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 <div class="doc_title"> LLVM Language Reference Manual </div>
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>
244 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
246 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
247 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
248 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
249 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
252 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
254 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
255 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
256 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
257 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
258 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
259 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
262 <li><a href="#int_fp16">Half Precision Floating Point Intrinsics</a>
264 <li><a href="#int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a></li>
265 <li><a href="#int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a></li>
268 <li><a href="#int_debugger">Debugger intrinsics</a></li>
269 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
270 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
272 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
275 <li><a href="#int_atomics">Atomic intrinsics</a>
277 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
278 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
279 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
280 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
281 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
282 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
283 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
284 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
285 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
286 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
287 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
288 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
289 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
292 <li><a href="#int_memorymarkers">Memory Use Markers</a>
294 <li><a href="#int_lifetime_start"><tt>llvm.lifetime.start</tt></a></li>
295 <li><a href="#int_lifetime_end"><tt>llvm.lifetime.end</tt></a></li>
296 <li><a href="#int_invariant_start"><tt>llvm.invariant.start</tt></a></li>
297 <li><a href="#int_invariant_end"><tt>llvm.invariant.end</tt></a></li>
300 <li><a href="#int_general">General intrinsics</a>
302 <li><a href="#int_var_annotation">
303 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
304 <li><a href="#int_annotation">
305 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
306 <li><a href="#int_trap">
307 '<tt>llvm.trap</tt>' Intrinsic</a></li>
308 <li><a href="#int_stackprotector">
309 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
310 <li><a href="#int_objectsize">
311 '<tt>llvm.objectsize</tt>' Intrinsic</a></li>
318 <div class="doc_author">
319 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
320 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
323 <!-- *********************************************************************** -->
324 <div class="doc_section"> <a name="abstract">Abstract </a></div>
325 <!-- *********************************************************************** -->
327 <div class="doc_text">
329 <p>This document is a reference manual for the LLVM assembly language. LLVM is
330 a Static Single Assignment (SSA) based representation that provides type
331 safety, low-level operations, flexibility, and the capability of representing
332 'all' high-level languages cleanly. It is the common code representation
333 used throughout all phases of the LLVM compilation strategy.</p>
337 <!-- *********************************************************************** -->
338 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
339 <!-- *********************************************************************** -->
341 <div class="doc_text">
343 <p>The LLVM code representation is designed to be used in three different forms:
344 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
345 for fast loading by a Just-In-Time compiler), and as a human readable
346 assembly language representation. This allows LLVM to provide a powerful
347 intermediate representation for efficient compiler transformations and
348 analysis, while providing a natural means to debug and visualize the
349 transformations. The three different forms of LLVM are all equivalent. This
350 document describes the human readable representation and notation.</p>
352 <p>The LLVM representation aims to be light-weight and low-level while being
353 expressive, typed, and extensible at the same time. It aims to be a
354 "universal IR" of sorts, by being at a low enough level that high-level ideas
355 may be cleanly mapped to it (similar to how microprocessors are "universal
356 IR's", allowing many source languages to be mapped to them). By providing
357 type information, LLVM can be used as the target of optimizations: for
358 example, through pointer analysis, it can be proven that a C automatic
359 variable is never accessed outside of the current function, allowing it to
360 be promoted to a simple SSA value instead of a memory location.</p>
364 <!-- _______________________________________________________________________ -->
365 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
367 <div class="doc_text">
369 <p>It is important to note that this document describes 'well formed' LLVM
370 assembly language. There is a difference between what the parser accepts and
371 what is considered 'well formed'. For example, the following instruction is
372 syntactically okay, but not well formed:</p>
374 <pre class="doc_code">
375 %x = <a href="#i_add">add</a> i32 1, %x
378 <p>because the definition of <tt>%x</tt> does not dominate all of its uses. The
379 LLVM infrastructure provides a verification pass that may be used to verify
380 that an LLVM module is well formed. This pass is automatically run by the
381 parser after parsing input assembly and by the optimizer before it outputs
382 bitcode. The violations pointed out by the verifier pass indicate bugs in
383 transformation passes or input to the parser.</p>
387 <!-- Describe the typesetting conventions here. -->
389 <!-- *********************************************************************** -->
390 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
391 <!-- *********************************************************************** -->
393 <div class="doc_text">
395 <p>LLVM identifiers come in two basic types: global and local. Global
396 identifiers (functions, global variables) begin with the <tt>'@'</tt>
397 character. Local identifiers (register names, types) begin with
398 the <tt>'%'</tt> character. Additionally, there are three different formats
399 for identifiers, for different purposes:</p>
402 <li>Named values are represented as a string of characters with their prefix.
403 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
404 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
405 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
406 other characters in their names can be surrounded with quotes. Special
407 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
408 ASCII code for the character in hexadecimal. In this way, any character
409 can be used in a name value, even quotes themselves.</li>
411 <li>Unnamed values are represented as an unsigned numeric value with their
412 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
414 <li>Constants, which are described in a <a href="#constants">section about
415 constants</a>, below.</li>
418 <p>LLVM requires that values start with a prefix for two reasons: Compilers
419 don't need to worry about name clashes with reserved words, and the set of
420 reserved words may be expanded in the future without penalty. Additionally,
421 unnamed identifiers allow a compiler to quickly come up with a temporary
422 variable without having to avoid symbol table conflicts.</p>
424 <p>Reserved words in LLVM are very similar to reserved words in other
425 languages. There are keywords for different opcodes
426 ('<tt><a href="#i_add">add</a></tt>',
427 '<tt><a href="#i_bitcast">bitcast</a></tt>',
428 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
429 ('<tt><a href="#t_void">void</a></tt>',
430 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
431 reserved words cannot conflict with variable names, because none of them
432 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
434 <p>Here is an example of LLVM code to multiply the integer variable
435 '<tt>%X</tt>' by 8:</p>
439 <pre class="doc_code">
440 %result = <a href="#i_mul">mul</a> i32 %X, 8
443 <p>After strength reduction:</p>
445 <pre class="doc_code">
446 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
449 <p>And the hard way:</p>
451 <pre class="doc_code">
452 %0 = <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
453 %1 = <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
454 %result = <a href="#i_add">add</a> i32 %1, %1
457 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
458 lexical features of LLVM:</p>
461 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
464 <li>Unnamed temporaries are created when the result of a computation is not
465 assigned to a named value.</li>
467 <li>Unnamed temporaries are numbered sequentially</li>
470 <p>It also shows a convention that we follow in this document. When
471 demonstrating instructions, we will follow an instruction with a comment that
472 defines the type and name of value produced. Comments are shown in italic
477 <!-- *********************************************************************** -->
478 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
479 <!-- *********************************************************************** -->
481 <!-- ======================================================================= -->
482 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
485 <div class="doc_text">
487 <p>LLVM programs are composed of "Module"s, each of which is a translation unit
488 of the input programs. Each module consists of functions, global variables,
489 and symbol table entries. Modules may be combined together with the LLVM
490 linker, which merges function (and global variable) definitions, resolves
491 forward declarations, and merges symbol table entries. Here is an example of
492 the "hello world" module:</p>
494 <pre class="doc_code">
495 <i>; Declare the string constant as a global constant.</i>
496 <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>
498 <i>; External declaration of the puts function</i>
499 <a href="#functionstructure">declare</a> i32 @puts(i8*) <i>; i32 (i8*)* </i>
501 <i>; Definition of main function</i>
502 define i32 @main() { <i>; i32()* </i>
503 <i>; Convert [13 x i8]* to i8 *...</i>
504 %cast210 = <a href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8*</i>
506 <i>; Call puts function to write out the string to stdout.</i>
507 <a href="#i_call">call</a> i32 @puts(i8* %cast210) <i>; i32</i>
508 <a href="#i_ret">ret</a> i32 0
511 <i>; Named metadata</i>
512 !1 = metadata !{i32 41}
516 <p>This example is made up of a <a href="#globalvars">global variable</a> named
517 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function,
518 a <a href="#functionstructure">function definition</a> for
519 "<tt>main</tt>" and <a href="#namedmetadatastructure">named metadata</a>
522 <p>In general, a module is made up of a list of global values, where both
523 functions and global variables are global values. Global values are
524 represented by a pointer to a memory location (in this case, a pointer to an
525 array of char, and a pointer to a function), and have one of the
526 following <a href="#linkage">linkage types</a>.</p>
530 <!-- ======================================================================= -->
531 <div class="doc_subsection">
532 <a name="linkage">Linkage Types</a>
535 <div class="doc_text">
537 <p>All Global Variables and Functions have one of the following types of
541 <dt><tt><b><a name="linkage_private">private</a></b></tt></dt>
542 <dd>Global values with "<tt>private</tt>" linkage are only directly accessible
543 by objects in the current module. In particular, linking code into a
544 module with an private global value may cause the private to be renamed as
545 necessary to avoid collisions. Because the symbol is private to the
546 module, all references can be updated. This doesn't show up in any symbol
547 table in the object file.</dd>
549 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt></dt>
550 <dd>Similar to <tt>private</tt>, but the symbol is passed through the
551 assembler and evaluated by the linker. Unlike normal strong symbols, they
552 are removed by the linker from the final linked image (executable or
553 dynamic library).</dd>
555 <dt><tt><b><a name="linkage_linker_private_weak">linker_private_weak</a></b></tt></dt>
556 <dd>Similar to "<tt>linker_private</tt>", but the symbol is weak. Note that
557 <tt>linker_private_weak</tt> symbols are subject to coalescing by the
558 linker. The symbols are removed by the linker from the final linked image
559 (executable or dynamic library).</dd>
561 <dt><tt><b><a name="linkage_linker_private_weak_def_auto">linker_private_weak_def_auto</a></b></tt></dt>
562 <dd>Similar to "<tt>linker_private_weak</tt>", but it's known that the address
563 of the object is not taken. For instance, functions that had an inline
564 definition, but the compiler decided not to inline it. Note,
565 unlike <tt>linker_private</tt> and <tt>linker_private_weak</tt>,
566 <tt>linker_private_weak_def_auto</tt> may have only <tt>default</tt>
567 visibility. The symbols are removed by the linker from the final linked
568 image (executable or dynamic library).</dd>
570 <dt><tt><b><a name="linkage_internal">internal</a></b></tt></dt>
571 <dd>Similar to private, but the value shows as a local symbol
572 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
573 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
575 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt></dt>
576 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
577 into the object file corresponding to the LLVM module. They exist to
578 allow inlining and other optimizations to take place given knowledge of
579 the definition of the global, which is known to be somewhere outside the
580 module. Globals with <tt>available_externally</tt> linkage are allowed to
581 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
582 This linkage type is only allowed on definitions, not declarations.</dd>
584 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt></dt>
585 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
586 the same name when linkage occurs. This can be used to implement
587 some forms of inline functions, templates, or other code which must be
588 generated in each translation unit that uses it, but where the body may
589 be overridden with a more definitive definition later. Unreferenced
590 <tt>linkonce</tt> globals are allowed to be discarded. Note that
591 <tt>linkonce</tt> linkage does not actually allow the optimizer to
592 inline the body of this function into callers because it doesn't know if
593 this definition of the function is the definitive definition within the
594 program or whether it will be overridden by a stronger definition.
595 To enable inlining and other optimizations, use "<tt>linkonce_odr</tt>"
598 <dt><tt><b><a name="linkage_weak">weak</a></b></tt></dt>
599 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
600 <tt>linkonce</tt> linkage, except that unreferenced globals with
601 <tt>weak</tt> linkage may not be discarded. This is used for globals that
602 are declared "weak" in C source code.</dd>
604 <dt><tt><b><a name="linkage_common">common</a></b></tt></dt>
605 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
606 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
608 Symbols with "<tt>common</tt>" linkage are merged in the same way as
609 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
610 <tt>common</tt> symbols may not have an explicit section,
611 must have a zero initializer, and may not be marked '<a
612 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
613 have common linkage.</dd>
616 <dt><tt><b><a name="linkage_appending">appending</a></b></tt></dt>
617 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
618 pointer to array type. When two global variables with appending linkage
619 are linked together, the two global arrays are appended together. This is
620 the LLVM, typesafe, equivalent of having the system linker append together
621 "sections" with identical names when .o files are linked.</dd>
623 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt></dt>
624 <dd>The semantics of this linkage follow the ELF object file model: the symbol
625 is weak until linked, if not linked, the symbol becomes null instead of
626 being an undefined reference.</dd>
628 <dt><tt><b><a name="linkage_linkonce_odr">linkonce_odr</a></b></tt></dt>
629 <dt><tt><b><a name="linkage_weak_odr">weak_odr</a></b></tt></dt>
630 <dd>Some languages allow differing globals to be merged, such as two functions
631 with different semantics. Other languages, such as <tt>C++</tt>, ensure
632 that only equivalent globals are ever merged (the "one definition rule"
633 — "ODR"). Such languages can use the <tt>linkonce_odr</tt>
634 and <tt>weak_odr</tt> linkage types to indicate that the global will only
635 be merged with equivalent globals. These linkage types are otherwise the
636 same as their non-<tt>odr</tt> versions.</dd>
638 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
639 <dd>If none of the above identifiers are used, the global is externally
640 visible, meaning that it participates in linkage and can be used to
641 resolve external symbol references.</dd>
644 <p>The next two types of linkage are targeted for Microsoft Windows platform
645 only. They are designed to support importing (exporting) symbols from (to)
646 DLLs (Dynamic Link Libraries).</p>
649 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt></dt>
650 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
651 or variable via a global pointer to a pointer that is set up by the DLL
652 exporting the symbol. On Microsoft Windows targets, the pointer name is
653 formed by combining <code>__imp_</code> and the function or variable
656 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt></dt>
657 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
658 pointer to a pointer in a DLL, so that it can be referenced with the
659 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
660 name is formed by combining <code>__imp_</code> and the function or
664 <p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
665 another module defined a "<tt>.LC0</tt>" variable and was linked with this
666 one, one of the two would be renamed, preventing a collision. Since
667 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
668 declarations), they are accessible outside of the current module.</p>
670 <p>It is illegal for a function <i>declaration</i> to have any linkage type
671 other than "externally visible", <tt>dllimport</tt>
672 or <tt>extern_weak</tt>.</p>
674 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
675 or <tt>weak_odr</tt> linkages.</p>
679 <!-- ======================================================================= -->
680 <div class="doc_subsection">
681 <a name="callingconv">Calling Conventions</a>
684 <div class="doc_text">
686 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
687 and <a href="#i_invoke">invokes</a> can all have an optional calling
688 convention specified for the call. The calling convention of any pair of
689 dynamic caller/callee must match, or the behavior of the program is
690 undefined. The following calling conventions are supported by LLVM, and more
691 may be added in the future:</p>
694 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
695 <dd>This calling convention (the default if no other calling convention is
696 specified) matches the target C calling conventions. This calling
697 convention supports varargs function calls and tolerates some mismatch in
698 the declared prototype and implemented declaration of the function (as
701 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
702 <dd>This calling convention attempts to make calls as fast as possible
703 (e.g. by passing things in registers). This calling convention allows the
704 target to use whatever tricks it wants to produce fast code for the
705 target, without having to conform to an externally specified ABI
706 (Application Binary Interface).
707 <a href="CodeGenerator.html#tailcallopt">Tail calls can only be optimized
708 when this or the GHC convention is used.</a> This calling convention
709 does not support varargs and requires the prototype of all callees to
710 exactly match the prototype of the function definition.</dd>
712 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
713 <dd>This calling convention attempts to make code in the caller as efficient
714 as possible under the assumption that the call is not commonly executed.
715 As such, these calls often preserve all registers so that the call does
716 not break any live ranges in the caller side. This calling convention
717 does not support varargs and requires the prototype of all callees to
718 exactly match the prototype of the function definition.</dd>
720 <dt><b>"<tt>cc <em>10</em></tt>" - GHC convention</b>:</dt>
721 <dd>This calling convention has been implemented specifically for use by the
722 <a href="http://www.haskell.org/ghc">Glasgow Haskell Compiler (GHC)</a>.
723 It passes everything in registers, going to extremes to achieve this by
724 disabling callee save registers. This calling convention should not be
725 used lightly but only for specific situations such as an alternative to
726 the <em>register pinning</em> performance technique often used when
727 implementing functional programming languages.At the moment only X86
728 supports this convention and it has the following limitations:
730 <li>On <em>X86-32</em> only supports up to 4 bit type parameters. No
731 floating point types are supported.</li>
732 <li>On <em>X86-64</em> only supports up to 10 bit type parameters and
733 6 floating point parameters.</li>
735 This calling convention supports
736 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> but
737 requires both the caller and callee are using it.
740 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
741 <dd>Any calling convention may be specified by number, allowing
742 target-specific calling conventions to be used. Target specific calling
743 conventions start at 64.</dd>
746 <p>More calling conventions can be added/defined on an as-needed basis, to
747 support Pascal conventions or any other well-known target-independent
752 <!-- ======================================================================= -->
753 <div class="doc_subsection">
754 <a name="visibility">Visibility Styles</a>
757 <div class="doc_text">
759 <p>All Global Variables and Functions have one of the following visibility
763 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
764 <dd>On targets that use the ELF object file format, default visibility means
765 that the declaration is visible to other modules and, in shared libraries,
766 means that the declared entity may be overridden. On Darwin, default
767 visibility means that the declaration is visible to other modules. Default
768 visibility corresponds to "external linkage" in the language.</dd>
770 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
771 <dd>Two declarations of an object with hidden visibility refer to the same
772 object if they are in the same shared object. Usually, hidden visibility
773 indicates that the symbol will not be placed into the dynamic symbol
774 table, so no other module (executable or shared library) can reference it
777 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
778 <dd>On ELF, protected visibility indicates that the symbol will be placed in
779 the dynamic symbol table, but that references within the defining module
780 will bind to the local symbol. That is, the symbol cannot be overridden by
786 <!-- ======================================================================= -->
787 <div class="doc_subsection">
788 <a name="namedtypes">Named Types</a>
791 <div class="doc_text">
793 <p>LLVM IR allows you to specify name aliases for certain types. This can make
794 it easier to read the IR and make the IR more condensed (particularly when
795 recursive types are involved). An example of a name specification is:</p>
797 <pre class="doc_code">
798 %mytype = type { %mytype*, i32 }
801 <p>You may give a name to any <a href="#typesystem">type</a> except
802 "<a href="#t_void">void</a>". Type name aliases may be used anywhere a type
803 is expected with the syntax "%mytype".</p>
805 <p>Note that type names are aliases for the structural type that they indicate,
806 and that you can therefore specify multiple names for the same type. This
807 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
808 uses structural typing, the name is not part of the type. When printing out
809 LLVM IR, the printer will pick <em>one name</em> to render all types of a
810 particular shape. This means that if you have code where two different
811 source types end up having the same LLVM type, that the dumper will sometimes
812 print the "wrong" or unexpected type. This is an important design point and
813 isn't going to change.</p>
817 <!-- ======================================================================= -->
818 <div class="doc_subsection">
819 <a name="globalvars">Global Variables</a>
822 <div class="doc_text">
824 <p>Global variables define regions of memory allocated at compilation time
825 instead of run-time. Global variables may optionally be initialized, may
826 have an explicit section to be placed in, and may have an optional explicit
827 alignment specified. A variable may be defined as "thread_local", which
828 means that it will not be shared by threads (each thread will have a
829 separated copy of the variable). A variable may be defined as a global
830 "constant," which indicates that the contents of the variable
831 will <b>never</b> be modified (enabling better optimization, allowing the
832 global data to be placed in the read-only section of an executable, etc).
833 Note that variables that need runtime initialization cannot be marked
834 "constant" as there is a store to the variable.</p>
836 <p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
837 constant, even if the final definition of the global is not. This capability
838 can be used to enable slightly better optimization of the program, but
839 requires the language definition to guarantee that optimizations based on the
840 'constantness' are valid for the translation units that do not include the
843 <p>As SSA values, global variables define pointer values that are in scope
844 (i.e. they dominate) all basic blocks in the program. Global variables
845 always define a pointer to their "content" type because they describe a
846 region of memory, and all memory objects in LLVM are accessed through
849 <p>A global variable may be declared to reside in a target-specific numbered
850 address space. For targets that support them, address spaces may affect how
851 optimizations are performed and/or what target instructions are used to
852 access the variable. The default address space is zero. The address space
853 qualifier must precede any other attributes.</p>
855 <p>LLVM allows an explicit section to be specified for globals. If the target
856 supports it, it will emit globals to the section specified.</p>
858 <p>An explicit alignment may be specified for a global, which must be a power
859 of 2. If not present, or if the alignment is set to zero, the alignment of
860 the global is set by the target to whatever it feels convenient. If an
861 explicit alignment is specified, the global is forced to have exactly that
862 alignment. Targets and optimizers are not allowed to over-align the global
863 if the global has an assigned section. In this case, the extra alignment
864 could be observable: for example, code could assume that the globals are
865 densely packed in their section and try to iterate over them as an array,
866 alignment padding would break this iteration.</p>
868 <p>For example, the following defines a global in a numbered address space with
869 an initializer, section, and alignment:</p>
871 <pre class="doc_code">
872 @G = addrspace(5) constant float 1.0, section "foo", align 4
878 <!-- ======================================================================= -->
879 <div class="doc_subsection">
880 <a name="functionstructure">Functions</a>
883 <div class="doc_text">
885 <p>LLVM function definitions consist of the "<tt>define</tt>" keyword, an
886 optional <a href="#linkage">linkage type</a>, an optional
887 <a href="#visibility">visibility style</a>, an optional
888 <a href="#callingconv">calling convention</a>, a return type, an optional
889 <a href="#paramattrs">parameter attribute</a> for the return type, a function
890 name, a (possibly empty) argument list (each with optional
891 <a href="#paramattrs">parameter attributes</a>), optional
892 <a href="#fnattrs">function attributes</a>, an optional section, an optional
893 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
894 curly brace, a list of basic blocks, and a closing curly brace.</p>
896 <p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
897 optional <a href="#linkage">linkage type</a>, an optional
898 <a href="#visibility">visibility style</a>, an optional
899 <a href="#callingconv">calling convention</a>, a return type, an optional
900 <a href="#paramattrs">parameter attribute</a> for the return type, a function
901 name, a possibly empty list of arguments, an optional alignment, and an
902 optional <a href="#gc">garbage collector name</a>.</p>
904 <p>A function definition contains a list of basic blocks, forming the CFG
905 (Control Flow Graph) for the function. Each basic block may optionally start
906 with a label (giving the basic block a symbol table entry), contains a list
907 of instructions, and ends with a <a href="#terminators">terminator</a>
908 instruction (such as a branch or function return).</p>
910 <p>The first basic block in a function is special in two ways: it is immediately
911 executed on entrance to the function, and it is not allowed to have
912 predecessor basic blocks (i.e. there can not be any branches to the entry
913 block of a function). Because the block can have no predecessors, it also
914 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
916 <p>LLVM allows an explicit section to be specified for functions. If the target
917 supports it, it will emit functions to the section specified.</p>
919 <p>An explicit alignment may be specified for a function. If not present, or if
920 the alignment is set to zero, the alignment of the function is set by the
921 target to whatever it feels convenient. If an explicit alignment is
922 specified, the function is forced to have at least that much alignment. All
923 alignments must be a power of 2.</p>
926 <pre class="doc_code">
927 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
928 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
929 <ResultType> @<FunctionName> ([argument list])
930 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
931 [<a href="#gc">gc</a>] { ... }
936 <!-- ======================================================================= -->
937 <div class="doc_subsection">
938 <a name="aliasstructure">Aliases</a>
941 <div class="doc_text">
943 <p>Aliases act as "second name" for the aliasee value (which can be either
944 function, global variable, another alias or bitcast of global value). Aliases
945 may have an optional <a href="#linkage">linkage type</a>, and an
946 optional <a href="#visibility">visibility style</a>.</p>
949 <pre class="doc_code">
950 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
955 <!-- ======================================================================= -->
956 <div class="doc_subsection">
957 <a name="namedmetadatastructure">Named Metadata</a>
960 <div class="doc_text">
962 <p>Named metadata is a collection of metadata. <a href="#metadata">Metadata
963 nodes</a> (but not metadata strings) are the only valid operands for
964 a named metadata.</p>
967 <pre class="doc_code">
968 ; Some unnamed metadata nodes, which are referenced by the named metadata.
969 !0 = metadata !{metadata !"zero"}
970 !1 = metadata !{metadata !"one"}
971 !2 = metadata !{metadata !"two"}
973 !name = !{!0, !1, !2}
978 <!-- ======================================================================= -->
979 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
981 <div class="doc_text">
983 <p>The return type and each parameter of a function type may have a set of
984 <i>parameter attributes</i> associated with them. Parameter attributes are
985 used to communicate additional information about the result or parameters of
986 a function. Parameter attributes are considered to be part of the function,
987 not of the function type, so functions with different parameter attributes
988 can have the same function type.</p>
990 <p>Parameter attributes are simple keywords that follow the type specified. If
991 multiple parameter attributes are needed, they are space separated. For
994 <pre class="doc_code">
995 declare i32 @printf(i8* noalias nocapture, ...)
996 declare i32 @atoi(i8 zeroext)
997 declare signext i8 @returns_signed_char()
1000 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
1001 <tt>readonly</tt>) come immediately after the argument list.</p>
1003 <p>Currently, only the following parameter attributes are defined:</p>
1006 <dt><tt><b>zeroext</b></tt></dt>
1007 <dd>This indicates to the code generator that the parameter or return value
1008 should be zero-extended to a 32-bit value by the caller (for a parameter)
1009 or the callee (for a return value).</dd>
1011 <dt><tt><b>signext</b></tt></dt>
1012 <dd>This indicates to the code generator that the parameter or return value
1013 should be sign-extended to a 32-bit value by the caller (for a parameter)
1014 or the callee (for a return value).</dd>
1016 <dt><tt><b>inreg</b></tt></dt>
1017 <dd>This indicates that this parameter or return value should be treated in a
1018 special target-dependent fashion during while emitting code for a function
1019 call or return (usually, by putting it in a register as opposed to memory,
1020 though some targets use it to distinguish between two different kinds of
1021 registers). Use of this attribute is target-specific.</dd>
1023 <dt><tt><b><a name="byval">byval</a></b></tt></dt>
1024 <dd><p>This indicates that the pointer parameter should really be passed by
1025 value to the function. The attribute implies that a hidden copy of the
1027 is made between the caller and the callee, so the callee is unable to
1028 modify the value in the callee. This attribute is only valid on LLVM
1029 pointer arguments. It is generally used to pass structs and arrays by
1030 value, but is also valid on pointers to scalars. The copy is considered
1031 to belong to the caller not the callee (for example,
1032 <tt><a href="#readonly">readonly</a></tt> functions should not write to
1033 <tt>byval</tt> parameters). This is not a valid attribute for return
1036 <p>The byval attribute also supports specifying an alignment with
1037 the align attribute. It indicates the alignment of the stack slot to
1038 form and the known alignment of the pointer specified to the call site. If
1039 the alignment is not specified, then the code generator makes a
1040 target-specific assumption.</p></dd>
1042 <dt><tt><b><a name="sret">sret</a></b></tt></dt>
1043 <dd>This indicates that the pointer parameter specifies the address of a
1044 structure that is the return value of the function in the source program.
1045 This pointer must be guaranteed by the caller to be valid: loads and
1046 stores to the structure may be assumed by the callee to not to trap. This
1047 may only be applied to the first parameter. This is not a valid attribute
1048 for return values. </dd>
1050 <dt><tt><b><a name="noalias">noalias</a></b></tt></dt>
1051 <dd>This indicates that pointer values
1052 <a href="#pointeraliasing"><i>based</i></a> on the argument or return
1053 value do not alias pointer values which are not <i>based</i> on it,
1054 ignoring certain "irrelevant" dependencies.
1055 For a call to the parent function, dependencies between memory
1056 references from before or after the call and from those during the call
1057 are "irrelevant" to the <tt>noalias</tt> keyword for the arguments and
1058 return value used in that call.
1059 The caller shares the responsibility with the callee for ensuring that
1060 these requirements are met.
1061 For further details, please see the discussion of the NoAlias response in
1062 <a href="AliasAnalysis.html#MustMayNo">alias analysis</a>.<br>
1064 Note that this definition of <tt>noalias</tt> is intentionally
1065 similar to the definition of <tt>restrict</tt> in C99 for function
1066 arguments, though it is slightly weaker.
1068 For function return values, C99's <tt>restrict</tt> is not meaningful,
1069 while LLVM's <tt>noalias</tt> is.
1072 <dt><tt><b><a name="nocapture">nocapture</a></b></tt></dt>
1073 <dd>This indicates that the callee does not make any copies of the pointer
1074 that outlive the callee itself. This is not a valid attribute for return
1077 <dt><tt><b><a name="nest">nest</a></b></tt></dt>
1078 <dd>This indicates that the pointer parameter can be excised using the
1079 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
1080 attribute for return values.</dd>
1085 <!-- ======================================================================= -->
1086 <div class="doc_subsection">
1087 <a name="gc">Garbage Collector Names</a>
1090 <div class="doc_text">
1092 <p>Each function may specify a garbage collector name, which is simply a
1095 <pre class="doc_code">
1096 define void @f() gc "name" { ... }
1099 <p>The compiler declares the supported values of <i>name</i>. Specifying a
1100 collector which will cause the compiler to alter its output in order to
1101 support the named garbage collection algorithm.</p>
1105 <!-- ======================================================================= -->
1106 <div class="doc_subsection">
1107 <a name="fnattrs">Function Attributes</a>
1110 <div class="doc_text">
1112 <p>Function attributes are set to communicate additional information about a
1113 function. Function attributes are considered to be part of the function, not
1114 of the function type, so functions with different parameter attributes can
1115 have the same function type.</p>
1117 <p>Function attributes are simple keywords that follow the type specified. If
1118 multiple attributes are needed, they are space separated. For example:</p>
1120 <pre class="doc_code">
1121 define void @f() noinline { ... }
1122 define void @f() alwaysinline { ... }
1123 define void @f() alwaysinline optsize { ... }
1124 define void @f() optsize { ... }
1128 <dt><tt><b>alignstack(<<em>n</em>>)</b></tt></dt>
1129 <dd>This attribute indicates that, when emitting the prologue and epilogue,
1130 the backend should forcibly align the stack pointer. Specify the
1131 desired alignment, which must be a power of two, in parentheses.
1133 <dt><tt><b>alwaysinline</b></tt></dt>
1134 <dd>This attribute indicates that the inliner should attempt to inline this
1135 function into callers whenever possible, ignoring any active inlining size
1136 threshold for this caller.</dd>
1138 <dt><tt><b>hotpatch</b></tt></dt>
1139 <dd>This attribute indicates that the function should be 'hotpatchable',
1140 meaning the function can be patched and/or hooked even while it is
1141 loaded into memory. On x86, the function prologue will be preceded
1142 by six bytes of padding and will begin with a two-byte instruction.
1143 Most of the functions in the Windows system DLLs in Windows XP SP2 or
1144 higher were compiled in this fashion.</dd>
1146 <dt><tt><b>inlinehint</b></tt></dt>
1147 <dd>This attribute indicates that the source code contained a hint that inlining
1148 this function is desirable (such as the "inline" keyword in C/C++). It
1149 is just a hint; it imposes no requirements on the inliner.</dd>
1151 <dt><tt><b>naked</b></tt></dt>
1152 <dd>This attribute disables prologue / epilogue emission for the function.
1153 This can have very system-specific consequences.</dd>
1155 <dt><tt><b>noimplicitfloat</b></tt></dt>
1156 <dd>This attributes disables implicit floating point instructions.</dd>
1158 <dt><tt><b>noinline</b></tt></dt>
1159 <dd>This attribute indicates that the inliner should never inline this
1160 function in any situation. This attribute may not be used together with
1161 the <tt>alwaysinline</tt> attribute.</dd>
1163 <dt><tt><b>noredzone</b></tt></dt>
1164 <dd>This attribute indicates that the code generator should not use a red
1165 zone, even if the target-specific ABI normally permits it.</dd>
1167 <dt><tt><b>noreturn</b></tt></dt>
1168 <dd>This function attribute indicates that the function never returns
1169 normally. This produces undefined behavior at runtime if the function
1170 ever does dynamically return.</dd>
1172 <dt><tt><b>nounwind</b></tt></dt>
1173 <dd>This function attribute indicates that the function never returns with an
1174 unwind or exceptional control flow. If the function does unwind, its
1175 runtime behavior is undefined.</dd>
1177 <dt><tt><b>optsize</b></tt></dt>
1178 <dd>This attribute suggests that optimization passes and code generator passes
1179 make choices that keep the code size of this function low, and otherwise
1180 do optimizations specifically to reduce code size.</dd>
1182 <dt><tt><b>readnone</b></tt></dt>
1183 <dd>This attribute indicates that the function computes its result (or decides
1184 to unwind an exception) based strictly on its arguments, without
1185 dereferencing any pointer arguments or otherwise accessing any mutable
1186 state (e.g. memory, control registers, etc) visible to caller functions.
1187 It does not write through any pointer arguments
1188 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1189 changes any state visible to callers. This means that it cannot unwind
1190 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1191 could use the <tt>unwind</tt> instruction.</dd>
1193 <dt><tt><b><a name="readonly">readonly</a></b></tt></dt>
1194 <dd>This attribute indicates that the function does not write through any
1195 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1196 arguments) or otherwise modify any state (e.g. memory, control registers,
1197 etc) visible to caller functions. It may dereference pointer arguments
1198 and read state that may be set in the caller. A readonly function always
1199 returns the same value (or unwinds an exception identically) when called
1200 with the same set of arguments and global state. It cannot unwind an
1201 exception by calling the <tt>C++</tt> exception throwing methods, but may
1202 use the <tt>unwind</tt> instruction.</dd>
1204 <dt><tt><b><a name="ssp">ssp</a></b></tt></dt>
1205 <dd>This attribute indicates that the function should emit a stack smashing
1206 protector. It is in the form of a "canary"—a random value placed on
1207 the stack before the local variables that's checked upon return from the
1208 function to see if it has been overwritten. A heuristic is used to
1209 determine if a function needs stack protectors or not.<br>
1211 If a function that has an <tt>ssp</tt> attribute is inlined into a
1212 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1213 function will have an <tt>ssp</tt> attribute.</dd>
1215 <dt><tt><b>sspreq</b></tt></dt>
1216 <dd>This attribute indicates that the function should <em>always</em> emit a
1217 stack smashing protector. This overrides
1218 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1220 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1221 function that doesn't have an <tt>sspreq</tt> attribute or which has
1222 an <tt>ssp</tt> attribute, then the resulting function will have
1223 an <tt>sspreq</tt> attribute.</dd>
1228 <!-- ======================================================================= -->
1229 <div class="doc_subsection">
1230 <a name="moduleasm">Module-Level Inline Assembly</a>
1233 <div class="doc_text">
1235 <p>Modules may contain "module-level inline asm" blocks, which corresponds to
1236 the GCC "file scope inline asm" blocks. These blocks are internally
1237 concatenated by LLVM and treated as a single unit, but may be separated in
1238 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
1240 <pre class="doc_code">
1241 module asm "inline asm code goes here"
1242 module asm "more can go here"
1245 <p>The strings can contain any character by escaping non-printable characters.
1246 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1249 <p>The inline asm code is simply printed to the machine code .s file when
1250 assembly code is generated.</p>
1254 <!-- ======================================================================= -->
1255 <div class="doc_subsection">
1256 <a name="datalayout">Data Layout</a>
1259 <div class="doc_text">
1261 <p>A module may specify a target specific data layout string that specifies how
1262 data is to be laid out in memory. The syntax for the data layout is
1265 <pre class="doc_code">
1266 target datalayout = "<i>layout specification</i>"
1269 <p>The <i>layout specification</i> consists of a list of specifications
1270 separated by the minus sign character ('-'). Each specification starts with
1271 a letter and may include other information after the letter to define some
1272 aspect of the data layout. The specifications accepted are as follows:</p>
1276 <dd>Specifies that the target lays out data in big-endian form. That is, the
1277 bits with the most significance have the lowest address location.</dd>
1280 <dd>Specifies that the target lays out data in little-endian form. That is,
1281 the bits with the least significance have the lowest address
1284 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1285 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1286 <i>preferred</i> alignments. All sizes are in bits. Specifying
1287 the <i>pref</i> alignment is optional. If omitted, the
1288 preceding <tt>:</tt> should be omitted too.</dd>
1290 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1291 <dd>This specifies the alignment for an integer type of a given bit
1292 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1294 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1295 <dd>This specifies the alignment for a vector type of a given bit
1298 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1299 <dd>This specifies the alignment for a floating point type of a given bit
1300 <i>size</i>. Only values of <i>size</i> that are supported by the target
1301 will work. 32 (float) and 64 (double) are supported on all targets;
1302 80 or 128 (different flavors of long double) are also supported on some
1305 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1306 <dd>This specifies the alignment for an aggregate type of a given bit
1309 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1310 <dd>This specifies the alignment for a stack object of a given bit
1313 <dt><tt>n<i>size1</i>:<i>size2</i>:<i>size3</i>...</tt></dt>
1314 <dd>This specifies a set of native integer widths for the target CPU
1315 in bits. For example, it might contain "n32" for 32-bit PowerPC,
1316 "n32:64" for PowerPC 64, or "n8:16:32:64" for X86-64. Elements of
1317 this set are considered to support most general arithmetic
1318 operations efficiently.</dd>
1321 <p>When constructing the data layout for a given target, LLVM starts with a
1322 default set of specifications which are then (possibly) overridden by the
1323 specifications in the <tt>datalayout</tt> keyword. The default specifications
1324 are given in this list:</p>
1327 <li><tt>E</tt> - big endian</li>
1328 <li><tt>p:64:64:64</tt> - 64-bit pointers with 64-bit alignment</li>
1329 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1330 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1331 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1332 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1333 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1334 alignment of 64-bits</li>
1335 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1336 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1337 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1338 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1339 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1340 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
1343 <p>When LLVM is determining the alignment for a given type, it uses the
1344 following rules:</p>
1347 <li>If the type sought is an exact match for one of the specifications, that
1348 specification is used.</li>
1350 <li>If no match is found, and the type sought is an integer type, then the
1351 smallest integer type that is larger than the bitwidth of the sought type
1352 is used. If none of the specifications are larger than the bitwidth then
1353 the the largest integer type is used. For example, given the default
1354 specifications above, the i7 type will use the alignment of i8 (next
1355 largest) while both i65 and i256 will use the alignment of i64 (largest
1358 <li>If no match is found, and the type sought is a vector type, then the
1359 largest vector type that is smaller than the sought vector type will be
1360 used as a fall back. This happens because <128 x double> can be
1361 implemented in terms of 64 <2 x double>, for example.</li>
1366 <!-- ======================================================================= -->
1367 <div class="doc_subsection">
1368 <a name="pointeraliasing">Pointer Aliasing Rules</a>
1371 <div class="doc_text">
1373 <p>Any memory access must be done through a pointer value associated
1374 with an address range of the memory access, otherwise the behavior
1375 is undefined. Pointer values are associated with address ranges
1376 according to the following rules:</p>
1379 <li>A pointer value is associated with the addresses associated with
1380 any value it is <i>based</i> on.
1381 <li>An address of a global variable is associated with the address
1382 range of the variable's storage.</li>
1383 <li>The result value of an allocation instruction is associated with
1384 the address range of the allocated storage.</li>
1385 <li>A null pointer in the default address-space is associated with
1387 <li>An integer constant other than zero or a pointer value returned
1388 from a function not defined within LLVM may be associated with address
1389 ranges allocated through mechanisms other than those provided by
1390 LLVM. Such ranges shall not overlap with any ranges of addresses
1391 allocated by mechanisms provided by LLVM.</li>
1394 <p>A pointer value is <i>based</i> on another pointer value according
1395 to the following rules:</p>
1398 <li>A pointer value formed from a
1399 <tt><a href="#i_getelementptr">getelementptr</a></tt> operation
1400 is <i>based</i> on the first operand of the <tt>getelementptr</tt>.</li>
1401 <li>The result value of a
1402 <tt><a href="#i_bitcast">bitcast</a></tt> is <i>based</i> on the operand
1403 of the <tt>bitcast</tt>.</li>
1404 <li>A pointer value formed by an
1405 <tt><a href="#i_inttoptr">inttoptr</a></tt> is <i>based</i> on all
1406 pointer values that contribute (directly or indirectly) to the
1407 computation of the pointer's value.</li>
1408 <li>The "<i>based</i> on" relationship is transitive.</li>
1411 <p>Note that this definition of <i>"based"</i> is intentionally
1412 similar to the definition of <i>"based"</i> in C99, though it is
1413 slightly weaker.</p>
1415 <p>LLVM IR does not associate types with memory. The result type of a
1416 <tt><a href="#i_load">load</a></tt> merely indicates the size and
1417 alignment of the memory from which to load, as well as the
1418 interpretation of the value. The first operand type of a
1419 <tt><a href="#i_store">store</a></tt> similarly only indicates the size
1420 and alignment of the store.</p>
1422 <p>Consequently, type-based alias analysis, aka TBAA, aka
1423 <tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1424 LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1425 additional information which specialized optimization passes may use
1426 to implement type-based alias analysis.</p>
1430 <!-- ======================================================================= -->
1431 <div class="doc_subsection">
1432 <a name="volatile">Volatile Memory Accesses</a>
1435 <div class="doc_text">
1437 <p>Certain memory accesses, such as <a href="#i_load"><tt>load</tt></a>s, <a
1438 href="#i_store"><tt>store</tt></a>s, and <a
1439 href="#int_memcpy"><tt>llvm.memcpy</tt></a>s may be marked <tt>volatile</tt>.
1440 The optimizers must not change the number of volatile operations or change their
1441 order of execution relative to other volatile operations. The optimizers
1442 <i>may</i> change the order of volatile operations relative to non-volatile
1443 operations. This is not Java's "volatile" and has no cross-thread
1444 synchronization behavior.</p>
1448 <!-- *********************************************************************** -->
1449 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1450 <!-- *********************************************************************** -->
1452 <div class="doc_text">
1454 <p>The LLVM type system is one of the most important features of the
1455 intermediate representation. Being typed enables a number of optimizations
1456 to be performed on the intermediate representation directly, without having
1457 to do extra analyses on the side before the transformation. A strong type
1458 system makes it easier to read the generated code and enables novel analyses
1459 and transformations that are not feasible to perform on normal three address
1460 code representations.</p>
1464 <!-- ======================================================================= -->
1465 <div class="doc_subsection"> <a name="t_classifications">Type
1466 Classifications</a> </div>
1468 <div class="doc_text">
1470 <p>The types fall into a few useful classifications:</p>
1472 <table border="1" cellspacing="0" cellpadding="4">
1474 <tr><th>Classification</th><th>Types</th></tr>
1476 <td><a href="#t_integer">integer</a></td>
1477 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1480 <td><a href="#t_floating">floating point</a></td>
1481 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1484 <td><a name="t_firstclass">first class</a></td>
1485 <td><a href="#t_integer">integer</a>,
1486 <a href="#t_floating">floating point</a>,
1487 <a href="#t_pointer">pointer</a>,
1488 <a href="#t_vector">vector</a>,
1489 <a href="#t_struct">structure</a>,
1490 <a href="#t_array">array</a>,
1491 <a href="#t_label">label</a>,
1492 <a href="#t_metadata">metadata</a>.
1496 <td><a href="#t_primitive">primitive</a></td>
1497 <td><a href="#t_label">label</a>,
1498 <a href="#t_void">void</a>,
1499 <a href="#t_floating">floating point</a>,
1500 <a href="#t_x86mmx">x86mmx</a>,
1501 <a href="#t_metadata">metadata</a>.</td>
1504 <td><a href="#t_derived">derived</a></td>
1505 <td><a href="#t_array">array</a>,
1506 <a href="#t_function">function</a>,
1507 <a href="#t_pointer">pointer</a>,
1508 <a href="#t_struct">structure</a>,
1509 <a href="#t_pstruct">packed structure</a>,
1510 <a href="#t_vector">vector</a>,
1511 <a href="#t_opaque">opaque</a>.
1517 <p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1518 important. Values of these types are the only ones which can be produced by
1523 <!-- ======================================================================= -->
1524 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1526 <div class="doc_text">
1528 <p>The primitive types are the fundamental building blocks of the LLVM
1533 <!-- _______________________________________________________________________ -->
1534 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1536 <div class="doc_text">
1539 <p>The integer type is a very simple type that simply specifies an arbitrary
1540 bit width for the integer type desired. Any bit width from 1 bit to
1541 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
1548 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1552 <table class="layout">
1554 <td class="left"><tt>i1</tt></td>
1555 <td class="left">a single-bit integer.</td>
1558 <td class="left"><tt>i32</tt></td>
1559 <td class="left">a 32-bit integer.</td>
1562 <td class="left"><tt>i1942652</tt></td>
1563 <td class="left">a really big integer of over 1 million bits.</td>
1569 <!-- _______________________________________________________________________ -->
1570 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1572 <div class="doc_text">
1576 <tr><th>Type</th><th>Description</th></tr>
1577 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1578 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1579 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1580 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1581 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1587 <!-- _______________________________________________________________________ -->
1588 <div class="doc_subsubsection"> <a name="t_x86mmx">X86mmx Type</a> </div>
1590 <div class="doc_text">
1593 <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>
1602 <!-- _______________________________________________________________________ -->
1603 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1605 <div class="doc_text">
1608 <p>The void type does not represent any value and has no size.</p>
1617 <!-- _______________________________________________________________________ -->
1618 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1620 <div class="doc_text">
1623 <p>The label type represents code labels.</p>
1632 <!-- _______________________________________________________________________ -->
1633 <div class="doc_subsubsection"> <a name="t_metadata">Metadata Type</a> </div>
1635 <div class="doc_text">
1638 <p>The metadata type represents embedded metadata. No derived types may be
1639 created from metadata except for <a href="#t_function">function</a>
1650 <!-- ======================================================================= -->
1651 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1653 <div class="doc_text">
1655 <p>The real power in LLVM comes from the derived types in the system. This is
1656 what allows a programmer to represent arrays, functions, pointers, and other
1657 useful types. Each of these types contain one or more element types which
1658 may be a primitive type, or another derived type. For example, it is
1659 possible to have a two dimensional array, using an array as the element type
1660 of another array.</p>
1665 <!-- _______________________________________________________________________ -->
1666 <div class="doc_subsubsection"> <a name="t_aggregate">Aggregate Types</a> </div>
1668 <div class="doc_text">
1670 <p>Aggregate Types are a subset of derived types that can contain multiple
1671 member types. <a href="#t_array">Arrays</a>,
1672 <a href="#t_struct">structs</a>, and <a href="#t_vector">vectors</a> are
1673 aggregate types.</p>
1677 <!-- _______________________________________________________________________ -->
1678 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1680 <div class="doc_text">
1683 <p>The array type is a very simple derived type that arranges elements
1684 sequentially in memory. The array type requires a size (number of elements)
1685 and an underlying data type.</p>
1689 [<# elements> x <elementtype>]
1692 <p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1693 be any type with a size.</p>
1696 <table class="layout">
1698 <td class="left"><tt>[40 x i32]</tt></td>
1699 <td class="left">Array of 40 32-bit integer values.</td>
1702 <td class="left"><tt>[41 x i32]</tt></td>
1703 <td class="left">Array of 41 32-bit integer values.</td>
1706 <td class="left"><tt>[4 x i8]</tt></td>
1707 <td class="left">Array of 4 8-bit integer values.</td>
1710 <p>Here are some examples of multidimensional arrays:</p>
1711 <table class="layout">
1713 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1714 <td class="left">3x4 array of 32-bit integer values.</td>
1717 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1718 <td class="left">12x10 array of single precision floating point values.</td>
1721 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1722 <td class="left">2x3x4 array of 16-bit integer values.</td>
1726 <p>There is no restriction on indexing beyond the end of the array implied by
1727 a static type (though there are restrictions on indexing beyond the bounds
1728 of an allocated object in some cases). This means that single-dimension
1729 'variable sized array' addressing can be implemented in LLVM with a zero
1730 length array type. An implementation of 'pascal style arrays' in LLVM could
1731 use the type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
1735 <!-- _______________________________________________________________________ -->
1736 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1738 <div class="doc_text">
1741 <p>The function type can be thought of as a function signature. It consists of
1742 a return type and a list of formal parameter types. The return type of a
1743 function type is a first class type or a void type.</p>
1747 <returntype> (<parameter list>)
1750 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1751 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1752 which indicates that the function takes a variable number of arguments.
1753 Variable argument functions can access their arguments with
1754 the <a href="#int_varargs">variable argument handling intrinsic</a>
1755 functions. '<tt><returntype></tt>' is any type except
1756 <a href="#t_label">label</a>.</p>
1759 <table class="layout">
1761 <td class="left"><tt>i32 (i32)</tt></td>
1762 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1764 </tr><tr class="layout">
1765 <td class="left"><tt>float (i16, i32 *) *
1767 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1768 an <tt>i16</tt> and a <a href="#t_pointer">pointer</a> to <tt>i32</tt>,
1769 returning <tt>float</tt>.
1771 </tr><tr class="layout">
1772 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1773 <td class="left">A vararg function that takes at least one
1774 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1775 which returns an integer. This is the signature for <tt>printf</tt> in
1778 </tr><tr class="layout">
1779 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1780 <td class="left">A function taking an <tt>i32</tt>, returning a
1781 <a href="#t_struct">structure</a> containing two <tt>i32</tt> values
1788 <!-- _______________________________________________________________________ -->
1789 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1791 <div class="doc_text">
1794 <p>The structure type is used to represent a collection of data members together
1795 in memory. The packing of the field types is defined to match the ABI of the
1796 underlying processor. The elements of a structure may be any type that has a
1799 <p>Structures in memory are accessed using '<tt><a href="#i_load">load</a></tt>'
1800 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field
1801 with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
1802 Structures in registers are accessed using the
1803 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' and
1804 '<tt><a href="#i_insertvalue">insertvalue</a></tt>' instructions.</p>
1807 { <type list> }
1811 <table class="layout">
1813 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1814 <td class="left">A triple of three <tt>i32</tt> values</td>
1815 </tr><tr class="layout">
1816 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1817 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1818 second element is a <a href="#t_pointer">pointer</a> to a
1819 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1820 an <tt>i32</tt>.</td>
1826 <!-- _______________________________________________________________________ -->
1827 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1830 <div class="doc_text">
1833 <p>The packed structure type is used to represent a collection of data members
1834 together in memory. There is no padding between fields. Further, the
1835 alignment of a packed structure is 1 byte. The elements of a packed
1836 structure may be any type that has a size.</p>
1838 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1839 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1840 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1844 < { <type list> } >
1848 <table class="layout">
1850 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1851 <td class="left">A triple of three <tt>i32</tt> values</td>
1852 </tr><tr class="layout">
1854 <tt>< { float, i32 (i32)* } ></tt></td>
1855 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1856 second element is a <a href="#t_pointer">pointer</a> to a
1857 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1858 an <tt>i32</tt>.</td>
1864 <!-- _______________________________________________________________________ -->
1865 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1867 <div class="doc_text">
1870 <p>The pointer type is used to specify memory locations.
1871 Pointers are commonly used to reference objects in memory.</p>
1873 <p>Pointer types may have an optional address space attribute defining the
1874 numbered address space where the pointed-to object resides. The default
1875 address space is number zero. The semantics of non-zero address
1876 spaces are target-specific.</p>
1878 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
1879 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
1887 <table class="layout">
1889 <td class="left"><tt>[4 x i32]*</tt></td>
1890 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1891 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1894 <td class="left"><tt>i32 (i32*) *</tt></td>
1895 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1896 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1900 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1901 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1902 that resides in address space #5.</td>
1908 <!-- _______________________________________________________________________ -->
1909 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1911 <div class="doc_text">
1914 <p>A vector type is a simple derived type that represents a vector of elements.
1915 Vector types are used when multiple primitive data are operated in parallel
1916 using a single instruction (SIMD). A vector type requires a size (number of
1917 elements) and an underlying primitive data type. Vector types are considered
1918 <a href="#t_firstclass">first class</a>.</p>
1922 < <# elements> x <elementtype> >
1925 <p>The number of elements is a constant integer value larger than 0; elementtype
1926 may be any integer or floating point type. Vectors of size zero are not
1927 allowed, and pointers are not allowed as the element type.</p>
1930 <table class="layout">
1932 <td class="left"><tt><4 x i32></tt></td>
1933 <td class="left">Vector of 4 32-bit integer values.</td>
1936 <td class="left"><tt><8 x float></tt></td>
1937 <td class="left">Vector of 8 32-bit floating-point values.</td>
1940 <td class="left"><tt><2 x i64></tt></td>
1941 <td class="left">Vector of 2 64-bit integer values.</td>
1947 <!-- _______________________________________________________________________ -->
1948 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1949 <div class="doc_text">
1952 <p>Opaque types are used to represent unknown types in the system. This
1953 corresponds (for example) to the C notion of a forward declared structure
1954 type. In LLVM, opaque types can eventually be resolved to any type (not just
1955 a structure type).</p>
1963 <table class="layout">
1965 <td class="left"><tt>opaque</tt></td>
1966 <td class="left">An opaque type.</td>
1972 <!-- ======================================================================= -->
1973 <div class="doc_subsection">
1974 <a name="t_uprefs">Type Up-references</a>
1977 <div class="doc_text">
1980 <p>An "up reference" allows you to refer to a lexically enclosing type without
1981 requiring it to have a name. For instance, a structure declaration may
1982 contain a pointer to any of the types it is lexically a member of. Example
1983 of up references (with their equivalent as named type declarations)
1987 { \2 * } %x = type { %x* }
1988 { \2 }* %y = type { %y }*
1992 <p>An up reference is needed by the asmprinter for printing out cyclic types
1993 when there is no declared name for a type in the cycle. Because the
1994 asmprinter does not want to print out an infinite type string, it needs a
1995 syntax to handle recursive types that have no names (all names are optional
2003 <p>The level is the count of the lexical type that is being referred to.</p>
2006 <table class="layout">
2008 <td class="left"><tt>\1*</tt></td>
2009 <td class="left">Self-referential pointer.</td>
2012 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
2013 <td class="left">Recursive structure where the upref refers to the out-most
2020 <!-- *********************************************************************** -->
2021 <div class="doc_section"> <a name="constants">Constants</a> </div>
2022 <!-- *********************************************************************** -->
2024 <div class="doc_text">
2026 <p>LLVM has several different basic types of constants. This section describes
2027 them all and their syntax.</p>
2031 <!-- ======================================================================= -->
2032 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
2034 <div class="doc_text">
2037 <dt><b>Boolean constants</b></dt>
2038 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
2039 constants of the <tt><a href="#t_integer">i1</a></tt> type.</dd>
2041 <dt><b>Integer constants</b></dt>
2042 <dd>Standard integers (such as '4') are constants of
2043 the <a href="#t_integer">integer</a> type. Negative numbers may be used
2044 with integer types.</dd>
2046 <dt><b>Floating point constants</b></dt>
2047 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
2048 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
2049 notation (see below). The assembler requires the exact decimal value of a
2050 floating-point constant. For example, the assembler accepts 1.25 but
2051 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
2052 constants must have a <a href="#t_floating">floating point</a> type. </dd>
2054 <dt><b>Null pointer constants</b></dt>
2055 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
2056 and must be of <a href="#t_pointer">pointer type</a>.</dd>
2059 <p>The one non-intuitive notation for constants is the hexadecimal form of
2060 floating point constants. For example, the form '<tt>double
2061 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
2062 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
2063 constants are required (and the only time that they are generated by the
2064 disassembler) is when a floating point constant must be emitted but it cannot
2065 be represented as a decimal floating point number in a reasonable number of
2066 digits. For example, NaN's, infinities, and other special values are
2067 represented in their IEEE hexadecimal format so that assembly and disassembly
2068 do not cause any bits to change in the constants.</p>
2070 <p>When using the hexadecimal form, constants of types float and double are
2071 represented using the 16-digit form shown above (which matches the IEEE754
2072 representation for double); float values must, however, be exactly
2073 representable as IEE754 single precision. Hexadecimal format is always used
2074 for long double, and there are three forms of long double. The 80-bit format
2075 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
2076 The 128-bit format used by PowerPC (two adjacent doubles) is represented
2077 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
2078 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
2079 currently supported target uses this format. Long doubles will only work if
2080 they match the long double format on your target. All hexadecimal formats
2081 are big-endian (sign bit at the left).</p>
2083 <p>There are no constants of type x86mmx.</p>
2086 <!-- ======================================================================= -->
2087 <div class="doc_subsection">
2088 <a name="aggregateconstants"></a> <!-- old anchor -->
2089 <a name="complexconstants">Complex Constants</a>
2092 <div class="doc_text">
2094 <p>Complex constants are a (potentially recursive) combination of simple
2095 constants and smaller complex constants.</p>
2098 <dt><b>Structure constants</b></dt>
2099 <dd>Structure constants are represented with notation similar to structure
2100 type definitions (a comma separated list of elements, surrounded by braces
2101 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
2102 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
2103 Structure constants must have <a href="#t_struct">structure type</a>, and
2104 the number and types of elements must match those specified by the
2107 <dt><b>Array constants</b></dt>
2108 <dd>Array constants are represented with notation similar to array type
2109 definitions (a comma separated list of elements, surrounded by square
2110 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
2111 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
2112 the number and types of elements must match those specified by the
2115 <dt><b>Vector constants</b></dt>
2116 <dd>Vector constants are represented with notation similar to vector type
2117 definitions (a comma separated list of elements, surrounded by
2118 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32
2119 42, i32 11, i32 74, i32 100 ></tt>". Vector constants must
2120 have <a href="#t_vector">vector type</a>, and the number and types of
2121 elements must match those specified by the type.</dd>
2123 <dt><b>Zero initialization</b></dt>
2124 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
2125 value to zero of <em>any</em> type, including scalar and
2126 <a href="#t_aggregate">aggregate</a> types.
2127 This is often used to avoid having to print large zero initializers
2128 (e.g. for large arrays) and is always exactly equivalent to using explicit
2129 zero initializers.</dd>
2131 <dt><b>Metadata node</b></dt>
2132 <dd>A metadata node is a structure-like constant with
2133 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
2134 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
2135 be interpreted as part of the instruction stream, metadata is a place to
2136 attach additional information such as debug info.</dd>
2141 <!-- ======================================================================= -->
2142 <div class="doc_subsection">
2143 <a name="globalconstants">Global Variable and Function Addresses</a>
2146 <div class="doc_text">
2148 <p>The addresses of <a href="#globalvars">global variables</a>
2149 and <a href="#functionstructure">functions</a> are always implicitly valid
2150 (link-time) constants. These constants are explicitly referenced when
2151 the <a href="#identifiers">identifier for the global</a> is used and always
2152 have <a href="#t_pointer">pointer</a> type. For example, the following is a
2153 legal LLVM file:</p>
2155 <pre class="doc_code">
2158 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2163 <!-- ======================================================================= -->
2164 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
2165 <div class="doc_text">
2167 <p>The string '<tt>undef</tt>' can be used anywhere a constant is expected, and
2168 indicates that the user of the value may receive an unspecified bit-pattern.
2169 Undefined values may be of any type (other than '<tt>label</tt>'
2170 or '<tt>void</tt>') and be used anywhere a constant is permitted.</p>
2172 <p>Undefined values are useful because they indicate to the compiler that the
2173 program is well defined no matter what value is used. This gives the
2174 compiler more freedom to optimize. Here are some examples of (potentially
2175 surprising) transformations that are valid (in pseudo IR):</p>
2178 <pre class="doc_code">
2188 <p>This is safe because all of the output bits are affected by the undef bits.
2189 Any output bit can have a zero or one depending on the input bits.</p>
2191 <pre class="doc_code">
2202 <p>These logical operations have bits that are not always affected by the input.
2203 For example, if <tt>%X</tt> has a zero bit, then the output of the
2204 '<tt>and</tt>' operation will always be a zero for that bit, no matter what
2205 the corresponding bit from the '<tt>undef</tt>' is. As such, it is unsafe to
2206 optimize or assume that the result of the '<tt>and</tt>' is '<tt>undef</tt>'.
2207 However, it is safe to assume that all bits of the '<tt>undef</tt>' could be
2208 0, and optimize the '<tt>and</tt>' to 0. Likewise, it is safe to assume that
2209 all the bits of the '<tt>undef</tt>' operand to the '<tt>or</tt>' could be
2210 set, allowing the '<tt>or</tt>' to be folded to -1.</p>
2212 <pre class="doc_code">
2213 %A = select undef, %X, %Y
2214 %B = select undef, 42, %Y
2215 %C = select %X, %Y, undef
2226 <p>This set of examples shows that undefined '<tt>select</tt>' (and conditional
2227 branch) conditions can go <em>either way</em>, but they have to come from one
2228 of the two operands. In the <tt>%A</tt> example, if <tt>%X</tt> and
2229 <tt>%Y</tt> were both known to have a clear low bit, then <tt>%A</tt> would
2230 have to have a cleared low bit. However, in the <tt>%C</tt> example, the
2231 optimizer is allowed to assume that the '<tt>undef</tt>' operand could be the
2232 same as <tt>%Y</tt>, allowing the whole '<tt>select</tt>' to be
2235 <pre class="doc_code">
2236 %A = xor undef, undef
2254 <p>This example points out that two '<tt>undef</tt>' operands are not
2255 necessarily the same. This can be surprising to people (and also matches C
2256 semantics) where they assume that "<tt>X^X</tt>" is always zero, even
2257 if <tt>X</tt> is undefined. This isn't true for a number of reasons, but the
2258 short answer is that an '<tt>undef</tt>' "variable" can arbitrarily change
2259 its value over its "live range". This is true because the variable doesn't
2260 actually <em>have a live range</em>. Instead, the value is logically read
2261 from arbitrary registers that happen to be around when needed, so the value
2262 is not necessarily consistent over time. In fact, <tt>%A</tt> and <tt>%C</tt>
2263 need to have the same semantics or the core LLVM "replace all uses with"
2264 concept would not hold.</p>
2266 <pre class="doc_code">
2274 <p>These examples show the crucial difference between an <em>undefined
2275 value</em> and <em>undefined behavior</em>. An undefined value (like
2276 '<tt>undef</tt>') is allowed to have an arbitrary bit-pattern. This means that
2277 the <tt>%A</tt> operation can be constant folded to '<tt>undef</tt>', because
2278 the '<tt>undef</tt>' could be an SNaN, and <tt>fdiv</tt> is not (currently)
2279 defined on SNaN's. However, in the second example, we can make a more
2280 aggressive assumption: because the <tt>undef</tt> is allowed to be an
2281 arbitrary value, we are allowed to assume that it could be zero. Since a
2282 divide by zero has <em>undefined behavior</em>, we are allowed to assume that
2283 the operation does not execute at all. This allows us to delete the divide and
2284 all code after it. Because the undefined operation "can't happen", the
2285 optimizer can assume that it occurs in dead code.</p>
2287 <pre class="doc_code">
2288 a: store undef -> %X
2289 b: store %X -> undef
2295 <p>These examples reiterate the <tt>fdiv</tt> example: a store <em>of</em> an
2296 undefined value can be assumed to not have any effect; we can assume that the
2297 value is overwritten with bits that happen to match what was already there.
2298 However, a store <em>to</em> an undefined location could clobber arbitrary
2299 memory, therefore, it has undefined behavior.</p>
2303 <!-- ======================================================================= -->
2304 <div class="doc_subsection"><a name="trapvalues">Trap Values</a></div>
2305 <div class="doc_text">
2307 <p>Trap values are similar to <a href="#undefvalues">undef values</a>, however
2308 instead of representing an unspecified bit pattern, they represent the
2309 fact that an instruction or constant expression which cannot evoke side
2310 effects has nevertheless detected a condition which results in undefined
2313 <p>There is currently no way of representing a trap value in the IR; they
2314 only exist when produced by operations such as
2315 <a href="#i_add"><tt>add</tt></a> with the <tt>nsw</tt> flag.</p>
2317 <p>Trap value behavior is defined in terms of value <i>dependence</i>:</p>
2320 <li>Values other than <a href="#i_phi"><tt>phi</tt></a> nodes depend on
2321 their operands.</li>
2323 <li><a href="#i_phi"><tt>Phi</tt></a> nodes depend on the operand corresponding
2324 to their dynamic predecessor basic block.</li>
2326 <li>Function arguments depend on the corresponding actual argument values in
2327 the dynamic callers of their functions.</li>
2329 <li><a href="#i_call"><tt>Call</tt></a> instructions depend on the
2330 <a href="#i_ret"><tt>ret</tt></a> instructions that dynamically transfer
2331 control back to them.</li>
2333 <li><a href="#i_invoke"><tt>Invoke</tt></a> instructions depend on the
2334 <a href="#i_ret"><tt>ret</tt></a>, <a href="#i_unwind"><tt>unwind</tt></a>,
2335 or exception-throwing call instructions that dynamically transfer control
2338 <li>Non-volatile loads and stores depend on the most recent stores to all of the
2339 referenced memory addresses, following the order in the IR
2340 (including loads and stores implied by intrinsics such as
2341 <a href="#int_memcpy"><tt>@llvm.memcpy</tt></a>.)</li>
2343 <!-- TODO: In the case of multiple threads, this only applies if the store
2344 "happens-before" the load or store. -->
2346 <!-- TODO: floating-point exception state -->
2348 <li>An instruction with externally visible side effects depends on the most
2349 recent preceding instruction with externally visible side effects, following
2350 the order in the IR. (This includes
2351 <a href="#volatile">volatile operations</a>.)</li>
2353 <li>An instruction <i>control-depends</i> on a
2354 <a href="#terminators">terminator instruction</a>
2355 if the terminator instruction has multiple successors and the instruction
2356 is always executed when control transfers to one of the successors, and
2357 may not be executed when control is transfered to another.</li>
2359 <li>Dependence is transitive.</li>
2363 <p>Whenever a trap value is generated, all values which depend on it evaluate
2364 to trap. If they have side effects, the evoke their side effects as if each
2365 operand with a trap value were undef. If they have externally-visible side
2366 effects, the behavior is undefined.</p>
2368 <p>Here are some examples:</p>
2370 <pre class="doc_code">
2372 %trap = sub nuw i32 0, 1 ; Results in a trap value.
2373 %still_trap = and i32 %trap, 0 ; Whereas (and i32 undef, 0) would return 0.
2374 %trap_yet_again = getelementptr i32* @h, i32 %still_trap
2375 store i32 0, i32* %trap_yet_again ; undefined behavior
2377 store i32 %trap, i32* @g ; Trap value conceptually stored to memory.
2378 %trap2 = load i32* @g ; Returns a trap value, not just undef.
2380 volatile store i32 %trap, i32* @g ; External observation; undefined behavior.
2382 %narrowaddr = bitcast i32* @g to i16*
2383 %wideaddr = bitcast i32* @g to i64*
2384 %trap3 = load 16* %narrowaddr ; Returns a trap value.
2385 %trap4 = load i64* %widaddr ; Returns a trap value.
2387 %cmp = icmp i32 slt %trap, 0 ; Returns a trap value.
2388 %br i1 %cmp, %true, %end ; Branch to either destination.
2391 volatile store i32 0, i32* @g ; This is control-dependent on %cmp, so
2392 ; it has undefined behavior.
2396 %p = phi i32 [ 0, %entry ], [ 1, %true ]
2397 ; Both edges into this PHI are
2398 ; control-dependent on %cmp, so this
2399 ; always results in a trap value.
2401 volatile store i32 0, i32* @g ; %end is control-equivalent to %entry
2402 ; so this is defined (ignoring earlier
2403 ; undefined behavior in this example).
2408 <!-- ======================================================================= -->
2409 <div class="doc_subsection"><a name="blockaddress">Addresses of Basic
2411 <div class="doc_text">
2413 <p><b><tt>blockaddress(@function, %block)</tt></b></p>
2415 <p>The '<tt>blockaddress</tt>' constant computes the address of the specified
2416 basic block in the specified function, and always has an i8* type. Taking
2417 the address of the entry block is illegal.</p>
2419 <p>This value only has defined behavior when used as an operand to the
2420 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>' instruction, or for
2421 comparisons against null. Pointer equality tests between labels addresses
2422 results in undefined behavior — though, again, comparison against null
2423 is ok, and no label is equal to the null pointer. This may be passed around
2424 as an opaque pointer sized value as long as the bits are not inspected. This
2425 allows <tt>ptrtoint</tt> and arithmetic to be performed on these values so
2426 long as the original value is reconstituted before the <tt>indirectbr</tt>
2429 <p>Finally, some targets may provide defined semantics when using the value as
2430 the operand to an inline assembly, but that is target specific.</p>
2435 <!-- ======================================================================= -->
2436 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
2439 <div class="doc_text">
2441 <p>Constant expressions are used to allow expressions involving other constants
2442 to be used as constants. Constant expressions may be of
2443 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2444 operation that does not have side effects (e.g. load and call are not
2445 supported). The following is the syntax for constant expressions:</p>
2448 <dt><b><tt>trunc (CST to TYPE)</tt></b></dt>
2449 <dd>Truncate a constant to another type. The bit size of CST must be larger
2450 than the bit size of TYPE. Both types must be integers.</dd>
2452 <dt><b><tt>zext (CST to TYPE)</tt></b></dt>
2453 <dd>Zero extend a constant to another type. The bit size of CST must be
2454 smaller than the bit size of TYPE. Both types must be integers.</dd>
2456 <dt><b><tt>sext (CST to TYPE)</tt></b></dt>
2457 <dd>Sign extend a constant to another type. The bit size of CST must be
2458 smaller than the bit size of TYPE. Both types must be integers.</dd>
2460 <dt><b><tt>fptrunc (CST to TYPE)</tt></b></dt>
2461 <dd>Truncate a floating point constant to another floating point type. The
2462 size of CST must be larger than the size of TYPE. Both types must be
2463 floating point.</dd>
2465 <dt><b><tt>fpext (CST to TYPE)</tt></b></dt>
2466 <dd>Floating point extend a constant to another type. The size of CST must be
2467 smaller or equal to the size of TYPE. Both types must be floating
2470 <dt><b><tt>fptoui (CST to TYPE)</tt></b></dt>
2471 <dd>Convert a floating point constant to the corresponding unsigned integer
2472 constant. TYPE must be a scalar or vector integer type. CST must be of
2473 scalar or vector floating point type. Both CST and TYPE must be scalars,
2474 or vectors of the same number of elements. If the value won't fit in the
2475 integer type, the results are undefined.</dd>
2477 <dt><b><tt>fptosi (CST to TYPE)</tt></b></dt>
2478 <dd>Convert a floating point constant to the corresponding signed integer
2479 constant. TYPE must be a scalar or vector integer type. CST must be of
2480 scalar or vector floating point type. Both CST and TYPE must be scalars,
2481 or vectors of the same number of elements. If the value won't fit in the
2482 integer type, the results are undefined.</dd>
2484 <dt><b><tt>uitofp (CST to TYPE)</tt></b></dt>
2485 <dd>Convert an unsigned integer constant to the corresponding floating point
2486 constant. TYPE must be a scalar or vector floating point type. CST must be
2487 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2488 vectors of the same number of elements. If the value won't fit in the
2489 floating point type, the results are undefined.</dd>
2491 <dt><b><tt>sitofp (CST to TYPE)</tt></b></dt>
2492 <dd>Convert a signed integer constant to the corresponding floating point
2493 constant. TYPE must be a scalar or vector floating point type. CST must be
2494 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2495 vectors of the same number of elements. If the value won't fit in the
2496 floating point type, the results are undefined.</dd>
2498 <dt><b><tt>ptrtoint (CST to TYPE)</tt></b></dt>
2499 <dd>Convert a pointer typed constant to the corresponding integer constant
2500 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2501 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2502 make it fit in <tt>TYPE</tt>.</dd>
2504 <dt><b><tt>inttoptr (CST to TYPE)</tt></b></dt>
2505 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2506 type. CST must be of integer type. The CST value is zero extended,
2507 truncated, or unchanged to make it fit in a pointer size. This one is
2508 <i>really</i> dangerous!</dd>
2510 <dt><b><tt>bitcast (CST to TYPE)</tt></b></dt>
2511 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2512 are the same as those for the <a href="#i_bitcast">bitcast
2513 instruction</a>.</dd>
2515 <dt><b><tt>getelementptr (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2516 <dt><b><tt>getelementptr inbounds (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2517 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2518 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2519 instruction, the index list may have zero or more indexes, which are
2520 required to make sense for the type of "CSTPTR".</dd>
2522 <dt><b><tt>select (COND, VAL1, VAL2)</tt></b></dt>
2523 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
2525 <dt><b><tt>icmp COND (VAL1, VAL2)</tt></b></dt>
2526 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2528 <dt><b><tt>fcmp COND (VAL1, VAL2)</tt></b></dt>
2529 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2531 <dt><b><tt>extractelement (VAL, IDX)</tt></b></dt>
2532 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2535 <dt><b><tt>insertelement (VAL, ELT, IDX)</tt></b></dt>
2536 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2539 <dt><b><tt>shufflevector (VEC1, VEC2, IDXMASK)</tt></b></dt>
2540 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2543 <dt><b><tt>extractvalue (VAL, IDX0, IDX1, ...)</tt></b></dt>
2544 <dd>Perform the <a href="#i_extractvalue">extractvalue operation</a> on
2545 constants. The index list is interpreted in a similar manner as indices in
2546 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2547 index value must be specified.</dd>
2549 <dt><b><tt>insertvalue (VAL, ELT, IDX0, IDX1, ...)</tt></b></dt>
2550 <dd>Perform the <a href="#i_insertvalue">insertvalue operation</a> on
2551 constants. The index list is interpreted in a similar manner as indices in
2552 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2553 index value must be specified.</dd>
2555 <dt><b><tt>OPCODE (LHS, RHS)</tt></b></dt>
2556 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2557 be any of the <a href="#binaryops">binary</a>
2558 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2559 on operands are the same as those for the corresponding instruction
2560 (e.g. no bitwise operations on floating point values are allowed).</dd>
2565 <!-- *********************************************************************** -->
2566 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2567 <!-- *********************************************************************** -->
2569 <!-- ======================================================================= -->
2570 <div class="doc_subsection">
2571 <a name="inlineasm">Inline Assembler Expressions</a>
2574 <div class="doc_text">
2576 <p>LLVM supports inline assembler expressions (as opposed
2577 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2578 a special value. This value represents the inline assembler as a string
2579 (containing the instructions to emit), a list of operand constraints (stored
2580 as a string), a flag that indicates whether or not the inline asm
2581 expression has side effects, and a flag indicating whether the function
2582 containing the asm needs to align its stack conservatively. An example
2583 inline assembler expression is:</p>
2585 <pre class="doc_code">
2586 i32 (i32) asm "bswap $0", "=r,r"
2589 <p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2590 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2593 <pre class="doc_code">
2594 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2597 <p>Inline asms with side effects not visible in the constraint list must be
2598 marked as having side effects. This is done through the use of the
2599 '<tt>sideeffect</tt>' keyword, like so:</p>
2601 <pre class="doc_code">
2602 call void asm sideeffect "eieio", ""()
2605 <p>In some cases inline asms will contain code that will not work unless the
2606 stack is aligned in some way, such as calls or SSE instructions on x86,
2607 yet will not contain code that does that alignment within the asm.
2608 The compiler should make conservative assumptions about what the asm might
2609 contain and should generate its usual stack alignment code in the prologue
2610 if the '<tt>alignstack</tt>' keyword is present:</p>
2612 <pre class="doc_code">
2613 call void asm alignstack "eieio", ""()
2616 <p>If both keywords appear the '<tt>sideeffect</tt>' keyword must come
2619 <p>TODO: The format of the asm and constraints string still need to be
2620 documented here. Constraints on what can be done (e.g. duplication, moving,
2621 etc need to be documented). This is probably best done by reference to
2622 another document that covers inline asm from a holistic perspective.</p>
2625 <div class="doc_subsubsection">
2626 <a name="inlineasm_md">Inline Asm Metadata</a>
2629 <div class="doc_text">
2631 <p>The call instructions that wrap inline asm nodes may have a "!srcloc" MDNode
2632 attached to it that contains a list of constant integers. If present, the
2633 code generator will use the integer as the location cookie value when report
2634 errors through the LLVMContext error reporting mechanisms. This allows a
2635 front-end to correlate backend errors that occur with inline asm back to the
2636 source code that produced it. For example:</p>
2638 <pre class="doc_code">
2639 call void asm sideeffect "something bad", ""()<b>, !srcloc !42</b>
2641 !42 = !{ i32 1234567 }
2644 <p>It is up to the front-end to make sense of the magic numbers it places in the
2645 IR. If the MDNode contains multiple constants, the code generator will use
2646 the one that corresponds to the line of the asm that the error occurs on.</p>
2650 <!-- ======================================================================= -->
2651 <div class="doc_subsection"><a name="metadata">Metadata Nodes and Metadata
2655 <div class="doc_text">
2657 <p>LLVM IR allows metadata to be attached to instructions in the program that
2658 can convey extra information about the code to the optimizers and code
2659 generator. One example application of metadata is source-level debug
2660 information. There are two metadata primitives: strings and nodes. All
2661 metadata has the <tt>metadata</tt> type and is identified in syntax by a
2662 preceding exclamation point ('<tt>!</tt>').</p>
2664 <p>A metadata string is a string surrounded by double quotes. It can contain
2665 any character by escaping non-printable characters with "\xx" where "xx" is
2666 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
2668 <p>Metadata nodes are represented with notation similar to structure constants
2669 (a comma separated list of elements, surrounded by braces and preceded by an
2670 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2671 10}</tt>". Metadata nodes can have any values as their operand.</p>
2673 <p>A <a href="#namedmetadatastructure">named metadata</a> is a collection of
2674 metadata nodes, which can be looked up in the module symbol table. For
2675 example: "<tt>!foo = metadata !{!4, !3}</tt>".
2677 <p>Metadata can be used as function arguments. Here <tt>llvm.dbg.value</tt>
2678 function is using two metadata arguments.</p>
2680 <pre class="doc_code">
2681 call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
2684 <p>Metadata can be attached with an instruction. Here metadata <tt>!21</tt> is
2685 attached with <tt>add</tt> instruction using <tt>!dbg</tt> identifier.</p>
2687 <pre class="doc_code">
2688 %indvar.next = add i64 %indvar, 1, !dbg !21
2693 <!-- *********************************************************************** -->
2694 <div class="doc_section">
2695 <a name="intrinsic_globals">Intrinsic Global Variables</a>
2697 <!-- *********************************************************************** -->
2699 <p>LLVM has a number of "magic" global variables that contain data that affect
2700 code generation or other IR semantics. These are documented here. All globals
2701 of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2702 section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2705 <!-- ======================================================================= -->
2706 <div class="doc_subsection">
2707 <a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
2710 <div class="doc_text">
2712 <p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2713 href="#linkage_appending">appending linkage</a>. This array contains a list of
2714 pointers to global variables and functions which may optionally have a pointer
2715 cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2721 @llvm.used = appending global [2 x i8*] [
2723 i8* bitcast (i32* @Y to i8*)
2724 ], section "llvm.metadata"
2727 <p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2728 compiler, assembler, and linker are required to treat the symbol as if there is
2729 a reference to the global that it cannot see. For example, if a variable has
2730 internal linkage and no references other than that from the <tt>@llvm.used</tt>
2731 list, it cannot be deleted. This is commonly used to represent references from
2732 inline asms and other things the compiler cannot "see", and corresponds to
2733 "attribute((used))" in GNU C.</p>
2735 <p>On some targets, the code generator must emit a directive to the assembler or
2736 object file to prevent the assembler and linker from molesting the symbol.</p>
2740 <!-- ======================================================================= -->
2741 <div class="doc_subsection">
2742 <a name="intg_compiler_used">The '<tt>llvm.compiler.used</tt>' Global Variable</a>
2745 <div class="doc_text">
2747 <p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2748 <tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2749 touching the symbol. On targets that support it, this allows an intelligent
2750 linker to optimize references to the symbol without being impeded as it would be
2751 by <tt>@llvm.used</tt>.</p>
2753 <p>This is a rare construct that should only be used in rare circumstances, and
2754 should not be exposed to source languages.</p>
2758 <!-- ======================================================================= -->
2759 <div class="doc_subsection">
2760 <a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
2763 <div class="doc_text">
2765 %0 = type { i32, void ()* }
2766 @llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor }]
2768 <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.
2773 <!-- ======================================================================= -->
2774 <div class="doc_subsection">
2775 <a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
2778 <div class="doc_text">
2780 %0 = type { i32, void ()* }
2781 @llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor }]
2784 <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.
2790 <!-- *********************************************************************** -->
2791 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2792 <!-- *********************************************************************** -->
2794 <div class="doc_text">
2796 <p>The LLVM instruction set consists of several different classifications of
2797 instructions: <a href="#terminators">terminator
2798 instructions</a>, <a href="#binaryops">binary instructions</a>,
2799 <a href="#bitwiseops">bitwise binary instructions</a>,
2800 <a href="#memoryops">memory instructions</a>, and
2801 <a href="#otherops">other instructions</a>.</p>
2805 <!-- ======================================================================= -->
2806 <div class="doc_subsection"> <a name="terminators">Terminator
2807 Instructions</a> </div>
2809 <div class="doc_text">
2811 <p>As mentioned <a href="#functionstructure">previously</a>, every basic block
2812 in a program ends with a "Terminator" instruction, which indicates which
2813 block should be executed after the current block is finished. These
2814 terminator instructions typically yield a '<tt>void</tt>' value: they produce
2815 control flow, not values (the one exception being the
2816 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2818 <p>There are seven different terminator instructions: the
2819 '<a href="#i_ret"><tt>ret</tt></a>' instruction, the
2820 '<a href="#i_br"><tt>br</tt></a>' instruction, the
2821 '<a href="#i_switch"><tt>switch</tt></a>' instruction, the
2822 '<a href="#i_indirectbr">'<tt>indirectbr</tt></a>' Instruction, the
2823 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the
2824 '<a href="#i_unwind"><tt>unwind</tt></a>' instruction, and the
2825 '<a href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2829 <!-- _______________________________________________________________________ -->
2830 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2831 Instruction</a> </div>
2833 <div class="doc_text">
2837 ret <type> <value> <i>; Return a value from a non-void function</i>
2838 ret void <i>; Return from void function</i>
2842 <p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
2843 a value) from a function back to the caller.</p>
2845 <p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
2846 value and then causes control flow, and one that just causes control flow to
2850 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
2851 return value. The type of the return value must be a
2852 '<a href="#t_firstclass">first class</a>' type.</p>
2854 <p>A function is not <a href="#wellformed">well formed</a> if it it has a
2855 non-void return type and contains a '<tt>ret</tt>' instruction with no return
2856 value or a return value with a type that does not match its type, or if it
2857 has a void return type and contains a '<tt>ret</tt>' instruction with a
2861 <p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
2862 the calling function's context. If the caller is a
2863 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
2864 instruction after the call. If the caller was an
2865 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
2866 the beginning of the "normal" destination block. If the instruction returns
2867 a value, that value shall set the call or invoke instruction's return
2872 ret i32 5 <i>; Return an integer value of 5</i>
2873 ret void <i>; Return from a void function</i>
2874 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
2878 <!-- _______________________________________________________________________ -->
2879 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2881 <div class="doc_text">
2885 br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2889 <p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
2890 different basic block in the current function. There are two forms of this
2891 instruction, corresponding to a conditional branch and an unconditional
2895 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
2896 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
2897 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
2901 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2902 argument is evaluated. If the value is <tt>true</tt>, control flows to the
2903 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2904 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2909 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
2910 br i1 %cond, label %IfEqual, label %IfUnequal
2912 <a href="#i_ret">ret</a> i32 1
2914 <a href="#i_ret">ret</a> i32 0
2919 <!-- _______________________________________________________________________ -->
2920 <div class="doc_subsubsection">
2921 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2924 <div class="doc_text">
2928 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2932 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2933 several different places. It is a generalization of the '<tt>br</tt>'
2934 instruction, allowing a branch to occur to one of many possible
2938 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2939 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
2940 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
2941 The table is not allowed to contain duplicate constant entries.</p>
2944 <p>The <tt>switch</tt> instruction specifies a table of values and
2945 destinations. When the '<tt>switch</tt>' instruction is executed, this table
2946 is searched for the given value. If the value is found, control flow is
2947 transferred to the corresponding destination; otherwise, control flow is
2948 transferred to the default destination.</p>
2950 <h5>Implementation:</h5>
2951 <p>Depending on properties of the target machine and the particular
2952 <tt>switch</tt> instruction, this instruction may be code generated in
2953 different ways. For example, it could be generated as a series of chained
2954 conditional branches or with a lookup table.</p>
2958 <i>; Emulate a conditional br instruction</i>
2959 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2960 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
2962 <i>; Emulate an unconditional br instruction</i>
2963 switch i32 0, label %dest [ ]
2965 <i>; Implement a jump table:</i>
2966 switch i32 %val, label %otherwise [ i32 0, label %onzero
2968 i32 2, label %ontwo ]
2974 <!-- _______________________________________________________________________ -->
2975 <div class="doc_subsubsection">
2976 <a name="i_indirectbr">'<tt>indirectbr</tt>' Instruction</a>
2979 <div class="doc_text">
2983 indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
2988 <p>The '<tt>indirectbr</tt>' instruction implements an indirect branch to a label
2989 within the current function, whose address is specified by
2990 "<tt>address</tt>". Address must be derived from a <a
2991 href="#blockaddress">blockaddress</a> constant.</p>
2995 <p>The '<tt>address</tt>' argument is the address of the label to jump to. The
2996 rest of the arguments indicate the full set of possible destinations that the
2997 address may point to. Blocks are allowed to occur multiple times in the
2998 destination list, though this isn't particularly useful.</p>
3000 <p>This destination list is required so that dataflow analysis has an accurate
3001 understanding of the CFG.</p>
3005 <p>Control transfers to the block specified in the address argument. All
3006 possible destination blocks must be listed in the label list, otherwise this
3007 instruction has undefined behavior. This implies that jumps to labels
3008 defined in other functions have undefined behavior as well.</p>
3010 <h5>Implementation:</h5>
3012 <p>This is typically implemented with a jump through a register.</p>
3016 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
3022 <!-- _______________________________________________________________________ -->
3023 <div class="doc_subsubsection">
3024 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
3027 <div class="doc_text">
3031 <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>]
3032 to label <normal label> unwind label <exception label>
3036 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
3037 function, with the possibility of control flow transfer to either the
3038 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
3039 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
3040 control flow will return to the "normal" label. If the callee (or any
3041 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
3042 instruction, control is interrupted and continued at the dynamically nearest
3043 "exception" label.</p>
3046 <p>This instruction requires several arguments:</p>
3049 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
3050 convention</a> the call should use. If none is specified, the call
3051 defaults to using C calling conventions.</li>
3053 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
3054 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
3055 '<tt>inreg</tt>' attributes are valid here.</li>
3057 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
3058 function value being invoked. In most cases, this is a direct function
3059 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
3060 off an arbitrary pointer to function value.</li>
3062 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
3063 function to be invoked. </li>
3065 <li>'<tt>function args</tt>': argument list whose types match the function
3066 signature argument types and parameter attributes. All arguments must be
3067 of <a href="#t_firstclass">first class</a> type. If the function
3068 signature indicates the function accepts a variable number of arguments,
3069 the extra arguments can be specified.</li>
3071 <li>'<tt>normal label</tt>': the label reached when the called function
3072 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
3074 <li>'<tt>exception label</tt>': the label reached when a callee returns with
3075 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
3077 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
3078 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
3079 '<tt>readnone</tt>' attributes are valid here.</li>
3083 <p>This instruction is designed to operate as a standard
3084 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
3085 primary difference is that it establishes an association with a label, which
3086 is used by the runtime library to unwind the stack.</p>
3088 <p>This instruction is used in languages with destructors to ensure that proper
3089 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
3090 exception. Additionally, this is important for implementation of
3091 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
3093 <p>For the purposes of the SSA form, the definition of the value returned by the
3094 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
3095 block to the "normal" label. If the callee unwinds then no return value is
3098 <p>Note that the code generator does not yet completely support unwind, and
3099 that the invoke/unwind semantics are likely to change in future versions.</p>
3103 %retval = invoke i32 @Test(i32 15) to label %Continue
3104 unwind label %TestCleanup <i>; {i32}:retval set</i>
3105 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
3106 unwind label %TestCleanup <i>; {i32}:retval set</i>
3111 <!-- _______________________________________________________________________ -->
3113 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
3114 Instruction</a> </div>
3116 <div class="doc_text">
3124 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
3125 at the first callee in the dynamic call stack which used
3126 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
3127 This is primarily used to implement exception handling.</p>
3130 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
3131 immediately halt. The dynamic call stack is then searched for the
3132 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
3133 Once found, execution continues at the "exceptional" destination block
3134 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
3135 instruction in the dynamic call chain, undefined behavior results.</p>
3137 <p>Note that the code generator does not yet completely support unwind, and
3138 that the invoke/unwind semantics are likely to change in future versions.</p>
3142 <!-- _______________________________________________________________________ -->
3144 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
3145 Instruction</a> </div>
3147 <div class="doc_text">
3155 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
3156 instruction is used to inform the optimizer that a particular portion of the
3157 code is not reachable. This can be used to indicate that the code after a
3158 no-return function cannot be reached, and other facts.</p>
3161 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
3165 <!-- ======================================================================= -->
3166 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
3168 <div class="doc_text">
3170 <p>Binary operators are used to do most of the computation in a program. They
3171 require two operands of the same type, execute an operation on them, and
3172 produce a single value. The operands might represent multiple data, as is
3173 the case with the <a href="#t_vector">vector</a> data type. The result value
3174 has the same type as its operands.</p>
3176 <p>There are several different binary operators:</p>
3180 <!-- _______________________________________________________________________ -->
3181 <div class="doc_subsubsection">
3182 <a name="i_add">'<tt>add</tt>' Instruction</a>
3185 <div class="doc_text">
3189 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3190 <result> = add nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3191 <result> = add nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3192 <result> = add nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3196 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
3199 <p>The two arguments to the '<tt>add</tt>' instruction must
3200 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3201 integer values. Both arguments must have identical types.</p>
3204 <p>The value produced is the integer sum of the two operands.</p>
3206 <p>If the sum has unsigned overflow, the result returned is the mathematical
3207 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
3209 <p>Because LLVM integers use a two's complement representation, this instruction
3210 is appropriate for both signed and unsigned integers.</p>
3212 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3213 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3214 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
3215 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3216 respectively, occurs.</p>
3220 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
3225 <!-- _______________________________________________________________________ -->
3226 <div class="doc_subsubsection">
3227 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
3230 <div class="doc_text">
3234 <result> = fadd <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3238 <p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
3241 <p>The two arguments to the '<tt>fadd</tt>' instruction must be
3242 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3243 floating point values. Both arguments must have identical types.</p>
3246 <p>The value produced is the floating point sum of the two operands.</p>
3250 <result> = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
3255 <!-- _______________________________________________________________________ -->
3256 <div class="doc_subsubsection">
3257 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
3260 <div class="doc_text">
3264 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3265 <result> = sub nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3266 <result> = sub nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3267 <result> = sub nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3271 <p>The '<tt>sub</tt>' instruction returns the difference of its two
3274 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
3275 '<tt>neg</tt>' instruction present in most other intermediate
3276 representations.</p>
3279 <p>The two arguments to the '<tt>sub</tt>' instruction must
3280 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3281 integer values. Both arguments must have identical types.</p>
3284 <p>The value produced is the integer difference of the two operands.</p>
3286 <p>If the difference has unsigned overflow, the result returned is the
3287 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
3290 <p>Because LLVM integers use a two's complement representation, this instruction
3291 is appropriate for both signed and unsigned integers.</p>
3293 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3294 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3295 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
3296 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3297 respectively, occurs.</p>
3301 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
3302 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
3307 <!-- _______________________________________________________________________ -->
3308 <div class="doc_subsubsection">
3309 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
3312 <div class="doc_text">
3316 <result> = fsub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3320 <p>The '<tt>fsub</tt>' instruction returns the difference of its two
3323 <p>Note that the '<tt>fsub</tt>' instruction is used to represent the
3324 '<tt>fneg</tt>' instruction present in most other intermediate
3325 representations.</p>
3328 <p>The two arguments to the '<tt>fsub</tt>' instruction must be
3329 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3330 floating point values. Both arguments must have identical types.</p>
3333 <p>The value produced is the floating point difference of the two operands.</p>
3337 <result> = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
3338 <result> = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
3343 <!-- _______________________________________________________________________ -->
3344 <div class="doc_subsubsection">
3345 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
3348 <div class="doc_text">
3352 <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3353 <result> = mul nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3354 <result> = mul nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3355 <result> = mul nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3359 <p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
3362 <p>The two arguments to the '<tt>mul</tt>' instruction must
3363 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3364 integer values. Both arguments must have identical types.</p>
3367 <p>The value produced is the integer product of the two operands.</p>
3369 <p>If the result of the multiplication has unsigned overflow, the result
3370 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
3371 width of the result.</p>
3373 <p>Because LLVM integers use a two's complement representation, and the result
3374 is the same width as the operands, this instruction returns the correct
3375 result for both signed and unsigned integers. If a full product
3376 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
3377 be sign-extended or zero-extended as appropriate to the width of the full
3380 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3381 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3382 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
3383 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3384 respectively, occurs.</p>
3388 <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
3393 <!-- _______________________________________________________________________ -->
3394 <div class="doc_subsubsection">
3395 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
3398 <div class="doc_text">
3402 <result> = fmul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3406 <p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
3409 <p>The two arguments to the '<tt>fmul</tt>' instruction must be
3410 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3411 floating point values. Both arguments must have identical types.</p>
3414 <p>The value produced is the floating point product of the two operands.</p>
3418 <result> = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
3423 <!-- _______________________________________________________________________ -->
3424 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
3427 <div class="doc_text">
3431 <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3435 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
3438 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
3439 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3440 values. Both arguments must have identical types.</p>
3443 <p>The value produced is the unsigned integer quotient of the two operands.</p>
3445 <p>Note that unsigned integer division and signed integer division are distinct
3446 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
3448 <p>Division by zero leads to undefined behavior.</p>
3452 <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3457 <!-- _______________________________________________________________________ -->
3458 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
3461 <div class="doc_text">
3465 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3466 <result> = sdiv exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3470 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
3473 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
3474 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3475 values. Both arguments must have identical types.</p>
3478 <p>The value produced is the signed integer quotient of the two operands rounded
3481 <p>Note that signed integer division and unsigned integer division are distinct
3482 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
3484 <p>Division by zero leads to undefined behavior. Overflow also leads to
3485 undefined behavior; this is a rare case, but can occur, for example, by doing
3486 a 32-bit division of -2147483648 by -1.</p>
3488 <p>If the <tt>exact</tt> keyword is present, the result value of the
3489 <tt>sdiv</tt> is a <a href="#trapvalues">trap value</a> if the result would
3494 <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3499 <!-- _______________________________________________________________________ -->
3500 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
3501 Instruction</a> </div>
3503 <div class="doc_text">
3507 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3511 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
3514 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
3515 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3516 floating point values. Both arguments must have identical types.</p>
3519 <p>The value produced is the floating point quotient of the two operands.</p>
3523 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
3528 <!-- _______________________________________________________________________ -->
3529 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
3532 <div class="doc_text">
3536 <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3540 <p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3541 division of its two arguments.</p>
3544 <p>The two arguments to the '<tt>urem</tt>' instruction must be
3545 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3546 values. Both arguments must have identical types.</p>
3549 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
3550 This instruction always performs an unsigned division to get the
3553 <p>Note that unsigned integer remainder and signed integer remainder are
3554 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3556 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
3560 <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3565 <!-- _______________________________________________________________________ -->
3566 <div class="doc_subsubsection">
3567 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
3570 <div class="doc_text">
3574 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3578 <p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3579 division of its two operands. This instruction can also take
3580 <a href="#t_vector">vector</a> versions of the values in which case the
3581 elements must be integers.</p>
3584 <p>The two arguments to the '<tt>srem</tt>' instruction must be
3585 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3586 values. Both arguments must have identical types.</p>
3589 <p>This instruction returns the <i>remainder</i> of a division (where the result
3590 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
3591 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
3592 a value. For more information about the difference,
3593 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3594 Math Forum</a>. For a table of how this is implemented in various languages,
3595 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3596 Wikipedia: modulo operation</a>.</p>
3598 <p>Note that signed integer remainder and unsigned integer remainder are
3599 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3601 <p>Taking the remainder of a division by zero leads to undefined behavior.
3602 Overflow also leads to undefined behavior; this is a rare case, but can
3603 occur, for example, by taking the remainder of a 32-bit division of
3604 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3605 lets srem be implemented using instructions that return both the result of
3606 the division and the remainder.)</p>
3610 <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3615 <!-- _______________________________________________________________________ -->
3616 <div class="doc_subsubsection">
3617 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
3619 <div class="doc_text">
3623 <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3627 <p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3628 its two operands.</p>
3631 <p>The two arguments to the '<tt>frem</tt>' instruction must be
3632 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3633 floating point values. Both arguments must have identical types.</p>
3636 <p>This instruction returns the <i>remainder</i> of a division. The remainder
3637 has the same sign as the dividend.</p>
3641 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
3646 <!-- ======================================================================= -->
3647 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
3648 Operations</a> </div>
3650 <div class="doc_text">
3652 <p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3653 program. They are generally very efficient instructions and can commonly be
3654 strength reduced from other instructions. They require two operands of the
3655 same type, execute an operation on them, and produce a single value. The
3656 resulting value is the same type as its operands.</p>
3660 <!-- _______________________________________________________________________ -->
3661 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
3662 Instruction</a> </div>
3664 <div class="doc_text">
3668 <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3672 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3673 a specified number of bits.</p>
3676 <p>Both arguments to the '<tt>shl</tt>' instruction must be the
3677 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3678 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3681 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3682 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3683 is (statically or dynamically) negative or equal to or larger than the number
3684 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3685 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3686 shift amount in <tt>op2</tt>.</p>
3690 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
3691 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
3692 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
3693 <result> = shl i32 1, 32 <i>; undefined</i>
3694 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
3699 <!-- _______________________________________________________________________ -->
3700 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
3701 Instruction</a> </div>
3703 <div class="doc_text">
3707 <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3711 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
3712 operand shifted to the right a specified number of bits with zero fill.</p>
3715 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
3716 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3717 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3720 <p>This instruction always performs a logical shift right operation. The most
3721 significant bits of the result will be filled with zero bits after the shift.
3722 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
3723 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3724 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3725 shift amount in <tt>op2</tt>.</p>
3729 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
3730 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
3731 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
3732 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
3733 <result> = lshr i32 1, 32 <i>; undefined</i>
3734 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
3739 <!-- _______________________________________________________________________ -->
3740 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
3741 Instruction</a> </div>
3742 <div class="doc_text">
3746 <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3750 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3751 operand shifted to the right a specified number of bits with sign
3755 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
3756 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3757 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3760 <p>This instruction always performs an arithmetic shift right operation, The
3761 most significant bits of the result will be filled with the sign bit
3762 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3763 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
3764 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
3765 the corresponding shift amount in <tt>op2</tt>.</p>
3769 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3770 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3771 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3772 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
3773 <result> = ashr i32 1, 32 <i>; undefined</i>
3774 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
3779 <!-- _______________________________________________________________________ -->
3780 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
3781 Instruction</a> </div>
3783 <div class="doc_text">
3787 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3791 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
3795 <p>The two arguments to the '<tt>and</tt>' instruction must be
3796 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3797 values. Both arguments must have identical types.</p>
3800 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
3802 <table border="1" cellspacing="0" cellpadding="4">
3834 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
3835 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
3836 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
3839 <!-- _______________________________________________________________________ -->
3840 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
3842 <div class="doc_text">
3846 <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3850 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
3854 <p>The two arguments to the '<tt>or</tt>' instruction must be
3855 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3856 values. Both arguments must have identical types.</p>
3859 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
3861 <table border="1" cellspacing="0" cellpadding="4">
3893 <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
3894 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
3895 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
3900 <!-- _______________________________________________________________________ -->
3901 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3902 Instruction</a> </div>
3904 <div class="doc_text">
3908 <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3912 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
3913 its two operands. The <tt>xor</tt> is used to implement the "one's
3914 complement" operation, which is the "~" operator in C.</p>
3917 <p>The two arguments to the '<tt>xor</tt>' instruction must be
3918 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3919 values. Both arguments must have identical types.</p>
3922 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
3924 <table border="1" cellspacing="0" cellpadding="4">
3956 <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
3957 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3958 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3959 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
3964 <!-- ======================================================================= -->
3965 <div class="doc_subsection">
3966 <a name="vectorops">Vector Operations</a>
3969 <div class="doc_text">
3971 <p>LLVM supports several instructions to represent vector operations in a
3972 target-independent manner. These instructions cover the element-access and
3973 vector-specific operations needed to process vectors effectively. While LLVM
3974 does directly support these vector operations, many sophisticated algorithms
3975 will want to use target-specific intrinsics to take full advantage of a
3976 specific target.</p>
3980 <!-- _______________________________________________________________________ -->
3981 <div class="doc_subsubsection">
3982 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3985 <div class="doc_text">
3989 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
3993 <p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
3994 from a vector at a specified index.</p>
3998 <p>The first operand of an '<tt>extractelement</tt>' instruction is a value
3999 of <a href="#t_vector">vector</a> type. The second operand is an index
4000 indicating the position from which to extract the element. The index may be
4004 <p>The result is a scalar of the same type as the element type of
4005 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
4006 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4007 results are undefined.</p>
4011 <result> = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
4016 <!-- _______________________________________________________________________ -->
4017 <div class="doc_subsubsection">
4018 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
4021 <div class="doc_text">
4025 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
4029 <p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
4030 vector at a specified index.</p>
4033 <p>The first operand of an '<tt>insertelement</tt>' instruction is a value
4034 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
4035 whose type must equal the element type of the first operand. The third
4036 operand is an index indicating the position at which to insert the value.
4037 The index may be a variable.</p>
4040 <p>The result is a vector of the same type as <tt>val</tt>. Its element values
4041 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
4042 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4043 results are undefined.</p>
4047 <result> = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
4052 <!-- _______________________________________________________________________ -->
4053 <div class="doc_subsubsection">
4054 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
4057 <div class="doc_text">
4061 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
4065 <p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
4066 from two input vectors, returning a vector with the same element type as the
4067 input and length that is the same as the shuffle mask.</p>
4070 <p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
4071 with types that match each other. The third argument is a shuffle mask whose
4072 element type is always 'i32'. The result of the instruction is a vector
4073 whose length is the same as the shuffle mask and whose element type is the
4074 same as the element type of the first two operands.</p>
4076 <p>The shuffle mask operand is required to be a constant vector with either
4077 constant integer or undef values.</p>
4080 <p>The elements of the two input vectors are numbered from left to right across
4081 both of the vectors. The shuffle mask operand specifies, for each element of
4082 the result vector, which element of the two input vectors the result element
4083 gets. The element selector may be undef (meaning "don't care") and the
4084 second operand may be undef if performing a shuffle from only one vector.</p>
4088 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4089 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
4090 <result> = shufflevector <4 x i32> %v1, <4 x i32> undef,
4091 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
4092 <result> = shufflevector <8 x i32> %v1, <8 x i32> undef,
4093 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
4094 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4095 <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>
4100 <!-- ======================================================================= -->
4101 <div class="doc_subsection">
4102 <a name="aggregateops">Aggregate Operations</a>
4105 <div class="doc_text">
4107 <p>LLVM supports several instructions for working with
4108 <a href="#t_aggregate">aggregate</a> values.</p>
4112 <!-- _______________________________________________________________________ -->
4113 <div class="doc_subsubsection">
4114 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
4117 <div class="doc_text">
4121 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
4125 <p>The '<tt>extractvalue</tt>' instruction extracts the value of a member field
4126 from an <a href="#t_aggregate">aggregate</a> value.</p>
4129 <p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
4130 of <a href="#t_struct">struct</a> or
4131 <a href="#t_array">array</a> type. The operands are constant indices to
4132 specify which value to extract in a similar manner as indices in a
4133 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
4134 <p>The major differences to <tt>getelementptr</tt> indexing are:</p>
4136 <li>Since the value being indexed is not a pointer, the first index is
4137 omitted and assumed to be zero.</li>
4138 <li>At least one index must be specified.</li>
4139 <li>Not only struct indices but also array indices must be in
4144 <p>The result is the value at the position in the aggregate specified by the
4149 <result> = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
4154 <!-- _______________________________________________________________________ -->
4155 <div class="doc_subsubsection">
4156 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
4159 <div class="doc_text">
4163 <result> = insertvalue <aggregate type> <val>, <ty> <elt>, <idx> <i>; yields <aggregate type></i>
4167 <p>The '<tt>insertvalue</tt>' instruction inserts a value into a member field
4168 in an <a href="#t_aggregate">aggregate</a> value.</p>
4171 <p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
4172 of <a href="#t_struct">struct</a> or
4173 <a href="#t_array">array</a> type. The second operand is a first-class
4174 value to insert. The following operands are constant indices indicating
4175 the position at which to insert the value in a similar manner as indices in a
4176 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' instruction. The
4177 value to insert must have the same type as the value identified by the
4181 <p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
4182 that of <tt>val</tt> except that the value at the position specified by the
4183 indices is that of <tt>elt</tt>.</p>
4187 %agg1 = insertvalue {i32, float} undef, i32 1, 0 <i>; yields {i32 1, float undef}</i>
4188 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 <i>; yields {i32 1, float %val}</i>
4194 <!-- ======================================================================= -->
4195 <div class="doc_subsection">
4196 <a name="memoryops">Memory Access and Addressing Operations</a>
4199 <div class="doc_text">
4201 <p>A key design point of an SSA-based representation is how it represents
4202 memory. In LLVM, no memory locations are in SSA form, which makes things
4203 very simple. This section describes how to read, write, and allocate
4208 <!-- _______________________________________________________________________ -->
4209 <div class="doc_subsubsection">
4210 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
4213 <div class="doc_text">
4217 <result> = alloca <type>[, <ty> <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
4221 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
4222 currently executing function, to be automatically released when this function
4223 returns to its caller. The object is always allocated in the generic address
4224 space (address space zero).</p>
4227 <p>The '<tt>alloca</tt>' instruction
4228 allocates <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the
4229 runtime stack, returning a pointer of the appropriate type to the program.
4230 If "NumElements" is specified, it is the number of elements allocated,
4231 otherwise "NumElements" is defaulted to be one. If a constant alignment is
4232 specified, the value result of the allocation is guaranteed to be aligned to
4233 at least that boundary. If not specified, or if zero, the target can choose
4234 to align the allocation on any convenient boundary compatible with the
4237 <p>'<tt>type</tt>' may be any sized type.</p>
4240 <p>Memory is allocated; a pointer is returned. The operation is undefined if
4241 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
4242 memory is automatically released when the function returns. The
4243 '<tt>alloca</tt>' instruction is commonly used to represent automatic
4244 variables that must have an address available. When the function returns
4245 (either with the <tt><a href="#i_ret">ret</a></tt>
4246 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
4247 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
4251 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
4252 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
4253 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
4254 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
4259 <!-- _______________________________________________________________________ -->
4260 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
4261 Instruction</a> </div>
4263 <div class="doc_text">
4267 <result> = load <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>]
4268 <result> = volatile load <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>]
4269 !<index> = !{ i32 1 }
4273 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
4276 <p>The argument to the '<tt>load</tt>' instruction specifies the memory address
4277 from which to load. The pointer must point to
4278 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
4279 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
4280 number or order of execution of this <tt>load</tt> with other <a
4281 href="#volatile">volatile operations</a>.</p>
4283 <p>The optional constant <tt>align</tt> argument specifies the alignment of the
4284 operation (that is, the alignment of the memory address). A value of 0 or an
4285 omitted <tt>align</tt> argument means that the operation has the preferential
4286 alignment for the target. It is the responsibility of the code emitter to
4287 ensure that the alignment information is correct. Overestimating the
4288 alignment results in undefined behavior. Underestimating the alignment may
4289 produce less efficient code. An alignment of 1 is always safe.</p>
4291 <p>The optional <tt>!nontemporal</tt> metadata must reference a single
4292 metatadata name <index> corresponding to a metadata node with
4293 one <tt>i32</tt> entry of value 1. The existence of
4294 the <tt>!nontemporal</tt> metatadata on the instruction tells the optimizer
4295 and code generator that this load is not expected to be reused in the cache.
4296 The code generator may select special instructions to save cache bandwidth,
4297 such as the <tt>MOVNT</tt> instruction on x86.</p>
4300 <p>The location of memory pointed to is loaded. If the value being loaded is of
4301 scalar type then the number of bytes read does not exceed the minimum number
4302 of bytes needed to hold all bits of the type. For example, loading an
4303 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
4304 <tt>i20</tt> with a size that is not an integral number of bytes, the result
4305 is undefined if the value was not originally written using a store of the
4310 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4311 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
4312 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
4317 <!-- _______________________________________________________________________ -->
4318 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
4319 Instruction</a> </div>
4321 <div class="doc_text">
4325 store <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] <i>; yields {void}</i>
4326 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] <i>; yields {void}</i>
4330 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
4333 <p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
4334 and an address at which to store it. The type of the
4335 '<tt><pointer></tt>' operand must be a pointer to
4336 the <a href="#t_firstclass">first class</a> type of the
4337 '<tt><value></tt>' operand. If the <tt>store</tt> is marked as
4338 <tt>volatile</tt>, then the optimizer is not allowed to modify the number or
4339 order of execution of this <tt>store</tt> with other <a
4340 href="#volatile">volatile operations</a>.</p>
4342 <p>The optional constant "align" argument specifies the alignment of the
4343 operation (that is, the alignment of the memory address). A value of 0 or an
4344 omitted "align" argument means that the operation has the preferential
4345 alignment for the target. It is the responsibility of the code emitter to
4346 ensure that the alignment information is correct. Overestimating the
4347 alignment results in an undefined behavior. Underestimating the alignment may
4348 produce less efficient code. An alignment of 1 is always safe.</p>
4350 <p>The optional !nontemporal metadata must reference a single metatadata
4351 name <index> corresponding to a metadata node with one i32 entry of
4352 value 1. The existence of the !nontemporal metatadata on the
4353 instruction tells the optimizer and code generator that this load is
4354 not expected to be reused in the cache. The code generator may
4355 select special instructions to save cache bandwidth, such as the
4356 MOVNT instruction on x86.</p>
4360 <p>The contents of memory are updated to contain '<tt><value></tt>' at the
4361 location specified by the '<tt><pointer></tt>' operand. If
4362 '<tt><value></tt>' is of scalar type then the number of bytes written
4363 does not exceed the minimum number of bytes needed to hold all bits of the
4364 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
4365 writing a value of a type like <tt>i20</tt> with a size that is not an
4366 integral number of bytes, it is unspecified what happens to the extra bits
4367 that do not belong to the type, but they will typically be overwritten.</p>
4371 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4372 store i32 3, i32* %ptr <i>; yields {void}</i>
4373 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
4378 <!-- _______________________________________________________________________ -->
4379 <div class="doc_subsubsection">
4380 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
4383 <div class="doc_text">
4387 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
4388 <result> = getelementptr inbounds <pty>* <ptrval>{, <ty> <idx>}*
4392 <p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
4393 subelement of an <a href="#t_aggregate">aggregate</a> data structure.
4394 It performs address calculation only and does not access memory.</p>
4397 <p>The first argument is always a pointer, and forms the basis of the
4398 calculation. The remaining arguments are indices that indicate which of the
4399 elements of the aggregate object are indexed. The interpretation of each
4400 index is dependent on the type being indexed into. The first index always
4401 indexes the pointer value given as the first argument, the second index
4402 indexes a value of the type pointed to (not necessarily the value directly
4403 pointed to, since the first index can be non-zero), etc. The first type
4404 indexed into must be a pointer value, subsequent types can be arrays,
4405 vectors, and structs. Note that subsequent types being indexed into
4406 can never be pointers, since that would require loading the pointer before
4407 continuing calculation.</p>
4409 <p>The type of each index argument depends on the type it is indexing into.
4410 When indexing into a (optionally packed) structure, only <tt>i32</tt>
4411 integer <b>constants</b> are allowed. When indexing into an array, pointer
4412 or vector, integers of any width are allowed, and they are not required to be
4415 <p>For example, let's consider a C code fragment and how it gets compiled to
4418 <pre class="doc_code">
4430 int *foo(struct ST *s) {
4431 return &s[1].Z.B[5][13];
4435 <p>The LLVM code generated by the GCC frontend is:</p>
4437 <pre class="doc_code">
4438 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
4439 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
4441 define i32* @foo(%ST* %s) {
4443 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
4449 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
4450 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
4451 }</tt>' type, a structure. The second index indexes into the third element
4452 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
4453 i8 }</tt>' type, another structure. The third index indexes into the second
4454 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
4455 array. The two dimensions of the array are subscripted into, yielding an
4456 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
4457 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
4459 <p>Note that it is perfectly legal to index partially through a structure,
4460 returning a pointer to an inner element. Because of this, the LLVM code for
4461 the given testcase is equivalent to:</p>
4464 define i32* @foo(%ST* %s) {
4465 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
4466 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
4467 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
4468 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
4469 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
4474 <p>If the <tt>inbounds</tt> keyword is present, the result value of the
4475 <tt>getelementptr</tt> is a <a href="#trapvalues">trap value</a> if the
4476 base pointer is not an <i>in bounds</i> address of an allocated object,
4477 or if any of the addresses that would be formed by successive addition of
4478 the offsets implied by the indices to the base address with infinitely
4479 precise arithmetic are not an <i>in bounds</i> address of that allocated
4480 object. The <i>in bounds</i> addresses for an allocated object are all
4481 the addresses that point into the object, plus the address one byte past
4484 <p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
4485 the base address with silently-wrapping two's complement arithmetic, and
4486 the result value of the <tt>getelementptr</tt> may be outside the object
4487 pointed to by the base pointer. The result value may not necessarily be
4488 used to access memory though, even if it happens to point into allocated
4489 storage. See the <a href="#pointeraliasing">Pointer Aliasing Rules</a>
4490 section for more information.</p>
4492 <p>The getelementptr instruction is often confusing. For some more insight into
4493 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
4497 <i>; yields [12 x i8]*:aptr</i>
4498 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
4499 <i>; yields i8*:vptr</i>
4500 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
4501 <i>; yields i8*:eptr</i>
4502 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
4503 <i>; yields i32*:iptr</i>
4504 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
4509 <!-- ======================================================================= -->
4510 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
4513 <div class="doc_text">
4515 <p>The instructions in this category are the conversion instructions (casting)
4516 which all take a single operand and a type. They perform various bit
4517 conversions on the operand.</p>
4521 <!-- _______________________________________________________________________ -->
4522 <div class="doc_subsubsection">
4523 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
4525 <div class="doc_text">
4529 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
4533 <p>The '<tt>trunc</tt>' instruction truncates its operand to the
4534 type <tt>ty2</tt>.</p>
4537 <p>The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
4538 be an <a href="#t_integer">integer</a> type, and a type that specifies the
4539 size and type of the result, which must be
4540 an <a href="#t_integer">integer</a> type. The bit size of <tt>value</tt> must
4541 be larger than the bit size of <tt>ty2</tt>. Equal sized types are not
4545 <p>The '<tt>trunc</tt>' instruction truncates the high order bits
4546 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
4547 source size must be larger than the destination size, <tt>trunc</tt> cannot
4548 be a <i>no-op cast</i>. It will always truncate bits.</p>
4552 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
4553 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
4554 %Z = trunc i32 122 to i1 <i>; yields i1:false</i>
4559 <!-- _______________________________________________________________________ -->
4560 <div class="doc_subsubsection">
4561 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
4563 <div class="doc_text">
4567 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
4571 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
4576 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
4577 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4578 also be of <a href="#t_integer">integer</a> type. The bit size of the
4579 <tt>value</tt> must be smaller than the bit size of the destination type,
4583 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
4584 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
4586 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
4590 %X = zext i32 257 to i64 <i>; yields i64:257</i>
4591 %Y = zext i1 true to i32 <i>; yields i32:1</i>
4596 <!-- _______________________________________________________________________ -->
4597 <div class="doc_subsubsection">
4598 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
4600 <div class="doc_text">
4604 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
4608 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
4611 <p>The '<tt>sext</tt>' instruction takes a value to cast, which must be of
4612 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4613 also be of <a href="#t_integer">integer</a> type. The bit size of the
4614 <tt>value</tt> must be smaller than the bit size of the destination type,
4618 <p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
4619 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
4620 of the type <tt>ty2</tt>.</p>
4622 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
4626 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
4627 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
4632 <!-- _______________________________________________________________________ -->
4633 <div class="doc_subsubsection">
4634 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
4637 <div class="doc_text">
4641 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
4645 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
4649 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
4650 point</a> value to cast and a <a href="#t_floating">floating point</a> type
4651 to cast it to. The size of <tt>value</tt> must be larger than the size of
4652 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
4653 <i>no-op cast</i>.</p>
4656 <p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
4657 <a href="#t_floating">floating point</a> type to a smaller
4658 <a href="#t_floating">floating point</a> type. If the value cannot fit
4659 within the destination type, <tt>ty2</tt>, then the results are
4664 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4665 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4670 <!-- _______________________________________________________________________ -->
4671 <div class="doc_subsubsection">
4672 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
4674 <div class="doc_text">
4678 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
4682 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
4683 floating point value.</p>
4686 <p>The '<tt>fpext</tt>' instruction takes a
4687 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
4688 a <a href="#t_floating">floating point</a> type to cast it to. The source
4689 type must be smaller than the destination type.</p>
4692 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
4693 <a href="#t_floating">floating point</a> type to a larger
4694 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4695 used to make a <i>no-op cast</i> because it always changes bits. Use
4696 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
4700 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
4701 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
4706 <!-- _______________________________________________________________________ -->
4707 <div class="doc_subsubsection">
4708 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
4710 <div class="doc_text">
4714 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
4718 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
4719 unsigned integer equivalent of type <tt>ty2</tt>.</p>
4722 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
4723 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4724 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4725 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4726 vector integer type with the same number of elements as <tt>ty</tt></p>
4729 <p>The '<tt>fptoui</tt>' instruction converts its
4730 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4731 towards zero) unsigned integer value. If the value cannot fit
4732 in <tt>ty2</tt>, the results are undefined.</p>
4736 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
4737 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
4738 %Z = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
4743 <!-- _______________________________________________________________________ -->
4744 <div class="doc_subsubsection">
4745 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
4747 <div class="doc_text">
4751 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
4755 <p>The '<tt>fptosi</tt>' instruction converts
4756 <a href="#t_floating">floating point</a> <tt>value</tt> to
4757 type <tt>ty2</tt>.</p>
4760 <p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
4761 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4762 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4763 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4764 vector integer type with the same number of elements as <tt>ty</tt></p>
4767 <p>The '<tt>fptosi</tt>' instruction converts its
4768 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4769 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4770 the results are undefined.</p>
4774 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
4775 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
4776 %Z = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
4781 <!-- _______________________________________________________________________ -->
4782 <div class="doc_subsubsection">
4783 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
4785 <div class="doc_text">
4789 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4793 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
4794 integer and converts that value to the <tt>ty2</tt> type.</p>
4797 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
4798 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4799 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4800 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4801 floating point type with the same number of elements as <tt>ty</tt></p>
4804 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
4805 integer quantity and converts it to the corresponding floating point
4806 value. If the value cannot fit in the floating point value, the results are
4811 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
4812 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
4817 <!-- _______________________________________________________________________ -->
4818 <div class="doc_subsubsection">
4819 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
4821 <div class="doc_text">
4825 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4829 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
4830 and converts that value to the <tt>ty2</tt> type.</p>
4833 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
4834 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4835 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4836 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4837 floating point type with the same number of elements as <tt>ty</tt></p>
4840 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
4841 quantity and converts it to the corresponding floating point value. If the
4842 value cannot fit in the floating point value, the results are undefined.</p>
4846 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
4847 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
4852 <!-- _______________________________________________________________________ -->
4853 <div class="doc_subsubsection">
4854 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4856 <div class="doc_text">
4860 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
4864 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4865 the integer type <tt>ty2</tt>.</p>
4868 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4869 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
4870 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
4873 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4874 <tt>ty2</tt> by interpreting the pointer value as an integer and either
4875 truncating or zero extending that value to the size of the integer type. If
4876 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4877 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4878 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4883 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4884 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4889 <!-- _______________________________________________________________________ -->
4890 <div class="doc_subsubsection">
4891 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4893 <div class="doc_text">
4897 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
4901 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
4902 pointer type, <tt>ty2</tt>.</p>
4905 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4906 value to cast, and a type to cast it to, which must be a
4907 <a href="#t_pointer">pointer</a> type.</p>
4910 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4911 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
4912 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4913 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
4914 than the size of a pointer then a zero extension is done. If they are the
4915 same size, nothing is done (<i>no-op cast</i>).</p>
4919 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4920 %Y = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4921 %Z = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4926 <!-- _______________________________________________________________________ -->
4927 <div class="doc_subsubsection">
4928 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4930 <div class="doc_text">
4934 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
4938 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4939 <tt>ty2</tt> without changing any bits.</p>
4942 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
4943 non-aggregate first class value, and a type to cast it to, which must also be
4944 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
4945 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
4946 identical. If the source type is a pointer, the destination type must also be
4947 a pointer. This instruction supports bitwise conversion of vectors to
4948 integers and to vectors of other types (as long as they have the same
4952 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4953 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4954 this conversion. The conversion is done as if the <tt>value</tt> had been
4955 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
4956 be converted to other pointer types with this instruction. To convert
4957 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
4958 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4962 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4963 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
4964 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
4969 <!-- ======================================================================= -->
4970 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4972 <div class="doc_text">
4974 <p>The instructions in this category are the "miscellaneous" instructions, which
4975 defy better classification.</p>
4979 <!-- _______________________________________________________________________ -->
4980 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4983 <div class="doc_text">
4987 <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4991 <p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
4992 boolean values based on comparison of its two integer, integer vector, or
4993 pointer operands.</p>
4996 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
4997 the condition code indicating the kind of comparison to perform. It is not a
4998 value, just a keyword. The possible condition code are:</p>
5001 <li><tt>eq</tt>: equal</li>
5002 <li><tt>ne</tt>: not equal </li>
5003 <li><tt>ugt</tt>: unsigned greater than</li>
5004 <li><tt>uge</tt>: unsigned greater or equal</li>
5005 <li><tt>ult</tt>: unsigned less than</li>
5006 <li><tt>ule</tt>: unsigned less or equal</li>
5007 <li><tt>sgt</tt>: signed greater than</li>
5008 <li><tt>sge</tt>: signed greater or equal</li>
5009 <li><tt>slt</tt>: signed less than</li>
5010 <li><tt>sle</tt>: signed less or equal</li>
5013 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
5014 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
5015 typed. They must also be identical types.</p>
5018 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
5019 condition code given as <tt>cond</tt>. The comparison performed always yields
5020 either an <a href="#t_integer"><tt>i1</tt></a> or vector of <tt>i1</tt>
5021 result, as follows:</p>
5024 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
5025 <tt>false</tt> otherwise. No sign interpretation is necessary or
5028 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
5029 <tt>false</tt> otherwise. No sign interpretation is necessary or
5032 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
5033 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5035 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
5036 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5037 to <tt>op2</tt>.</li>
5039 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
5040 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5042 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
5043 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5045 <li><tt>sgt</tt>: interprets the operands as signed values and yields
5046 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5048 <li><tt>sge</tt>: interprets the operands as signed values and yields
5049 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5050 to <tt>op2</tt>.</li>
5052 <li><tt>slt</tt>: interprets the operands as signed values and yields
5053 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5055 <li><tt>sle</tt>: interprets the operands as signed values and yields
5056 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5059 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
5060 values are compared as if they were integers.</p>
5062 <p>If the operands are integer vectors, then they are compared element by
5063 element. The result is an <tt>i1</tt> vector with the same number of elements
5064 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
5068 <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
5069 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
5070 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
5071 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
5072 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
5073 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
5076 <p>Note that the code generator does not yet support vector types with
5077 the <tt>icmp</tt> instruction.</p>
5081 <!-- _______________________________________________________________________ -->
5082 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
5085 <div class="doc_text">
5089 <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
5093 <p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
5094 values based on comparison of its operands.</p>
5096 <p>If the operands are floating point scalars, then the result type is a boolean
5097 (<a href="#t_integer"><tt>i1</tt></a>).</p>
5099 <p>If the operands are floating point vectors, then the result type is a vector
5100 of boolean with the same number of elements as the operands being
5104 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
5105 the condition code indicating the kind of comparison to perform. It is not a
5106 value, just a keyword. The possible condition code are:</p>
5109 <li><tt>false</tt>: no comparison, always returns false</li>
5110 <li><tt>oeq</tt>: ordered and equal</li>
5111 <li><tt>ogt</tt>: ordered and greater than </li>
5112 <li><tt>oge</tt>: ordered and greater than or equal</li>
5113 <li><tt>olt</tt>: ordered and less than </li>
5114 <li><tt>ole</tt>: ordered and less than or equal</li>
5115 <li><tt>one</tt>: ordered and not equal</li>
5116 <li><tt>ord</tt>: ordered (no nans)</li>
5117 <li><tt>ueq</tt>: unordered or equal</li>
5118 <li><tt>ugt</tt>: unordered or greater than </li>
5119 <li><tt>uge</tt>: unordered or greater than or equal</li>
5120 <li><tt>ult</tt>: unordered or less than </li>
5121 <li><tt>ule</tt>: unordered or less than or equal</li>
5122 <li><tt>une</tt>: unordered or not equal</li>
5123 <li><tt>uno</tt>: unordered (either nans)</li>
5124 <li><tt>true</tt>: no comparison, always returns true</li>
5127 <p><i>Ordered</i> means that neither operand is a QNAN while
5128 <i>unordered</i> means that either operand may be a QNAN.</p>
5130 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
5131 a <a href="#t_floating">floating point</a> type or
5132 a <a href="#t_vector">vector</a> of floating point type. They must have
5133 identical types.</p>
5136 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
5137 according to the condition code given as <tt>cond</tt>. If the operands are
5138 vectors, then the vectors are compared element by element. Each comparison
5139 performed always yields an <a href="#t_integer">i1</a> result, as
5143 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
5145 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5146 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5148 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5149 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5151 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5152 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5154 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5155 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5157 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5158 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5160 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5161 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5163 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
5165 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
5166 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5168 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
5169 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5171 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
5172 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5174 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
5175 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5177 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
5178 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5180 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
5181 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5183 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
5185 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
5190 <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
5191 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
5192 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
5193 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
5196 <p>Note that the code generator does not yet support vector types with
5197 the <tt>fcmp</tt> instruction.</p>
5201 <!-- _______________________________________________________________________ -->
5202 <div class="doc_subsubsection">
5203 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
5206 <div class="doc_text">
5210 <result> = phi <ty> [ <val0>, <label0>], ...
5214 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in the
5215 SSA graph representing the function.</p>
5218 <p>The type of the incoming values is specified with the first type field. After
5219 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
5220 one pair for each predecessor basic block of the current block. Only values
5221 of <a href="#t_firstclass">first class</a> type may be used as the value
5222 arguments to the PHI node. Only labels may be used as the label
5225 <p>There must be no non-phi instructions between the start of a basic block and
5226 the PHI instructions: i.e. PHI instructions must be first in a basic
5229 <p>For the purposes of the SSA form, the use of each incoming value is deemed to
5230 occur on the edge from the corresponding predecessor block to the current
5231 block (but after any definition of an '<tt>invoke</tt>' instruction's return
5232 value on the same edge).</p>
5235 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
5236 specified by the pair corresponding to the predecessor basic block that
5237 executed just prior to the current block.</p>
5241 Loop: ; Infinite loop that counts from 0 on up...
5242 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
5243 %nextindvar = add i32 %indvar, 1
5249 <!-- _______________________________________________________________________ -->
5250 <div class="doc_subsubsection">
5251 <a name="i_select">'<tt>select</tt>' Instruction</a>
5254 <div class="doc_text">
5258 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
5260 <i>selty</i> is either i1 or {<N x i1>}
5264 <p>The '<tt>select</tt>' instruction is used to choose one value based on a
5265 condition, without branching.</p>
5269 <p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
5270 values indicating the condition, and two values of the
5271 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
5272 vectors and the condition is a scalar, then entire vectors are selected, not
5273 individual elements.</p>
5276 <p>If the condition is an i1 and it evaluates to 1, the instruction returns the
5277 first value argument; otherwise, it returns the second value argument.</p>
5279 <p>If the condition is a vector of i1, then the value arguments must be vectors
5280 of the same size, and the selection is done element by element.</p>
5284 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
5287 <p>Note that the code generator does not yet support conditions
5288 with vector type.</p>
5292 <!-- _______________________________________________________________________ -->
5293 <div class="doc_subsubsection">
5294 <a name="i_call">'<tt>call</tt>' Instruction</a>
5297 <div class="doc_text">
5301 <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>]
5305 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
5308 <p>This instruction requires several arguments:</p>
5311 <li>The optional "tail" marker indicates that the callee function does not
5312 access any allocas or varargs in the caller. Note that calls may be
5313 marked "tail" even if they do not occur before
5314 a <a href="#i_ret"><tt>ret</tt></a> instruction. If the "tail" marker is
5315 present, the function call is eligible for tail call optimization,
5316 but <a href="CodeGenerator.html#tailcallopt">might not in fact be
5317 optimized into a jump</a>. The code generator may optimize calls marked
5318 "tail" with either 1) automatic <a href="CodeGenerator.html#sibcallopt">
5319 sibling call optimization</a> when the caller and callee have
5320 matching signatures, or 2) forced tail call optimization when the
5321 following extra requirements are met:
5323 <li>Caller and callee both have the calling
5324 convention <tt>fastcc</tt>.</li>
5325 <li>The call is in tail position (ret immediately follows call and ret
5326 uses value of call or is void).</li>
5327 <li>Option <tt>-tailcallopt</tt> is enabled,
5328 or <code>llvm::GuaranteedTailCallOpt</code> is <code>true</code>.</li>
5329 <li><a href="CodeGenerator.html#tailcallopt">Platform specific
5330 constraints are met.</a></li>
5334 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
5335 convention</a> the call should use. If none is specified, the call
5336 defaults to using C calling conventions. The calling convention of the
5337 call must match the calling convention of the target function, or else the
5338 behavior is undefined.</li>
5340 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
5341 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
5342 '<tt>inreg</tt>' attributes are valid here.</li>
5344 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
5345 type of the return value. Functions that return no value are marked
5346 <tt><a href="#t_void">void</a></tt>.</li>
5348 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
5349 being invoked. The argument types must match the types implied by this
5350 signature. This type can be omitted if the function is not varargs and if
5351 the function type does not return a pointer to a function.</li>
5353 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
5354 be invoked. In most cases, this is a direct function invocation, but
5355 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
5356 to function value.</li>
5358 <li>'<tt>function args</tt>': argument list whose types match the function
5359 signature argument types and parameter attributes. All arguments must be
5360 of <a href="#t_firstclass">first class</a> type. If the function
5361 signature indicates the function accepts a variable number of arguments,
5362 the extra arguments can be specified.</li>
5364 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
5365 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
5366 '<tt>readnone</tt>' attributes are valid here.</li>
5370 <p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
5371 a specified function, with its incoming arguments bound to the specified
5372 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
5373 function, control flow continues with the instruction after the function
5374 call, and the return value of the function is bound to the result
5379 %retval = call i32 @test(i32 %argc)
5380 call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42) <i>; yields i32</i>
5381 %X = tail call i32 @foo() <i>; yields i32</i>
5382 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
5383 call void %foo(i8 97 signext)
5385 %struct.A = type { i32, i8 }
5386 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
5387 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
5388 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
5389 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
5390 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
5393 <p>llvm treats calls to some functions with names and arguments that match the
5394 standard C99 library as being the C99 library functions, and may perform
5395 optimizations or generate code for them under that assumption. This is
5396 something we'd like to change in the future to provide better support for
5397 freestanding environments and non-C-based languages.</p>
5401 <!-- _______________________________________________________________________ -->
5402 <div class="doc_subsubsection">
5403 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
5406 <div class="doc_text">
5410 <resultval> = va_arg <va_list*> <arglist>, <argty>
5414 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
5415 the "variable argument" area of a function call. It is used to implement the
5416 <tt>va_arg</tt> macro in C.</p>
5419 <p>This instruction takes a <tt>va_list*</tt> value and the type of the
5420 argument. It returns a value of the specified argument type and increments
5421 the <tt>va_list</tt> to point to the next argument. The actual type
5422 of <tt>va_list</tt> is target specific.</p>
5425 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
5426 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
5427 to the next argument. For more information, see the variable argument
5428 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
5430 <p>It is legal for this instruction to be called in a function which does not
5431 take a variable number of arguments, for example, the <tt>vfprintf</tt>
5434 <p><tt>va_arg</tt> is an LLVM instruction instead of
5435 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
5439 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
5441 <p>Note that the code generator does not yet fully support va_arg on many
5442 targets. Also, it does not currently support va_arg with aggregate types on
5447 <!-- *********************************************************************** -->
5448 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
5449 <!-- *********************************************************************** -->
5451 <div class="doc_text">
5453 <p>LLVM supports the notion of an "intrinsic function". These functions have
5454 well known names and semantics and are required to follow certain
5455 restrictions. Overall, these intrinsics represent an extension mechanism for
5456 the LLVM language that does not require changing all of the transformations
5457 in LLVM when adding to the language (or the bitcode reader/writer, the
5458 parser, etc...).</p>
5460 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
5461 prefix is reserved in LLVM for intrinsic names; thus, function names may not
5462 begin with this prefix. Intrinsic functions must always be external
5463 functions: you cannot define the body of intrinsic functions. Intrinsic
5464 functions may only be used in call or invoke instructions: it is illegal to
5465 take the address of an intrinsic function. Additionally, because intrinsic
5466 functions are part of the LLVM language, it is required if any are added that
5467 they be documented here.</p>
5469 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
5470 family of functions that perform the same operation but on different data
5471 types. Because LLVM can represent over 8 million different integer types,
5472 overloading is used commonly to allow an intrinsic function to operate on any
5473 integer type. One or more of the argument types or the result type can be
5474 overloaded to accept any integer type. Argument types may also be defined as
5475 exactly matching a previous argument's type or the result type. This allows
5476 an intrinsic function which accepts multiple arguments, but needs all of them
5477 to be of the same type, to only be overloaded with respect to a single
5478 argument or the result.</p>
5480 <p>Overloaded intrinsics will have the names of its overloaded argument types
5481 encoded into its function name, each preceded by a period. Only those types
5482 which are overloaded result in a name suffix. Arguments whose type is matched
5483 against another type do not. For example, the <tt>llvm.ctpop</tt> function
5484 can take an integer of any width and returns an integer of exactly the same
5485 integer width. This leads to a family of functions such as
5486 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
5487 %val)</tt>. Only one type, the return type, is overloaded, and only one type
5488 suffix is required. Because the argument's type is matched against the return
5489 type, it does not require its own name suffix.</p>
5491 <p>To learn how to add an intrinsic function, please see the
5492 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
5496 <!-- ======================================================================= -->
5497 <div class="doc_subsection">
5498 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
5501 <div class="doc_text">
5503 <p>Variable argument support is defined in LLVM with
5504 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
5505 intrinsic functions. These functions are related to the similarly named
5506 macros defined in the <tt><stdarg.h></tt> header file.</p>
5508 <p>All of these functions operate on arguments that use a target-specific value
5509 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
5510 not define what this type is, so all transformations should be prepared to
5511 handle these functions regardless of the type used.</p>
5513 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
5514 instruction and the variable argument handling intrinsic functions are
5517 <pre class="doc_code">
5518 define i32 @test(i32 %X, ...) {
5519 ; Initialize variable argument processing
5521 %ap2 = bitcast i8** %ap to i8*
5522 call void @llvm.va_start(i8* %ap2)
5524 ; Read a single integer argument
5525 %tmp = va_arg i8** %ap, i32
5527 ; Demonstrate usage of llvm.va_copy and llvm.va_end
5529 %aq2 = bitcast i8** %aq to i8*
5530 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
5531 call void @llvm.va_end(i8* %aq2)
5533 ; Stop processing of arguments.
5534 call void @llvm.va_end(i8* %ap2)
5538 declare void @llvm.va_start(i8*)
5539 declare void @llvm.va_copy(i8*, i8*)
5540 declare void @llvm.va_end(i8*)
5545 <!-- _______________________________________________________________________ -->
5546 <div class="doc_subsubsection">
5547 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
5551 <div class="doc_text">
5555 declare void %llvm.va_start(i8* <arglist>)
5559 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*<arglist></tt>
5560 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
5563 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
5566 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
5567 macro available in C. In a target-dependent way, it initializes
5568 the <tt>va_list</tt> element to which the argument points, so that the next
5569 call to <tt>va_arg</tt> will produce the first variable argument passed to
5570 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
5571 need to know the last argument of the function as the compiler can figure
5576 <!-- _______________________________________________________________________ -->
5577 <div class="doc_subsubsection">
5578 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
5581 <div class="doc_text">
5585 declare void @llvm.va_end(i8* <arglist>)
5589 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
5590 which has been initialized previously
5591 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
5592 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
5595 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
5598 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
5599 macro available in C. In a target-dependent way, it destroys
5600 the <tt>va_list</tt> element to which the argument points. Calls
5601 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
5602 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
5603 with calls to <tt>llvm.va_end</tt>.</p>
5607 <!-- _______________________________________________________________________ -->
5608 <div class="doc_subsubsection">
5609 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
5612 <div class="doc_text">
5616 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
5620 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
5621 from the source argument list to the destination argument list.</p>
5624 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
5625 The second argument is a pointer to a <tt>va_list</tt> element to copy
5629 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
5630 macro available in C. In a target-dependent way, it copies the
5631 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
5632 element. This intrinsic is necessary because
5633 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
5634 arbitrarily complex and require, for example, memory allocation.</p>
5638 <!-- ======================================================================= -->
5639 <div class="doc_subsection">
5640 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
5643 <div class="doc_text">
5645 <p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
5646 Collection</a> (GC) requires the implementation and generation of these
5647 intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
5648 roots on the stack</a>, as well as garbage collector implementations that
5649 require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
5650 barriers. Front-ends for type-safe garbage collected languages should generate
5651 these intrinsics to make use of the LLVM garbage collectors. For more details,
5652 see <a href="GarbageCollection.html">Accurate Garbage Collection with
5655 <p>The garbage collection intrinsics only operate on objects in the generic
5656 address space (address space zero).</p>
5660 <!-- _______________________________________________________________________ -->
5661 <div class="doc_subsubsection">
5662 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
5665 <div class="doc_text">
5669 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
5673 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
5674 the code generator, and allows some metadata to be associated with it.</p>
5677 <p>The first argument specifies the address of a stack object that contains the
5678 root pointer. The second pointer (which must be either a constant or a
5679 global value address) contains the meta-data to be associated with the
5683 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
5684 location. At compile-time, the code generator generates information to allow
5685 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5686 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5691 <!-- _______________________________________________________________________ -->
5692 <div class="doc_subsubsection">
5693 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
5696 <div class="doc_text">
5700 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
5704 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
5705 locations, allowing garbage collector implementations that require read
5709 <p>The second argument is the address to read from, which should be an address
5710 allocated from the garbage collector. The first object is a pointer to the
5711 start of the referenced object, if needed by the language runtime (otherwise
5715 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
5716 instruction, but may be replaced with substantially more complex code by the
5717 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5718 may only be used in a function which <a href="#gc">specifies a GC
5723 <!-- _______________________________________________________________________ -->
5724 <div class="doc_subsubsection">
5725 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
5728 <div class="doc_text">
5732 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
5736 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
5737 locations, allowing garbage collector implementations that require write
5738 barriers (such as generational or reference counting collectors).</p>
5741 <p>The first argument is the reference to store, the second is the start of the
5742 object to store it to, and the third is the address of the field of Obj to
5743 store to. If the runtime does not require a pointer to the object, Obj may
5747 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
5748 instruction, but may be replaced with substantially more complex code by the
5749 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5750 may only be used in a function which <a href="#gc">specifies a GC
5755 <!-- ======================================================================= -->
5756 <div class="doc_subsection">
5757 <a name="int_codegen">Code Generator Intrinsics</a>
5760 <div class="doc_text">
5762 <p>These intrinsics are provided by LLVM to expose special features that may
5763 only be implemented with code generator support.</p>
5767 <!-- _______________________________________________________________________ -->
5768 <div class="doc_subsubsection">
5769 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
5772 <div class="doc_text">
5776 declare i8 *@llvm.returnaddress(i32 <level>)
5780 <p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5781 target-specific value indicating the return address of the current function
5782 or one of its callers.</p>
5785 <p>The argument to this intrinsic indicates which function to return the address
5786 for. Zero indicates the calling function, one indicates its caller, etc.
5787 The argument is <b>required</b> to be a constant integer value.</p>
5790 <p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
5791 indicating the return address of the specified call frame, or zero if it
5792 cannot be identified. The value returned by this intrinsic is likely to be
5793 incorrect or 0 for arguments other than zero, so it should only be used for
5794 debugging purposes.</p>
5796 <p>Note that calling this intrinsic does not prevent function inlining or other
5797 aggressive transformations, so the value returned may not be that of the
5798 obvious source-language caller.</p>
5802 <!-- _______________________________________________________________________ -->
5803 <div class="doc_subsubsection">
5804 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5807 <div class="doc_text">
5811 declare i8* @llvm.frameaddress(i32 <level>)
5815 <p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5816 target-specific frame pointer value for the specified stack frame.</p>
5819 <p>The argument to this intrinsic indicates which function to return the frame
5820 pointer for. Zero indicates the calling function, one indicates its caller,
5821 etc. The argument is <b>required</b> to be a constant integer value.</p>
5824 <p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
5825 indicating the frame address of the specified call frame, or zero if it
5826 cannot be identified. The value returned by this intrinsic is likely to be
5827 incorrect or 0 for arguments other than zero, so it should only be used for
5828 debugging purposes.</p>
5830 <p>Note that calling this intrinsic does not prevent function inlining or other
5831 aggressive transformations, so the value returned may not be that of the
5832 obvious source-language caller.</p>
5836 <!-- _______________________________________________________________________ -->
5837 <div class="doc_subsubsection">
5838 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5841 <div class="doc_text">
5845 declare i8* @llvm.stacksave()
5849 <p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
5850 of the function stack, for use
5851 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
5852 useful for implementing language features like scoped automatic variable
5853 sized arrays in C99.</p>
5856 <p>This intrinsic returns a opaque pointer value that can be passed
5857 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
5858 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
5859 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
5860 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
5861 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
5862 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
5866 <!-- _______________________________________________________________________ -->
5867 <div class="doc_subsubsection">
5868 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5871 <div class="doc_text">
5875 declare void @llvm.stackrestore(i8* %ptr)
5879 <p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5880 the function stack to the state it was in when the
5881 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
5882 executed. This is useful for implementing language features like scoped
5883 automatic variable sized arrays in C99.</p>
5886 <p>See the description
5887 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
5891 <!-- _______________________________________________________________________ -->
5892 <div class="doc_subsubsection">
5893 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5896 <div class="doc_text">
5900 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
5904 <p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
5905 insert a prefetch instruction if supported; otherwise, it is a noop.
5906 Prefetches have no effect on the behavior of the program but can change its
5907 performance characteristics.</p>
5910 <p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
5911 specifier determining if the fetch should be for a read (0) or write (1),
5912 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5913 locality, to (3) - extremely local keep in cache. The <tt>rw</tt>
5914 and <tt>locality</tt> arguments must be constant integers.</p>
5917 <p>This intrinsic does not modify the behavior of the program. In particular,
5918 prefetches cannot trap and do not produce a value. On targets that support
5919 this intrinsic, the prefetch can provide hints to the processor cache for
5920 better performance.</p>
5924 <!-- _______________________________________________________________________ -->
5925 <div class="doc_subsubsection">
5926 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5929 <div class="doc_text">
5933 declare void @llvm.pcmarker(i32 <id>)
5937 <p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
5938 Counter (PC) in a region of code to simulators and other tools. The method
5939 is target specific, but it is expected that the marker will use exported
5940 symbols to transmit the PC of the marker. The marker makes no guarantees
5941 that it will remain with any specific instruction after optimizations. It is
5942 possible that the presence of a marker will inhibit optimizations. The
5943 intended use is to be inserted after optimizations to allow correlations of
5944 simulation runs.</p>
5947 <p><tt>id</tt> is a numerical id identifying the marker.</p>
5950 <p>This intrinsic does not modify the behavior of the program. Backends that do
5951 not support this intrinsic may ignore it.</p>
5955 <!-- _______________________________________________________________________ -->
5956 <div class="doc_subsubsection">
5957 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5960 <div class="doc_text">
5964 declare i64 @llvm.readcyclecounter()
5968 <p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5969 counter register (or similar low latency, high accuracy clocks) on those
5970 targets that support it. On X86, it should map to RDTSC. On Alpha, it
5971 should map to RPCC. As the backing counters overflow quickly (on the order
5972 of 9 seconds on alpha), this should only be used for small timings.</p>
5975 <p>When directly supported, reading the cycle counter should not modify any
5976 memory. Implementations are allowed to either return a application specific
5977 value or a system wide value. On backends without support, this is lowered
5978 to a constant 0.</p>
5982 <!-- ======================================================================= -->
5983 <div class="doc_subsection">
5984 <a name="int_libc">Standard C Library Intrinsics</a>
5987 <div class="doc_text">
5989 <p>LLVM provides intrinsics for a few important standard C library functions.
5990 These intrinsics allow source-language front-ends to pass information about
5991 the alignment of the pointer arguments to the code generator, providing
5992 opportunity for more efficient code generation.</p>
5996 <!-- _______________________________________________________________________ -->
5997 <div class="doc_subsubsection">
5998 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
6001 <div class="doc_text">
6004 <p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
6005 integer bit width and for different address spaces. Not all targets support
6006 all bit widths however.</p>
6009 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
6010 i32 <len>, i32 <align>, i1 <isvolatile>)
6011 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
6012 i64 <len>, i32 <align>, i1 <isvolatile>)
6016 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6017 source location to the destination location.</p>
6019 <p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
6020 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6021 and the pointers can be in specified address spaces.</p>
6025 <p>The first argument is a pointer to the destination, the second is a pointer
6026 to the source. The third argument is an integer argument specifying the
6027 number of bytes to copy, the fourth argument is the alignment of the
6028 source and destination locations, and the fifth is a boolean indicating a
6029 volatile access.</p>
6031 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6032 then the caller guarantees that both the source and destination pointers are
6033 aligned to that boundary.</p>
6035 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6036 <tt>llvm.memcpy</tt> call is a <a href="#volatile">volatile operation</a>.
6037 The detailed access behavior is not very cleanly specified and it is unwise
6038 to depend on it.</p>
6042 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6043 source location to the destination location, which are not allowed to
6044 overlap. It copies "len" bytes of memory over. If the argument is known to
6045 be aligned to some boundary, this can be specified as the fourth argument,
6046 otherwise it should be set to 0 or 1.</p>
6050 <!-- _______________________________________________________________________ -->
6051 <div class="doc_subsubsection">
6052 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
6055 <div class="doc_text">
6058 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
6059 width and for different address space. Not all targets support all bit
6063 declare void @llvm.memmove.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
6064 i32 <len>, i32 <align>, i1 <isvolatile>)
6065 declare void @llvm.memmove.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
6066 i64 <len>, i32 <align>, i1 <isvolatile>)
6070 <p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
6071 source location to the destination location. It is similar to the
6072 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
6075 <p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
6076 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6077 and the pointers can be in specified address spaces.</p>
6081 <p>The first argument is a pointer to the destination, the second is a pointer
6082 to the source. The third argument is an integer argument specifying the
6083 number of bytes to copy, the fourth argument is the alignment of the
6084 source and destination locations, and the fifth is a boolean indicating a
6085 volatile access.</p>
6087 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6088 then the caller guarantees that the source and destination pointers are
6089 aligned to that boundary.</p>
6091 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6092 <tt>llvm.memmove</tt> call is a <a href="#volatile">volatile operation</a>.
6093 The detailed access behavior is not very cleanly specified and it is unwise
6094 to depend on it.</p>
6098 <p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
6099 source location to the destination location, which may overlap. It copies
6100 "len" bytes of memory over. If the argument is known to be aligned to some
6101 boundary, this can be specified as the fourth argument, otherwise it should
6102 be set to 0 or 1.</p>
6106 <!-- _______________________________________________________________________ -->
6107 <div class="doc_subsubsection">
6108 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
6111 <div class="doc_text">
6114 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
6115 width and for different address spaces. However, not all targets support all
6119 declare void @llvm.memset.p0i8.i32(i8* <dest>, i8 <val>,
6120 i32 <len>, i32 <align>, i1 <isvolatile>)
6121 declare void @llvm.memset.p0i8.i64(i8* <dest>, i8 <val>,
6122 i64 <len>, i32 <align>, i1 <isvolatile>)
6126 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
6127 particular byte value.</p>
6129 <p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
6130 intrinsic does not return a value and takes extra alignment/volatile
6131 arguments. Also, the destination can be in an arbitrary address space.</p>
6134 <p>The first argument is a pointer to the destination to fill, the second is the
6135 byte value with which to fill it, the third argument is an integer argument
6136 specifying the number of bytes to fill, and the fourth argument is the known
6137 alignment of the destination location.</p>
6139 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6140 then the caller guarantees that the destination pointer is aligned to that
6143 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6144 <tt>llvm.memset</tt> call is a <a href="#volatile">volatile operation</a>.
6145 The detailed access behavior is not very cleanly specified and it is unwise
6146 to depend on it.</p>
6149 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
6150 at the destination location. If the argument is known to be aligned to some
6151 boundary, this can be specified as the fourth argument, otherwise it should
6152 be set to 0 or 1.</p>
6156 <!-- _______________________________________________________________________ -->
6157 <div class="doc_subsubsection">
6158 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
6161 <div class="doc_text">
6164 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
6165 floating point or vector of floating point type. Not all targets support all
6169 declare float @llvm.sqrt.f32(float %Val)
6170 declare double @llvm.sqrt.f64(double %Val)
6171 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
6172 declare fp128 @llvm.sqrt.f128(fp128 %Val)
6173 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
6177 <p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
6178 returning the same value as the libm '<tt>sqrt</tt>' functions would.
6179 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
6180 behavior for negative numbers other than -0.0 (which allows for better
6181 optimization, because there is no need to worry about errno being
6182 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
6185 <p>The argument and return value are floating point numbers of the same
6189 <p>This function returns the sqrt of the specified operand if it is a
6190 nonnegative floating point number.</p>
6194 <!-- _______________________________________________________________________ -->
6195 <div class="doc_subsubsection">
6196 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
6199 <div class="doc_text">
6202 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
6203 floating point or vector of floating point type. Not all targets support all
6207 declare float @llvm.powi.f32(float %Val, i32 %power)
6208 declare double @llvm.powi.f64(double %Val, i32 %power)
6209 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
6210 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
6211 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
6215 <p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
6216 specified (positive or negative) power. The order of evaluation of
6217 multiplications is not defined. When a vector of floating point type is
6218 used, the second argument remains a scalar integer value.</p>
6221 <p>The second argument is an integer power, and the first is a value to raise to
6225 <p>This function returns the first value raised to the second power with an
6226 unspecified sequence of rounding operations.</p>
6230 <!-- _______________________________________________________________________ -->
6231 <div class="doc_subsubsection">
6232 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
6235 <div class="doc_text">
6238 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
6239 floating point or vector of floating point type. Not all targets support all
6243 declare float @llvm.sin.f32(float %Val)
6244 declare double @llvm.sin.f64(double %Val)
6245 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
6246 declare fp128 @llvm.sin.f128(fp128 %Val)
6247 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
6251 <p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
6254 <p>The argument and return value are floating point numbers of the same
6258 <p>This function returns the sine of the specified operand, returning the same
6259 values as the libm <tt>sin</tt> functions would, and handles error conditions
6260 in the same way.</p>
6264 <!-- _______________________________________________________________________ -->
6265 <div class="doc_subsubsection">
6266 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
6269 <div class="doc_text">
6272 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
6273 floating point or vector of floating point type. Not all targets support all
6277 declare float @llvm.cos.f32(float %Val)
6278 declare double @llvm.cos.f64(double %Val)
6279 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
6280 declare fp128 @llvm.cos.f128(fp128 %Val)
6281 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
6285 <p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
6288 <p>The argument and return value are floating point numbers of the same
6292 <p>This function returns the cosine of the specified operand, returning the same
6293 values as the libm <tt>cos</tt> functions would, and handles error conditions
6294 in the same way.</p>
6298 <!-- _______________________________________________________________________ -->
6299 <div class="doc_subsubsection">
6300 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
6303 <div class="doc_text">
6306 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
6307 floating point or vector of floating point type. Not all targets support all
6311 declare float @llvm.pow.f32(float %Val, float %Power)
6312 declare double @llvm.pow.f64(double %Val, double %Power)
6313 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
6314 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
6315 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
6319 <p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
6320 specified (positive or negative) power.</p>
6323 <p>The second argument is a floating point power, and the first is a value to
6324 raise to that power.</p>
6327 <p>This function returns the first value raised to the second power, returning
6328 the same values as the libm <tt>pow</tt> functions would, and handles error
6329 conditions in the same way.</p>
6333 <!-- ======================================================================= -->
6334 <div class="doc_subsection">
6335 <a name="int_manip">Bit Manipulation Intrinsics</a>
6338 <div class="doc_text">
6340 <p>LLVM provides intrinsics for a few important bit manipulation operations.
6341 These allow efficient code generation for some algorithms.</p>
6345 <!-- _______________________________________________________________________ -->
6346 <div class="doc_subsubsection">
6347 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
6350 <div class="doc_text">
6353 <p>This is an overloaded intrinsic function. You can use bswap on any integer
6354 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
6357 declare i16 @llvm.bswap.i16(i16 <id>)
6358 declare i32 @llvm.bswap.i32(i32 <id>)
6359 declare i64 @llvm.bswap.i64(i64 <id>)
6363 <p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
6364 values with an even number of bytes (positive multiple of 16 bits). These
6365 are useful for performing operations on data that is not in the target's
6366 native byte order.</p>
6369 <p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
6370 and low byte of the input i16 swapped. Similarly,
6371 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
6372 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
6373 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
6374 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
6375 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
6376 more, respectively).</p>
6380 <!-- _______________________________________________________________________ -->
6381 <div class="doc_subsubsection">
6382 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
6385 <div class="doc_text">
6388 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
6389 width. Not all targets support all bit widths however.</p>
6392 declare i8 @llvm.ctpop.i8(i8 <src>)
6393 declare i16 @llvm.ctpop.i16(i16 <src>)
6394 declare i32 @llvm.ctpop.i32(i32 <src>)
6395 declare i64 @llvm.ctpop.i64(i64 <src>)
6396 declare i256 @llvm.ctpop.i256(i256 <src>)
6400 <p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
6404 <p>The only argument is the value to be counted. The argument may be of any
6405 integer type. The return type must match the argument type.</p>
6408 <p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.</p>
6412 <!-- _______________________________________________________________________ -->
6413 <div class="doc_subsubsection">
6414 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
6417 <div class="doc_text">
6420 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
6421 integer bit width. Not all targets support all bit widths however.</p>
6424 declare i8 @llvm.ctlz.i8 (i8 <src>)
6425 declare i16 @llvm.ctlz.i16(i16 <src>)
6426 declare i32 @llvm.ctlz.i32(i32 <src>)
6427 declare i64 @llvm.ctlz.i64(i64 <src>)
6428 declare i256 @llvm.ctlz.i256(i256 <src>)
6432 <p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
6433 leading zeros in a variable.</p>
6436 <p>The only argument is the value to be counted. The argument may be of any
6437 integer type. The return type must match the argument type.</p>
6440 <p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
6441 zeros in a variable. If the src == 0 then the result is the size in bits of
6442 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
6446 <!-- _______________________________________________________________________ -->
6447 <div class="doc_subsubsection">
6448 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
6451 <div class="doc_text">
6454 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
6455 integer bit width. Not all targets support all bit widths however.</p>
6458 declare i8 @llvm.cttz.i8 (i8 <src>)
6459 declare i16 @llvm.cttz.i16(i16 <src>)
6460 declare i32 @llvm.cttz.i32(i32 <src>)
6461 declare i64 @llvm.cttz.i64(i64 <src>)
6462 declare i256 @llvm.cttz.i256(i256 <src>)
6466 <p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
6470 <p>The only argument is the value to be counted. The argument may be of any
6471 integer type. The return type must match the argument type.</p>
6474 <p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
6475 zeros in a variable. If the src == 0 then the result is the size in bits of
6476 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
6480 <!-- ======================================================================= -->
6481 <div class="doc_subsection">
6482 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
6485 <div class="doc_text">
6487 <p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
6491 <!-- _______________________________________________________________________ -->
6492 <div class="doc_subsubsection">
6493 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
6496 <div class="doc_text">
6499 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
6500 on any integer bit width.</p>
6503 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
6504 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6505 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6509 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6510 a signed addition of the two arguments, and indicate whether an overflow
6511 occurred during the signed summation.</p>
6514 <p>The arguments (%a and %b) and the first element of the result structure may
6515 be of integer types of any bit width, but they must have the same bit
6516 width. The second element of the result structure must be of
6517 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6518 undergo signed addition.</p>
6521 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6522 a signed addition of the two variables. They return a structure — the
6523 first element of which is the signed summation, and the second element of
6524 which is a bit specifying if the signed summation resulted in an
6529 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6530 %sum = extractvalue {i32, i1} %res, 0
6531 %obit = extractvalue {i32, i1} %res, 1
6532 br i1 %obit, label %overflow, label %normal
6537 <!-- _______________________________________________________________________ -->
6538 <div class="doc_subsubsection">
6539 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
6542 <div class="doc_text">
6545 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
6546 on any integer bit width.</p>
6549 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6550 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6551 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6555 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6556 an unsigned addition of the two arguments, and indicate whether a carry
6557 occurred during the unsigned summation.</p>
6560 <p>The arguments (%a and %b) and the first element of the result structure may
6561 be of integer types of any bit width, but they must have the same bit
6562 width. The second element of the result structure must be of
6563 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6564 undergo unsigned addition.</p>
6567 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6568 an unsigned addition of the two arguments. They return a structure —
6569 the first element of which is the sum, and the second element of which is a
6570 bit specifying if the unsigned summation resulted in a carry.</p>
6574 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6575 %sum = extractvalue {i32, i1} %res, 0
6576 %obit = extractvalue {i32, i1} %res, 1
6577 br i1 %obit, label %carry, label %normal
6582 <!-- _______________________________________________________________________ -->
6583 <div class="doc_subsubsection">
6584 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
6587 <div class="doc_text">
6590 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
6591 on any integer bit width.</p>
6594 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6595 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6596 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6600 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6601 a signed subtraction of the two arguments, and indicate whether an overflow
6602 occurred during the signed subtraction.</p>
6605 <p>The arguments (%a and %b) and the first element of the result structure may
6606 be of integer types of any bit width, but they must have the same bit
6607 width. The second element of the result structure must be of
6608 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6609 undergo signed subtraction.</p>
6612 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6613 a signed subtraction of the two arguments. They return a structure —
6614 the first element of which is the subtraction, and the second element of
6615 which is a bit specifying if the signed subtraction resulted in an
6620 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6621 %sum = extractvalue {i32, i1} %res, 0
6622 %obit = extractvalue {i32, i1} %res, 1
6623 br i1 %obit, label %overflow, label %normal
6628 <!-- _______________________________________________________________________ -->
6629 <div class="doc_subsubsection">
6630 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
6633 <div class="doc_text">
6636 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
6637 on any integer bit width.</p>
6640 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6641 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6642 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6646 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6647 an unsigned subtraction of the two arguments, and indicate whether an
6648 overflow occurred during the unsigned subtraction.</p>
6651 <p>The arguments (%a and %b) and the first element of the result structure may
6652 be of integer types of any bit width, but they must have the same bit
6653 width. The second element of the result structure must be of
6654 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6655 undergo unsigned subtraction.</p>
6658 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6659 an unsigned subtraction of the two arguments. They return a structure —
6660 the first element of which is the subtraction, and the second element of
6661 which is a bit specifying if the unsigned subtraction resulted in an
6666 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6667 %sum = extractvalue {i32, i1} %res, 0
6668 %obit = extractvalue {i32, i1} %res, 1
6669 br i1 %obit, label %overflow, label %normal
6674 <!-- _______________________________________________________________________ -->
6675 <div class="doc_subsubsection">
6676 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
6679 <div class="doc_text">
6682 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
6683 on any integer bit width.</p>
6686 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6687 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6688 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6693 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6694 a signed multiplication of the two arguments, and indicate whether an
6695 overflow occurred during the signed multiplication.</p>
6698 <p>The arguments (%a and %b) and the first element of the result structure may
6699 be of integer types of any bit width, but they must have the same bit
6700 width. The second element of the result structure must be of
6701 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6702 undergo signed multiplication.</p>
6705 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6706 a signed multiplication of the two arguments. They return a structure —
6707 the first element of which is the multiplication, and the second element of
6708 which is a bit specifying if the signed multiplication resulted in an
6713 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6714 %sum = extractvalue {i32, i1} %res, 0
6715 %obit = extractvalue {i32, i1} %res, 1
6716 br i1 %obit, label %overflow, label %normal
6721 <!-- _______________________________________________________________________ -->
6722 <div class="doc_subsubsection">
6723 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6726 <div class="doc_text">
6729 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6730 on any integer bit width.</p>
6733 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6734 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6735 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6739 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6740 a unsigned multiplication of the two arguments, and indicate whether an
6741 overflow occurred during the unsigned multiplication.</p>
6744 <p>The arguments (%a and %b) and the first element of the result structure may
6745 be of integer types of any bit width, but they must have the same bit
6746 width. The second element of the result structure must be of
6747 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6748 undergo unsigned multiplication.</p>
6751 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6752 an unsigned multiplication of the two arguments. They return a structure
6753 — the first element of which is the multiplication, and the second
6754 element of which is a bit specifying if the unsigned multiplication resulted
6759 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6760 %sum = extractvalue {i32, i1} %res, 0
6761 %obit = extractvalue {i32, i1} %res, 1
6762 br i1 %obit, label %overflow, label %normal
6767 <!-- ======================================================================= -->
6768 <div class="doc_subsection">
6769 <a name="int_fp16">Half Precision Floating Point Intrinsics</a>
6772 <div class="doc_text">
6774 <p>Half precision floating point is a storage-only format. This means that it is
6775 a dense encoding (in memory) but does not support computation in the
6778 <p>This means that code must first load the half-precision floating point
6779 value as an i16, then convert it to float with <a
6780 href="#int_convert_from_fp16"><tt>llvm.convert.from.fp16</tt></a>.
6781 Computation can then be performed on the float value (including extending to
6782 double etc). To store the value back to memory, it is first converted to
6783 float if needed, then converted to i16 with
6784 <a href="#int_convert_to_fp16"><tt>llvm.convert.to.fp16</tt></a>, then
6785 storing as an i16 value.</p>
6788 <!-- _______________________________________________________________________ -->
6789 <div class="doc_subsubsection">
6790 <a name="int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a>
6793 <div class="doc_text">
6797 declare i16 @llvm.convert.to.fp16(f32 %a)
6801 <p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
6802 a conversion from single precision floating point format to half precision
6803 floating point format.</p>
6806 <p>The intrinsic function contains single argument - the value to be
6810 <p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
6811 a conversion from single precision floating point format to half precision
6812 floating point format. The return value is an <tt>i16</tt> which
6813 contains the converted number.</p>
6817 %res = call i16 @llvm.convert.to.fp16(f32 %a)
6818 store i16 %res, i16* @x, align 2
6823 <!-- _______________________________________________________________________ -->
6824 <div class="doc_subsubsection">
6825 <a name="int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a>
6828 <div class="doc_text">
6832 declare f32 @llvm.convert.from.fp16(i16 %a)
6836 <p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs
6837 a conversion from half precision floating point format to single precision
6838 floating point format.</p>
6841 <p>The intrinsic function contains single argument - the value to be
6845 <p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs a
6846 conversion from half single precision floating point format to single
6847 precision floating point format. The input half-float value is represented by
6848 an <tt>i16</tt> value.</p>
6852 %a = load i16* @x, align 2
6853 %res = call f32 @llvm.convert.from.fp16(i16 %a)
6858 <!-- ======================================================================= -->
6859 <div class="doc_subsection">
6860 <a name="int_debugger">Debugger Intrinsics</a>
6863 <div class="doc_text">
6865 <p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
6866 prefix), are described in
6867 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
6868 Level Debugging</a> document.</p>
6872 <!-- ======================================================================= -->
6873 <div class="doc_subsection">
6874 <a name="int_eh">Exception Handling Intrinsics</a>
6877 <div class="doc_text">
6879 <p>The LLVM exception handling intrinsics (which all start with
6880 <tt>llvm.eh.</tt> prefix), are described in
6881 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6882 Handling</a> document.</p>
6886 <!-- ======================================================================= -->
6887 <div class="doc_subsection">
6888 <a name="int_trampoline">Trampoline Intrinsic</a>
6891 <div class="doc_text">
6893 <p>This intrinsic makes it possible to excise one parameter, marked with
6894 the <a href="#nest"><tt>nest</tt></a> attribute, from a function.
6895 The result is a callable
6896 function pointer lacking the nest parameter - the caller does not need to
6897 provide a value for it. Instead, the value to use is stored in advance in a
6898 "trampoline", a block of memory usually allocated on the stack, which also
6899 contains code to splice the nest value into the argument list. This is used
6900 to implement the GCC nested function address extension.</p>
6902 <p>For example, if the function is
6903 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
6904 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
6907 <pre class="doc_code">
6908 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6909 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6910 %p = call i8* @llvm.init.trampoline(i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval)
6911 %fp = bitcast i8* %p to i32 (i32, i32)*
6914 <p>The call <tt>%val = call i32 %fp(i32 %x, i32 %y)</tt> is then equivalent
6915 to <tt>%val = call i32 %f(i8* %nval, i32 %x, i32 %y)</tt>.</p>
6919 <!-- _______________________________________________________________________ -->
6920 <div class="doc_subsubsection">
6921 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6924 <div class="doc_text">
6928 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
6932 <p>This fills the memory pointed to by <tt>tramp</tt> with code and returns a
6933 function pointer suitable for executing it.</p>
6936 <p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6937 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
6938 sufficiently aligned block of memory; this memory is written to by the
6939 intrinsic. Note that the size and the alignment are target-specific - LLVM
6940 currently provides no portable way of determining them, so a front-end that
6941 generates this intrinsic needs to have some target-specific knowledge.
6942 The <tt>func</tt> argument must hold a function bitcast to
6943 an <tt>i8*</tt>.</p>
6946 <p>The block of memory pointed to by <tt>tramp</tt> is filled with target
6947 dependent code, turning it into a function. A pointer to this function is
6948 returned, but needs to be bitcast to an <a href="#int_trampoline">appropriate
6949 function pointer type</a> before being called. The new function's signature
6950 is the same as that of <tt>func</tt> with any arguments marked with
6951 the <tt>nest</tt> attribute removed. At most one such <tt>nest</tt> argument
6952 is allowed, and it must be of pointer type. Calling the new function is
6953 equivalent to calling <tt>func</tt> with the same argument list, but
6954 with <tt>nval</tt> used for the missing <tt>nest</tt> argument. If, after
6955 calling <tt>llvm.init.trampoline</tt>, the memory pointed to
6956 by <tt>tramp</tt> is modified, then the effect of any later call to the
6957 returned function pointer is undefined.</p>
6961 <!-- ======================================================================= -->
6962 <div class="doc_subsection">
6963 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6966 <div class="doc_text">
6968 <p>These intrinsic functions expand the "universal IR" of LLVM to represent
6969 hardware constructs for atomic operations and memory synchronization. This
6970 provides an interface to the hardware, not an interface to the programmer. It
6971 is aimed at a low enough level to allow any programming models or APIs
6972 (Application Programming Interfaces) which need atomic behaviors to map
6973 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
6974 hardware provides a "universal IR" for source languages, it also provides a
6975 starting point for developing a "universal" atomic operation and
6976 synchronization IR.</p>
6978 <p>These do <em>not</em> form an API such as high-level threading libraries,
6979 software transaction memory systems, atomic primitives, and intrinsic
6980 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6981 application libraries. The hardware interface provided by LLVM should allow
6982 a clean implementation of all of these APIs and parallel programming models.
6983 No one model or paradigm should be selected above others unless the hardware
6984 itself ubiquitously does so.</p>
6988 <!-- _______________________________________________________________________ -->
6989 <div class="doc_subsubsection">
6990 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6992 <div class="doc_text">
6995 declare void @llvm.memory.barrier(i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>, i1 <device>)
6999 <p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
7000 specific pairs of memory access types.</p>
7003 <p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
7004 The first four arguments enables a specific barrier as listed below. The
7005 fifth argument specifies that the barrier applies to io or device or uncached
7009 <li><tt>ll</tt>: load-load barrier</li>
7010 <li><tt>ls</tt>: load-store barrier</li>
7011 <li><tt>sl</tt>: store-load barrier</li>
7012 <li><tt>ss</tt>: store-store barrier</li>
7013 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
7017 <p>This intrinsic causes the system to enforce some ordering constraints upon
7018 the loads and stores of the program. This barrier does not
7019 indicate <em>when</em> any events will occur, it only enforces
7020 an <em>order</em> in which they occur. For any of the specified pairs of load
7021 and store operations (f.ex. load-load, or store-load), all of the first
7022 operations preceding the barrier will complete before any of the second
7023 operations succeeding the barrier begin. Specifically the semantics for each
7024 pairing is as follows:</p>
7027 <li><tt>ll</tt>: All loads before the barrier must complete before any load
7028 after the barrier begins.</li>
7029 <li><tt>ls</tt>: All loads before the barrier must complete before any
7030 store after the barrier begins.</li>
7031 <li><tt>ss</tt>: All stores before the barrier must complete before any
7032 store after the barrier begins.</li>
7033 <li><tt>sl</tt>: All stores before the barrier must complete before any
7034 load after the barrier begins.</li>
7037 <p>These semantics are applied with a logical "and" behavior when more than one
7038 is enabled in a single memory barrier intrinsic.</p>
7040 <p>Backends may implement stronger barriers than those requested when they do
7041 not support as fine grained a barrier as requested. Some architectures do
7042 not need all types of barriers and on such architectures, these become
7047 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7048 %ptr = bitcast i8* %mallocP to i32*
7051 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
7052 call void @llvm.memory.barrier(i1 false, i1 true, i1 false, i1 false)
7053 <i>; guarantee the above finishes</i>
7054 store i32 8, %ptr <i>; before this begins</i>
7059 <!-- _______________________________________________________________________ -->
7060 <div class="doc_subsubsection">
7061 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
7064 <div class="doc_text">
7067 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
7068 any integer bit width and for different address spaces. Not all targets
7069 support all bit widths however.</p>
7072 declare i8 @llvm.atomic.cmp.swap.i8.p0i8(i8* <ptr>, i8 <cmp>, i8 <val>)
7073 declare i16 @llvm.atomic.cmp.swap.i16.p0i16(i16* <ptr>, i16 <cmp>, i16 <val>)
7074 declare i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* <ptr>, i32 <cmp>, i32 <val>)
7075 declare i64 @llvm.atomic.cmp.swap.i64.p0i64(i64* <ptr>, i64 <cmp>, i64 <val>)
7079 <p>This loads a value in memory and compares it to a given value. If they are
7080 equal, it stores a new value into the memory.</p>
7083 <p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
7084 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
7085 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
7086 this integer type. While any bit width integer may be used, targets may only
7087 lower representations they support in hardware.</p>
7090 <p>This entire intrinsic must be executed atomically. It first loads the value
7091 in memory pointed to by <tt>ptr</tt> and compares it with the
7092 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
7093 memory. The loaded value is yielded in all cases. This provides the
7094 equivalent of an atomic compare-and-swap operation within the SSA
7099 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7100 %ptr = bitcast i8* %mallocP to i32*
7103 %val1 = add i32 4, 4
7104 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 4, %val1)
7105 <i>; yields {i32}:result1 = 4</i>
7106 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7107 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7109 %val2 = add i32 1, 1
7110 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 5, %val2)
7111 <i>; yields {i32}:result2 = 8</i>
7112 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
7114 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
7119 <!-- _______________________________________________________________________ -->
7120 <div class="doc_subsubsection">
7121 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
7123 <div class="doc_text">
7126 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
7127 integer bit width. Not all targets support all bit widths however.</p>
7130 declare i8 @llvm.atomic.swap.i8.p0i8(i8* <ptr>, i8 <val>)
7131 declare i16 @llvm.atomic.swap.i16.p0i16(i16* <ptr>, i16 <val>)
7132 declare i32 @llvm.atomic.swap.i32.p0i32(i32* <ptr>, i32 <val>)
7133 declare i64 @llvm.atomic.swap.i64.p0i64(i64* <ptr>, i64 <val>)
7137 <p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
7138 the value from memory. It then stores the value in <tt>val</tt> in the memory
7139 at <tt>ptr</tt>.</p>
7142 <p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
7143 the <tt>val</tt> argument and the result must be integers of the same bit
7144 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
7145 integer type. The targets may only lower integer representations they
7149 <p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
7150 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
7151 equivalent of an atomic swap operation within the SSA framework.</p>
7155 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7156 %ptr = bitcast i8* %mallocP to i32*
7159 %val1 = add i32 4, 4
7160 %result1 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val1)
7161 <i>; yields {i32}:result1 = 4</i>
7162 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7163 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7165 %val2 = add i32 1, 1
7166 %result2 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val2)
7167 <i>; yields {i32}:result2 = 8</i>
7169 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
7170 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
7175 <!-- _______________________________________________________________________ -->
7176 <div class="doc_subsubsection">
7177 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
7181 <div class="doc_text">
7184 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
7185 any integer bit width. Not all targets support all bit widths however.</p>
7188 declare i8 @llvm.atomic.load.add.i8.p0i8(i8* <ptr>, i8 <delta>)
7189 declare i16 @llvm.atomic.load.add.i16.p0i16(i16* <ptr>, i16 <delta>)
7190 declare i32 @llvm.atomic.load.add.i32.p0i32(i32* <ptr>, i32 <delta>)
7191 declare i64 @llvm.atomic.load.add.i64.p0i64(i64* <ptr>, i64 <delta>)
7195 <p>This intrinsic adds <tt>delta</tt> to the value stored in memory
7196 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7199 <p>The intrinsic takes two arguments, the first a pointer to an integer value
7200 and the second an integer value. The result is also an integer value. These
7201 integer types can have any bit width, but they must all have the same bit
7202 width. The targets may only lower integer representations they support.</p>
7205 <p>This intrinsic does a series of operations atomically. It first loads the
7206 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
7207 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
7211 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7212 %ptr = bitcast i8* %mallocP to i32*
7214 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 4)
7215 <i>; yields {i32}:result1 = 4</i>
7216 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 2)
7217 <i>; yields {i32}:result2 = 8</i>
7218 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 5)
7219 <i>; yields {i32}:result3 = 10</i>
7220 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
7225 <!-- _______________________________________________________________________ -->
7226 <div class="doc_subsubsection">
7227 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
7231 <div class="doc_text">
7234 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
7235 any integer bit width and for different address spaces. Not all targets
7236 support all bit widths however.</p>
7239 declare i8 @llvm.atomic.load.sub.i8.p0i32(i8* <ptr>, i8 <delta>)
7240 declare i16 @llvm.atomic.load.sub.i16.p0i32(i16* <ptr>, i16 <delta>)
7241 declare i32 @llvm.atomic.load.sub.i32.p0i32(i32* <ptr>, i32 <delta>)
7242 declare i64 @llvm.atomic.load.sub.i64.p0i32(i64* <ptr>, i64 <delta>)
7246 <p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
7247 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7250 <p>The intrinsic takes two arguments, the first a pointer to an integer value
7251 and the second an integer value. The result is also an integer value. These
7252 integer types can have any bit width, but they must all have the same bit
7253 width. The targets may only lower integer representations they support.</p>
7256 <p>This intrinsic does a series of operations atomically. It first loads the
7257 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
7258 result to <tt>ptr</tt>. It yields the original value stored
7259 at <tt>ptr</tt>.</p>
7263 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7264 %ptr = bitcast i8* %mallocP to i32*
7266 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 4)
7267 <i>; yields {i32}:result1 = 8</i>
7268 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 2)
7269 <i>; yields {i32}:result2 = 4</i>
7270 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 5)
7271 <i>; yields {i32}:result3 = 2</i>
7272 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
7277 <!-- _______________________________________________________________________ -->
7278 <div class="doc_subsubsection">
7279 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
7280 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
7281 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
7282 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
7285 <div class="doc_text">
7288 <p>These are overloaded intrinsics. You can
7289 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
7290 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
7291 bit width and for different address spaces. Not all targets support all bit
7295 declare i8 @llvm.atomic.load.and.i8.p0i8(i8* <ptr>, i8 <delta>)
7296 declare i16 @llvm.atomic.load.and.i16.p0i16(i16* <ptr>, i16 <delta>)
7297 declare i32 @llvm.atomic.load.and.i32.p0i32(i32* <ptr>, i32 <delta>)
7298 declare i64 @llvm.atomic.load.and.i64.p0i64(i64* <ptr>, i64 <delta>)
7302 declare i8 @llvm.atomic.load.or.i8.p0i8(i8* <ptr>, i8 <delta>)
7303 declare i16 @llvm.atomic.load.or.i16.p0i16(i16* <ptr>, i16 <delta>)
7304 declare i32 @llvm.atomic.load.or.i32.p0i32(i32* <ptr>, i32 <delta>)
7305 declare i64 @llvm.atomic.load.or.i64.p0i64(i64* <ptr>, i64 <delta>)
7309 declare i8 @llvm.atomic.load.nand.i8.p0i32(i8* <ptr>, i8 <delta>)
7310 declare i16 @llvm.atomic.load.nand.i16.p0i32(i16* <ptr>, i16 <delta>)
7311 declare i32 @llvm.atomic.load.nand.i32.p0i32(i32* <ptr>, i32 <delta>)
7312 declare i64 @llvm.atomic.load.nand.i64.p0i32(i64* <ptr>, i64 <delta>)
7316 declare i8 @llvm.atomic.load.xor.i8.p0i32(i8* <ptr>, i8 <delta>)
7317 declare i16 @llvm.atomic.load.xor.i16.p0i32(i16* <ptr>, i16 <delta>)
7318 declare i32 @llvm.atomic.load.xor.i32.p0i32(i32* <ptr>, i32 <delta>)
7319 declare i64 @llvm.atomic.load.xor.i64.p0i32(i64* <ptr>, i64 <delta>)
7323 <p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
7324 the value stored in memory at <tt>ptr</tt>. It yields the original value
7325 at <tt>ptr</tt>.</p>
7328 <p>These intrinsics take two arguments, the first a pointer to an integer value
7329 and the second an integer value. The result is also an integer value. These
7330 integer types can have any bit width, but they must all have the same bit
7331 width. The targets may only lower integer representations they support.</p>
7334 <p>These intrinsics does a series of operations atomically. They first load the
7335 value stored at <tt>ptr</tt>. They then do the bitwise
7336 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
7337 original value stored at <tt>ptr</tt>.</p>
7341 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7342 %ptr = bitcast i8* %mallocP to i32*
7343 store i32 0x0F0F, %ptr
7344 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32(i32* %ptr, i32 0xFF)
7345 <i>; yields {i32}:result0 = 0x0F0F</i>
7346 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32(i32* %ptr, i32 0xFF)
7347 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
7348 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32(i32* %ptr, i32 0F)
7349 <i>; yields {i32}:result2 = 0xF0</i>
7350 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32(i32* %ptr, i32 0F)
7351 <i>; yields {i32}:result3 = FF</i>
7352 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
7357 <!-- _______________________________________________________________________ -->
7358 <div class="doc_subsubsection">
7359 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
7360 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
7361 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
7362 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
7365 <div class="doc_text">
7368 <p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
7369 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
7370 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
7371 address spaces. Not all targets support all bit widths however.</p>
7374 declare i8 @llvm.atomic.load.max.i8.p0i8(i8* <ptr>, i8 <delta>)
7375 declare i16 @llvm.atomic.load.max.i16.p0i16(i16* <ptr>, i16 <delta>)
7376 declare i32 @llvm.atomic.load.max.i32.p0i32(i32* <ptr>, i32 <delta>)
7377 declare i64 @llvm.atomic.load.max.i64.p0i64(i64* <ptr>, i64 <delta>)
7381 declare i8 @llvm.atomic.load.min.i8.p0i8(i8* <ptr>, i8 <delta>)
7382 declare i16 @llvm.atomic.load.min.i16.p0i16(i16* <ptr>, i16 <delta>)
7383 declare i32 @llvm.atomic.load.min.i32.p0i32(i32* <ptr>, i32 <delta>)
7384 declare i64 @llvm.atomic.load.min.i64.p0i64(i64* <ptr>, i64 <delta>)
7388 declare i8 @llvm.atomic.load.umax.i8.p0i8(i8* <ptr>, i8 <delta>)
7389 declare i16 @llvm.atomic.load.umax.i16.p0i16(i16* <ptr>, i16 <delta>)
7390 declare i32 @llvm.atomic.load.umax.i32.p0i32(i32* <ptr>, i32 <delta>)
7391 declare i64 @llvm.atomic.load.umax.i64.p0i64(i64* <ptr>, i64 <delta>)
7395 declare i8 @llvm.atomic.load.umin.i8.p0i8(i8* <ptr>, i8 <delta>)
7396 declare i16 @llvm.atomic.load.umin.i16.p0i16(i16* <ptr>, i16 <delta>)
7397 declare i32 @llvm.atomic.load.umin.i32.p0i32(i32* <ptr>, i32 <delta>)
7398 declare i64 @llvm.atomic.load.umin.i64.p0i64(i64* <ptr>, i64 <delta>)
7402 <p>These intrinsics takes the signed or unsigned minimum or maximum of
7403 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
7404 original value at <tt>ptr</tt>.</p>
7407 <p>These intrinsics take two arguments, the first a pointer to an integer value
7408 and the second an integer value. The result is also an integer value. These
7409 integer types can have any bit width, but they must all have the same bit
7410 width. The targets may only lower integer representations they support.</p>
7413 <p>These intrinsics does a series of operations atomically. They first load the
7414 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
7415 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
7416 yield the original value stored at <tt>ptr</tt>.</p>
7420 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7421 %ptr = bitcast i8* %mallocP to i32*
7423 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32(i32* %ptr, i32 -2)
7424 <i>; yields {i32}:result0 = 7</i>
7425 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32(i32* %ptr, i32 8)
7426 <i>; yields {i32}:result1 = -2</i>
7427 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32(i32* %ptr, i32 10)
7428 <i>; yields {i32}:result2 = 8</i>
7429 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32(i32* %ptr, i32 30)
7430 <i>; yields {i32}:result3 = 8</i>
7431 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
7437 <!-- ======================================================================= -->
7438 <div class="doc_subsection">
7439 <a name="int_memorymarkers">Memory Use Markers</a>
7442 <div class="doc_text">
7444 <p>This class of intrinsics exists to information about the lifetime of memory
7445 objects and ranges where variables are immutable.</p>
7449 <!-- _______________________________________________________________________ -->
7450 <div class="doc_subsubsection">
7451 <a name="int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a>
7454 <div class="doc_text">
7458 declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
7462 <p>The '<tt>llvm.lifetime.start</tt>' intrinsic specifies the start of a memory
7463 object's lifetime.</p>
7466 <p>The first argument is a constant integer representing the size of the
7467 object, or -1 if it is variable sized. The second argument is a pointer to
7471 <p>This intrinsic indicates that before this point in the code, the value of the
7472 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7473 never be used and has an undefined value. A load from the pointer that
7474 precedes this intrinsic can be replaced with
7475 <tt>'<a href="#undefvalues">undef</a>'</tt>.</p>
7479 <!-- _______________________________________________________________________ -->
7480 <div class="doc_subsubsection">
7481 <a name="int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a>
7484 <div class="doc_text">
7488 declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
7492 <p>The '<tt>llvm.lifetime.end</tt>' intrinsic specifies the end of a memory
7493 object's lifetime.</p>
7496 <p>The first argument is a constant integer representing the size of the
7497 object, or -1 if it is variable sized. The second argument is a pointer to
7501 <p>This intrinsic indicates that after this point in the code, the value of the
7502 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7503 never be used and has an undefined value. Any stores into the memory object
7504 following this intrinsic may be removed as dead.
7508 <!-- _______________________________________________________________________ -->
7509 <div class="doc_subsubsection">
7510 <a name="int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a>
7513 <div class="doc_text">
7517 declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>)
7521 <p>The '<tt>llvm.invariant.start</tt>' intrinsic specifies that the contents of
7522 a memory object will not change.</p>
7525 <p>The first argument is a constant integer representing the size of the
7526 object, or -1 if it is variable sized. The second argument is a pointer to
7530 <p>This intrinsic indicates that until an <tt>llvm.invariant.end</tt> that uses
7531 the return value, the referenced memory location is constant and
7536 <!-- _______________________________________________________________________ -->
7537 <div class="doc_subsubsection">
7538 <a name="int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a>
7541 <div class="doc_text">
7545 declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
7549 <p>The '<tt>llvm.invariant.end</tt>' intrinsic specifies that the contents of
7550 a memory object are mutable.</p>
7553 <p>The first argument is the matching <tt>llvm.invariant.start</tt> intrinsic.
7554 The second argument is a constant integer representing the size of the
7555 object, or -1 if it is variable sized and the third argument is a pointer
7559 <p>This intrinsic indicates that the memory is mutable again.</p>
7563 <!-- ======================================================================= -->
7564 <div class="doc_subsection">
7565 <a name="int_general">General Intrinsics</a>
7568 <div class="doc_text">
7570 <p>This class of intrinsics is designed to be generic and has no specific
7575 <!-- _______________________________________________________________________ -->
7576 <div class="doc_subsubsection">
7577 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
7580 <div class="doc_text">
7584 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int>)
7588 <p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
7591 <p>The first argument is a pointer to a value, the second is a pointer to a
7592 global string, the third is a pointer to a global string which is the source
7593 file name, and the last argument is the line number.</p>
7596 <p>This intrinsic allows annotation of local variables with arbitrary strings.
7597 This can be useful for special purpose optimizations that want to look for
7598 these annotations. These have no other defined use, they are ignored by code
7599 generation and optimization.</p>
7603 <!-- _______________________________________________________________________ -->
7604 <div class="doc_subsubsection">
7605 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
7608 <div class="doc_text">
7611 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
7612 any integer bit width.</p>
7615 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int>)
7616 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int>)
7617 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int>)
7618 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int>)
7619 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int>)
7623 <p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
7626 <p>The first argument is an integer value (result of some expression), the
7627 second is a pointer to a global string, the third is a pointer to a global
7628 string which is the source file name, and the last argument is the line
7629 number. It returns the value of the first argument.</p>
7632 <p>This intrinsic allows annotations to be put on arbitrary expressions with
7633 arbitrary strings. This can be useful for special purpose optimizations that
7634 want to look for these annotations. These have no other defined use, they
7635 are ignored by code generation and optimization.</p>
7639 <!-- _______________________________________________________________________ -->
7640 <div class="doc_subsubsection">
7641 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
7644 <div class="doc_text">
7648 declare void @llvm.trap()
7652 <p>The '<tt>llvm.trap</tt>' intrinsic.</p>
7658 <p>This intrinsics is lowered to the target dependent trap instruction. If the
7659 target does not have a trap instruction, this intrinsic will be lowered to
7660 the call of the <tt>abort()</tt> function.</p>
7664 <!-- _______________________________________________________________________ -->
7665 <div class="doc_subsubsection">
7666 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
7669 <div class="doc_text">
7673 declare void @llvm.stackprotector(i8* <guard>, i8** <slot>)
7677 <p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
7678 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
7679 ensure that it is placed on the stack before local variables.</p>
7682 <p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
7683 arguments. The first argument is the value loaded from the stack
7684 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
7685 that has enough space to hold the value of the guard.</p>
7688 <p>This intrinsic causes the prologue/epilogue inserter to force the position of
7689 the <tt>AllocaInst</tt> stack slot to be before local variables on the
7690 stack. This is to ensure that if a local variable on the stack is
7691 overwritten, it will destroy the value of the guard. When the function exits,
7692 the guard on the stack is checked against the original guard. If they are
7693 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
7698 <!-- _______________________________________________________________________ -->
7699 <div class="doc_subsubsection">
7700 <a name="int_objectsize">'<tt>llvm.objectsize</tt>' Intrinsic</a>
7703 <div class="doc_text">
7707 declare i32 @llvm.objectsize.i32(i8* <object>, i1 <type>)
7708 declare i64 @llvm.objectsize.i64(i8* <object>, i1 <type>)
7712 <p>The <tt>llvm.objectsize</tt> intrinsic is designed to provide information to
7713 the optimizers to determine at compile time whether a) an operation (like
7714 memcpy) will overflow a buffer that corresponds to an object, or b) that a
7715 runtime check for overflow isn't necessary. An object in this context means
7716 an allocation of a specific class, structure, array, or other object.</p>
7719 <p>The <tt>llvm.objectsize</tt> intrinsic takes two arguments. The first
7720 argument is a pointer to or into the <tt>object</tt>. The second argument
7721 is a boolean 0 or 1. This argument determines whether you want the
7722 maximum (0) or minimum (1) bytes remaining. This needs to be a literal 0 or
7723 1, variables are not allowed.</p>
7726 <p>The <tt>llvm.objectsize</tt> intrinsic is lowered to either a constant
7727 representing the size of the object concerned, or <tt>i32/i64 -1 or 0</tt>,
7728 depending on the <tt>type</tt> argument, if the size cannot be determined at
7733 <!-- *********************************************************************** -->
7736 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
7737 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
7738 <a href="http://validator.w3.org/check/referer"><img
7739 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
7741 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
7742 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
7743 Last modified: $Date$