1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>LLVM Assembly Language Reference Manual</title>
6 <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.">
10 <link rel="stylesheet" href="llvm.css" type="text/css">
15 <h1>LLVM Language Reference Manual</h1>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
20 <li><a href="#highlevel">High Level Structure</a>
22 <li><a href="#modulestructure">Module Structure</a></li>
23 <li><a href="#linkage">Linkage Types</a>
25 <li><a href="#linkage_private">'<tt>private</tt>' Linkage</a></li>
26 <li><a href="#linkage_linker_private">'<tt>linker_private</tt>' Linkage</a></li>
27 <li><a href="#linkage_linker_private_weak">'<tt>linker_private_weak</tt>' Linkage</a></li>
28 <li><a href="#linkage_linker_private_weak_def_auto">'<tt>linker_private_weak_def_auto</tt>' Linkage</a></li>
29 <li><a href="#linkage_internal">'<tt>internal</tt>' Linkage</a></li>
30 <li><a href="#linkage_available_externally">'<tt>available_externally</tt>' Linkage</a></li>
31 <li><a href="#linkage_linkonce">'<tt>linkonce</tt>' Linkage</a></li>
32 <li><a href="#linkage_common">'<tt>common</tt>' Linkage</a></li>
33 <li><a href="#linkage_weak">'<tt>weak</tt>' Linkage</a></li>
34 <li><a href="#linkage_appending">'<tt>appending</tt>' Linkage</a></li>
35 <li><a href="#linkage_externweak">'<tt>extern_weak</tt>' Linkage</a></li>
36 <li><a href="#linkage_linkonce_odr">'<tt>linkonce_odr</tt>' Linkage</a></li>
37 <li><a href="#linkage_weak">'<tt>weak_odr</tt>' Linkage</a></li>
38 <li><a href="#linkage_external">'<tt>externally visible</tt>' Linkage</a></li>
39 <li><a href="#linkage_dllimport">'<tt>dllimport</tt>' Linkage</a></li>
40 <li><a href="#linkage_dllexport">'<tt>dllexport</tt>' Linkage</a></li>
43 <li><a href="#callingconv">Calling Conventions</a></li>
44 <li><a href="#namedtypes">Named Types</a></li>
45 <li><a href="#globalvars">Global Variables</a></li>
46 <li><a href="#functionstructure">Functions</a></li>
47 <li><a href="#aliasstructure">Aliases</a></li>
48 <li><a href="#namedmetadatastructure">Named Metadata</a></li>
49 <li><a href="#paramattrs">Parameter Attributes</a></li>
50 <li><a href="#fnattrs">Function Attributes</a></li>
51 <li><a href="#gc">Garbage Collector Names</a></li>
52 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
53 <li><a href="#datalayout">Data Layout</a></li>
54 <li><a href="#pointeraliasing">Pointer Aliasing Rules</a></li>
55 <li><a href="#volatile">Volatile Memory Accesses</a></li>
58 <li><a href="#typesystem">Type System</a>
60 <li><a href="#t_classifications">Type Classifications</a></li>
61 <li><a href="#t_primitive">Primitive Types</a>
63 <li><a href="#t_integer">Integer Type</a></li>
64 <li><a href="#t_floating">Floating Point Types</a></li>
65 <li><a href="#t_x86mmx">X86mmx Type</a></li>
66 <li><a href="#t_void">Void Type</a></li>
67 <li><a href="#t_label">Label Type</a></li>
68 <li><a href="#t_metadata">Metadata Type</a></li>
71 <li><a href="#t_derived">Derived Types</a>
73 <li><a href="#t_aggregate">Aggregate Types</a>
75 <li><a href="#t_array">Array Type</a></li>
76 <li><a href="#t_struct">Structure Type</a></li>
77 <li><a href="#t_pstruct">Packed Structure Type</a></li>
78 <li><a href="#t_vector">Vector Type</a></li>
81 <li><a href="#t_function">Function Type</a></li>
82 <li><a href="#t_pointer">Pointer Type</a></li>
83 <li><a href="#t_opaque">Opaque Type</a></li>
86 <li><a href="#t_uprefs">Type Up-references</a></li>
89 <li><a href="#constants">Constants</a>
91 <li><a href="#simpleconstants">Simple Constants</a></li>
92 <li><a href="#complexconstants">Complex Constants</a></li>
93 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
94 <li><a href="#undefvalues">Undefined Values</a></li>
95 <li><a href="#trapvalues">Trap Values</a></li>
96 <li><a href="#blockaddress">Addresses of Basic Blocks</a></li>
97 <li><a href="#constantexprs">Constant Expressions</a></li>
100 <li><a href="#othervalues">Other Values</a>
102 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
103 <li><a href="#metadata">Metadata Nodes and Metadata Strings</a></li>
106 <li><a href="#intrinsic_globals">Intrinsic Global Variables</a>
108 <li><a href="#intg_used">The '<tt>llvm.used</tt>' Global Variable</a></li>
109 <li><a href="#intg_compiler_used">The '<tt>llvm.compiler.used</tt>'
110 Global Variable</a></li>
111 <li><a href="#intg_global_ctors">The '<tt>llvm.global_ctors</tt>'
112 Global Variable</a></li>
113 <li><a href="#intg_global_dtors">The '<tt>llvm.global_dtors</tt>'
114 Global Variable</a></li>
117 <li><a href="#instref">Instruction Reference</a>
119 <li><a href="#terminators">Terminator Instructions</a>
121 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
122 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
123 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
124 <li><a href="#i_indirectbr">'<tt>indirectbr</tt>' Instruction</a></li>
125 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
126 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
127 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
130 <li><a href="#binaryops">Binary Operations</a>
132 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
133 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
134 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
135 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
136 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
137 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
138 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
139 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
140 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
141 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
142 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
143 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
146 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
148 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
149 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
150 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
151 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
152 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
153 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
156 <li><a href="#vectorops">Vector Operations</a>
158 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
159 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
160 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
163 <li><a href="#aggregateops">Aggregate Operations</a>
165 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
166 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
169 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
171 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
172 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
173 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
174 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
177 <li><a href="#convertops">Conversion Operations</a>
179 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
180 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
181 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
182 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
183 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
184 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
185 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
186 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
187 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
188 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
189 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
190 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
193 <li><a href="#otherops">Other Operations</a>
195 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
196 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
197 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
198 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
199 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
200 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
205 <li><a href="#intrinsics">Intrinsic Functions</a>
207 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
209 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
210 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
211 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
214 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
216 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
217 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
218 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
221 <li><a href="#int_codegen">Code Generator Intrinsics</a>
223 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
224 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
225 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
226 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
227 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
228 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
229 <li><a href="#int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
232 <li><a href="#int_libc">Standard C Library Intrinsics</a>
234 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
235 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
236 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
237 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
238 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
239 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
240 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
241 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
242 <li><a href="#int_exp">'<tt>llvm.exp.*</tt>' Intrinsic</a></li>
243 <li><a href="#int_log">'<tt>llvm.log.*</tt>' Intrinsic</a></li>
246 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
248 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
249 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
250 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
251 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
254 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
256 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
257 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
258 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
259 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
260 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
261 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
264 <li><a href="#int_fp16">Half Precision Floating Point Intrinsics</a>
266 <li><a href="#int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a></li>
267 <li><a href="#int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a></li>
270 <li><a href="#int_debugger">Debugger intrinsics</a></li>
271 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
272 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
274 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
277 <li><a href="#int_atomics">Atomic intrinsics</a>
279 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
280 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
281 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
282 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
283 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
284 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
285 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
286 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
287 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
288 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
289 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
290 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
291 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
294 <li><a href="#int_memorymarkers">Memory Use Markers</a>
296 <li><a href="#int_lifetime_start"><tt>llvm.lifetime.start</tt></a></li>
297 <li><a href="#int_lifetime_end"><tt>llvm.lifetime.end</tt></a></li>
298 <li><a href="#int_invariant_start"><tt>llvm.invariant.start</tt></a></li>
299 <li><a href="#int_invariant_end"><tt>llvm.invariant.end</tt></a></li>
302 <li><a href="#int_general">General intrinsics</a>
304 <li><a href="#int_var_annotation">
305 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
306 <li><a href="#int_annotation">
307 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
308 <li><a href="#int_trap">
309 '<tt>llvm.trap</tt>' Intrinsic</a></li>
310 <li><a href="#int_stackprotector">
311 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
312 <li><a href="#int_objectsize">
313 '<tt>llvm.objectsize</tt>' Intrinsic</a></li>
320 <div class="doc_author">
321 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
322 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
325 <!-- *********************************************************************** -->
326 <h2><a name="abstract">Abstract</a></h2>
327 <!-- *********************************************************************** -->
331 <p>This document is a reference manual for the LLVM assembly language. LLVM is
332 a Static Single Assignment (SSA) based representation that provides type
333 safety, low-level operations, flexibility, and the capability of representing
334 'all' high-level languages cleanly. It is the common code representation
335 used throughout all phases of the LLVM compilation strategy.</p>
339 <!-- *********************************************************************** -->
340 <h2><a name="introduction">Introduction</a></h2>
341 <!-- *********************************************************************** -->
345 <p>The LLVM code representation is designed to be used in three different forms:
346 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
347 for fast loading by a Just-In-Time compiler), and as a human readable
348 assembly language representation. This allows LLVM to provide a powerful
349 intermediate representation for efficient compiler transformations and
350 analysis, while providing a natural means to debug and visualize the
351 transformations. The three different forms of LLVM are all equivalent. This
352 document describes the human readable representation and notation.</p>
354 <p>The LLVM representation aims to be light-weight and low-level while being
355 expressive, typed, and extensible at the same time. It aims to be a
356 "universal IR" of sorts, by being at a low enough level that high-level ideas
357 may be cleanly mapped to it (similar to how microprocessors are "universal
358 IR's", allowing many source languages to be mapped to them). By providing
359 type information, LLVM can be used as the target of optimizations: for
360 example, through pointer analysis, it can be proven that a C automatic
361 variable is never accessed outside of the current function, allowing it to
362 be promoted to a simple SSA value instead of a memory location.</p>
364 <!-- _______________________________________________________________________ -->
366 <a name="wellformed">Well-Formedness</a>
371 <p>It is important to note that this document describes 'well formed' LLVM
372 assembly language. There is a difference between what the parser accepts and
373 what is considered 'well formed'. For example, the following instruction is
374 syntactically okay, but not well formed:</p>
376 <pre class="doc_code">
377 %x = <a href="#i_add">add</a> i32 1, %x
380 <p>because the definition of <tt>%x</tt> does not dominate all of its uses. The
381 LLVM infrastructure provides a verification pass that may be used to verify
382 that an LLVM module is well formed. This pass is automatically run by the
383 parser after parsing input assembly and by the optimizer before it outputs
384 bitcode. The violations pointed out by the verifier pass indicate bugs in
385 transformation passes or input to the parser.</p>
391 <!-- Describe the typesetting conventions here. -->
393 <!-- *********************************************************************** -->
394 <h2><a name="identifiers">Identifiers</a></h2>
395 <!-- *********************************************************************** -->
399 <p>LLVM identifiers come in two basic types: global and local. Global
400 identifiers (functions, global variables) begin with the <tt>'@'</tt>
401 character. Local identifiers (register names, types) begin with
402 the <tt>'%'</tt> character. Additionally, there are three different formats
403 for identifiers, for different purposes:</p>
406 <li>Named values are represented as a string of characters with their prefix.
407 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
408 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
409 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
410 other characters in their names can be surrounded with quotes. Special
411 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
412 ASCII code for the character in hexadecimal. In this way, any character
413 can be used in a name value, even quotes themselves.</li>
415 <li>Unnamed values are represented as an unsigned numeric value with their
416 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
418 <li>Constants, which are described in a <a href="#constants">section about
419 constants</a>, below.</li>
422 <p>LLVM requires that values start with a prefix for two reasons: Compilers
423 don't need to worry about name clashes with reserved words, and the set of
424 reserved words may be expanded in the future without penalty. Additionally,
425 unnamed identifiers allow a compiler to quickly come up with a temporary
426 variable without having to avoid symbol table conflicts.</p>
428 <p>Reserved words in LLVM are very similar to reserved words in other
429 languages. There are keywords for different opcodes
430 ('<tt><a href="#i_add">add</a></tt>',
431 '<tt><a href="#i_bitcast">bitcast</a></tt>',
432 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
433 ('<tt><a href="#t_void">void</a></tt>',
434 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
435 reserved words cannot conflict with variable names, because none of them
436 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
438 <p>Here is an example of LLVM code to multiply the integer variable
439 '<tt>%X</tt>' by 8:</p>
443 <pre class="doc_code">
444 %result = <a href="#i_mul">mul</a> i32 %X, 8
447 <p>After strength reduction:</p>
449 <pre class="doc_code">
450 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
453 <p>And the hard way:</p>
455 <pre class="doc_code">
456 %0 = <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
457 %1 = <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
458 %result = <a href="#i_add">add</a> i32 %1, %1
461 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
462 lexical features of LLVM:</p>
465 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
468 <li>Unnamed temporaries are created when the result of a computation is not
469 assigned to a named value.</li>
471 <li>Unnamed temporaries are numbered sequentially</li>
474 <p>It also shows a convention that we follow in this document. When
475 demonstrating instructions, we will follow an instruction with a comment that
476 defines the type and name of value produced. Comments are shown in italic
481 <!-- *********************************************************************** -->
482 <h2><a name="highlevel">High Level Structure</a></h2>
483 <!-- *********************************************************************** -->
485 <!-- ======================================================================= -->
487 <a name="modulestructure">Module Structure</a>
492 <p>LLVM programs are composed of "Module"s, each of which is a translation unit
493 of the input programs. Each module consists of functions, global variables,
494 and symbol table entries. Modules may be combined together with the LLVM
495 linker, which merges function (and global variable) definitions, resolves
496 forward declarations, and merges symbol table entries. Here is an example of
497 the "hello world" module:</p>
499 <pre class="doc_code">
500 <i>; Declare the string constant as a global constant.</i>
501 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
503 <i>; External declaration of the puts function</i>
504 <a href="#functionstructure">declare</a> i32 @puts(i8*) <i>; i32 (i8*)* </i>
506 <i>; Definition of main function</i>
507 define i32 @main() { <i>; i32()* </i>
508 <i>; Convert [13 x i8]* to i8 *...</i>
509 %cast210 = <a href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8*</i>
511 <i>; Call puts function to write out the string to stdout.</i>
512 <a href="#i_call">call</a> i32 @puts(i8* %cast210) <i>; i32</i>
513 <a href="#i_ret">ret</a> i32 0
516 <i>; Named metadata</i>
517 !1 = metadata !{i32 41}
521 <p>This example is made up of a <a href="#globalvars">global variable</a> named
522 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function,
523 a <a href="#functionstructure">function definition</a> for
524 "<tt>main</tt>" and <a href="#namedmetadatastructure">named metadata</a>
527 <p>In general, a module is made up of a list of global values, where both
528 functions and global variables are global values. Global values are
529 represented by a pointer to a memory location (in this case, a pointer to an
530 array of char, and a pointer to a function), and have one of the
531 following <a href="#linkage">linkage types</a>.</p>
535 <!-- ======================================================================= -->
537 <a name="linkage">Linkage Types</a>
542 <p>All Global Variables and Functions have one of the following types of
546 <dt><tt><b><a name="linkage_private">private</a></b></tt></dt>
547 <dd>Global values with "<tt>private</tt>" linkage are only directly accessible
548 by objects in the current module. In particular, linking code into a
549 module with an private global value may cause the private to be renamed as
550 necessary to avoid collisions. Because the symbol is private to the
551 module, all references can be updated. This doesn't show up in any symbol
552 table in the object file.</dd>
554 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt></dt>
555 <dd>Similar to <tt>private</tt>, but the symbol is passed through the
556 assembler and evaluated by the linker. Unlike normal strong symbols, they
557 are removed by the linker from the final linked image (executable or
558 dynamic library).</dd>
560 <dt><tt><b><a name="linkage_linker_private_weak">linker_private_weak</a></b></tt></dt>
561 <dd>Similar to "<tt>linker_private</tt>", but the symbol is weak. Note that
562 <tt>linker_private_weak</tt> symbols are subject to coalescing by the
563 linker. The symbols are removed by the linker from the final linked image
564 (executable or dynamic library).</dd>
566 <dt><tt><b><a name="linkage_linker_private_weak_def_auto">linker_private_weak_def_auto</a></b></tt></dt>
567 <dd>Similar to "<tt>linker_private_weak</tt>", but it's known that the address
568 of the object is not taken. For instance, functions that had an inline
569 definition, but the compiler decided not to inline it. Note,
570 unlike <tt>linker_private</tt> and <tt>linker_private_weak</tt>,
571 <tt>linker_private_weak_def_auto</tt> may have only <tt>default</tt>
572 visibility. The symbols are removed by the linker from the final linked
573 image (executable or dynamic library).</dd>
575 <dt><tt><b><a name="linkage_internal">internal</a></b></tt></dt>
576 <dd>Similar to private, but the value shows as a local symbol
577 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
578 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
580 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt></dt>
581 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
582 into the object file corresponding to the LLVM module. They exist to
583 allow inlining and other optimizations to take place given knowledge of
584 the definition of the global, which is known to be somewhere outside the
585 module. Globals with <tt>available_externally</tt> linkage are allowed to
586 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
587 This linkage type is only allowed on definitions, not declarations.</dd>
589 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt></dt>
590 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
591 the same name when linkage occurs. This can be used to implement
592 some forms of inline functions, templates, or other code which must be
593 generated in each translation unit that uses it, but where the body may
594 be overridden with a more definitive definition later. Unreferenced
595 <tt>linkonce</tt> globals are allowed to be discarded. Note that
596 <tt>linkonce</tt> linkage does not actually allow the optimizer to
597 inline the body of this function into callers because it doesn't know if
598 this definition of the function is the definitive definition within the
599 program or whether it will be overridden by a stronger definition.
600 To enable inlining and other optimizations, use "<tt>linkonce_odr</tt>"
603 <dt><tt><b><a name="linkage_weak">weak</a></b></tt></dt>
604 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
605 <tt>linkonce</tt> linkage, except that unreferenced globals with
606 <tt>weak</tt> linkage may not be discarded. This is used for globals that
607 are declared "weak" in C source code.</dd>
609 <dt><tt><b><a name="linkage_common">common</a></b></tt></dt>
610 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
611 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
613 Symbols with "<tt>common</tt>" linkage are merged in the same way as
614 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
615 <tt>common</tt> symbols may not have an explicit section,
616 must have a zero initializer, and may not be marked '<a
617 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
618 have common linkage.</dd>
621 <dt><tt><b><a name="linkage_appending">appending</a></b></tt></dt>
622 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
623 pointer to array type. When two global variables with appending linkage
624 are linked together, the two global arrays are appended together. This is
625 the LLVM, typesafe, equivalent of having the system linker append together
626 "sections" with identical names when .o files are linked.</dd>
628 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt></dt>
629 <dd>The semantics of this linkage follow the ELF object file model: the symbol
630 is weak until linked, if not linked, the symbol becomes null instead of
631 being an undefined reference.</dd>
633 <dt><tt><b><a name="linkage_linkonce_odr">linkonce_odr</a></b></tt></dt>
634 <dt><tt><b><a name="linkage_weak_odr">weak_odr</a></b></tt></dt>
635 <dd>Some languages allow differing globals to be merged, such as two functions
636 with different semantics. Other languages, such as <tt>C++</tt>, ensure
637 that only equivalent globals are ever merged (the "one definition rule"
638 — "ODR"). Such languages can use the <tt>linkonce_odr</tt>
639 and <tt>weak_odr</tt> linkage types to indicate that the global will only
640 be merged with equivalent globals. These linkage types are otherwise the
641 same as their non-<tt>odr</tt> versions.</dd>
643 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
644 <dd>If none of the above identifiers are used, the global is externally
645 visible, meaning that it participates in linkage and can be used to
646 resolve external symbol references.</dd>
649 <p>The next two types of linkage are targeted for Microsoft Windows platform
650 only. They are designed to support importing (exporting) symbols from (to)
651 DLLs (Dynamic Link Libraries).</p>
654 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt></dt>
655 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
656 or variable via a global pointer to a pointer that is set up by the DLL
657 exporting the symbol. On Microsoft Windows targets, the pointer name is
658 formed by combining <code>__imp_</code> and the function or variable
661 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt></dt>
662 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
663 pointer to a pointer in a DLL, so that it can be referenced with the
664 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
665 name is formed by combining <code>__imp_</code> and the function or
669 <p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
670 another module defined a "<tt>.LC0</tt>" variable and was linked with this
671 one, one of the two would be renamed, preventing a collision. Since
672 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
673 declarations), they are accessible outside of the current module.</p>
675 <p>It is illegal for a function <i>declaration</i> to have any linkage type
676 other than "externally visible", <tt>dllimport</tt>
677 or <tt>extern_weak</tt>.</p>
679 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
680 or <tt>weak_odr</tt> linkages.</p>
684 <!-- ======================================================================= -->
686 <a name="callingconv">Calling Conventions</a>
691 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
692 and <a href="#i_invoke">invokes</a> can all have an optional calling
693 convention specified for the call. The calling convention of any pair of
694 dynamic caller/callee must match, or the behavior of the program is
695 undefined. The following calling conventions are supported by LLVM, and more
696 may be added in the future:</p>
699 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
700 <dd>This calling convention (the default if no other calling convention is
701 specified) matches the target C calling conventions. This calling
702 convention supports varargs function calls and tolerates some mismatch in
703 the declared prototype and implemented declaration of the function (as
706 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
707 <dd>This calling convention attempts to make calls as fast as possible
708 (e.g. by passing things in registers). This calling convention allows the
709 target to use whatever tricks it wants to produce fast code for the
710 target, without having to conform to an externally specified ABI
711 (Application Binary Interface).
712 <a href="CodeGenerator.html#tailcallopt">Tail calls can only be optimized
713 when this or the GHC convention is used.</a> This calling convention
714 does not support varargs and requires the prototype of all callees to
715 exactly match the prototype of the function definition.</dd>
717 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
718 <dd>This calling convention attempts to make code in the caller as efficient
719 as possible under the assumption that the call is not commonly executed.
720 As such, these calls often preserve all registers so that the call does
721 not break any live ranges in the caller side. This calling convention
722 does not support varargs and requires the prototype of all callees to
723 exactly match the prototype of the function definition.</dd>
725 <dt><b>"<tt>cc <em>10</em></tt>" - GHC convention</b>:</dt>
726 <dd>This calling convention has been implemented specifically for use by the
727 <a href="http://www.haskell.org/ghc">Glasgow Haskell Compiler (GHC)</a>.
728 It passes everything in registers, going to extremes to achieve this by
729 disabling callee save registers. This calling convention should not be
730 used lightly but only for specific situations such as an alternative to
731 the <em>register pinning</em> performance technique often used when
732 implementing functional programming languages.At the moment only X86
733 supports this convention and it has the following limitations:
735 <li>On <em>X86-32</em> only supports up to 4 bit type parameters. No
736 floating point types are supported.</li>
737 <li>On <em>X86-64</em> only supports up to 10 bit type parameters and
738 6 floating point parameters.</li>
740 This calling convention supports
741 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> but
742 requires both the caller and callee are using it.
745 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
746 <dd>Any calling convention may be specified by number, allowing
747 target-specific calling conventions to be used. Target specific calling
748 conventions start at 64.</dd>
751 <p>More calling conventions can be added/defined on an as-needed basis, to
752 support Pascal conventions or any other well-known target-independent
757 <!-- ======================================================================= -->
759 <a name="visibility">Visibility Styles</a>
764 <p>All Global Variables and Functions have one of the following visibility
768 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
769 <dd>On targets that use the ELF object file format, default visibility means
770 that the declaration is visible to other modules and, in shared libraries,
771 means that the declared entity may be overridden. On Darwin, default
772 visibility means that the declaration is visible to other modules. Default
773 visibility corresponds to "external linkage" in the language.</dd>
775 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
776 <dd>Two declarations of an object with hidden visibility refer to the same
777 object if they are in the same shared object. Usually, hidden visibility
778 indicates that the symbol will not be placed into the dynamic symbol
779 table, so no other module (executable or shared library) can reference it
782 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
783 <dd>On ELF, protected visibility indicates that the symbol will be placed in
784 the dynamic symbol table, but that references within the defining module
785 will bind to the local symbol. That is, the symbol cannot be overridden by
791 <!-- ======================================================================= -->
793 <a name="namedtypes">Named Types</a>
798 <p>LLVM IR allows you to specify name aliases for certain types. This can make
799 it easier to read the IR and make the IR more condensed (particularly when
800 recursive types are involved). An example of a name specification is:</p>
802 <pre class="doc_code">
803 %mytype = type { %mytype*, i32 }
806 <p>You may give a name to any <a href="#typesystem">type</a> except
807 "<a href="#t_void">void</a>". Type name aliases may be used anywhere a type
808 is expected with the syntax "%mytype".</p>
810 <p>Note that type names are aliases for the structural type that they indicate,
811 and that you can therefore specify multiple names for the same type. This
812 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
813 uses structural typing, the name is not part of the type. When printing out
814 LLVM IR, the printer will pick <em>one name</em> to render all types of a
815 particular shape. This means that if you have code where two different
816 source types end up having the same LLVM type, that the dumper will sometimes
817 print the "wrong" or unexpected type. This is an important design point and
818 isn't going to change.</p>
822 <!-- ======================================================================= -->
824 <a name="globalvars">Global Variables</a>
829 <p>Global variables define regions of memory allocated at compilation time
830 instead of run-time. Global variables may optionally be initialized, may
831 have an explicit section to be placed in, and may have an optional explicit
832 alignment specified. A variable may be defined as "thread_local", which
833 means that it will not be shared by threads (each thread will have a
834 separated copy of the variable). A variable may be defined as a global
835 "constant," which indicates that the contents of the variable
836 will <b>never</b> be modified (enabling better optimization, allowing the
837 global data to be placed in the read-only section of an executable, etc).
838 Note that variables that need runtime initialization cannot be marked
839 "constant" as there is a store to the variable.</p>
841 <p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
842 constant, even if the final definition of the global is not. This capability
843 can be used to enable slightly better optimization of the program, but
844 requires the language definition to guarantee that optimizations based on the
845 'constantness' are valid for the translation units that do not include the
848 <p>As SSA values, global variables define pointer values that are in scope
849 (i.e. they dominate) all basic blocks in the program. Global variables
850 always define a pointer to their "content" type because they describe a
851 region of memory, and all memory objects in LLVM are accessed through
854 <p>Global variables can be marked with <tt>unnamed_addr</tt> which indicates
855 that the address is not significant, only the content. Constants marked
856 like this can be merged with other constants if they have the same
857 initializer. Note that a constant with significant address <em>can</em>
858 be merged with a <tt>unnamed_addr</tt> constant, the result being a
859 constant whose address is significant.</p>
861 <p>A global variable may be declared to reside in a target-specific numbered
862 address space. For targets that support them, address spaces may affect how
863 optimizations are performed and/or what target instructions are used to
864 access the variable. The default address space is zero. The address space
865 qualifier must precede any other attributes.</p>
867 <p>LLVM allows an explicit section to be specified for globals. If the target
868 supports it, it will emit globals to the section specified.</p>
870 <p>An explicit alignment may be specified for a global, which must be a power
871 of 2. If not present, or if the alignment is set to zero, the alignment of
872 the global is set by the target to whatever it feels convenient. If an
873 explicit alignment is specified, the global is forced to have exactly that
874 alignment. Targets and optimizers are not allowed to over-align the global
875 if the global has an assigned section. In this case, the extra alignment
876 could be observable: for example, code could assume that the globals are
877 densely packed in their section and try to iterate over them as an array,
878 alignment padding would break this iteration.</p>
880 <p>For example, the following defines a global in a numbered address space with
881 an initializer, section, and alignment:</p>
883 <pre class="doc_code">
884 @G = addrspace(5) constant float 1.0, section "foo", align 4
890 <!-- ======================================================================= -->
892 <a name="functionstructure">Functions</a>
897 <p>LLVM function definitions consist of the "<tt>define</tt>" keyword, an
898 optional <a href="#linkage">linkage type</a>, an optional
899 <a href="#visibility">visibility style</a>, an optional
900 <a href="#callingconv">calling convention</a>,
901 an optional <tt>unnamed_addr</tt> attribute, a return type, an optional
902 <a href="#paramattrs">parameter attribute</a> for the return type, a function
903 name, a (possibly empty) argument list (each with optional
904 <a href="#paramattrs">parameter attributes</a>), optional
905 <a href="#fnattrs">function attributes</a>, an optional section, an optional
906 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
907 curly brace, a list of basic blocks, and a closing curly brace.</p>
909 <p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
910 optional <a href="#linkage">linkage type</a>, an optional
911 <a href="#visibility">visibility style</a>, an optional
912 <a href="#callingconv">calling convention</a>,
913 an optional <tt>unnamed_addr</tt> attribute, a return type, an optional
914 <a href="#paramattrs">parameter attribute</a> for the return type, a function
915 name, a possibly empty list of arguments, an optional alignment, and an
916 optional <a href="#gc">garbage collector name</a>.</p>
918 <p>A function definition contains a list of basic blocks, forming the CFG
919 (Control Flow Graph) for the function. Each basic block may optionally start
920 with a label (giving the basic block a symbol table entry), contains a list
921 of instructions, and ends with a <a href="#terminators">terminator</a>
922 instruction (such as a branch or function return).</p>
924 <p>The first basic block in a function is special in two ways: it is immediately
925 executed on entrance to the function, and it is not allowed to have
926 predecessor basic blocks (i.e. there can not be any branches to the entry
927 block of a function). Because the block can have no predecessors, it also
928 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
930 <p>LLVM allows an explicit section to be specified for functions. If the target
931 supports it, it will emit functions to the section specified.</p>
933 <p>An explicit alignment may be specified for a function. If not present, or if
934 the alignment is set to zero, the alignment of the function is set by the
935 target to whatever it feels convenient. If an explicit alignment is
936 specified, the function is forced to have at least that much alignment. All
937 alignments must be a power of 2.</p>
939 <p>If the <tt>unnamed_addr</tt> attribute is given, the address is know to not
940 be significant and two identical functions can be merged</p>.
943 <pre class="doc_code">
944 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
945 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
946 <ResultType> @<FunctionName> ([argument list])
947 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
948 [<a href="#gc">gc</a>] { ... }
953 <!-- ======================================================================= -->
955 <a name="aliasstructure">Aliases</a>
960 <p>Aliases act as "second name" for the aliasee value (which can be either
961 function, global variable, another alias or bitcast of global value). Aliases
962 may have an optional <a href="#linkage">linkage type</a>, and an
963 optional <a href="#visibility">visibility style</a>.</p>
966 <pre class="doc_code">
967 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
972 <!-- ======================================================================= -->
974 <a name="namedmetadatastructure">Named Metadata</a>
979 <p>Named metadata is a collection of metadata. <a href="#metadata">Metadata
980 nodes</a> (but not metadata strings) are the only valid operands for
981 a named metadata.</p>
984 <pre class="doc_code">
985 ; Some unnamed metadata nodes, which are referenced by the named metadata.
986 !0 = metadata !{metadata !"zero"}
987 !1 = metadata !{metadata !"one"}
988 !2 = metadata !{metadata !"two"}
990 !name = !{!0, !1, !2}
995 <!-- ======================================================================= -->
997 <a name="paramattrs">Parameter Attributes</a>
1002 <p>The return type and each parameter of a function type may have a set of
1003 <i>parameter attributes</i> associated with them. Parameter attributes are
1004 used to communicate additional information about the result or parameters of
1005 a function. Parameter attributes are considered to be part of the function,
1006 not of the function type, so functions with different parameter attributes
1007 can have the same function type.</p>
1009 <p>Parameter attributes are simple keywords that follow the type specified. If
1010 multiple parameter attributes are needed, they are space separated. For
1013 <pre class="doc_code">
1014 declare i32 @printf(i8* noalias nocapture, ...)
1015 declare i32 @atoi(i8 zeroext)
1016 declare signext i8 @returns_signed_char()
1019 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
1020 <tt>readonly</tt>) come immediately after the argument list.</p>
1022 <p>Currently, only the following parameter attributes are defined:</p>
1025 <dt><tt><b>zeroext</b></tt></dt>
1026 <dd>This indicates to the code generator that the parameter or return value
1027 should be zero-extended to the extent required by the target's ABI (which
1028 is usually 32-bits, but is 8-bits for a i1 on x86-64) by the caller (for a
1029 parameter) or the callee (for a return value).</dd>
1031 <dt><tt><b>signext</b></tt></dt>
1032 <dd>This indicates to the code generator that the parameter or return value
1033 should be sign-extended to the extent required by the target's ABI (which
1034 is usually 32-bits) by the caller (for a parameter) or the callee (for a
1037 <dt><tt><b>inreg</b></tt></dt>
1038 <dd>This indicates that this parameter or return value should be treated in a
1039 special target-dependent fashion during while emitting code for a function
1040 call or return (usually, by putting it in a register as opposed to memory,
1041 though some targets use it to distinguish between two different kinds of
1042 registers). Use of this attribute is target-specific.</dd>
1044 <dt><tt><b><a name="byval">byval</a></b></tt></dt>
1045 <dd><p>This indicates that the pointer parameter should really be passed by
1046 value to the function. The attribute implies that a hidden copy of the
1048 is made between the caller and the callee, so the callee is unable to
1049 modify the value in the callee. This attribute is only valid on LLVM
1050 pointer arguments. It is generally used to pass structs and arrays by
1051 value, but is also valid on pointers to scalars. The copy is considered
1052 to belong to the caller not the callee (for example,
1053 <tt><a href="#readonly">readonly</a></tt> functions should not write to
1054 <tt>byval</tt> parameters). This is not a valid attribute for return
1057 <p>The byval attribute also supports specifying an alignment with
1058 the align attribute. It indicates the alignment of the stack slot to
1059 form and the known alignment of the pointer specified to the call site. If
1060 the alignment is not specified, then the code generator makes a
1061 target-specific assumption.</p></dd>
1063 <dt><tt><b><a name="sret">sret</a></b></tt></dt>
1064 <dd>This indicates that the pointer parameter specifies the address of a
1065 structure that is the return value of the function in the source program.
1066 This pointer must be guaranteed by the caller to be valid: loads and
1067 stores to the structure may be assumed by the callee to not to trap. This
1068 may only be applied to the first parameter. This is not a valid attribute
1069 for return values. </dd>
1071 <dt><tt><b><a name="noalias">noalias</a></b></tt></dt>
1072 <dd>This indicates that pointer values
1073 <a href="#pointeraliasing"><i>based</i></a> on the argument or return
1074 value do not alias pointer values which are not <i>based</i> on it,
1075 ignoring certain "irrelevant" dependencies.
1076 For a call to the parent function, dependencies between memory
1077 references from before or after the call and from those during the call
1078 are "irrelevant" to the <tt>noalias</tt> keyword for the arguments and
1079 return value used in that call.
1080 The caller shares the responsibility with the callee for ensuring that
1081 these requirements are met.
1082 For further details, please see the discussion of the NoAlias response in
1083 <a href="AliasAnalysis.html#MustMayNo">alias analysis</a>.<br>
1085 Note that this definition of <tt>noalias</tt> is intentionally
1086 similar to the definition of <tt>restrict</tt> in C99 for function
1087 arguments, though it is slightly weaker.
1089 For function return values, C99's <tt>restrict</tt> is not meaningful,
1090 while LLVM's <tt>noalias</tt> is.
1093 <dt><tt><b><a name="nocapture">nocapture</a></b></tt></dt>
1094 <dd>This indicates that the callee does not make any copies of the pointer
1095 that outlive the callee itself. This is not a valid attribute for return
1098 <dt><tt><b><a name="nest">nest</a></b></tt></dt>
1099 <dd>This indicates that the pointer parameter can be excised using the
1100 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
1101 attribute for return values.</dd>
1106 <!-- ======================================================================= -->
1108 <a name="gc">Garbage Collector Names</a>
1113 <p>Each function may specify a garbage collector name, which is simply a
1116 <pre class="doc_code">
1117 define void @f() gc "name" { ... }
1120 <p>The compiler declares the supported values of <i>name</i>. Specifying a
1121 collector which will cause the compiler to alter its output in order to
1122 support the named garbage collection algorithm.</p>
1126 <!-- ======================================================================= -->
1128 <a name="fnattrs">Function Attributes</a>
1133 <p>Function attributes are set to communicate additional information about a
1134 function. Function attributes are considered to be part of the function, not
1135 of the function type, so functions with different parameter attributes can
1136 have the same function type.</p>
1138 <p>Function attributes are simple keywords that follow the type specified. If
1139 multiple attributes are needed, they are space separated. For example:</p>
1141 <pre class="doc_code">
1142 define void @f() noinline { ... }
1143 define void @f() alwaysinline { ... }
1144 define void @f() alwaysinline optsize { ... }
1145 define void @f() optsize { ... }
1149 <dt><tt><b>alignstack(<<em>n</em>>)</b></tt></dt>
1150 <dd>This attribute indicates that, when emitting the prologue and epilogue,
1151 the backend should forcibly align the stack pointer. Specify the
1152 desired alignment, which must be a power of two, in parentheses.
1154 <dt><tt><b>alwaysinline</b></tt></dt>
1155 <dd>This attribute indicates that the inliner should attempt to inline this
1156 function into callers whenever possible, ignoring any active inlining size
1157 threshold for this caller.</dd>
1159 <dt><tt><b>hotpatch</b></tt></dt>
1160 <dd>This attribute indicates that the function should be 'hotpatchable',
1161 meaning the function can be patched and/or hooked even while it is
1162 loaded into memory. On x86, the function prologue will be preceded
1163 by six bytes of padding and will begin with a two-byte instruction.
1164 Most of the functions in the Windows system DLLs in Windows XP SP2 or
1165 higher were compiled in this fashion.</dd>
1167 <dt><tt><b>nonlazybind</b></tt></dt>
1168 <dd>This attribute suppresses lazy symbol binding for the function. This
1169 may make calls to the function faster, at the cost of extra program
1170 startup time if the function is not called during program startup.</dd>
1172 <dt><tt><b>inlinehint</b></tt></dt>
1173 <dd>This attribute indicates that the source code contained a hint that inlining
1174 this function is desirable (such as the "inline" keyword in C/C++). It
1175 is just a hint; it imposes no requirements on the inliner.</dd>
1177 <dt><tt><b>naked</b></tt></dt>
1178 <dd>This attribute disables prologue / epilogue emission for the function.
1179 This can have very system-specific consequences.</dd>
1181 <dt><tt><b>noimplicitfloat</b></tt></dt>
1182 <dd>This attributes disables implicit floating point instructions.</dd>
1184 <dt><tt><b>noinline</b></tt></dt>
1185 <dd>This attribute indicates that the inliner should never inline this
1186 function in any situation. This attribute may not be used together with
1187 the <tt>alwaysinline</tt> attribute.</dd>
1189 <dt><tt><b>noredzone</b></tt></dt>
1190 <dd>This attribute indicates that the code generator should not use a red
1191 zone, even if the target-specific ABI normally permits it.</dd>
1193 <dt><tt><b>noreturn</b></tt></dt>
1194 <dd>This function attribute indicates that the function never returns
1195 normally. This produces undefined behavior at runtime if the function
1196 ever does dynamically return.</dd>
1198 <dt><tt><b>nounwind</b></tt></dt>
1199 <dd>This function attribute indicates that the function never returns with an
1200 unwind or exceptional control flow. If the function does unwind, its
1201 runtime behavior is undefined.</dd>
1203 <dt><tt><b>optsize</b></tt></dt>
1204 <dd>This attribute suggests that optimization passes and code generator passes
1205 make choices that keep the code size of this function low, and otherwise
1206 do optimizations specifically to reduce code size.</dd>
1208 <dt><tt><b>readnone</b></tt></dt>
1209 <dd>This attribute indicates that the function computes its result (or decides
1210 to unwind an exception) based strictly on its arguments, without
1211 dereferencing any pointer arguments or otherwise accessing any mutable
1212 state (e.g. memory, control registers, etc) visible to caller functions.
1213 It does not write through any pointer arguments
1214 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1215 changes any state visible to callers. This means that it cannot unwind
1216 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1217 could use the <tt>unwind</tt> instruction.</dd>
1219 <dt><tt><b><a name="readonly">readonly</a></b></tt></dt>
1220 <dd>This attribute indicates that the function does not write through any
1221 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1222 arguments) or otherwise modify any state (e.g. memory, control registers,
1223 etc) visible to caller functions. It may dereference pointer arguments
1224 and read state that may be set in the caller. A readonly function always
1225 returns the same value (or unwinds an exception identically) when called
1226 with the same set of arguments and global state. It cannot unwind an
1227 exception by calling the <tt>C++</tt> exception throwing methods, but may
1228 use the <tt>unwind</tt> instruction.</dd>
1230 <dt><tt><b><a name="ssp">ssp</a></b></tt></dt>
1231 <dd>This attribute indicates that the function should emit a stack smashing
1232 protector. It is in the form of a "canary"—a random value placed on
1233 the stack before the local variables that's checked upon return from the
1234 function to see if it has been overwritten. A heuristic is used to
1235 determine if a function needs stack protectors or not.<br>
1237 If a function that has an <tt>ssp</tt> attribute is inlined into a
1238 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1239 function will have an <tt>ssp</tt> attribute.</dd>
1241 <dt><tt><b>sspreq</b></tt></dt>
1242 <dd>This attribute indicates that the function should <em>always</em> emit a
1243 stack smashing protector. This overrides
1244 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1246 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1247 function that doesn't have an <tt>sspreq</tt> attribute or which has
1248 an <tt>ssp</tt> attribute, then the resulting function will have
1249 an <tt>sspreq</tt> attribute.</dd>
1254 <!-- ======================================================================= -->
1256 <a name="moduleasm">Module-Level Inline Assembly</a>
1261 <p>Modules may contain "module-level inline asm" blocks, which corresponds to
1262 the GCC "file scope inline asm" blocks. These blocks are internally
1263 concatenated by LLVM and treated as a single unit, but may be separated in
1264 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
1266 <pre class="doc_code">
1267 module asm "inline asm code goes here"
1268 module asm "more can go here"
1271 <p>The strings can contain any character by escaping non-printable characters.
1272 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1275 <p>The inline asm code is simply printed to the machine code .s file when
1276 assembly code is generated.</p>
1280 <!-- ======================================================================= -->
1282 <a name="datalayout">Data Layout</a>
1287 <p>A module may specify a target specific data layout string that specifies how
1288 data is to be laid out in memory. The syntax for the data layout is
1291 <pre class="doc_code">
1292 target datalayout = "<i>layout specification</i>"
1295 <p>The <i>layout specification</i> consists of a list of specifications
1296 separated by the minus sign character ('-'). Each specification starts with
1297 a letter and may include other information after the letter to define some
1298 aspect of the data layout. The specifications accepted are as follows:</p>
1302 <dd>Specifies that the target lays out data in big-endian form. That is, the
1303 bits with the most significance have the lowest address location.</dd>
1306 <dd>Specifies that the target lays out data in little-endian form. That is,
1307 the bits with the least significance have the lowest address
1310 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1311 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1312 <i>preferred</i> alignments. All sizes are in bits. Specifying
1313 the <i>pref</i> alignment is optional. If omitted, the
1314 preceding <tt>:</tt> should be omitted too.</dd>
1316 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1317 <dd>This specifies the alignment for an integer type of a given bit
1318 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1320 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1321 <dd>This specifies the alignment for a vector type of a given bit
1324 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1325 <dd>This specifies the alignment for a floating point type of a given bit
1326 <i>size</i>. Only values of <i>size</i> that are supported by the target
1327 will work. 32 (float) and 64 (double) are supported on all targets;
1328 80 or 128 (different flavors of long double) are also supported on some
1331 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1332 <dd>This specifies the alignment for an aggregate type of a given bit
1335 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1336 <dd>This specifies the alignment for a stack object of a given bit
1339 <dt><tt>n<i>size1</i>:<i>size2</i>:<i>size3</i>...</tt></dt>
1340 <dd>This specifies a set of native integer widths for the target CPU
1341 in bits. For example, it might contain "n32" for 32-bit PowerPC,
1342 "n32:64" for PowerPC 64, or "n8:16:32:64" for X86-64. Elements of
1343 this set are considered to support most general arithmetic
1344 operations efficiently.</dd>
1347 <p>When constructing the data layout for a given target, LLVM starts with a
1348 default set of specifications which are then (possibly) overridden by the
1349 specifications in the <tt>datalayout</tt> keyword. The default specifications
1350 are given in this list:</p>
1353 <li><tt>E</tt> - big endian</li>
1354 <li><tt>p:64:64:64</tt> - 64-bit pointers with 64-bit alignment</li>
1355 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1356 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1357 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1358 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1359 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1360 alignment of 64-bits</li>
1361 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1362 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1363 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1364 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1365 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1366 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
1369 <p>When LLVM is determining the alignment for a given type, it uses the
1370 following rules:</p>
1373 <li>If the type sought is an exact match for one of the specifications, that
1374 specification is used.</li>
1376 <li>If no match is found, and the type sought is an integer type, then the
1377 smallest integer type that is larger than the bitwidth of the sought type
1378 is used. If none of the specifications are larger than the bitwidth then
1379 the the largest integer type is used. For example, given the default
1380 specifications above, the i7 type will use the alignment of i8 (next
1381 largest) while both i65 and i256 will use the alignment of i64 (largest
1384 <li>If no match is found, and the type sought is a vector type, then the
1385 largest vector type that is smaller than the sought vector type will be
1386 used as a fall back. This happens because <128 x double> can be
1387 implemented in terms of 64 <2 x double>, for example.</li>
1392 <!-- ======================================================================= -->
1394 <a name="pointeraliasing">Pointer Aliasing Rules</a>
1399 <p>Any memory access must be done through a pointer value associated
1400 with an address range of the memory access, otherwise the behavior
1401 is undefined. Pointer values are associated with address ranges
1402 according to the following rules:</p>
1405 <li>A pointer value is associated with the addresses associated with
1406 any value it is <i>based</i> on.
1407 <li>An address of a global variable is associated with the address
1408 range of the variable's storage.</li>
1409 <li>The result value of an allocation instruction is associated with
1410 the address range of the allocated storage.</li>
1411 <li>A null pointer in the default address-space is associated with
1413 <li>An integer constant other than zero or a pointer value returned
1414 from a function not defined within LLVM may be associated with address
1415 ranges allocated through mechanisms other than those provided by
1416 LLVM. Such ranges shall not overlap with any ranges of addresses
1417 allocated by mechanisms provided by LLVM.</li>
1420 <p>A pointer value is <i>based</i> on another pointer value according
1421 to the following rules:</p>
1424 <li>A pointer value formed from a
1425 <tt><a href="#i_getelementptr">getelementptr</a></tt> operation
1426 is <i>based</i> on the first operand of the <tt>getelementptr</tt>.</li>
1427 <li>The result value of a
1428 <tt><a href="#i_bitcast">bitcast</a></tt> is <i>based</i> on the operand
1429 of the <tt>bitcast</tt>.</li>
1430 <li>A pointer value formed by an
1431 <tt><a href="#i_inttoptr">inttoptr</a></tt> is <i>based</i> on all
1432 pointer values that contribute (directly or indirectly) to the
1433 computation of the pointer's value.</li>
1434 <li>The "<i>based</i> on" relationship is transitive.</li>
1437 <p>Note that this definition of <i>"based"</i> is intentionally
1438 similar to the definition of <i>"based"</i> in C99, though it is
1439 slightly weaker.</p>
1441 <p>LLVM IR does not associate types with memory. The result type of a
1442 <tt><a href="#i_load">load</a></tt> merely indicates the size and
1443 alignment of the memory from which to load, as well as the
1444 interpretation of the value. The first operand type of a
1445 <tt><a href="#i_store">store</a></tt> similarly only indicates the size
1446 and alignment of the store.</p>
1448 <p>Consequently, type-based alias analysis, aka TBAA, aka
1449 <tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1450 LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1451 additional information which specialized optimization passes may use
1452 to implement type-based alias analysis.</p>
1456 <!-- ======================================================================= -->
1458 <a name="volatile">Volatile Memory Accesses</a>
1463 <p>Certain memory accesses, such as <a href="#i_load"><tt>load</tt></a>s, <a
1464 href="#i_store"><tt>store</tt></a>s, and <a
1465 href="#int_memcpy"><tt>llvm.memcpy</tt></a>s may be marked <tt>volatile</tt>.
1466 The optimizers must not change the number of volatile operations or change their
1467 order of execution relative to other volatile operations. The optimizers
1468 <i>may</i> change the order of volatile operations relative to non-volatile
1469 operations. This is not Java's "volatile" and has no cross-thread
1470 synchronization behavior.</p>
1476 <!-- *********************************************************************** -->
1477 <h2><a name="typesystem">Type System</a></h2>
1478 <!-- *********************************************************************** -->
1482 <p>The LLVM type system is one of the most important features of the
1483 intermediate representation. Being typed enables a number of optimizations
1484 to be performed on the intermediate representation directly, without having
1485 to do extra analyses on the side before the transformation. A strong type
1486 system makes it easier to read the generated code and enables novel analyses
1487 and transformations that are not feasible to perform on normal three address
1488 code representations.</p>
1490 <!-- ======================================================================= -->
1492 <a name="t_classifications">Type Classifications</a>
1497 <p>The types fall into a few useful classifications:</p>
1499 <table border="1" cellspacing="0" cellpadding="4">
1501 <tr><th>Classification</th><th>Types</th></tr>
1503 <td><a href="#t_integer">integer</a></td>
1504 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1507 <td><a href="#t_floating">floating point</a></td>
1508 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1511 <td><a name="t_firstclass">first class</a></td>
1512 <td><a href="#t_integer">integer</a>,
1513 <a href="#t_floating">floating point</a>,
1514 <a href="#t_pointer">pointer</a>,
1515 <a href="#t_vector">vector</a>,
1516 <a href="#t_struct">structure</a>,
1517 <a href="#t_array">array</a>,
1518 <a href="#t_label">label</a>,
1519 <a href="#t_metadata">metadata</a>.
1523 <td><a href="#t_primitive">primitive</a></td>
1524 <td><a href="#t_label">label</a>,
1525 <a href="#t_void">void</a>,
1526 <a href="#t_integer">integer</a>,
1527 <a href="#t_floating">floating point</a>,
1528 <a href="#t_x86mmx">x86mmx</a>,
1529 <a href="#t_metadata">metadata</a>.</td>
1532 <td><a href="#t_derived">derived</a></td>
1533 <td><a href="#t_array">array</a>,
1534 <a href="#t_function">function</a>,
1535 <a href="#t_pointer">pointer</a>,
1536 <a href="#t_struct">structure</a>,
1537 <a href="#t_pstruct">packed structure</a>,
1538 <a href="#t_vector">vector</a>,
1539 <a href="#t_opaque">opaque</a>.
1545 <p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1546 important. Values of these types are the only ones which can be produced by
1551 <!-- ======================================================================= -->
1553 <a name="t_primitive">Primitive Types</a>
1558 <p>The primitive types are the fundamental building blocks of the LLVM
1561 <!-- _______________________________________________________________________ -->
1563 <a name="t_integer">Integer Type</a>
1569 <p>The integer type is a very simple type that simply specifies an arbitrary
1570 bit width for the integer type desired. Any bit width from 1 bit to
1571 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
1578 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1582 <table class="layout">
1584 <td class="left"><tt>i1</tt></td>
1585 <td class="left">a single-bit integer.</td>
1588 <td class="left"><tt>i32</tt></td>
1589 <td class="left">a 32-bit integer.</td>
1592 <td class="left"><tt>i1942652</tt></td>
1593 <td class="left">a really big integer of over 1 million bits.</td>
1599 <!-- _______________________________________________________________________ -->
1601 <a name="t_floating">Floating Point Types</a>
1608 <tr><th>Type</th><th>Description</th></tr>
1609 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1610 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1611 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1612 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1613 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1619 <!-- _______________________________________________________________________ -->
1621 <a name="t_x86mmx">X86mmx Type</a>
1627 <p>The x86mmx type represents a value held in an MMX register on an x86 machine. The operations allowed on it are quite limited: parameters and return values, load and store, and bitcast. User-specified MMX instructions are represented as intrinsic or asm calls with arguments and/or results of this type. There are no arrays, vectors or constants of this type.</p>
1636 <!-- _______________________________________________________________________ -->
1638 <a name="t_void">Void Type</a>
1644 <p>The void type does not represent any value and has no size.</p>
1653 <!-- _______________________________________________________________________ -->
1655 <a name="t_label">Label Type</a>
1661 <p>The label type represents code labels.</p>
1670 <!-- _______________________________________________________________________ -->
1672 <a name="t_metadata">Metadata Type</a>
1678 <p>The metadata type represents embedded metadata. No derived types may be
1679 created from metadata except for <a href="#t_function">function</a>
1691 <!-- ======================================================================= -->
1693 <a name="t_derived">Derived Types</a>
1698 <p>The real power in LLVM comes from the derived types in the system. This is
1699 what allows a programmer to represent arrays, functions, pointers, and other
1700 useful types. Each of these types contain one or more element types which
1701 may be a primitive type, or another derived type. For example, it is
1702 possible to have a two dimensional array, using an array as the element type
1703 of another array.</p>
1706 <!-- _______________________________________________________________________ -->
1708 <a name="t_aggregate">Aggregate Types</a>
1713 <p>Aggregate Types are a subset of derived types that can contain multiple
1714 member types. <a href="#t_array">Arrays</a>,
1715 <a href="#t_struct">structs</a>, and <a href="#t_vector">vectors</a> are
1716 aggregate types.</p>
1720 <!-- _______________________________________________________________________ -->
1722 <a name="t_array">Array Type</a>
1728 <p>The array type is a very simple derived type that arranges elements
1729 sequentially in memory. The array type requires a size (number of elements)
1730 and an underlying data type.</p>
1734 [<# elements> x <elementtype>]
1737 <p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1738 be any type with a size.</p>
1741 <table class="layout">
1743 <td class="left"><tt>[40 x i32]</tt></td>
1744 <td class="left">Array of 40 32-bit integer values.</td>
1747 <td class="left"><tt>[41 x i32]</tt></td>
1748 <td class="left">Array of 41 32-bit integer values.</td>
1751 <td class="left"><tt>[4 x i8]</tt></td>
1752 <td class="left">Array of 4 8-bit integer values.</td>
1755 <p>Here are some examples of multidimensional arrays:</p>
1756 <table class="layout">
1758 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1759 <td class="left">3x4 array of 32-bit integer values.</td>
1762 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1763 <td class="left">12x10 array of single precision floating point values.</td>
1766 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1767 <td class="left">2x3x4 array of 16-bit integer values.</td>
1771 <p>There is no restriction on indexing beyond the end of the array implied by
1772 a static type (though there are restrictions on indexing beyond the bounds
1773 of an allocated object in some cases). This means that single-dimension
1774 'variable sized array' addressing can be implemented in LLVM with a zero
1775 length array type. An implementation of 'pascal style arrays' in LLVM could
1776 use the type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
1780 <!-- _______________________________________________________________________ -->
1782 <a name="t_function">Function Type</a>
1788 <p>The function type can be thought of as a function signature. It consists of
1789 a return type and a list of formal parameter types. The return type of a
1790 function type is a first class type or a void type.</p>
1794 <returntype> (<parameter list>)
1797 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1798 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1799 which indicates that the function takes a variable number of arguments.
1800 Variable argument functions can access their arguments with
1801 the <a href="#int_varargs">variable argument handling intrinsic</a>
1802 functions. '<tt><returntype></tt>' is any type except
1803 <a href="#t_label">label</a>.</p>
1806 <table class="layout">
1808 <td class="left"><tt>i32 (i32)</tt></td>
1809 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1811 </tr><tr class="layout">
1812 <td class="left"><tt>float (i16, i32 *) *
1814 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1815 an <tt>i16</tt> and a <a href="#t_pointer">pointer</a> to <tt>i32</tt>,
1816 returning <tt>float</tt>.
1818 </tr><tr class="layout">
1819 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1820 <td class="left">A vararg function that takes at least one
1821 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1822 which returns an integer. This is the signature for <tt>printf</tt> in
1825 </tr><tr class="layout">
1826 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1827 <td class="left">A function taking an <tt>i32</tt>, returning a
1828 <a href="#t_struct">structure</a> containing two <tt>i32</tt> values
1835 <!-- _______________________________________________________________________ -->
1837 <a name="t_struct">Structure Type</a>
1843 <p>The structure type is used to represent a collection of data members together
1844 in memory. The packing of the field types is defined to match the ABI of the
1845 underlying processor. The elements of a structure may be any type that has a
1848 <p>Structures in memory are accessed using '<tt><a href="#i_load">load</a></tt>'
1849 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field
1850 with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
1851 Structures in registers are accessed using the
1852 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' and
1853 '<tt><a href="#i_insertvalue">insertvalue</a></tt>' instructions.</p>
1856 { <type list> }
1860 <table class="layout">
1862 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1863 <td class="left">A triple of three <tt>i32</tt> values</td>
1864 </tr><tr class="layout">
1865 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1866 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1867 second element is a <a href="#t_pointer">pointer</a> to a
1868 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1869 an <tt>i32</tt>.</td>
1875 <!-- _______________________________________________________________________ -->
1877 <a name="t_pstruct">Packed Structure Type</a>
1883 <p>The packed structure type is used to represent a collection of data members
1884 together in memory. There is no padding between fields. Further, the
1885 alignment of a packed structure is 1 byte. The elements of a packed
1886 structure may be any type that has a size.</p>
1888 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1889 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1890 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1894 < { <type list> } >
1898 <table class="layout">
1900 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1901 <td class="left">A triple of three <tt>i32</tt> values</td>
1902 </tr><tr class="layout">
1904 <tt>< { float, i32 (i32)* } ></tt></td>
1905 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1906 second element is a <a href="#t_pointer">pointer</a> to a
1907 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1908 an <tt>i32</tt>.</td>
1914 <!-- _______________________________________________________________________ -->
1916 <a name="t_pointer">Pointer Type</a>
1922 <p>The pointer type is used to specify memory locations.
1923 Pointers are commonly used to reference objects in memory.</p>
1925 <p>Pointer types may have an optional address space attribute defining the
1926 numbered address space where the pointed-to object resides. The default
1927 address space is number zero. The semantics of non-zero address
1928 spaces are target-specific.</p>
1930 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
1931 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
1939 <table class="layout">
1941 <td class="left"><tt>[4 x i32]*</tt></td>
1942 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1943 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1946 <td class="left"><tt>i32 (i32*) *</tt></td>
1947 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1948 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1952 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1953 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1954 that resides in address space #5.</td>
1960 <!-- _______________________________________________________________________ -->
1962 <a name="t_vector">Vector Type</a>
1968 <p>A vector type is a simple derived type that represents a vector of elements.
1969 Vector types are used when multiple primitive data are operated in parallel
1970 using a single instruction (SIMD). A vector type requires a size (number of
1971 elements) and an underlying primitive data type. Vector types are considered
1972 <a href="#t_firstclass">first class</a>.</p>
1976 < <# elements> x <elementtype> >
1979 <p>The number of elements is a constant integer value larger than 0; elementtype
1980 may be any integer or floating point type. Vectors of size zero are not
1981 allowed, and pointers are not allowed as the element type.</p>
1984 <table class="layout">
1986 <td class="left"><tt><4 x i32></tt></td>
1987 <td class="left">Vector of 4 32-bit integer values.</td>
1990 <td class="left"><tt><8 x float></tt></td>
1991 <td class="left">Vector of 8 32-bit floating-point values.</td>
1994 <td class="left"><tt><2 x i64></tt></td>
1995 <td class="left">Vector of 2 64-bit integer values.</td>
2001 <!-- _______________________________________________________________________ -->
2003 <a name="t_opaque">Opaque Type</a>
2009 <p>Opaque types are used to represent unknown types in the system. This
2010 corresponds (for example) to the C notion of a forward declared structure
2011 type. In LLVM, opaque types can eventually be resolved to any type (not just
2012 a structure type).</p>
2020 <table class="layout">
2022 <td class="left"><tt>opaque</tt></td>
2023 <td class="left">An opaque type.</td>
2031 <!-- ======================================================================= -->
2033 <a name="t_uprefs">Type Up-references</a>
2039 <p>An "up reference" allows you to refer to a lexically enclosing type without
2040 requiring it to have a name. For instance, a structure declaration may
2041 contain a pointer to any of the types it is lexically a member of. Example
2042 of up references (with their equivalent as named type declarations)
2046 { \2 * } %x = type { %x* }
2047 { \2 }* %y = type { %y }*
2051 <p>An up reference is needed by the asmprinter for printing out cyclic types
2052 when there is no declared name for a type in the cycle. Because the
2053 asmprinter does not want to print out an infinite type string, it needs a
2054 syntax to handle recursive types that have no names (all names are optional
2062 <p>The level is the count of the lexical type that is being referred to.</p>
2065 <table class="layout">
2067 <td class="left"><tt>\1*</tt></td>
2068 <td class="left">Self-referential pointer.</td>
2071 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
2072 <td class="left">Recursive structure where the upref refers to the out-most
2081 <!-- *********************************************************************** -->
2082 <h2><a name="constants">Constants</a></h2>
2083 <!-- *********************************************************************** -->
2087 <p>LLVM has several different basic types of constants. This section describes
2088 them all and their syntax.</p>
2090 <!-- ======================================================================= -->
2092 <a name="simpleconstants">Simple Constants</a>
2098 <dt><b>Boolean constants</b></dt>
2099 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
2100 constants of the <tt><a href="#t_integer">i1</a></tt> type.</dd>
2102 <dt><b>Integer constants</b></dt>
2103 <dd>Standard integers (such as '4') are constants of
2104 the <a href="#t_integer">integer</a> type. Negative numbers may be used
2105 with integer types.</dd>
2107 <dt><b>Floating point constants</b></dt>
2108 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
2109 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
2110 notation (see below). The assembler requires the exact decimal value of a
2111 floating-point constant. For example, the assembler accepts 1.25 but
2112 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
2113 constants must have a <a href="#t_floating">floating point</a> type. </dd>
2115 <dt><b>Null pointer constants</b></dt>
2116 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
2117 and must be of <a href="#t_pointer">pointer type</a>.</dd>
2120 <p>The one non-intuitive notation for constants is the hexadecimal form of
2121 floating point constants. For example, the form '<tt>double
2122 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
2123 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
2124 constants are required (and the only time that they are generated by the
2125 disassembler) is when a floating point constant must be emitted but it cannot
2126 be represented as a decimal floating point number in a reasonable number of
2127 digits. For example, NaN's, infinities, and other special values are
2128 represented in their IEEE hexadecimal format so that assembly and disassembly
2129 do not cause any bits to change in the constants.</p>
2131 <p>When using the hexadecimal form, constants of types float and double are
2132 represented using the 16-digit form shown above (which matches the IEEE754
2133 representation for double); float values must, however, be exactly
2134 representable as IEE754 single precision. Hexadecimal format is always used
2135 for long double, and there are three forms of long double. The 80-bit format
2136 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
2137 The 128-bit format used by PowerPC (two adjacent doubles) is represented
2138 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
2139 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
2140 currently supported target uses this format. Long doubles will only work if
2141 they match the long double format on your target. All hexadecimal formats
2142 are big-endian (sign bit at the left).</p>
2144 <p>There are no constants of type x86mmx.</p>
2147 <!-- ======================================================================= -->
2149 <a name="aggregateconstants"></a> <!-- old anchor -->
2150 <a name="complexconstants">Complex Constants</a>
2155 <p>Complex constants are a (potentially recursive) combination of simple
2156 constants and smaller complex constants.</p>
2159 <dt><b>Structure constants</b></dt>
2160 <dd>Structure constants are represented with notation similar to structure
2161 type definitions (a comma separated list of elements, surrounded by braces
2162 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
2163 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
2164 Structure constants must have <a href="#t_struct">structure type</a>, and
2165 the number and types of elements must match those specified by the
2168 <dt><b>Array constants</b></dt>
2169 <dd>Array constants are represented with notation similar to array type
2170 definitions (a comma separated list of elements, surrounded by square
2171 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
2172 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
2173 the number and types of elements must match those specified by the
2176 <dt><b>Vector constants</b></dt>
2177 <dd>Vector constants are represented with notation similar to vector type
2178 definitions (a comma separated list of elements, surrounded by
2179 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32
2180 42, i32 11, i32 74, i32 100 ></tt>". Vector constants must
2181 have <a href="#t_vector">vector type</a>, and the number and types of
2182 elements must match those specified by the type.</dd>
2184 <dt><b>Zero initialization</b></dt>
2185 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
2186 value to zero of <em>any</em> type, including scalar and
2187 <a href="#t_aggregate">aggregate</a> types.
2188 This is often used to avoid having to print large zero initializers
2189 (e.g. for large arrays) and is always exactly equivalent to using explicit
2190 zero initializers.</dd>
2192 <dt><b>Metadata node</b></dt>
2193 <dd>A metadata node is a structure-like constant with
2194 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
2195 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
2196 be interpreted as part of the instruction stream, metadata is a place to
2197 attach additional information such as debug info.</dd>
2202 <!-- ======================================================================= -->
2204 <a name="globalconstants">Global Variable and Function Addresses</a>
2209 <p>The addresses of <a href="#globalvars">global variables</a>
2210 and <a href="#functionstructure">functions</a> are always implicitly valid
2211 (link-time) constants. These constants are explicitly referenced when
2212 the <a href="#identifiers">identifier for the global</a> is used and always
2213 have <a href="#t_pointer">pointer</a> type. For example, the following is a
2214 legal LLVM file:</p>
2216 <pre class="doc_code">
2219 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2224 <!-- ======================================================================= -->
2226 <a name="undefvalues">Undefined Values</a>
2231 <p>The string '<tt>undef</tt>' can be used anywhere a constant is expected, and
2232 indicates that the user of the value may receive an unspecified bit-pattern.
2233 Undefined values may be of any type (other than '<tt>label</tt>'
2234 or '<tt>void</tt>') and be used anywhere a constant is permitted.</p>
2236 <p>Undefined values are useful because they indicate to the compiler that the
2237 program is well defined no matter what value is used. This gives the
2238 compiler more freedom to optimize. Here are some examples of (potentially
2239 surprising) transformations that are valid (in pseudo IR):</p>
2242 <pre class="doc_code">
2252 <p>This is safe because all of the output bits are affected by the undef bits.
2253 Any output bit can have a zero or one depending on the input bits.</p>
2255 <pre class="doc_code">
2266 <p>These logical operations have bits that are not always affected by the input.
2267 For example, if <tt>%X</tt> has a zero bit, then the output of the
2268 '<tt>and</tt>' operation will always be a zero for that bit, no matter what
2269 the corresponding bit from the '<tt>undef</tt>' is. As such, it is unsafe to
2270 optimize or assume that the result of the '<tt>and</tt>' is '<tt>undef</tt>'.
2271 However, it is safe to assume that all bits of the '<tt>undef</tt>' could be
2272 0, and optimize the '<tt>and</tt>' to 0. Likewise, it is safe to assume that
2273 all the bits of the '<tt>undef</tt>' operand to the '<tt>or</tt>' could be
2274 set, allowing the '<tt>or</tt>' to be folded to -1.</p>
2276 <pre class="doc_code">
2277 %A = select undef, %X, %Y
2278 %B = select undef, 42, %Y
2279 %C = select %X, %Y, undef
2290 <p>This set of examples shows that undefined '<tt>select</tt>' (and conditional
2291 branch) conditions can go <em>either way</em>, but they have to come from one
2292 of the two operands. In the <tt>%A</tt> example, if <tt>%X</tt> and
2293 <tt>%Y</tt> were both known to have a clear low bit, then <tt>%A</tt> would
2294 have to have a cleared low bit. However, in the <tt>%C</tt> example, the
2295 optimizer is allowed to assume that the '<tt>undef</tt>' operand could be the
2296 same as <tt>%Y</tt>, allowing the whole '<tt>select</tt>' to be
2299 <pre class="doc_code">
2300 %A = xor undef, undef
2318 <p>This example points out that two '<tt>undef</tt>' operands are not
2319 necessarily the same. This can be surprising to people (and also matches C
2320 semantics) where they assume that "<tt>X^X</tt>" is always zero, even
2321 if <tt>X</tt> is undefined. This isn't true for a number of reasons, but the
2322 short answer is that an '<tt>undef</tt>' "variable" can arbitrarily change
2323 its value over its "live range". This is true because the variable doesn't
2324 actually <em>have a live range</em>. Instead, the value is logically read
2325 from arbitrary registers that happen to be around when needed, so the value
2326 is not necessarily consistent over time. In fact, <tt>%A</tt> and <tt>%C</tt>
2327 need to have the same semantics or the core LLVM "replace all uses with"
2328 concept would not hold.</p>
2330 <pre class="doc_code">
2338 <p>These examples show the crucial difference between an <em>undefined
2339 value</em> and <em>undefined behavior</em>. An undefined value (like
2340 '<tt>undef</tt>') is allowed to have an arbitrary bit-pattern. This means that
2341 the <tt>%A</tt> operation can be constant folded to '<tt>undef</tt>', because
2342 the '<tt>undef</tt>' could be an SNaN, and <tt>fdiv</tt> is not (currently)
2343 defined on SNaN's. However, in the second example, we can make a more
2344 aggressive assumption: because the <tt>undef</tt> is allowed to be an
2345 arbitrary value, we are allowed to assume that it could be zero. Since a
2346 divide by zero has <em>undefined behavior</em>, we are allowed to assume that
2347 the operation does not execute at all. This allows us to delete the divide and
2348 all code after it. Because the undefined operation "can't happen", the
2349 optimizer can assume that it occurs in dead code.</p>
2351 <pre class="doc_code">
2352 a: store undef -> %X
2353 b: store %X -> undef
2359 <p>These examples reiterate the <tt>fdiv</tt> example: a store <em>of</em> an
2360 undefined value can be assumed to not have any effect; we can assume that the
2361 value is overwritten with bits that happen to match what was already there.
2362 However, a store <em>to</em> an undefined location could clobber arbitrary
2363 memory, therefore, it has undefined behavior.</p>
2367 <!-- ======================================================================= -->
2369 <a name="trapvalues">Trap Values</a>
2374 <p>Trap values are similar to <a href="#undefvalues">undef values</a>, however
2375 instead of representing an unspecified bit pattern, they represent the
2376 fact that an instruction or constant expression which cannot evoke side
2377 effects has nevertheless detected a condition which results in undefined
2380 <p>There is currently no way of representing a trap value in the IR; they
2381 only exist when produced by operations such as
2382 <a href="#i_add"><tt>add</tt></a> with the <tt>nsw</tt> flag.</p>
2384 <p>Trap value behavior is defined in terms of value <i>dependence</i>:</p>
2387 <li>Values other than <a href="#i_phi"><tt>phi</tt></a> nodes depend on
2388 their operands.</li>
2390 <li><a href="#i_phi"><tt>Phi</tt></a> nodes depend on the operand corresponding
2391 to their dynamic predecessor basic block.</li>
2393 <li>Function arguments depend on the corresponding actual argument values in
2394 the dynamic callers of their functions.</li>
2396 <li><a href="#i_call"><tt>Call</tt></a> instructions depend on the
2397 <a href="#i_ret"><tt>ret</tt></a> instructions that dynamically transfer
2398 control back to them.</li>
2400 <li><a href="#i_invoke"><tt>Invoke</tt></a> instructions depend on the
2401 <a href="#i_ret"><tt>ret</tt></a>, <a href="#i_unwind"><tt>unwind</tt></a>,
2402 or exception-throwing call instructions that dynamically transfer control
2405 <li>Non-volatile loads and stores depend on the most recent stores to all of the
2406 referenced memory addresses, following the order in the IR
2407 (including loads and stores implied by intrinsics such as
2408 <a href="#int_memcpy"><tt>@llvm.memcpy</tt></a>.)</li>
2410 <!-- TODO: In the case of multiple threads, this only applies if the store
2411 "happens-before" the load or store. -->
2413 <!-- TODO: floating-point exception state -->
2415 <li>An instruction with externally visible side effects depends on the most
2416 recent preceding instruction with externally visible side effects, following
2417 the order in the IR. (This includes
2418 <a href="#volatile">volatile operations</a>.)</li>
2420 <li>An instruction <i>control-depends</i> on a
2421 <a href="#terminators">terminator instruction</a>
2422 if the terminator instruction has multiple successors and the instruction
2423 is always executed when control transfers to one of the successors, and
2424 may not be executed when control is transferred to another.</li>
2426 <li>Additionally, an instruction also <i>control-depends</i> on a terminator
2427 instruction if the set of instructions it otherwise depends on would be
2428 different if the terminator had transferred control to a different
2431 <li>Dependence is transitive.</li>
2435 <p>Whenever a trap value is generated, all values which depend on it evaluate
2436 to trap. If they have side effects, the evoke their side effects as if each
2437 operand with a trap value were undef. If they have externally-visible side
2438 effects, the behavior is undefined.</p>
2440 <p>Here are some examples:</p>
2442 <pre class="doc_code">
2444 %trap = sub nuw i32 0, 1 ; Results in a trap value.
2445 %still_trap = and i32 %trap, 0 ; Whereas (and i32 undef, 0) would return 0.
2446 %trap_yet_again = getelementptr i32* @h, i32 %still_trap
2447 store i32 0, i32* %trap_yet_again ; undefined behavior
2449 store i32 %trap, i32* @g ; Trap value conceptually stored to memory.
2450 %trap2 = load i32* @g ; Returns a trap value, not just undef.
2452 volatile store i32 %trap, i32* @g ; External observation; undefined behavior.
2454 %narrowaddr = bitcast i32* @g to i16*
2455 %wideaddr = bitcast i32* @g to i64*
2456 %trap3 = load i16* %narrowaddr ; Returns a trap value.
2457 %trap4 = load i64* %wideaddr ; Returns a trap value.
2459 %cmp = icmp slt i32 %trap, 0 ; Returns a trap value.
2460 br i1 %cmp, label %true, label %end ; Branch to either destination.
2463 volatile store i32 0, i32* @g ; This is control-dependent on %cmp, so
2464 ; it has undefined behavior.
2468 %p = phi i32 [ 0, %entry ], [ 1, %true ]
2469 ; Both edges into this PHI are
2470 ; control-dependent on %cmp, so this
2471 ; always results in a trap value.
2473 volatile store i32 0, i32* @g ; This would depend on the store in %true
2474 ; if %cmp is true, or the store in %entry
2475 ; otherwise, so this is undefined behavior.
2477 br i1 %cmp, label %second_true, label %second_end
2478 ; The same branch again, but this time the
2479 ; true block doesn't have side effects.
2486 volatile store i32 0, i32* @g ; This time, the instruction always depends
2487 ; on the store in %end. Also, it is
2488 ; control-equivalent to %end, so this is
2489 ; well-defined (again, ignoring earlier
2490 ; undefined behavior in this example).
2495 <!-- ======================================================================= -->
2497 <a name="blockaddress">Addresses of Basic Blocks</a>
2502 <p><b><tt>blockaddress(@function, %block)</tt></b></p>
2504 <p>The '<tt>blockaddress</tt>' constant computes the address of the specified
2505 basic block in the specified function, and always has an i8* type. Taking
2506 the address of the entry block is illegal.</p>
2508 <p>This value only has defined behavior when used as an operand to the
2509 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>' instruction, or for
2510 comparisons against null. Pointer equality tests between labels addresses
2511 results in undefined behavior — though, again, comparison against null
2512 is ok, and no label is equal to the null pointer. This may be passed around
2513 as an opaque pointer sized value as long as the bits are not inspected. This
2514 allows <tt>ptrtoint</tt> and arithmetic to be performed on these values so
2515 long as the original value is reconstituted before the <tt>indirectbr</tt>
2518 <p>Finally, some targets may provide defined semantics when using the value as
2519 the operand to an inline assembly, but that is target specific.</p>
2524 <!-- ======================================================================= -->
2526 <a name="constantexprs">Constant Expressions</a>
2531 <p>Constant expressions are used to allow expressions involving other constants
2532 to be used as constants. Constant expressions may be of
2533 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2534 operation that does not have side effects (e.g. load and call are not
2535 supported). The following is the syntax for constant expressions:</p>
2538 <dt><b><tt>trunc (CST to TYPE)</tt></b></dt>
2539 <dd>Truncate a constant to another type. The bit size of CST must be larger
2540 than the bit size of TYPE. Both types must be integers.</dd>
2542 <dt><b><tt>zext (CST to TYPE)</tt></b></dt>
2543 <dd>Zero extend a constant to another type. The bit size of CST must be
2544 smaller than the bit size of TYPE. Both types must be integers.</dd>
2546 <dt><b><tt>sext (CST to TYPE)</tt></b></dt>
2547 <dd>Sign extend a constant to another type. The bit size of CST must be
2548 smaller than the bit size of TYPE. Both types must be integers.</dd>
2550 <dt><b><tt>fptrunc (CST to TYPE)</tt></b></dt>
2551 <dd>Truncate a floating point constant to another floating point type. The
2552 size of CST must be larger than the size of TYPE. Both types must be
2553 floating point.</dd>
2555 <dt><b><tt>fpext (CST to TYPE)</tt></b></dt>
2556 <dd>Floating point extend a constant to another type. The size of CST must be
2557 smaller or equal to the size of TYPE. Both types must be floating
2560 <dt><b><tt>fptoui (CST to TYPE)</tt></b></dt>
2561 <dd>Convert a floating point constant to the corresponding unsigned integer
2562 constant. TYPE must be a scalar or vector integer type. CST must be of
2563 scalar or vector floating point type. Both CST and TYPE must be scalars,
2564 or vectors of the same number of elements. If the value won't fit in the
2565 integer type, the results are undefined.</dd>
2567 <dt><b><tt>fptosi (CST to TYPE)</tt></b></dt>
2568 <dd>Convert a floating point constant to the corresponding signed integer
2569 constant. TYPE must be a scalar or vector integer type. CST must be of
2570 scalar or vector floating point type. Both CST and TYPE must be scalars,
2571 or vectors of the same number of elements. If the value won't fit in the
2572 integer type, the results are undefined.</dd>
2574 <dt><b><tt>uitofp (CST to TYPE)</tt></b></dt>
2575 <dd>Convert an unsigned integer constant to the corresponding floating point
2576 constant. TYPE must be a scalar or vector floating point type. CST must be
2577 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2578 vectors of the same number of elements. If the value won't fit in the
2579 floating point type, the results are undefined.</dd>
2581 <dt><b><tt>sitofp (CST to TYPE)</tt></b></dt>
2582 <dd>Convert a signed integer constant to the corresponding floating point
2583 constant. TYPE must be a scalar or vector floating point type. CST must be
2584 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2585 vectors of the same number of elements. If the value won't fit in the
2586 floating point type, the results are undefined.</dd>
2588 <dt><b><tt>ptrtoint (CST to TYPE)</tt></b></dt>
2589 <dd>Convert a pointer typed constant to the corresponding integer constant
2590 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2591 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2592 make it fit in <tt>TYPE</tt>.</dd>
2594 <dt><b><tt>inttoptr (CST to TYPE)</tt></b></dt>
2595 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2596 type. CST must be of integer type. The CST value is zero extended,
2597 truncated, or unchanged to make it fit in a pointer size. This one is
2598 <i>really</i> dangerous!</dd>
2600 <dt><b><tt>bitcast (CST to TYPE)</tt></b></dt>
2601 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2602 are the same as those for the <a href="#i_bitcast">bitcast
2603 instruction</a>.</dd>
2605 <dt><b><tt>getelementptr (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2606 <dt><b><tt>getelementptr inbounds (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2607 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2608 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2609 instruction, the index list may have zero or more indexes, which are
2610 required to make sense for the type of "CSTPTR".</dd>
2612 <dt><b><tt>select (COND, VAL1, VAL2)</tt></b></dt>
2613 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
2615 <dt><b><tt>icmp COND (VAL1, VAL2)</tt></b></dt>
2616 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2618 <dt><b><tt>fcmp COND (VAL1, VAL2)</tt></b></dt>
2619 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2621 <dt><b><tt>extractelement (VAL, IDX)</tt></b></dt>
2622 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2625 <dt><b><tt>insertelement (VAL, ELT, IDX)</tt></b></dt>
2626 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2629 <dt><b><tt>shufflevector (VEC1, VEC2, IDXMASK)</tt></b></dt>
2630 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2633 <dt><b><tt>extractvalue (VAL, IDX0, IDX1, ...)</tt></b></dt>
2634 <dd>Perform the <a href="#i_extractvalue">extractvalue operation</a> on
2635 constants. The index list is interpreted in a similar manner as indices in
2636 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2637 index value must be specified.</dd>
2639 <dt><b><tt>insertvalue (VAL, ELT, IDX0, IDX1, ...)</tt></b></dt>
2640 <dd>Perform the <a href="#i_insertvalue">insertvalue operation</a> on
2641 constants. The index list is interpreted in a similar manner as indices in
2642 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2643 index value must be specified.</dd>
2645 <dt><b><tt>OPCODE (LHS, RHS)</tt></b></dt>
2646 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2647 be any of the <a href="#binaryops">binary</a>
2648 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2649 on operands are the same as those for the corresponding instruction
2650 (e.g. no bitwise operations on floating point values are allowed).</dd>
2657 <!-- *********************************************************************** -->
2658 <h2><a name="othervalues">Other Values</a></h2>
2659 <!-- *********************************************************************** -->
2661 <!-- ======================================================================= -->
2663 <a name="inlineasm">Inline Assembler Expressions</a>
2668 <p>LLVM supports inline assembler expressions (as opposed
2669 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2670 a special value. This value represents the inline assembler as a string
2671 (containing the instructions to emit), a list of operand constraints (stored
2672 as a string), a flag that indicates whether or not the inline asm
2673 expression has side effects, and a flag indicating whether the function
2674 containing the asm needs to align its stack conservatively. An example
2675 inline assembler expression is:</p>
2677 <pre class="doc_code">
2678 i32 (i32) asm "bswap $0", "=r,r"
2681 <p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2682 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2685 <pre class="doc_code">
2686 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2689 <p>Inline asms with side effects not visible in the constraint list must be
2690 marked as having side effects. This is done through the use of the
2691 '<tt>sideeffect</tt>' keyword, like so:</p>
2693 <pre class="doc_code">
2694 call void asm sideeffect "eieio", ""()
2697 <p>In some cases inline asms will contain code that will not work unless the
2698 stack is aligned in some way, such as calls or SSE instructions on x86,
2699 yet will not contain code that does that alignment within the asm.
2700 The compiler should make conservative assumptions about what the asm might
2701 contain and should generate its usual stack alignment code in the prologue
2702 if the '<tt>alignstack</tt>' keyword is present:</p>
2704 <pre class="doc_code">
2705 call void asm alignstack "eieio", ""()
2708 <p>If both keywords appear the '<tt>sideeffect</tt>' keyword must come
2711 <p>TODO: The format of the asm and constraints string still need to be
2712 documented here. Constraints on what can be done (e.g. duplication, moving,
2713 etc need to be documented). This is probably best done by reference to
2714 another document that covers inline asm from a holistic perspective.</p>
2717 <a name="inlineasm_md">Inline Asm Metadata</a>
2722 <p>The call instructions that wrap inline asm nodes may have a "!srcloc" MDNode
2723 attached to it that contains a list of constant integers. If present, the
2724 code generator will use the integer as the location cookie value when report
2725 errors through the LLVMContext error reporting mechanisms. This allows a
2726 front-end to correlate backend errors that occur with inline asm back to the
2727 source code that produced it. For example:</p>
2729 <pre class="doc_code">
2730 call void asm sideeffect "something bad", ""()<b>, !srcloc !42</b>
2732 !42 = !{ i32 1234567 }
2735 <p>It is up to the front-end to make sense of the magic numbers it places in the
2736 IR. If the MDNode contains multiple constants, the code generator will use
2737 the one that corresponds to the line of the asm that the error occurs on.</p>
2743 <!-- ======================================================================= -->
2745 <a name="metadata">Metadata Nodes and Metadata Strings</a>
2750 <p>LLVM IR allows metadata to be attached to instructions in the program that
2751 can convey extra information about the code to the optimizers and code
2752 generator. One example application of metadata is source-level debug
2753 information. There are two metadata primitives: strings and nodes. All
2754 metadata has the <tt>metadata</tt> type and is identified in syntax by a
2755 preceding exclamation point ('<tt>!</tt>').</p>
2757 <p>A metadata string is a string surrounded by double quotes. It can contain
2758 any character by escaping non-printable characters with "\xx" where "xx" is
2759 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
2761 <p>Metadata nodes are represented with notation similar to structure constants
2762 (a comma separated list of elements, surrounded by braces and preceded by an
2763 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2764 10}</tt>". Metadata nodes can have any values as their operand.</p>
2766 <p>A <a href="#namedmetadatastructure">named metadata</a> is a collection of
2767 metadata nodes, which can be looked up in the module symbol table. For
2768 example: "<tt>!foo = metadata !{!4, !3}</tt>".
2770 <p>Metadata can be used as function arguments. Here <tt>llvm.dbg.value</tt>
2771 function is using two metadata arguments.</p>
2773 <div class="doc_code">
2775 call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
2779 <p>Metadata can be attached with an instruction. Here metadata <tt>!21</tt> is
2780 attached with <tt>add</tt> instruction using <tt>!dbg</tt> identifier.</p>
2782 <div class="doc_code">
2784 %indvar.next = add i64 %indvar, 1, !dbg !21
2792 <!-- *********************************************************************** -->
2794 <a name="intrinsic_globals">Intrinsic Global Variables</a>
2796 <!-- *********************************************************************** -->
2798 <p>LLVM has a number of "magic" global variables that contain data that affect
2799 code generation or other IR semantics. These are documented here. All globals
2800 of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2801 section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2804 <!-- ======================================================================= -->
2806 <a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
2811 <p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2812 href="#linkage_appending">appending linkage</a>. This array contains a list of
2813 pointers to global variables and functions which may optionally have a pointer
2814 cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2820 @llvm.used = appending global [2 x i8*] [
2822 i8* bitcast (i32* @Y to i8*)
2823 ], section "llvm.metadata"
2826 <p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2827 compiler, assembler, and linker are required to treat the symbol as if there is
2828 a reference to the global that it cannot see. For example, if a variable has
2829 internal linkage and no references other than that from the <tt>@llvm.used</tt>
2830 list, it cannot be deleted. This is commonly used to represent references from
2831 inline asms and other things the compiler cannot "see", and corresponds to
2832 "attribute((used))" in GNU C.</p>
2834 <p>On some targets, the code generator must emit a directive to the assembler or
2835 object file to prevent the assembler and linker from molesting the symbol.</p>
2839 <!-- ======================================================================= -->
2841 <a name="intg_compiler_used">
2842 The '<tt>llvm.compiler.used</tt>' Global Variable
2848 <p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2849 <tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2850 touching the symbol. On targets that support it, this allows an intelligent
2851 linker to optimize references to the symbol without being impeded as it would be
2852 by <tt>@llvm.used</tt>.</p>
2854 <p>This is a rare construct that should only be used in rare circumstances, and
2855 should not be exposed to source languages.</p>
2859 <!-- ======================================================================= -->
2861 <a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
2866 %0 = type { i32, void ()* }
2867 @llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor }]
2869 <p>The <tt>@llvm.global_ctors</tt> array contains a list of constructor functions and associated priorities. The functions referenced by this array will be called in ascending order of priority (i.e. lowest first) when the module is loaded. The order of functions with the same priority is not defined.
2874 <!-- ======================================================================= -->
2876 <a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
2881 %0 = type { i32, void ()* }
2882 @llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor }]
2885 <p>The <tt>@llvm.global_dtors</tt> array contains a list of destructor functions and associated priorities. The functions referenced by this array will be called in descending order of priority (i.e. highest first) when the module is loaded. The order of functions with the same priority is not defined.
2892 <!-- *********************************************************************** -->
2893 <h2><a name="instref">Instruction Reference</a></h2>
2894 <!-- *********************************************************************** -->
2898 <p>The LLVM instruction set consists of several different classifications of
2899 instructions: <a href="#terminators">terminator
2900 instructions</a>, <a href="#binaryops">binary instructions</a>,
2901 <a href="#bitwiseops">bitwise binary instructions</a>,
2902 <a href="#memoryops">memory instructions</a>, and
2903 <a href="#otherops">other instructions</a>.</p>
2905 <!-- ======================================================================= -->
2907 <a name="terminators">Terminator Instructions</a>
2912 <p>As mentioned <a href="#functionstructure">previously</a>, every basic block
2913 in a program ends with a "Terminator" instruction, which indicates which
2914 block should be executed after the current block is finished. These
2915 terminator instructions typically yield a '<tt>void</tt>' value: they produce
2916 control flow, not values (the one exception being the
2917 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2919 <p>There are seven different terminator instructions: the
2920 '<a href="#i_ret"><tt>ret</tt></a>' instruction, the
2921 '<a href="#i_br"><tt>br</tt></a>' instruction, the
2922 '<a href="#i_switch"><tt>switch</tt></a>' instruction, the
2923 '<a href="#i_indirectbr">'<tt>indirectbr</tt></a>' Instruction, the
2924 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the
2925 '<a href="#i_unwind"><tt>unwind</tt></a>' instruction, and the
2926 '<a href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2928 <!-- _______________________________________________________________________ -->
2930 <a name="i_ret">'<tt>ret</tt>' Instruction</a>
2937 ret <type> <value> <i>; Return a value from a non-void function</i>
2938 ret void <i>; Return from void function</i>
2942 <p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
2943 a value) from a function back to the caller.</p>
2945 <p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
2946 value and then causes control flow, and one that just causes control flow to
2950 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
2951 return value. The type of the return value must be a
2952 '<a href="#t_firstclass">first class</a>' type.</p>
2954 <p>A function is not <a href="#wellformed">well formed</a> if it it has a
2955 non-void return type and contains a '<tt>ret</tt>' instruction with no return
2956 value or a return value with a type that does not match its type, or if it
2957 has a void return type and contains a '<tt>ret</tt>' instruction with a
2961 <p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
2962 the calling function's context. If the caller is a
2963 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
2964 instruction after the call. If the caller was an
2965 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
2966 the beginning of the "normal" destination block. If the instruction returns
2967 a value, that value shall set the call or invoke instruction's return
2972 ret i32 5 <i>; Return an integer value of 5</i>
2973 ret void <i>; Return from a void function</i>
2974 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
2978 <!-- _______________________________________________________________________ -->
2980 <a name="i_br">'<tt>br</tt>' Instruction</a>
2987 br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2991 <p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
2992 different basic block in the current function. There are two forms of this
2993 instruction, corresponding to a conditional branch and an unconditional
2997 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
2998 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
2999 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
3003 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
3004 argument is evaluated. If the value is <tt>true</tt>, control flows to the
3005 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
3006 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
3011 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
3012 br i1 %cond, label %IfEqual, label %IfUnequal
3014 <a href="#i_ret">ret</a> i32 1
3016 <a href="#i_ret">ret</a> i32 0
3021 <!-- _______________________________________________________________________ -->
3023 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
3030 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
3034 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
3035 several different places. It is a generalization of the '<tt>br</tt>'
3036 instruction, allowing a branch to occur to one of many possible
3040 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
3041 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
3042 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
3043 The table is not allowed to contain duplicate constant entries.</p>
3046 <p>The <tt>switch</tt> instruction specifies a table of values and
3047 destinations. When the '<tt>switch</tt>' instruction is executed, this table
3048 is searched for the given value. If the value is found, control flow is
3049 transferred to the corresponding destination; otherwise, control flow is
3050 transferred to the default destination.</p>
3052 <h5>Implementation:</h5>
3053 <p>Depending on properties of the target machine and the particular
3054 <tt>switch</tt> instruction, this instruction may be code generated in
3055 different ways. For example, it could be generated as a series of chained
3056 conditional branches or with a lookup table.</p>
3060 <i>; Emulate a conditional br instruction</i>
3061 %Val = <a href="#i_zext">zext</a> i1 %value to i32
3062 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
3064 <i>; Emulate an unconditional br instruction</i>
3065 switch i32 0, label %dest [ ]
3067 <i>; Implement a jump table:</i>
3068 switch i32 %val, label %otherwise [ i32 0, label %onzero
3070 i32 2, label %ontwo ]
3076 <!-- _______________________________________________________________________ -->
3078 <a name="i_indirectbr">'<tt>indirectbr</tt>' Instruction</a>
3085 indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
3090 <p>The '<tt>indirectbr</tt>' instruction implements an indirect branch to a label
3091 within the current function, whose address is specified by
3092 "<tt>address</tt>". Address must be derived from a <a
3093 href="#blockaddress">blockaddress</a> constant.</p>
3097 <p>The '<tt>address</tt>' argument is the address of the label to jump to. The
3098 rest of the arguments indicate the full set of possible destinations that the
3099 address may point to. Blocks are allowed to occur multiple times in the
3100 destination list, though this isn't particularly useful.</p>
3102 <p>This destination list is required so that dataflow analysis has an accurate
3103 understanding of the CFG.</p>
3107 <p>Control transfers to the block specified in the address argument. All
3108 possible destination blocks must be listed in the label list, otherwise this
3109 instruction has undefined behavior. This implies that jumps to labels
3110 defined in other functions have undefined behavior as well.</p>
3112 <h5>Implementation:</h5>
3114 <p>This is typically implemented with a jump through a register.</p>
3118 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
3124 <!-- _______________________________________________________________________ -->
3126 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
3133 <result> = invoke [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>] <ptr to function ty> <function ptr val>(<function args>) [<a href="#fnattrs">fn attrs</a>]
3134 to label <normal label> unwind label <exception label>
3138 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
3139 function, with the possibility of control flow transfer to either the
3140 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
3141 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
3142 control flow will return to the "normal" label. If the callee (or any
3143 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
3144 instruction, control is interrupted and continued at the dynamically nearest
3145 "exception" label.</p>
3148 <p>This instruction requires several arguments:</p>
3151 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
3152 convention</a> the call should use. If none is specified, the call
3153 defaults to using C calling conventions.</li>
3155 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
3156 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
3157 '<tt>inreg</tt>' attributes are valid here.</li>
3159 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
3160 function value being invoked. In most cases, this is a direct function
3161 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
3162 off an arbitrary pointer to function value.</li>
3164 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
3165 function to be invoked. </li>
3167 <li>'<tt>function args</tt>': argument list whose types match the function
3168 signature argument types and parameter attributes. All arguments must be
3169 of <a href="#t_firstclass">first class</a> type. If the function
3170 signature indicates the function accepts a variable number of arguments,
3171 the extra arguments can be specified.</li>
3173 <li>'<tt>normal label</tt>': the label reached when the called function
3174 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
3176 <li>'<tt>exception label</tt>': the label reached when a callee returns with
3177 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
3179 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
3180 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
3181 '<tt>readnone</tt>' attributes are valid here.</li>
3185 <p>This instruction is designed to operate as a standard
3186 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
3187 primary difference is that it establishes an association with a label, which
3188 is used by the runtime library to unwind the stack.</p>
3190 <p>This instruction is used in languages with destructors to ensure that proper
3191 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
3192 exception. Additionally, this is important for implementation of
3193 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
3195 <p>For the purposes of the SSA form, the definition of the value returned by the
3196 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
3197 block to the "normal" label. If the callee unwinds then no return value is
3200 <p>Note that the code generator does not yet completely support unwind, and
3201 that the invoke/unwind semantics are likely to change in future versions.</p>
3205 %retval = invoke i32 @Test(i32 15) to label %Continue
3206 unwind label %TestCleanup <i>; {i32}:retval set</i>
3207 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
3208 unwind label %TestCleanup <i>; {i32}:retval set</i>
3213 <!-- _______________________________________________________________________ -->
3216 <a name="i_unwind">'<tt>unwind</tt>' Instruction</a>
3227 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
3228 at the first callee in the dynamic call stack which used
3229 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
3230 This is primarily used to implement exception handling.</p>
3233 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
3234 immediately halt. The dynamic call stack is then searched for the
3235 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
3236 Once found, execution continues at the "exceptional" destination block
3237 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
3238 instruction in the dynamic call chain, undefined behavior results.</p>
3240 <p>Note that the code generator does not yet completely support unwind, and
3241 that the invoke/unwind semantics are likely to change in future versions.</p>
3245 <!-- _______________________________________________________________________ -->
3248 <a name="i_unreachable">'<tt>unreachable</tt>' Instruction</a>
3259 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
3260 instruction is used to inform the optimizer that a particular portion of the
3261 code is not reachable. This can be used to indicate that the code after a
3262 no-return function cannot be reached, and other facts.</p>
3265 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
3271 <!-- ======================================================================= -->
3273 <a name="binaryops">Binary Operations</a>
3278 <p>Binary operators are used to do most of the computation in a program. They
3279 require two operands of the same type, execute an operation on them, and
3280 produce a single value. The operands might represent multiple data, as is
3281 the case with the <a href="#t_vector">vector</a> data type. The result value
3282 has the same type as its operands.</p>
3284 <p>There are several different binary operators:</p>
3286 <!-- _______________________________________________________________________ -->
3288 <a name="i_add">'<tt>add</tt>' Instruction</a>
3295 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3296 <result> = add nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3297 <result> = add nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3298 <result> = add nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3302 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
3305 <p>The two arguments to the '<tt>add</tt>' instruction must
3306 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3307 integer values. Both arguments must have identical types.</p>
3310 <p>The value produced is the integer sum of the two operands.</p>
3312 <p>If the sum has unsigned overflow, the result returned is the mathematical
3313 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
3315 <p>Because LLVM integers use a two's complement representation, this instruction
3316 is appropriate for both signed and unsigned integers.</p>
3318 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3319 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3320 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
3321 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3322 respectively, occurs.</p>
3326 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
3331 <!-- _______________________________________________________________________ -->
3333 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
3340 <result> = fadd <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3344 <p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
3347 <p>The two arguments to the '<tt>fadd</tt>' instruction must be
3348 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3349 floating point values. Both arguments must have identical types.</p>
3352 <p>The value produced is the floating point sum of the two operands.</p>
3356 <result> = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
3361 <!-- _______________________________________________________________________ -->
3363 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
3370 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3371 <result> = sub nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3372 <result> = sub nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3373 <result> = sub nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3377 <p>The '<tt>sub</tt>' instruction returns the difference of its two
3380 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
3381 '<tt>neg</tt>' instruction present in most other intermediate
3382 representations.</p>
3385 <p>The two arguments to the '<tt>sub</tt>' instruction must
3386 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3387 integer values. Both arguments must have identical types.</p>
3390 <p>The value produced is the integer difference of the two operands.</p>
3392 <p>If the difference has unsigned overflow, the result returned is the
3393 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
3396 <p>Because LLVM integers use a two's complement representation, this instruction
3397 is appropriate for both signed and unsigned integers.</p>
3399 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3400 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3401 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
3402 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3403 respectively, occurs.</p>
3407 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
3408 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
3413 <!-- _______________________________________________________________________ -->
3415 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
3422 <result> = fsub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3426 <p>The '<tt>fsub</tt>' instruction returns the difference of its two
3429 <p>Note that the '<tt>fsub</tt>' instruction is used to represent the
3430 '<tt>fneg</tt>' instruction present in most other intermediate
3431 representations.</p>
3434 <p>The two arguments to the '<tt>fsub</tt>' instruction must be
3435 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3436 floating point values. Both arguments must have identical types.</p>
3439 <p>The value produced is the floating point difference of the two operands.</p>
3443 <result> = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
3444 <result> = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
3449 <!-- _______________________________________________________________________ -->
3451 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
3458 <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3459 <result> = mul nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3460 <result> = mul nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3461 <result> = mul nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3465 <p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
3468 <p>The two arguments to the '<tt>mul</tt>' instruction must
3469 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3470 integer values. Both arguments must have identical types.</p>
3473 <p>The value produced is the integer product of the two operands.</p>
3475 <p>If the result of the multiplication has unsigned overflow, the result
3476 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
3477 width of the result.</p>
3479 <p>Because LLVM integers use a two's complement representation, and the result
3480 is the same width as the operands, this instruction returns the correct
3481 result for both signed and unsigned integers. If a full product
3482 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
3483 be sign-extended or zero-extended as appropriate to the width of the full
3486 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3487 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3488 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
3489 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3490 respectively, occurs.</p>
3494 <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
3499 <!-- _______________________________________________________________________ -->
3501 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
3508 <result> = fmul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3512 <p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
3515 <p>The two arguments to the '<tt>fmul</tt>' instruction must be
3516 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3517 floating point values. Both arguments must have identical types.</p>
3520 <p>The value produced is the floating point product of the two operands.</p>
3524 <result> = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
3529 <!-- _______________________________________________________________________ -->
3531 <a name="i_udiv">'<tt>udiv</tt>' Instruction</a>
3538 <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3539 <result> = udiv exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3543 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
3546 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
3547 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3548 values. Both arguments must have identical types.</p>
3551 <p>The value produced is the unsigned integer quotient of the two operands.</p>
3553 <p>Note that unsigned integer division and signed integer division are distinct
3554 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
3556 <p>Division by zero leads to undefined behavior.</p>
3558 <p>If the <tt>exact</tt> keyword is present, the result value of the
3559 <tt>udiv</tt> is a <a href="#trapvalues">trap value</a> if %op1 is not a
3560 multiple of %op2 (as such, "((a udiv exact b) mul b) == a").</p>
3565 <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3570 <!-- _______________________________________________________________________ -->
3572 <a name="i_sdiv">'<tt>sdiv</tt>' Instruction</a>
3579 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3580 <result> = sdiv exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3584 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
3587 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
3588 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3589 values. Both arguments must have identical types.</p>
3592 <p>The value produced is the signed integer quotient of the two operands rounded
3595 <p>Note that signed integer division and unsigned integer division are distinct
3596 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
3598 <p>Division by zero leads to undefined behavior. Overflow also leads to
3599 undefined behavior; this is a rare case, but can occur, for example, by doing
3600 a 32-bit division of -2147483648 by -1.</p>
3602 <p>If the <tt>exact</tt> keyword is present, the result value of the
3603 <tt>sdiv</tt> is a <a href="#trapvalues">trap value</a> if the result would
3608 <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3613 <!-- _______________________________________________________________________ -->
3615 <a name="i_fdiv">'<tt>fdiv</tt>' Instruction</a>
3622 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3626 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
3629 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
3630 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3631 floating point values. Both arguments must have identical types.</p>
3634 <p>The value produced is the floating point quotient of the two operands.</p>
3638 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
3643 <!-- _______________________________________________________________________ -->
3645 <a name="i_urem">'<tt>urem</tt>' Instruction</a>
3652 <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3656 <p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3657 division of its two arguments.</p>
3660 <p>The two arguments to the '<tt>urem</tt>' instruction must be
3661 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3662 values. Both arguments must have identical types.</p>
3665 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
3666 This instruction always performs an unsigned division to get the
3669 <p>Note that unsigned integer remainder and signed integer remainder are
3670 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3672 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
3676 <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3681 <!-- _______________________________________________________________________ -->
3683 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
3690 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3694 <p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3695 division of its two operands. This instruction can also take
3696 <a href="#t_vector">vector</a> versions of the values in which case the
3697 elements must be integers.</p>
3700 <p>The two arguments to the '<tt>srem</tt>' instruction must be
3701 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3702 values. Both arguments must have identical types.</p>
3705 <p>This instruction returns the <i>remainder</i> of a division (where the result
3706 is either zero or has the same sign as the dividend, <tt>op1</tt>), not the
3707 <i>modulo</i> operator (where the result is either zero or has the same sign
3708 as the divisor, <tt>op2</tt>) of a value.
3709 For more information about the difference,
3710 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3711 Math Forum</a>. For a table of how this is implemented in various languages,
3712 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3713 Wikipedia: modulo operation</a>.</p>
3715 <p>Note that signed integer remainder and unsigned integer remainder are
3716 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3718 <p>Taking the remainder of a division by zero leads to undefined behavior.
3719 Overflow also leads to undefined behavior; this is a rare case, but can
3720 occur, for example, by taking the remainder of a 32-bit division of
3721 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3722 lets srem be implemented using instructions that return both the result of
3723 the division and the remainder.)</p>
3727 <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3732 <!-- _______________________________________________________________________ -->
3734 <a name="i_frem">'<tt>frem</tt>' Instruction</a>
3741 <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3745 <p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3746 its two operands.</p>
3749 <p>The two arguments to the '<tt>frem</tt>' instruction must be
3750 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3751 floating point values. Both arguments must have identical types.</p>
3754 <p>This instruction returns the <i>remainder</i> of a division. The remainder
3755 has the same sign as the dividend.</p>
3759 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
3766 <!-- ======================================================================= -->
3768 <a name="bitwiseops">Bitwise Binary Operations</a>
3773 <p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3774 program. They are generally very efficient instructions and can commonly be
3775 strength reduced from other instructions. They require two operands of the
3776 same type, execute an operation on them, and produce a single value. The
3777 resulting value is the same type as its operands.</p>
3779 <!-- _______________________________________________________________________ -->
3781 <a name="i_shl">'<tt>shl</tt>' Instruction</a>
3788 <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3789 <result> = shl nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3790 <result> = shl nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3791 <result> = shl nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3795 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3796 a specified number of bits.</p>
3799 <p>Both arguments to the '<tt>shl</tt>' instruction must be the
3800 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3801 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3804 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3805 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3806 is (statically or dynamically) negative or equal to or larger than the number
3807 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3808 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3809 shift amount in <tt>op2</tt>.</p>
3811 <p>If the <tt>nuw</tt> keyword is present, then the shift produces a
3812 <a href="#trapvalues">trap value</a> if it shifts out any non-zero bits. If
3813 the <tt>nsw</tt> keyword is present, then the shift produces a
3814 <a href="#trapvalues">trap value</a> if it shifts out any bits that disagree
3815 with the resultant sign bit. As such, NUW/NSW have the same semantics as
3816 they would if the shift were expressed as a mul instruction with the same
3817 nsw/nuw bits in (mul %op1, (shl 1, %op2)).</p>
3821 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
3822 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
3823 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
3824 <result> = shl i32 1, 32 <i>; undefined</i>
3825 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
3830 <!-- _______________________________________________________________________ -->
3832 <a name="i_lshr">'<tt>lshr</tt>' Instruction</a>
3839 <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3840 <result> = lshr exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3844 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
3845 operand shifted to the right a specified number of bits with zero fill.</p>
3848 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
3849 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3850 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3853 <p>This instruction always performs a logical shift right operation. The most
3854 significant bits of the result will be filled with zero bits after the shift.
3855 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
3856 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3857 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3858 shift amount in <tt>op2</tt>.</p>
3860 <p>If the <tt>exact</tt> keyword is present, the result value of the
3861 <tt>lshr</tt> is a <a href="#trapvalues">trap value</a> if any of the bits
3862 shifted out are non-zero.</p>
3867 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
3868 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
3869 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
3870 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
3871 <result> = lshr i32 1, 32 <i>; undefined</i>
3872 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
3877 <!-- _______________________________________________________________________ -->
3879 <a name="i_ashr">'<tt>ashr</tt>' Instruction</a>
3886 <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3887 <result> = ashr exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3891 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3892 operand shifted to the right a specified number of bits with sign
3896 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
3897 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3898 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3901 <p>This instruction always performs an arithmetic shift right operation, The
3902 most significant bits of the result will be filled with the sign bit
3903 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3904 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
3905 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
3906 the corresponding shift amount in <tt>op2</tt>.</p>
3908 <p>If the <tt>exact</tt> keyword is present, the result value of the
3909 <tt>ashr</tt> is a <a href="#trapvalues">trap value</a> if any of the bits
3910 shifted out are non-zero.</p>
3914 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3915 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3916 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3917 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
3918 <result> = ashr i32 1, 32 <i>; undefined</i>
3919 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
3924 <!-- _______________________________________________________________________ -->
3926 <a name="i_and">'<tt>and</tt>' Instruction</a>
3933 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3937 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
3941 <p>The two arguments to the '<tt>and</tt>' instruction must be
3942 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3943 values. Both arguments must have identical types.</p>
3946 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
3948 <table border="1" cellspacing="0" cellpadding="4">
3980 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
3981 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
3982 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
3985 <!-- _______________________________________________________________________ -->
3987 <a name="i_or">'<tt>or</tt>' Instruction</a>
3994 <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3998 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
4002 <p>The two arguments to the '<tt>or</tt>' instruction must be
4003 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4004 values. Both arguments must have identical types.</p>
4007 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
4009 <table border="1" cellspacing="0" cellpadding="4">
4041 <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
4042 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
4043 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
4048 <!-- _______________________________________________________________________ -->
4050 <a name="i_xor">'<tt>xor</tt>' Instruction</a>
4057 <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4061 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
4062 its two operands. The <tt>xor</tt> is used to implement the "one's
4063 complement" operation, which is the "~" operator in C.</p>
4066 <p>The two arguments to the '<tt>xor</tt>' instruction must be
4067 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
4068 values. Both arguments must have identical types.</p>
4071 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
4073 <table border="1" cellspacing="0" cellpadding="4">
4105 <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
4106 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
4107 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
4108 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
4115 <!-- ======================================================================= -->
4117 <a name="vectorops">Vector Operations</a>
4122 <p>LLVM supports several instructions to represent vector operations in a
4123 target-independent manner. These instructions cover the element-access and
4124 vector-specific operations needed to process vectors effectively. While LLVM
4125 does directly support these vector operations, many sophisticated algorithms
4126 will want to use target-specific intrinsics to take full advantage of a
4127 specific target.</p>
4129 <!-- _______________________________________________________________________ -->
4131 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
4138 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
4142 <p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
4143 from a vector at a specified index.</p>
4147 <p>The first operand of an '<tt>extractelement</tt>' instruction is a value
4148 of <a href="#t_vector">vector</a> type. The second operand is an index
4149 indicating the position from which to extract the element. The index may be
4153 <p>The result is a scalar of the same type as the element type of
4154 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
4155 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4156 results are undefined.</p>
4160 <result> = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
4165 <!-- _______________________________________________________________________ -->
4167 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
4174 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
4178 <p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
4179 vector at a specified index.</p>
4182 <p>The first operand of an '<tt>insertelement</tt>' instruction is a value
4183 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
4184 whose type must equal the element type of the first operand. The third
4185 operand is an index indicating the position at which to insert the value.
4186 The index may be a variable.</p>
4189 <p>The result is a vector of the same type as <tt>val</tt>. Its element values
4190 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
4191 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4192 results are undefined.</p>
4196 <result> = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
4201 <!-- _______________________________________________________________________ -->
4203 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
4210 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
4214 <p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
4215 from two input vectors, returning a vector with the same element type as the
4216 input and length that is the same as the shuffle mask.</p>
4219 <p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
4220 with types that match each other. The third argument is a shuffle mask whose
4221 element type is always 'i32'. The result of the instruction is a vector
4222 whose length is the same as the shuffle mask and whose element type is the
4223 same as the element type of the first two operands.</p>
4225 <p>The shuffle mask operand is required to be a constant vector with either
4226 constant integer or undef values.</p>
4229 <p>The elements of the two input vectors are numbered from left to right across
4230 both of the vectors. The shuffle mask operand specifies, for each element of
4231 the result vector, which element of the two input vectors the result element
4232 gets. The element selector may be undef (meaning "don't care") and the
4233 second operand may be undef if performing a shuffle from only one vector.</p>
4237 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4238 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
4239 <result> = shufflevector <4 x i32> %v1, <4 x i32> undef,
4240 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
4241 <result> = shufflevector <8 x i32> %v1, <8 x i32> undef,
4242 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
4243 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4244 <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7 > <i>; yields <8 x i32></i>
4251 <!-- ======================================================================= -->
4253 <a name="aggregateops">Aggregate Operations</a>
4258 <p>LLVM supports several instructions for working with
4259 <a href="#t_aggregate">aggregate</a> values.</p>
4261 <!-- _______________________________________________________________________ -->
4263 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
4270 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
4274 <p>The '<tt>extractvalue</tt>' instruction extracts the value of a member field
4275 from an <a href="#t_aggregate">aggregate</a> value.</p>
4278 <p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
4279 of <a href="#t_struct">struct</a> or
4280 <a href="#t_array">array</a> type. The operands are constant indices to
4281 specify which value to extract in a similar manner as indices in a
4282 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
4283 <p>The major differences to <tt>getelementptr</tt> indexing are:</p>
4285 <li>Since the value being indexed is not a pointer, the first index is
4286 omitted and assumed to be zero.</li>
4287 <li>At least one index must be specified.</li>
4288 <li>Not only struct indices but also array indices must be in
4293 <p>The result is the value at the position in the aggregate specified by the
4298 <result> = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
4303 <!-- _______________________________________________________________________ -->
4305 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
4312 <result> = insertvalue <aggregate type> <val>, <ty> <elt>, <idx>{, <idx>}* <i>; yields <aggregate type></i>
4316 <p>The '<tt>insertvalue</tt>' instruction inserts a value into a member field
4317 in an <a href="#t_aggregate">aggregate</a> value.</p>
4320 <p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
4321 of <a href="#t_struct">struct</a> or
4322 <a href="#t_array">array</a> type. The second operand is a first-class
4323 value to insert. The following operands are constant indices indicating
4324 the position at which to insert the value in a similar manner as indices in a
4325 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' instruction. The
4326 value to insert must have the same type as the value identified by the
4330 <p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
4331 that of <tt>val</tt> except that the value at the position specified by the
4332 indices is that of <tt>elt</tt>.</p>
4336 %agg1 = insertvalue {i32, float} undef, i32 1, 0 <i>; yields {i32 1, float undef}</i>
4337 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 <i>; yields {i32 1, float %val}</i>
4338 %agg3 = insertvalue {i32, {float}} %agg1, float %val, 1, 0 <i>; yields {i32 1, float %val}</i>
4345 <!-- ======================================================================= -->
4347 <a name="memoryops">Memory Access and Addressing Operations</a>
4352 <p>A key design point of an SSA-based representation is how it represents
4353 memory. In LLVM, no memory locations are in SSA form, which makes things
4354 very simple. This section describes how to read, write, and allocate
4357 <!-- _______________________________________________________________________ -->
4359 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
4366 <result> = alloca <type>[, <ty> <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
4370 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
4371 currently executing function, to be automatically released when this function
4372 returns to its caller. The object is always allocated in the generic address
4373 space (address space zero).</p>
4376 <p>The '<tt>alloca</tt>' instruction
4377 allocates <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the
4378 runtime stack, returning a pointer of the appropriate type to the program.
4379 If "NumElements" is specified, it is the number of elements allocated,
4380 otherwise "NumElements" is defaulted to be one. If a constant alignment is
4381 specified, the value result of the allocation is guaranteed to be aligned to
4382 at least that boundary. If not specified, or if zero, the target can choose
4383 to align the allocation on any convenient boundary compatible with the
4386 <p>'<tt>type</tt>' may be any sized type.</p>
4389 <p>Memory is allocated; a pointer is returned. The operation is undefined if
4390 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
4391 memory is automatically released when the function returns. The
4392 '<tt>alloca</tt>' instruction is commonly used to represent automatic
4393 variables that must have an address available. When the function returns
4394 (either with the <tt><a href="#i_ret">ret</a></tt>
4395 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
4396 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
4400 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
4401 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
4402 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
4403 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
4408 <!-- _______________________________________________________________________ -->
4410 <a name="i_load">'<tt>load</tt>' Instruction</a>
4417 <result> = load <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>]
4418 <result> = volatile load <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>]
4419 !<index> = !{ i32 1 }
4423 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
4426 <p>The argument to the '<tt>load</tt>' instruction specifies the memory address
4427 from which to load. The pointer must point to
4428 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
4429 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
4430 number or order of execution of this <tt>load</tt> with other <a
4431 href="#volatile">volatile operations</a>.</p>
4433 <p>The optional constant <tt>align</tt> argument specifies the alignment of the
4434 operation (that is, the alignment of the memory address). A value of 0 or an
4435 omitted <tt>align</tt> argument means that the operation has the preferential
4436 alignment for the target. It is the responsibility of the code emitter to
4437 ensure that the alignment information is correct. Overestimating the
4438 alignment results in undefined behavior. Underestimating the alignment may
4439 produce less efficient code. An alignment of 1 is always safe.</p>
4441 <p>The optional <tt>!nontemporal</tt> metadata must reference a single
4442 metatadata name <index> corresponding to a metadata node with
4443 one <tt>i32</tt> entry of value 1. The existence of
4444 the <tt>!nontemporal</tt> metatadata on the instruction tells the optimizer
4445 and code generator that this load is not expected to be reused in the cache.
4446 The code generator may select special instructions to save cache bandwidth,
4447 such as the <tt>MOVNT</tt> instruction on x86.</p>
4450 <p>The location of memory pointed to is loaded. If the value being loaded is of
4451 scalar type then the number of bytes read does not exceed the minimum number
4452 of bytes needed to hold all bits of the type. For example, loading an
4453 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
4454 <tt>i20</tt> with a size that is not an integral number of bytes, the result
4455 is undefined if the value was not originally written using a store of the
4460 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4461 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
4462 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
4467 <!-- _______________________________________________________________________ -->
4469 <a name="i_store">'<tt>store</tt>' Instruction</a>
4476 store <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] <i>; yields {void}</i>
4477 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] <i>; yields {void}</i>
4481 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
4484 <p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
4485 and an address at which to store it. The type of the
4486 '<tt><pointer></tt>' operand must be a pointer to
4487 the <a href="#t_firstclass">first class</a> type of the
4488 '<tt><value></tt>' operand. If the <tt>store</tt> is marked as
4489 <tt>volatile</tt>, then the optimizer is not allowed to modify the number or
4490 order of execution of this <tt>store</tt> with other <a
4491 href="#volatile">volatile operations</a>.</p>
4493 <p>The optional constant "align" argument specifies the alignment of the
4494 operation (that is, the alignment of the memory address). A value of 0 or an
4495 omitted "align" argument means that the operation has the preferential
4496 alignment for the target. It is the responsibility of the code emitter to
4497 ensure that the alignment information is correct. Overestimating the
4498 alignment results in an undefined behavior. Underestimating the alignment may
4499 produce less efficient code. An alignment of 1 is always safe.</p>
4501 <p>The optional !nontemporal metadata must reference a single metatadata
4502 name <index> corresponding to a metadata node with one i32 entry of
4503 value 1. The existence of the !nontemporal metatadata on the
4504 instruction tells the optimizer and code generator that this load is
4505 not expected to be reused in the cache. The code generator may
4506 select special instructions to save cache bandwidth, such as the
4507 MOVNT instruction on x86.</p>
4511 <p>The contents of memory are updated to contain '<tt><value></tt>' at the
4512 location specified by the '<tt><pointer></tt>' operand. If
4513 '<tt><value></tt>' is of scalar type then the number of bytes written
4514 does not exceed the minimum number of bytes needed to hold all bits of the
4515 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
4516 writing a value of a type like <tt>i20</tt> with a size that is not an
4517 integral number of bytes, it is unspecified what happens to the extra bits
4518 that do not belong to the type, but they will typically be overwritten.</p>
4522 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4523 store i32 3, i32* %ptr <i>; yields {void}</i>
4524 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
4529 <!-- _______________________________________________________________________ -->
4531 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
4538 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
4539 <result> = getelementptr inbounds <pty>* <ptrval>{, <ty> <idx>}*
4543 <p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
4544 subelement of an <a href="#t_aggregate">aggregate</a> data structure.
4545 It performs address calculation only and does not access memory.</p>
4548 <p>The first argument is always a pointer, and forms the basis of the
4549 calculation. The remaining arguments are indices that indicate which of the
4550 elements of the aggregate object are indexed. The interpretation of each
4551 index is dependent on the type being indexed into. The first index always
4552 indexes the pointer value given as the first argument, the second index
4553 indexes a value of the type pointed to (not necessarily the value directly
4554 pointed to, since the first index can be non-zero), etc. The first type
4555 indexed into must be a pointer value, subsequent types can be arrays,
4556 vectors, and structs. Note that subsequent types being indexed into
4557 can never be pointers, since that would require loading the pointer before
4558 continuing calculation.</p>
4560 <p>The type of each index argument depends on the type it is indexing into.
4561 When indexing into a (optionally packed) structure, only <tt>i32</tt>
4562 integer <b>constants</b> are allowed. When indexing into an array, pointer
4563 or vector, integers of any width are allowed, and they are not required to be
4566 <p>For example, let's consider a C code fragment and how it gets compiled to
4569 <pre class="doc_code">
4581 int *foo(struct ST *s) {
4582 return &s[1].Z.B[5][13];
4586 <p>The LLVM code generated by the GCC frontend is:</p>
4588 <pre class="doc_code">
4589 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
4590 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
4592 define i32* @foo(%ST* %s) {
4594 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
4600 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
4601 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
4602 }</tt>' type, a structure. The second index indexes into the third element
4603 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
4604 i8 }</tt>' type, another structure. The third index indexes into the second
4605 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
4606 array. The two dimensions of the array are subscripted into, yielding an
4607 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
4608 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
4610 <p>Note that it is perfectly legal to index partially through a structure,
4611 returning a pointer to an inner element. Because of this, the LLVM code for
4612 the given testcase is equivalent to:</p>
4615 define i32* @foo(%ST* %s) {
4616 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
4617 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
4618 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
4619 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
4620 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
4625 <p>If the <tt>inbounds</tt> keyword is present, the result value of the
4626 <tt>getelementptr</tt> is a <a href="#trapvalues">trap value</a> if the
4627 base pointer is not an <i>in bounds</i> address of an allocated object,
4628 or if any of the addresses that would be formed by successive addition of
4629 the offsets implied by the indices to the base address with infinitely
4630 precise arithmetic are not an <i>in bounds</i> address of that allocated
4631 object. The <i>in bounds</i> addresses for an allocated object are all
4632 the addresses that point into the object, plus the address one byte past
4635 <p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
4636 the base address with silently-wrapping two's complement arithmetic, and
4637 the result value of the <tt>getelementptr</tt> may be outside the object
4638 pointed to by the base pointer. The result value may not necessarily be
4639 used to access memory though, even if it happens to point into allocated
4640 storage. See the <a href="#pointeraliasing">Pointer Aliasing Rules</a>
4641 section for more information.</p>
4643 <p>The getelementptr instruction is often confusing. For some more insight into
4644 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
4648 <i>; yields [12 x i8]*:aptr</i>
4649 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
4650 <i>; yields i8*:vptr</i>
4651 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
4652 <i>; yields i8*:eptr</i>
4653 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
4654 <i>; yields i32*:iptr</i>
4655 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
4662 <!-- ======================================================================= -->
4664 <a name="convertops">Conversion Operations</a>
4669 <p>The instructions in this category are the conversion instructions (casting)
4670 which all take a single operand and a type. They perform various bit
4671 conversions on the operand.</p>
4673 <!-- _______________________________________________________________________ -->
4675 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
4682 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
4686 <p>The '<tt>trunc</tt>' instruction truncates its operand to the
4687 type <tt>ty2</tt>.</p>
4690 <p>The '<tt>trunc</tt>' instruction takes a value to trunc, and a type to trunc it to.
4691 Both types must be of <a href="#t_integer">integer</a> types, or vectors
4692 of the same number of integers.
4693 The bit size of the <tt>value</tt> must be larger than
4694 the bit size of the destination type, <tt>ty2</tt>.
4695 Equal sized types are not allowed.</p>
4698 <p>The '<tt>trunc</tt>' instruction truncates the high order bits
4699 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
4700 source size must be larger than the destination size, <tt>trunc</tt> cannot
4701 be a <i>no-op cast</i>. It will always truncate bits.</p>
4705 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
4706 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
4707 %Z = trunc i32 122 to i1 <i>; yields i1:false</i>
4708 %W = trunc <2 x i16> <i16 8, i16 7> to <2 x i8> <i>; yields <i8 8, i8 7></i>
4713 <!-- _______________________________________________________________________ -->
4715 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
4722 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
4726 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
4731 <p>The '<tt>zext</tt>' instruction takes a value to cast, and a type to cast it to.
4732 Both types must be of <a href="#t_integer">integer</a> types, or vectors
4733 of the same number of integers.
4734 The bit size of the <tt>value</tt> must be smaller than
4735 the bit size of the destination type,
4739 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
4740 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
4742 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
4746 %X = zext i32 257 to i64 <i>; yields i64:257</i>
4747 %Y = zext i1 true to i32 <i>; yields i32:1</i>
4748 %Z = zext <2 x i16> <i16 8, i16 7> to <2 x i32> <i>; yields <i32 8, i32 7></i>
4753 <!-- _______________________________________________________________________ -->
4755 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
4762 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
4766 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
4769 <p>The '<tt>sext</tt>' instruction takes a value to cast, and a type to cast it to.
4770 Both types must be of <a href="#t_integer">integer</a> types, or vectors
4771 of the same number of integers.
4772 The bit size of the <tt>value</tt> must be smaller than
4773 the bit size of the destination type,
4777 <p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
4778 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
4779 of the type <tt>ty2</tt>.</p>
4781 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
4785 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
4786 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
4787 %Z = sext <2 x i16> <i16 8, i16 7> to <2 x i32> <i>; yields <i32 8, i32 7></i>
4792 <!-- _______________________________________________________________________ -->
4794 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
4801 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
4805 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
4809 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
4810 point</a> value to cast and a <a href="#t_floating">floating point</a> type
4811 to cast it to. The size of <tt>value</tt> must be larger than the size of
4812 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
4813 <i>no-op cast</i>.</p>
4816 <p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
4817 <a href="#t_floating">floating point</a> type to a smaller
4818 <a href="#t_floating">floating point</a> type. If the value cannot fit
4819 within the destination type, <tt>ty2</tt>, then the results are
4824 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4825 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4830 <!-- _______________________________________________________________________ -->
4832 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
4839 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
4843 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
4844 floating point value.</p>
4847 <p>The '<tt>fpext</tt>' instruction takes a
4848 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
4849 a <a href="#t_floating">floating point</a> type to cast it to. The source
4850 type must be smaller than the destination type.</p>
4853 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
4854 <a href="#t_floating">floating point</a> type to a larger
4855 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4856 used to make a <i>no-op cast</i> because it always changes bits. Use
4857 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
4861 %X = fpext float 3.125 to double <i>; yields double:3.125000e+00</i>
4862 %Y = fpext double %X to fp128 <i>; yields fp128:0xL00000000000000004000900000000000</i>
4867 <!-- _______________________________________________________________________ -->
4869 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
4876 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
4880 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
4881 unsigned integer equivalent of type <tt>ty2</tt>.</p>
4884 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
4885 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4886 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4887 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4888 vector integer type with the same number of elements as <tt>ty</tt></p>
4891 <p>The '<tt>fptoui</tt>' instruction converts its
4892 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4893 towards zero) unsigned integer value. If the value cannot fit
4894 in <tt>ty2</tt>, the results are undefined.</p>
4898 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
4899 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
4900 %Z = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
4905 <!-- _______________________________________________________________________ -->
4907 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
4914 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
4918 <p>The '<tt>fptosi</tt>' instruction converts
4919 <a href="#t_floating">floating point</a> <tt>value</tt> to
4920 type <tt>ty2</tt>.</p>
4923 <p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
4924 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4925 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4926 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4927 vector integer type with the same number of elements as <tt>ty</tt></p>
4930 <p>The '<tt>fptosi</tt>' instruction converts its
4931 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4932 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4933 the results are undefined.</p>
4937 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
4938 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
4939 %Z = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
4944 <!-- _______________________________________________________________________ -->
4946 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
4953 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4957 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
4958 integer and converts that value to the <tt>ty2</tt> type.</p>
4961 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
4962 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4963 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4964 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4965 floating point type with the same number of elements as <tt>ty</tt></p>
4968 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
4969 integer quantity and converts it to the corresponding floating point
4970 value. If the value cannot fit in the floating point value, the results are
4975 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
4976 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
4981 <!-- _______________________________________________________________________ -->
4983 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
4990 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4994 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
4995 and converts that value to the <tt>ty2</tt> type.</p>
4998 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
4999 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
5000 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
5001 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
5002 floating point type with the same number of elements as <tt>ty</tt></p>
5005 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
5006 quantity and converts it to the corresponding floating point value. If the
5007 value cannot fit in the floating point value, the results are undefined.</p>
5011 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
5012 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
5017 <!-- _______________________________________________________________________ -->
5019 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
5026 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
5030 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
5031 the integer type <tt>ty2</tt>.</p>
5034 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
5035 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
5036 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
5039 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
5040 <tt>ty2</tt> by interpreting the pointer value as an integer and either
5041 truncating or zero extending that value to the size of the integer type. If
5042 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
5043 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
5044 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
5049 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
5050 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
5055 <!-- _______________________________________________________________________ -->
5057 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
5064 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
5068 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
5069 pointer type, <tt>ty2</tt>.</p>
5072 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
5073 value to cast, and a type to cast it to, which must be a
5074 <a href="#t_pointer">pointer</a> type.</p>
5077 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
5078 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
5079 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
5080 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
5081 than the size of a pointer then a zero extension is done. If they are the
5082 same size, nothing is done (<i>no-op cast</i>).</p>
5086 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
5087 %Y = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
5088 %Z = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
5093 <!-- _______________________________________________________________________ -->
5095 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
5102 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
5106 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
5107 <tt>ty2</tt> without changing any bits.</p>
5110 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
5111 non-aggregate first class value, and a type to cast it to, which must also be
5112 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
5113 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
5114 identical. If the source type is a pointer, the destination type must also be
5115 a pointer. This instruction supports bitwise conversion of vectors to
5116 integers and to vectors of other types (as long as they have the same
5120 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
5121 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
5122 this conversion. The conversion is done as if the <tt>value</tt> had been
5123 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
5124 be converted to other pointer types with this instruction. To convert
5125 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
5126 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
5130 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
5131 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
5132 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
5139 <!-- ======================================================================= -->
5141 <a name="otherops">Other Operations</a>
5146 <p>The instructions in this category are the "miscellaneous" instructions, which
5147 defy better classification.</p>
5149 <!-- _______________________________________________________________________ -->
5151 <a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
5158 <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
5162 <p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
5163 boolean values based on comparison of its two integer, integer vector, or
5164 pointer operands.</p>
5167 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
5168 the condition code indicating the kind of comparison to perform. It is not a
5169 value, just a keyword. The possible condition code are:</p>
5172 <li><tt>eq</tt>: equal</li>
5173 <li><tt>ne</tt>: not equal </li>
5174 <li><tt>ugt</tt>: unsigned greater than</li>
5175 <li><tt>uge</tt>: unsigned greater or equal</li>
5176 <li><tt>ult</tt>: unsigned less than</li>
5177 <li><tt>ule</tt>: unsigned less or equal</li>
5178 <li><tt>sgt</tt>: signed greater than</li>
5179 <li><tt>sge</tt>: signed greater or equal</li>
5180 <li><tt>slt</tt>: signed less than</li>
5181 <li><tt>sle</tt>: signed less or equal</li>
5184 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
5185 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
5186 typed. They must also be identical types.</p>
5189 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
5190 condition code given as <tt>cond</tt>. The comparison performed always yields
5191 either an <a href="#t_integer"><tt>i1</tt></a> or vector of <tt>i1</tt>
5192 result, as follows:</p>
5195 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
5196 <tt>false</tt> otherwise. No sign interpretation is necessary or
5199 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
5200 <tt>false</tt> otherwise. No sign interpretation is necessary or
5203 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
5204 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5206 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
5207 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5208 to <tt>op2</tt>.</li>
5210 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
5211 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5213 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
5214 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5216 <li><tt>sgt</tt>: interprets the operands as signed values and yields
5217 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5219 <li><tt>sge</tt>: interprets the operands as signed values and yields
5220 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5221 to <tt>op2</tt>.</li>
5223 <li><tt>slt</tt>: interprets the operands as signed values and yields
5224 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5226 <li><tt>sle</tt>: interprets the operands as signed values and yields
5227 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5230 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
5231 values are compared as if they were integers.</p>
5233 <p>If the operands are integer vectors, then they are compared element by
5234 element. The result is an <tt>i1</tt> vector with the same number of elements
5235 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
5239 <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
5240 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
5241 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
5242 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
5243 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
5244 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
5247 <p>Note that the code generator does not yet support vector types with
5248 the <tt>icmp</tt> instruction.</p>
5252 <!-- _______________________________________________________________________ -->
5254 <a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
5261 <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
5265 <p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
5266 values based on comparison of its operands.</p>
5268 <p>If the operands are floating point scalars, then the result type is a boolean
5269 (<a href="#t_integer"><tt>i1</tt></a>).</p>
5271 <p>If the operands are floating point vectors, then the result type is a vector
5272 of boolean with the same number of elements as the operands being
5276 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
5277 the condition code indicating the kind of comparison to perform. It is not a
5278 value, just a keyword. The possible condition code are:</p>
5281 <li><tt>false</tt>: no comparison, always returns false</li>
5282 <li><tt>oeq</tt>: ordered and equal</li>
5283 <li><tt>ogt</tt>: ordered and greater than </li>
5284 <li><tt>oge</tt>: ordered and greater than or equal</li>
5285 <li><tt>olt</tt>: ordered and less than </li>
5286 <li><tt>ole</tt>: ordered and less than or equal</li>
5287 <li><tt>one</tt>: ordered and not equal</li>
5288 <li><tt>ord</tt>: ordered (no nans)</li>
5289 <li><tt>ueq</tt>: unordered or equal</li>
5290 <li><tt>ugt</tt>: unordered or greater than </li>
5291 <li><tt>uge</tt>: unordered or greater than or equal</li>
5292 <li><tt>ult</tt>: unordered or less than </li>
5293 <li><tt>ule</tt>: unordered or less than or equal</li>
5294 <li><tt>une</tt>: unordered or not equal</li>
5295 <li><tt>uno</tt>: unordered (either nans)</li>
5296 <li><tt>true</tt>: no comparison, always returns true</li>
5299 <p><i>Ordered</i> means that neither operand is a QNAN while
5300 <i>unordered</i> means that either operand may be a QNAN.</p>
5302 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
5303 a <a href="#t_floating">floating point</a> type or
5304 a <a href="#t_vector">vector</a> of floating point type. They must have
5305 identical types.</p>
5308 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
5309 according to the condition code given as <tt>cond</tt>. If the operands are
5310 vectors, then the vectors are compared element by element. Each comparison
5311 performed always yields an <a href="#t_integer">i1</a> result, as
5315 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
5317 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5318 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5320 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5321 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5323 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5324 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5326 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5327 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5329 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5330 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5332 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5333 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5335 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
5337 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
5338 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5340 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
5341 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5343 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
5344 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5346 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
5347 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5349 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
5350 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5352 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
5353 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5355 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
5357 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
5362 <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
5363 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
5364 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
5365 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
5368 <p>Note that the code generator does not yet support vector types with
5369 the <tt>fcmp</tt> instruction.</p>
5373 <!-- _______________________________________________________________________ -->
5375 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
5382 <result> = phi <ty> [ <val0>, <label0>], ...
5386 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in the
5387 SSA graph representing the function.</p>
5390 <p>The type of the incoming values is specified with the first type field. After
5391 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
5392 one pair for each predecessor basic block of the current block. Only values
5393 of <a href="#t_firstclass">first class</a> type may be used as the value
5394 arguments to the PHI node. Only labels may be used as the label
5397 <p>There must be no non-phi instructions between the start of a basic block and
5398 the PHI instructions: i.e. PHI instructions must be first in a basic
5401 <p>For the purposes of the SSA form, the use of each incoming value is deemed to
5402 occur on the edge from the corresponding predecessor block to the current
5403 block (but after any definition of an '<tt>invoke</tt>' instruction's return
5404 value on the same edge).</p>
5407 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
5408 specified by the pair corresponding to the predecessor basic block that
5409 executed just prior to the current block.</p>
5413 Loop: ; Infinite loop that counts from 0 on up...
5414 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
5415 %nextindvar = add i32 %indvar, 1
5421 <!-- _______________________________________________________________________ -->
5423 <a name="i_select">'<tt>select</tt>' Instruction</a>
5430 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
5432 <i>selty</i> is either i1 or {<N x i1>}
5436 <p>The '<tt>select</tt>' instruction is used to choose one value based on a
5437 condition, without branching.</p>
5441 <p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
5442 values indicating the condition, and two values of the
5443 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
5444 vectors and the condition is a scalar, then entire vectors are selected, not
5445 individual elements.</p>
5448 <p>If the condition is an i1 and it evaluates to 1, the instruction returns the
5449 first value argument; otherwise, it returns the second value argument.</p>
5451 <p>If the condition is a vector of i1, then the value arguments must be vectors
5452 of the same size, and the selection is done element by element.</p>
5456 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
5459 <p>Note that the code generator does not yet support conditions
5460 with vector type.</p>
5464 <!-- _______________________________________________________________________ -->
5466 <a name="i_call">'<tt>call</tt>' Instruction</a>
5473 <result> = [tail] call [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>] <ty> [<fnty>*] <fnptrval>(<function args>) [<a href="#fnattrs">fn attrs</a>]
5477 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
5480 <p>This instruction requires several arguments:</p>
5483 <li>The optional "tail" marker indicates that the callee function does not
5484 access any allocas or varargs in the caller. Note that calls may be
5485 marked "tail" even if they do not occur before
5486 a <a href="#i_ret"><tt>ret</tt></a> instruction. If the "tail" marker is
5487 present, the function call is eligible for tail call optimization,
5488 but <a href="CodeGenerator.html#tailcallopt">might not in fact be
5489 optimized into a jump</a>. The code generator may optimize calls marked
5490 "tail" with either 1) automatic <a href="CodeGenerator.html#sibcallopt">
5491 sibling call optimization</a> when the caller and callee have
5492 matching signatures, or 2) forced tail call optimization when the
5493 following extra requirements are met:
5495 <li>Caller and callee both have the calling
5496 convention <tt>fastcc</tt>.</li>
5497 <li>The call is in tail position (ret immediately follows call and ret
5498 uses value of call or is void).</li>
5499 <li>Option <tt>-tailcallopt</tt> is enabled,
5500 or <code>llvm::GuaranteedTailCallOpt</code> is <code>true</code>.</li>
5501 <li><a href="CodeGenerator.html#tailcallopt">Platform specific
5502 constraints are met.</a></li>
5506 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
5507 convention</a> the call should use. If none is specified, the call
5508 defaults to using C calling conventions. The calling convention of the
5509 call must match the calling convention of the target function, or else the
5510 behavior is undefined.</li>
5512 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
5513 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
5514 '<tt>inreg</tt>' attributes are valid here.</li>
5516 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
5517 type of the return value. Functions that return no value are marked
5518 <tt><a href="#t_void">void</a></tt>.</li>
5520 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
5521 being invoked. The argument types must match the types implied by this
5522 signature. This type can be omitted if the function is not varargs and if
5523 the function type does not return a pointer to a function.</li>
5525 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
5526 be invoked. In most cases, this is a direct function invocation, but
5527 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
5528 to function value.</li>
5530 <li>'<tt>function args</tt>': argument list whose types match the function
5531 signature argument types and parameter attributes. All arguments must be
5532 of <a href="#t_firstclass">first class</a> type. If the function
5533 signature indicates the function accepts a variable number of arguments,
5534 the extra arguments can be specified.</li>
5536 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
5537 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
5538 '<tt>readnone</tt>' attributes are valid here.</li>
5542 <p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
5543 a specified function, with its incoming arguments bound to the specified
5544 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
5545 function, control flow continues with the instruction after the function
5546 call, and the return value of the function is bound to the result
5551 %retval = call i32 @test(i32 %argc)
5552 call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42) <i>; yields i32</i>
5553 %X = tail call i32 @foo() <i>; yields i32</i>
5554 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
5555 call void %foo(i8 97 signext)
5557 %struct.A = type { i32, i8 }
5558 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
5559 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
5560 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
5561 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
5562 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
5565 <p>llvm treats calls to some functions with names and arguments that match the
5566 standard C99 library as being the C99 library functions, and may perform
5567 optimizations or generate code for them under that assumption. This is
5568 something we'd like to change in the future to provide better support for
5569 freestanding environments and non-C-based languages.</p>
5573 <!-- _______________________________________________________________________ -->
5575 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
5582 <resultval> = va_arg <va_list*> <arglist>, <argty>
5586 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
5587 the "variable argument" area of a function call. It is used to implement the
5588 <tt>va_arg</tt> macro in C.</p>
5591 <p>This instruction takes a <tt>va_list*</tt> value and the type of the
5592 argument. It returns a value of the specified argument type and increments
5593 the <tt>va_list</tt> to point to the next argument. The actual type
5594 of <tt>va_list</tt> is target specific.</p>
5597 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
5598 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
5599 to the next argument. For more information, see the variable argument
5600 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
5602 <p>It is legal for this instruction to be called in a function which does not
5603 take a variable number of arguments, for example, the <tt>vfprintf</tt>
5606 <p><tt>va_arg</tt> is an LLVM instruction instead of
5607 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
5611 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
5613 <p>Note that the code generator does not yet fully support va_arg on many
5614 targets. Also, it does not currently support va_arg with aggregate types on
5623 <!-- *********************************************************************** -->
5624 <h2><a name="intrinsics">Intrinsic Functions</a></h2>
5625 <!-- *********************************************************************** -->
5629 <p>LLVM supports the notion of an "intrinsic function". These functions have
5630 well known names and semantics and are required to follow certain
5631 restrictions. Overall, these intrinsics represent an extension mechanism for
5632 the LLVM language that does not require changing all of the transformations
5633 in LLVM when adding to the language (or the bitcode reader/writer, the
5634 parser, etc...).</p>
5636 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
5637 prefix is reserved in LLVM for intrinsic names; thus, function names may not
5638 begin with this prefix. Intrinsic functions must always be external
5639 functions: you cannot define the body of intrinsic functions. Intrinsic
5640 functions may only be used in call or invoke instructions: it is illegal to
5641 take the address of an intrinsic function. Additionally, because intrinsic
5642 functions are part of the LLVM language, it is required if any are added that
5643 they be documented here.</p>
5645 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
5646 family of functions that perform the same operation but on different data
5647 types. Because LLVM can represent over 8 million different integer types,
5648 overloading is used commonly to allow an intrinsic function to operate on any
5649 integer type. One or more of the argument types or the result type can be
5650 overloaded to accept any integer type. Argument types may also be defined as
5651 exactly matching a previous argument's type or the result type. This allows
5652 an intrinsic function which accepts multiple arguments, but needs all of them
5653 to be of the same type, to only be overloaded with respect to a single
5654 argument or the result.</p>
5656 <p>Overloaded intrinsics will have the names of its overloaded argument types
5657 encoded into its function name, each preceded by a period. Only those types
5658 which are overloaded result in a name suffix. Arguments whose type is matched
5659 against another type do not. For example, the <tt>llvm.ctpop</tt> function
5660 can take an integer of any width and returns an integer of exactly the same
5661 integer width. This leads to a family of functions such as
5662 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
5663 %val)</tt>. Only one type, the return type, is overloaded, and only one type
5664 suffix is required. Because the argument's type is matched against the return
5665 type, it does not require its own name suffix.</p>
5667 <p>To learn how to add an intrinsic function, please see the
5668 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
5670 <!-- ======================================================================= -->
5672 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
5677 <p>Variable argument support is defined in LLVM with
5678 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
5679 intrinsic functions. These functions are related to the similarly named
5680 macros defined in the <tt><stdarg.h></tt> header file.</p>
5682 <p>All of these functions operate on arguments that use a target-specific value
5683 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
5684 not define what this type is, so all transformations should be prepared to
5685 handle these functions regardless of the type used.</p>
5687 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
5688 instruction and the variable argument handling intrinsic functions are
5691 <pre class="doc_code">
5692 define i32 @test(i32 %X, ...) {
5693 ; Initialize variable argument processing
5695 %ap2 = bitcast i8** %ap to i8*
5696 call void @llvm.va_start(i8* %ap2)
5698 ; Read a single integer argument
5699 %tmp = va_arg i8** %ap, i32
5701 ; Demonstrate usage of llvm.va_copy and llvm.va_end
5703 %aq2 = bitcast i8** %aq to i8*
5704 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
5705 call void @llvm.va_end(i8* %aq2)
5707 ; Stop processing of arguments.
5708 call void @llvm.va_end(i8* %ap2)
5712 declare void @llvm.va_start(i8*)
5713 declare void @llvm.va_copy(i8*, i8*)
5714 declare void @llvm.va_end(i8*)
5717 <!-- _______________________________________________________________________ -->
5719 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
5727 declare void %llvm.va_start(i8* <arglist>)
5731 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*<arglist></tt>
5732 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
5735 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
5738 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
5739 macro available in C. In a target-dependent way, it initializes
5740 the <tt>va_list</tt> element to which the argument points, so that the next
5741 call to <tt>va_arg</tt> will produce the first variable argument passed to
5742 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
5743 need to know the last argument of the function as the compiler can figure
5748 <!-- _______________________________________________________________________ -->
5750 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
5757 declare void @llvm.va_end(i8* <arglist>)
5761 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
5762 which has been initialized previously
5763 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
5764 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
5767 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
5770 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
5771 macro available in C. In a target-dependent way, it destroys
5772 the <tt>va_list</tt> element to which the argument points. Calls
5773 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
5774 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
5775 with calls to <tt>llvm.va_end</tt>.</p>
5779 <!-- _______________________________________________________________________ -->
5781 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
5788 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
5792 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
5793 from the source argument list to the destination argument list.</p>
5796 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
5797 The second argument is a pointer to a <tt>va_list</tt> element to copy
5801 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
5802 macro available in C. In a target-dependent way, it copies the
5803 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
5804 element. This intrinsic is necessary because
5805 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
5806 arbitrarily complex and require, for example, memory allocation.</p>
5812 <!-- ======================================================================= -->
5814 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
5819 <p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
5820 Collection</a> (GC) requires the implementation and generation of these
5821 intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
5822 roots on the stack</a>, as well as garbage collector implementations that
5823 require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
5824 barriers. Front-ends for type-safe garbage collected languages should generate
5825 these intrinsics to make use of the LLVM garbage collectors. For more details,
5826 see <a href="GarbageCollection.html">Accurate Garbage Collection with
5829 <p>The garbage collection intrinsics only operate on objects in the generic
5830 address space (address space zero).</p>
5832 <!-- _______________________________________________________________________ -->
5834 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
5841 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
5845 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
5846 the code generator, and allows some metadata to be associated with it.</p>
5849 <p>The first argument specifies the address of a stack object that contains the
5850 root pointer. The second pointer (which must be either a constant or a
5851 global value address) contains the meta-data to be associated with the
5855 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
5856 location. At compile-time, the code generator generates information to allow
5857 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5858 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5863 <!-- _______________________________________________________________________ -->
5865 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
5872 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
5876 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
5877 locations, allowing garbage collector implementations that require read
5881 <p>The second argument is the address to read from, which should be an address
5882 allocated from the garbage collector. The first object is a pointer to the
5883 start of the referenced object, if needed by the language runtime (otherwise
5887 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
5888 instruction, but may be replaced with substantially more complex code by the
5889 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5890 may only be used in a function which <a href="#gc">specifies a GC
5895 <!-- _______________________________________________________________________ -->
5897 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
5904 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
5908 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
5909 locations, allowing garbage collector implementations that require write
5910 barriers (such as generational or reference counting collectors).</p>
5913 <p>The first argument is the reference to store, the second is the start of the
5914 object to store it to, and the third is the address of the field of Obj to
5915 store to. If the runtime does not require a pointer to the object, Obj may
5919 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
5920 instruction, but may be replaced with substantially more complex code by the
5921 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5922 may only be used in a function which <a href="#gc">specifies a GC
5929 <!-- ======================================================================= -->
5931 <a name="int_codegen">Code Generator Intrinsics</a>
5936 <p>These intrinsics are provided by LLVM to expose special features that may
5937 only be implemented with code generator support.</p>
5939 <!-- _______________________________________________________________________ -->
5941 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
5948 declare i8 *@llvm.returnaddress(i32 <level>)
5952 <p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5953 target-specific value indicating the return address of the current function
5954 or one of its callers.</p>
5957 <p>The argument to this intrinsic indicates which function to return the address
5958 for. Zero indicates the calling function, one indicates its caller, etc.
5959 The argument is <b>required</b> to be a constant integer value.</p>
5962 <p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
5963 indicating the return address of the specified call frame, or zero if it
5964 cannot be identified. The value returned by this intrinsic is likely to be
5965 incorrect or 0 for arguments other than zero, so it should only be used for
5966 debugging purposes.</p>
5968 <p>Note that calling this intrinsic does not prevent function inlining or other
5969 aggressive transformations, so the value returned may not be that of the
5970 obvious source-language caller.</p>
5974 <!-- _______________________________________________________________________ -->
5976 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5983 declare i8* @llvm.frameaddress(i32 <level>)
5987 <p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5988 target-specific frame pointer value for the specified stack frame.</p>
5991 <p>The argument to this intrinsic indicates which function to return the frame
5992 pointer for. Zero indicates the calling function, one indicates its caller,
5993 etc. The argument is <b>required</b> to be a constant integer value.</p>
5996 <p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
5997 indicating the frame address of the specified call frame, or zero if it
5998 cannot be identified. The value returned by this intrinsic is likely to be
5999 incorrect or 0 for arguments other than zero, so it should only be used for
6000 debugging purposes.</p>
6002 <p>Note that calling this intrinsic does not prevent function inlining or other
6003 aggressive transformations, so the value returned may not be that of the
6004 obvious source-language caller.</p>
6008 <!-- _______________________________________________________________________ -->
6010 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
6017 declare i8* @llvm.stacksave()
6021 <p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
6022 of the function stack, for use
6023 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
6024 useful for implementing language features like scoped automatic variable
6025 sized arrays in C99.</p>
6028 <p>This intrinsic returns a opaque pointer value that can be passed
6029 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
6030 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
6031 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
6032 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
6033 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
6034 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
6038 <!-- _______________________________________________________________________ -->
6040 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
6047 declare void @llvm.stackrestore(i8* %ptr)
6051 <p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
6052 the function stack to the state it was in when the
6053 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
6054 executed. This is useful for implementing language features like scoped
6055 automatic variable sized arrays in C99.</p>
6058 <p>See the description
6059 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
6063 <!-- _______________________________________________________________________ -->
6065 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
6072 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>, i32 <cache type>)
6076 <p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
6077 insert a prefetch instruction if supported; otherwise, it is a noop.
6078 Prefetches have no effect on the behavior of the program but can change its
6079 performance characteristics.</p>
6082 <p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
6083 specifier determining if the fetch should be for a read (0) or write (1),
6084 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
6085 locality, to (3) - extremely local keep in cache. The <tt>cache type</tt>
6086 specifies whether the prefetch is performed on the data (1) or instruction (0)
6087 cache. The <tt>rw</tt>, <tt>locality</tt> and <tt>cache type</tt> arguments
6088 must be constant integers.</p>
6091 <p>This intrinsic does not modify the behavior of the program. In particular,
6092 prefetches cannot trap and do not produce a value. On targets that support
6093 this intrinsic, the prefetch can provide hints to the processor cache for
6094 better performance.</p>
6098 <!-- _______________________________________________________________________ -->
6100 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
6107 declare void @llvm.pcmarker(i32 <id>)
6111 <p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
6112 Counter (PC) in a region of code to simulators and other tools. The method
6113 is target specific, but it is expected that the marker will use exported
6114 symbols to transmit the PC of the marker. The marker makes no guarantees
6115 that it will remain with any specific instruction after optimizations. It is
6116 possible that the presence of a marker will inhibit optimizations. The
6117 intended use is to be inserted after optimizations to allow correlations of
6118 simulation runs.</p>
6121 <p><tt>id</tt> is a numerical id identifying the marker.</p>
6124 <p>This intrinsic does not modify the behavior of the program. Backends that do
6125 not support this intrinsic may ignore it.</p>
6129 <!-- _______________________________________________________________________ -->
6131 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
6138 declare i64 @llvm.readcyclecounter()
6142 <p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
6143 counter register (or similar low latency, high accuracy clocks) on those
6144 targets that support it. On X86, it should map to RDTSC. On Alpha, it
6145 should map to RPCC. As the backing counters overflow quickly (on the order
6146 of 9 seconds on alpha), this should only be used for small timings.</p>
6149 <p>When directly supported, reading the cycle counter should not modify any
6150 memory. Implementations are allowed to either return a application specific
6151 value or a system wide value. On backends without support, this is lowered
6152 to a constant 0.</p>
6158 <!-- ======================================================================= -->
6160 <a name="int_libc">Standard C Library Intrinsics</a>
6165 <p>LLVM provides intrinsics for a few important standard C library functions.
6166 These intrinsics allow source-language front-ends to pass information about
6167 the alignment of the pointer arguments to the code generator, providing
6168 opportunity for more efficient code generation.</p>
6170 <!-- _______________________________________________________________________ -->
6172 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
6178 <p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
6179 integer bit width and for different address spaces. Not all targets support
6180 all bit widths however.</p>
6183 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
6184 i32 <len>, i32 <align>, i1 <isvolatile>)
6185 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
6186 i64 <len>, i32 <align>, i1 <isvolatile>)
6190 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6191 source location to the destination location.</p>
6193 <p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
6194 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6195 and the pointers can be in specified address spaces.</p>
6199 <p>The first argument is a pointer to the destination, the second is a pointer
6200 to the source. The third argument is an integer argument specifying the
6201 number of bytes to copy, the fourth argument is the alignment of the
6202 source and destination locations, and the fifth is a boolean indicating a
6203 volatile access.</p>
6205 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6206 then the caller guarantees that both the source and destination pointers are
6207 aligned to that boundary.</p>
6209 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6210 <tt>llvm.memcpy</tt> call is a <a href="#volatile">volatile operation</a>.
6211 The detailed access behavior is not very cleanly specified and it is unwise
6212 to depend on it.</p>
6216 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6217 source location to the destination location, which are not allowed to
6218 overlap. It copies "len" bytes of memory over. If the argument is known to
6219 be aligned to some boundary, this can be specified as the fourth argument,
6220 otherwise it should be set to 0 or 1.</p>
6224 <!-- _______________________________________________________________________ -->
6226 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
6232 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
6233 width and for different address space. Not all targets support all bit
6237 declare void @llvm.memmove.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
6238 i32 <len>, i32 <align>, i1 <isvolatile>)
6239 declare void @llvm.memmove.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
6240 i64 <len>, i32 <align>, i1 <isvolatile>)
6244 <p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
6245 source location to the destination location. It is similar to the
6246 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
6249 <p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
6250 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6251 and the pointers can be in specified address spaces.</p>
6255 <p>The first argument is a pointer to the destination, the second is a pointer
6256 to the source. The third argument is an integer argument specifying the
6257 number of bytes to copy, the fourth argument is the alignment of the
6258 source and destination locations, and the fifth is a boolean indicating a
6259 volatile access.</p>
6261 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6262 then the caller guarantees that the source and destination pointers are
6263 aligned to that boundary.</p>
6265 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6266 <tt>llvm.memmove</tt> call is a <a href="#volatile">volatile operation</a>.
6267 The detailed access behavior is not very cleanly specified and it is unwise
6268 to depend on it.</p>
6272 <p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
6273 source location to the destination location, which may overlap. It copies
6274 "len" bytes of memory over. If the argument is known to be aligned to some
6275 boundary, this can be specified as the fourth argument, otherwise it should
6276 be set to 0 or 1.</p>
6280 <!-- _______________________________________________________________________ -->
6282 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
6288 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
6289 width and for different address spaces. However, not all targets support all
6293 declare void @llvm.memset.p0i8.i32(i8* <dest>, i8 <val>,
6294 i32 <len>, i32 <align>, i1 <isvolatile>)
6295 declare void @llvm.memset.p0i8.i64(i8* <dest>, i8 <val>,
6296 i64 <len>, i32 <align>, i1 <isvolatile>)
6300 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
6301 particular byte value.</p>
6303 <p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
6304 intrinsic does not return a value and takes extra alignment/volatile
6305 arguments. Also, the destination can be in an arbitrary address space.</p>
6308 <p>The first argument is a pointer to the destination to fill, the second is the
6309 byte value with which to fill it, the third argument is an integer argument
6310 specifying the number of bytes to fill, and the fourth argument is the known
6311 alignment of the destination location.</p>
6313 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6314 then the caller guarantees that the destination pointer is aligned to that
6317 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6318 <tt>llvm.memset</tt> call is a <a href="#volatile">volatile operation</a>.
6319 The detailed access behavior is not very cleanly specified and it is unwise
6320 to depend on it.</p>
6323 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
6324 at the destination location. If the argument is known to be aligned to some
6325 boundary, this can be specified as the fourth argument, otherwise it should
6326 be set to 0 or 1.</p>
6330 <!-- _______________________________________________________________________ -->
6332 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
6338 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
6339 floating point or vector of floating point type. Not all targets support all
6343 declare float @llvm.sqrt.f32(float %Val)
6344 declare double @llvm.sqrt.f64(double %Val)
6345 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
6346 declare fp128 @llvm.sqrt.f128(fp128 %Val)
6347 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
6351 <p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
6352 returning the same value as the libm '<tt>sqrt</tt>' functions would.
6353 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
6354 behavior for negative numbers other than -0.0 (which allows for better
6355 optimization, because there is no need to worry about errno being
6356 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
6359 <p>The argument and return value are floating point numbers of the same
6363 <p>This function returns the sqrt of the specified operand if it is a
6364 nonnegative floating point number.</p>
6368 <!-- _______________________________________________________________________ -->
6370 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
6376 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
6377 floating point or vector of floating point type. Not all targets support all
6381 declare float @llvm.powi.f32(float %Val, i32 %power)
6382 declare double @llvm.powi.f64(double %Val, i32 %power)
6383 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
6384 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
6385 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
6389 <p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
6390 specified (positive or negative) power. The order of evaluation of
6391 multiplications is not defined. When a vector of floating point type is
6392 used, the second argument remains a scalar integer value.</p>
6395 <p>The second argument is an integer power, and the first is a value to raise to
6399 <p>This function returns the first value raised to the second power with an
6400 unspecified sequence of rounding operations.</p>
6404 <!-- _______________________________________________________________________ -->
6406 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
6412 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
6413 floating point or vector of floating point type. Not all targets support all
6417 declare float @llvm.sin.f32(float %Val)
6418 declare double @llvm.sin.f64(double %Val)
6419 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
6420 declare fp128 @llvm.sin.f128(fp128 %Val)
6421 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
6425 <p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
6428 <p>The argument and return value are floating point numbers of the same
6432 <p>This function returns the sine of the specified operand, returning the same
6433 values as the libm <tt>sin</tt> functions would, and handles error conditions
6434 in the same way.</p>
6438 <!-- _______________________________________________________________________ -->
6440 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
6446 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
6447 floating point or vector of floating point type. Not all targets support all
6451 declare float @llvm.cos.f32(float %Val)
6452 declare double @llvm.cos.f64(double %Val)
6453 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
6454 declare fp128 @llvm.cos.f128(fp128 %Val)
6455 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
6459 <p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
6462 <p>The argument and return value are floating point numbers of the same
6466 <p>This function returns the cosine of the specified operand, returning the same
6467 values as the libm <tt>cos</tt> functions would, and handles error conditions
6468 in the same way.</p>
6472 <!-- _______________________________________________________________________ -->
6474 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
6480 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
6481 floating point or vector of floating point type. Not all targets support all
6485 declare float @llvm.pow.f32(float %Val, float %Power)
6486 declare double @llvm.pow.f64(double %Val, double %Power)
6487 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
6488 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
6489 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
6493 <p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
6494 specified (positive or negative) power.</p>
6497 <p>The second argument is a floating point power, and the first is a value to
6498 raise to that power.</p>
6501 <p>This function returns the first value raised to the second power, returning
6502 the same values as the libm <tt>pow</tt> functions would, and handles error
6503 conditions in the same way.</p>
6509 <!-- _______________________________________________________________________ -->
6511 <a name="int_exp">'<tt>llvm.exp.*</tt>' Intrinsic</a>
6517 <p>This is an overloaded intrinsic. You can use <tt>llvm.exp</tt> on any
6518 floating point or vector of floating point type. Not all targets support all
6522 declare float @llvm.exp.f32(float %Val)
6523 declare double @llvm.exp.f64(double %Val)
6524 declare x86_fp80 @llvm.exp.f80(x86_fp80 %Val)
6525 declare fp128 @llvm.exp.f128(fp128 %Val)
6526 declare ppc_fp128 @llvm.exp.ppcf128(ppc_fp128 %Val)
6530 <p>The '<tt>llvm.exp.*</tt>' intrinsics perform the exp function.</p>
6533 <p>The argument and return value are floating point numbers of the same
6537 <p>This function returns the same values as the libm <tt>exp</tt> functions
6538 would, and handles error conditions in the same way.</p>
6542 <!-- _______________________________________________________________________ -->
6544 <a name="int_log">'<tt>llvm.log.*</tt>' Intrinsic</a>
6550 <p>This is an overloaded intrinsic. You can use <tt>llvm.log</tt> on any
6551 floating point or vector of floating point type. Not all targets support all
6555 declare float @llvm.log.f32(float %Val)
6556 declare double @llvm.log.f64(double %Val)
6557 declare x86_fp80 @llvm.log.f80(x86_fp80 %Val)
6558 declare fp128 @llvm.log.f128(fp128 %Val)
6559 declare ppc_fp128 @llvm.log.ppcf128(ppc_fp128 %Val)
6563 <p>The '<tt>llvm.log.*</tt>' intrinsics perform the log function.</p>
6566 <p>The argument and return value are floating point numbers of the same
6570 <p>This function returns the same values as the libm <tt>log</tt> functions
6571 would, and handles error conditions in the same way.</p>
6575 <!-- ======================================================================= -->
6577 <a name="int_manip">Bit Manipulation Intrinsics</a>
6582 <p>LLVM provides intrinsics for a few important bit manipulation operations.
6583 These allow efficient code generation for some algorithms.</p>
6585 <!-- _______________________________________________________________________ -->
6587 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
6593 <p>This is an overloaded intrinsic function. You can use bswap on any integer
6594 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
6597 declare i16 @llvm.bswap.i16(i16 <id>)
6598 declare i32 @llvm.bswap.i32(i32 <id>)
6599 declare i64 @llvm.bswap.i64(i64 <id>)
6603 <p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
6604 values with an even number of bytes (positive multiple of 16 bits). These
6605 are useful for performing operations on data that is not in the target's
6606 native byte order.</p>
6609 <p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
6610 and low byte of the input i16 swapped. Similarly,
6611 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
6612 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
6613 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
6614 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
6615 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
6616 more, respectively).</p>
6620 <!-- _______________________________________________________________________ -->
6622 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
6628 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
6629 width, or on any vector with integer elements. Not all targets support all
6630 bit widths or vector types, however.</p>
6633 declare i8 @llvm.ctpop.i8(i8 <src>)
6634 declare i16 @llvm.ctpop.i16(i16 <src>)
6635 declare i32 @llvm.ctpop.i32(i32 <src>)
6636 declare i64 @llvm.ctpop.i64(i64 <src>)
6637 declare i256 @llvm.ctpop.i256(i256 <src>)
6638 declare <2 x i32> @llvm.ctpop.v2i32(<2 x i32> <src>)
6642 <p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
6646 <p>The only argument is the value to be counted. The argument may be of any
6647 integer type, or a vector with integer elements.
6648 The return type must match the argument type.</p>
6651 <p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable, or within each
6652 element of a vector.</p>
6656 <!-- _______________________________________________________________________ -->
6658 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
6664 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
6665 integer bit width, or any vector whose elements are integers. Not all
6666 targets support all bit widths or vector types, however.</p>
6669 declare i8 @llvm.ctlz.i8 (i8 <src>)
6670 declare i16 @llvm.ctlz.i16(i16 <src>)
6671 declare i32 @llvm.ctlz.i32(i32 <src>)
6672 declare i64 @llvm.ctlz.i64(i64 <src>)
6673 declare i256 @llvm.ctlz.i256(i256 <src>)
6674 declare <2 x i32> @llvm.ctlz.v2i32(<2 x i32> <src;gt)
6678 <p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
6679 leading zeros in a variable.</p>
6682 <p>The only argument is the value to be counted. The argument may be of any
6683 integer type, or any vector type with integer element type.
6684 The return type must match the argument type.</p>
6687 <p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
6688 zeros in a variable, or within each element of the vector if the operation
6689 is of vector type. If the src == 0 then the result is the size in bits of
6690 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
6694 <!-- _______________________________________________________________________ -->
6696 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
6702 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
6703 integer bit width, or any vector of integer elements. Not all targets
6704 support all bit widths or vector types, however.</p>
6707 declare i8 @llvm.cttz.i8 (i8 <src>)
6708 declare i16 @llvm.cttz.i16(i16 <src>)
6709 declare i32 @llvm.cttz.i32(i32 <src>)
6710 declare i64 @llvm.cttz.i64(i64 <src>)
6711 declare i256 @llvm.cttz.i256(i256 <src>)
6712 declase <2 x i32> @llvm.cttz.v2i32(<2 x i32> <src>)
6716 <p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
6720 <p>The only argument is the value to be counted. The argument may be of any
6721 integer type, or a vectory with integer element type.. The return type
6722 must match the argument type.</p>
6725 <p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
6726 zeros in a variable, or within each element of a vector.
6727 If the src == 0 then the result is the size in bits of
6728 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
6734 <!-- ======================================================================= -->
6736 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
6741 <p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
6743 <!-- _______________________________________________________________________ -->
6745 <a name="int_sadd_overflow">
6746 '<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics
6753 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
6754 on any integer bit width.</p>
6757 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
6758 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6759 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6763 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6764 a signed addition of the two arguments, and indicate whether an overflow
6765 occurred during the signed summation.</p>
6768 <p>The arguments (%a and %b) and the first element of the result structure may
6769 be of integer types of any bit width, but they must have the same bit
6770 width. The second element of the result structure must be of
6771 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6772 undergo signed addition.</p>
6775 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6776 a signed addition of the two variables. They return a structure — the
6777 first element of which is the signed summation, and the second element of
6778 which is a bit specifying if the signed summation resulted in an
6783 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6784 %sum = extractvalue {i32, i1} %res, 0
6785 %obit = extractvalue {i32, i1} %res, 1
6786 br i1 %obit, label %overflow, label %normal
6791 <!-- _______________________________________________________________________ -->
6793 <a name="int_uadd_overflow">
6794 '<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics
6801 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
6802 on any integer bit width.</p>
6805 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6806 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6807 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6811 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6812 an unsigned addition of the two arguments, and indicate whether a carry
6813 occurred during the unsigned summation.</p>
6816 <p>The arguments (%a and %b) and the first element of the result structure may
6817 be of integer types of any bit width, but they must have the same bit
6818 width. The second element of the result structure must be of
6819 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6820 undergo unsigned addition.</p>
6823 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6824 an unsigned addition of the two arguments. They return a structure —
6825 the first element of which is the sum, and the second element of which is a
6826 bit specifying if the unsigned summation resulted in a carry.</p>
6830 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6831 %sum = extractvalue {i32, i1} %res, 0
6832 %obit = extractvalue {i32, i1} %res, 1
6833 br i1 %obit, label %carry, label %normal
6838 <!-- _______________________________________________________________________ -->
6840 <a name="int_ssub_overflow">
6841 '<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics
6848 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
6849 on any integer bit width.</p>
6852 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6853 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6854 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6858 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6859 a signed subtraction of the two arguments, and indicate whether an overflow
6860 occurred during the signed subtraction.</p>
6863 <p>The arguments (%a and %b) and the first element of the result structure may
6864 be of integer types of any bit width, but they must have the same bit
6865 width. The second element of the result structure must be of
6866 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6867 undergo signed subtraction.</p>
6870 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6871 a signed subtraction of the two arguments. They return a structure —
6872 the first element of which is the subtraction, and the second element of
6873 which is a bit specifying if the signed subtraction resulted in an
6878 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6879 %sum = extractvalue {i32, i1} %res, 0
6880 %obit = extractvalue {i32, i1} %res, 1
6881 br i1 %obit, label %overflow, label %normal
6886 <!-- _______________________________________________________________________ -->
6888 <a name="int_usub_overflow">
6889 '<tt>llvm.usub.with.overflow.*</tt>' Intrinsics
6896 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
6897 on any integer bit width.</p>
6900 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6901 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6902 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6906 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6907 an unsigned subtraction of the two arguments, and indicate whether an
6908 overflow occurred during the unsigned subtraction.</p>
6911 <p>The arguments (%a and %b) and the first element of the result structure may
6912 be of integer types of any bit width, but they must have the same bit
6913 width. The second element of the result structure must be of
6914 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6915 undergo unsigned subtraction.</p>
6918 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6919 an unsigned subtraction of the two arguments. They return a structure —
6920 the first element of which is the subtraction, and the second element of
6921 which is a bit specifying if the unsigned subtraction resulted in an
6926 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6927 %sum = extractvalue {i32, i1} %res, 0
6928 %obit = extractvalue {i32, i1} %res, 1
6929 br i1 %obit, label %overflow, label %normal
6934 <!-- _______________________________________________________________________ -->
6936 <a name="int_smul_overflow">
6937 '<tt>llvm.smul.with.overflow.*</tt>' Intrinsics
6944 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
6945 on any integer bit width.</p>
6948 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6949 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6950 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6955 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6956 a signed multiplication of the two arguments, and indicate whether an
6957 overflow occurred during the signed multiplication.</p>
6960 <p>The arguments (%a and %b) and the first element of the result structure may
6961 be of integer types of any bit width, but they must have the same bit
6962 width. The second element of the result structure must be of
6963 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6964 undergo signed multiplication.</p>
6967 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6968 a signed multiplication of the two arguments. They return a structure —
6969 the first element of which is the multiplication, and the second element of
6970 which is a bit specifying if the signed multiplication resulted in an
6975 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6976 %sum = extractvalue {i32, i1} %res, 0
6977 %obit = extractvalue {i32, i1} %res, 1
6978 br i1 %obit, label %overflow, label %normal
6983 <!-- _______________________________________________________________________ -->
6985 <a name="int_umul_overflow">
6986 '<tt>llvm.umul.with.overflow.*</tt>' Intrinsics
6993 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6994 on any integer bit width.</p>
6997 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6998 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6999 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
7003 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
7004 a unsigned multiplication of the two arguments, and indicate whether an
7005 overflow occurred during the unsigned multiplication.</p>
7008 <p>The arguments (%a and %b) and the first element of the result structure may
7009 be of integer types of any bit width, but they must have the same bit
7010 width. The second element of the result structure must be of
7011 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
7012 undergo unsigned multiplication.</p>
7015 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
7016 an unsigned multiplication of the two arguments. They return a structure
7017 — the first element of which is the multiplication, and the second
7018 element of which is a bit specifying if the unsigned multiplication resulted
7023 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
7024 %sum = extractvalue {i32, i1} %res, 0
7025 %obit = extractvalue {i32, i1} %res, 1
7026 br i1 %obit, label %overflow, label %normal
7033 <!-- ======================================================================= -->
7035 <a name="int_fp16">Half Precision Floating Point Intrinsics</a>
7040 <p>Half precision floating point is a storage-only format. This means that it is
7041 a dense encoding (in memory) but does not support computation in the
7044 <p>This means that code must first load the half-precision floating point
7045 value as an i16, then convert it to float with <a
7046 href="#int_convert_from_fp16"><tt>llvm.convert.from.fp16</tt></a>.
7047 Computation can then be performed on the float value (including extending to
7048 double etc). To store the value back to memory, it is first converted to
7049 float if needed, then converted to i16 with
7050 <a href="#int_convert_to_fp16"><tt>llvm.convert.to.fp16</tt></a>, then
7051 storing as an i16 value.</p>
7053 <!-- _______________________________________________________________________ -->
7055 <a name="int_convert_to_fp16">
7056 '<tt>llvm.convert.to.fp16</tt>' Intrinsic
7064 declare i16 @llvm.convert.to.fp16(f32 %a)
7068 <p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
7069 a conversion from single precision floating point format to half precision
7070 floating point format.</p>
7073 <p>The intrinsic function contains single argument - the value to be
7077 <p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
7078 a conversion from single precision floating point format to half precision
7079 floating point format. The return value is an <tt>i16</tt> which
7080 contains the converted number.</p>
7084 %res = call i16 @llvm.convert.to.fp16(f32 %a)
7085 store i16 %res, i16* @x, align 2
7090 <!-- _______________________________________________________________________ -->
7092 <a name="int_convert_from_fp16">
7093 '<tt>llvm.convert.from.fp16</tt>' Intrinsic
7101 declare f32 @llvm.convert.from.fp16(i16 %a)
7105 <p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs
7106 a conversion from half precision floating point format to single precision
7107 floating point format.</p>
7110 <p>The intrinsic function contains single argument - the value to be
7114 <p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs a
7115 conversion from half single precision floating point format to single
7116 precision floating point format. The input half-float value is represented by
7117 an <tt>i16</tt> value.</p>
7121 %a = load i16* @x, align 2
7122 %res = call f32 @llvm.convert.from.fp16(i16 %a)
7129 <!-- ======================================================================= -->
7131 <a name="int_debugger">Debugger Intrinsics</a>
7136 <p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
7137 prefix), are described in
7138 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
7139 Level Debugging</a> document.</p>
7143 <!-- ======================================================================= -->
7145 <a name="int_eh">Exception Handling Intrinsics</a>
7150 <p>The LLVM exception handling intrinsics (which all start with
7151 <tt>llvm.eh.</tt> prefix), are described in
7152 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
7153 Handling</a> document.</p>
7157 <!-- ======================================================================= -->
7159 <a name="int_trampoline">Trampoline Intrinsic</a>
7164 <p>This intrinsic makes it possible to excise one parameter, marked with
7165 the <a href="#nest"><tt>nest</tt></a> attribute, from a function.
7166 The result is a callable
7167 function pointer lacking the nest parameter - the caller does not need to
7168 provide a value for it. Instead, the value to use is stored in advance in a
7169 "trampoline", a block of memory usually allocated on the stack, which also
7170 contains code to splice the nest value into the argument list. This is used
7171 to implement the GCC nested function address extension.</p>
7173 <p>For example, if the function is
7174 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
7175 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
7178 <pre class="doc_code">
7179 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
7180 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
7181 %p = call i8* @llvm.init.trampoline(i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval)
7182 %fp = bitcast i8* %p to i32 (i32, i32)*
7185 <p>The call <tt>%val = call i32 %fp(i32 %x, i32 %y)</tt> is then equivalent
7186 to <tt>%val = call i32 %f(i8* %nval, i32 %x, i32 %y)</tt>.</p>
7188 <!-- _______________________________________________________________________ -->
7191 '<tt>llvm.init.trampoline</tt>' Intrinsic
7199 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
7203 <p>This fills the memory pointed to by <tt>tramp</tt> with code and returns a
7204 function pointer suitable for executing it.</p>
7207 <p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
7208 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
7209 sufficiently aligned block of memory; this memory is written to by the
7210 intrinsic. Note that the size and the alignment are target-specific - LLVM
7211 currently provides no portable way of determining them, so a front-end that
7212 generates this intrinsic needs to have some target-specific knowledge.
7213 The <tt>func</tt> argument must hold a function bitcast to
7214 an <tt>i8*</tt>.</p>
7217 <p>The block of memory pointed to by <tt>tramp</tt> is filled with target
7218 dependent code, turning it into a function. A pointer to this function is
7219 returned, but needs to be bitcast to an <a href="#int_trampoline">appropriate
7220 function pointer type</a> before being called. The new function's signature
7221 is the same as that of <tt>func</tt> with any arguments marked with
7222 the <tt>nest</tt> attribute removed. At most one such <tt>nest</tt> argument
7223 is allowed, and it must be of pointer type. Calling the new function is
7224 equivalent to calling <tt>func</tt> with the same argument list, but
7225 with <tt>nval</tt> used for the missing <tt>nest</tt> argument. If, after
7226 calling <tt>llvm.init.trampoline</tt>, the memory pointed to
7227 by <tt>tramp</tt> is modified, then the effect of any later call to the
7228 returned function pointer is undefined.</p>
7234 <!-- ======================================================================= -->
7236 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
7241 <p>These intrinsic functions expand the "universal IR" of LLVM to represent
7242 hardware constructs for atomic operations and memory synchronization. This
7243 provides an interface to the hardware, not an interface to the programmer. It
7244 is aimed at a low enough level to allow any programming models or APIs
7245 (Application Programming Interfaces) which need atomic behaviors to map
7246 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
7247 hardware provides a "universal IR" for source languages, it also provides a
7248 starting point for developing a "universal" atomic operation and
7249 synchronization IR.</p>
7251 <p>These do <em>not</em> form an API such as high-level threading libraries,
7252 software transaction memory systems, atomic primitives, and intrinsic
7253 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
7254 application libraries. The hardware interface provided by LLVM should allow
7255 a clean implementation of all of these APIs and parallel programming models.
7256 No one model or paradigm should be selected above others unless the hardware
7257 itself ubiquitously does so.</p>
7259 <!-- _______________________________________________________________________ -->
7261 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
7267 declare void @llvm.memory.barrier(i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>, i1 <device>)
7271 <p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
7272 specific pairs of memory access types.</p>
7275 <p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
7276 The first four arguments enables a specific barrier as listed below. The
7277 fifth argument specifies that the barrier applies to io or device or uncached
7281 <li><tt>ll</tt>: load-load barrier</li>
7282 <li><tt>ls</tt>: load-store barrier</li>
7283 <li><tt>sl</tt>: store-load barrier</li>
7284 <li><tt>ss</tt>: store-store barrier</li>
7285 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
7289 <p>This intrinsic causes the system to enforce some ordering constraints upon
7290 the loads and stores of the program. This barrier does not
7291 indicate <em>when</em> any events will occur, it only enforces
7292 an <em>order</em> in which they occur. For any of the specified pairs of load
7293 and store operations (f.ex. load-load, or store-load), all of the first
7294 operations preceding the barrier will complete before any of the second
7295 operations succeeding the barrier begin. Specifically the semantics for each
7296 pairing is as follows:</p>
7299 <li><tt>ll</tt>: All loads before the barrier must complete before any load
7300 after the barrier begins.</li>
7301 <li><tt>ls</tt>: All loads before the barrier must complete before any
7302 store after the barrier begins.</li>
7303 <li><tt>ss</tt>: All stores before the barrier must complete before any
7304 store after the barrier begins.</li>
7305 <li><tt>sl</tt>: All stores before the barrier must complete before any
7306 load after the barrier begins.</li>
7309 <p>These semantics are applied with a logical "and" behavior when more than one
7310 is enabled in a single memory barrier intrinsic.</p>
7312 <p>Backends may implement stronger barriers than those requested when they do
7313 not support as fine grained a barrier as requested. Some architectures do
7314 not need all types of barriers and on such architectures, these become
7319 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7320 %ptr = bitcast i8* %mallocP to i32*
7323 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
7324 call void @llvm.memory.barrier(i1 false, i1 true, i1 false, i1 false, i1 true)
7325 <i>; guarantee the above finishes</i>
7326 store i32 8, %ptr <i>; before this begins</i>
7331 <!-- _______________________________________________________________________ -->
7333 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
7339 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
7340 any integer bit width and for different address spaces. Not all targets
7341 support all bit widths however.</p>
7344 declare i8 @llvm.atomic.cmp.swap.i8.p0i8(i8* <ptr>, i8 <cmp>, i8 <val>)
7345 declare i16 @llvm.atomic.cmp.swap.i16.p0i16(i16* <ptr>, i16 <cmp>, i16 <val>)
7346 declare i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* <ptr>, i32 <cmp>, i32 <val>)
7347 declare i64 @llvm.atomic.cmp.swap.i64.p0i64(i64* <ptr>, i64 <cmp>, i64 <val>)
7351 <p>This loads a value in memory and compares it to a given value. If they are
7352 equal, it stores a new value into the memory.</p>
7355 <p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
7356 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
7357 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
7358 this integer type. While any bit width integer may be used, targets may only
7359 lower representations they support in hardware.</p>
7362 <p>This entire intrinsic must be executed atomically. It first loads the value
7363 in memory pointed to by <tt>ptr</tt> and compares it with the
7364 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
7365 memory. The loaded value is yielded in all cases. This provides the
7366 equivalent of an atomic compare-and-swap operation within the SSA
7371 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7372 %ptr = bitcast i8* %mallocP to i32*
7375 %val1 = add i32 4, 4
7376 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 4, %val1)
7377 <i>; yields {i32}:result1 = 4</i>
7378 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7379 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7381 %val2 = add i32 1, 1
7382 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 5, %val2)
7383 <i>; yields {i32}:result2 = 8</i>
7384 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
7386 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
7391 <!-- _______________________________________________________________________ -->
7393 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
7399 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
7400 integer bit width. Not all targets support all bit widths however.</p>
7403 declare i8 @llvm.atomic.swap.i8.p0i8(i8* <ptr>, i8 <val>)
7404 declare i16 @llvm.atomic.swap.i16.p0i16(i16* <ptr>, i16 <val>)
7405 declare i32 @llvm.atomic.swap.i32.p0i32(i32* <ptr>, i32 <val>)
7406 declare i64 @llvm.atomic.swap.i64.p0i64(i64* <ptr>, i64 <val>)
7410 <p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
7411 the value from memory. It then stores the value in <tt>val</tt> in the memory
7412 at <tt>ptr</tt>.</p>
7415 <p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
7416 the <tt>val</tt> argument and the result must be integers of the same bit
7417 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
7418 integer type. The targets may only lower integer representations they
7422 <p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
7423 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
7424 equivalent of an atomic swap operation within the SSA framework.</p>
7428 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7429 %ptr = bitcast i8* %mallocP to i32*
7432 %val1 = add i32 4, 4
7433 %result1 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val1)
7434 <i>; yields {i32}:result1 = 4</i>
7435 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7436 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7438 %val2 = add i32 1, 1
7439 %result2 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val2)
7440 <i>; yields {i32}:result2 = 8</i>
7442 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
7443 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
7448 <!-- _______________________________________________________________________ -->
7450 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
7456 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
7457 any integer bit width. Not all targets support all bit widths however.</p>
7460 declare i8 @llvm.atomic.load.add.i8.p0i8(i8* <ptr>, i8 <delta>)
7461 declare i16 @llvm.atomic.load.add.i16.p0i16(i16* <ptr>, i16 <delta>)
7462 declare i32 @llvm.atomic.load.add.i32.p0i32(i32* <ptr>, i32 <delta>)
7463 declare i64 @llvm.atomic.load.add.i64.p0i64(i64* <ptr>, i64 <delta>)
7467 <p>This intrinsic adds <tt>delta</tt> to the value stored in memory
7468 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7471 <p>The intrinsic takes two arguments, the first a pointer to an integer value
7472 and the second an integer value. The result is also an integer value. These
7473 integer types can have any bit width, but they must all have the same bit
7474 width. The targets may only lower integer representations they support.</p>
7477 <p>This intrinsic does a series of operations atomically. It first loads the
7478 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
7479 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
7483 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7484 %ptr = bitcast i8* %mallocP to i32*
7486 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 4)
7487 <i>; yields {i32}:result1 = 4</i>
7488 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 2)
7489 <i>; yields {i32}:result2 = 8</i>
7490 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 5)
7491 <i>; yields {i32}:result3 = 10</i>
7492 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
7497 <!-- _______________________________________________________________________ -->
7499 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
7505 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
7506 any integer bit width and for different address spaces. Not all targets
7507 support all bit widths however.</p>
7510 declare i8 @llvm.atomic.load.sub.i8.p0i32(i8* <ptr>, i8 <delta>)
7511 declare i16 @llvm.atomic.load.sub.i16.p0i32(i16* <ptr>, i16 <delta>)
7512 declare i32 @llvm.atomic.load.sub.i32.p0i32(i32* <ptr>, i32 <delta>)
7513 declare i64 @llvm.atomic.load.sub.i64.p0i32(i64* <ptr>, i64 <delta>)
7517 <p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
7518 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7521 <p>The intrinsic takes two arguments, the first a pointer to an integer value
7522 and the second an integer value. The result is also an integer value. These
7523 integer types can have any bit width, but they must all have the same bit
7524 width. The targets may only lower integer representations they support.</p>
7527 <p>This intrinsic does a series of operations atomically. It first loads the
7528 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
7529 result to <tt>ptr</tt>. It yields the original value stored
7530 at <tt>ptr</tt>.</p>
7534 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7535 %ptr = bitcast i8* %mallocP to i32*
7537 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 4)
7538 <i>; yields {i32}:result1 = 8</i>
7539 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 2)
7540 <i>; yields {i32}:result2 = 4</i>
7541 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 5)
7542 <i>; yields {i32}:result3 = 2</i>
7543 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
7548 <!-- _______________________________________________________________________ -->
7550 <a name="int_atomic_load_and">
7551 '<tt>llvm.atomic.load.and.*</tt>' Intrinsic
7554 <a name="int_atomic_load_nand">
7555 '<tt>llvm.atomic.load.nand.*</tt>' Intrinsic
7558 <a name="int_atomic_load_or">
7559 '<tt>llvm.atomic.load.or.*</tt>' Intrinsic
7562 <a name="int_atomic_load_xor">
7563 '<tt>llvm.atomic.load.xor.*</tt>' Intrinsic
7570 <p>These are overloaded intrinsics. You can
7571 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
7572 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
7573 bit width and for different address spaces. Not all targets support all bit
7577 declare i8 @llvm.atomic.load.and.i8.p0i8(i8* <ptr>, i8 <delta>)
7578 declare i16 @llvm.atomic.load.and.i16.p0i16(i16* <ptr>, i16 <delta>)
7579 declare i32 @llvm.atomic.load.and.i32.p0i32(i32* <ptr>, i32 <delta>)
7580 declare i64 @llvm.atomic.load.and.i64.p0i64(i64* <ptr>, i64 <delta>)
7584 declare i8 @llvm.atomic.load.or.i8.p0i8(i8* <ptr>, i8 <delta>)
7585 declare i16 @llvm.atomic.load.or.i16.p0i16(i16* <ptr>, i16 <delta>)
7586 declare i32 @llvm.atomic.load.or.i32.p0i32(i32* <ptr>, i32 <delta>)
7587 declare i64 @llvm.atomic.load.or.i64.p0i64(i64* <ptr>, i64 <delta>)
7591 declare i8 @llvm.atomic.load.nand.i8.p0i32(i8* <ptr>, i8 <delta>)
7592 declare i16 @llvm.atomic.load.nand.i16.p0i32(i16* <ptr>, i16 <delta>)
7593 declare i32 @llvm.atomic.load.nand.i32.p0i32(i32* <ptr>, i32 <delta>)
7594 declare i64 @llvm.atomic.load.nand.i64.p0i32(i64* <ptr>, i64 <delta>)
7598 declare i8 @llvm.atomic.load.xor.i8.p0i32(i8* <ptr>, i8 <delta>)
7599 declare i16 @llvm.atomic.load.xor.i16.p0i32(i16* <ptr>, i16 <delta>)
7600 declare i32 @llvm.atomic.load.xor.i32.p0i32(i32* <ptr>, i32 <delta>)
7601 declare i64 @llvm.atomic.load.xor.i64.p0i32(i64* <ptr>, i64 <delta>)
7605 <p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
7606 the value stored in memory at <tt>ptr</tt>. It yields the original value
7607 at <tt>ptr</tt>.</p>
7610 <p>These intrinsics take two arguments, the first a pointer to an integer value
7611 and the second an integer value. The result is also an integer value. These
7612 integer types can have any bit width, but they must all have the same bit
7613 width. The targets may only lower integer representations they support.</p>
7616 <p>These intrinsics does a series of operations atomically. They first load the
7617 value stored at <tt>ptr</tt>. They then do the bitwise
7618 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
7619 original value stored at <tt>ptr</tt>.</p>
7623 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7624 %ptr = bitcast i8* %mallocP to i32*
7625 store i32 0x0F0F, %ptr
7626 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32(i32* %ptr, i32 0xFF)
7627 <i>; yields {i32}:result0 = 0x0F0F</i>
7628 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32(i32* %ptr, i32 0xFF)
7629 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
7630 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32(i32* %ptr, i32 0F)
7631 <i>; yields {i32}:result2 = 0xF0</i>
7632 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32(i32* %ptr, i32 0F)
7633 <i>; yields {i32}:result3 = FF</i>
7634 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
7639 <!-- _______________________________________________________________________ -->
7641 <a name="int_atomic_load_max">
7642 '<tt>llvm.atomic.load.max.*</tt>' Intrinsic
7645 <a name="int_atomic_load_min">
7646 '<tt>llvm.atomic.load.min.*</tt>' Intrinsic
7649 <a name="int_atomic_load_umax">
7650 '<tt>llvm.atomic.load.umax.*</tt>' Intrinsic
7653 <a name="int_atomic_load_umin">
7654 '<tt>llvm.atomic.load.umin.*</tt>' Intrinsic
7661 <p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
7662 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
7663 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
7664 address spaces. Not all targets support all bit widths however.</p>
7667 declare i8 @llvm.atomic.load.max.i8.p0i8(i8* <ptr>, i8 <delta>)
7668 declare i16 @llvm.atomic.load.max.i16.p0i16(i16* <ptr>, i16 <delta>)
7669 declare i32 @llvm.atomic.load.max.i32.p0i32(i32* <ptr>, i32 <delta>)
7670 declare i64 @llvm.atomic.load.max.i64.p0i64(i64* <ptr>, i64 <delta>)
7674 declare i8 @llvm.atomic.load.min.i8.p0i8(i8* <ptr>, i8 <delta>)
7675 declare i16 @llvm.atomic.load.min.i16.p0i16(i16* <ptr>, i16 <delta>)
7676 declare i32 @llvm.atomic.load.min.i32.p0i32(i32* <ptr>, i32 <delta>)
7677 declare i64 @llvm.atomic.load.min.i64.p0i64(i64* <ptr>, i64 <delta>)
7681 declare i8 @llvm.atomic.load.umax.i8.p0i8(i8* <ptr>, i8 <delta>)
7682 declare i16 @llvm.atomic.load.umax.i16.p0i16(i16* <ptr>, i16 <delta>)
7683 declare i32 @llvm.atomic.load.umax.i32.p0i32(i32* <ptr>, i32 <delta>)
7684 declare i64 @llvm.atomic.load.umax.i64.p0i64(i64* <ptr>, i64 <delta>)
7688 declare i8 @llvm.atomic.load.umin.i8.p0i8(i8* <ptr>, i8 <delta>)
7689 declare i16 @llvm.atomic.load.umin.i16.p0i16(i16* <ptr>, i16 <delta>)
7690 declare i32 @llvm.atomic.load.umin.i32.p0i32(i32* <ptr>, i32 <delta>)
7691 declare i64 @llvm.atomic.load.umin.i64.p0i64(i64* <ptr>, i64 <delta>)
7695 <p>These intrinsics takes the signed or unsigned minimum or maximum of
7696 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
7697 original value at <tt>ptr</tt>.</p>
7700 <p>These intrinsics take two arguments, the first a pointer to an integer value
7701 and the second an integer value. The result is also an integer value. These
7702 integer types can have any bit width, but they must all have the same bit
7703 width. The targets may only lower integer representations they support.</p>
7706 <p>These intrinsics does a series of operations atomically. They first load the
7707 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
7708 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
7709 yield the original value stored at <tt>ptr</tt>.</p>
7713 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7714 %ptr = bitcast i8* %mallocP to i32*
7716 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32(i32* %ptr, i32 -2)
7717 <i>; yields {i32}:result0 = 7</i>
7718 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32(i32* %ptr, i32 8)
7719 <i>; yields {i32}:result1 = -2</i>
7720 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32(i32* %ptr, i32 10)
7721 <i>; yields {i32}:result2 = 8</i>
7722 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32(i32* %ptr, i32 30)
7723 <i>; yields {i32}:result3 = 8</i>
7724 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
7731 <!-- ======================================================================= -->
7733 <a name="int_memorymarkers">Memory Use Markers</a>
7738 <p>This class of intrinsics exists to information about the lifetime of memory
7739 objects and ranges where variables are immutable.</p>
7741 <!-- _______________________________________________________________________ -->
7743 <a name="int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a>
7750 declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
7754 <p>The '<tt>llvm.lifetime.start</tt>' intrinsic specifies the start of a memory
7755 object's lifetime.</p>
7758 <p>The first argument is a constant integer representing the size of the
7759 object, or -1 if it is variable sized. The second argument is a pointer to
7763 <p>This intrinsic indicates that before this point in the code, the value of the
7764 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7765 never be used and has an undefined value. A load from the pointer that
7766 precedes this intrinsic can be replaced with
7767 <tt>'<a href="#undefvalues">undef</a>'</tt>.</p>
7771 <!-- _______________________________________________________________________ -->
7773 <a name="int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a>
7780 declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
7784 <p>The '<tt>llvm.lifetime.end</tt>' intrinsic specifies the end of a memory
7785 object's lifetime.</p>
7788 <p>The first argument is a constant integer representing the size of the
7789 object, or -1 if it is variable sized. The second argument is a pointer to
7793 <p>This intrinsic indicates that after this point in the code, the value of the
7794 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7795 never be used and has an undefined value. Any stores into the memory object
7796 following this intrinsic may be removed as dead.
7800 <!-- _______________________________________________________________________ -->
7802 <a name="int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a>
7809 declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>)
7813 <p>The '<tt>llvm.invariant.start</tt>' intrinsic specifies that the contents of
7814 a memory object will not change.</p>
7817 <p>The first argument is a constant integer representing the size of the
7818 object, or -1 if it is variable sized. The second argument is a pointer to
7822 <p>This intrinsic indicates that until an <tt>llvm.invariant.end</tt> that uses
7823 the return value, the referenced memory location is constant and
7828 <!-- _______________________________________________________________________ -->
7830 <a name="int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a>
7837 declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
7841 <p>The '<tt>llvm.invariant.end</tt>' intrinsic specifies that the contents of
7842 a memory object are mutable.</p>
7845 <p>The first argument is the matching <tt>llvm.invariant.start</tt> intrinsic.
7846 The second argument is a constant integer representing the size of the
7847 object, or -1 if it is variable sized and the third argument is a pointer
7851 <p>This intrinsic indicates that the memory is mutable again.</p>
7857 <!-- ======================================================================= -->
7859 <a name="int_general">General Intrinsics</a>
7864 <p>This class of intrinsics is designed to be generic and has no specific
7867 <!-- _______________________________________________________________________ -->
7869 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
7876 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int>)
7880 <p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
7883 <p>The first argument is a pointer to a value, the second is a pointer to a
7884 global string, the third is a pointer to a global string which is the source
7885 file name, and the last argument is the line number.</p>
7888 <p>This intrinsic allows annotation of local variables with arbitrary strings.
7889 This can be useful for special purpose optimizations that want to look for
7890 these annotations. These have no other defined use, they are ignored by code
7891 generation and optimization.</p>
7895 <!-- _______________________________________________________________________ -->
7897 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
7903 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
7904 any integer bit width.</p>
7907 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int>)
7908 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int>)
7909 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int>)
7910 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int>)
7911 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int>)
7915 <p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
7918 <p>The first argument is an integer value (result of some expression), the
7919 second is a pointer to a global string, the third is a pointer to a global
7920 string which is the source file name, and the last argument is the line
7921 number. It returns the value of the first argument.</p>
7924 <p>This intrinsic allows annotations to be put on arbitrary expressions with
7925 arbitrary strings. This can be useful for special purpose optimizations that
7926 want to look for these annotations. These have no other defined use, they
7927 are ignored by code generation and optimization.</p>
7931 <!-- _______________________________________________________________________ -->
7933 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
7940 declare void @llvm.trap()
7944 <p>The '<tt>llvm.trap</tt>' intrinsic.</p>
7950 <p>This intrinsics is lowered to the target dependent trap instruction. If the
7951 target does not have a trap instruction, this intrinsic will be lowered to
7952 the call of the <tt>abort()</tt> function.</p>
7956 <!-- _______________________________________________________________________ -->
7958 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
7965 declare void @llvm.stackprotector(i8* <guard>, i8** <slot>)
7969 <p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
7970 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
7971 ensure that it is placed on the stack before local variables.</p>
7974 <p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
7975 arguments. The first argument is the value loaded from the stack
7976 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
7977 that has enough space to hold the value of the guard.</p>
7980 <p>This intrinsic causes the prologue/epilogue inserter to force the position of
7981 the <tt>AllocaInst</tt> stack slot to be before local variables on the
7982 stack. This is to ensure that if a local variable on the stack is
7983 overwritten, it will destroy the value of the guard. When the function exits,
7984 the guard on the stack is checked against the original guard. If they are
7985 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
7990 <!-- _______________________________________________________________________ -->
7992 <a name="int_objectsize">'<tt>llvm.objectsize</tt>' Intrinsic</a>
7999 declare i32 @llvm.objectsize.i32(i8* <object>, i1 <type>)
8000 declare i64 @llvm.objectsize.i64(i8* <object>, i1 <type>)
8004 <p>The <tt>llvm.objectsize</tt> intrinsic is designed to provide information to
8005 the optimizers to determine at compile time whether a) an operation (like
8006 memcpy) will overflow a buffer that corresponds to an object, or b) that a
8007 runtime check for overflow isn't necessary. An object in this context means
8008 an allocation of a specific class, structure, array, or other object.</p>
8011 <p>The <tt>llvm.objectsize</tt> intrinsic takes two arguments. The first
8012 argument is a pointer to or into the <tt>object</tt>. The second argument
8013 is a boolean 0 or 1. This argument determines whether you want the
8014 maximum (0) or minimum (1) bytes remaining. This needs to be a literal 0 or
8015 1, variables are not allowed.</p>
8018 <p>The <tt>llvm.objectsize</tt> intrinsic is lowered to either a constant
8019 representing the size of the object concerned, or <tt>i32/i64 -1 or 0</tt>,
8020 depending on the <tt>type</tt> argument, if the size cannot be determined at
8029 <!-- *********************************************************************** -->
8032 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
8033 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
8034 <a href="http://validator.w3.org/check/referer"><img
8035 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
8037 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
8038 <a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br>
8039 Last modified: $Date$