006-11-09 21:48:10 +00003886
3887<h5>Example:</h5>
3888<pre>
3889 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3890 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3891</pre>
3892</div>
3893
3894<!-- _______________________________________________________________________ -->
3895<div class="doc_subsubsection">
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003896 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3897</div>
3898<div class="doc_text">
3899
3900<h5>Syntax:</h5>
3901<pre>
3902 &lt;result&gt; = fpext &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
3903</pre>
3904
3905<h5>Overview:</h5>
3906<p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3907floating point value.</p>
3908
3909<h5>Arguments:</h5>
3910<p>The '<tt>fpext</tt>' instruction takes a
3911<a href="#t_floating">floating point</a> <tt>value</tt> to cast,
Reid Spencer51b07252006-11-09 23:03:26 +00003912and a <a href="#t_floating">floating point</a> type to cast it to. The source
3913type must be smaller than the destination type.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003914
3915<h5>Semantics:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00003916<p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
Duncan Sands16f122e2007-03-30 12:22:09 +00003917<a href="#t_floating">floating point</a> type to a larger
3918<a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
Reid Spencer51b07252006-11-09 23:03:26 +00003919used to make a <i>no-op cast</i> because it always changes bits. Use
Reid Spencer5b950642006-11-11 23:08:07 +00003920<tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003921
3922<h5>Example:</h5>
3923<pre>
3924 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3925 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3926</pre>
3927</div>
3928
3929<!-- _______________________________________________________________________ -->
3930<div class="doc_subsubsection">
Reid Spencer2eadb532007-01-21 00:29:26 +00003931 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003932</div>
3933<div class="doc_text">
3934
3935<h5>Syntax:</h5>
3936<pre>
Reid Spencer753163d2007-07-31 14:40:14 +00003937 &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 +00003938</pre>
3939
3940<h5>Overview:</h5>
Reid Spencer753163d2007-07-31 14:40:14 +00003941<p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003942unsigned integer equivalent of type <tt>ty2</tt>.
3943</p>
3944
3945<h5>Arguments:</h5>
Reid Spencer753163d2007-07-31 14:40:14 +00003946<p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
Nate Begemand4d45c22007-11-17 03:58:34 +00003947scalar or vector <a href="#t_floating">floating point</a> value, and a type
3948to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3949type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3950vector integer type with the same number of elements as <tt>ty</tt></p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003951
3952<h5>Semantics:</h5>
Reid Spencer753163d2007-07-31 14:40:14 +00003953<p> The '<tt>fptoui</tt>' instruction converts its
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003954<a href="#t_floating">floating point</a> operand into the nearest (rounding
3955towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3956the results are undefined.</p>
3957
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003958<h5>Example:</h5>
3959<pre>
Reid Spencer753163d2007-07-31 14:40:14 +00003960 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
Chris Lattner5b95a172007-09-22 03:17:52 +00003961 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
Reid Spencer753163d2007-07-31 14:40:14 +00003962 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003963</pre>
3964</div>
3965
3966<!-- _______________________________________________________________________ -->
3967<div class="doc_subsubsection">
Reid Spencer51b07252006-11-09 23:03:26 +00003968 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003969</div>
3970<div class="doc_text">
3971
3972<h5>Syntax:</h5>
3973<pre>
Reid Spencer51b07252006-11-09 23:03:26 +00003974 &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 +00003975</pre>
3976
3977<h5>Overview:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00003978<p>The '<tt>fptosi</tt>' instruction converts
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003979<a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
Chris Lattnerb53c28d2004-03-12 05:50:16 +00003980</p>
3981
Chris Lattnera8292f32002-05-06 22:08:29 +00003982<h5>Arguments:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00003983<p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
Nate Begemand4d45c22007-11-17 03:58:34 +00003984scalar or vector <a href="#t_floating">floating point</a> value, and a type
3985to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3986type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3987vector integer type with the same number of elements as <tt>ty</tt></p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00003988
Chris Lattnera8292f32002-05-06 22:08:29 +00003989<h5>Semantics:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00003990<p>The '<tt>fptosi</tt>' instruction converts its
Reid Spencer59b6b7d2006-11-08 01:11:31 +00003991<a href="#t_floating">floating point</a> operand into the nearest (rounding
3992towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3993the results are undefined.</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00003994
Chris Lattner70de6632001-07-09 00:26:23 +00003995<h5>Example:</h5>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00003996<pre>
Reid Spencer36a15422007-01-12 03:35:51 +00003997 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
Chris Lattner5b95a172007-09-22 03:17:52 +00003998 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00003999 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004000</pre>
4001</div>
4002
4003<!-- _______________________________________________________________________ -->
4004<div class="doc_subsubsection">
Reid Spencer51b07252006-11-09 23:03:26 +00004005 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004006</div>
4007<div class="doc_text">
4008
4009<h5>Syntax:</h5>
4010<pre>
Reid Spencer51b07252006-11-09 23:03:26 +00004011 &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 +00004012</pre>
4013
4014<h5>Overview:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00004015<p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004016integer and converts that value to the <tt>ty2</tt> type.</p>
4017
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004018<h5>Arguments:</h5>
Nate Begemand4d45c22007-11-17 03:58:34 +00004019<p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
4020scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
4021to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4022type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4023floating point type with the same number of elements as <tt>ty</tt></p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004024
4025<h5>Semantics:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00004026<p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004027integer quantity and converts it to the corresponding floating point value. If
Jeff Cohenbeccb742007-04-22 14:56:37 +00004028the value cannot fit in the floating point value, the results are undefined.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004029
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004030<h5>Example:</h5>
4031<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004032 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohmanef9462f2008-10-14 16:51:45 +00004033 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004034</pre>
4035</div>
4036
4037<!-- _______________________________________________________________________ -->
4038<div class="doc_subsubsection">
Reid Spencer51b07252006-11-09 23:03:26 +00004039 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004040</div>
4041<div class="doc_text">
4042
4043<h5>Syntax:</h5>
4044<pre>
Reid Spencer51b07252006-11-09 23:03:26 +00004045 &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 +00004046</pre>
4047
4048<h5>Overview:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00004049<p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004050integer and converts that value to the <tt>ty2</tt> type.</p>
4051
4052<h5>Arguments:</h5>
Nate Begemand4d45c22007-11-17 03:58:34 +00004053<p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
4054scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
4055to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4056type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4057floating point type with the same number of elements as <tt>ty</tt></p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004058
4059<h5>Semantics:</h5>
Reid Spencer51b07252006-11-09 23:03:26 +00004060<p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004061integer quantity and converts it to the corresponding floating point value. If
Jeff Cohenbeccb742007-04-22 14:56:37 +00004062the value cannot fit in the floating point value, the results are undefined.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004063
4064<h5>Example:</h5>
4065<pre>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004066 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
Dan Gohmanef9462f2008-10-14 16:51:45 +00004067 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004068</pre>
4069</div>
4070
4071<!-- _______________________________________________________________________ -->
4072<div class="doc_subsubsection">
Reid Spencerb7344ff2006-11-11 21:00:47 +00004073 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4074</div>
4075<div class="doc_text">
4076
4077<h5>Syntax:</h5>
4078<pre>
4079 &lt;result&gt; = ptrtoint &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4080</pre>
4081
4082<h5>Overview:</h5>
4083<p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4084the integer type <tt>ty2</tt>.</p>
4085
4086<h5>Arguments:</h5>
4087<p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
Duncan Sands16f122e2007-03-30 12:22:09 +00004088must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
Dan Gohmanef9462f2008-10-14 16:51:45 +00004089<tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
Reid Spencerb7344ff2006-11-11 21:00:47 +00004090
4091<h5>Semantics:</h5>
4092<p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4093<tt>ty2</tt> by interpreting the pointer value as an integer and either
4094truncating or zero extending that value to the size of the integer type. If
4095<tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4096<tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
Jeff Cohen222a8a42007-04-29 01:07:00 +00004097are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4098change.</p>
Reid Spencerb7344ff2006-11-11 21:00:47 +00004099
4100<h5>Example:</h5>
4101<pre>
Jeff Cohen222a8a42007-04-29 01:07:00 +00004102 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4103 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
Reid Spencerb7344ff2006-11-11 21:00:47 +00004104</pre>
4105</div>
4106
4107<!-- _______________________________________________________________________ -->
4108<div class="doc_subsubsection">
4109 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4110</div>
4111<div class="doc_text">
4112
4113<h5>Syntax:</h5>
4114<pre>
4115 &lt;result&gt; = inttoptr &lt;ty&gt; &lt;value&gt; to &lt;ty2&gt; <i>; yields ty2</i>
4116</pre>
4117
4118<h5>Overview:</h5>
4119<p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
4120a pointer type, <tt>ty2</tt>.</p>
4121
4122<h5>Arguments:</h5>
Duncan Sands16f122e2007-03-30 12:22:09 +00004123<p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
Reid Spencerb7344ff2006-11-11 21:00:47 +00004124value to cast, and a type to cast it to, which must be a
Dan Gohmanef9462f2008-10-14 16:51:45 +00004125<a href="#t_pointer">pointer</a> type.</p>
Reid Spencerb7344ff2006-11-11 21:00:47 +00004126
4127<h5>Semantics:</h5>
4128<p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4129<tt>ty2</tt> by applying either a zero extension or a truncation depending on
4130the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4131size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
4132the size of a pointer then a zero extension is done. If they are the same size,
4133nothing is done (<i>no-op cast</i>).</p>
4134
4135<h5>Example:</h5>
4136<pre>
Jeff Cohen222a8a42007-04-29 01:07:00 +00004137 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4138 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4139 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
Reid Spencerb7344ff2006-11-11 21:00:47 +00004140</pre>
4141</div>
4142
4143<!-- _______________________________________________________________________ -->
4144<div class="doc_subsubsection">
Reid Spencer5b950642006-11-11 23:08:07 +00004145 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004146</div>
4147<div class="doc_text">
4148
4149<h5>Syntax:</h5>
4150<pre>
Reid Spencer5b950642006-11-11 23:08:07 +00004151 &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 +00004152</pre>
4153
4154<h5>Overview:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00004155
Reid Spencer5b950642006-11-11 23:08:07 +00004156<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004157<tt>ty2</tt> without changing any bits.</p>
4158
4159<h5>Arguments:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00004160
Reid Spencer5b950642006-11-11 23:08:07 +00004161<p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
Dan Gohmanc05dca92008-09-08 16:45:59 +00004162a non-aggregate first class value, and a type to cast it to, which must also be
4163a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
4164<tt>value</tt>
Reid Spencere3db84c2007-01-09 20:08:58 +00004165and the destination type, <tt>ty2</tt>, must be identical. If the source
Chris Lattnerb54c30f2008-05-20 20:48:21 +00004166type is a pointer, the destination type must also be a pointer. This
4167instruction supports bitwise conversion of vectors to integers and to vectors
4168of other types (as long as they have the same size).</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004169
4170<h5>Semantics:</h5>
Reid Spencer5b950642006-11-11 23:08:07 +00004171<p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
Reid Spencerb7344ff2006-11-11 21:00:47 +00004172<tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4173this conversion. The conversion is done as if the <tt>value</tt> had been
4174stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
4175converted to other pointer types with this instruction. To convert pointers to
4176other types, use the <a href="#i_inttoptr">inttoptr</a> or
4177<a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
Reid Spencer59b6b7d2006-11-08 01:11:31 +00004178
4179<h5>Example:</h5>
4180<pre>
Jeff Cohen222a8a42007-04-29 01:07:00 +00004181 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004182 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
Dan Gohmanef9462f2008-10-14 16:51:45 +00004183 %Z = bitcast &lt;2 x int&gt; %V to i64; <i>; yields i64: %V</i>
Chris Lattner70de6632001-07-09 00:26:23 +00004184</pre>
Misha Brukman76307852003-11-08 01:05:38 +00004185</div>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00004186
Reid Spencer97c5fa42006-11-08 01:18:52 +00004187<!-- ======================================================================= -->
4188<div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4189<div class="doc_text">
4190<p>The instructions in this category are the "miscellaneous"
4191instructions, which defy better classification.</p>
4192</div>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004193
4194<!-- _______________________________________________________________________ -->
4195<div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4196</div>
4197<div class="doc_text">
4198<h5>Syntax:</h5>
Dan Gohmanef9462f2008-10-14 16:51:45 +00004199<pre> &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 +00004200</pre>
4201<h5>Overview:</h5>
Dan Gohmanc579d972008-09-09 01:02:47 +00004202<p>The '<tt>icmp</tt>' instruction returns a boolean value or
4203a vector of boolean values based on comparison
4204of its two integer, integer vector, or pointer operands.</p>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004205<h5>Arguments:</h5>
4206<p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
Jeff Cohen222a8a42007-04-29 01:07:00 +00004207the condition code indicating the kind of comparison to perform. It is not
4208a value, just a keyword. The possible condition code are:
Dan Gohmanef9462f2008-10-14 16:51:45 +00004209</p>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004210<ol>
4211 <li><tt>eq</tt>: equal</li>
4212 <li><tt>ne</tt>: not equal </li>
4213 <li><tt>ugt</tt>: unsigned greater than</li>
4214 <li><tt>uge</tt>: unsigned greater or equal</li>
4215 <li><tt>ult</tt>: unsigned less than</li>
4216 <li><tt>ule</tt>: unsigned less or equal</li>
4217 <li><tt>sgt</tt>: signed greater than</li>
4218 <li><tt>sge</tt>: signed greater or equal</li>
4219 <li><tt>slt</tt>: signed less than</li>
4220 <li><tt>sle</tt>: signed less or equal</li>
4221</ol>
Chris Lattnerc0f423a2007-01-15 01:54:13 +00004222<p>The remaining two arguments must be <a href="#t_integer">integer</a> or
Dan Gohmanc579d972008-09-09 01:02:47 +00004223<a href="#t_pointer">pointer</a>
4224or integer <a href="#t_vector">vector</a> typed.
4225They must also be identical types.</p>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004226<h5>Semantics:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +00004227<p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to
Reid Spencerc828a0e2006-11-18 21:50:54 +00004228the condition code given as <tt>cond</tt>. The comparison performed always
Dan Gohmanc579d972008-09-09 01:02:47 +00004229yields either an <a href="#t_primitive"><tt>i1</tt></a> or vector of <tt>i1</tt> result, as follows:
Dan Gohmanef9462f2008-10-14 16:51:45 +00004230</p>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004231<ol>
4232 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
4233 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
4234 </li>
4235 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
Dan Gohmanef9462f2008-10-14 16:51:45 +00004236 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004237 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
Gabor Greif0f75ad02008-08-07 21:46:00 +00004238 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004239 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
Gabor Greif0f75ad02008-08-07 21:46:00 +00004240 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004241 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
Gabor Greif0f75ad02008-08-07 21:46:00 +00004242 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004243 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
Gabor Greif0f75ad02008-08-07 21:46:00 +00004244 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004245 <li><tt>sgt</tt>: interprets the operands as signed values and yields
Gabor Greif0f75ad02008-08-07 21:46:00 +00004246 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004247 <li><tt>sge</tt>: interprets the operands as signed values and yields
Gabor Greif0f75ad02008-08-07 21:46:00 +00004248 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004249 <li><tt>slt</tt>: interprets the operands as signed values and yields
Gabor Greif0f75ad02008-08-07 21:46:00 +00004250 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004251 <li><tt>sle</tt>: interprets the operands as signed values and yields
Gabor Greif0f75ad02008-08-07 21:46:00 +00004252 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004253</ol>
4254<p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
Jeff Cohen222a8a42007-04-29 01:07:00 +00004255values are compared as if they were integers.</p>
Dan Gohmanc579d972008-09-09 01:02:47 +00004256<p>If the operands are integer vectors, then they are compared
4257element by element. The result is an <tt>i1</tt> vector with
4258the same number of elements as the values being compared.
4259Otherwise, the result is an <tt>i1</tt>.
4260</p>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004261
4262<h5>Example:</h5>
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004263<pre> &lt;result&gt; = icmp eq i32 4, 5 <i>; yields: result=false</i>
4264 &lt;result&gt; = icmp ne float* %X, %X <i>; yields: result=false</i>
4265 &lt;result&gt; = icmp ult i16 4, 5 <i>; yields: result=true</i>
4266 &lt;result&gt; = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4267 &lt;result&gt; = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4268 &lt;result&gt; = icmp sge i16 4, 5 <i>; yields: result=false</i>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004269</pre>
Dan Gohmana5127ab2009-01-22 01:39:38 +00004270
4271<p>Note that the code generator does not yet support vector types with
4272 the <tt>icmp</tt> instruction.</p>
4273
Reid Spencerc828a0e2006-11-18 21:50:54 +00004274</div>
4275
4276<!-- _______________________________________________________________________ -->
4277<div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4278</div>
4279<div class="doc_text">
4280<h5>Syntax:</h5>
Dan Gohmanef9462f2008-10-14 16:51:45 +00004281<pre> &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 +00004282</pre>
4283<h5>Overview:</h5>
Dan Gohmanc579d972008-09-09 01:02:47 +00004284<p>The '<tt>fcmp</tt>' instruction returns a boolean value
4285or vector of boolean values based on comparison
Dan Gohmanef9462f2008-10-14 16:51:45 +00004286of its operands.</p>
Dan Gohmanc579d972008-09-09 01:02:47 +00004287<p>
4288If the operands are floating point scalars, then the result
4289type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
4290</p>
4291<p>If the operands are floating point vectors, then the result type
4292is a vector of boolean with the same number of elements as the
4293operands being compared.</p>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004294<h5>Arguments:</h5>
4295<p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
Jeff Cohen222a8a42007-04-29 01:07:00 +00004296the condition code indicating the kind of comparison to perform. It is not
Dan Gohmanef9462f2008-10-14 16:51:45 +00004297a value, just a keyword. The possible condition code are:</p>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004298<ol>
Reid Spencerf69acf32006-11-19 03:00:14 +00004299 <li><tt>false</tt>: no comparison, always returns false</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004300 <li><tt>oeq</tt>: ordered and equal</li>
4301 <li><tt>ogt</tt>: ordered and greater than </li>
4302 <li><tt>oge</tt>: ordered and greater than or equal</li>
4303 <li><tt>olt</tt>: ordered and less than </li>
4304 <li><tt>ole</tt>: ordered and less than or equal</li>
4305 <li><tt>one</tt>: ordered and not equal</li>
4306 <li><tt>ord</tt>: ordered (no nans)</li>
4307 <li><tt>ueq</tt>: unordered or equal</li>
4308 <li><tt>ugt</tt>: unordered or greater than </li>
4309 <li><tt>uge</tt>: unordered or greater than or equal</li>
4310 <li><tt>ult</tt>: unordered or less than </li>
4311 <li><tt>ule</tt>: unordered or less than or equal</li>
4312 <li><tt>une</tt>: unordered or not equal</li>
4313 <li><tt>uno</tt>: unordered (either nans)</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00004314 <li><tt>true</tt>: no comparison, always returns true</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004315</ol>
Jeff Cohen222a8a42007-04-29 01:07:00 +00004316<p><i>Ordered</i> means that neither operand is a QNAN while
Reid Spencer02e0d1d2006-12-06 07:08:07 +00004317<i>unordered</i> means that either operand may be a QNAN.</p>
Dan Gohmanc579d972008-09-09 01:02:47 +00004318<p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
4319either a <a href="#t_floating">floating point</a> type
4320or a <a href="#t_vector">vector</a> of floating point type.
4321They must have identical types.</p>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004322<h5>Semantics:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +00004323<p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Dan Gohmanc579d972008-09-09 01:02:47 +00004324according to the condition code given as <tt>cond</tt>.
4325If the operands are vectors, then the vectors are compared
4326element by element.
4327Each comparison performed
Dan Gohmanef9462f2008-10-14 16:51:45 +00004328always yields an <a href="#t_primitive">i1</a> result, as follows:</p>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004329<ol>
4330 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00004331 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Gabor Greif0f75ad02008-08-07 21:46:00 +00004332 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00004333 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Gabor Greif0f75ad02008-08-07 21:46:00 +00004334 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00004335 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Gabor Greif0f75ad02008-08-07 21:46:00 +00004336 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00004337 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Gabor Greif0f75ad02008-08-07 21:46:00 +00004338 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00004339 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Gabor Greif0f75ad02008-08-07 21:46:00 +00004340 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00004341 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
Gabor Greif0f75ad02008-08-07 21:46:00 +00004342 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00004343 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4344 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
Gabor Greif0f75ad02008-08-07 21:46:00 +00004345 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00004346 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
Gabor Greif0f75ad02008-08-07 21:46:00 +00004347 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00004348 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
Gabor Greif0f75ad02008-08-07 21:46:00 +00004349 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00004350 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
Gabor Greif0f75ad02008-08-07 21:46:00 +00004351 <tt>op1</tt> is less than <tt>op2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00004352 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
Gabor Greif0f75ad02008-08-07 21:46:00 +00004353 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00004354 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
Gabor Greif0f75ad02008-08-07 21:46:00 +00004355 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
Reid Spencerf69acf32006-11-19 03:00:14 +00004356 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004357 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4358</ol>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004359
4360<h5>Example:</h5>
4361<pre> &lt;result&gt; = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
Dan Gohmanc579d972008-09-09 01:02:47 +00004362 &lt;result&gt; = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4363 &lt;result&gt; = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4364 &lt;result&gt; = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
Reid Spencerc828a0e2006-11-18 21:50:54 +00004365</pre>
Dan Gohmana5127ab2009-01-22 01:39:38 +00004366
4367<p>Note that the code generator does not yet support vector types with
4368 the <tt>fcmp</tt> instruction.</p>
4369
Reid Spencerc828a0e2006-11-18 21:50:54 +00004370</div>
4371
Reid Spencer97c5fa42006-11-08 01:18:52 +00004372<!-- _______________________________________________________________________ -->
Nate Begemand2195702008-05-12 19:01:56 +00004373<div class="doc_subsubsection">
4374 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
4375</div>
4376<div class="doc_text">
4377<h5>Syntax:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +00004378<pre> &lt;result&gt; = vicmp &lt;cond&gt; &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt; <i>; yields {ty}:result</i>
Nate Begemand2195702008-05-12 19:01:56 +00004379</pre>
4380<h5>Overview:</h5>
4381<p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
4382element-wise comparison of its two integer vector operands.</p>
4383<h5>Arguments:</h5>
4384<p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
4385the condition code indicating the kind of comparison to perform. It is not
Dan Gohmanef9462f2008-10-14 16:51:45 +00004386a value, just a keyword. The possible condition code are:</p>
Nate Begemand2195702008-05-12 19:01:56 +00004387<ol>
4388 <li><tt>eq</tt>: equal</li>
4389 <li><tt>ne</tt>: not equal </li>
4390 <li><tt>ugt</tt>: unsigned greater than</li>
4391 <li><tt>uge</tt>: unsigned greater or equal</li>
4392 <li><tt>ult</tt>: unsigned less than</li>
4393 <li><tt>ule</tt>: unsigned less or equal</li>
4394 <li><tt>sgt</tt>: signed greater than</li>
4395 <li><tt>sge</tt>: signed greater or equal</li>
4396 <li><tt>slt</tt>: signed less than</li>
4397 <li><tt>sle</tt>: signed less or equal</li>
4398</ol>
Dan Gohmanc579d972008-09-09 01:02:47 +00004399<p>The remaining two arguments must be <a href="#t_vector">vector</a> or
Nate Begemand2195702008-05-12 19:01:56 +00004400<a href="#t_integer">integer</a> typed. They must also be identical types.</p>
4401<h5>Semantics:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +00004402<p>The '<tt>vicmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begemand2195702008-05-12 19:01:56 +00004403according to the condition code given as <tt>cond</tt>. The comparison yields a
4404<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
4405identical type as the values being compared. The most significant bit in each
4406element is 1 if the element-wise comparison evaluates to true, and is 0
4407otherwise. All other bits of the result are undefined. The condition codes
4408are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
Dan Gohmanef9462f2008-10-14 16:51:45 +00004409instruction</a>.</p>
Nate Begemand2195702008-05-12 19:01:56 +00004410
4411<h5>Example:</h5>
4412<pre>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00004413 &lt;result&gt; = vicmp eq &lt;2 x i32&gt; &lt; i32 4, i32 0&gt;, &lt; i32 5, i32 0&gt; <i>; yields: result=&lt;2 x i32&gt; &lt; i32 0, i32 -1 &gt;</i>
4414 &lt;result&gt; = vicmp ult &lt;2 x i8 &gt; &lt; i8 1, i8 2&gt;, &lt; i8 2, i8 2 &gt; <i>; yields: result=&lt;2 x i8&gt; &lt; i8 -1, i8 0 &gt;</i>
Nate Begemand2195702008-05-12 19:01:56 +00004415</pre>
4416</div>
4417
4418<!-- _______________________________________________________________________ -->
4419<div class="doc_subsubsection">
4420 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4421</div>
4422<div class="doc_text">
4423<h5>Syntax:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +00004424<pre> &lt;result&gt; = vfcmp &lt;cond&gt; &lt;ty&gt; &lt;op1&gt;, &lt;op2&gt;</pre>
Nate Begemand2195702008-05-12 19:01:56 +00004425<h5>Overview:</h5>
4426<p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4427element-wise comparison of its two floating point vector operands. The output
4428elements have the same width as the input elements.</p>
4429<h5>Arguments:</h5>
4430<p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4431the condition code indicating the kind of comparison to perform. It is not
Dan Gohmanef9462f2008-10-14 16:51:45 +00004432a value, just a keyword. The possible condition code are:</p>
Nate Begemand2195702008-05-12 19:01:56 +00004433<ol>
4434 <li><tt>false</tt>: no comparison, always returns false</li>
4435 <li><tt>oeq</tt>: ordered and equal</li>
4436 <li><tt>ogt</tt>: ordered and greater than </li>
4437 <li><tt>oge</tt>: ordered and greater than or equal</li>
4438 <li><tt>olt</tt>: ordered and less than </li>
4439 <li><tt>ole</tt>: ordered and less than or equal</li>
4440 <li><tt>one</tt>: ordered and not equal</li>
4441 <li><tt>ord</tt>: ordered (no nans)</li>
4442 <li><tt>ueq</tt>: unordered or equal</li>
4443 <li><tt>ugt</tt>: unordered or greater than </li>
4444 <li><tt>uge</tt>: unordered or greater than or equal</li>
4445 <li><tt>ult</tt>: unordered or less than </li>
4446 <li><tt>ule</tt>: unordered or less than or equal</li>
4447 <li><tt>une</tt>: unordered or not equal</li>
4448 <li><tt>uno</tt>: unordered (either nans)</li>
4449 <li><tt>true</tt>: no comparison, always returns true</li>
4450</ol>
4451<p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4452<a href="#t_floating">floating point</a> typed. They must also be identical
4453types.</p>
4454<h5>Semantics:</h5>
Gabor Greif0f75ad02008-08-07 21:46:00 +00004455<p>The '<tt>vfcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
Nate Begemand2195702008-05-12 19:01:56 +00004456according to the condition code given as <tt>cond</tt>. The comparison yields a
4457<a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4458an identical number of elements as the values being compared, and each element
4459having identical with to the width of the floating point elements. The most
4460significant bit in each element is 1 if the element-wise comparison evaluates to
4461true, and is 0 otherwise. All other bits of the result are undefined. The
4462condition codes are evaluated identically to the
Dan Gohmanef9462f2008-10-14 16:51:45 +00004463<a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.</p>
Nate Begemand2195702008-05-12 19:01:56 +00004464
4465<h5>Example:</h5>
4466<pre>
Chris Lattner0ae02092008-10-13 16:55:18 +00004467 <i>; yields: result=&lt;2 x i32&gt; &lt; i32 0, i32 -1 &gt;</i>
4468 &lt;result&gt; = vfcmp oeq &lt;2 x float&gt; &lt; float 4, float 0 &gt;, &lt; float 5, float 0 &gt;
4469
4470 <i>; yields: result=&lt;2 x i64&gt; &lt; i64 -1, i64 0 &gt;</i>
4471 &lt;result&gt; = vfcmp ult &lt;2 x double&gt; &lt; double 1, double 2 &gt;, &lt; double 2, double 2&gt;
Nate Begemand2195702008-05-12 19:01:56 +00004472</pre>
4473</div>
4474
4475<!-- _______________________________________________________________________ -->
Chris Lattnerb54c30f2008-05-20 20:48:21 +00004476<div class="doc_subsubsection">
4477 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4478</div>
4479
Reid Spencer97c5fa42006-11-08 01:18:52 +00004480<div class="doc_text">
Chris Lattnerb54c30f2008-05-20 20:48:21 +00004481
Reid Spencer97c5fa42006-11-08 01:18:52 +00004482<h5>Syntax:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00004483
Reid Spencer97c5fa42006-11-08 01:18:52 +00004484<pre> &lt;result&gt; = phi &lt;ty&gt; [ &lt;val0&gt;, &lt;label0&gt;], ...<br></pre>
4485<h5>Overview:</h5>
4486<p>The '<tt>phi</tt>' instruction is used to implement the &#966; node in
4487the SSA graph representing the function.</p>
4488<h5>Arguments:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00004489
Jeff Cohen222a8a42007-04-29 01:07:00 +00004490<p>The type of the incoming values is specified with the first type
Reid Spencer97c5fa42006-11-08 01:18:52 +00004491field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4492as arguments, with one pair for each predecessor basic block of the
4493current block. Only values of <a href="#t_firstclass">first class</a>
4494type may be used as the value arguments to the PHI node. Only labels
4495may be used as the label arguments.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00004496
Reid Spencer97c5fa42006-11-08 01:18:52 +00004497<p>There must be no non-phi instructions between the start of a basic
4498block and the PHI instructions: i.e. PHI instructions must be first in
4499a basic block.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00004500
Reid Spencer97c5fa42006-11-08 01:18:52 +00004501<h5>Semantics:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00004502
Jeff Cohen222a8a42007-04-29 01:07:00 +00004503<p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4504specified by the pair corresponding to the predecessor basic block that executed
4505just prior to the current block.</p>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00004506
Reid Spencer97c5fa42006-11-08 01:18:52 +00004507<h5>Example:</h5>
Chris Lattnerb54c30f2008-05-20 20:48:21 +00004508<pre>
4509Loop: ; Infinite loop that counts from 0 on up...
4510 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4511 %nextindvar = add i32 %indvar, 1
4512 br label %Loop
4513</pre>
Reid Spencer97c5fa42006-11-08 01:18:52 +00004514</div>
4515
Chris Lattnerb53c28d2004-03-12 05:50:16 +00004516<!-- _______________________________________________________________________ -->
4517<div class="doc_subsubsection">
4518 <a name="i_select">'<tt>select</tt>' Instruction</a>
4519</div>
4520
4521<div class="doc_text">
4522
4523<h5>Syntax:</h5>
4524
4525<pre>
Dan Gohmanc579d972008-09-09 01:02:47 +00004526 &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>
4527
Dan Gohmanef9462f2008-10-14 16:51:45 +00004528 <i>selty</i> is either i1 or {&lt;N x i1&gt;}
Chris Lattnerb53c28d2004-03-12 05:50:16 +00004529</pre>
4530
4531<h5>Overview:</h5>
4532
4533<p>
4534The '<tt>select</tt>' instruction is used to choose one value based on a
4535condition, without branching.
4536</p>
4537
4538
4539<h5>Arguments:</h5>
4540
4541<p>
Dan Gohmanc579d972008-09-09 01:02:47 +00004542The '<tt>select</tt>' instruction requires an 'i1' value or
4543a vector of 'i1' values indicating the
Chris Lattnerb54c30f2008-05-20 20:48:21 +00004544condition, and two values of the same <a href="#t_firstclass">first class</a>
Dan Gohmanc579d972008-09-09 01:02:47 +00004545type. If the val1/val2 are vectors and
4546the condition is a scalar, then entire vectors are selected, not
Chris Lattnerb54c30f2008-05-20 20:48:21 +00004547individual elements.
Chris Lattnerb53c28d2004-03-12 05:50:16 +00004548</p>
4549
4550<h5>Semantics:</h5>
4551
4552<p>
Dan Gohmanc579d972008-09-09 01:02:47 +00004553If the condition is an i1 and it evaluates to 1, the instruction returns the first
John Criswell88190562005-05-16 16:17:45 +00004554value argument; otherwise, it returns the second value argument.
Chris Lattnerb53c28d2004-03-12 05:50:16 +00004555</p>
Dan Gohmanc579d972008-09-09 01:02:47 +00004556<p>
4557If the condition is a vector of i1, then the value arguments must
4558be vectors of the same size, and the selection is done element
4559by element.
4560</p>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00004561
4562<h5>Example:</h5>
4563
4564<pre>
Reid Spencer36a15422007-01-12 03:35:51 +00004565 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
Chris Lattnerb53c28d2004-03-12 05:50:16 +00004566</pre>
Dan Gohmana5127ab2009-01-22 01:39:38 +00004567
4568<p>Note that the code generator does not yet support conditions
4569 with vector type.</p>
4570
Chris Lattnerb53c28d2004-03-12 05:50:16 +00004571</div>
4572
Robert Bocchinof72fdfe2006-01-15 20:48:27 +00004573
4574<!-- _______________________________________________________________________ -->
4575<div class="doc_subsubsection">
Chris Lattnere23c1392005-05-06 05:47:36 +00004576 <a name="i_call">'<tt>call</tt>' Instruction</a>
4577</div>
4578
Misha Brukman76307852003-11-08 01:05:38 +00004579<div class="doc_text">
Chris Lattnere23c1392005-05-06 05:47:36 +00004580
Chris Lattner2f7c9632001-06-06 20:29:01 +00004581<h5>Syntax:</h5>
Chris Lattnere23c1392005-05-06 05:47:36 +00004582<pre>
Devang Patel02256232008-10-07 17:48:33 +00004583 &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 +00004584</pre>
4585
Chris Lattner2f7c9632001-06-06 20:29:01 +00004586<h5>Overview:</h5>
Chris Lattnere23c1392005-05-06 05:47:36 +00004587
Misha Brukman76307852003-11-08 01:05:38 +00004588<p>The '<tt>call</tt>' instruction represents a simple function call.</p>
Chris Lattnere23c1392005-05-06 05:47:36 +00004589
Chris Lattner2f7c9632001-06-06 20:29:01 +00004590<h5>Arguments:</h5>
Chris Lattnere23c1392005-05-06 05:47:36 +00004591
Misha Brukman76307852003-11-08 01:05:38 +00004592<p>This instruction requires several arguments:</p>
Chris Lattnere23c1392005-05-06 05:47:36 +00004593
Chris Lattnera8292f32002-05-06 22:08:29 +00004594<ol>
Chris Lattner48b383b02003-11-25 01:02:51 +00004595 <li>
Chris Lattner0132aff2005-05-06 22:57:40 +00004596 <p>The optional "tail" marker indicates whether the callee function accesses
4597 any allocas or varargs in the caller. If the "tail" marker is present, the
Chris Lattnere23c1392005-05-06 05:47:36 +00004598 function call is eligible for tail call optimization. Note that calls may
4599 be marked "tail" even if they do not occur before a <a
Dan Gohmanef9462f2008-10-14 16:51:45 +00004600 href="#i_ret"><tt>ret</tt></a> instruction.</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00004601 </li>
4602 <li>
Duncan Sands16f122e2007-03-30 12:22:09 +00004603 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
Chris Lattner0132aff2005-05-06 22:57:40 +00004604 convention</a> the call should use. If none is specified, the call defaults
Dan Gohmanef9462f2008-10-14 16:51:45 +00004605 to using C calling conventions.</p>
Chris Lattner0132aff2005-05-06 22:57:40 +00004606 </li>
Devang Patel7e9b05e2008-10-06 18:50:38 +00004607
4608 <li>
4609 <p>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4610 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
4611 and '<tt>inreg</tt>' attributes are valid here.</p>
4612 </li>
4613
Chris Lattner0132aff2005-05-06 22:57:40 +00004614 <li>
Nick Lewyckya9b13d52007-09-08 13:57:50 +00004615 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4616 the type of the return value. Functions that return no value are marked
4617 <tt><a href="#t_void">void</a></tt>.</p>
4618 </li>
4619 <li>
4620 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4621 value being invoked. The argument types must match the types implied by
4622 this signature. This type can be omitted if the function is not varargs
4623 and if the function type does not return a pointer to a function.</p>
Chris Lattnere23c1392005-05-06 05:47:36 +00004624 </li>
4625 <li>
4626 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4627 be invoked. In most cases, this is a direct function invocation, but
4628 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
John Criswell88190562005-05-16 16:17:45 +00004629 to function value.</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00004630 </li>
4631 <li>
4632 <p>'<tt>function args</tt>': argument list whose types match the
Reid Spencerd845d162005-05-01 22:22:57 +00004633 function signature argument types. All arguments must be of
4634 <a href="#t_firstclass">first class</a> type. If the function signature
4635 indicates the function accepts a variable number of arguments, the extra
4636 arguments can be specified.</p>
Chris Lattner48b383b02003-11-25 01:02:51 +00004637 </li>
Devang Patel7e9b05e2008-10-06 18:50:38 +00004638 <li>
Devang Patel02256232008-10-07 17:48:33 +00004639 <p>The optional <a href="#fnattrs">function attributes</a> list. Only
Devang Patel7e9b05e2008-10-06 18:50:38 +00004640 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
4641 '<tt>readnone</tt>' attributes are valid here.</p>
4642 </li>
Chris Lattnera8292f32002-05-06 22:08:29 +00004643</ol>
Chris Lattnere23c1392005-05-06 05:47:36 +00004644
Chris Lattner2f7c9632001-06-06 20:29:01 +00004645<h5>Semantics:</h5>
Chris Lattnere23c1392005-05-06 05:47:36 +00004646
Chris Lattner48b383b02003-11-25 01:02:51 +00004647<p>The '<tt>call</tt>' instruction is used to cause control flow to
4648transfer to a specified function, with its incoming arguments bound to
4649the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4650instruction in the called function, control flow continues with the
4651instruction after the function call, and the return value of the
Dan Gohmanef9462f2008-10-14 16:51:45 +00004652function is bound to the result argument.</p>
Chris Lattnere23c1392005-05-06 05:47:36 +00004653
Chris Lattner2f7c9632001-06-06 20:29:01 +00004654<h5>Example:</h5>
Chris Lattnere23c1392005-05-06 05:47:36 +00004655
4656<pre>
Nick Lewyckya9b13d52007-09-08 13:57:50 +00004657 %retval = call i32 @test(i32 %argc)
Chris Lattnerfb7c88d2008-03-21 17:24:17 +00004658 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4659 %X = tail call i32 @foo() <i>; yields i32</i>
4660 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4661 call void %foo(i8 97 signext)
Devang Pateld6cff512008-03-10 20:49:15 +00004662
4663 %struct.A = type { i32, i8 }
Devang Patel7e9b05e2008-10-06 18:50:38 +00004664 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
Dan Gohmancc3132e2008-10-04 19:00:07 +00004665 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4666 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
Chris Lattner6cbe8e92008-10-08 06:26:11 +00004667 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
Matthijs Kooijmaneefa7df2008-10-07 10:03:45 +00004668 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
Chris Lattnere23c1392005-05-06 05:47:36 +00004669</pre>
4670
Misha Brukman76307852003-11-08 01:05:38 +00004671</div>
Chris Lattner6a4a0492004-09-27 21:51:25 +00004672
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004673<!-- _______________________________________________________________________ -->
Chris Lattner6a4a0492004-09-27 21:51:25 +00004674<div class="doc_subsubsection">
Chris Lattner33337472006-01-13 23:26:01 +00004675 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
Chris Lattner6a4a0492004-09-27 21:51:25 +00004676</div>
4677
Misha Brukman76307852003-11-08 01:05:38 +00004678<div class="doc_text">
Chris Lattner6a4a0492004-09-27 21:51:25 +00004679
Chris Lattner26ca62e2003-10-18 05:51:36 +00004680<h5>Syntax:</h5>
Chris Lattner6a4a0492004-09-27 21:51:25 +00004681
4682<pre>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00004683 &lt;resultval&gt; = va_arg &lt;va_list*&gt; &lt;arglist&gt;, &lt;argty&gt;
Chris Lattner6a4a0492004-09-27 21:51:25 +00004684</pre>
4685
Chris Lattner26ca62e2003-10-18 05:51:36 +00004686<h5>Overview:</h5>
Chris Lattner6a4a0492004-09-27 21:51:25 +00004687
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00004688<p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
Chris Lattner6a4a0492004-09-27 21:51:25 +00004689the "variable argument" area of a function call. It is used to implement the
4690<tt>va_arg</tt> macro in C.</p>
4691
Chris Lattner26ca62e2003-10-18 05:51:36 +00004692<h5>Arguments:</h5>
Chris Lattner6a4a0492004-09-27 21:51:25 +00004693
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00004694<p>This instruction takes a <tt>va_list*</tt> value and the type of
4695the argument. It returns a value of the specified argument type and
Jeff Cohen222a8a42007-04-29 01:07:00 +00004696increments the <tt>va_list</tt> to point to the next argument. The
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00004697actual type of <tt>va_list</tt> is target specific.</p>
Chris Lattner6a4a0492004-09-27 21:51:25 +00004698
Chris Lattner26ca62e2003-10-18 05:51:36 +00004699<h5>Semantics:</h5>
Chris Lattner6a4a0492004-09-27 21:51:25 +00004700
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00004701<p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4702type from the specified <tt>va_list</tt> and causes the
4703<tt>va_list</tt> to point to the next argument. For more information,
4704see the variable argument handling <a href="#int_varargs">Intrinsic
4705Functions</a>.</p>
Chris Lattner6a4a0492004-09-27 21:51:25 +00004706
4707<p>It is legal for this instruction to be called in a function which does not
4708take a variable number of arguments, for example, the <tt>vfprintf</tt>
Misha Brukman76307852003-11-08 01:05:38 +00004709function.</p>
Chris Lattner6a4a0492004-09-27 21:51:25 +00004710
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00004711<p><tt>va_arg</tt> is an LLVM instruction instead of an <a
John Criswell88190562005-05-16 16:17:45 +00004712href="#intrinsics">intrinsic function</a> because it takes a type as an
Chris Lattner6a4a0492004-09-27 21:51:25 +00004713argument.</p>
4714
Chris Lattner26ca62e2003-10-18 05:51:36 +00004715<h5>Example:</h5>
Chris Lattner6a4a0492004-09-27 21:51:25 +00004716
4717<p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4718
Dan Gohman3065b612009-01-12 23:12:39 +00004719<p>Note that the code generator does not yet fully support va_arg
4720 on many targets. Also, it does not currently support va_arg with
4721 aggregate types on any target.</p>
4722
Misha Brukman76307852003-11-08 01:05:38 +00004723</div>
Chris Lattner941515c2004-01-06 05:31:32 +00004724
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004725<!-- *********************************************************************** -->
Chris Lattner48b383b02003-11-25 01:02:51 +00004726<div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4727<!-- *********************************************************************** -->
Chris Lattner941515c2004-01-06 05:31:32 +00004728
Misha Brukman76307852003-11-08 01:05:38 +00004729<div class="doc_text">
Chris Lattnerfee11462004-02-12 17:01:32 +00004730
4731<p>LLVM supports the notion of an "intrinsic function". These functions have
Reid Spencer4eefaab2007-04-01 08:04:23 +00004732well known names and semantics and are required to follow certain restrictions.
4733Overall, these intrinsics represent an extension mechanism for the LLVM
Jeff Cohen222a8a42007-04-29 01:07:00 +00004734language that does not require changing all of the transformations in LLVM when
Gabor Greifa54634a2007-07-06 22:07:22 +00004735adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
Chris Lattnerfee11462004-02-12 17:01:32 +00004736
John Criswell88190562005-05-16 16:17:45 +00004737<p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
Jeff Cohen222a8a42007-04-29 01:07:00 +00004738prefix is reserved in LLVM for intrinsic names; thus, function names may not
4739begin with this prefix. Intrinsic functions must always be external functions:
4740you cannot define the body of intrinsic functions. Intrinsic functions may
4741only be used in call or invoke instructions: it is illegal to take the address
4742of an intrinsic function. Additionally, because intrinsic functions are part
4743of the LLVM language, it is required if any are added that they be documented
4744here.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +00004745
Chandler Carruth7132e002007-08-04 01:51:18 +00004746<p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4747a family of functions that perform the same operation but on different data
4748types. Because LLVM can represent over 8 million different integer types,
4749overloading is used commonly to allow an intrinsic function to operate on any
4750integer type. One or more of the argument types or the result type can be
4751overloaded to accept any integer type. Argument types may also be defined as
4752exactly matching a previous argument's type or the result type. This allows an
4753intrinsic function which accepts multiple arguments, but needs all of them to
4754be of the same type, to only be overloaded with respect to a single argument or
4755the result.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +00004756
Chandler Carruth7132e002007-08-04 01:51:18 +00004757<p>Overloaded intrinsics will have the names of its overloaded argument types
4758encoded into its function name, each preceded by a period. Only those types
4759which are overloaded result in a name suffix. Arguments whose type is matched
4760against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4761take an integer of any width and returns an integer of exactly the same integer
4762width. This leads to a family of functions such as
4763<tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4764Only one type, the return type, is overloaded, and only one type suffix is
4765required. Because the argument's type is matched against the return type, it
4766does not require its own name suffix.</p>
Reid Spencer4eefaab2007-04-01 08:04:23 +00004767
4768<p>To learn how to add an intrinsic function, please see the
4769<a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
Chris Lattnerfee11462004-02-12 17:01:32 +00004770</p>
4771
Misha Brukman76307852003-11-08 01:05:38 +00004772</div>
Chris Lattner941515c2004-01-06 05:31:32 +00004773
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004774<!-- ======================================================================= -->
Chris Lattner941515c2004-01-06 05:31:32 +00004775<div class="doc_subsection">
4776 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4777</div>
4778
Misha Brukman76307852003-11-08 01:05:38 +00004779<div class="doc_text">
Chris Lattner757528b0b2004-05-23 21:06:01 +00004780
Misha Brukman76307852003-11-08 01:05:38 +00004781<p>Variable argument support is defined in LLVM with the <a
Chris Lattner33337472006-01-13 23:26:01 +00004782 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
Chris Lattner48b383b02003-11-25 01:02:51 +00004783intrinsic functions. These functions are related to the similarly
4784named macros defined in the <tt>&lt;stdarg.h&gt;</tt> header file.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00004785
Chris Lattner48b383b02003-11-25 01:02:51 +00004786<p>All of these functions operate on arguments that use a
4787target-specific value type "<tt>va_list</tt>". The LLVM assembly
4788language reference manual does not define what this type is, so all
Jeff Cohen222a8a42007-04-29 01:07:00 +00004789transformations should be prepared to handle these functions regardless of
4790the type used.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00004791
Chris Lattner30b868d2006-05-15 17:26:46 +00004792<p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
Chris Lattner48b383b02003-11-25 01:02:51 +00004793instruction and the variable argument handling intrinsic functions are
4794used.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00004795
Bill Wendling3716c5d2007-05-29 09:04:49 +00004796<div class="doc_code">
Chris Lattnerfee11462004-02-12 17:01:32 +00004797<pre>
Anton Korobeynikov640bbe02007-03-21 23:58:04 +00004798define i32 @test(i32 %X, ...) {
Chris Lattnerfee11462004-02-12 17:01:32 +00004799 ; Initialize variable argument processing
Jeff Cohen222a8a42007-04-29 01:07:00 +00004800 %ap = alloca i8*
Chris Lattnerdb0790c2007-01-08 07:55:15 +00004801 %ap2 = bitcast i8** %ap to i8*
Anton Korobeynikov640bbe02007-03-21 23:58:04 +00004802 call void @llvm.va_start(i8* %ap2)
Chris Lattnerfee11462004-02-12 17:01:32 +00004803
4804 ; Read a single integer argument
Jeff Cohen222a8a42007-04-29 01:07:00 +00004805 %tmp = va_arg i8** %ap, i32
Chris Lattnerfee11462004-02-12 17:01:32 +00004806
4807 ; Demonstrate usage of llvm.va_copy and llvm.va_end
Jeff Cohen222a8a42007-04-29 01:07:00 +00004808 %aq = alloca i8*
Chris Lattnerdb0790c2007-01-08 07:55:15 +00004809 %aq2 = bitcast i8** %aq to i8*
Jeff Cohen222a8a42007-04-29 01:07:00 +00004810 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
Anton Korobeynikov640bbe02007-03-21 23:58:04 +00004811 call void @llvm.va_end(i8* %aq2)
Chris Lattnerfee11462004-02-12 17:01:32 +00004812
4813 ; Stop processing of arguments.
Anton Korobeynikov640bbe02007-03-21 23:58:04 +00004814 call void @llvm.va_end(i8* %ap2)
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00004815 ret i32 %tmp
Chris Lattnerfee11462004-02-12 17:01:32 +00004816}
Anton Korobeynikov640bbe02007-03-21 23:58:04 +00004817
4818declare void @llvm.va_start(i8*)
4819declare void @llvm.va_copy(i8*, i8*)
4820declare void @llvm.va_end(i8*)
Chris Lattnerfee11462004-02-12 17:01:32 +00004821</pre>
Misha Brukman76307852003-11-08 01:05:38 +00004822</div>
Chris Lattner941515c2004-01-06 05:31:32 +00004823
Bill Wendling3716c5d2007-05-29 09:04:49 +00004824</div>
4825
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004826<!-- _______________________________________________________________________ -->
Chris Lattner941515c2004-01-06 05:31:32 +00004827<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004828 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
Chris Lattner941515c2004-01-06 05:31:32 +00004829</div>
4830
4831
Misha Brukman76307852003-11-08 01:05:38 +00004832<div class="doc_text">
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004833<h5>Syntax:</h5>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00004834<pre> declare void %llvm.va_start(i8* &lt;arglist&gt;)<br></pre>
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004835<h5>Overview:</h5>
Dan Gohmanef9462f2008-10-14 16:51:45 +00004836<p>The '<tt>llvm.va_start</tt>' intrinsic initializes
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00004837<tt>*&lt;arglist&gt;</tt> for subsequent use by <tt><a
4838href="#i_va_arg">va_arg</a></tt>.</p>
4839
4840<h5>Arguments:</h5>
4841
Dan Gohmanef9462f2008-10-14 16:51:45 +00004842<p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00004843
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004844<h5>Semantics:</h5>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00004845
Dan Gohmanef9462f2008-10-14 16:51:45 +00004846<p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00004847macro available in C. In a target-dependent way, it initializes the
Jeff Cohen222a8a42007-04-29 01:07:00 +00004848<tt>va_list</tt> element to which the argument points, so that the next call to
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00004849<tt>va_arg</tt> will produce the first variable argument passed to the function.
4850Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
Jeff Cohen222a8a42007-04-29 01:07:00 +00004851last argument of the function as the compiler can figure that out.</p>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00004852
Misha Brukman76307852003-11-08 01:05:38 +00004853</div>
Chris Lattner941515c2004-01-06 05:31:32 +00004854
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004855<!-- _______________________________________________________________________ -->
Chris Lattner941515c2004-01-06 05:31:32 +00004856<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004857 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
Chris Lattner941515c2004-01-06 05:31:32 +00004858</div>
4859
Misha Brukman76307852003-11-08 01:05:38 +00004860<div class="doc_text">
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004861<h5>Syntax:</h5>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004862<pre> declare void @llvm.va_end(i8* &lt;arglist&gt;)<br></pre>
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004863<h5>Overview:</h5>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00004864
Jeff Cohen222a8a42007-04-29 01:07:00 +00004865<p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*&lt;arglist&gt;</tt>,
Reid Spencer96a5f022007-04-04 02:42:35 +00004866which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
Chris Lattner48b383b02003-11-25 01:02:51 +00004867or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00004868
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004869<h5>Arguments:</h5>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00004870
Jeff Cohen222a8a42007-04-29 01:07:00 +00004871<p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00004872
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004873<h5>Semantics:</h5>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00004874
Misha Brukman76307852003-11-08 01:05:38 +00004875<p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
Jeff Cohen222a8a42007-04-29 01:07:00 +00004876macro available in C. In a target-dependent way, it destroys the
4877<tt>va_list</tt> element to which the argument points. Calls to <a
4878href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4879<tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4880<tt>llvm.va_end</tt>.</p>
Chris Lattnerdb0790c2007-01-08 07:55:15 +00004881
Misha Brukman76307852003-11-08 01:05:38 +00004882</div>
Chris Lattner941515c2004-01-06 05:31:32 +00004883
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004884<!-- _______________________________________________________________________ -->
Chris Lattner941515c2004-01-06 05:31:32 +00004885<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004886 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
Chris Lattner941515c2004-01-06 05:31:32 +00004887</div>
4888
Misha Brukman76307852003-11-08 01:05:38 +00004889<div class="doc_text">
Chris Lattner757528b0b2004-05-23 21:06:01 +00004890
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004891<h5>Syntax:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00004892
4893<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00004894 declare void @llvm.va_copy(i8* &lt;destarglist&gt;, i8* &lt;srcarglist&gt;)
Chris Lattner757528b0b2004-05-23 21:06:01 +00004895</pre>
4896
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004897<h5>Overview:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00004898
Jeff Cohen222a8a42007-04-29 01:07:00 +00004899<p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4900from the source argument list to the destination argument list.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00004901
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004902<h5>Arguments:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00004903
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00004904<p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
Andrew Lenharth5305ea52005-06-22 20:38:11 +00004905The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
Andrew Lenharth5fb787c2005-06-18 18:28:17 +00004906
Chris Lattner757528b0b2004-05-23 21:06:01 +00004907
Chris Lattnerbd64b4e2003-05-08 04:57:36 +00004908<h5>Semantics:</h5>
Chris Lattner757528b0b2004-05-23 21:06:01 +00004909
Jeff Cohen222a8a42007-04-29 01:07:00 +00004910<p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4911macro available in C. In a target-dependent way, it copies the source
4912<tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4913intrinsic is necessary because the <tt><a href="#int_va_start">
4914llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4915example, memory allocation.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00004916
Misha Brukman76307852003-11-08 01:05:38 +00004917</div>
Chris Lattner941515c2004-01-06 05:31:32 +00004918
Chris Lattnerfee11462004-02-12 17:01:32 +00004919<!-- ======================================================================= -->
4920<div class="doc_subsection">
Chris Lattner757528b0b2004-05-23 21:06:01 +00004921 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4922</div>
4923
4924<div class="doc_text">
4925
4926<p>
4927LLVM support for <a href="GarbageCollection.html">Accurate Garbage
Chris Lattner67c37d12008-08-05 18:29:16 +00004928Collection</a> (GC) requires the implementation and generation of these
4929intrinsics.
Reid Spencer96a5f022007-04-04 02:42:35 +00004930These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
Chris Lattner757528b0b2004-05-23 21:06:01 +00004931stack</a>, as well as garbage collector implementations that require <a
Reid Spencer96a5f022007-04-04 02:42:35 +00004932href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
Chris Lattner757528b0b2004-05-23 21:06:01 +00004933Front-ends for type-safe garbage collected languages should generate these
4934intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4935href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4936</p>
Christopher Lamb55c6d4f2007-12-17 01:00:21 +00004937
4938<p>The garbage collection intrinsics only operate on objects in the generic
4939 address space (address space zero).</p>
4940
Chris Lattner757528b0b2004-05-23 21:06:01 +00004941</div>
4942
4943<!-- _______________________________________________________________________ -->
4944<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004945 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
Chris Lattner757528b0b2004-05-23 21:06:01 +00004946</div>
4947
4948<div class="doc_text">
4949
4950<h5>Syntax:</h5>
4951
4952<pre>
Chris Lattner12477732007-09-21 17:30:40 +00004953 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
Chris Lattner757528b0b2004-05-23 21:06:01 +00004954</pre>
4955
4956<h5>Overview:</h5>
4957
John Criswelldfe6a862004-12-10 15:51:16 +00004958<p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
Chris Lattner757528b0b2004-05-23 21:06:01 +00004959the code generator, and allows some metadata to be associated with it.</p>
4960
4961<h5>Arguments:</h5>
4962
4963<p>The first argument specifies the address of a stack object that contains the
4964root pointer. The second pointer (which must be either a constant or a global
4965value address) contains the meta-data to be associated with the root.</p>
4966
4967<h5>Semantics:</h5>
4968
Chris Lattner851b7712008-04-24 05:59:56 +00004969<p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
Chris Lattner757528b0b2004-05-23 21:06:01 +00004970location. At compile-time, the code generator generates information to allow
Gordon Henriksenfb56bde2007-12-25 02:31:26 +00004971the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4972intrinsic may only be used in a function which <a href="#gc">specifies a GC
4973algorithm</a>.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00004974
4975</div>
4976
4977
4978<!-- _______________________________________________________________________ -->
4979<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00004980 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
Chris Lattner757528b0b2004-05-23 21:06:01 +00004981</div>
4982
4983<div class="doc_text">
4984
4985<h5>Syntax:</h5>
4986
4987<pre>
Chris Lattner12477732007-09-21 17:30:40 +00004988 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
Chris Lattner757528b0b2004-05-23 21:06:01 +00004989</pre>
4990
4991<h5>Overview:</h5>
4992
4993<p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4994locations, allowing garbage collector implementations that require read
4995barriers.</p>
4996
4997<h5>Arguments:</h5>
4998
Chris Lattnerf9228072006-03-14 20:02:51 +00004999<p>The second argument is the address to read from, which should be an address
5000allocated from the garbage collector. The first object is a pointer to the
5001start of the referenced object, if needed by the language runtime (otherwise
5002null).</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005003
5004<h5>Semantics:</h5>
5005
5006<p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
5007instruction, but may be replaced with substantially more complex code by the
Gordon Henriksenfb56bde2007-12-25 02:31:26 +00005008garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5009may only be used in a function which <a href="#gc">specifies a GC
5010algorithm</a>.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005011
5012</div>
5013
5014
5015<!-- _______________________________________________________________________ -->
5016<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005017 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005018</div>
5019
5020<div class="doc_text">
5021
5022<h5>Syntax:</h5>
5023
5024<pre>
Chris Lattner12477732007-09-21 17:30:40 +00005025 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
Chris Lattner757528b0b2004-05-23 21:06:01 +00005026</pre>
5027
5028<h5>Overview:</h5>
5029
5030<p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
5031locations, allowing garbage collector implementations that require write
5032barriers (such as generational or reference counting collectors).</p>
5033
5034<h5>Arguments:</h5>
5035
Chris Lattnerf9228072006-03-14 20:02:51 +00005036<p>The first argument is the reference to store, the second is the start of the
5037object to store it to, and the third is the address of the field of Obj to
5038store to. If the runtime does not require a pointer to the object, Obj may be
5039null.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005040
5041<h5>Semantics:</h5>
5042
5043<p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
5044instruction, but may be replaced with substantially more complex code by the
Gordon Henriksenfb56bde2007-12-25 02:31:26 +00005045garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5046may only be used in a function which <a href="#gc">specifies a GC
5047algorithm</a>.</p>
Chris Lattner757528b0b2004-05-23 21:06:01 +00005048
5049</div>
5050
5051
5052
5053<!-- ======================================================================= -->
5054<div class="doc_subsection">
Chris Lattner3649c3a2004-02-14 04:08:35 +00005055 <a name="int_codegen">Code Generator Intrinsics</a>
5056</div>
5057
5058<div class="doc_text">
5059<p>
5060These intrinsics are provided by LLVM to expose special features that may only
5061be implemented with code generator support.
5062</p>
5063
5064</div>
5065
5066<!-- _______________________________________________________________________ -->
5067<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005068 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
Chris Lattner3649c3a2004-02-14 04:08:35 +00005069</div>
5070
5071<div class="doc_text">
5072
5073<h5>Syntax:</h5>
5074<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00005075 declare i8 *@llvm.returnaddress(i32 &lt;level&gt;)
Chris Lattner3649c3a2004-02-14 04:08:35 +00005076</pre>
5077
5078<h5>Overview:</h5>
5079
5080<p>
Chris Lattnerc1fb4262006-10-15 20:05:59 +00005081The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5082target-specific value indicating the return address of the current function
5083or one of its callers.
Chris Lattner3649c3a2004-02-14 04:08:35 +00005084</p>
5085
5086<h5>Arguments:</h5>
5087
5088<p>
5089The argument to this intrinsic indicates which function to return the address
5090for. Zero indicates the calling function, one indicates its caller, etc. The
5091argument is <b>required</b> to be a constant integer value.
5092</p>
5093
5094<h5>Semantics:</h5>
5095
5096<p>
5097The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
5098the return address of the specified call frame, or zero if it cannot be
5099identified. The value returned by this intrinsic is likely to be incorrect or 0
5100for arguments other than zero, so it should only be used for debugging purposes.
5101</p>
5102
5103<p>
5104Note that calling this intrinsic does not prevent function inlining or other
Chris Lattner2e6eb5f2005-03-07 20:30:51 +00005105aggressive transformations, so the value returned may not be that of the obvious
Chris Lattner3649c3a2004-02-14 04:08:35 +00005106source-language caller.
5107</p>
5108</div>
5109
5110
5111<!-- _______________________________________________________________________ -->
5112<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005113 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
Chris Lattner3649c3a2004-02-14 04:08:35 +00005114</div>
5115
5116<div class="doc_text">
5117
5118<h5>Syntax:</h5>
5119<pre>
Chris Lattner12477732007-09-21 17:30:40 +00005120 declare i8 *@llvm.frameaddress(i32 &lt;level&gt;)
Chris Lattner3649c3a2004-02-14 04:08:35 +00005121</pre>
5122
5123<h5>Overview:</h5>
5124
5125<p>
Chris Lattnerc1fb4262006-10-15 20:05:59 +00005126The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5127target-specific frame pointer value for the specified stack frame.
Chris Lattner3649c3a2004-02-14 04:08:35 +00005128</p>
5129
5130<h5>Arguments:</h5>
5131
5132<p>
5133The argument to this intrinsic indicates which function to return the frame
5134pointer for. Zero indicates the calling function, one indicates its caller,
5135etc. The argument is <b>required</b> to be a constant integer value.
5136</p>
5137
5138<h5>Semantics:</h5>
5139
5140<p>
5141The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
5142the frame address of the specified call frame, or zero if it cannot be
5143identified. The value returned by this intrinsic is likely to be incorrect or 0
5144for arguments other than zero, so it should only be used for debugging purposes.
5145</p>
5146
5147<p>
5148Note that calling this intrinsic does not prevent function inlining or other
Chris Lattner2e6eb5f2005-03-07 20:30:51 +00005149aggressive transformations, so the value returned may not be that of the obvious
Chris Lattner3649c3a2004-02-14 04:08:35 +00005150source-language caller.
5151</p>
5152</div>
5153
Chris Lattnerc8a2c222005-02-28 19:24:19 +00005154<!-- _______________________________________________________________________ -->
5155<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005156 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
Chris Lattner2f0f0012006-01-13 02:03:13 +00005157</div>
5158
5159<div class="doc_text">
5160
5161<h5>Syntax:</h5>
5162<pre>
Chris Lattner12477732007-09-21 17:30:40 +00005163 declare i8 *@llvm.stacksave()
Chris Lattner2f0f0012006-01-13 02:03:13 +00005164</pre>
5165
5166<h5>Overview:</h5>
5167
5168<p>
5169The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
Reid Spencer96a5f022007-04-04 02:42:35 +00005170the function stack, for use with <a href="#int_stackrestore">
Chris Lattner2f0f0012006-01-13 02:03:13 +00005171<tt>llvm.stackrestore</tt></a>. This is useful for implementing language
5172features like scoped automatic variable sized arrays in C99.
5173</p>
5174
5175<h5>Semantics:</h5>
5176
5177<p>
5178This intrinsic returns a opaque pointer value that can be passed to <a
Reid Spencer96a5f022007-04-04 02:42:35 +00005179href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
Chris Lattner2f0f0012006-01-13 02:03:13 +00005180<tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
5181<tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
5182state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
5183practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
5184that were allocated after the <tt>llvm.stacksave</tt> was executed.
5185</p>
5186
5187</div>
5188
5189<!-- _______________________________________________________________________ -->
5190<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005191 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
Chris Lattner2f0f0012006-01-13 02:03:13 +00005192</div>
5193
5194<div class="doc_text">
5195
5196<h5>Syntax:</h5>
5197<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00005198 declare void @llvm.stackrestore(i8 * %ptr)
Chris Lattner2f0f0012006-01-13 02:03:13 +00005199</pre>
5200
5201<h5>Overview:</h5>
5202
5203<p>
5204The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5205the function stack to the state it was in when the corresponding <a
Reid Spencer96a5f022007-04-04 02:42:35 +00005206href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
Chris Lattner2f0f0012006-01-13 02:03:13 +00005207useful for implementing language features like scoped automatic variable sized
5208arrays in C99.
5209</p>
5210
5211<h5>Semantics:</h5>
5212
5213<p>
Reid Spencer96a5f022007-04-04 02:42:35 +00005214See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
Chris Lattner2f0f0012006-01-13 02:03:13 +00005215</p>
5216
5217</div>
5218
5219
5220<!-- _______________________________________________________________________ -->
5221<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005222 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
Chris Lattnerc8a2c222005-02-28 19:24:19 +00005223</div>
5224
5225<div class="doc_text">
5226
5227<h5>Syntax:</h5>
5228<pre>
Chris Lattner12477732007-09-21 17:30:40 +00005229 declare void @llvm.prefetch(i8* &lt;address&gt;, i32 &lt;rw&gt;, i32 &lt;locality&gt;)
Chris Lattnerc8a2c222005-02-28 19:24:19 +00005230</pre>
5231
5232<h5>Overview:</h5>
5233
5234
5235<p>
5236The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
John Criswell88190562005-05-16 16:17:45 +00005237a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
5238no
5239effect on the behavior of the program but can change its performance
Chris Lattnerff851072005-02-28 19:47:14 +00005240characteristics.
Chris Lattnerc8a2c222005-02-28 19:24:19 +00005241</p>
5242
5243<h5>Arguments:</h5>
5244
5245<p>
5246<tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
5247determining if the fetch should be for a read (0) or write (1), and
5248<tt>locality</tt> is a temporal locality specifier ranging from (0) - no
Chris Lattnerd3e641c2005-03-07 20:31:38 +00005249locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
Chris Lattnerc8a2c222005-02-28 19:24:19 +00005250<tt>locality</tt> arguments must be constant integers.
5251</p>
5252
5253<h5>Semantics:</h5>
5254
5255<p>
5256This intrinsic does not modify the behavior of the program. In particular,
5257prefetches cannot trap and do not produce a value. On targets that support this
5258intrinsic, the prefetch can provide hints to the processor cache for better
5259performance.
5260</p>
5261
5262</div>
5263
Andrew Lenharthb4427912005-03-28 20:05:49 +00005264<!-- _______________________________________________________________________ -->
5265<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005266 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
Andrew Lenharthb4427912005-03-28 20:05:49 +00005267</div>
5268
5269<div class="doc_text">
5270
5271<h5>Syntax:</h5>
5272<pre>
Chris Lattner12477732007-09-21 17:30:40 +00005273 declare void @llvm.pcmarker(i32 &lt;id&gt;)
Andrew Lenharthb4427912005-03-28 20:05:49 +00005274</pre>
5275
5276<h5>Overview:</h5>
5277
5278
5279<p>
John Criswell88190562005-05-16 16:17:45 +00005280The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
Chris Lattner67c37d12008-08-05 18:29:16 +00005281(PC) in a region of
5282code to simulators and other tools. The method is target specific, but it is
5283expected that the marker will use exported symbols to transmit the PC of the
5284marker.
5285The marker makes no guarantees that it will remain with any specific instruction
5286after optimizations. It is possible that the presence of a marker will inhibit
Chris Lattnerb40261e2006-03-24 07:16:10 +00005287optimizations. The intended use is to be inserted after optimizations to allow
John Criswell88190562005-05-16 16:17:45 +00005288correlations of simulation runs.
Andrew Lenharthb4427912005-03-28 20:05:49 +00005289</p>
5290
5291<h5>Arguments:</h5>
5292
5293<p>
5294<tt>id</tt> is a numerical id identifying the marker.
5295</p>
5296
5297<h5>Semantics:</h5>
5298
5299<p>
5300This intrinsic does not modify the behavior of the program. Backends that do not
5301support this intrinisic may ignore it.
5302</p>
5303
5304</div>
5305
Andrew Lenharth01aa5632005-11-11 16:47:30 +00005306<!-- _______________________________________________________________________ -->
5307<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005308 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
Andrew Lenharth01aa5632005-11-11 16:47:30 +00005309</div>
5310
5311<div class="doc_text">
5312
5313<h5>Syntax:</h5>
5314<pre>
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00005315 declare i64 @llvm.readcyclecounter( )
Andrew Lenharth01aa5632005-11-11 16:47:30 +00005316</pre>
5317
5318<h5>Overview:</h5>
5319
5320
5321<p>
5322The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5323counter register (or similar low latency, high accuracy clocks) on those targets
5324that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
5325As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
5326should only be used for small timings.
5327</p>
5328
5329<h5>Semantics:</h5>
5330
5331<p>
5332When directly supported, reading the cycle counter should not modify any memory.
5333Implementations are allowed to either return a application specific value or a
5334system wide value. On backends without support, this is lowered to a constant 0.
5335</p>
5336
5337</div>
5338
Chris Lattner3649c3a2004-02-14 04:08:35 +00005339<!-- ======================================================================= -->
5340<div class="doc_subsection">
Chris Lattnerfee11462004-02-12 17:01:32 +00005341 <a name="int_libc">Standard C Library Intrinsics</a>
5342</div>
5343
5344<div class="doc_text">
5345<p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00005346LLVM provides intrinsics for a few important standard C library functions.
5347These intrinsics allow source-language front-ends to pass information about the
5348alignment of the pointer arguments to the code generator, providing opportunity
5349for more efficient code generation.
Chris Lattnerfee11462004-02-12 17:01:32 +00005350</p>
5351
5352</div>
5353
5354<!-- _______________________________________________________________________ -->
5355<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005356 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
Chris Lattnerfee11462004-02-12 17:01:32 +00005357</div>
5358
5359<div class="doc_text">
5360
5361<h5>Syntax:</h5>
Chris Lattnerdd708342008-11-21 16:42:48 +00005362<p>This is an overloaded intrinsic. You can use llvm.memcpy on any integer bit
5363width. Not all targets support all bit widths however.</p>
Chris Lattnerfee11462004-02-12 17:01:32 +00005364<pre>
Chris Lattnerdd708342008-11-21 16:42:48 +00005365 declare void @llvm.memcpy.i8(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5366 i8 &lt;len&gt;, i32 &lt;align&gt;)
5367 declare void @llvm.memcpy.i16(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5368 i16 &lt;len&gt;, i32 &lt;align&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00005369 declare void @llvm.memcpy.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00005370 i32 &lt;len&gt;, i32 &lt;align&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00005371 declare void @llvm.memcpy.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00005372 i64 &lt;len&gt;, i32 &lt;align&gt;)
Chris Lattnerfee11462004-02-12 17:01:32 +00005373</pre>
5374
5375<h5>Overview:</h5>
5376
5377<p>
Chris Lattner0c8b2592006-03-03 00:07:20 +00005378The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
Chris Lattnerfee11462004-02-12 17:01:32 +00005379location to the destination location.
5380</p>
5381
5382<p>
Chris Lattner0c8b2592006-03-03 00:07:20 +00005383Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5384intrinsics do not return a value, and takes an extra alignment argument.
Chris Lattnerfee11462004-02-12 17:01:32 +00005385</p>
5386
5387<h5>Arguments:</h5>
5388
5389<p>
5390The first argument is a pointer to the destination, the second is a pointer to
Chris Lattner0c8b2592006-03-03 00:07:20 +00005391the source. The third argument is an integer argument
Chris Lattnerfee11462004-02-12 17:01:32 +00005392specifying the number of bytes to copy, and the fourth argument is the alignment
5393of the source and destination locations.
5394</p>
5395
Chris Lattner4c67c482004-02-12 21:18:15 +00005396<p>
5397If the call to this intrinisic has an alignment value that is not 0 or 1, then
Chris Lattner5316e5d2006-03-04 00:02:10 +00005398the caller guarantees that both the source and destination pointers are aligned
5399to that boundary.
Chris Lattner4c67c482004-02-12 21:18:15 +00005400</p>
5401
Chris Lattnerfee11462004-02-12 17:01:32 +00005402<h5>Semantics:</h5>
5403
5404<p>
Chris Lattner0c8b2592006-03-03 00:07:20 +00005405The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
Chris Lattnerfee11462004-02-12 17:01:32 +00005406location to the destination location, which are not allowed to overlap. It
5407copies "len" bytes of memory over. If the argument is known to be aligned to
5408some boundary, this can be specified as the fourth argument, otherwise it should
5409be set to 0 or 1.
5410</p>
5411</div>
5412
5413
Chris Lattnerf30152e2004-02-12 18:10:10 +00005414<!-- _______________________________________________________________________ -->
5415<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005416 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
Chris Lattnerf30152e2004-02-12 18:10:10 +00005417</div>
5418
5419<div class="doc_text">
5420
5421<h5>Syntax:</h5>
Chris Lattnerdd708342008-11-21 16:42:48 +00005422<p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5423width. Not all targets support all bit widths however.</p>
Chris Lattnerf30152e2004-02-12 18:10:10 +00005424<pre>
Chris Lattnerdd708342008-11-21 16:42:48 +00005425 declare void @llvm.memmove.i8(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5426 i8 &lt;len&gt;, i32 &lt;align&gt;)
5427 declare void @llvm.memmove.i16(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
5428 i16 &lt;len&gt;, i32 &lt;align&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00005429 declare void @llvm.memmove.i32(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00005430 i32 &lt;len&gt;, i32 &lt;align&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00005431 declare void @llvm.memmove.i64(i8 * &lt;dest&gt;, i8 * &lt;src&gt;,
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00005432 i64 &lt;len&gt;, i32 &lt;align&gt;)
Chris Lattnerf30152e2004-02-12 18:10:10 +00005433</pre>
5434
5435<h5>Overview:</h5>
5436
5437<p>
Chris Lattner0c8b2592006-03-03 00:07:20 +00005438The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5439location to the destination location. It is similar to the
Chris Lattnerec564022008-01-06 19:51:52 +00005440'<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
Chris Lattnerf30152e2004-02-12 18:10:10 +00005441</p>
5442
5443<p>
Chris Lattner0c8b2592006-03-03 00:07:20 +00005444Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5445intrinsics do not return a value, and takes an extra alignment argument.
Chris Lattnerf30152e2004-02-12 18:10:10 +00005446</p>
5447
5448<h5>Arguments:</h5>
5449
5450<p>
5451The first argument is a pointer to the destination, the second is a pointer to
Chris Lattner0c8b2592006-03-03 00:07:20 +00005452the source. The third argument is an integer argument
Chris Lattnerf30152e2004-02-12 18:10:10 +00005453specifying the number of bytes to copy, and the fourth argument is the alignment
5454of the source and destination locations.
5455</p>
5456
Chris Lattner4c67c482004-02-12 21:18:15 +00005457<p>
5458If the call to this intrinisic has an alignment value that is not 0 or 1, then
Chris Lattner5316e5d2006-03-04 00:02:10 +00005459the caller guarantees that the source and destination pointers are aligned to
5460that boundary.
Chris Lattner4c67c482004-02-12 21:18:15 +00005461</p>
5462
Chris Lattnerf30152e2004-02-12 18:10:10 +00005463<h5>Semantics:</h5>
5464
5465<p>
Chris Lattner0c8b2592006-03-03 00:07:20 +00005466The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
Chris Lattnerf30152e2004-02-12 18:10:10 +00005467location to the destination location, which may overlap. It
5468copies "len" bytes of memory over. If the argument is known to be aligned to
5469some boundary, this can be specified as the fourth argument, otherwise it should
5470be set to 0 or 1.
5471</p>
5472</div>
5473
Chris Lattner941515c2004-01-06 05:31:32 +00005474
Chris Lattner3649c3a2004-02-14 04:08:35 +00005475<!-- _______________________________________________________________________ -->
5476<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005477 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
Chris Lattner3649c3a2004-02-14 04:08:35 +00005478</div>
5479
5480<div class="doc_text">
5481
5482<h5>Syntax:</h5>
Chris Lattnerdd708342008-11-21 16:42:48 +00005483<p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5484width. Not all targets support all bit widths however.</p>
Chris Lattner3649c3a2004-02-14 04:08:35 +00005485<pre>
Chris Lattnerdd708342008-11-21 16:42:48 +00005486 declare void @llvm.memset.i8(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5487 i8 &lt;len&gt;, i32 &lt;align&gt;)
5488 declare void @llvm.memset.i16(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
5489 i16 &lt;len&gt;, i32 &lt;align&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00005490 declare void @llvm.memset.i32(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00005491 i32 &lt;len&gt;, i32 &lt;align&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00005492 declare void @llvm.memset.i64(i8 * &lt;dest&gt;, i8 &lt;val&gt;,
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00005493 i64 &lt;len&gt;, i32 &lt;align&gt;)
Chris Lattner3649c3a2004-02-14 04:08:35 +00005494</pre>
5495
5496<h5>Overview:</h5>
5497
5498<p>
Chris Lattner0c8b2592006-03-03 00:07:20 +00005499The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
Chris Lattner3649c3a2004-02-14 04:08:35 +00005500byte value.
5501</p>
5502
5503<p>
5504Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5505does not return a value, and takes an extra alignment argument.
5506</p>
5507
5508<h5>Arguments:</h5>
5509
5510<p>
5511The first argument is a pointer to the destination to fill, the second is the
Chris Lattner0c8b2592006-03-03 00:07:20 +00005512byte value to fill it with, the third argument is an integer
Chris Lattner3649c3a2004-02-14 04:08:35 +00005513argument specifying the number of bytes to fill, and the fourth argument is the
5514known alignment of destination location.
5515</p>
5516
5517<p>
5518If the call to this intrinisic has an alignment value that is not 0 or 1, then
Chris Lattner5316e5d2006-03-04 00:02:10 +00005519the caller guarantees that the destination pointer is aligned to that boundary.
Chris Lattner3649c3a2004-02-14 04:08:35 +00005520</p>
5521
5522<h5>Semantics:</h5>
5523
5524<p>
Chris Lattner0c8b2592006-03-03 00:07:20 +00005525The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5526the
Chris Lattner3649c3a2004-02-14 04:08:35 +00005527destination location. If the argument is known to be aligned to some boundary,
5528this can be specified as the fourth argument, otherwise it should be set to 0 or
55291.
5530</p>
5531</div>
5532
5533
Chris Lattner3b4f4372004-06-11 02:28:03 +00005534<!-- _______________________________________________________________________ -->
5535<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005536 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
Chris Lattner8a8f2e52005-07-21 01:29:16 +00005537</div>
5538
5539<div class="doc_text">
5540
5541<h5>Syntax:</h5>
Dale Johannesendd89d272007-10-02 17:47:38 +00005542<p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
Dan Gohmanb6324c12007-10-15 20:30:11 +00005543floating point or vector of floating point type. Not all targets support all
Dan Gohmanef9462f2008-10-14 16:51:45 +00005544types however.</p>
Chris Lattner8a8f2e52005-07-21 01:29:16 +00005545<pre>
Dale Johannesendd89d272007-10-02 17:47:38 +00005546 declare float @llvm.sqrt.f32(float %Val)
5547 declare double @llvm.sqrt.f64(double %Val)
5548 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5549 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5550 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
Chris Lattner8a8f2e52005-07-21 01:29:16 +00005551</pre>
5552
5553<h5>Overview:</h5>
5554
5555<p>
Reid Spencerb4f9a6f2006-01-16 21:12:35 +00005556The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
Dan Gohmanb6324c12007-10-15 20:30:11 +00005557returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
Chris Lattner8a8f2e52005-07-21 01:29:16 +00005558<tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
Chris Lattner00d7cb92008-01-29 07:00:44 +00005559negative numbers other than -0.0 (which allows for better optimization, because
5560there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5561defined to return -0.0 like IEEE sqrt.
Chris Lattner8a8f2e52005-07-21 01:29:16 +00005562</p>
5563
5564<h5>Arguments:</h5>
5565
5566<p>
5567The argument and return value are floating point numbers of the same type.
5568</p>
5569
5570<h5>Semantics:</h5>
5571
5572<p>
Dan Gohman33988db2007-07-16 14:37:41 +00005573This function returns the sqrt of the specified operand if it is a nonnegative
Chris Lattner8a8f2e52005-07-21 01:29:16 +00005574floating point number.
5575</p>
5576</div>
5577
Chris Lattner33b73f92006-09-08 06:34:02 +00005578<!-- _______________________________________________________________________ -->
5579<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005580 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
Chris Lattner33b73f92006-09-08 06:34:02 +00005581</div>
5582
5583<div class="doc_text">
5584
5585<h5>Syntax:</h5>
Dale Johannesendd89d272007-10-02 17:47:38 +00005586<p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
Dan Gohmanb6324c12007-10-15 20:30:11 +00005587floating point or vector of floating point type. Not all targets support all
Dan Gohmanef9462f2008-10-14 16:51:45 +00005588types however.</p>
Chris Lattner33b73f92006-09-08 06:34:02 +00005589<pre>
Dale Johannesendd89d272007-10-02 17:47:38 +00005590 declare float @llvm.powi.f32(float %Val, i32 %power)
5591 declare double @llvm.powi.f64(double %Val, i32 %power)
5592 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5593 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5594 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
Chris Lattner33b73f92006-09-08 06:34:02 +00005595</pre>
5596
5597<h5>Overview:</h5>
5598
5599<p>
5600The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5601specified (positive or negative) power. The order of evaluation of
Dan Gohmanb6324c12007-10-15 20:30:11 +00005602multiplications is not defined. When a vector of floating point type is
5603used, the second argument remains a scalar integer value.
Chris Lattner33b73f92006-09-08 06:34:02 +00005604</p>
5605
5606<h5>Arguments:</h5>
5607
5608<p>
5609The second argument is an integer power, and the first is a value to raise to
5610that power.
5611</p>
5612
5613<h5>Semantics:</h5>
5614
5615<p>
5616This function returns the first value raised to the second power with an
5617unspecified sequence of rounding operations.</p>
5618</div>
5619
Dan Gohmanb6324c12007-10-15 20:30:11 +00005620<!-- _______________________________________________________________________ -->
5621<div class="doc_subsubsection">
5622 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5623</div>
5624
5625<div class="doc_text">
5626
5627<h5>Syntax:</h5>
5628<p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5629floating point or vector of floating point type. Not all targets support all
Dan Gohmanef9462f2008-10-14 16:51:45 +00005630types however.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +00005631<pre>
5632 declare float @llvm.sin.f32(float %Val)
5633 declare double @llvm.sin.f64(double %Val)
5634 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5635 declare fp128 @llvm.sin.f128(fp128 %Val)
5636 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5637</pre>
5638
5639<h5>Overview:</h5>
5640
5641<p>
5642The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5643</p>
5644
5645<h5>Arguments:</h5>
5646
5647<p>
5648The argument and return value are floating point numbers of the same type.
5649</p>
5650
5651<h5>Semantics:</h5>
5652
5653<p>
5654This function returns the sine of the specified operand, returning the
5655same values as the libm <tt>sin</tt> functions would, and handles error
Dan Gohmand0806a02007-10-17 18:05:13 +00005656conditions in the same way.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +00005657</div>
5658
5659<!-- _______________________________________________________________________ -->
5660<div class="doc_subsubsection">
5661 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5662</div>
5663
5664<div class="doc_text">
5665
5666<h5>Syntax:</h5>
5667<p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5668floating point or vector of floating point type. Not all targets support all
Dan Gohmanef9462f2008-10-14 16:51:45 +00005669types however.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +00005670<pre>
5671 declare float @llvm.cos.f32(float %Val)
5672 declare double @llvm.cos.f64(double %Val)
5673 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5674 declare fp128 @llvm.cos.f128(fp128 %Val)
5675 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5676</pre>
5677
5678<h5>Overview:</h5>
5679
5680<p>
5681The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5682</p>
5683
5684<h5>Arguments:</h5>
5685
5686<p>
5687The argument and return value are floating point numbers of the same type.
5688</p>
5689
5690<h5>Semantics:</h5>
5691
5692<p>
5693This function returns the cosine of the specified operand, returning the
5694same values as the libm <tt>cos</tt> functions would, and handles error
Dan Gohmand0806a02007-10-17 18:05:13 +00005695conditions in the same way.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +00005696</div>
5697
5698<!-- _______________________________________________________________________ -->
5699<div class="doc_subsubsection">
5700 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5701</div>
5702
5703<div class="doc_text">
5704
5705<h5>Syntax:</h5>
5706<p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5707floating point or vector of floating point type. Not all targets support all
Dan Gohmanef9462f2008-10-14 16:51:45 +00005708types however.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +00005709<pre>
5710 declare float @llvm.pow.f32(float %Val, float %Power)
5711 declare double @llvm.pow.f64(double %Val, double %Power)
5712 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5713 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5714 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5715</pre>
5716
5717<h5>Overview:</h5>
5718
5719<p>
5720The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5721specified (positive or negative) power.
5722</p>
5723
5724<h5>Arguments:</h5>
5725
5726<p>
5727The second argument is a floating point power, and the first is a value to
5728raise to that power.
5729</p>
5730
5731<h5>Semantics:</h5>
5732
5733<p>
5734This function returns the first value raised to the second power,
5735returning the
5736same values as the libm <tt>pow</tt> functions would, and handles error
Dan Gohmand0806a02007-10-17 18:05:13 +00005737conditions in the same way.</p>
Dan Gohmanb6324c12007-10-15 20:30:11 +00005738</div>
5739
Chris Lattner33b73f92006-09-08 06:34:02 +00005740
Andrew Lenharth1d463522005-05-03 18:01:48 +00005741<!-- ======================================================================= -->
5742<div class="doc_subsection">
Nate Begeman0f223bb2006-01-13 23:26:38 +00005743 <a name="int_manip">Bit Manipulation Intrinsics</a>
Andrew Lenharth1d463522005-05-03 18:01:48 +00005744</div>
5745
5746<div class="doc_text">
5747<p>
Nate Begeman0f223bb2006-01-13 23:26:38 +00005748LLVM provides intrinsics for a few important bit manipulation operations.
Andrew Lenharth1d463522005-05-03 18:01:48 +00005749These allow efficient code generation for some algorithms.
5750</p>
5751
5752</div>
5753
5754<!-- _______________________________________________________________________ -->
5755<div class="doc_subsubsection">
Reid Spencer96a5f022007-04-04 02:42:35 +00005756 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
Nate Begeman0f223bb2006-01-13 23:26:38 +00005757</div>
5758
5759<div class="doc_text">
5760
5761<h5>Syntax:</h5>
Reid Spencer4eefaab2007-04-01 08:04:23 +00005762<p>This is an overloaded intrinsic function. You can use bswap on any integer
Dan Gohmanef9462f2008-10-14 16:51:45 +00005763type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
Nate Begeman0f223bb2006-01-13 23:26:38 +00005764<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +00005765 declare i16 @llvm.bswap.i16(i16 &lt;id&gt;)
5766 declare i32 @llvm.bswap.i32(i32 &lt;id&gt;)
5767 declare i64 @llvm.bswap.i64(i64 &lt;id&gt;)
Nate Begeman0f223bb2006-01-13 23:26:38 +00005768</pre>
5769
5770<h5>Overview:</h5>
5771
5772<p>
Reid Spencerf361c4f2007-04-02 02:25:19 +00005773The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
Reid Spencer4eefaab2007-04-01 08:04:23 +00005774values with an even number of bytes (positive multiple of 16 bits). These are
5775useful for performing operations on data that is not in the target's native
5776byte order.
Nate Begeman0f223bb2006-01-13 23:26:38 +00005777</p>
5778
5779<h5>Semantics:</h5>
5780
5781<p>
Chandler Carruth7132e002007-08-04 01:51:18 +00005782The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00005783and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5784intrinsic returns an i32 value that has the four bytes of the input i32
5785swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
Chandler Carruth7132e002007-08-04 01:51:18 +00005786i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5787<tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
Reid Spencer4eefaab2007-04-01 08:04:23 +00005788additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
Nate Begeman0f223bb2006-01-13 23:26:38 +00005789</p>
5790
5791</div>
5792
5793<!-- _______________________________________________________________________ -->
5794<div class="doc_subsubsection">
Reid Spencerb4f9a6f2006-01-16 21:12:35 +00005795 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
Andrew Lenharth1d463522005-05-03 18:01:48 +00005796</div>
5797
5798<div class="doc_text">
5799
5800<h5>Syntax:</h5>
Reid Spencer4eefaab2007-04-01 08:04:23 +00005801<p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
Dan Gohmanef9462f2008-10-14 16:51:45 +00005802width. Not all targets support all bit widths however.</p>
Andrew Lenharth1d463522005-05-03 18:01:48 +00005803<pre>
Bill Wendlingf4d70622009-02-08 01:40:31 +00005804 declare i8 @llvm.ctpop.i8(i8 &lt;src&gt;)
Chandler Carruth7132e002007-08-04 01:51:18 +00005805 declare i16 @llvm.ctpop.i16(i16 &lt;src&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00005806 declare i32 @llvm.ctpop.i32(i32 &lt;src&gt;)
Chandler Carruth7132e002007-08-04 01:51:18 +00005807 declare i64 @llvm.ctpop.i64(i64 &lt;src&gt;)
5808 declare i256 @llvm.ctpop.i256(i256 &lt;src&gt;)
Andrew Lenharth1d463522005-05-03 18:01:48 +00005809</pre>
5810
5811<h5>Overview:</h5>
5812
5813<p>
Chris Lattner069b5bd2006-01-16 22:38:59 +00005814The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5815value.
Andrew Lenharth1d463522005-05-03 18:01:48 +00005816</p>
5817
5818<h5>Arguments:</h5>
5819
5820<p>
Chris Lattner573f64e2005-05-07 01:46:40 +00005821The only argument is the value to be counted. The argument may be of any
Reid Spencer3e628eb92007-01-04 16:43:23 +00005822integer type. The return type must match the argument type.
Andrew Lenharth1d463522005-05-03 18:01:48 +00005823</p>
5824
5825<h5>Semantics:</h5>
5826
5827<p>
5828The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5829</p>
5830</div>
5831
5832<!-- _______________________________________________________________________ -->
5833<div class="doc_subsubsection">
Chris Lattnerb748c672006-01-16 22:34:14 +00005834 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
Andrew Lenharth1d463522005-05-03 18:01:48 +00005835</div>
5836
5837<div class="doc_text">
5838
5839<h5>Syntax:</h5>
Reid Spencer4eefaab2007-04-01 08:04:23 +00005840<p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
Dan Gohmanef9462f2008-10-14 16:51:45 +00005841integer bit width. Not all targets support all bit widths however.</p>
Andrew Lenharth1d463522005-05-03 18:01:48 +00005842<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +00005843 declare i8 @llvm.ctlz.i8 (i8 &lt;src&gt;)
5844 declare i16 @llvm.ctlz.i16(i16 &lt;src&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00005845 declare i32 @llvm.ctlz.i32(i32 &lt;src&gt;)
Chandler Carruth7132e002007-08-04 01:51:18 +00005846 declare i64 @llvm.ctlz.i64(i64 &lt;src&gt;)
5847 declare i256 @llvm.ctlz.i256(i256 &lt;src&gt;)
Andrew Lenharth1d463522005-05-03 18:01:48 +00005848</pre>
5849
5850<h5>Overview:</h5>
5851
5852<p>
Reid Spencerb4f9a6f2006-01-16 21:12:35 +00005853The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5854leading zeros in a variable.
Andrew Lenharth1d463522005-05-03 18:01:48 +00005855</p>
5856
5857<h5>Arguments:</h5>
5858
5859<p>
Chris Lattner573f64e2005-05-07 01:46:40 +00005860The only argument is the value to be counted. The argument may be of any
Reid Spencer3e628eb92007-01-04 16:43:23 +00005861integer type. The return type must match the argument type.
Andrew Lenharth1d463522005-05-03 18:01:48 +00005862</p>
5863
5864<h5>Semantics:</h5>
5865
5866<p>
Chris Lattnerefa20fa2005-05-15 19:39:26 +00005867The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5868in a variable. If the src == 0 then the result is the size in bits of the type
Reid Spencerb5ebf3d2006-12-31 07:07:53 +00005869of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
Andrew Lenharth1d463522005-05-03 18:01:48 +00005870</p>
5871</div>
Chris Lattner3b4f4372004-06-11 02:28:03 +00005872
5873
Chris Lattnerefa20fa2005-05-15 19:39:26 +00005874
5875<!-- _______________________________________________________________________ -->
5876<div class="doc_subsubsection">
Chris Lattnerb748c672006-01-16 22:34:14 +00005877 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
Chris Lattnerefa20fa2005-05-15 19:39:26 +00005878</div>
5879
5880<div class="doc_text">
5881
5882<h5>Syntax:</h5>
Reid Spencer4eefaab2007-04-01 08:04:23 +00005883<p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
Dan Gohmanef9462f2008-10-14 16:51:45 +00005884integer bit width. Not all targets support all bit widths however.</p>
Chris Lattnerefa20fa2005-05-15 19:39:26 +00005885<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +00005886 declare i8 @llvm.cttz.i8 (i8 &lt;src&gt;)
5887 declare i16 @llvm.cttz.i16(i16 &lt;src&gt;)
Anton Korobeynikovbe9c93c2007-03-22 00:02:17 +00005888 declare i32 @llvm.cttz.i32(i32 &lt;src&gt;)
Chandler Carruth7132e002007-08-04 01:51:18 +00005889 declare i64 @llvm.cttz.i64(i64 &lt;src&gt;)
5890 declare i256 @llvm.cttz.i256(i256 &lt;src&gt;)
Chris Lattnerefa20fa2005-05-15 19:39:26 +00005891</pre>
5892
5893<h5>Overview:</h5>
5894
5895<p>
Reid Spencerb4f9a6f2006-01-16 21:12:35 +00005896The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5897trailing zeros.
Chris Lattnerefa20fa2005-05-15 19:39:26 +00005898</p>
5899
5900<h5>Arguments:</h5>
5901
5902<p>
5903The only argument is the value to be counted. The argument may be of any
Reid Spencer3e628eb92007-01-04 16:43:23 +00005904integer type. The return type must match the argument type.
Chris Lattnerefa20fa2005-05-15 19:39:26 +00005905</p>
5906
5907<h5>Semantics:</h5>
5908
5909<p>
5910The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5911in a variable. If the src == 0 then the result is the size in bits of the type
5912of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5913</p>
5914</div>
5915
Reid Spencer8a5799f2007-04-01 08:27:01 +00005916<!-- _______________________________________________________________________ -->
5917<div class="doc_subsubsection">
Reid Spencerea2945e2007-04-10 02:51:31 +00005918 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
Reid Spencer8bc7d952007-04-01 19:00:37 +00005919</div>
5920
5921<div class="doc_text">
5922
5923<h5>Syntax:</h5>
Reid Spencerea2945e2007-04-10 02:51:31 +00005924<p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
Dan Gohmanef9462f2008-10-14 16:51:45 +00005925on any integer bit width.</p>
Reid Spencer8bc7d952007-04-01 19:00:37 +00005926<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +00005927 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5928 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
Reid Spencer8bc7d952007-04-01 19:00:37 +00005929</pre>
5930
5931<h5>Overview:</h5>
Reid Spencerea2945e2007-04-10 02:51:31 +00005932<p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
Reid Spencer8bc7d952007-04-01 19:00:37 +00005933range of bits from an integer value and returns them in the same bit width as
5934the original value.</p>
5935
5936<h5>Arguments:</h5>
5937<p>The first argument, <tt>%val</tt> and the result may be integer types of
5938any bit width but they must have the same bit width. The second and third
Reid Spencer96a5f022007-04-04 02:42:35 +00005939arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
Reid Spencer8bc7d952007-04-01 19:00:37 +00005940
5941<h5>Semantics:</h5>
Reid Spencerea2945e2007-04-10 02:51:31 +00005942<p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
Reid Spencer96a5f022007-04-04 02:42:35 +00005943of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5944<tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5945operates in forward mode.</p>
5946<p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5947right by <tt>%loBit</tt> bits and then ANDing it with a mask with
Reid Spencer8bc7d952007-04-01 19:00:37 +00005948only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5949<ol>
5950 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5951 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5952 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5953 to determine the number of bits to retain.</li>
5954 <li>A mask of the retained bits is created by shifting a -1 value.</li>
Dan Gohmanef9462f2008-10-14 16:51:45 +00005955 <li>The mask is ANDed with <tt>%val</tt> to produce the result.</li>
Reid Spencer8bc7d952007-04-01 19:00:37 +00005956</ol>
Reid Spencer70845c02007-05-14 16:14:57 +00005957<p>In reverse mode, a similar computation is made except that the bits are
5958returned in the reverse order. So, for example, if <tt>X</tt> has the value
5959<tt>i16 0x0ACF (101011001111)</tt> and we apply
5960<tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5961<tt>i16 0x0026 (000000100110)</tt>.</p>
Reid Spencer8bc7d952007-04-01 19:00:37 +00005962</div>
5963
Reid Spencer5bf54c82007-04-11 23:23:49 +00005964<div class="doc_subsubsection">
5965 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5966</div>
5967
5968<div class="doc_text">
5969
5970<h5>Syntax:</h5>
5971<p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
Dan Gohmanef9462f2008-10-14 16:51:45 +00005972on any integer bit width.</p>
Reid Spencer5bf54c82007-04-11 23:23:49 +00005973<pre>
Chandler Carruth7132e002007-08-04 01:51:18 +00005974 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5975 declare i29 @llvm.part.set.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
Reid Spencer5bf54c82007-04-11 23:23:49 +00005976</pre>
5977
5978<h5>Overview:</h5>
5979<p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5980of bits in an integer value with another integer value. It returns the integer
5981with the replaced bits.</p>
5982
5983<h5>Arguments:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00005984<p>The first argument, <tt>%val</tt>, and the result may be integer types of
5985any bit width, but they must have the same bit width. <tt>%val</tt> is the value
Reid Spencer5bf54c82007-04-11 23:23:49 +00005986whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5987integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5988type since they specify only a bit index.</p>
5989
5990<h5>Semantics:</h5>
5991<p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5992of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5993<tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5994operates in forward mode.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00005995
Reid Spencer5bf54c82007-04-11 23:23:49 +00005996<p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5997truncating it down to the size of the replacement area or zero extending it
5998up to that size.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00005999
Reid Spencer5bf54c82007-04-11 23:23:49 +00006000<p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
6001are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
6002in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
Dan Gohmanef9462f2008-10-14 16:51:45 +00006003to the <tt>%hi</tt>th bit.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006004
Reid Spencer146281c2007-05-14 16:50:20 +00006005<p>In reverse mode, a similar computation is made except that the bits are
6006reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
Dan Gohmanef9462f2008-10-14 16:51:45 +00006007<tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006008
Reid Spencer5bf54c82007-04-11 23:23:49 +00006009<h5>Examples:</h5>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006010
Reid Spencer5bf54c82007-04-11 23:23:49 +00006011<pre>
Reid Spencerc70afc32007-04-12 01:03:03 +00006012 llvm.part.set(0xFFFF, 0, 4, 7) -&gt; 0xFF0F
Reid Spencer146281c2007-05-14 16:50:20 +00006013 llvm.part.set(0xFFFF, 0, 7, 4) -&gt; 0xFF0F
6014 llvm.part.set(0xFFFF, 1, 7, 4) -&gt; 0xFF8F
6015 llvm.part.set(0xFFFF, F, 8, 3) -&gt; 0xFFE7
Reid Spencerc70afc32007-04-12 01:03:03 +00006016 llvm.part.set(0xFFFF, 0, 3, 8) -&gt; 0xFE07
Reid Spencer7972c472007-04-11 23:49:50 +00006017</pre>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006018
6019</div>
6020
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006021<!-- ======================================================================= -->
6022<div class="doc_subsection">
6023 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
6024</div>
6025
6026<div class="doc_text">
6027<p>
6028LLVM provides intrinsics for some arithmetic with overflow operations.
6029</p>
6030
6031</div>
6032
Bill Wendlingf4d70622009-02-08 01:40:31 +00006033<!-- _______________________________________________________________________ -->
6034<div class="doc_subsubsection">
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006035 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006036</div>
6037
6038<div class="doc_text">
6039
6040<h5>Syntax:</h5>
6041
6042<p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006043on any integer bit width.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006044
6045<pre>
6046 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
6047 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6048 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6049</pre>
6050
6051<h5>Overview:</h5>
6052
6053<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6054a signed addition of the two arguments, and indicate whether an overflow
6055occurred during the signed summation.</p>
6056
6057<h5>Arguments:</h5>
6058
6059<p>The arguments (%a and %b) and the first element of the result structure may
6060be of integer types of any bit width, but they must have the same bit width. The
6061second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6062and <tt>%b</tt> are the two values that will undergo signed addition.</p>
6063
6064<h5>Semantics:</h5>
6065
6066<p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6067a signed addition of the two variables. They return a structure &mdash; the
6068first element of which is the signed summation, and the second element of which
6069is a bit specifying if the signed summation resulted in an overflow.</p>
6070
6071<h5>Examples:</h5>
6072<pre>
6073 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6074 %sum = extractvalue {i32, i1} %res, 0
6075 %obit = extractvalue {i32, i1} %res, 1
6076 br i1 %obit, label %overflow, label %normal
6077</pre>
6078
6079</div>
6080
6081<!-- _______________________________________________________________________ -->
6082<div class="doc_subsubsection">
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006083 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006084</div>
6085
6086<div class="doc_text">
6087
6088<h5>Syntax:</h5>
6089
6090<p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006091on any integer bit width.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006092
6093<pre>
6094 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6095 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6096 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6097</pre>
6098
6099<h5>Overview:</h5>
6100
6101<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6102an unsigned addition of the two arguments, and indicate whether a carry occurred
6103during the unsigned summation.</p>
6104
6105<h5>Arguments:</h5>
6106
6107<p>The arguments (%a and %b) and the first element of the result structure may
6108be of integer types of any bit width, but they must have the same bit width. The
6109second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6110and <tt>%b</tt> are the two values that will undergo unsigned addition.</p>
6111
6112<h5>Semantics:</h5>
6113
6114<p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6115an unsigned addition of the two arguments. They return a structure &mdash; the
6116first element of which is the sum, and the second element of which is a bit
6117specifying if the unsigned summation resulted in a carry.</p>
6118
6119<h5>Examples:</h5>
6120<pre>
6121 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6122 %sum = extractvalue {i32, i1} %res, 0
6123 %obit = extractvalue {i32, i1} %res, 1
6124 br i1 %obit, label %carry, label %normal
6125</pre>
6126
6127</div>
6128
6129<!-- _______________________________________________________________________ -->
6130<div class="doc_subsubsection">
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006131 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006132</div>
6133
6134<div class="doc_text">
6135
6136<h5>Syntax:</h5>
6137
6138<p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006139on any integer bit width.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006140
6141<pre>
6142 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6143 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6144 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6145</pre>
6146
6147<h5>Overview:</h5>
6148
6149<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6150a signed subtraction of the two arguments, and indicate whether an overflow
6151occurred during the signed subtraction.</p>
6152
6153<h5>Arguments:</h5>
6154
6155<p>The arguments (%a and %b) and the first element of the result structure may
6156be of integer types of any bit width, but they must have the same bit width. The
6157second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6158and <tt>%b</tt> are the two values that will undergo signed subtraction.</p>
6159
6160<h5>Semantics:</h5>
6161
6162<p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6163a signed subtraction of the two arguments. They return a structure &mdash; the
6164first element of which is the subtraction, and the second element of which is a bit
6165specifying if the signed subtraction resulted in an overflow.</p>
6166
6167<h5>Examples:</h5>
6168<pre>
6169 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6170 %sum = extractvalue {i32, i1} %res, 0
6171 %obit = extractvalue {i32, i1} %res, 1
6172 br i1 %obit, label %overflow, label %normal
6173</pre>
6174
6175</div>
6176
6177<!-- _______________________________________________________________________ -->
6178<div class="doc_subsubsection">
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006179 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006180</div>
6181
6182<div class="doc_text">
6183
6184<h5>Syntax:</h5>
6185
6186<p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006187on any integer bit width.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006188
6189<pre>
6190 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6191 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6192 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6193</pre>
6194
6195<h5>Overview:</h5>
6196
6197<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6198an unsigned subtraction of the two arguments, and indicate whether an overflow
6199occurred during the unsigned subtraction.</p>
6200
6201<h5>Arguments:</h5>
6202
6203<p>The arguments (%a and %b) and the first element of the result structure may
6204be of integer types of any bit width, but they must have the same bit width. The
6205second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6206and <tt>%b</tt> are the two values that will undergo unsigned subtraction.</p>
6207
6208<h5>Semantics:</h5>
6209
6210<p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6211an unsigned subtraction of the two arguments. They return a structure &mdash; the
6212first element of which is the subtraction, and the second element of which is a bit
6213specifying if the unsigned subtraction resulted in an overflow.</p>
6214
6215<h5>Examples:</h5>
6216<pre>
6217 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6218 %sum = extractvalue {i32, i1} %res, 0
6219 %obit = extractvalue {i32, i1} %res, 1
6220 br i1 %obit, label %overflow, label %normal
6221</pre>
6222
6223</div>
6224
6225<!-- _______________________________________________________________________ -->
6226<div class="doc_subsubsection">
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006227 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006228</div>
6229
6230<div class="doc_text">
6231
6232<h5>Syntax:</h5>
6233
6234<p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
Bill Wendlingfd2bd722009-02-08 04:04:40 +00006235on any integer bit width.</p>
Bill Wendlingf4d70622009-02-08 01:40:31 +00006236
6237<pre>
6238 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6239 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6240 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6241</pre>
6242
6243<h5>Overview:</h5>
6244
6245<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6246a signed multiplication of the two arguments, and indicate whether an overflow
6247occurred during the signed multiplication.</p>
6248
6249<h5>Arguments:</h5>
6250
6251<p>The arguments (%a and %b) and the first element of the result structure may
6252be of integer types of any bit width, but they must have the same bit width. The
6253second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6254and <tt>%b</tt> are the two values that will undergo signed multiplication.</p>
6255
6256<h5>Semantics:</h5>
6257
6258<p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6259a signed multiplication of the two arguments. They return a structure &mdash;
6260the first element of which is the multiplication, and the second element of
6261which is a bit specifying if the signed multiplication resulted in an
6262overflow.</p>
6263
6264<h5>Examples:</h5>
6265<pre>
6266 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6267 %sum = extractvalue {i32, i1} %res, 0
6268 %obit = extractvalue {i32, i1} %res, 1
6269 br i1 %obit, label %overflow, label %normal
6270</pre>
6271
Reid Spencer5bf54c82007-04-11 23:23:49 +00006272</div>
6273
Bill Wendlingb9a73272009-02-08 23:00:09 +00006274<!-- _______________________________________________________________________ -->
6275<div class="doc_subsubsection">
6276 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6277</div>
6278
6279<div class="doc_text">
6280
6281<h5>Syntax:</h5>
6282
6283<p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6284on any integer bit width.</p>
6285
6286<pre>
6287 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6288 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6289 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6290</pre>
6291
6292<h5>Overview:</h5>
6293
6294<p><i><b>Warning:</b> '<tt>llvm.umul.with.overflow</tt>' is badly broken. It is
6295actively being fixed, but it should not currently be used!</i></p>
6296
6297<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6298a unsigned multiplication of the two arguments, and indicate whether an overflow
6299occurred during the unsigned multiplication.</p>
6300
6301<h5>Arguments:</h5>
6302
6303<p>The arguments (%a and %b) and the first element of the result structure may
6304be of integer types of any bit width, but they must have the same bit width. The
6305second element of the result structure must be of type <tt>i1</tt>. <tt>%a</tt>
6306and <tt>%b</tt> are the two values that will undergo unsigned
6307multiplication.</p>
6308
6309<h5>Semantics:</h5>
6310
6311<p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6312an unsigned multiplication of the two arguments. They return a structure &mdash;
6313the first element of which is the multiplication, and the second element of
6314which is a bit specifying if the unsigned multiplication resulted in an
6315overflow.</p>
6316
6317<h5>Examples:</h5>
6318<pre>
6319 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6320 %sum = extractvalue {i32, i1} %res, 0
6321 %obit = extractvalue {i32, i1} %res, 1
6322 br i1 %obit, label %overflow, label %normal
6323</pre>
6324
6325</div>
6326
Chris Lattner941515c2004-01-06 05:31:32 +00006327<!-- ======================================================================= -->
6328<div class="doc_subsection">
6329 <a name="int_debugger">Debugger Intrinsics</a>
6330</div>
6331
6332<div class="doc_text">
6333<p>
6334The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
6335are described in the <a
6336href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
6337Debugging</a> document.
6338</p>
6339</div>
6340
6341
Jim Laskey2211f492007-03-14 19:31:19 +00006342<!-- ======================================================================= -->
6343<div class="doc_subsection">
6344 <a name="int_eh">Exception Handling Intrinsics</a>
6345</div>
6346
6347<div class="doc_text">
6348<p> The LLVM exception handling intrinsics (which all start with
6349<tt>llvm.eh.</tt> prefix), are described in the <a
6350href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6351Handling</a> document. </p>
6352</div>
6353
Tanya Lattnercb1b9602007-06-15 20:50:54 +00006354<!-- ======================================================================= -->
6355<div class="doc_subsection">
Duncan Sands86e01192007-09-11 14:10:23 +00006356 <a name="int_trampoline">Trampoline Intrinsic</a>
Duncan Sands644f9172007-07-27 12:58:54 +00006357</div>
6358
6359<div class="doc_text">
6360<p>
Duncan Sands86e01192007-09-11 14:10:23 +00006361 This intrinsic makes it possible to excise one parameter, marked with
Duncan Sands644f9172007-07-27 12:58:54 +00006362 the <tt>nest</tt> attribute, from a function. The result is a callable
6363 function pointer lacking the nest parameter - the caller does not need
6364 to provide a value for it. Instead, the value to use is stored in
6365 advance in a "trampoline", a block of memory usually allocated
6366 on the stack, which also contains code to splice the nest value into the
6367 argument list. This is used to implement the GCC nested function address
6368 extension.
6369</p>
6370<p>
6371 For example, if the function is
6372 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
Bill Wendling252570f2007-09-22 09:23:55 +00006373 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:</p>
Duncan Sands644f9172007-07-27 12:58:54 +00006374<pre>
Duncan Sands86e01192007-09-11 14:10:23 +00006375 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6376 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6377 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
6378 %fp = bitcast i8* %p to i32 (i32, i32)*
Duncan Sands644f9172007-07-27 12:58:54 +00006379</pre>
Bill Wendling252570f2007-09-22 09:23:55 +00006380 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
6381 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
Duncan Sands644f9172007-07-27 12:58:54 +00006382</div>
6383
6384<!-- _______________________________________________________________________ -->
6385<div class="doc_subsubsection">
6386 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6387</div>
6388<div class="doc_text">
6389<h5>Syntax:</h5>
6390<pre>
Duncan Sands86e01192007-09-11 14:10:23 +00006391declare i8* @llvm.init.trampoline(i8* &lt;tramp&gt;, i8* &lt;func&gt;, i8* &lt;nval&gt;)
Duncan Sands644f9172007-07-27 12:58:54 +00006392</pre>
6393<h5>Overview:</h5>
6394<p>
Duncan Sands86e01192007-09-11 14:10:23 +00006395 This fills the memory pointed to by <tt>tramp</tt> with code
6396 and returns a function pointer suitable for executing it.
Duncan Sands644f9172007-07-27 12:58:54 +00006397</p>
6398<h5>Arguments:</h5>
6399<p>
6400 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6401 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
6402 and sufficiently aligned block of memory; this memory is written to by the
Duncan Sandsf2bcd372007-08-22 23:39:54 +00006403 intrinsic. Note that the size and the alignment are target-specific - LLVM
6404 currently provides no portable way of determining them, so a front-end that
6405 generates this intrinsic needs to have some target-specific knowledge.
6406 The <tt>func</tt> argument must hold a function bitcast to an <tt>i8*</tt>.
Duncan Sands644f9172007-07-27 12:58:54 +00006407</p>
6408<h5>Semantics:</h5>
6409<p>
6410 The block of memory pointed to by <tt>tramp</tt> is filled with target
Duncan Sands86e01192007-09-11 14:10:23 +00006411 dependent code, turning it into a function. A pointer to this function is
6412 returned, but needs to be bitcast to an
Duncan Sands644f9172007-07-27 12:58:54 +00006413 <a href="#int_trampoline">appropriate function pointer type</a>
Duncan Sands86e01192007-09-11 14:10:23 +00006414 before being called. The new function's signature is the same as that of
6415 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
6416 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
6417 of pointer type. Calling the new function is equivalent to calling
6418 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
6419 missing <tt>nest</tt> argument. If, after calling
6420 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
6421 modified, then the effect of any later call to the returned function pointer is
6422 undefined.
Duncan Sands644f9172007-07-27 12:58:54 +00006423</p>
6424</div>
6425
6426<!-- ======================================================================= -->
6427<div class="doc_subsection">
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00006428 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6429</div>
6430
6431<div class="doc_text">
6432<p>
6433 These intrinsic functions expand the "universal IR" of LLVM to represent
6434 hardware constructs for atomic operations and memory synchronization. This
6435 provides an interface to the hardware, not an interface to the programmer. It
Chris Lattner67c37d12008-08-05 18:29:16 +00006436 is aimed at a low enough level to allow any programming models or APIs
6437 (Application Programming Interfaces) which
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00006438 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
6439 hardware behavior. Just as hardware provides a "universal IR" for source
6440 languages, it also provides a starting point for developing a "universal"
6441 atomic operation and synchronization IR.
6442</p>
6443<p>
6444 These do <em>not</em> form an API such as high-level threading libraries,
6445 software transaction memory systems, atomic primitives, and intrinsic
6446 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6447 application libraries. The hardware interface provided by LLVM should allow
6448 a clean implementation of all of these APIs and parallel programming models.
6449 No one model or paradigm should be selected above others unless the hardware
6450 itself ubiquitously does so.
6451
6452</p>
6453</div>
6454
6455<!-- _______________________________________________________________________ -->
6456<div class="doc_subsubsection">
6457 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6458</div>
6459<div class="doc_text">
6460<h5>Syntax:</h5>
6461<pre>
6462declare void @llvm.memory.barrier( i1 &lt;ll&gt;, i1 &lt;ls&gt;, i1 &lt;sl&gt;, i1 &lt;ss&gt;,
6463i1 &lt;device&gt; )
6464
6465</pre>
6466<h5>Overview:</h5>
6467<p>
6468 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6469 specific pairs of memory access types.
6470</p>
6471<h5>Arguments:</h5>
6472<p>
6473 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6474 The first four arguments enables a specific barrier as listed below. The fith
6475 argument specifies that the barrier applies to io or device or uncached memory.
6476
6477</p>
6478 <ul>
6479 <li><tt>ll</tt>: load-load barrier</li>
6480 <li><tt>ls</tt>: load-store barrier</li>
6481 <li><tt>sl</tt>: store-load barrier</li>
6482 <li><tt>ss</tt>: store-store barrier</li>
Dan Gohmanef9462f2008-10-14 16:51:45 +00006483 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00006484 </ul>
6485<h5>Semantics:</h5>
6486<p>
6487 This intrinsic causes the system to enforce some ordering constraints upon
6488 the loads and stores of the program. This barrier does not indicate
6489 <em>when</em> any events will occur, it only enforces an <em>order</em> in
6490 which they occur. For any of the specified pairs of load and store operations
6491 (f.ex. load-load, or store-load), all of the first operations preceding the
6492 barrier will complete before any of the second operations succeeding the
6493 barrier begin. Specifically the semantics for each pairing is as follows:
6494</p>
6495 <ul>
6496 <li><tt>ll</tt>: All loads before the barrier must complete before any load
6497 after the barrier begins.</li>
6498
6499 <li><tt>ls</tt>: All loads before the barrier must complete before any
6500 store after the barrier begins.</li>
6501 <li><tt>ss</tt>: All stores before the barrier must complete before any
6502 store after the barrier begins.</li>
6503 <li><tt>sl</tt>: All stores before the barrier must complete before any
6504 load after the barrier begins.</li>
6505 </ul>
6506<p>
6507 These semantics are applied with a logical "and" behavior when more than one
6508 is enabled in a single memory barrier intrinsic.
6509</p>
6510<p>
6511 Backends may implement stronger barriers than those requested when they do not
6512 support as fine grained a barrier as requested. Some architectures do not
6513 need all types of barriers and on such architectures, these become noops.
6514</p>
6515<h5>Example:</h5>
6516<pre>
6517%ptr = malloc i32
6518 store i32 4, %ptr
6519
6520%result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
6521 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
6522 <i>; guarantee the above finishes</i>
6523 store i32 8, %ptr <i>; before this begins</i>
6524</pre>
6525</div>
6526
Andrew Lenharth95528942008-02-21 06:45:13 +00006527<!-- _______________________________________________________________________ -->
6528<div class="doc_subsubsection">
Mon P Wang6a490372008-06-25 08:15:39 +00006529 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
Andrew Lenharth95528942008-02-21 06:45:13 +00006530</div>
6531<div class="doc_text">
6532<h5>Syntax:</h5>
6533<p>
Mon P Wang2c839d42008-07-30 04:36:53 +00006534 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
6535 any integer bit width and for different address spaces. Not all targets
6536 support all bit widths however.</p>
Andrew Lenharth95528942008-02-21 06:45:13 +00006537
6538<pre>
Mon P Wang2c839d42008-07-30 04:36:53 +00006539declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;cmp&gt;, i8 &lt;val&gt; )
6540declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;cmp&gt;, i16 &lt;val&gt; )
6541declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;cmp&gt;, i32 &lt;val&gt; )
6542declare 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 +00006543
6544</pre>
6545<h5>Overview:</h5>
6546<p>
6547 This loads a value in memory and compares it to a given value. If they are
6548 equal, it stores a new value into the memory.
6549</p>
6550<h5>Arguments:</h5>
6551<p>
Mon P Wang6a490372008-06-25 08:15:39 +00006552 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
Andrew Lenharth95528942008-02-21 06:45:13 +00006553 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6554 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6555 this integer type. While any bit width integer may be used, targets may only
6556 lower representations they support in hardware.
6557
6558</p>
6559<h5>Semantics:</h5>
6560<p>
6561 This entire intrinsic must be executed atomically. It first loads the value
6562 in memory pointed to by <tt>ptr</tt> and compares it with the value
6563 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
6564 loaded value is yielded in all cases. This provides the equivalent of an
6565 atomic compare-and-swap operation within the SSA framework.
6566</p>
6567<h5>Examples:</h5>
6568
6569<pre>
6570%ptr = malloc i32
6571 store i32 4, %ptr
6572
6573%val1 = add i32 4, 4
Mon P Wang2c839d42008-07-30 04:36:53 +00006574%result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
Andrew Lenharth95528942008-02-21 06:45:13 +00006575 <i>; yields {i32}:result1 = 4</i>
6576%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6577%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6578
6579%val2 = add i32 1, 1
Mon P Wang2c839d42008-07-30 04:36:53 +00006580%result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
Andrew Lenharth95528942008-02-21 06:45:13 +00006581 <i>; yields {i32}:result2 = 8</i>
6582%stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6583
6584%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6585</pre>
6586</div>
6587
6588<!-- _______________________________________________________________________ -->
6589<div class="doc_subsubsection">
6590 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6591</div>
6592<div class="doc_text">
6593<h5>Syntax:</h5>
6594
6595<p>
6596 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6597 integer bit width. Not all targets support all bit widths however.</p>
6598<pre>
Mon P Wang2c839d42008-07-30 04:36:53 +00006599declare i8 @llvm.atomic.swap.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;val&gt; )
6600declare i16 @llvm.atomic.swap.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;val&gt; )
6601declare i32 @llvm.atomic.swap.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;val&gt; )
6602declare i64 @llvm.atomic.swap.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;val&gt; )
Andrew Lenharth95528942008-02-21 06:45:13 +00006603
6604</pre>
6605<h5>Overview:</h5>
6606<p>
6607 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6608 the value from memory. It then stores the value in <tt>val</tt> in the memory
6609 at <tt>ptr</tt>.
6610</p>
6611<h5>Arguments:</h5>
6612
6613<p>
Mon P Wang6a490372008-06-25 08:15:39 +00006614 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
Andrew Lenharth95528942008-02-21 06:45:13 +00006615 <tt>val</tt> argument and the result must be integers of the same bit width.
6616 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6617 integer type. The targets may only lower integer representations they
6618 support.
6619</p>
6620<h5>Semantics:</h5>
6621<p>
6622 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6623 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6624 equivalent of an atomic swap operation within the SSA framework.
6625
6626</p>
6627<h5>Examples:</h5>
6628<pre>
6629%ptr = malloc i32
6630 store i32 4, %ptr
6631
6632%val1 = add i32 4, 4
Mon P Wang2c839d42008-07-30 04:36:53 +00006633%result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
Andrew Lenharth95528942008-02-21 06:45:13 +00006634 <i>; yields {i32}:result1 = 4</i>
6635%stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6636%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6637
6638%val2 = add i32 1, 1
Mon P Wang2c839d42008-07-30 04:36:53 +00006639%result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
Andrew Lenharth95528942008-02-21 06:45:13 +00006640 <i>; yields {i32}:result2 = 8</i>
6641
6642%stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6643%memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6644</pre>
6645</div>
6646
6647<!-- _______________________________________________________________________ -->
6648<div class="doc_subsubsection">
Mon P Wang6a490372008-06-25 08:15:39 +00006649 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
Andrew Lenharth95528942008-02-21 06:45:13 +00006650
6651</div>
6652<div class="doc_text">
6653<h5>Syntax:</h5>
6654<p>
Mon P Wang6a490372008-06-25 08:15:39 +00006655 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on any
Andrew Lenharth95528942008-02-21 06:45:13 +00006656 integer bit width. Not all targets support all bit widths however.</p>
6657<pre>
Mon P Wang2c839d42008-07-30 04:36:53 +00006658declare i8 @llvm.atomic.load.add.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6659declare i16 @llvm.atomic.load.add.i16..p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6660declare i32 @llvm.atomic.load.add.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6661declare i64 @llvm.atomic.load.add.i64..p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Andrew Lenharth95528942008-02-21 06:45:13 +00006662
6663</pre>
6664<h5>Overview:</h5>
6665<p>
6666 This intrinsic adds <tt>delta</tt> to the value stored in memory at
6667 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6668</p>
6669<h5>Arguments:</h5>
6670<p>
6671
6672 The intrinsic takes two arguments, the first a pointer to an integer value
6673 and the second an integer value. The result is also an integer value. These
6674 integer types can have any bit width, but they must all have the same bit
6675 width. The targets may only lower integer representations they support.
6676</p>
6677<h5>Semantics:</h5>
6678<p>
6679 This intrinsic does a series of operations atomically. It first loads the
6680 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6681 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6682</p>
6683
6684<h5>Examples:</h5>
6685<pre>
6686%ptr = malloc i32
6687 store i32 4, %ptr
Mon P Wang2c839d42008-07-30 04:36:53 +00006688%result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
Andrew Lenharth95528942008-02-21 06:45:13 +00006689 <i>; yields {i32}:result1 = 4</i>
Mon P Wang2c839d42008-07-30 04:36:53 +00006690%result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
Andrew Lenharth95528942008-02-21 06:45:13 +00006691 <i>; yields {i32}:result2 = 8</i>
Mon P Wang2c839d42008-07-30 04:36:53 +00006692%result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
Andrew Lenharth95528942008-02-21 06:45:13 +00006693 <i>; yields {i32}:result3 = 10</i>
Mon P Wang6a490372008-06-25 08:15:39 +00006694%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
Andrew Lenharth95528942008-02-21 06:45:13 +00006695</pre>
6696</div>
6697
Mon P Wang6a490372008-06-25 08:15:39 +00006698<!-- _______________________________________________________________________ -->
6699<div class="doc_subsubsection">
6700 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6701
6702</div>
6703<div class="doc_text">
6704<h5>Syntax:</h5>
6705<p>
6706 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
Mon P Wang2c839d42008-07-30 04:36:53 +00006707 any integer bit width and for different address spaces. Not all targets
6708 support all bit widths however.</p>
Mon P Wang6a490372008-06-25 08:15:39 +00006709<pre>
Mon P Wang2c839d42008-07-30 04:36:53 +00006710declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6711declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6712declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6713declare i64 @llvm.atomic.load.sub.i64.p0i32( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6a490372008-06-25 08:15:39 +00006714
6715</pre>
6716<h5>Overview:</h5>
6717<p>
6718 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6719 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6720</p>
6721<h5>Arguments:</h5>
6722<p>
6723
6724 The intrinsic takes two arguments, the first a pointer to an integer value
6725 and the second an integer value. The result is also an integer value. These
6726 integer types can have any bit width, but they must all have the same bit
6727 width. The targets may only lower integer representations they support.
6728</p>
6729<h5>Semantics:</h5>
6730<p>
6731 This intrinsic does a series of operations atomically. It first loads the
6732 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6733 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6734</p>
6735
6736<h5>Examples:</h5>
6737<pre>
6738%ptr = malloc i32
6739 store i32 8, %ptr
Mon P Wang2c839d42008-07-30 04:36:53 +00006740%result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
Mon P Wang6a490372008-06-25 08:15:39 +00006741 <i>; yields {i32}:result1 = 8</i>
Mon P Wang2c839d42008-07-30 04:36:53 +00006742%result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
Mon P Wang6a490372008-06-25 08:15:39 +00006743 <i>; yields {i32}:result2 = 4</i>
Mon P Wang2c839d42008-07-30 04:36:53 +00006744%result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
Mon P Wang6a490372008-06-25 08:15:39 +00006745 <i>; yields {i32}:result3 = 2</i>
6746%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6747</pre>
6748</div>
6749
6750<!-- _______________________________________________________________________ -->
6751<div class="doc_subsubsection">
6752 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6753 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6754 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6755 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6756
6757</div>
6758<div class="doc_text">
6759<h5>Syntax:</h5>
6760<p>
6761 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6762 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
Mon P Wang2c839d42008-07-30 04:36:53 +00006763 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6764 address spaces. Not all targets support all bit widths however.</p>
Mon P Wang6a490372008-06-25 08:15:39 +00006765<pre>
Mon P Wang2c839d42008-07-30 04:36:53 +00006766declare i8 @llvm.atomic.load.and.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6767declare i16 @llvm.atomic.load.and.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6768declare i32 @llvm.atomic.load.and.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6769declare i64 @llvm.atomic.load.and.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6a490372008-06-25 08:15:39 +00006770
6771</pre>
6772
6773<pre>
Mon P Wang2c839d42008-07-30 04:36:53 +00006774declare i8 @llvm.atomic.load.or.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6775declare i16 @llvm.atomic.load.or.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6776declare i32 @llvm.atomic.load.or.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6777declare i64 @llvm.atomic.load.or.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6a490372008-06-25 08:15:39 +00006778
6779</pre>
6780
6781<pre>
Mon P Wang2c839d42008-07-30 04:36:53 +00006782declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6783declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6784declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6785declare i64 @llvm.atomic.load.nand.i64.p0i32( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6a490372008-06-25 08:15:39 +00006786
6787</pre>
6788
6789<pre>
Mon P Wang2c839d42008-07-30 04:36:53 +00006790declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6791declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6792declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6793declare i64 @llvm.atomic.load.xor.i64.p0i32( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6a490372008-06-25 08:15:39 +00006794
6795</pre>
6796<h5>Overview:</h5>
6797<p>
6798 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6799 the value stored in memory at <tt>ptr</tt>. It yields the original value
6800 at <tt>ptr</tt>.
6801</p>
6802<h5>Arguments:</h5>
6803<p>
6804
6805 These intrinsics take two arguments, the first a pointer to an integer value
6806 and the second an integer value. The result is also an integer value. These
6807 integer types can have any bit width, but they must all have the same bit
6808 width. The targets may only lower integer representations they support.
6809</p>
6810<h5>Semantics:</h5>
6811<p>
6812 These intrinsics does a series of operations atomically. They first load the
6813 value stored at <tt>ptr</tt>. They then do the bitwise operation
6814 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6815 value stored at <tt>ptr</tt>.
6816</p>
6817
6818<h5>Examples:</h5>
6819<pre>
6820%ptr = malloc i32
6821 store i32 0x0F0F, %ptr
Mon P Wang2c839d42008-07-30 04:36:53 +00006822%result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6a490372008-06-25 08:15:39 +00006823 <i>; yields {i32}:result0 = 0x0F0F</i>
Mon P Wang2c839d42008-07-30 04:36:53 +00006824%result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
Mon P Wang6a490372008-06-25 08:15:39 +00006825 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
Mon P Wang2c839d42008-07-30 04:36:53 +00006826%result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6a490372008-06-25 08:15:39 +00006827 <i>; yields {i32}:result2 = 0xF0</i>
Mon P Wang2c839d42008-07-30 04:36:53 +00006828%result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
Mon P Wang6a490372008-06-25 08:15:39 +00006829 <i>; yields {i32}:result3 = FF</i>
6830%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6831</pre>
6832</div>
6833
6834
6835<!-- _______________________________________________________________________ -->
6836<div class="doc_subsubsection">
6837 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6838 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6839 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6840 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6841
6842</div>
6843<div class="doc_text">
6844<h5>Syntax:</h5>
6845<p>
6846 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6847 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
Mon P Wang2c839d42008-07-30 04:36:53 +00006848 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6849 address spaces. Not all targets
Mon P Wang6a490372008-06-25 08:15:39 +00006850 support all bit widths however.</p>
6851<pre>
Mon P Wang2c839d42008-07-30 04:36:53 +00006852declare i8 @llvm.atomic.load.max.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6853declare i16 @llvm.atomic.load.max.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6854declare i32 @llvm.atomic.load.max.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6855declare i64 @llvm.atomic.load.max.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6a490372008-06-25 08:15:39 +00006856
6857</pre>
6858
6859<pre>
Mon P Wang2c839d42008-07-30 04:36:53 +00006860declare i8 @llvm.atomic.load.min.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6861declare i16 @llvm.atomic.load.min.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6862declare i32 @llvm.atomic.load.min.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6863declare i64 @llvm.atomic.load.min.i64..p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6a490372008-06-25 08:15:39 +00006864
6865</pre>
6866
6867<pre>
Mon P Wang2c839d42008-07-30 04:36:53 +00006868declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6869declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6870declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6871declare i64 @llvm.atomic.load.umax.i64.p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6a490372008-06-25 08:15:39 +00006872
6873</pre>
6874
6875<pre>
Mon P Wang2c839d42008-07-30 04:36:53 +00006876declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* &lt;ptr&gt;, i8 &lt;delta&gt; )
6877declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* &lt;ptr&gt;, i16 &lt;delta&gt; )
6878declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* &lt;ptr&gt;, i32 &lt;delta&gt; )
6879declare i64 @llvm.atomic.load.umin.i64..p0i64( i64* &lt;ptr&gt;, i64 &lt;delta&gt; )
Mon P Wang6a490372008-06-25 08:15:39 +00006880
6881</pre>
6882<h5>Overview:</h5>
6883<p>
6884 These intrinsics takes the signed or unsigned minimum or maximum of
6885 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6886 original value at <tt>ptr</tt>.
6887</p>
6888<h5>Arguments:</h5>
6889<p>
6890
6891 These intrinsics take two arguments, the first a pointer to an integer value
6892 and the second an integer value. The result is also an integer value. These
6893 integer types can have any bit width, but they must all have the same bit
6894 width. The targets may only lower integer representations they support.
6895</p>
6896<h5>Semantics:</h5>
6897<p>
6898 These intrinsics does a series of operations atomically. They first load the
6899 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6900 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6901 the original value stored at <tt>ptr</tt>.
6902</p>
6903
6904<h5>Examples:</h5>
6905<pre>
6906%ptr = malloc i32
6907 store i32 7, %ptr
Mon P Wang2c839d42008-07-30 04:36:53 +00006908%result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
Mon P Wang6a490372008-06-25 08:15:39 +00006909 <i>; yields {i32}:result0 = 7</i>
Mon P Wang2c839d42008-07-30 04:36:53 +00006910%result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
Mon P Wang6a490372008-06-25 08:15:39 +00006911 <i>; yields {i32}:result1 = -2</i>
Mon P Wang2c839d42008-07-30 04:36:53 +00006912%result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
Mon P Wang6a490372008-06-25 08:15:39 +00006913 <i>; yields {i32}:result2 = 8</i>
Mon P Wang2c839d42008-07-30 04:36:53 +00006914%result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
Mon P Wang6a490372008-06-25 08:15:39 +00006915 <i>; yields {i32}:result3 = 8</i>
6916%memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6917</pre>
6918</div>
Andrew Lenharth9b254ee2008-02-16 01:24:58 +00006919
6920<!-- ======================================================================= -->
6921<div class="doc_subsection">
Tanya Lattnercb1b9602007-06-15 20:50:54 +00006922 <a name="int_general">General Intrinsics</a>
6923</div>
6924
6925<div class="doc_text">
6926<p> This class of intrinsics is designed to be generic and has
6927no specific purpose. </p>
6928</div>
6929
6930<!-- _______________________________________________________________________ -->
6931<div class="doc_subsubsection">
6932 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6933</div>
6934
6935<div class="doc_text">
6936
6937<h5>Syntax:</h5>
6938<pre>
Tanya Lattnerbed1d4d2007-06-18 23:42:37 +00006939 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 +00006940</pre>
6941
6942<h5>Overview:</h5>
6943
6944<p>
6945The '<tt>llvm.var.annotation</tt>' intrinsic
6946</p>
6947
6948<h5>Arguments:</h5>
6949
6950<p>
Tanya Lattnerbed1d4d2007-06-18 23:42:37 +00006951The first argument is a pointer to a value, the second is a pointer to a
6952global string, the third is a pointer to a global string which is the source
6953file name, and the last argument is the line number.
Tanya Lattnercb1b9602007-06-15 20:50:54 +00006954</p>
6955
6956<h5>Semantics:</h5>
6957
6958<p>
Anton Korobeynikov06cbb652008-01-15 22:31:34 +00006959This intrinsic allows annotation of local variables with arbitrary strings.
Tanya Lattnercb1b9602007-06-15 20:50:54 +00006960This can be useful for special purpose optimizations that want to look for these
Anton Korobeynikov06cbb652008-01-15 22:31:34 +00006961annotations. These have no other defined use, they are ignored by code
6962generation and optimization.
6963</p>
Tanya Lattnercb1b9602007-06-15 20:50:54 +00006964</div>
6965
Tanya Lattner293c0372007-09-21 22:59:12 +00006966<!-- _______________________________________________________________________ -->
6967<div class="doc_subsubsection">
Tanya Lattner0186a652007-09-21 23:57:59 +00006968 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
Tanya Lattner293c0372007-09-21 22:59:12 +00006969</div>
6970
6971<div class="doc_text">
6972
6973<h5>Syntax:</h5>
Tanya Lattner23dbd572007-09-21 23:56:27 +00006974<p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6975any integer bit width.
6976</p>
Tanya Lattner293c0372007-09-21 22:59:12 +00006977<pre>
Tanya Lattnercf3e26f2007-09-22 00:03:01 +00006978 declare i8 @llvm.annotation.i8(i8 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6979 declare i16 @llvm.annotation.i16(i16 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6980 declare i32 @llvm.annotation.i32(i32 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6981 declare i64 @llvm.annotation.i64(i64 &lt;val&gt;, i8* &lt;str&gt;, i8* &lt;str&gt;, i32 &lt;int&gt; )
6982 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 +00006983</pre>
6984
6985<h5>Overview:</h5>
Tanya Lattner23dbd572007-09-21 23:56:27 +00006986
6987<p>
6988The '<tt>llvm.annotation</tt>' intrinsic.
Tanya Lattner293c0372007-09-21 22:59:12 +00006989</p>
6990
6991<h5>Arguments:</h5>
6992
6993<p>
6994The first argument is an integer value (result of some expression),
6995the second is a pointer to a global string, the third is a pointer to a global
6996string which is the source file name, and the last argument is the line number.
Tanya Lattner23dbd572007-09-21 23:56:27 +00006997It returns the value of the first argument.
Tanya Lattner293c0372007-09-21 22:59:12 +00006998</p>
6999
7000<h5>Semantics:</h5>
7001
7002<p>
7003This intrinsic allows annotations to be put on arbitrary expressions
7004with arbitrary strings. This can be useful for special purpose optimizations
7005that want to look for these annotations. These have no other defined use, they
7006are ignored by code generation and optimization.
Dan Gohmanef9462f2008-10-14 16:51:45 +00007007</p>
Tanya Lattner293c0372007-09-21 22:59:12 +00007008</div>
Jim Laskey2211f492007-03-14 19:31:19 +00007009
Anton Korobeynikov06cbb652008-01-15 22:31:34 +00007010<!-- _______________________________________________________________________ -->
7011<div class="doc_subsubsection">
7012 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
7013</div>
7014
7015<div class="doc_text">
7016
7017<h5>Syntax:</h5>
7018<pre>
7019 declare void @llvm.trap()
7020</pre>
7021
7022<h5>Overview:</h5>
7023
7024<p>
7025The '<tt>llvm.trap</tt>' intrinsic
7026</p>
7027
7028<h5>Arguments:</h5>
7029
7030<p>
7031None
7032</p>
7033
7034<h5>Semantics:</h5>
7035
7036<p>
7037This intrinsics is lowered to the target dependent trap instruction. If the
7038target does not have a trap instruction, this intrinsic will be lowered to the
7039call of the abort() function.
7040</p>
7041</div>
7042
Bill Wendling14313312008-11-19 05:56:17 +00007043<!-- _______________________________________________________________________ -->
7044<div class="doc_subsubsection">
Misha Brukman50de2b22008-11-22 23:55:29 +00007045 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
Bill Wendling14313312008-11-19 05:56:17 +00007046</div>
7047<div class="doc_text">
7048<h5>Syntax:</h5>
7049<pre>
7050declare void @llvm.stackprotector( i8* &lt;guard&gt;, i8** &lt;slot&gt; )
7051
7052</pre>
7053<h5>Overview:</h5>
7054<p>
7055 The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and stores
7056 it onto the stack at <tt>slot</tt>. The stack slot is adjusted to ensure that
7057 it is placed on the stack before local variables.
7058</p>
7059<h5>Arguments:</h5>
7060<p>
7061 The <tt>llvm.stackprotector</tt> intrinsic requires two pointer arguments. The
7062 first argument is the value loaded from the stack guard
7063 <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt> that
7064 has enough space to hold the value of the guard.
7065</p>
7066<h5>Semantics:</h5>
7067<p>
7068 This intrinsic causes the prologue/epilogue inserter to force the position of
7069 the <tt>AllocaInst</tt> stack slot to be before local variables on the
7070 stack. This is to ensure that if a local variable on the stack is overwritten,
7071 it will destroy the value of the guard. When the function exits, the guard on
7072 the stack is checked against the original guard. If they're different, then
7073 the program aborts by calling the <tt>__stack_chk_fail()</tt> function.
7074</p>
7075</div>
7076
Chris Lattner2f7c9632001-06-06 20:29:01 +00007077<!-- *********************************************************************** -->
Chris Lattner2f7c9632001-06-06 20:29:01 +00007078<hr>
Misha Brukmanc501f552004-03-01 17:47:27 +00007079<address>
7080 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman86242e12008-12-11 17:34:48 +00007081 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Misha Brukmanc501f552004-03-01 17:47:27 +00007082 <a href="http://validator.w3.org/check/referer"><img
Misha Brukman86242e12008-12-11 17:34:48 +00007083 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
Misha Brukmanc501f552004-03-01 17:47:27 +00007084
7085 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
Reid Spencerca058542006-03-14 05:39:39 +00007086 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
Misha Brukmanc501f552004-03-01 17:47:27 +00007087 Last modified: $Date$
7088</address>
Chris Lattnerb8f816e2008-01-04 04:33:49 +00007089
Misha Brukman76307852003-11-08 01:05:38 +00007090</body>
7091</html>