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_internal">'<tt>internal</tt>' Linkage</a></li>
29 <li><a href="#linkage_available_externally">'<tt>available_externally</tt>' Linkage</a></li>
30 <li><a href="#linkage_linkonce">'<tt>linkonce</tt>' Linkage</a></li>
31 <li><a href="#linkage_common">'<tt>common</tt>' Linkage</a></li>
32 <li><a href="#linkage_weak">'<tt>weak</tt>' Linkage</a></li>
33 <li><a href="#linkage_appending">'<tt>appending</tt>' Linkage</a></li>
34 <li><a href="#linkage_externweak">'<tt>extern_weak</tt>' Linkage</a></li>
35 <li><a href="#linkage_linkonce_odr">'<tt>linkonce_odr</tt>' Linkage</a></li>
36 <li><a href="#linkage_weak">'<tt>weak_odr</tt>' Linkage</a></li>
37 <li><a href="#linkage_external">'<tt>externally visible</tt>' Linkage</a></li>
38 <li><a href="#linkage_dllimport">'<tt>dllimport</tt>' Linkage</a></li>
39 <li><a href="#linkage_dllexport">'<tt>dllexport</tt>' Linkage</a></li>
42 <li><a href="#callingconv">Calling Conventions</a></li>
43 <li><a href="#namedtypes">Named Types</a></li>
44 <li><a href="#globalvars">Global Variables</a></li>
45 <li><a href="#functionstructure">Functions</a></li>
46 <li><a href="#aliasstructure">Aliases</a></li>
47 <li><a href="#namedmetadatastructure">Named Metadata</a></li>
48 <li><a href="#paramattrs">Parameter Attributes</a></li>
49 <li><a href="#fnattrs">Function Attributes</a></li>
50 <li><a href="#gc">Garbage Collector Names</a></li>
51 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
52 <li><a href="#datalayout">Data Layout</a></li>
53 <li><a href="#pointeraliasing">Pointer Aliasing Rules</a></li>
54 <li><a href="#volatile">Volatile Memory Accesses</a></li>
57 <li><a href="#typesystem">Type System</a>
59 <li><a href="#t_classifications">Type Classifications</a></li>
60 <li><a href="#t_primitive">Primitive Types</a>
62 <li><a href="#t_integer">Integer Type</a></li>
63 <li><a href="#t_floating">Floating Point Types</a></li>
64 <li><a href="#t_void">Void Type</a></li>
65 <li><a href="#t_label">Label Type</a></li>
66 <li><a href="#t_metadata">Metadata Type</a></li>
69 <li><a href="#t_derived">Derived Types</a>
71 <li><a href="#t_aggregate">Aggregate Types</a>
73 <li><a href="#t_array">Array Type</a></li>
74 <li><a href="#t_struct">Structure Type</a></li>
75 <li><a href="#t_pstruct">Packed Structure Type</a></li>
76 <li><a href="#t_union">Union Type</a></li>
77 <li><a href="#t_vector">Vector Type</a></li>
80 <li><a href="#t_function">Function Type</a></li>
81 <li><a href="#t_pointer">Pointer Type</a></li>
82 <li><a href="#t_opaque">Opaque Type</a></li>
85 <li><a href="#t_uprefs">Type Up-references</a></li>
88 <li><a href="#constants">Constants</a>
90 <li><a href="#simpleconstants">Simple Constants</a></li>
91 <li><a href="#complexconstants">Complex Constants</a></li>
92 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
93 <li><a href="#undefvalues">Undefined Values</a></li>
94 <li><a href="#trapvalues">Trap Values</a></li>
95 <li><a href="#blockaddress">Addresses of Basic Blocks</a></li>
96 <li><a href="#constantexprs">Constant Expressions</a></li>
99 <li><a href="#othervalues">Other Values</a>
101 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
102 <li><a href="#metadata">Metadata Nodes and Metadata Strings</a></li>
105 <li><a href="#intrinsic_globals">Intrinsic Global Variables</a>
107 <li><a href="#intg_used">The '<tt>llvm.used</tt>' Global Variable</a></li>
108 <li><a href="#intg_compiler_used">The '<tt>llvm.compiler.used</tt>'
109 Global Variable</a></li>
110 <li><a href="#intg_global_ctors">The '<tt>llvm.global_ctors</tt>'
111 Global Variable</a></li>
112 <li><a href="#intg_global_dtors">The '<tt>llvm.global_dtors</tt>'
113 Global Variable</a></li>
116 <li><a href="#instref">Instruction Reference</a>
118 <li><a href="#terminators">Terminator Instructions</a>
120 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
121 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
122 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
123 <li><a href="#i_indirectbr">'<tt>indirectbr</tt>' Instruction</a></li>
124 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
125 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
126 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
129 <li><a href="#binaryops">Binary Operations</a>
131 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
132 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
133 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
134 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
135 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
136 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
137 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
138 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
139 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
140 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
141 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
142 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
145 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
147 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
148 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
149 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
150 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
151 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
152 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
155 <li><a href="#vectorops">Vector Operations</a>
157 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
158 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
159 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
162 <li><a href="#aggregateops">Aggregate Operations</a>
164 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
165 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
168 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
170 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
171 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
172 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
173 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
176 <li><a href="#convertops">Conversion Operations</a>
178 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
179 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
180 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
181 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
182 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
183 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
184 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
185 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
186 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
187 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
188 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
189 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
192 <li><a href="#otherops">Other Operations</a>
194 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
195 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
196 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
197 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
198 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
199 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
204 <li><a href="#intrinsics">Intrinsic Functions</a>
206 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
208 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
209 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
210 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
213 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
215 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
216 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
217 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
220 <li><a href="#int_codegen">Code Generator Intrinsics</a>
222 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
223 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
224 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
225 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
226 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
227 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
228 <li><a href="#int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
231 <li><a href="#int_libc">Standard C Library Intrinsics</a>
233 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
234 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
235 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
236 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
237 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
238 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
239 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
240 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
243 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
245 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
246 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
247 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
248 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
251 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
253 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
254 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
255 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
256 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
257 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
258 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
261 <li><a href="#int_fp16">Half Precision Floating Point Intrinsics</a>
263 <li><a href="#int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a></li>
264 <li><a href="#int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a></li>
267 <li><a href="#int_debugger">Debugger intrinsics</a></li>
268 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
269 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
271 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
274 <li><a href="#int_atomics">Atomic intrinsics</a>
276 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
277 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
278 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
279 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
280 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
281 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
282 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
283 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
284 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
285 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
286 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
287 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
288 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
291 <li><a href="#int_memorymarkers">Memory Use Markers</a>
293 <li><a href="#int_lifetime_start"><tt>llvm.lifetime.start</tt></a></li>
294 <li><a href="#int_lifetime_end"><tt>llvm.lifetime.end</tt></a></li>
295 <li><a href="#int_invariant_start"><tt>llvm.invariant.start</tt></a></li>
296 <li><a href="#int_invariant_end"><tt>llvm.invariant.end</tt></a></li>
299 <li><a href="#int_general">General intrinsics</a>
301 <li><a href="#int_var_annotation">
302 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
303 <li><a href="#int_annotation">
304 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
305 <li><a href="#int_trap">
306 '<tt>llvm.trap</tt>' Intrinsic</a></li>
307 <li><a href="#int_stackprotector">
308 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
309 <li><a href="#int_objectsize">
310 '<tt>llvm.objectsize</tt>' Intrinsic</a></li>
317 <div class="doc_author">
318 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
319 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
322 <!-- *********************************************************************** -->
323 <div class="doc_section"> <a name="abstract">Abstract </a></div>
324 <!-- *********************************************************************** -->
326 <div class="doc_text">
328 <p>This document is a reference manual for the LLVM assembly language. LLVM is
329 a Static Single Assignment (SSA) based representation that provides type
330 safety, low-level operations, flexibility, and the capability of representing
331 'all' high-level languages cleanly. It is the common code representation
332 used throughout all phases of the LLVM compilation strategy.</p>
336 <!-- *********************************************************************** -->
337 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
338 <!-- *********************************************************************** -->
340 <div class="doc_text">
342 <p>The LLVM code representation is designed to be used in three different forms:
343 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
344 for fast loading by a Just-In-Time compiler), and as a human readable
345 assembly language representation. This allows LLVM to provide a powerful
346 intermediate representation for efficient compiler transformations and
347 analysis, while providing a natural means to debug and visualize the
348 transformations. The three different forms of LLVM are all equivalent. This
349 document describes the human readable representation and notation.</p>
351 <p>The LLVM representation aims to be light-weight and low-level while being
352 expressive, typed, and extensible at the same time. It aims to be a
353 "universal IR" of sorts, by being at a low enough level that high-level ideas
354 may be cleanly mapped to it (similar to how microprocessors are "universal
355 IR's", allowing many source languages to be mapped to them). By providing
356 type information, LLVM can be used as the target of optimizations: for
357 example, through pointer analysis, it can be proven that a C automatic
358 variable is never accessed outside of the current function, allowing it to
359 be promoted to a simple SSA value instead of a memory location.</p>
363 <!-- _______________________________________________________________________ -->
364 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
366 <div class="doc_text">
368 <p>It is important to note that this document describes 'well formed' LLVM
369 assembly language. There is a difference between what the parser accepts and
370 what is considered 'well formed'. For example, the following instruction is
371 syntactically okay, but not well formed:</p>
373 <pre class="doc_code">
374 %x = <a href="#i_add">add</a> i32 1, %x
377 <p>because the definition of <tt>%x</tt> does not dominate all of its uses. The
378 LLVM infrastructure provides a verification pass that may be used to verify
379 that an LLVM module is well formed. This pass is automatically run by the
380 parser after parsing input assembly and by the optimizer before it outputs
381 bitcode. The violations pointed out by the verifier pass indicate bugs in
382 transformation passes or input to the parser.</p>
386 <!-- Describe the typesetting conventions here. -->
388 <!-- *********************************************************************** -->
389 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
390 <!-- *********************************************************************** -->
392 <div class="doc_text">
394 <p>LLVM identifiers come in two basic types: global and local. Global
395 identifiers (functions, global variables) begin with the <tt>'@'</tt>
396 character. Local identifiers (register names, types) begin with
397 the <tt>'%'</tt> character. Additionally, there are three different formats
398 for identifiers, for different purposes:</p>
401 <li>Named values are represented as a string of characters with their prefix.
402 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
403 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
404 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
405 other characters in their names can be surrounded with quotes. Special
406 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
407 ASCII code for the character in hexadecimal. In this way, any character
408 can be used in a name value, even quotes themselves.</li>
410 <li>Unnamed values are represented as an unsigned numeric value with their
411 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
413 <li>Constants, which are described in a <a href="#constants">section about
414 constants</a>, below.</li>
417 <p>LLVM requires that values start with a prefix for two reasons: Compilers
418 don't need to worry about name clashes with reserved words, and the set of
419 reserved words may be expanded in the future without penalty. Additionally,
420 unnamed identifiers allow a compiler to quickly come up with a temporary
421 variable without having to avoid symbol table conflicts.</p>
423 <p>Reserved words in LLVM are very similar to reserved words in other
424 languages. There are keywords for different opcodes
425 ('<tt><a href="#i_add">add</a></tt>',
426 '<tt><a href="#i_bitcast">bitcast</a></tt>',
427 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
428 ('<tt><a href="#t_void">void</a></tt>',
429 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
430 reserved words cannot conflict with variable names, because none of them
431 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
433 <p>Here is an example of LLVM code to multiply the integer variable
434 '<tt>%X</tt>' by 8:</p>
438 <pre class="doc_code">
439 %result = <a href="#i_mul">mul</a> i32 %X, 8
442 <p>After strength reduction:</p>
444 <pre class="doc_code">
445 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
448 <p>And the hard way:</p>
450 <pre class="doc_code">
451 %0 = <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
452 %1 = <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
453 %result = <a href="#i_add">add</a> i32 %1, %1
456 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
457 lexical features of LLVM:</p>
460 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
463 <li>Unnamed temporaries are created when the result of a computation is not
464 assigned to a named value.</li>
466 <li>Unnamed temporaries are numbered sequentially</li>
469 <p>It also shows a convention that we follow in this document. When
470 demonstrating instructions, we will follow an instruction with a comment that
471 defines the type and name of value produced. Comments are shown in italic
476 <!-- *********************************************************************** -->
477 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
478 <!-- *********************************************************************** -->
480 <!-- ======================================================================= -->
481 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
484 <div class="doc_text">
486 <p>LLVM programs are composed of "Module"s, each of which is a translation unit
487 of the input programs. Each module consists of functions, global variables,
488 and symbol table entries. Modules may be combined together with the LLVM
489 linker, which merges function (and global variable) definitions, resolves
490 forward declarations, and merges symbol table entries. Here is an example of
491 the "hello world" module:</p>
493 <pre class="doc_code">
494 <i>; Declare the string constant as a global constant.</i>
495 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
497 <i>; External declaration of the puts function</i>
498 <a href="#functionstructure">declare</a> i32 @puts(i8*) <i>; i32 (i8*)* </i>
500 <i>; Definition of main function</i>
501 define i32 @main() { <i>; i32()* </i>
502 <i>; Convert [13 x i8]* to i8 *...</i>
503 %cast210 = <a href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8*</i>
505 <i>; Call puts function to write out the string to stdout.</i>
506 <a href="#i_call">call</a> i32 @puts(i8* %cast210) <i>; i32</i>
507 <a href="#i_ret">ret</a> i32 0<br>}
509 <i>; Named metadata</i>
510 !1 = metadata !{i32 41}
514 <p>This example is made up of a <a href="#globalvars">global variable</a> named
515 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function,
516 a <a href="#functionstructure">function definition</a> for
517 "<tt>main</tt>" and <a href="#namedmetadatastructure">named metadata</a>
520 <p>In general, a module is made up of a list of global values, where both
521 functions and global variables are global values. Global values are
522 represented by a pointer to a memory location (in this case, a pointer to an
523 array of char, and a pointer to a function), and have one of the
524 following <a href="#linkage">linkage types</a>.</p>
528 <!-- ======================================================================= -->
529 <div class="doc_subsection">
530 <a name="linkage">Linkage Types</a>
533 <div class="doc_text">
535 <p>All Global Variables and Functions have one of the following types of
539 <dt><tt><b><a name="linkage_private">private</a></b></tt></dt>
540 <dd>Global values with "<tt>private</tt>" linkage are only directly accessible
541 by objects in the current module. In particular, linking code into a
542 module with an private global value may cause the private to be renamed as
543 necessary to avoid collisions. Because the symbol is private to the
544 module, all references can be updated. This doesn't show up in any symbol
545 table in the object file.</dd>
547 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt></dt>
548 <dd>Similar to <tt>private</tt>, but the symbol is passed through the
549 assembler and evaluated by the linker. Unlike normal strong symbols, they
550 are removed by the linker from the final linked image (executable or
551 dynamic library).</dd>
553 <dt><tt><b><a name="linkage_linker_private_weak">linker_private_weak</a></b></tt></dt>
554 <dd>Similar to "<tt>linker_private</tt>", but the symbol is weak. Note that
555 <tt>linker_private_weak</tt> symbols are subject to coalescing by the
556 linker. The symbols are removed by the linker from the final linked image
557 (executable or dynamic library).</dd>
559 <dt><tt><b><a name="linkage_internal">internal</a></b></tt></dt>
560 <dd>Similar to private, but the value shows as a local symbol
561 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
562 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
564 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt></dt>
565 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
566 into the object file corresponding to the LLVM module. They exist to
567 allow inlining and other optimizations to take place given knowledge of
568 the definition of the global, which is known to be somewhere outside the
569 module. Globals with <tt>available_externally</tt> linkage are allowed to
570 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
571 This linkage type is only allowed on definitions, not declarations.</dd>
573 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt></dt>
574 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
575 the same name when linkage occurs. This can be used to implement
576 some forms of inline functions, templates, or other code which must be
577 generated in each translation unit that uses it, but where the body may
578 be overridden with a more definitive definition later. Unreferenced
579 <tt>linkonce</tt> globals are allowed to be discarded. Note that
580 <tt>linkonce</tt> linkage does not actually allow the optimizer to
581 inline the body of this function into callers because it doesn't know if
582 this definition of the function is the definitive definition within the
583 program or whether it will be overridden by a stronger definition.
584 To enable inlining and other optimizations, use "<tt>linkonce_odr</tt>"
587 <dt><tt><b><a name="linkage_weak">weak</a></b></tt></dt>
588 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
589 <tt>linkonce</tt> linkage, except that unreferenced globals with
590 <tt>weak</tt> linkage may not be discarded. This is used for globals that
591 are declared "weak" in C source code.</dd>
593 <dt><tt><b><a name="linkage_common">common</a></b></tt></dt>
594 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
595 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
597 Symbols with "<tt>common</tt>" linkage are merged in the same way as
598 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
599 <tt>common</tt> symbols may not have an explicit section,
600 must have a zero initializer, and may not be marked '<a
601 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
602 have common linkage.</dd>
605 <dt><tt><b><a name="linkage_appending">appending</a></b></tt></dt>
606 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
607 pointer to array type. When two global variables with appending linkage
608 are linked together, the two global arrays are appended together. This is
609 the LLVM, typesafe, equivalent of having the system linker append together
610 "sections" with identical names when .o files are linked.</dd>
612 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt></dt>
613 <dd>The semantics of this linkage follow the ELF object file model: the symbol
614 is weak until linked, if not linked, the symbol becomes null instead of
615 being an undefined reference.</dd>
617 <dt><tt><b><a name="linkage_linkonce_odr">linkonce_odr</a></b></tt></dt>
618 <dt><tt><b><a name="linkage_weak_odr">weak_odr</a></b></tt></dt>
619 <dd>Some languages allow differing globals to be merged, such as two functions
620 with different semantics. Other languages, such as <tt>C++</tt>, ensure
621 that only equivalent globals are ever merged (the "one definition rule"
622 — "ODR"). Such languages can use the <tt>linkonce_odr</tt>
623 and <tt>weak_odr</tt> linkage types to indicate that the global will only
624 be merged with equivalent globals. These linkage types are otherwise the
625 same as their non-<tt>odr</tt> versions.</dd>
627 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
628 <dd>If none of the above identifiers are used, the global is externally
629 visible, meaning that it participates in linkage and can be used to
630 resolve external symbol references.</dd>
633 <p>The next two types of linkage are targeted for Microsoft Windows platform
634 only. They are designed to support importing (exporting) symbols from (to)
635 DLLs (Dynamic Link Libraries).</p>
638 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt></dt>
639 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
640 or variable via a global pointer to a pointer that is set up by the DLL
641 exporting the symbol. On Microsoft Windows targets, the pointer name is
642 formed by combining <code>__imp_</code> and the function or variable
645 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt></dt>
646 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
647 pointer to a pointer in a DLL, so that it can be referenced with the
648 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
649 name is formed by combining <code>__imp_</code> and the function or
653 <p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
654 another module defined a "<tt>.LC0</tt>" variable and was linked with this
655 one, one of the two would be renamed, preventing a collision. Since
656 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
657 declarations), they are accessible outside of the current module.</p>
659 <p>It is illegal for a function <i>declaration</i> to have any linkage type
660 other than "externally visible", <tt>dllimport</tt>
661 or <tt>extern_weak</tt>.</p>
663 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
664 or <tt>weak_odr</tt> linkages.</p>
668 <!-- ======================================================================= -->
669 <div class="doc_subsection">
670 <a name="callingconv">Calling Conventions</a>
673 <div class="doc_text">
675 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
676 and <a href="#i_invoke">invokes</a> can all have an optional calling
677 convention specified for the call. The calling convention of any pair of
678 dynamic caller/callee must match, or the behavior of the program is
679 undefined. The following calling conventions are supported by LLVM, and more
680 may be added in the future:</p>
683 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
684 <dd>This calling convention (the default if no other calling convention is
685 specified) matches the target C calling conventions. This calling
686 convention supports varargs function calls and tolerates some mismatch in
687 the declared prototype and implemented declaration of the function (as
690 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
691 <dd>This calling convention attempts to make calls as fast as possible
692 (e.g. by passing things in registers). This calling convention allows the
693 target to use whatever tricks it wants to produce fast code for the
694 target, without having to conform to an externally specified ABI
695 (Application Binary Interface).
696 <a href="CodeGenerator.html#tailcallopt">Tail calls can only be optimized
697 when this or the GHC convention is used.</a> This calling convention
698 does not support varargs and requires the prototype of all callees to
699 exactly match the prototype of the function definition.</dd>
701 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
702 <dd>This calling convention attempts to make code in the caller as efficient
703 as possible under the assumption that the call is not commonly executed.
704 As such, these calls often preserve all registers so that the call does
705 not break any live ranges in the caller side. This calling convention
706 does not support varargs and requires the prototype of all callees to
707 exactly match the prototype of the function definition.</dd>
709 <dt><b>"<tt>cc <em>10</em></tt>" - GHC convention</b>:</dt>
710 <dd>This calling convention has been implemented specifically for use by the
711 <a href="http://www.haskell.org/ghc">Glasgow Haskell Compiler (GHC)</a>.
712 It passes everything in registers, going to extremes to achieve this by
713 disabling callee save registers. This calling convention should not be
714 used lightly but only for specific situations such as an alternative to
715 the <em>register pinning</em> performance technique often used when
716 implementing functional programming languages.At the moment only X86
717 supports this convention and it has the following limitations:
719 <li>On <em>X86-32</em> only supports up to 4 bit type parameters. No
720 floating point types are supported.</li>
721 <li>On <em>X86-64</em> only supports up to 10 bit type parameters and
722 6 floating point parameters.</li>
724 This calling convention supports
725 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> but
726 requires both the caller and callee are using it.
729 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
730 <dd>Any calling convention may be specified by number, allowing
731 target-specific calling conventions to be used. Target specific calling
732 conventions start at 64.</dd>
735 <p>More calling conventions can be added/defined on an as-needed basis, to
736 support Pascal conventions or any other well-known target-independent
741 <!-- ======================================================================= -->
742 <div class="doc_subsection">
743 <a name="visibility">Visibility Styles</a>
746 <div class="doc_text">
748 <p>All Global Variables and Functions have one of the following visibility
752 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
753 <dd>On targets that use the ELF object file format, default visibility means
754 that the declaration is visible to other modules and, in shared libraries,
755 means that the declared entity may be overridden. On Darwin, default
756 visibility means that the declaration is visible to other modules. Default
757 visibility corresponds to "external linkage" in the language.</dd>
759 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
760 <dd>Two declarations of an object with hidden visibility refer to the same
761 object if they are in the same shared object. Usually, hidden visibility
762 indicates that the symbol will not be placed into the dynamic symbol
763 table, so no other module (executable or shared library) can reference it
766 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
767 <dd>On ELF, protected visibility indicates that the symbol will be placed in
768 the dynamic symbol table, but that references within the defining module
769 will bind to the local symbol. That is, the symbol cannot be overridden by
775 <!-- ======================================================================= -->
776 <div class="doc_subsection">
777 <a name="namedtypes">Named Types</a>
780 <div class="doc_text">
782 <p>LLVM IR allows you to specify name aliases for certain types. This can make
783 it easier to read the IR and make the IR more condensed (particularly when
784 recursive types are involved). An example of a name specification is:</p>
786 <pre class="doc_code">
787 %mytype = type { %mytype*, i32 }
790 <p>You may give a name to any <a href="#typesystem">type</a> except
791 "<a href="t_void">void</a>". Type name aliases may be used anywhere a type
792 is expected with the syntax "%mytype".</p>
794 <p>Note that type names are aliases for the structural type that they indicate,
795 and that you can therefore specify multiple names for the same type. This
796 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
797 uses structural typing, the name is not part of the type. When printing out
798 LLVM IR, the printer will pick <em>one name</em> to render all types of a
799 particular shape. This means that if you have code where two different
800 source types end up having the same LLVM type, that the dumper will sometimes
801 print the "wrong" or unexpected type. This is an important design point and
802 isn't going to change.</p>
806 <!-- ======================================================================= -->
807 <div class="doc_subsection">
808 <a name="globalvars">Global Variables</a>
811 <div class="doc_text">
813 <p>Global variables define regions of memory allocated at compilation time
814 instead of run-time. Global variables may optionally be initialized, may
815 have an explicit section to be placed in, and may have an optional explicit
816 alignment specified. A variable may be defined as "thread_local", which
817 means that it will not be shared by threads (each thread will have a
818 separated copy of the variable). A variable may be defined as a global
819 "constant," which indicates that the contents of the variable
820 will <b>never</b> be modified (enabling better optimization, allowing the
821 global data to be placed in the read-only section of an executable, etc).
822 Note that variables that need runtime initialization cannot be marked
823 "constant" as there is a store to the variable.</p>
825 <p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
826 constant, even if the final definition of the global is not. This capability
827 can be used to enable slightly better optimization of the program, but
828 requires the language definition to guarantee that optimizations based on the
829 'constantness' are valid for the translation units that do not include the
832 <p>As SSA values, global variables define pointer values that are in scope
833 (i.e. they dominate) all basic blocks in the program. Global variables
834 always define a pointer to their "content" type because they describe a
835 region of memory, and all memory objects in LLVM are accessed through
838 <p>A global variable may be declared to reside in a target-specific numbered
839 address space. For targets that support them, address spaces may affect how
840 optimizations are performed and/or what target instructions are used to
841 access the variable. The default address space is zero. The address space
842 qualifier must precede any other attributes.</p>
844 <p>LLVM allows an explicit section to be specified for globals. If the target
845 supports it, it will emit globals to the section specified.</p>
847 <p>An explicit alignment may be specified for a global, which must be a power
848 of 2. If not present, or if the alignment is set to zero, the alignment of
849 the global is set by the target to whatever it feels convenient. If an
850 explicit alignment is specified, the global is forced to have exactly that
851 alignment. Targets and optimizers are not allowed to over-align the global
852 if the global has an assigned section. In this case, the extra alignment
853 could be observable: for example, code could assume that the globals are
854 densely packed in their section and try to iterate over them as an array,
855 alignment padding would break this iteration.</p>
857 <p>For example, the following defines a global in a numbered address space with
858 an initializer, section, and alignment:</p>
860 <pre class="doc_code">
861 @G = addrspace(5) constant float 1.0, section "foo", align 4
867 <!-- ======================================================================= -->
868 <div class="doc_subsection">
869 <a name="functionstructure">Functions</a>
872 <div class="doc_text">
874 <p>LLVM function definitions consist of the "<tt>define</tt>" keyword, an
875 optional <a href="#linkage">linkage type</a>, an optional
876 <a href="#visibility">visibility style</a>, an optional
877 <a href="#callingconv">calling convention</a>, a return type, an optional
878 <a href="#paramattrs">parameter attribute</a> for the return type, a function
879 name, a (possibly empty) argument list (each with optional
880 <a href="#paramattrs">parameter attributes</a>), optional
881 <a href="#fnattrs">function attributes</a>, an optional section, an optional
882 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
883 curly brace, a list of basic blocks, and a closing curly brace.</p>
885 <p>LLVM function declarations consist of the "<tt>declare</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 list of arguments, an optional alignment, and an
891 optional <a href="#gc">garbage collector name</a>.</p>
893 <p>A function definition contains a list of basic blocks, forming the CFG
894 (Control Flow Graph) for the function. Each basic block may optionally start
895 with a label (giving the basic block a symbol table entry), contains a list
896 of instructions, and ends with a <a href="#terminators">terminator</a>
897 instruction (such as a branch or function return).</p>
899 <p>The first basic block in a function is special in two ways: it is immediately
900 executed on entrance to the function, and it is not allowed to have
901 predecessor basic blocks (i.e. there can not be any branches to the entry
902 block of a function). Because the block can have no predecessors, it also
903 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
905 <p>LLVM allows an explicit section to be specified for functions. If the target
906 supports it, it will emit functions to the section specified.</p>
908 <p>An explicit alignment may be specified for a function. If not present, or if
909 the alignment is set to zero, the alignment of the function is set by the
910 target to whatever it feels convenient. If an explicit alignment is
911 specified, the function is forced to have at least that much alignment. All
912 alignments must be a power of 2.</p>
915 <pre class="doc_code">
916 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
917 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
918 <ResultType> @<FunctionName> ([argument list])
919 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
920 [<a href="#gc">gc</a>] { ... }
925 <!-- ======================================================================= -->
926 <div class="doc_subsection">
927 <a name="aliasstructure">Aliases</a>
930 <div class="doc_text">
932 <p>Aliases act as "second name" for the aliasee value (which can be either
933 function, global variable, another alias or bitcast of global value). Aliases
934 may have an optional <a href="#linkage">linkage type</a>, and an
935 optional <a href="#visibility">visibility style</a>.</p>
938 <pre class="doc_code">
939 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
944 <!-- ======================================================================= -->
945 <div class="doc_subsection">
946 <a name="namedmetadatastructure">Named Metadata</a>
949 <div class="doc_text">
951 <p>Named metadata is a collection of metadata. <a href="#metadata">Metadata
952 nodes</a> (but not metadata strings) and null are the only valid operands for
953 a named metadata.</p>
956 <pre class="doc_code">
957 ; An unnamed metadata node, which is referenced by the named metadata.
958 !1 = metadata !{metadata !"one"}
965 <!-- ======================================================================= -->
966 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
968 <div class="doc_text">
970 <p>The return type and each parameter of a function type may have a set of
971 <i>parameter attributes</i> associated with them. Parameter attributes are
972 used to communicate additional information about the result or parameters of
973 a function. Parameter attributes are considered to be part of the function,
974 not of the function type, so functions with different parameter attributes
975 can have the same function type.</p>
977 <p>Parameter attributes are simple keywords that follow the type specified. If
978 multiple parameter attributes are needed, they are space separated. For
981 <pre class="doc_code">
982 declare i32 @printf(i8* noalias nocapture, ...)
983 declare i32 @atoi(i8 zeroext)
984 declare signext i8 @returns_signed_char()
987 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
988 <tt>readonly</tt>) come immediately after the argument list.</p>
990 <p>Currently, only the following parameter attributes are defined:</p>
993 <dt><tt><b>zeroext</b></tt></dt>
994 <dd>This indicates to the code generator that the parameter or return value
995 should be zero-extended to a 32-bit value by the caller (for a parameter)
996 or the callee (for a return value).</dd>
998 <dt><tt><b>signext</b></tt></dt>
999 <dd>This indicates to the code generator that the parameter or return value
1000 should be sign-extended to a 32-bit value by the caller (for a parameter)
1001 or the callee (for a return value).</dd>
1003 <dt><tt><b>inreg</b></tt></dt>
1004 <dd>This indicates that this parameter or return value should be treated in a
1005 special target-dependent fashion during while emitting code for a function
1006 call or return (usually, by putting it in a register as opposed to memory,
1007 though some targets use it to distinguish between two different kinds of
1008 registers). Use of this attribute is target-specific.</dd>
1010 <dt><tt><b><a name="byval">byval</a></b></tt></dt>
1011 <dd>This indicates that the pointer parameter should really be passed by value
1012 to the function. The attribute implies that a hidden copy of the pointee
1013 is made between the caller and the callee, so the callee is unable to
1014 modify the value in the callee. This attribute is only valid on LLVM
1015 pointer arguments. It is generally used to pass structs and arrays by
1016 value, but is also valid on pointers to scalars. The copy is considered
1017 to belong to the caller not the callee (for example,
1018 <tt><a href="#readonly">readonly</a></tt> functions should not write to
1019 <tt>byval</tt> parameters). This is not a valid attribute for return
1020 values. The byval attribute also supports specifying an alignment with
1021 the align attribute. This has a target-specific effect on the code
1022 generator that usually indicates a desired alignment for the synthesized
1025 <dt><tt><b><a name="sret">sret</a></b></tt></dt>
1026 <dd>This indicates that the pointer parameter specifies the address of a
1027 structure that is the return value of the function in the source program.
1028 This pointer must be guaranteed by the caller to be valid: loads and
1029 stores to the structure may be assumed by the callee to not to trap. This
1030 may only be applied to the first parameter. This is not a valid attribute
1031 for return values. </dd>
1033 <dt><tt><b><a name="noalias">noalias</a></b></tt></dt>
1034 <dd>This indicates that pointer values
1035 <a href="#pointeraliasing"><i>based</i></a> on the argument or return
1036 value do not alias pointer values which are not <i>based</i> on it,
1037 ignoring certain "irrelevant" dependencies.
1038 For a call to the parent function, dependencies between memory
1039 references from before or after the call and from those during the call
1040 are "irrelevant" to the <tt>noalias</tt> keyword for the arguments and
1041 return value used in that call.
1042 The caller shares the responsibility with the callee for ensuring that
1043 these requirements are met.
1044 For further details, please see the discussion of the NoAlias response in
1045 <a href="AliasAnalysis.html#MustMayNo">alias analysis</a>.<br>
1047 Note that this definition of <tt>noalias</tt> is intentionally
1048 similar to the definition of <tt>restrict</tt> in C99 for function
1049 arguments, though it is slightly weaker.
1051 For function return values, C99's <tt>restrict</tt> is not meaningful,
1052 while LLVM's <tt>noalias</tt> is.
1055 <dt><tt><b><a name="nocapture">nocapture</a></b></tt></dt>
1056 <dd>This indicates that the callee does not make any copies of the pointer
1057 that outlive the callee itself. This is not a valid attribute for return
1060 <dt><tt><b><a name="nest">nest</a></b></tt></dt>
1061 <dd>This indicates that the pointer parameter can be excised using the
1062 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
1063 attribute for return values.</dd>
1068 <!-- ======================================================================= -->
1069 <div class="doc_subsection">
1070 <a name="gc">Garbage Collector Names</a>
1073 <div class="doc_text">
1075 <p>Each function may specify a garbage collector name, which is simply a
1078 <pre class="doc_code">
1079 define void @f() gc "name" { ... }
1082 <p>The compiler declares the supported values of <i>name</i>. Specifying a
1083 collector which will cause the compiler to alter its output in order to
1084 support the named garbage collection algorithm.</p>
1088 <!-- ======================================================================= -->
1089 <div class="doc_subsection">
1090 <a name="fnattrs">Function Attributes</a>
1093 <div class="doc_text">
1095 <p>Function attributes are set to communicate additional information about a
1096 function. Function attributes are considered to be part of the function, not
1097 of the function type, so functions with different parameter attributes can
1098 have the same function type.</p>
1100 <p>Function attributes are simple keywords that follow the type specified. If
1101 multiple attributes are needed, they are space separated. For example:</p>
1103 <pre class="doc_code">
1104 define void @f() noinline { ... }
1105 define void @f() alwaysinline { ... }
1106 define void @f() alwaysinline optsize { ... }
1107 define void @f() optsize { ... }
1111 <dt><tt><b>alignstack(<<em>n</em>>)</b></tt></dt>
1112 <dd>This attribute indicates that, when emitting the prologue and epilogue,
1113 the backend should forcibly align the stack pointer. Specify the
1114 desired alignment, which must be a power of two, in parentheses.
1116 <dt><tt><b>alwaysinline</b></tt></dt>
1117 <dd>This attribute indicates that the inliner should attempt to inline this
1118 function into callers whenever possible, ignoring any active inlining size
1119 threshold for this caller.</dd>
1121 <dt><tt><b>inlinehint</b></tt></dt>
1122 <dd>This attribute indicates that the source code contained a hint that inlining
1123 this function is desirable (such as the "inline" keyword in C/C++). It
1124 is just a hint; it imposes no requirements on the inliner.</dd>
1126 <dt><tt><b>naked</b></tt></dt>
1127 <dd>This attribute disables prologue / epilogue emission for the function.
1128 This can have very system-specific consequences.</dd>
1130 <dt><tt><b>noimplicitfloat</b></tt></dt>
1131 <dd>This attributes disables implicit floating point instructions.</dd>
1133 <dt><tt><b>noinline</b></tt></dt>
1134 <dd>This attribute indicates that the inliner should never inline this
1135 function in any situation. This attribute may not be used together with
1136 the <tt>alwaysinline</tt> attribute.</dd>
1138 <dt><tt><b>noredzone</b></tt></dt>
1139 <dd>This attribute indicates that the code generator should not use a red
1140 zone, even if the target-specific ABI normally permits it.</dd>
1142 <dt><tt><b>noreturn</b></tt></dt>
1143 <dd>This function attribute indicates that the function never returns
1144 normally. This produces undefined behavior at runtime if the function
1145 ever does dynamically return.</dd>
1147 <dt><tt><b>nounwind</b></tt></dt>
1148 <dd>This function attribute indicates that the function never returns with an
1149 unwind or exceptional control flow. If the function does unwind, its
1150 runtime behavior is undefined.</dd>
1152 <dt><tt><b>optsize</b></tt></dt>
1153 <dd>This attribute suggests that optimization passes and code generator passes
1154 make choices that keep the code size of this function low, and otherwise
1155 do optimizations specifically to reduce code size.</dd>
1157 <dt><tt><b>readnone</b></tt></dt>
1158 <dd>This attribute indicates that the function computes its result (or decides
1159 to unwind an exception) based strictly on its arguments, without
1160 dereferencing any pointer arguments or otherwise accessing any mutable
1161 state (e.g. memory, control registers, etc) visible to caller functions.
1162 It does not write through any pointer arguments
1163 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1164 changes any state visible to callers. This means that it cannot unwind
1165 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1166 could use the <tt>unwind</tt> instruction.</dd>
1168 <dt><tt><b><a name="readonly">readonly</a></b></tt></dt>
1169 <dd>This attribute indicates that the function does not write through any
1170 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1171 arguments) or otherwise modify any state (e.g. memory, control registers,
1172 etc) visible to caller functions. It may dereference pointer arguments
1173 and read state that may be set in the caller. A readonly function always
1174 returns the same value (or unwinds an exception identically) when called
1175 with the same set of arguments and global state. It cannot unwind an
1176 exception by calling the <tt>C++</tt> exception throwing methods, but may
1177 use the <tt>unwind</tt> instruction.</dd>
1179 <dt><tt><b><a name="ssp">ssp</a></b></tt></dt>
1180 <dd>This attribute indicates that the function should emit a stack smashing
1181 protector. It is in the form of a "canary"—a random value placed on
1182 the stack before the local variables that's checked upon return from the
1183 function to see if it has been overwritten. A heuristic is used to
1184 determine if a function needs stack protectors or not.<br>
1186 If a function that has an <tt>ssp</tt> attribute is inlined into a
1187 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1188 function will have an <tt>ssp</tt> attribute.</dd>
1190 <dt><tt><b>sspreq</b></tt></dt>
1191 <dd>This attribute indicates that the function should <em>always</em> emit a
1192 stack smashing protector. This overrides
1193 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1195 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1196 function that doesn't have an <tt>sspreq</tt> attribute or which has
1197 an <tt>ssp</tt> attribute, then the resulting function will have
1198 an <tt>sspreq</tt> attribute.</dd>
1203 <!-- ======================================================================= -->
1204 <div class="doc_subsection">
1205 <a name="moduleasm">Module-Level Inline Assembly</a>
1208 <div class="doc_text">
1210 <p>Modules may contain "module-level inline asm" blocks, which corresponds to
1211 the GCC "file scope inline asm" blocks. These blocks are internally
1212 concatenated by LLVM and treated as a single unit, but may be separated in
1213 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
1215 <pre class="doc_code">
1216 module asm "inline asm code goes here"
1217 module asm "more can go here"
1220 <p>The strings can contain any character by escaping non-printable characters.
1221 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1224 <p>The inline asm code is simply printed to the machine code .s file when
1225 assembly code is generated.</p>
1229 <!-- ======================================================================= -->
1230 <div class="doc_subsection">
1231 <a name="datalayout">Data Layout</a>
1234 <div class="doc_text">
1236 <p>A module may specify a target specific data layout string that specifies how
1237 data is to be laid out in memory. The syntax for the data layout is
1240 <pre class="doc_code">
1241 target datalayout = "<i>layout specification</i>"
1244 <p>The <i>layout specification</i> consists of a list of specifications
1245 separated by the minus sign character ('-'). Each specification starts with
1246 a letter and may include other information after the letter to define some
1247 aspect of the data layout. The specifications accepted are as follows:</p>
1251 <dd>Specifies that the target lays out data in big-endian form. That is, the
1252 bits with the most significance have the lowest address location.</dd>
1255 <dd>Specifies that the target lays out data in little-endian form. That is,
1256 the bits with the least significance have the lowest address
1259 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1260 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1261 <i>preferred</i> alignments. All sizes are in bits. Specifying
1262 the <i>pref</i> alignment is optional. If omitted, the
1263 preceding <tt>:</tt> should be omitted too.</dd>
1265 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1266 <dd>This specifies the alignment for an integer type of a given bit
1267 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1269 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1270 <dd>This specifies the alignment for a vector type of a given bit
1273 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1274 <dd>This specifies the alignment for a floating point type of a given bit
1275 <i>size</i>. Only values of <i>size</i> that are supported by the target
1276 will work. 32 (float) and 64 (double) are supported on all targets;
1277 80 or 128 (different flavors of long double) are also supported on some
1280 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1281 <dd>This specifies the alignment for an aggregate type of a given bit
1284 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1285 <dd>This specifies the alignment for a stack object of a given bit
1288 <dt><tt>n<i>size1</i>:<i>size2</i>:<i>size3</i>...</tt></dt>
1289 <dd>This specifies a set of native integer widths for the target CPU
1290 in bits. For example, it might contain "n32" for 32-bit PowerPC,
1291 "n32:64" for PowerPC 64, or "n8:16:32:64" for X86-64. Elements of
1292 this set are considered to support most general arithmetic
1293 operations efficiently.</dd>
1296 <p>When constructing the data layout for a given target, LLVM starts with a
1297 default set of specifications which are then (possibly) overridden by the
1298 specifications in the <tt>datalayout</tt> keyword. The default specifications
1299 are given in this list:</p>
1302 <li><tt>E</tt> - big endian</li>
1303 <li><tt>p:64:64:64</tt> - 64-bit pointers with 64-bit alignment</li>
1304 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1305 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1306 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1307 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1308 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1309 alignment of 64-bits</li>
1310 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1311 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1312 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1313 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1314 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1315 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
1318 <p>When LLVM is determining the alignment for a given type, it uses the
1319 following rules:</p>
1322 <li>If the type sought is an exact match for one of the specifications, that
1323 specification is used.</li>
1325 <li>If no match is found, and the type sought is an integer type, then the
1326 smallest integer type that is larger than the bitwidth of the sought type
1327 is used. If none of the specifications are larger than the bitwidth then
1328 the the largest integer type is used. For example, given the default
1329 specifications above, the i7 type will use the alignment of i8 (next
1330 largest) while both i65 and i256 will use the alignment of i64 (largest
1333 <li>If no match is found, and the type sought is a vector type, then the
1334 largest vector type that is smaller than the sought vector type will be
1335 used as a fall back. This happens because <128 x double> can be
1336 implemented in terms of 64 <2 x double>, for example.</li>
1341 <!-- ======================================================================= -->
1342 <div class="doc_subsection">
1343 <a name="pointeraliasing">Pointer Aliasing Rules</a>
1346 <div class="doc_text">
1348 <p>Any memory access must be done through a pointer value associated
1349 with an address range of the memory access, otherwise the behavior
1350 is undefined. Pointer values are associated with address ranges
1351 according to the following rules:</p>
1354 <li>A pointer value is associated with the addresses associated with
1355 any value it is <i>based</i> on.
1356 <li>An address of a global variable is associated with the address
1357 range of the variable's storage.</li>
1358 <li>The result value of an allocation instruction is associated with
1359 the address range of the allocated storage.</li>
1360 <li>A null pointer in the default address-space is associated with
1362 <li>An integer constant other than zero or a pointer value returned
1363 from a function not defined within LLVM may be associated with address
1364 ranges allocated through mechanisms other than those provided by
1365 LLVM. Such ranges shall not overlap with any ranges of addresses
1366 allocated by mechanisms provided by LLVM.</li>
1369 <p>A pointer value is <i>based</i> on another pointer value according
1370 to the following rules:</p>
1373 <li>A pointer value formed from a
1374 <tt><a href="#i_getelementptr">getelementptr</a></tt> operation
1375 is <i>based</i> on the first operand of the <tt>getelementptr</tt>.</li>
1376 <li>The result value of a
1377 <tt><a href="#i_bitcast">bitcast</a></tt> is <i>based</i> on the operand
1378 of the <tt>bitcast</tt>.</li>
1379 <li>A pointer value formed by an
1380 <tt><a href="#i_inttoptr">inttoptr</a></tt> is <i>based</i> on all
1381 pointer values that contribute (directly or indirectly) to the
1382 computation of the pointer's value.</li>
1383 <li>The "<i>based</i> on" relationship is transitive.</li>
1386 <p>Note that this definition of <i>"based"</i> is intentionally
1387 similar to the definition of <i>"based"</i> in C99, though it is
1388 slightly weaker.</p>
1390 <p>LLVM IR does not associate types with memory. The result type of a
1391 <tt><a href="#i_load">load</a></tt> merely indicates the size and
1392 alignment of the memory from which to load, as well as the
1393 interpretation of the value. The first operand type of a
1394 <tt><a href="#i_store">store</a></tt> similarly only indicates the size
1395 and alignment of the store.</p>
1397 <p>Consequently, type-based alias analysis, aka TBAA, aka
1398 <tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1399 LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1400 additional information which specialized optimization passes may use
1401 to implement type-based alias analysis.</p>
1405 <!-- ======================================================================= -->
1406 <div class="doc_subsection">
1407 <a name="volatile">Volatile Memory Accesses</a>
1410 <div class="doc_text">
1412 <p>Certain memory accesses, such as <a href="#i_load"><tt>load</tt></a>s, <a
1413 href="#i_store"><tt>store</tt></a>s, and <a
1414 href="#int_memcpy"><tt>llvm.memcpy</tt></a>s may be marked <tt>volatile</tt>.
1415 The optimizers must not change the number of volatile operations or change their
1416 order of execution relative to other volatile operations. The optimizers
1417 <i>may</i> change the order of volatile operations relative to non-volatile
1418 operations. This is not Java's "volatile" and has no cross-thread
1419 synchronization behavior.</p>
1423 <!-- *********************************************************************** -->
1424 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1425 <!-- *********************************************************************** -->
1427 <div class="doc_text">
1429 <p>The LLVM type system is one of the most important features of the
1430 intermediate representation. Being typed enables a number of optimizations
1431 to be performed on the intermediate representation directly, without having
1432 to do extra analyses on the side before the transformation. A strong type
1433 system makes it easier to read the generated code and enables novel analyses
1434 and transformations that are not feasible to perform on normal three address
1435 code representations.</p>
1439 <!-- ======================================================================= -->
1440 <div class="doc_subsection"> <a name="t_classifications">Type
1441 Classifications</a> </div>
1443 <div class="doc_text">
1445 <p>The types fall into a few useful classifications:</p>
1447 <table border="1" cellspacing="0" cellpadding="4">
1449 <tr><th>Classification</th><th>Types</th></tr>
1451 <td><a href="#t_integer">integer</a></td>
1452 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1455 <td><a href="#t_floating">floating point</a></td>
1456 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1459 <td><a name="t_firstclass">first class</a></td>
1460 <td><a href="#t_integer">integer</a>,
1461 <a href="#t_floating">floating point</a>,
1462 <a href="#t_pointer">pointer</a>,
1463 <a href="#t_vector">vector</a>,
1464 <a href="#t_struct">structure</a>,
1465 <a href="#t_union">union</a>,
1466 <a href="#t_array">array</a>,
1467 <a href="#t_label">label</a>,
1468 <a href="#t_metadata">metadata</a>.
1472 <td><a href="#t_primitive">primitive</a></td>
1473 <td><a href="#t_label">label</a>,
1474 <a href="#t_void">void</a>,
1475 <a href="#t_floating">floating point</a>,
1476 <a href="#t_metadata">metadata</a>.</td>
1479 <td><a href="#t_derived">derived</a></td>
1480 <td><a href="#t_array">array</a>,
1481 <a href="#t_function">function</a>,
1482 <a href="#t_pointer">pointer</a>,
1483 <a href="#t_struct">structure</a>,
1484 <a href="#t_pstruct">packed structure</a>,
1485 <a href="#t_union">union</a>,
1486 <a href="#t_vector">vector</a>,
1487 <a href="#t_opaque">opaque</a>.
1493 <p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1494 important. Values of these types are the only ones which can be produced by
1499 <!-- ======================================================================= -->
1500 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1502 <div class="doc_text">
1504 <p>The primitive types are the fundamental building blocks of the LLVM
1509 <!-- _______________________________________________________________________ -->
1510 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1512 <div class="doc_text">
1515 <p>The integer type is a very simple type that simply specifies an arbitrary
1516 bit width for the integer type desired. Any bit width from 1 bit to
1517 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
1524 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1528 <table class="layout">
1530 <td class="left"><tt>i1</tt></td>
1531 <td class="left">a single-bit integer.</td>
1534 <td class="left"><tt>i32</tt></td>
1535 <td class="left">a 32-bit integer.</td>
1538 <td class="left"><tt>i1942652</tt></td>
1539 <td class="left">a really big integer of over 1 million bits.</td>
1545 <!-- _______________________________________________________________________ -->
1546 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1548 <div class="doc_text">
1552 <tr><th>Type</th><th>Description</th></tr>
1553 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1554 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1555 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1556 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1557 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1563 <!-- _______________________________________________________________________ -->
1564 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1566 <div class="doc_text">
1569 <p>The void type does not represent any value and has no size.</p>
1578 <!-- _______________________________________________________________________ -->
1579 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1581 <div class="doc_text">
1584 <p>The label type represents code labels.</p>
1593 <!-- _______________________________________________________________________ -->
1594 <div class="doc_subsubsection"> <a name="t_metadata">Metadata Type</a> </div>
1596 <div class="doc_text">
1599 <p>The metadata type represents embedded metadata. No derived types may be
1600 created from metadata except for <a href="#t_function">function</a>
1611 <!-- ======================================================================= -->
1612 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1614 <div class="doc_text">
1616 <p>The real power in LLVM comes from the derived types in the system. This is
1617 what allows a programmer to represent arrays, functions, pointers, and other
1618 useful types. Each of these types contain one or more element types which
1619 may be a primitive type, or another derived type. For example, it is
1620 possible to have a two dimensional array, using an array as the element type
1621 of another array.</p>
1626 <!-- _______________________________________________________________________ -->
1627 <div class="doc_subsubsection"> <a name="t_aggregate">Aggregate Types</a> </div>
1629 <div class="doc_text">
1631 <p>Aggregate Types are a subset of derived types that can contain multiple
1632 member types. <a href="#t_array">Arrays</a>,
1633 <a href="#t_struct">structs</a>, <a href="#t_vector">vectors</a> and
1634 <a href="#t_union">unions</a> are aggregate types.</p>
1638 <!-- _______________________________________________________________________ -->
1639 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1641 <div class="doc_text">
1644 <p>The array type is a very simple derived type that arranges elements
1645 sequentially in memory. The array type requires a size (number of elements)
1646 and an underlying data type.</p>
1650 [<# elements> x <elementtype>]
1653 <p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1654 be any type with a size.</p>
1657 <table class="layout">
1659 <td class="left"><tt>[40 x i32]</tt></td>
1660 <td class="left">Array of 40 32-bit integer values.</td>
1663 <td class="left"><tt>[41 x i32]</tt></td>
1664 <td class="left">Array of 41 32-bit integer values.</td>
1667 <td class="left"><tt>[4 x i8]</tt></td>
1668 <td class="left">Array of 4 8-bit integer values.</td>
1671 <p>Here are some examples of multidimensional arrays:</p>
1672 <table class="layout">
1674 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1675 <td class="left">3x4 array of 32-bit integer values.</td>
1678 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1679 <td class="left">12x10 array of single precision floating point values.</td>
1682 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1683 <td class="left">2x3x4 array of 16-bit integer values.</td>
1687 <p>There is no restriction on indexing beyond the end of the array implied by
1688 a static type (though there are restrictions on indexing beyond the bounds
1689 of an allocated object in some cases). This means that single-dimension
1690 'variable sized array' addressing can be implemented in LLVM with a zero
1691 length array type. An implementation of 'pascal style arrays' in LLVM could
1692 use the type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
1696 <!-- _______________________________________________________________________ -->
1697 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1699 <div class="doc_text">
1702 <p>The function type can be thought of as a function signature. It consists of
1703 a return type and a list of formal parameter types. The return type of a
1704 function type is a scalar type, a void type, a struct type, or a union
1705 type. If the return type is a struct type then all struct elements must be
1706 of first class types, and the struct must have at least one element.</p>
1710 <returntype> (<parameter list>)
1713 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1714 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1715 which indicates that the function takes a variable number of arguments.
1716 Variable argument functions can access their arguments with
1717 the <a href="#int_varargs">variable argument handling intrinsic</a>
1718 functions. '<tt><returntype></tt>' is any type except
1719 <a href="#t_label">label</a>.</p>
1722 <table class="layout">
1724 <td class="left"><tt>i32 (i32)</tt></td>
1725 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1727 </tr><tr class="layout">
1728 <td class="left"><tt>float (i16, i32 *) *
1730 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1731 an <tt>i16</tt> and a <a href="#t_pointer">pointer</a> to <tt>i32</tt>,
1732 returning <tt>float</tt>.
1734 </tr><tr class="layout">
1735 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1736 <td class="left">A vararg function that takes at least one
1737 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1738 which returns an integer. This is the signature for <tt>printf</tt> in
1741 </tr><tr class="layout">
1742 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1743 <td class="left">A function taking an <tt>i32</tt>, returning a
1744 <a href="#t_struct">structure</a> containing two <tt>i32</tt> values
1751 <!-- _______________________________________________________________________ -->
1752 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1754 <div class="doc_text">
1757 <p>The structure type is used to represent a collection of data members together
1758 in memory. The packing of the field types is defined to match the ABI of the
1759 underlying processor. The elements of a structure may be any type that has a
1762 <p>Structures in memory are accessed using '<tt><a href="#i_load">load</a></tt>'
1763 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field
1764 with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
1765 Structures in registers are accessed using the
1766 '<tt><a href="#i_extractvalue">extractvalue</a></tt>' and
1767 '<tt><a href="#i_insertvalue">insertvalue</a></tt>' instructions.</p>
1770 { <type list> }
1774 <table class="layout">
1776 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1777 <td class="left">A triple of three <tt>i32</tt> values</td>
1778 </tr><tr class="layout">
1779 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1780 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1781 second element is a <a href="#t_pointer">pointer</a> to a
1782 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1783 an <tt>i32</tt>.</td>
1789 <!-- _______________________________________________________________________ -->
1790 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1793 <div class="doc_text">
1796 <p>The packed structure type is used to represent a collection of data members
1797 together in memory. There is no padding between fields. Further, the
1798 alignment of a packed structure is 1 byte. The elements of a packed
1799 structure may be any type that has a size.</p>
1801 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1802 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1803 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</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">
1817 <tt>< { float, i32 (i32)* } ></tt></td>
1818 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1819 second element is a <a href="#t_pointer">pointer</a> to a
1820 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1821 an <tt>i32</tt>.</td>
1827 <!-- _______________________________________________________________________ -->
1828 <div class="doc_subsubsection"> <a name="t_union">Union Type</a> </div>
1830 <div class="doc_text">
1833 <p>A union type describes an object with size and alignment suitable for
1834 an object of any one of a given set of types (also known as an "untagged"
1835 union). It is similar in concept and usage to a
1836 <a href="#t_struct">struct</a>, except that all members of the union
1837 have an offset of zero. The elements of a union may be any type that has a
1838 size. Unions must have at least one member - empty unions are not allowed.
1841 <p>The size of the union as a whole will be the size of its largest member,
1842 and the alignment requirements of the union as a whole will be the largest
1843 alignment requirement of any member.</p>
1845 <p>Union members are accessed using '<tt><a href="#i_load">load</a></tt> and
1846 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1847 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
1848 Since all members are at offset zero, the getelementptr instruction does
1849 not affect the address, only the type of the resulting pointer.</p>
1853 union { <type list> }
1857 <table class="layout">
1859 <td class="left"><tt>union { i32, i32*, float }</tt></td>
1860 <td class="left">A union of three types: an <tt>i32</tt>, a pointer to
1861 an <tt>i32</tt>, and a <tt>float</tt>.</td>
1862 </tr><tr class="layout">
1864 <tt>union { float, i32 (i32) * }</tt></td>
1865 <td class="left">A union, where the first element is a <tt>float</tt> and the
1866 second element is a <a href="#t_pointer">pointer</a> to a
1867 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1868 an <tt>i32</tt>.</td>
1874 <!-- _______________________________________________________________________ -->
1875 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1877 <div class="doc_text">
1880 <p>The pointer type is used to specify memory locations.
1881 Pointers are commonly used to reference objects in memory.</p>
1883 <p>Pointer types may have an optional address space attribute defining the
1884 numbered address space where the pointed-to object resides. The default
1885 address space is number zero. The semantics of non-zero address
1886 spaces are target-specific.</p>
1888 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
1889 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
1897 <table class="layout">
1899 <td class="left"><tt>[4 x i32]*</tt></td>
1900 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1901 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1904 <td class="left"><tt>i32 (i32*) *</tt></td>
1905 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1906 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1910 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1911 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1912 that resides in address space #5.</td>
1918 <!-- _______________________________________________________________________ -->
1919 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1921 <div class="doc_text">
1924 <p>A vector type is a simple derived type that represents a vector of elements.
1925 Vector types are used when multiple primitive data are operated in parallel
1926 using a single instruction (SIMD). A vector type requires a size (number of
1927 elements) and an underlying primitive data type. Vector types are considered
1928 <a href="#t_firstclass">first class</a>.</p>
1932 < <# elements> x <elementtype> >
1935 <p>The number of elements is a constant integer value; elementtype may be any
1936 integer or floating point type.</p>
1939 <table class="layout">
1941 <td class="left"><tt><4 x i32></tt></td>
1942 <td class="left">Vector of 4 32-bit integer values.</td>
1945 <td class="left"><tt><8 x float></tt></td>
1946 <td class="left">Vector of 8 32-bit floating-point values.</td>
1949 <td class="left"><tt><2 x i64></tt></td>
1950 <td class="left">Vector of 2 64-bit integer values.</td>
1956 <!-- _______________________________________________________________________ -->
1957 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1958 <div class="doc_text">
1961 <p>Opaque types are used to represent unknown types in the system. This
1962 corresponds (for example) to the C notion of a forward declared structure
1963 type. In LLVM, opaque types can eventually be resolved to any type (not just
1964 a structure type).</p>
1972 <table class="layout">
1974 <td class="left"><tt>opaque</tt></td>
1975 <td class="left">An opaque type.</td>
1981 <!-- ======================================================================= -->
1982 <div class="doc_subsection">
1983 <a name="t_uprefs">Type Up-references</a>
1986 <div class="doc_text">
1989 <p>An "up reference" allows you to refer to a lexically enclosing type without
1990 requiring it to have a name. For instance, a structure declaration may
1991 contain a pointer to any of the types it is lexically a member of. Example
1992 of up references (with their equivalent as named type declarations)
1996 { \2 * } %x = type { %x* }
1997 { \2 }* %y = type { %y }*
2001 <p>An up reference is needed by the asmprinter for printing out cyclic types
2002 when there is no declared name for a type in the cycle. Because the
2003 asmprinter does not want to print out an infinite type string, it needs a
2004 syntax to handle recursive types that have no names (all names are optional
2012 <p>The level is the count of the lexical type that is being referred to.</p>
2015 <table class="layout">
2017 <td class="left"><tt>\1*</tt></td>
2018 <td class="left">Self-referential pointer.</td>
2021 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
2022 <td class="left">Recursive structure where the upref refers to the out-most
2029 <!-- *********************************************************************** -->
2030 <div class="doc_section"> <a name="constants">Constants</a> </div>
2031 <!-- *********************************************************************** -->
2033 <div class="doc_text">
2035 <p>LLVM has several different basic types of constants. This section describes
2036 them all and their syntax.</p>
2040 <!-- ======================================================================= -->
2041 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
2043 <div class="doc_text">
2046 <dt><b>Boolean constants</b></dt>
2047 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
2048 constants of the <tt><a href="#t_integer">i1</a></tt> type.</dd>
2050 <dt><b>Integer constants</b></dt>
2051 <dd>Standard integers (such as '4') are constants of
2052 the <a href="#t_integer">integer</a> type. Negative numbers may be used
2053 with integer types.</dd>
2055 <dt><b>Floating point constants</b></dt>
2056 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
2057 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
2058 notation (see below). The assembler requires the exact decimal value of a
2059 floating-point constant. For example, the assembler accepts 1.25 but
2060 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
2061 constants must have a <a href="#t_floating">floating point</a> type. </dd>
2063 <dt><b>Null pointer constants</b></dt>
2064 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
2065 and must be of <a href="#t_pointer">pointer type</a>.</dd>
2068 <p>The one non-intuitive notation for constants is the hexadecimal form of
2069 floating point constants. For example, the form '<tt>double
2070 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
2071 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
2072 constants are required (and the only time that they are generated by the
2073 disassembler) is when a floating point constant must be emitted but it cannot
2074 be represented as a decimal floating point number in a reasonable number of
2075 digits. For example, NaN's, infinities, and other special values are
2076 represented in their IEEE hexadecimal format so that assembly and disassembly
2077 do not cause any bits to change in the constants.</p>
2079 <p>When using the hexadecimal form, constants of types float and double are
2080 represented using the 16-digit form shown above (which matches the IEEE754
2081 representation for double); float values must, however, be exactly
2082 representable as IEE754 single precision. Hexadecimal format is always used
2083 for long double, and there are three forms of long double. The 80-bit format
2084 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
2085 The 128-bit format used by PowerPC (two adjacent doubles) is represented
2086 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
2087 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
2088 currently supported target uses this format. Long doubles will only work if
2089 they match the long double format on your target. All hexadecimal formats
2090 are big-endian (sign bit at the left).</p>
2094 <!-- ======================================================================= -->
2095 <div class="doc_subsection">
2096 <a name="aggregateconstants"></a> <!-- old anchor -->
2097 <a name="complexconstants">Complex Constants</a>
2100 <div class="doc_text">
2102 <p>Complex constants are a (potentially recursive) combination of simple
2103 constants and smaller complex constants.</p>
2106 <dt><b>Structure constants</b></dt>
2107 <dd>Structure constants are represented with notation similar to structure
2108 type definitions (a comma separated list of elements, surrounded by braces
2109 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
2110 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
2111 Structure constants must have <a href="#t_struct">structure type</a>, and
2112 the number and types of elements must match those specified by the
2115 <dt><b>Union constants</b></dt>
2116 <dd>Union constants are represented with notation similar to a structure with
2117 a single element - that is, a single typed element surrounded
2118 by braces (<tt>{}</tt>)). For example: "<tt>{ i32 4 }</tt>". The
2119 <a href="#t_union">union type</a> can be initialized with a single-element
2120 struct as long as the type of the struct element matches the type of
2121 one of the union members.</dd>
2123 <dt><b>Array constants</b></dt>
2124 <dd>Array constants are represented with notation similar to array type
2125 definitions (a comma separated list of elements, surrounded by square
2126 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
2127 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
2128 the number and types of elements must match those specified by the
2131 <dt><b>Vector constants</b></dt>
2132 <dd>Vector constants are represented with notation similar to vector type
2133 definitions (a comma separated list of elements, surrounded by
2134 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32
2135 42, i32 11, i32 74, i32 100 ></tt>". Vector constants must
2136 have <a href="#t_vector">vector type</a>, and the number and types of
2137 elements must match those specified by the type.</dd>
2139 <dt><b>Zero initialization</b></dt>
2140 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
2141 value to zero of <em>any</em> type, including scalar and
2142 <a href="#t_aggregate">aggregate</a> types.
2143 This is often used to avoid having to print large zero initializers
2144 (e.g. for large arrays) and is always exactly equivalent to using explicit
2145 zero initializers.</dd>
2147 <dt><b>Metadata node</b></dt>
2148 <dd>A metadata node is a structure-like constant with
2149 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
2150 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
2151 be interpreted as part of the instruction stream, metadata is a place to
2152 attach additional information such as debug info.</dd>
2157 <!-- ======================================================================= -->
2158 <div class="doc_subsection">
2159 <a name="globalconstants">Global Variable and Function Addresses</a>
2162 <div class="doc_text">
2164 <p>The addresses of <a href="#globalvars">global variables</a>
2165 and <a href="#functionstructure">functions</a> are always implicitly valid
2166 (link-time) constants. These constants are explicitly referenced when
2167 the <a href="#identifiers">identifier for the global</a> is used and always
2168 have <a href="#t_pointer">pointer</a> type. For example, the following is a
2169 legal LLVM file:</p>
2171 <pre class="doc_code">
2174 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2179 <!-- ======================================================================= -->
2180 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
2181 <div class="doc_text">
2183 <p>The string '<tt>undef</tt>' can be used anywhere a constant is expected, and
2184 indicates that the user of the value may receive an unspecified bit-pattern.
2185 Undefined values may be of any type (other than label or void) and be used
2186 anywhere a constant is permitted.</p>
2188 <p>Undefined values are useful because they indicate to the compiler that the
2189 program is well defined no matter what value is used. This gives the
2190 compiler more freedom to optimize. Here are some examples of (potentially
2191 surprising) transformations that are valid (in pseudo IR):</p>
2194 <pre class="doc_code">
2204 <p>This is safe because all of the output bits are affected by the undef bits.
2205 Any output bit can have a zero or one depending on the input bits.</p>
2207 <pre class="doc_code">
2218 <p>These logical operations have bits that are not always affected by the input.
2219 For example, if "%X" has a zero bit, then the output of the 'and' operation will
2220 always be a zero, no matter what the corresponding bit from the undef is. As
2221 such, it is unsafe to optimize or assume that the result of the and is undef.
2222 However, it is safe to assume that all bits of the undef could be 0, and
2223 optimize the and to 0. Likewise, it is safe to assume that all the bits of
2224 the undef operand to the or could be set, allowing the or to be folded to
2227 <pre class="doc_code">
2228 %A = select undef, %X, %Y
2229 %B = select undef, 42, %Y
2230 %C = select %X, %Y, undef
2241 <p>This set of examples show that undefined select (and conditional branch)
2242 conditions can go "either way" but they have to come from one of the two
2243 operands. In the %A example, if %X and %Y were both known to have a clear low
2244 bit, then %A would have to have a cleared low bit. However, in the %C example,
2245 the optimizer is allowed to assume that the undef operand could be the same as
2246 %Y, allowing the whole select to be eliminated.</p>
2249 <pre class="doc_code">
2250 %A = xor undef, undef
2268 <p>This example points out that two undef operands are not necessarily the same.
2269 This can be surprising to people (and also matches C semantics) where they
2270 assume that "X^X" is always zero, even if X is undef. This isn't true for a
2271 number of reasons, but the short answer is that an undef "variable" can
2272 arbitrarily change its value over its "live range". This is true because the
2273 "variable" doesn't actually <em>have a live range</em>. Instead, the value is
2274 logically read from arbitrary registers that happen to be around when needed,
2275 so the value is not necessarily consistent over time. In fact, %A and %C need
2276 to have the same semantics or the core LLVM "replace all uses with" concept
2279 <pre class="doc_code">
2287 <p>These examples show the crucial difference between an <em>undefined
2288 value</em> and <em>undefined behavior</em>. An undefined value (like undef) is
2289 allowed to have an arbitrary bit-pattern. This means that the %A operation
2290 can be constant folded to undef because the undef could be an SNaN, and fdiv is
2291 not (currently) defined on SNaN's. However, in the second example, we can make
2292 a more aggressive assumption: because the undef is allowed to be an arbitrary
2293 value, we are allowed to assume that it could be zero. Since a divide by zero
2294 has <em>undefined behavior</em>, we are allowed to assume that the operation
2295 does not execute at all. This allows us to delete the divide and all code after
2296 it: since the undefined operation "can't happen", the optimizer can assume that
2297 it occurs in dead code.
2300 <pre class="doc_code">
2301 a: store undef -> %X
2302 b: store %X -> undef
2308 <p>These examples reiterate the fdiv example: a store "of" an undefined value
2309 can be assumed to not have any effect: we can assume that the value is
2310 overwritten with bits that happen to match what was already there. However, a
2311 store "to" an undefined location could clobber arbitrary memory, therefore, it
2312 has undefined behavior.</p>
2316 <!-- ======================================================================= -->
2317 <div class="doc_subsection"><a name="trapvalues">Trap Values</a></div>
2318 <div class="doc_text">
2320 <p>Trap values are similar to <a href="#undefvalues">undef values</a>, however
2321 instead of representing an unspecified bit pattern, they represent the
2322 fact that an instruction or constant expression which cannot evoke side
2323 effects has nevertheless detected a condition which results in undefined
2326 <p>There is currently no way of representing a trap value in the IR; they
2327 only exist when produced by operations such as
2328 <a href="#i_add"><tt>add</tt></a> with the <tt>nsw</tt> flag.</p>
2330 <p>Trap value behavior is defined in terms of value <i>dependence</i>:</p>
2333 <li>Values other than <a href="#i_phi"><tt>phi</tt></a> nodes depend on
2334 their operands.</li>
2336 <li><a href="#i_phi"><tt>Phi</tt></a> nodes depend on the operand corresponding
2337 to their dynamic predecessor basic block.</li>
2339 <li>Function arguments depend on the corresponding actual argument values in
2340 the dynamic callers of their functions.</li>
2342 <li><a href="#i_call"><tt>Call</tt></a> instructions depend on the
2343 <a href="#i_ret"><tt>ret</tt></a> instructions that dynamically transfer
2344 control back to them.</li>
2346 <li><a href="#i_invoke"><tt>Invoke</tt></a> instructions depend on the
2347 <a href="#i_ret"><tt>ret</tt></a>, <a href="#i_unwind"><tt>unwind</tt></a>,
2348 or exception-throwing call instructions that dynamically transfer control
2351 <li>Non-volatile loads and stores depend on the most recent stores to all of the
2352 referenced memory addresses, following the order in the IR
2353 (including loads and stores implied by intrinsics such as
2354 <a href="#int_memcpy"><tt>@llvm.memcpy</tt></a>.)</li>
2356 <!-- TODO: In the case of multiple threads, this only applies if the store
2357 "happens-before" the load or store. -->
2359 <!-- TODO: floating-point exception state -->
2361 <li>An instruction with externally visible side effects depends on the most
2362 recent preceding instruction with externally visible side effects, following
2363 the order in the IR. (This includes
2364 <a href="#volatile">volatile operations</a>.)</li>
2366 <li>An instruction <i>control-depends</i> on a
2367 <a href="#terminators">terminator instruction</a>
2368 if the terminator instruction has multiple successors and the instruction
2369 is always executed when control transfers to one of the successors, and
2370 may not be executed when control is transfered to another.</li>
2372 <li>Dependence is transitive.</li>
2376 <p>Whenever a trap value is generated, all values which depend on it evaluate
2377 to trap. If they have side effects, the evoke their side effects as if each
2378 operand with a trap value were undef. If they have externally-visible side
2379 effects, the behavior is undefined.</p>
2381 <p>Here are some examples:</p>
2383 <pre class="doc_code">
2385 %trap = sub nuw i32 0, 1 ; Results in a trap value.
2386 %still_trap = and i32 %trap, 0 ; Whereas (and i32 undef, 0) would return 0.
2387 %trap_yet_again = getelementptr i32* @h, i32 %still_trap
2388 store i32 0, i32* %trap_yet_again ; undefined behavior
2390 store i32 %trap, i32* @g ; Trap value conceptually stored to memory.
2391 %trap2 = load i32* @g ; Returns a trap value, not just undef.
2393 volatile store i32 %trap, i32* @g ; External observation; undefined behavior.
2395 %narrowaddr = bitcast i32* @g to i16*
2396 %wideaddr = bitcast i32* @g to i64*
2397 %trap3 = load 16* %narrowaddr ; Returns a trap value.
2398 %trap4 = load i64* %widaddr ; Returns a trap value.
2400 %cmp = icmp i32 slt %trap, 0 ; Returns a trap value.
2401 %br i1 %cmp, %true, %end ; Branch to either destination.
2404 volatile store i32 0, i32* @g ; This is control-dependent on %cmp, so
2405 ; it has undefined behavior.
2409 %p = phi i32 [ 0, %entry ], [ 1, %true ]
2410 ; Both edges into this PHI are
2411 ; control-dependent on %cmp, so this
2412 ; always results in a trap value.
2414 volatile store i32 0, i32* @g ; %end is control-equivalent to %entry
2415 ; so this is defined (ignoring earlier
2416 ; undefined behavior in this example).
2421 <!-- ======================================================================= -->
2422 <div class="doc_subsection"><a name="blockaddress">Addresses of Basic
2424 <div class="doc_text">
2426 <p><b><tt>blockaddress(@function, %block)</tt></b></p>
2428 <p>The '<tt>blockaddress</tt>' constant computes the address of the specified
2429 basic block in the specified function, and always has an i8* type. Taking
2430 the address of the entry block is illegal.</p>
2432 <p>This value only has defined behavior when used as an operand to the
2433 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>' instruction or for comparisons
2434 against null. Pointer equality tests between labels addresses is undefined
2435 behavior - though, again, comparison against null is ok, and no label is
2436 equal to the null pointer. This may also be passed around as an opaque
2437 pointer sized value as long as the bits are not inspected. This allows
2438 <tt>ptrtoint</tt> and arithmetic to be performed on these values so long as
2439 the original value is reconstituted before the <tt>indirectbr</tt>.</p>
2441 <p>Finally, some targets may provide defined semantics when
2442 using the value as the operand to an inline assembly, but that is target
2449 <!-- ======================================================================= -->
2450 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
2453 <div class="doc_text">
2455 <p>Constant expressions are used to allow expressions involving other constants
2456 to be used as constants. Constant expressions may be of
2457 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2458 operation that does not have side effects (e.g. load and call are not
2459 supported). The following is the syntax for constant expressions:</p>
2462 <dt><b><tt>trunc (CST to TYPE)</tt></b></dt>
2463 <dd>Truncate a constant to another type. The bit size of CST must be larger
2464 than the bit size of TYPE. Both types must be integers.</dd>
2466 <dt><b><tt>zext (CST to TYPE)</tt></b></dt>
2467 <dd>Zero extend a constant to another type. The bit size of CST must be
2468 smaller than the bit size of TYPE. Both types must be integers.</dd>
2470 <dt><b><tt>sext (CST to TYPE)</tt></b></dt>
2471 <dd>Sign extend a constant to another type. The bit size of CST must be
2472 smaller than the bit size of TYPE. Both types must be integers.</dd>
2474 <dt><b><tt>fptrunc (CST to TYPE)</tt></b></dt>
2475 <dd>Truncate a floating point constant to another floating point type. The
2476 size of CST must be larger than the size of TYPE. Both types must be
2477 floating point.</dd>
2479 <dt><b><tt>fpext (CST to TYPE)</tt></b></dt>
2480 <dd>Floating point extend a constant to another type. The size of CST must be
2481 smaller or equal to the size of TYPE. Both types must be floating
2484 <dt><b><tt>fptoui (CST to TYPE)</tt></b></dt>
2485 <dd>Convert a floating point constant to the corresponding unsigned integer
2486 constant. TYPE must be a scalar or vector integer type. CST must be of
2487 scalar or vector floating point type. Both CST and TYPE must be scalars,
2488 or vectors of the same number of elements. If the value won't fit in the
2489 integer type, the results are undefined.</dd>
2491 <dt><b><tt>fptosi (CST to TYPE)</tt></b></dt>
2492 <dd>Convert a floating point constant to the corresponding signed integer
2493 constant. TYPE must be a scalar or vector integer type. CST must be of
2494 scalar or vector floating point type. Both CST and TYPE must be scalars,
2495 or vectors of the same number of elements. If the value won't fit in the
2496 integer type, the results are undefined.</dd>
2498 <dt><b><tt>uitofp (CST to TYPE)</tt></b></dt>
2499 <dd>Convert an unsigned integer constant to the corresponding floating point
2500 constant. TYPE must be a scalar or vector floating point type. CST must be
2501 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2502 vectors of the same number of elements. If the value won't fit in the
2503 floating point type, the results are undefined.</dd>
2505 <dt><b><tt>sitofp (CST to TYPE)</tt></b></dt>
2506 <dd>Convert a signed integer constant to the corresponding floating point
2507 constant. TYPE must be a scalar or vector floating point type. CST must be
2508 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2509 vectors of the same number of elements. If the value won't fit in the
2510 floating point type, the results are undefined.</dd>
2512 <dt><b><tt>ptrtoint (CST to TYPE)</tt></b></dt>
2513 <dd>Convert a pointer typed constant to the corresponding integer constant
2514 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2515 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2516 make it fit in <tt>TYPE</tt>.</dd>
2518 <dt><b><tt>inttoptr (CST to TYPE)</tt></b></dt>
2519 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2520 type. CST must be of integer type. The CST value is zero extended,
2521 truncated, or unchanged to make it fit in a pointer size. This one is
2522 <i>really</i> dangerous!</dd>
2524 <dt><b><tt>bitcast (CST to TYPE)</tt></b></dt>
2525 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2526 are the same as those for the <a href="#i_bitcast">bitcast
2527 instruction</a>.</dd>
2529 <dt><b><tt>getelementptr (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2530 <dt><b><tt>getelementptr inbounds (CSTPTR, IDX0, IDX1, ...)</tt></b></dt>
2531 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2532 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2533 instruction, the index list may have zero or more indexes, which are
2534 required to make sense for the type of "CSTPTR".</dd>
2536 <dt><b><tt>select (COND, VAL1, VAL2)</tt></b></dt>
2537 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
2539 <dt><b><tt>icmp COND (VAL1, VAL2)</tt></b></dt>
2540 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2542 <dt><b><tt>fcmp COND (VAL1, VAL2)</tt></b></dt>
2543 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2545 <dt><b><tt>extractelement (VAL, IDX)</tt></b></dt>
2546 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2549 <dt><b><tt>insertelement (VAL, ELT, IDX)</tt></b></dt>
2550 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2553 <dt><b><tt>shufflevector (VEC1, VEC2, IDXMASK)</tt></b></dt>
2554 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2557 <dt><b><tt>extractvalue (VAL, IDX0, IDX1, ...)</tt></b></dt>
2558 <dd>Perform the <a href="#i_extractvalue">extractvalue operation</a> on
2559 constants. The index list is interpreted in a similar manner as indices in
2560 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2561 index value must be specified.</dd>
2563 <dt><b><tt>insertvalue (VAL, ELT, IDX0, IDX1, ...)</tt></b></dt>
2564 <dd>Perform the <a href="#i_insertvalue">insertvalue operation</a> on
2565 constants. The index list is interpreted in a similar manner as indices in
2566 a '<a href="#i_getelementptr">getelementptr</a>' operation. At least one
2567 index value must be specified.</dd>
2569 <dt><b><tt>OPCODE (LHS, RHS)</tt></b></dt>
2570 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2571 be any of the <a href="#binaryops">binary</a>
2572 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2573 on operands are the same as those for the corresponding instruction
2574 (e.g. no bitwise operations on floating point values are allowed).</dd>
2579 <!-- *********************************************************************** -->
2580 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2581 <!-- *********************************************************************** -->
2583 <!-- ======================================================================= -->
2584 <div class="doc_subsection">
2585 <a name="inlineasm">Inline Assembler Expressions</a>
2588 <div class="doc_text">
2590 <p>LLVM supports inline assembler expressions (as opposed
2591 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2592 a special value. This value represents the inline assembler as a string
2593 (containing the instructions to emit), a list of operand constraints (stored
2594 as a string), a flag that indicates whether or not the inline asm
2595 expression has side effects, and a flag indicating whether the function
2596 containing the asm needs to align its stack conservatively. An example
2597 inline assembler expression is:</p>
2599 <pre class="doc_code">
2600 i32 (i32) asm "bswap $0", "=r,r"
2603 <p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2604 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2607 <pre class="doc_code">
2608 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2611 <p>Inline asms with side effects not visible in the constraint list must be
2612 marked as having side effects. This is done through the use of the
2613 '<tt>sideeffect</tt>' keyword, like so:</p>
2615 <pre class="doc_code">
2616 call void asm sideeffect "eieio", ""()
2619 <p>In some cases inline asms will contain code that will not work unless the
2620 stack is aligned in some way, such as calls or SSE instructions on x86,
2621 yet will not contain code that does that alignment within the asm.
2622 The compiler should make conservative assumptions about what the asm might
2623 contain and should generate its usual stack alignment code in the prologue
2624 if the '<tt>alignstack</tt>' keyword is present:</p>
2626 <pre class="doc_code">
2627 call void asm alignstack "eieio", ""()
2630 <p>If both keywords appear the '<tt>sideeffect</tt>' keyword must come
2633 <p>TODO: The format of the asm and constraints string still need to be
2634 documented here. Constraints on what can be done (e.g. duplication, moving,
2635 etc need to be documented). This is probably best done by reference to
2636 another document that covers inline asm from a holistic perspective.</p>
2639 <div class="doc_subsubsection">
2640 <a name="inlineasm_md">Inline Asm Metadata</a>
2643 <div class="doc_text">
2645 <p>The call instructions that wrap inline asm nodes may have a "!srcloc" MDNode
2646 attached to it that contains a constant integer. If present, the code
2647 generator will use the integer as the location cookie value when report
2648 errors through the LLVMContext error reporting mechanisms. This allows a
2649 front-end to correlate backend errors that occur with inline asm back to the
2650 source code that produced it. For example:</p>
2652 <pre class="doc_code">
2653 call void asm sideeffect "something bad", ""()<b>, !srcloc !42</b>
2655 !42 = !{ i32 1234567 }
2658 <p>It is up to the front-end to make sense of the magic numbers it places in the
2663 <!-- ======================================================================= -->
2664 <div class="doc_subsection"><a name="metadata">Metadata Nodes and Metadata
2668 <div class="doc_text">
2670 <p>LLVM IR allows metadata to be attached to instructions in the program that
2671 can convey extra information about the code to the optimizers and code
2672 generator. One example application of metadata is source-level debug
2673 information. There are two metadata primitives: strings and nodes. All
2674 metadata has the <tt>metadata</tt> type and is identified in syntax by a
2675 preceding exclamation point ('<tt>!</tt>').</p>
2677 <p>A metadata string is a string surrounded by double quotes. It can contain
2678 any character by escaping non-printable characters with "\xx" where "xx" is
2679 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
2681 <p>Metadata nodes are represented with notation similar to structure constants
2682 (a comma separated list of elements, surrounded by braces and preceded by an
2683 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2684 10}</tt>". Metadata nodes can have any values as their operand.</p>
2686 <p>A <a href="#namedmetadatastructure">named metadata</a> is a collection of
2687 metadata nodes, which can be looked up in the module symbol table. For
2688 example: "<tt>!foo = metadata !{!4, !3}</tt>".
2690 <p>Metadata can be used as function arguments. Here <tt>llvm.dbg.value</tt>
2691 function is using two metadata arguments.</p>
2693 <pre class="doc_code">
2694 call void @llvm.dbg.value(metadata !24, i64 0, metadata !25)
2697 <p>Metadata can be attached with an instruction. Here metadata <tt>!21</tt> is
2698 attached with <tt>add</tt> instruction using <tt>!dbg</tt> identifier.</p>
2700 <pre class="doc_code">
2701 %indvar.next = add i64 %indvar, 1, !dbg !21
2706 <!-- *********************************************************************** -->
2707 <div class="doc_section">
2708 <a name="intrinsic_globals">Intrinsic Global Variables</a>
2710 <!-- *********************************************************************** -->
2712 <p>LLVM has a number of "magic" global variables that contain data that affect
2713 code generation or other IR semantics. These are documented here. All globals
2714 of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2715 section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2718 <!-- ======================================================================= -->
2719 <div class="doc_subsection">
2720 <a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
2723 <div class="doc_text">
2725 <p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2726 href="#linkage_appending">appending linkage</a>. This array contains a list of
2727 pointers to global variables and functions which may optionally have a pointer
2728 cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2734 @llvm.used = appending global [2 x i8*] [
2736 i8* bitcast (i32* @Y to i8*)
2737 ], section "llvm.metadata"
2740 <p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2741 compiler, assembler, and linker are required to treat the symbol as if there is
2742 a reference to the global that it cannot see. For example, if a variable has
2743 internal linkage and no references other than that from the <tt>@llvm.used</tt>
2744 list, it cannot be deleted. This is commonly used to represent references from
2745 inline asms and other things the compiler cannot "see", and corresponds to
2746 "attribute((used))" in GNU C.</p>
2748 <p>On some targets, the code generator must emit a directive to the assembler or
2749 object file to prevent the assembler and linker from molesting the symbol.</p>
2753 <!-- ======================================================================= -->
2754 <div class="doc_subsection">
2755 <a name="intg_compiler_used">The '<tt>llvm.compiler.used</tt>' Global Variable</a>
2758 <div class="doc_text">
2760 <p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2761 <tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2762 touching the symbol. On targets that support it, this allows an intelligent
2763 linker to optimize references to the symbol without being impeded as it would be
2764 by <tt>@llvm.used</tt>.</p>
2766 <p>This is a rare construct that should only be used in rare circumstances, and
2767 should not be exposed to source languages.</p>
2771 <!-- ======================================================================= -->
2772 <div class="doc_subsection">
2773 <a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
2776 <div class="doc_text">
2778 %0 = type { i32, void ()* }
2779 @llvm.global_ctors = appending global [1 x %0] [%0 { i32 65535, void ()* @ctor }]
2781 <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.
2786 <!-- ======================================================================= -->
2787 <div class="doc_subsection">
2788 <a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
2791 <div class="doc_text">
2793 %0 = type { i32, void ()* }
2794 @llvm.global_dtors = appending global [1 x %0] [%0 { i32 65535, void ()* @dtor }]
2797 <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.
2803 <!-- *********************************************************************** -->
2804 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2805 <!-- *********************************************************************** -->
2807 <div class="doc_text">
2809 <p>The LLVM instruction set consists of several different classifications of
2810 instructions: <a href="#terminators">terminator
2811 instructions</a>, <a href="#binaryops">binary instructions</a>,
2812 <a href="#bitwiseops">bitwise binary instructions</a>,
2813 <a href="#memoryops">memory instructions</a>, and
2814 <a href="#otherops">other instructions</a>.</p>
2818 <!-- ======================================================================= -->
2819 <div class="doc_subsection"> <a name="terminators">Terminator
2820 Instructions</a> </div>
2822 <div class="doc_text">
2824 <p>As mentioned <a href="#functionstructure">previously</a>, every basic block
2825 in a program ends with a "Terminator" instruction, which indicates which
2826 block should be executed after the current block is finished. These
2827 terminator instructions typically yield a '<tt>void</tt>' value: they produce
2828 control flow, not values (the one exception being the
2829 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2831 <p>There are seven different terminator instructions: the
2832 '<a href="#i_ret"><tt>ret</tt></a>' instruction, the
2833 '<a href="#i_br"><tt>br</tt></a>' instruction, the
2834 '<a href="#i_switch"><tt>switch</tt></a>' instruction, the
2835 '<a href="#i_indirectbr">'<tt>indirectbr</tt></a>' Instruction, the
2836 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the
2837 '<a href="#i_unwind"><tt>unwind</tt></a>' instruction, and the
2838 '<a href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2842 <!-- _______________________________________________________________________ -->
2843 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2844 Instruction</a> </div>
2846 <div class="doc_text">
2850 ret <type> <value> <i>; Return a value from a non-void function</i>
2851 ret void <i>; Return from void function</i>
2855 <p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
2856 a value) from a function back to the caller.</p>
2858 <p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
2859 value and then causes control flow, and one that just causes control flow to
2863 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
2864 return value. The type of the return value must be a
2865 '<a href="#t_firstclass">first class</a>' type.</p>
2867 <p>A function is not <a href="#wellformed">well formed</a> if it it has a
2868 non-void return type and contains a '<tt>ret</tt>' instruction with no return
2869 value or a return value with a type that does not match its type, or if it
2870 has a void return type and contains a '<tt>ret</tt>' instruction with a
2874 <p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
2875 the calling function's context. If the caller is a
2876 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
2877 instruction after the call. If the caller was an
2878 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
2879 the beginning of the "normal" destination block. If the instruction returns
2880 a value, that value shall set the call or invoke instruction's return
2885 ret i32 5 <i>; Return an integer value of 5</i>
2886 ret void <i>; Return from a void function</i>
2887 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
2891 <!-- _______________________________________________________________________ -->
2892 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2894 <div class="doc_text">
2898 br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2902 <p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
2903 different basic block in the current function. There are two forms of this
2904 instruction, corresponding to a conditional branch and an unconditional
2908 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
2909 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
2910 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
2914 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2915 argument is evaluated. If the value is <tt>true</tt>, control flows to the
2916 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2917 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2922 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
2923 br i1 %cond, label %IfEqual, label %IfUnequal
2925 <a href="#i_ret">ret</a> i32 1
2927 <a href="#i_ret">ret</a> i32 0
2932 <!-- _______________________________________________________________________ -->
2933 <div class="doc_subsubsection">
2934 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2937 <div class="doc_text">
2941 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2945 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2946 several different places. It is a generalization of the '<tt>br</tt>'
2947 instruction, allowing a branch to occur to one of many possible
2951 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2952 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
2953 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
2954 The table is not allowed to contain duplicate constant entries.</p>
2957 <p>The <tt>switch</tt> instruction specifies a table of values and
2958 destinations. When the '<tt>switch</tt>' instruction is executed, this table
2959 is searched for the given value. If the value is found, control flow is
2960 transferred to the corresponding destination; otherwise, control flow is
2961 transferred to the default destination.</p>
2963 <h5>Implementation:</h5>
2964 <p>Depending on properties of the target machine and the particular
2965 <tt>switch</tt> instruction, this instruction may be code generated in
2966 different ways. For example, it could be generated as a series of chained
2967 conditional branches or with a lookup table.</p>
2971 <i>; Emulate a conditional br instruction</i>
2972 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2973 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
2975 <i>; Emulate an unconditional br instruction</i>
2976 switch i32 0, label %dest [ ]
2978 <i>; Implement a jump table:</i>
2979 switch i32 %val, label %otherwise [ i32 0, label %onzero
2981 i32 2, label %ontwo ]
2987 <!-- _______________________________________________________________________ -->
2988 <div class="doc_subsubsection">
2989 <a name="i_indirectbr">'<tt>indirectbr</tt>' Instruction</a>
2992 <div class="doc_text">
2996 indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
3001 <p>The '<tt>indirectbr</tt>' instruction implements an indirect branch to a label
3002 within the current function, whose address is specified by
3003 "<tt>address</tt>". Address must be derived from a <a
3004 href="#blockaddress">blockaddress</a> constant.</p>
3008 <p>The '<tt>address</tt>' argument is the address of the label to jump to. The
3009 rest of the arguments indicate the full set of possible destinations that the
3010 address may point to. Blocks are allowed to occur multiple times in the
3011 destination list, though this isn't particularly useful.</p>
3013 <p>This destination list is required so that dataflow analysis has an accurate
3014 understanding of the CFG.</p>
3018 <p>Control transfers to the block specified in the address argument. All
3019 possible destination blocks must be listed in the label list, otherwise this
3020 instruction has undefined behavior. This implies that jumps to labels
3021 defined in other functions have undefined behavior as well.</p>
3023 <h5>Implementation:</h5>
3025 <p>This is typically implemented with a jump through a register.</p>
3029 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
3035 <!-- _______________________________________________________________________ -->
3036 <div class="doc_subsubsection">
3037 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
3040 <div class="doc_text">
3044 <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>]
3045 to label <normal label> unwind label <exception label>
3049 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
3050 function, with the possibility of control flow transfer to either the
3051 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
3052 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
3053 control flow will return to the "normal" label. If the callee (or any
3054 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
3055 instruction, control is interrupted and continued at the dynamically nearest
3056 "exception" label.</p>
3059 <p>This instruction requires several arguments:</p>
3062 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
3063 convention</a> the call should use. If none is specified, the call
3064 defaults to using C calling conventions.</li>
3066 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
3067 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
3068 '<tt>inreg</tt>' attributes are valid here.</li>
3070 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
3071 function value being invoked. In most cases, this is a direct function
3072 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
3073 off an arbitrary pointer to function value.</li>
3075 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
3076 function to be invoked. </li>
3078 <li>'<tt>function args</tt>': argument list whose types match the function
3079 signature argument types and parameter attributes. All arguments must be
3080 of <a href="#t_firstclass">first class</a> type. If the function
3081 signature indicates the function accepts a variable number of arguments,
3082 the extra arguments can be specified.</li>
3084 <li>'<tt>normal label</tt>': the label reached when the called function
3085 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
3087 <li>'<tt>exception label</tt>': the label reached when a callee returns with
3088 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
3090 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
3091 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
3092 '<tt>readnone</tt>' attributes are valid here.</li>
3096 <p>This instruction is designed to operate as a standard
3097 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
3098 primary difference is that it establishes an association with a label, which
3099 is used by the runtime library to unwind the stack.</p>
3101 <p>This instruction is used in languages with destructors to ensure that proper
3102 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
3103 exception. Additionally, this is important for implementation of
3104 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
3106 <p>For the purposes of the SSA form, the definition of the value returned by the
3107 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
3108 block to the "normal" label. If the callee unwinds then no return value is
3111 <p>Note that the code generator does not yet completely support unwind, and
3112 that the invoke/unwind semantics are likely to change in future versions.</p>
3116 %retval = invoke i32 @Test(i32 15) to label %Continue
3117 unwind label %TestCleanup <i>; {i32}:retval set</i>
3118 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
3119 unwind label %TestCleanup <i>; {i32}:retval set</i>
3124 <!-- _______________________________________________________________________ -->
3126 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
3127 Instruction</a> </div>
3129 <div class="doc_text">
3137 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
3138 at the first callee in the dynamic call stack which used
3139 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
3140 This is primarily used to implement exception handling.</p>
3143 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
3144 immediately halt. The dynamic call stack is then searched for the
3145 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
3146 Once found, execution continues at the "exceptional" destination block
3147 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
3148 instruction in the dynamic call chain, undefined behavior results.</p>
3150 <p>Note that the code generator does not yet completely support unwind, and
3151 that the invoke/unwind semantics are likely to change in future versions.</p>
3155 <!-- _______________________________________________________________________ -->
3157 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
3158 Instruction</a> </div>
3160 <div class="doc_text">
3168 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
3169 instruction is used to inform the optimizer that a particular portion of the
3170 code is not reachable. This can be used to indicate that the code after a
3171 no-return function cannot be reached, and other facts.</p>
3174 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
3178 <!-- ======================================================================= -->
3179 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
3181 <div class="doc_text">
3183 <p>Binary operators are used to do most of the computation in a program. They
3184 require two operands of the same type, execute an operation on them, and
3185 produce a single value. The operands might represent multiple data, as is
3186 the case with the <a href="#t_vector">vector</a> data type. The result value
3187 has the same type as its operands.</p>
3189 <p>There are several different binary operators:</p>
3193 <!-- _______________________________________________________________________ -->
3194 <div class="doc_subsubsection">
3195 <a name="i_add">'<tt>add</tt>' Instruction</a>
3198 <div class="doc_text">
3202 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3203 <result> = add nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3204 <result> = add nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3205 <result> = add nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3209 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
3212 <p>The two arguments to the '<tt>add</tt>' instruction must
3213 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3214 integer values. Both arguments must have identical types.</p>
3217 <p>The value produced is the integer sum of the two operands.</p>
3219 <p>If the sum has unsigned overflow, the result returned is the mathematical
3220 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
3222 <p>Because LLVM integers use a two's complement representation, this instruction
3223 is appropriate for both signed and unsigned integers.</p>
3225 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3226 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3227 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
3228 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3229 respectively, occurs.</p>
3233 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
3238 <!-- _______________________________________________________________________ -->
3239 <div class="doc_subsubsection">
3240 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
3243 <div class="doc_text">
3247 <result> = fadd <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3251 <p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
3254 <p>The two arguments to the '<tt>fadd</tt>' instruction must be
3255 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3256 floating point values. Both arguments must have identical types.</p>
3259 <p>The value produced is the floating point sum of the two operands.</p>
3263 <result> = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
3268 <!-- _______________________________________________________________________ -->
3269 <div class="doc_subsubsection">
3270 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
3273 <div class="doc_text">
3277 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3278 <result> = sub nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3279 <result> = sub nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3280 <result> = sub nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3284 <p>The '<tt>sub</tt>' instruction returns the difference of its two
3287 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
3288 '<tt>neg</tt>' instruction present in most other intermediate
3289 representations.</p>
3292 <p>The two arguments to the '<tt>sub</tt>' instruction must
3293 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3294 integer values. Both arguments must have identical types.</p>
3297 <p>The value produced is the integer difference of the two operands.</p>
3299 <p>If the difference has unsigned overflow, the result returned is the
3300 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
3303 <p>Because LLVM integers use a two's complement representation, this instruction
3304 is appropriate for both signed and unsigned integers.</p>
3306 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3307 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3308 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
3309 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3310 respectively, occurs.</p>
3314 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
3315 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
3320 <!-- _______________________________________________________________________ -->
3321 <div class="doc_subsubsection">
3322 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
3325 <div class="doc_text">
3329 <result> = fsub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3333 <p>The '<tt>fsub</tt>' instruction returns the difference of its two
3336 <p>Note that the '<tt>fsub</tt>' instruction is used to represent the
3337 '<tt>fneg</tt>' instruction present in most other intermediate
3338 representations.</p>
3341 <p>The two arguments to the '<tt>fsub</tt>' instruction must be
3342 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3343 floating point values. Both arguments must have identical types.</p>
3346 <p>The value produced is the floating point difference of the two operands.</p>
3350 <result> = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
3351 <result> = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
3356 <!-- _______________________________________________________________________ -->
3357 <div class="doc_subsubsection">
3358 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
3361 <div class="doc_text">
3365 <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3366 <result> = mul nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3367 <result> = mul nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3368 <result> = mul nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3372 <p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
3375 <p>The two arguments to the '<tt>mul</tt>' instruction must
3376 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3377 integer values. Both arguments must have identical types.</p>
3380 <p>The value produced is the integer product of the two operands.</p>
3382 <p>If the result of the multiplication has unsigned overflow, the result
3383 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
3384 width of the result.</p>
3386 <p>Because LLVM integers use a two's complement representation, and the result
3387 is the same width as the operands, this instruction returns the correct
3388 result for both signed and unsigned integers. If a full product
3389 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
3390 be sign-extended or zero-extended as appropriate to the width of the full
3393 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3394 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3395 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
3396 is a <a href="#trapvalues">trap value</a> if unsigned and/or signed overflow,
3397 respectively, occurs.</p>
3401 <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
3406 <!-- _______________________________________________________________________ -->
3407 <div class="doc_subsubsection">
3408 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
3411 <div class="doc_text">
3415 <result> = fmul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3419 <p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
3422 <p>The two arguments to the '<tt>fmul</tt>' instruction must be
3423 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3424 floating point values. Both arguments must have identical types.</p>
3427 <p>The value produced is the floating point product of the two operands.</p>
3431 <result> = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
3436 <!-- _______________________________________________________________________ -->
3437 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
3440 <div class="doc_text">
3444 <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3448 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
3451 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
3452 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3453 values. Both arguments must have identical types.</p>
3456 <p>The value produced is the unsigned integer quotient of the two operands.</p>
3458 <p>Note that unsigned integer division and signed integer division are distinct
3459 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
3461 <p>Division by zero leads to undefined behavior.</p>
3465 <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3470 <!-- _______________________________________________________________________ -->
3471 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
3474 <div class="doc_text">
3478 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3479 <result> = sdiv exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3483 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
3486 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
3487 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3488 values. Both arguments must have identical types.</p>
3491 <p>The value produced is the signed integer quotient of the two operands rounded
3494 <p>Note that signed integer division and unsigned integer division are distinct
3495 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
3497 <p>Division by zero leads to undefined behavior. Overflow also leads to
3498 undefined behavior; this is a rare case, but can occur, for example, by doing
3499 a 32-bit division of -2147483648 by -1.</p>
3501 <p>If the <tt>exact</tt> keyword is present, the result value of the
3502 <tt>sdiv</tt> is a <a href="#trapvalues">trap value</a> if the result would
3507 <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3512 <!-- _______________________________________________________________________ -->
3513 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
3514 Instruction</a> </div>
3516 <div class="doc_text">
3520 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3524 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
3527 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
3528 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3529 floating point values. Both arguments must have identical types.</p>
3532 <p>The value produced is the floating point quotient of the two operands.</p>
3536 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
3541 <!-- _______________________________________________________________________ -->
3542 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
3545 <div class="doc_text">
3549 <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3553 <p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3554 division of its two arguments.</p>
3557 <p>The two arguments to the '<tt>urem</tt>' instruction must be
3558 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3559 values. Both arguments must have identical types.</p>
3562 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
3563 This instruction always performs an unsigned division to get the
3566 <p>Note that unsigned integer remainder and signed integer remainder are
3567 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3569 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
3573 <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3578 <!-- _______________________________________________________________________ -->
3579 <div class="doc_subsubsection">
3580 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
3583 <div class="doc_text">
3587 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3591 <p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3592 division of its two operands. This instruction can also take
3593 <a href="#t_vector">vector</a> versions of the values in which case the
3594 elements must be integers.</p>
3597 <p>The two arguments to the '<tt>srem</tt>' instruction must be
3598 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3599 values. Both arguments must have identical types.</p>
3602 <p>This instruction returns the <i>remainder</i> of a division (where the result
3603 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
3604 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
3605 a value. For more information about the difference,
3606 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3607 Math Forum</a>. For a table of how this is implemented in various languages,
3608 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3609 Wikipedia: modulo operation</a>.</p>
3611 <p>Note that signed integer remainder and unsigned integer remainder are
3612 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3614 <p>Taking the remainder of a division by zero leads to undefined behavior.
3615 Overflow also leads to undefined behavior; this is a rare case, but can
3616 occur, for example, by taking the remainder of a 32-bit division of
3617 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3618 lets srem be implemented using instructions that return both the result of
3619 the division and the remainder.)</p>
3623 <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3628 <!-- _______________________________________________________________________ -->
3629 <div class="doc_subsubsection">
3630 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
3632 <div class="doc_text">
3636 <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3640 <p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3641 its two operands.</p>
3644 <p>The two arguments to the '<tt>frem</tt>' instruction must be
3645 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3646 floating point values. Both arguments must have identical types.</p>
3649 <p>This instruction returns the <i>remainder</i> of a division. The remainder
3650 has the same sign as the dividend.</p>
3654 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
3659 <!-- ======================================================================= -->
3660 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
3661 Operations</a> </div>
3663 <div class="doc_text">
3665 <p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3666 program. They are generally very efficient instructions and can commonly be
3667 strength reduced from other instructions. They require two operands of the
3668 same type, execute an operation on them, and produce a single value. The
3669 resulting value is the same type as its operands.</p>
3673 <!-- _______________________________________________________________________ -->
3674 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
3675 Instruction</a> </div>
3677 <div class="doc_text">
3681 <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3685 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3686 a specified number of bits.</p>
3689 <p>Both arguments to the '<tt>shl</tt>' instruction must be the
3690 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3691 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3694 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3695 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3696 is (statically or dynamically) negative or equal to or larger than the number
3697 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3698 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3699 shift amount in <tt>op2</tt>.</p>
3703 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
3704 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
3705 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
3706 <result> = shl i32 1, 32 <i>; undefined</i>
3707 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
3712 <!-- _______________________________________________________________________ -->
3713 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
3714 Instruction</a> </div>
3716 <div class="doc_text">
3720 <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3724 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
3725 operand shifted to the right a specified number of bits with zero fill.</p>
3728 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
3729 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3730 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3733 <p>This instruction always performs a logical shift right operation. The most
3734 significant bits of the result will be filled with zero bits after the shift.
3735 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
3736 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3737 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3738 shift amount in <tt>op2</tt>.</p>
3742 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
3743 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
3744 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
3745 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
3746 <result> = lshr i32 1, 32 <i>; undefined</i>
3747 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
3752 <!-- _______________________________________________________________________ -->
3753 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
3754 Instruction</a> </div>
3755 <div class="doc_text">
3759 <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3763 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3764 operand shifted to the right a specified number of bits with sign
3768 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
3769 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3770 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3773 <p>This instruction always performs an arithmetic shift right operation, The
3774 most significant bits of the result will be filled with the sign bit
3775 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3776 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
3777 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
3778 the corresponding shift amount in <tt>op2</tt>.</p>
3782 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3783 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3784 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3785 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
3786 <result> = ashr i32 1, 32 <i>; undefined</i>
3787 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
3792 <!-- _______________________________________________________________________ -->
3793 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
3794 Instruction</a> </div>
3796 <div class="doc_text">
3800 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3804 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
3808 <p>The two arguments to the '<tt>and</tt>' instruction must be
3809 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3810 values. Both arguments must have identical types.</p>
3813 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
3815 <table border="1" cellspacing="0" cellpadding="4">
3847 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
3848 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
3849 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
3852 <!-- _______________________________________________________________________ -->
3853 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
3855 <div class="doc_text">
3859 <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3863 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
3867 <p>The two arguments to the '<tt>or</tt>' instruction must be
3868 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3869 values. Both arguments must have identical types.</p>
3872 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
3874 <table border="1" cellspacing="0" cellpadding="4">
3906 <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
3907 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
3908 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
3913 <!-- _______________________________________________________________________ -->
3914 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3915 Instruction</a> </div>
3917 <div class="doc_text">
3921 <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3925 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
3926 its two operands. The <tt>xor</tt> is used to implement the "one's
3927 complement" operation, which is the "~" operator in C.</p>
3930 <p>The two arguments to the '<tt>xor</tt>' instruction must be
3931 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3932 values. Both arguments must have identical types.</p>
3935 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
3937 <table border="1" cellspacing="0" cellpadding="4">
3969 <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
3970 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3971 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3972 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
3977 <!-- ======================================================================= -->
3978 <div class="doc_subsection">
3979 <a name="vectorops">Vector Operations</a>
3982 <div class="doc_text">
3984 <p>LLVM supports several instructions to represent vector operations in a
3985 target-independent manner. These instructions cover the element-access and
3986 vector-specific operations needed to process vectors effectively. While LLVM
3987 does directly support these vector operations, many sophisticated algorithms
3988 will want to use target-specific intrinsics to take full advantage of a
3989 specific target.</p>
3993 <!-- _______________________________________________________________________ -->
3994 <div class="doc_subsubsection">
3995 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3998 <div class="doc_text">
4002 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
4006 <p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
4007 from a vector at a specified index.</p>
4011 <p>The first operand of an '<tt>extractelement</tt>' instruction is a value
4012 of <a href="#t_vector">vector</a> type. The second operand is an index
4013 indicating the position from which to extract the element. The index may be
4017 <p>The result is a scalar of the same type as the element type of
4018 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
4019 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4020 results are undefined.</p>
4024 <result> = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
4029 <!-- _______________________________________________________________________ -->
4030 <div class="doc_subsubsection">
4031 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
4034 <div class="doc_text">
4038 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
4042 <p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
4043 vector at a specified index.</p>
4046 <p>The first operand of an '<tt>insertelement</tt>' instruction is a value
4047 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
4048 whose type must equal the element type of the first operand. The third
4049 operand is an index indicating the position at which to insert the value.
4050 The index may be a variable.</p>
4053 <p>The result is a vector of the same type as <tt>val</tt>. Its element values
4054 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
4055 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
4056 results are undefined.</p>
4060 <result> = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
4065 <!-- _______________________________________________________________________ -->
4066 <div class="doc_subsubsection">
4067 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
4070 <div class="doc_text">
4074 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
4078 <p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
4079 from two input vectors, returning a vector with the same element type as the
4080 input and length that is the same as the shuffle mask.</p>
4083 <p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
4084 with types that match each other. The third argument is a shuffle mask whose
4085 element type is always 'i32'. The result of the instruction is a vector
4086 whose length is the same as the shuffle mask and whose element type is the
4087 same as the element type of the first two operands.</p>
4089 <p>The shuffle mask operand is required to be a constant vector with either
4090 constant integer or undef values.</p>
4093 <p>The elements of the two input vectors are numbered from left to right across
4094 both of the vectors. The shuffle mask operand specifies, for each element of
4095 the result vector, which element of the two input vectors the result element
4096 gets. The element selector may be undef (meaning "don't care") and the
4097 second operand may be undef if performing a shuffle from only one vector.</p>
4101 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4102 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
4103 <result> = shufflevector <4 x i32> %v1, <4 x i32> undef,
4104 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
4105 <result> = shufflevector <8 x i32> %v1, <8 x i32> undef,
4106 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
4107 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
4108 <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>
4113 <!-- ======================================================================= -->
4114 <div class="doc_subsection">
4115 <a name="aggregateops">Aggregate Operations</a>
4118 <div class="doc_text">
4120 <p>LLVM supports several instructions for working with
4121 <a href="#t_aggregate">aggregate</a> values.</p>
4125 <!-- _______________________________________________________________________ -->
4126 <div class="doc_subsubsection">
4127 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
4130 <div class="doc_text">
4134 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
4138 <p>The '<tt>extractvalue</tt>' instruction extracts the value of a member field
4139 from an <a href="#t_aggregate">aggregate</a> value.</p>
4142 <p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
4143 of <a href="#t_struct">struct</a>, <a href="#t_union">union</a> or
4144 <a href="#t_array">array</a> type. The operands are constant indices to
4145 specify which value to extract in a similar manner as indices in a
4146 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
4149 <p>The result is the value at the position in the aggregate specified by the
4154 <result> = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
4159 <!-- _______________________________________________________________________ -->
4160 <div class="doc_subsubsection">
4161 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
4164 <div class="doc_text">
4168 <result> = insertvalue <aggregate type> <val>, <ty> <elt>, <idx> <i>; yields <aggregate type></i>
4172 <p>The '<tt>insertvalue</tt>' instruction inserts a value into a member field
4173 in an <a href="#t_aggregate">aggregate</a> value.</p>
4176 <p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
4177 of <a href="#t_struct">struct</a>, <a href="#t_union">union</a> or
4178 <a href="#t_array">array</a> type. The second operand is a first-class
4179 value to insert. The following operands are constant indices indicating
4180 the position at which to insert the value in a similar manner as indices in a
4181 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction. The
4182 value to insert must have the same type as the value identified by the
4186 <p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
4187 that of <tt>val</tt> except that the value at the position specified by the
4188 indices is that of <tt>elt</tt>.</p>
4192 %agg1 = insertvalue {i32, float} undef, i32 1, 0 <i>; yields {i32 1, float undef}</i>
4193 %agg2 = insertvalue {i32, float} %agg1, float %val, 1 <i>; yields {i32 1, float %val}</i>
4199 <!-- ======================================================================= -->
4200 <div class="doc_subsection">
4201 <a name="memoryops">Memory Access and Addressing Operations</a>
4204 <div class="doc_text">
4206 <p>A key design point of an SSA-based representation is how it represents
4207 memory. In LLVM, no memory locations are in SSA form, which makes things
4208 very simple. This section describes how to read, write, and allocate
4213 <!-- _______________________________________________________________________ -->
4214 <div class="doc_subsubsection">
4215 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
4218 <div class="doc_text">
4222 <result> = alloca <type>[, <ty> <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
4226 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
4227 currently executing function, to be automatically released when this function
4228 returns to its caller. The object is always allocated in the generic address
4229 space (address space zero).</p>
4232 <p>The '<tt>alloca</tt>' instruction
4233 allocates <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the
4234 runtime stack, returning a pointer of the appropriate type to the program.
4235 If "NumElements" is specified, it is the number of elements allocated,
4236 otherwise "NumElements" is defaulted to be one. If a constant alignment is
4237 specified, the value result of the allocation is guaranteed to be aligned to
4238 at least that boundary. If not specified, or if zero, the target can choose
4239 to align the allocation on any convenient boundary compatible with the
4242 <p>'<tt>type</tt>' may be any sized type.</p>
4245 <p>Memory is allocated; a pointer is returned. The operation is undefined if
4246 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
4247 memory is automatically released when the function returns. The
4248 '<tt>alloca</tt>' instruction is commonly used to represent automatic
4249 variables that must have an address available. When the function returns
4250 (either with the <tt><a href="#i_ret">ret</a></tt>
4251 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
4252 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
4256 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
4257 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
4258 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
4259 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
4264 <!-- _______________________________________________________________________ -->
4265 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
4266 Instruction</a> </div>
4268 <div class="doc_text">
4272 <result> = load <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>]
4273 <result> = volatile load <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>]
4274 !<index> = !{ i32 1 }
4278 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
4281 <p>The argument to the '<tt>load</tt>' instruction specifies the memory address
4282 from which to load. The pointer must point to
4283 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
4284 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
4285 number or order of execution of this <tt>load</tt> with other <a
4286 href="#volatile">volatile operations</a>.</p>
4288 <p>The optional constant <tt>align</tt> argument specifies the alignment of the
4289 operation (that is, the alignment of the memory address). A value of 0 or an
4290 omitted <tt>align</tt> argument means that the operation has the preferential
4291 alignment for the target. It is the responsibility of the code emitter to
4292 ensure that the alignment information is correct. Overestimating the
4293 alignment results in undefined behavior. Underestimating the alignment may
4294 produce less efficient code. An alignment of 1 is always safe.</p>
4296 <p>The optional <tt>!nontemporal</tt> metadata must reference a single
4297 metatadata name <index> corresponding to a metadata node with
4298 one <tt>i32</tt> entry of value 1. The existence of
4299 the <tt>!nontemporal</tt> metatadata on the instruction tells the optimizer
4300 and code generator that this load is not expected to be reused in the cache.
4301 The code generator may select special instructions to save cache bandwidth,
4302 such as the <tt>MOVNT</tt> instruction on x86.</p>
4305 <p>The location of memory pointed to is loaded. If the value being loaded is of
4306 scalar type then the number of bytes read does not exceed the minimum number
4307 of bytes needed to hold all bits of the type. For example, loading an
4308 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
4309 <tt>i20</tt> with a size that is not an integral number of bytes, the result
4310 is undefined if the value was not originally written using a store of the
4315 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4316 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
4317 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
4322 <!-- _______________________________________________________________________ -->
4323 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
4324 Instruction</a> </div>
4326 <div class="doc_text">
4330 store <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] <i>; yields {void}</i>
4331 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>][, !nontemporal !<index>] <i>; yields {void}</i>
4335 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
4338 <p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
4339 and an address at which to store it. The type of the
4340 '<tt><pointer></tt>' operand must be a pointer to
4341 the <a href="#t_firstclass">first class</a> type of the
4342 '<tt><value></tt>' operand. If the <tt>store</tt> is marked as
4343 <tt>volatile</tt>, then the optimizer is not allowed to modify the number or
4344 order of execution of this <tt>store</tt> with other <a
4345 href="#volatile">volatile operations</a>.</p>
4347 <p>The optional constant "align" argument specifies the alignment of the
4348 operation (that is, the alignment of the memory address). A value of 0 or an
4349 omitted "align" argument means that the operation has the preferential
4350 alignment for the target. It is the responsibility of the code emitter to
4351 ensure that the alignment information is correct. Overestimating the
4352 alignment results in an undefined behavior. Underestimating the alignment may
4353 produce less efficient code. An alignment of 1 is always safe.</p>
4355 <p>The optional !nontemporal metadata must reference a single metatadata
4356 name <index> corresponding to a metadata node with one i32 entry of
4357 value 1. The existence of the !nontemporal metatadata on the
4358 instruction tells the optimizer and code generator that this load is
4359 not expected to be reused in the cache. The code generator may
4360 select special instructions to save cache bandwidth, such as the
4361 MOVNT instruction on x86.</p>
4365 <p>The contents of memory are updated to contain '<tt><value></tt>' at the
4366 location specified by the '<tt><pointer></tt>' operand. If
4367 '<tt><value></tt>' is of scalar type then the number of bytes written
4368 does not exceed the minimum number of bytes needed to hold all bits of the
4369 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
4370 writing a value of a type like <tt>i20</tt> with a size that is not an
4371 integral number of bytes, it is unspecified what happens to the extra bits
4372 that do not belong to the type, but they will typically be overwritten.</p>
4376 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4377 store i32 3, i32* %ptr <i>; yields {void}</i>
4378 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
4383 <!-- _______________________________________________________________________ -->
4384 <div class="doc_subsubsection">
4385 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
4388 <div class="doc_text">
4392 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
4393 <result> = getelementptr inbounds <pty>* <ptrval>{, <ty> <idx>}*
4397 <p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
4398 subelement of an <a href="#t_aggregate">aggregate</a> data structure.
4399 It performs address calculation only and does not access memory.</p>
4402 <p>The first argument is always a pointer, and forms the basis of the
4403 calculation. The remaining arguments are indices that indicate which of the
4404 elements of the aggregate object are indexed. The interpretation of each
4405 index is dependent on the type being indexed into. The first index always
4406 indexes the pointer value given as the first argument, the second index
4407 indexes a value of the type pointed to (not necessarily the value directly
4408 pointed to, since the first index can be non-zero), etc. The first type
4409 indexed into must be a pointer value, subsequent types can be arrays,
4410 vectors, structs and unions. Note that subsequent types being indexed into
4411 can never be pointers, since that would require loading the pointer before
4412 continuing calculation.</p>
4414 <p>The type of each index argument depends on the type it is indexing into.
4415 When indexing into a (optionally packed) structure or union, only <tt>i32</tt>
4416 integer <b>constants</b> are allowed. When indexing into an array, pointer
4417 or vector, integers of any width are allowed, and they are not required to be
4420 <p>For example, let's consider a C code fragment and how it gets compiled to
4423 <pre class="doc_code">
4435 int *foo(struct ST *s) {
4436 return &s[1].Z.B[5][13];
4440 <p>The LLVM code generated by the GCC frontend is:</p>
4442 <pre class="doc_code">
4443 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
4444 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
4446 define i32* @foo(%ST* %s) {
4448 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
4454 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
4455 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
4456 }</tt>' type, a structure. The second index indexes into the third element
4457 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
4458 i8 }</tt>' type, another structure. The third index indexes into the second
4459 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
4460 array. The two dimensions of the array are subscripted into, yielding an
4461 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
4462 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
4464 <p>Note that it is perfectly legal to index partially through a structure,
4465 returning a pointer to an inner element. Because of this, the LLVM code for
4466 the given testcase is equivalent to:</p>
4469 define i32* @foo(%ST* %s) {
4470 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
4471 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
4472 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
4473 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
4474 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
4479 <p>If the <tt>inbounds</tt> keyword is present, the result value of the
4480 <tt>getelementptr</tt> is a <a href="#trapvalues">trap value</a> if the
4481 base pointer is not an <i>in bounds</i> address of an allocated object,
4482 or if any of the addresses that would be formed by successive addition of
4483 the offsets implied by the indices to the base address with infinitely
4484 precise arithmetic are not an <i>in bounds</i> address of that allocated
4485 object. The <i>in bounds</i> addresses for an allocated object are all
4486 the addresses that point into the object, plus the address one byte past
4489 <p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
4490 the base address with silently-wrapping two's complement arithmetic, and
4491 the result value of the <tt>getelementptr</tt> may be outside the object
4492 pointed to by the base pointer. The result value may not necessarily be
4493 used to access memory though, even if it happens to point into allocated
4494 storage. See the <a href="#pointeraliasing">Pointer Aliasing Rules</a>
4495 section for more information.</p>
4497 <p>The getelementptr instruction is often confusing. For some more insight into
4498 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
4502 <i>; yields [12 x i8]*:aptr</i>
4503 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
4504 <i>; yields i8*:vptr</i>
4505 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
4506 <i>; yields i8*:eptr</i>
4507 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
4508 <i>; yields i32*:iptr</i>
4509 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
4514 <!-- ======================================================================= -->
4515 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
4518 <div class="doc_text">
4520 <p>The instructions in this category are the conversion instructions (casting)
4521 which all take a single operand and a type. They perform various bit
4522 conversions on the operand.</p>
4526 <!-- _______________________________________________________________________ -->
4527 <div class="doc_subsubsection">
4528 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
4530 <div class="doc_text">
4534 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
4538 <p>The '<tt>trunc</tt>' instruction truncates its operand to the
4539 type <tt>ty2</tt>.</p>
4542 <p>The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
4543 be an <a href="#t_integer">integer</a> type, and a type that specifies the
4544 size and type of the result, which must be
4545 an <a href="#t_integer">integer</a> type. The bit size of <tt>value</tt> must
4546 be larger than the bit size of <tt>ty2</tt>. Equal sized types are not
4550 <p>The '<tt>trunc</tt>' instruction truncates the high order bits
4551 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
4552 source size must be larger than the destination size, <tt>trunc</tt> cannot
4553 be a <i>no-op cast</i>. It will always truncate bits.</p>
4557 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
4558 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
4559 %Z = trunc i32 122 to i1 <i>; yields i1:false</i>
4564 <!-- _______________________________________________________________________ -->
4565 <div class="doc_subsubsection">
4566 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
4568 <div class="doc_text">
4572 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
4576 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
4581 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
4582 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4583 also be of <a href="#t_integer">integer</a> type. The bit size of the
4584 <tt>value</tt> must be smaller than the bit size of the destination type,
4588 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
4589 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
4591 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
4595 %X = zext i32 257 to i64 <i>; yields i64:257</i>
4596 %Y = zext i1 true to i32 <i>; yields i32:1</i>
4601 <!-- _______________________________________________________________________ -->
4602 <div class="doc_subsubsection">
4603 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
4605 <div class="doc_text">
4609 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
4613 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
4616 <p>The '<tt>sext</tt>' instruction takes a value to cast, which must be of
4617 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4618 also be of <a href="#t_integer">integer</a> type. The bit size of the
4619 <tt>value</tt> must be smaller than the bit size of the destination type,
4623 <p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
4624 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
4625 of the type <tt>ty2</tt>.</p>
4627 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
4631 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
4632 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
4637 <!-- _______________________________________________________________________ -->
4638 <div class="doc_subsubsection">
4639 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
4642 <div class="doc_text">
4646 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
4650 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
4654 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
4655 point</a> value to cast and a <a href="#t_floating">floating point</a> type
4656 to cast it to. The size of <tt>value</tt> must be larger than the size of
4657 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
4658 <i>no-op cast</i>.</p>
4661 <p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
4662 <a href="#t_floating">floating point</a> type to a smaller
4663 <a href="#t_floating">floating point</a> type. If the value cannot fit
4664 within the destination type, <tt>ty2</tt>, then the results are
4669 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4670 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4675 <!-- _______________________________________________________________________ -->
4676 <div class="doc_subsubsection">
4677 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
4679 <div class="doc_text">
4683 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
4687 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
4688 floating point value.</p>
4691 <p>The '<tt>fpext</tt>' instruction takes a
4692 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
4693 a <a href="#t_floating">floating point</a> type to cast it to. The source
4694 type must be smaller than the destination type.</p>
4697 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
4698 <a href="#t_floating">floating point</a> type to a larger
4699 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4700 used to make a <i>no-op cast</i> because it always changes bits. Use
4701 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
4705 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
4706 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
4711 <!-- _______________________________________________________________________ -->
4712 <div class="doc_subsubsection">
4713 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
4715 <div class="doc_text">
4719 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
4723 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
4724 unsigned integer equivalent of type <tt>ty2</tt>.</p>
4727 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
4728 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4729 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4730 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4731 vector integer type with the same number of elements as <tt>ty</tt></p>
4734 <p>The '<tt>fptoui</tt>' instruction converts its
4735 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4736 towards zero) unsigned integer value. If the value cannot fit
4737 in <tt>ty2</tt>, the results are undefined.</p>
4741 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
4742 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
4743 %Z = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
4748 <!-- _______________________________________________________________________ -->
4749 <div class="doc_subsubsection">
4750 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
4752 <div class="doc_text">
4756 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
4760 <p>The '<tt>fptosi</tt>' instruction converts
4761 <a href="#t_floating">floating point</a> <tt>value</tt> to
4762 type <tt>ty2</tt>.</p>
4765 <p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
4766 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4767 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4768 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4769 vector integer type with the same number of elements as <tt>ty</tt></p>
4772 <p>The '<tt>fptosi</tt>' instruction converts its
4773 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4774 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4775 the results are undefined.</p>
4779 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
4780 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
4781 %Z = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
4786 <!-- _______________________________________________________________________ -->
4787 <div class="doc_subsubsection">
4788 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
4790 <div class="doc_text">
4794 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4798 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
4799 integer and converts that value to the <tt>ty2</tt> type.</p>
4802 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
4803 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4804 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4805 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4806 floating point type with the same number of elements as <tt>ty</tt></p>
4809 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
4810 integer quantity and converts it to the corresponding floating point
4811 value. If the value cannot fit in the floating point value, the results are
4816 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
4817 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
4822 <!-- _______________________________________________________________________ -->
4823 <div class="doc_subsubsection">
4824 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
4826 <div class="doc_text">
4830 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4834 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
4835 and converts that value to the <tt>ty2</tt> type.</p>
4838 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
4839 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4840 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4841 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4842 floating point type with the same number of elements as <tt>ty</tt></p>
4845 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
4846 quantity and converts it to the corresponding floating point value. If the
4847 value cannot fit in the floating point value, the results are undefined.</p>
4851 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
4852 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
4857 <!-- _______________________________________________________________________ -->
4858 <div class="doc_subsubsection">
4859 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4861 <div class="doc_text">
4865 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
4869 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4870 the integer type <tt>ty2</tt>.</p>
4873 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4874 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
4875 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
4878 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4879 <tt>ty2</tt> by interpreting the pointer value as an integer and either
4880 truncating or zero extending that value to the size of the integer type. If
4881 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4882 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4883 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4888 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4889 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4894 <!-- _______________________________________________________________________ -->
4895 <div class="doc_subsubsection">
4896 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4898 <div class="doc_text">
4902 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
4906 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
4907 pointer type, <tt>ty2</tt>.</p>
4910 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4911 value to cast, and a type to cast it to, which must be a
4912 <a href="#t_pointer">pointer</a> type.</p>
4915 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4916 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
4917 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4918 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
4919 than the size of a pointer then a zero extension is done. If they are the
4920 same size, nothing is done (<i>no-op cast</i>).</p>
4924 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4925 %Y = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4926 %Z = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4931 <!-- _______________________________________________________________________ -->
4932 <div class="doc_subsubsection">
4933 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4935 <div class="doc_text">
4939 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
4943 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4944 <tt>ty2</tt> without changing any bits.</p>
4947 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
4948 non-aggregate first class value, and a type to cast it to, which must also be
4949 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
4950 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
4951 identical. If the source type is a pointer, the destination type must also be
4952 a pointer. This instruction supports bitwise conversion of vectors to
4953 integers and to vectors of other types (as long as they have the same
4957 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4958 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4959 this conversion. The conversion is done as if the <tt>value</tt> had been
4960 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
4961 be converted to other pointer types with this instruction. To convert
4962 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
4963 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4967 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4968 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
4969 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
4974 <!-- ======================================================================= -->
4975 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4977 <div class="doc_text">
4979 <p>The instructions in this category are the "miscellaneous" instructions, which
4980 defy better classification.</p>
4984 <!-- _______________________________________________________________________ -->
4985 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4988 <div class="doc_text">
4992 <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4996 <p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
4997 boolean values based on comparison of its two integer, integer vector, or
4998 pointer operands.</p>
5001 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
5002 the condition code indicating the kind of comparison to perform. It is not a
5003 value, just a keyword. The possible condition code are:</p>
5006 <li><tt>eq</tt>: equal</li>
5007 <li><tt>ne</tt>: not equal </li>
5008 <li><tt>ugt</tt>: unsigned greater than</li>
5009 <li><tt>uge</tt>: unsigned greater or equal</li>
5010 <li><tt>ult</tt>: unsigned less than</li>
5011 <li><tt>ule</tt>: unsigned less or equal</li>
5012 <li><tt>sgt</tt>: signed greater than</li>
5013 <li><tt>sge</tt>: signed greater or equal</li>
5014 <li><tt>slt</tt>: signed less than</li>
5015 <li><tt>sle</tt>: signed less or equal</li>
5018 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
5019 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
5020 typed. They must also be identical types.</p>
5023 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
5024 condition code given as <tt>cond</tt>. The comparison performed always yields
5025 either an <a href="#t_integer"><tt>i1</tt></a> or vector of <tt>i1</tt>
5026 result, as follows:</p>
5029 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
5030 <tt>false</tt> otherwise. No sign interpretation is necessary or
5033 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
5034 <tt>false</tt> otherwise. No sign interpretation is necessary or
5037 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
5038 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5040 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
5041 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5042 to <tt>op2</tt>.</li>
5044 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
5045 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5047 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
5048 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5050 <li><tt>sgt</tt>: interprets the operands as signed values and yields
5051 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5053 <li><tt>sge</tt>: interprets the operands as signed values and yields
5054 <tt>true</tt> if <tt>op1</tt> is greater than or equal
5055 to <tt>op2</tt>.</li>
5057 <li><tt>slt</tt>: interprets the operands as signed values and yields
5058 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
5060 <li><tt>sle</tt>: interprets the operands as signed values and yields
5061 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5064 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
5065 values are compared as if they were integers.</p>
5067 <p>If the operands are integer vectors, then they are compared element by
5068 element. The result is an <tt>i1</tt> vector with the same number of elements
5069 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
5073 <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
5074 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
5075 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
5076 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
5077 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
5078 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
5081 <p>Note that the code generator does not yet support vector types with
5082 the <tt>icmp</tt> instruction.</p>
5086 <!-- _______________________________________________________________________ -->
5087 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
5090 <div class="doc_text">
5094 <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
5098 <p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
5099 values based on comparison of its operands.</p>
5101 <p>If the operands are floating point scalars, then the result type is a boolean
5102 (<a href="#t_integer"><tt>i1</tt></a>).</p>
5104 <p>If the operands are floating point vectors, then the result type is a vector
5105 of boolean with the same number of elements as the operands being
5109 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
5110 the condition code indicating the kind of comparison to perform. It is not a
5111 value, just a keyword. The possible condition code are:</p>
5114 <li><tt>false</tt>: no comparison, always returns false</li>
5115 <li><tt>oeq</tt>: ordered and equal</li>
5116 <li><tt>ogt</tt>: ordered and greater than </li>
5117 <li><tt>oge</tt>: ordered and greater than or equal</li>
5118 <li><tt>olt</tt>: ordered and less than </li>
5119 <li><tt>ole</tt>: ordered and less than or equal</li>
5120 <li><tt>one</tt>: ordered and not equal</li>
5121 <li><tt>ord</tt>: ordered (no nans)</li>
5122 <li><tt>ueq</tt>: unordered or equal</li>
5123 <li><tt>ugt</tt>: unordered or greater than </li>
5124 <li><tt>uge</tt>: unordered or greater than or equal</li>
5125 <li><tt>ult</tt>: unordered or less than </li>
5126 <li><tt>ule</tt>: unordered or less than or equal</li>
5127 <li><tt>une</tt>: unordered or not equal</li>
5128 <li><tt>uno</tt>: unordered (either nans)</li>
5129 <li><tt>true</tt>: no comparison, always returns true</li>
5132 <p><i>Ordered</i> means that neither operand is a QNAN while
5133 <i>unordered</i> means that either operand may be a QNAN.</p>
5135 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
5136 a <a href="#t_floating">floating point</a> type or
5137 a <a href="#t_vector">vector</a> of floating point type. They must have
5138 identical types.</p>
5141 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
5142 according to the condition code given as <tt>cond</tt>. If the operands are
5143 vectors, then the vectors are compared element by element. Each comparison
5144 performed always yields an <a href="#t_integer">i1</a> result, as
5148 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
5150 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5151 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5153 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5154 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5156 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5157 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5159 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5160 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5162 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5163 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5165 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
5166 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5168 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
5170 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
5171 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
5173 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
5174 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
5176 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
5177 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
5179 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
5180 <tt>op1</tt> is less than <tt>op2</tt>.</li>
5182 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
5183 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
5185 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
5186 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
5188 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
5190 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
5195 <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
5196 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
5197 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
5198 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
5201 <p>Note that the code generator does not yet support vector types with
5202 the <tt>fcmp</tt> instruction.</p>
5206 <!-- _______________________________________________________________________ -->
5207 <div class="doc_subsubsection">
5208 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
5211 <div class="doc_text">
5215 <result> = phi <ty> [ <val0>, <label0>], ...
5219 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in the
5220 SSA graph representing the function.</p>
5223 <p>The type of the incoming values is specified with the first type field. After
5224 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
5225 one pair for each predecessor basic block of the current block. Only values
5226 of <a href="#t_firstclass">first class</a> type may be used as the value
5227 arguments to the PHI node. Only labels may be used as the label
5230 <p>There must be no non-phi instructions between the start of a basic block and
5231 the PHI instructions: i.e. PHI instructions must be first in a basic
5234 <p>For the purposes of the SSA form, the use of each incoming value is deemed to
5235 occur on the edge from the corresponding predecessor block to the current
5236 block (but after any definition of an '<tt>invoke</tt>' instruction's return
5237 value on the same edge).</p>
5240 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
5241 specified by the pair corresponding to the predecessor basic block that
5242 executed just prior to the current block.</p>
5246 Loop: ; Infinite loop that counts from 0 on up...
5247 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
5248 %nextindvar = add i32 %indvar, 1
5254 <!-- _______________________________________________________________________ -->
5255 <div class="doc_subsubsection">
5256 <a name="i_select">'<tt>select</tt>' Instruction</a>
5259 <div class="doc_text">
5263 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
5265 <i>selty</i> is either i1 or {<N x i1>}
5269 <p>The '<tt>select</tt>' instruction is used to choose one value based on a
5270 condition, without branching.</p>
5274 <p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
5275 values indicating the condition, and two values of the
5276 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
5277 vectors and the condition is a scalar, then entire vectors are selected, not
5278 individual elements.</p>
5281 <p>If the condition is an i1 and it evaluates to 1, the instruction returns the
5282 first value argument; otherwise, it returns the second value argument.</p>
5284 <p>If the condition is a vector of i1, then the value arguments must be vectors
5285 of the same size, and the selection is done element by element.</p>
5289 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
5292 <p>Note that the code generator does not yet support conditions
5293 with vector type.</p>
5297 <!-- _______________________________________________________________________ -->
5298 <div class="doc_subsubsection">
5299 <a name="i_call">'<tt>call</tt>' Instruction</a>
5302 <div class="doc_text">
5306 <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>]
5310 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
5313 <p>This instruction requires several arguments:</p>
5316 <li>The optional "tail" marker indicates that the callee function does not
5317 access any allocas or varargs in the caller. Note that calls may be
5318 marked "tail" even if they do not occur before
5319 a <a href="#i_ret"><tt>ret</tt></a> instruction. If the "tail" marker is
5320 present, the function call is eligible for tail call optimization,
5321 but <a href="CodeGenerator.html#tailcallopt">might not in fact be
5322 optimized into a jump</a>. The code generator may optimize calls marked
5323 "tail" with either 1) automatic <a href="CodeGenerator.html#sibcallopt">
5324 sibling call optimization</a> when the caller and callee have
5325 matching signatures, or 2) forced tail call optimization when the
5326 following extra requirements are met:
5328 <li>Caller and callee both have the calling
5329 convention <tt>fastcc</tt>.</li>
5330 <li>The call is in tail position (ret immediately follows call and ret
5331 uses value of call or is void).</li>
5332 <li>Option <tt>-tailcallopt</tt> is enabled,
5333 or <code>llvm::GuaranteedTailCallOpt</code> is <code>true</code>.</li>
5334 <li><a href="CodeGenerator.html#tailcallopt">Platform specific
5335 constraints are met.</a></li>
5339 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
5340 convention</a> the call should use. If none is specified, the call
5341 defaults to using C calling conventions. The calling convention of the
5342 call must match the calling convention of the target function, or else the
5343 behavior is undefined.</li>
5345 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
5346 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
5347 '<tt>inreg</tt>' attributes are valid here.</li>
5349 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
5350 type of the return value. Functions that return no value are marked
5351 <tt><a href="#t_void">void</a></tt>.</li>
5353 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
5354 being invoked. The argument types must match the types implied by this
5355 signature. This type can be omitted if the function is not varargs and if
5356 the function type does not return a pointer to a function.</li>
5358 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
5359 be invoked. In most cases, this is a direct function invocation, but
5360 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
5361 to function value.</li>
5363 <li>'<tt>function args</tt>': argument list whose types match the function
5364 signature argument types and parameter attributes. All arguments must be
5365 of <a href="#t_firstclass">first class</a> type. If the function
5366 signature indicates the function accepts a variable number of arguments,
5367 the extra arguments can be specified.</li>
5369 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
5370 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
5371 '<tt>readnone</tt>' attributes are valid here.</li>
5375 <p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
5376 a specified function, with its incoming arguments bound to the specified
5377 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
5378 function, control flow continues with the instruction after the function
5379 call, and the return value of the function is bound to the result
5384 %retval = call i32 @test(i32 %argc)
5385 call i32 (i8*, ...)* @printf(i8* %msg, i32 12, i8 42) <i>; yields i32</i>
5386 %X = tail call i32 @foo() <i>; yields i32</i>
5387 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
5388 call void %foo(i8 97 signext)
5390 %struct.A = type { i32, i8 }
5391 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
5392 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
5393 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
5394 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
5395 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
5398 <p>llvm treats calls to some functions with names and arguments that match the
5399 standard C99 library as being the C99 library functions, and may perform
5400 optimizations or generate code for them under that assumption. This is
5401 something we'd like to change in the future to provide better support for
5402 freestanding environments and non-C-based languages.</p>
5406 <!-- _______________________________________________________________________ -->
5407 <div class="doc_subsubsection">
5408 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
5411 <div class="doc_text">
5415 <resultval> = va_arg <va_list*> <arglist>, <argty>
5419 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
5420 the "variable argument" area of a function call. It is used to implement the
5421 <tt>va_arg</tt> macro in C.</p>
5424 <p>This instruction takes a <tt>va_list*</tt> value and the type of the
5425 argument. It returns a value of the specified argument type and increments
5426 the <tt>va_list</tt> to point to the next argument. The actual type
5427 of <tt>va_list</tt> is target specific.</p>
5430 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
5431 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
5432 to the next argument. For more information, see the variable argument
5433 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
5435 <p>It is legal for this instruction to be called in a function which does not
5436 take a variable number of arguments, for example, the <tt>vfprintf</tt>
5439 <p><tt>va_arg</tt> is an LLVM instruction instead of
5440 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
5444 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
5446 <p>Note that the code generator does not yet fully support va_arg on many
5447 targets. Also, it does not currently support va_arg with aggregate types on
5452 <!-- *********************************************************************** -->
5453 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
5454 <!-- *********************************************************************** -->
5456 <div class="doc_text">
5458 <p>LLVM supports the notion of an "intrinsic function". These functions have
5459 well known names and semantics and are required to follow certain
5460 restrictions. Overall, these intrinsics represent an extension mechanism for
5461 the LLVM language that does not require changing all of the transformations
5462 in LLVM when adding to the language (or the bitcode reader/writer, the
5463 parser, etc...).</p>
5465 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
5466 prefix is reserved in LLVM for intrinsic names; thus, function names may not
5467 begin with this prefix. Intrinsic functions must always be external
5468 functions: you cannot define the body of intrinsic functions. Intrinsic
5469 functions may only be used in call or invoke instructions: it is illegal to
5470 take the address of an intrinsic function. Additionally, because intrinsic
5471 functions are part of the LLVM language, it is required if any are added that
5472 they be documented here.</p>
5474 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
5475 family of functions that perform the same operation but on different data
5476 types. Because LLVM can represent over 8 million different integer types,
5477 overloading is used commonly to allow an intrinsic function to operate on any
5478 integer type. One or more of the argument types or the result type can be
5479 overloaded to accept any integer type. Argument types may also be defined as
5480 exactly matching a previous argument's type or the result type. This allows
5481 an intrinsic function which accepts multiple arguments, but needs all of them
5482 to be of the same type, to only be overloaded with respect to a single
5483 argument or the result.</p>
5485 <p>Overloaded intrinsics will have the names of its overloaded argument types
5486 encoded into its function name, each preceded by a period. Only those types
5487 which are overloaded result in a name suffix. Arguments whose type is matched
5488 against another type do not. For example, the <tt>llvm.ctpop</tt> function
5489 can take an integer of any width and returns an integer of exactly the same
5490 integer width. This leads to a family of functions such as
5491 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
5492 %val)</tt>. Only one type, the return type, is overloaded, and only one type
5493 suffix is required. Because the argument's type is matched against the return
5494 type, it does not require its own name suffix.</p>
5496 <p>To learn how to add an intrinsic function, please see the
5497 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
5501 <!-- ======================================================================= -->
5502 <div class="doc_subsection">
5503 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
5506 <div class="doc_text">
5508 <p>Variable argument support is defined in LLVM with
5509 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
5510 intrinsic functions. These functions are related to the similarly named
5511 macros defined in the <tt><stdarg.h></tt> header file.</p>
5513 <p>All of these functions operate on arguments that use a target-specific value
5514 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
5515 not define what this type is, so all transformations should be prepared to
5516 handle these functions regardless of the type used.</p>
5518 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
5519 instruction and the variable argument handling intrinsic functions are
5522 <pre class="doc_code">
5523 define i32 @test(i32 %X, ...) {
5524 ; Initialize variable argument processing
5526 %ap2 = bitcast i8** %ap to i8*
5527 call void @llvm.va_start(i8* %ap2)
5529 ; Read a single integer argument
5530 %tmp = va_arg i8** %ap, i32
5532 ; Demonstrate usage of llvm.va_copy and llvm.va_end
5534 %aq2 = bitcast i8** %aq to i8*
5535 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
5536 call void @llvm.va_end(i8* %aq2)
5538 ; Stop processing of arguments.
5539 call void @llvm.va_end(i8* %ap2)
5543 declare void @llvm.va_start(i8*)
5544 declare void @llvm.va_copy(i8*, i8*)
5545 declare void @llvm.va_end(i8*)
5550 <!-- _______________________________________________________________________ -->
5551 <div class="doc_subsubsection">
5552 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
5556 <div class="doc_text">
5560 declare void %llvm.va_start(i8* <arglist>)
5564 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*<arglist></tt>
5565 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
5568 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
5571 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
5572 macro available in C. In a target-dependent way, it initializes
5573 the <tt>va_list</tt> element to which the argument points, so that the next
5574 call to <tt>va_arg</tt> will produce the first variable argument passed to
5575 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
5576 need to know the last argument of the function as the compiler can figure
5581 <!-- _______________________________________________________________________ -->
5582 <div class="doc_subsubsection">
5583 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
5586 <div class="doc_text">
5590 declare void @llvm.va_end(i8* <arglist>)
5594 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
5595 which has been initialized previously
5596 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
5597 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
5600 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
5603 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
5604 macro available in C. In a target-dependent way, it destroys
5605 the <tt>va_list</tt> element to which the argument points. Calls
5606 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
5607 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
5608 with calls to <tt>llvm.va_end</tt>.</p>
5612 <!-- _______________________________________________________________________ -->
5613 <div class="doc_subsubsection">
5614 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
5617 <div class="doc_text">
5621 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
5625 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
5626 from the source argument list to the destination argument list.</p>
5629 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
5630 The second argument is a pointer to a <tt>va_list</tt> element to copy
5634 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
5635 macro available in C. In a target-dependent way, it copies the
5636 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
5637 element. This intrinsic is necessary because
5638 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
5639 arbitrarily complex and require, for example, memory allocation.</p>
5643 <!-- ======================================================================= -->
5644 <div class="doc_subsection">
5645 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
5648 <div class="doc_text">
5650 <p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
5651 Collection</a> (GC) requires the implementation and generation of these
5652 intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
5653 roots on the stack</a>, as well as garbage collector implementations that
5654 require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
5655 barriers. Front-ends for type-safe garbage collected languages should generate
5656 these intrinsics to make use of the LLVM garbage collectors. For more details,
5657 see <a href="GarbageCollection.html">Accurate Garbage Collection with
5660 <p>The garbage collection intrinsics only operate on objects in the generic
5661 address space (address space zero).</p>
5665 <!-- _______________________________________________________________________ -->
5666 <div class="doc_subsubsection">
5667 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
5670 <div class="doc_text">
5674 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
5678 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
5679 the code generator, and allows some metadata to be associated with it.</p>
5682 <p>The first argument specifies the address of a stack object that contains the
5683 root pointer. The second pointer (which must be either a constant or a
5684 global value address) contains the meta-data to be associated with the
5688 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
5689 location. At compile-time, the code generator generates information to allow
5690 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5691 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5696 <!-- _______________________________________________________________________ -->
5697 <div class="doc_subsubsection">
5698 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
5701 <div class="doc_text">
5705 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
5709 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
5710 locations, allowing garbage collector implementations that require read
5714 <p>The second argument is the address to read from, which should be an address
5715 allocated from the garbage collector. The first object is a pointer to the
5716 start of the referenced object, if needed by the language runtime (otherwise
5720 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
5721 instruction, but may be replaced with substantially more complex code by the
5722 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5723 may only be used in a function which <a href="#gc">specifies a GC
5728 <!-- _______________________________________________________________________ -->
5729 <div class="doc_subsubsection">
5730 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
5733 <div class="doc_text">
5737 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
5741 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
5742 locations, allowing garbage collector implementations that require write
5743 barriers (such as generational or reference counting collectors).</p>
5746 <p>The first argument is the reference to store, the second is the start of the
5747 object to store it to, and the third is the address of the field of Obj to
5748 store to. If the runtime does not require a pointer to the object, Obj may
5752 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
5753 instruction, but may be replaced with substantially more complex code by the
5754 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5755 may only be used in a function which <a href="#gc">specifies a GC
5760 <!-- ======================================================================= -->
5761 <div class="doc_subsection">
5762 <a name="int_codegen">Code Generator Intrinsics</a>
5765 <div class="doc_text">
5767 <p>These intrinsics are provided by LLVM to expose special features that may
5768 only be implemented with code generator support.</p>
5772 <!-- _______________________________________________________________________ -->
5773 <div class="doc_subsubsection">
5774 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
5777 <div class="doc_text">
5781 declare i8 *@llvm.returnaddress(i32 <level>)
5785 <p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5786 target-specific value indicating the return address of the current function
5787 or one of its callers.</p>
5790 <p>The argument to this intrinsic indicates which function to return the address
5791 for. Zero indicates the calling function, one indicates its caller, etc.
5792 The argument is <b>required</b> to be a constant integer value.</p>
5795 <p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
5796 indicating the return address of the specified call frame, or zero if it
5797 cannot be identified. The value returned by this intrinsic is likely to be
5798 incorrect or 0 for arguments other than zero, so it should only be used for
5799 debugging purposes.</p>
5801 <p>Note that calling this intrinsic does not prevent function inlining or other
5802 aggressive transformations, so the value returned may not be that of the
5803 obvious source-language caller.</p>
5807 <!-- _______________________________________________________________________ -->
5808 <div class="doc_subsubsection">
5809 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5812 <div class="doc_text">
5816 declare i8* @llvm.frameaddress(i32 <level>)
5820 <p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5821 target-specific frame pointer value for the specified stack frame.</p>
5824 <p>The argument to this intrinsic indicates which function to return the frame
5825 pointer for. Zero indicates the calling function, one indicates its caller,
5826 etc. The argument is <b>required</b> to be a constant integer value.</p>
5829 <p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
5830 indicating the frame address of the specified call frame, or zero if it
5831 cannot be identified. The value returned by this intrinsic is likely to be
5832 incorrect or 0 for arguments other than zero, so it should only be used for
5833 debugging purposes.</p>
5835 <p>Note that calling this intrinsic does not prevent function inlining or other
5836 aggressive transformations, so the value returned may not be that of the
5837 obvious source-language caller.</p>
5841 <!-- _______________________________________________________________________ -->
5842 <div class="doc_subsubsection">
5843 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5846 <div class="doc_text">
5850 declare i8* @llvm.stacksave()
5854 <p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
5855 of the function stack, for use
5856 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
5857 useful for implementing language features like scoped automatic variable
5858 sized arrays in C99.</p>
5861 <p>This intrinsic returns a opaque pointer value that can be passed
5862 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
5863 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
5864 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
5865 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
5866 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
5867 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
5871 <!-- _______________________________________________________________________ -->
5872 <div class="doc_subsubsection">
5873 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5876 <div class="doc_text">
5880 declare void @llvm.stackrestore(i8* %ptr)
5884 <p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5885 the function stack to the state it was in when the
5886 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
5887 executed. This is useful for implementing language features like scoped
5888 automatic variable sized arrays in C99.</p>
5891 <p>See the description
5892 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
5896 <!-- _______________________________________________________________________ -->
5897 <div class="doc_subsubsection">
5898 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5901 <div class="doc_text">
5905 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
5909 <p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
5910 insert a prefetch instruction if supported; otherwise, it is a noop.
5911 Prefetches have no effect on the behavior of the program but can change its
5912 performance characteristics.</p>
5915 <p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
5916 specifier determining if the fetch should be for a read (0) or write (1),
5917 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5918 locality, to (3) - extremely local keep in cache. The <tt>rw</tt>
5919 and <tt>locality</tt> arguments must be constant integers.</p>
5922 <p>This intrinsic does not modify the behavior of the program. In particular,
5923 prefetches cannot trap and do not produce a value. On targets that support
5924 this intrinsic, the prefetch can provide hints to the processor cache for
5925 better performance.</p>
5929 <!-- _______________________________________________________________________ -->
5930 <div class="doc_subsubsection">
5931 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5934 <div class="doc_text">
5938 declare void @llvm.pcmarker(i32 <id>)
5942 <p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
5943 Counter (PC) in a region of code to simulators and other tools. The method
5944 is target specific, but it is expected that the marker will use exported
5945 symbols to transmit the PC of the marker. The marker makes no guarantees
5946 that it will remain with any specific instruction after optimizations. It is
5947 possible that the presence of a marker will inhibit optimizations. The
5948 intended use is to be inserted after optimizations to allow correlations of
5949 simulation runs.</p>
5952 <p><tt>id</tt> is a numerical id identifying the marker.</p>
5955 <p>This intrinsic does not modify the behavior of the program. Backends that do
5956 not support this intrinsic may ignore it.</p>
5960 <!-- _______________________________________________________________________ -->
5961 <div class="doc_subsubsection">
5962 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5965 <div class="doc_text">
5969 declare i64 @llvm.readcyclecounter()
5973 <p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5974 counter register (or similar low latency, high accuracy clocks) on those
5975 targets that support it. On X86, it should map to RDTSC. On Alpha, it
5976 should map to RPCC. As the backing counters overflow quickly (on the order
5977 of 9 seconds on alpha), this should only be used for small timings.</p>
5980 <p>When directly supported, reading the cycle counter should not modify any
5981 memory. Implementations are allowed to either return a application specific
5982 value or a system wide value. On backends without support, this is lowered
5983 to a constant 0.</p>
5987 <!-- ======================================================================= -->
5988 <div class="doc_subsection">
5989 <a name="int_libc">Standard C Library Intrinsics</a>
5992 <div class="doc_text">
5994 <p>LLVM provides intrinsics for a few important standard C library functions.
5995 These intrinsics allow source-language front-ends to pass information about
5996 the alignment of the pointer arguments to the code generator, providing
5997 opportunity for more efficient code generation.</p>
6001 <!-- _______________________________________________________________________ -->
6002 <div class="doc_subsubsection">
6003 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
6006 <div class="doc_text">
6009 <p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
6010 integer bit width and for different address spaces. Not all targets support
6011 all bit widths however.</p>
6014 declare void @llvm.memcpy.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
6015 i32 <len>, i32 <align>, i1 <isvolatile>)
6016 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
6017 i64 <len>, i32 <align>, i1 <isvolatile>)
6021 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6022 source location to the destination location.</p>
6024 <p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
6025 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6026 and the pointers can be in specified address spaces.</p>
6030 <p>The first argument is a pointer to the destination, the second is a pointer
6031 to the source. The third argument is an integer argument specifying the
6032 number of bytes to copy, the fourth argument is the alignment of the
6033 source and destination locations, and the fifth is a boolean indicating a
6034 volatile access.</p>
6036 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6037 then the caller guarantees that both the source and destination pointers are
6038 aligned to that boundary.</p>
6040 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6041 <tt>llvm.memcpy</tt> call is a <a href="#volatile">volatile operation</a>.
6042 The detailed access behavior is not very cleanly specified and it is unwise
6043 to depend on it.</p>
6047 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
6048 source location to the destination location, which are not allowed to
6049 overlap. It copies "len" bytes of memory over. If the argument is known to
6050 be aligned to some boundary, this can be specified as the fourth argument,
6051 otherwise it should be set to 0 or 1.</p>
6055 <!-- _______________________________________________________________________ -->
6056 <div class="doc_subsubsection">
6057 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
6060 <div class="doc_text">
6063 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
6064 width and for different address space. Not all targets support all bit
6068 declare void @llvm.memmove.p0i8.p0i8.i32(i8* <dest>, i8* <src>,
6069 i32 <len>, i32 <align>, i1 <isvolatile>)
6070 declare void @llvm.memmove.p0i8.p0i8.i64(i8* <dest>, i8* <src>,
6071 i64 <len>, i32 <align>, i1 <isvolatile>)
6075 <p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
6076 source location to the destination location. It is similar to the
6077 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
6080 <p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
6081 intrinsics do not return a value, takes extra alignment/isvolatile arguments
6082 and the pointers can be in specified address spaces.</p>
6086 <p>The first argument is a pointer to the destination, the second is a pointer
6087 to the source. The third argument is an integer argument specifying the
6088 number of bytes to copy, the fourth argument is the alignment of the
6089 source and destination locations, and the fifth is a boolean indicating a
6090 volatile access.</p>
6092 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6093 then the caller guarantees that the source and destination pointers are
6094 aligned to that boundary.</p>
6096 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6097 <tt>llvm.memmove</tt> call is a <a href="#volatile">volatile operation</a>.
6098 The detailed access behavior is not very cleanly specified and it is unwise
6099 to depend on it.</p>
6103 <p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
6104 source location to the destination location, which may overlap. It copies
6105 "len" bytes of memory over. If the argument is known to be aligned to some
6106 boundary, this can be specified as the fourth argument, otherwise it should
6107 be set to 0 or 1.</p>
6111 <!-- _______________________________________________________________________ -->
6112 <div class="doc_subsubsection">
6113 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
6116 <div class="doc_text">
6119 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
6120 width and for different address spaces. Not all targets support all bit
6124 declare void @llvm.memset.p0i8.i32(i8* <dest>, i8 <val>,
6125 i32 <len>, i32 <align>, i1 <isvolatile>)
6126 declare void @llvm.memset.p0i8.i64(i8* <dest>, i8 <val>,
6127 i64 <len>, i32 <align>, i1 <isvolatile>)
6131 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
6132 particular byte value.</p>
6134 <p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
6135 intrinsic does not return a value, takes extra alignment/volatile arguments,
6136 and the destination can be in an arbitrary address space.</p>
6139 <p>The first argument is a pointer to the destination to fill, the second is the
6140 byte value to fill it with, the third argument is an integer argument
6141 specifying the number of bytes to fill, and the fourth argument is the known
6142 alignment of destination location.</p>
6144 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
6145 then the caller guarantees that the destination pointer is aligned to that
6148 <p>If the <tt>isvolatile</tt> parameter is <tt>true</tt>, the
6149 <tt>llvm.memset</tt> call is a <a href="#volatile">volatile operation</a>.
6150 The detailed access behavior is not very cleanly specified and it is unwise
6151 to depend on it.</p>
6154 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
6155 at the destination location. If the argument is known to be aligned to some
6156 boundary, this can be specified as the fourth argument, otherwise it should
6157 be set to 0 or 1.</p>
6161 <!-- _______________________________________________________________________ -->
6162 <div class="doc_subsubsection">
6163 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
6166 <div class="doc_text">
6169 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
6170 floating point or vector of floating point type. Not all targets support all
6174 declare float @llvm.sqrt.f32(float %Val)
6175 declare double @llvm.sqrt.f64(double %Val)
6176 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
6177 declare fp128 @llvm.sqrt.f128(fp128 %Val)
6178 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
6182 <p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
6183 returning the same value as the libm '<tt>sqrt</tt>' functions would.
6184 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
6185 behavior for negative numbers other than -0.0 (which allows for better
6186 optimization, because there is no need to worry about errno being
6187 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
6190 <p>The argument and return value are floating point numbers of the same
6194 <p>This function returns the sqrt of the specified operand if it is a
6195 nonnegative floating point number.</p>
6199 <!-- _______________________________________________________________________ -->
6200 <div class="doc_subsubsection">
6201 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
6204 <div class="doc_text">
6207 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
6208 floating point or vector of floating point type. Not all targets support all
6212 declare float @llvm.powi.f32(float %Val, i32 %power)
6213 declare double @llvm.powi.f64(double %Val, i32 %power)
6214 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
6215 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
6216 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
6220 <p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
6221 specified (positive or negative) power. The order of evaluation of
6222 multiplications is not defined. When a vector of floating point type is
6223 used, the second argument remains a scalar integer value.</p>
6226 <p>The second argument is an integer power, and the first is a value to raise to
6230 <p>This function returns the first value raised to the second power with an
6231 unspecified sequence of rounding operations.</p>
6235 <!-- _______________________________________________________________________ -->
6236 <div class="doc_subsubsection">
6237 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
6240 <div class="doc_text">
6243 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
6244 floating point or vector of floating point type. Not all targets support all
6248 declare float @llvm.sin.f32(float %Val)
6249 declare double @llvm.sin.f64(double %Val)
6250 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
6251 declare fp128 @llvm.sin.f128(fp128 %Val)
6252 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
6256 <p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
6259 <p>The argument and return value are floating point numbers of the same
6263 <p>This function returns the sine of the specified operand, returning the same
6264 values as the libm <tt>sin</tt> functions would, and handles error conditions
6265 in the same way.</p>
6269 <!-- _______________________________________________________________________ -->
6270 <div class="doc_subsubsection">
6271 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
6274 <div class="doc_text">
6277 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
6278 floating point or vector of floating point type. Not all targets support all
6282 declare float @llvm.cos.f32(float %Val)
6283 declare double @llvm.cos.f64(double %Val)
6284 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
6285 declare fp128 @llvm.cos.f128(fp128 %Val)
6286 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
6290 <p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
6293 <p>The argument and return value are floating point numbers of the same
6297 <p>This function returns the cosine of the specified operand, returning the same
6298 values as the libm <tt>cos</tt> functions would, and handles error conditions
6299 in the same way.</p>
6303 <!-- _______________________________________________________________________ -->
6304 <div class="doc_subsubsection">
6305 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
6308 <div class="doc_text">
6311 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
6312 floating point or vector of floating point type. Not all targets support all
6316 declare float @llvm.pow.f32(float %Val, float %Power)
6317 declare double @llvm.pow.f64(double %Val, double %Power)
6318 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
6319 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
6320 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
6324 <p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
6325 specified (positive or negative) power.</p>
6328 <p>The second argument is a floating point power, and the first is a value to
6329 raise to that power.</p>
6332 <p>This function returns the first value raised to the second power, returning
6333 the same values as the libm <tt>pow</tt> functions would, and handles error
6334 conditions in the same way.</p>
6338 <!-- ======================================================================= -->
6339 <div class="doc_subsection">
6340 <a name="int_manip">Bit Manipulation Intrinsics</a>
6343 <div class="doc_text">
6345 <p>LLVM provides intrinsics for a few important bit manipulation operations.
6346 These allow efficient code generation for some algorithms.</p>
6350 <!-- _______________________________________________________________________ -->
6351 <div class="doc_subsubsection">
6352 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
6355 <div class="doc_text">
6358 <p>This is an overloaded intrinsic function. You can use bswap on any integer
6359 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
6362 declare i16 @llvm.bswap.i16(i16 <id>)
6363 declare i32 @llvm.bswap.i32(i32 <id>)
6364 declare i64 @llvm.bswap.i64(i64 <id>)
6368 <p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
6369 values with an even number of bytes (positive multiple of 16 bits). These
6370 are useful for performing operations on data that is not in the target's
6371 native byte order.</p>
6374 <p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
6375 and low byte of the input i16 swapped. Similarly,
6376 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
6377 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
6378 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
6379 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
6380 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
6381 more, respectively).</p>
6385 <!-- _______________________________________________________________________ -->
6386 <div class="doc_subsubsection">
6387 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
6390 <div class="doc_text">
6393 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
6394 width. Not all targets support all bit widths however.</p>
6397 declare i8 @llvm.ctpop.i8(i8 <src>)
6398 declare i16 @llvm.ctpop.i16(i16 <src>)
6399 declare i32 @llvm.ctpop.i32(i32 <src>)
6400 declare i64 @llvm.ctpop.i64(i64 <src>)
6401 declare i256 @llvm.ctpop.i256(i256 <src>)
6405 <p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
6409 <p>The only argument is the value to be counted. The argument may be of any
6410 integer type. The return type must match the argument type.</p>
6413 <p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.</p>
6417 <!-- _______________________________________________________________________ -->
6418 <div class="doc_subsubsection">
6419 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
6422 <div class="doc_text">
6425 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
6426 integer bit width. Not all targets support all bit widths however.</p>
6429 declare i8 @llvm.ctlz.i8 (i8 <src>)
6430 declare i16 @llvm.ctlz.i16(i16 <src>)
6431 declare i32 @llvm.ctlz.i32(i32 <src>)
6432 declare i64 @llvm.ctlz.i64(i64 <src>)
6433 declare i256 @llvm.ctlz.i256(i256 <src>)
6437 <p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
6438 leading zeros in a variable.</p>
6441 <p>The only argument is the value to be counted. The argument may be of any
6442 integer type. The return type must match the argument type.</p>
6445 <p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
6446 zeros in a variable. If the src == 0 then the result is the size in bits of
6447 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
6451 <!-- _______________________________________________________________________ -->
6452 <div class="doc_subsubsection">
6453 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
6456 <div class="doc_text">
6459 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
6460 integer bit width. Not all targets support all bit widths however.</p>
6463 declare i8 @llvm.cttz.i8 (i8 <src>)
6464 declare i16 @llvm.cttz.i16(i16 <src>)
6465 declare i32 @llvm.cttz.i32(i32 <src>)
6466 declare i64 @llvm.cttz.i64(i64 <src>)
6467 declare i256 @llvm.cttz.i256(i256 <src>)
6471 <p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
6475 <p>The only argument is the value to be counted. The argument may be of any
6476 integer type. The return type must match the argument type.</p>
6479 <p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
6480 zeros in a variable. If the src == 0 then the result is the size in bits of
6481 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
6485 <!-- ======================================================================= -->
6486 <div class="doc_subsection">
6487 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
6490 <div class="doc_text">
6492 <p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
6496 <!-- _______________________________________________________________________ -->
6497 <div class="doc_subsubsection">
6498 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
6501 <div class="doc_text">
6504 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
6505 on any integer bit width.</p>
6508 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
6509 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6510 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6514 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6515 a signed addition of the two arguments, and indicate whether an overflow
6516 occurred during the signed summation.</p>
6519 <p>The arguments (%a and %b) and the first element of the result structure may
6520 be of integer types of any bit width, but they must have the same bit
6521 width. The second element of the result structure must be of
6522 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6523 undergo signed addition.</p>
6526 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6527 a signed addition of the two variables. They return a structure — the
6528 first element of which is the signed summation, and the second element of
6529 which is a bit specifying if the signed summation resulted in an
6534 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6535 %sum = extractvalue {i32, i1} %res, 0
6536 %obit = extractvalue {i32, i1} %res, 1
6537 br i1 %obit, label %overflow, label %normal
6542 <!-- _______________________________________________________________________ -->
6543 <div class="doc_subsubsection">
6544 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
6547 <div class="doc_text">
6550 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
6551 on any integer bit width.</p>
6554 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6555 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6556 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6560 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6561 an unsigned addition of the two arguments, and indicate whether a carry
6562 occurred during the unsigned summation.</p>
6565 <p>The arguments (%a and %b) and the first element of the result structure may
6566 be of integer types of any bit width, but they must have the same bit
6567 width. The second element of the result structure must be of
6568 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6569 undergo unsigned addition.</p>
6572 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6573 an unsigned addition of the two arguments. They return a structure —
6574 the first element of which is the sum, and the second element of which is a
6575 bit specifying if the unsigned summation resulted in a carry.</p>
6579 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6580 %sum = extractvalue {i32, i1} %res, 0
6581 %obit = extractvalue {i32, i1} %res, 1
6582 br i1 %obit, label %carry, label %normal
6587 <!-- _______________________________________________________________________ -->
6588 <div class="doc_subsubsection">
6589 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
6592 <div class="doc_text">
6595 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
6596 on any integer bit width.</p>
6599 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6600 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6601 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6605 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6606 a signed subtraction of the two arguments, and indicate whether an overflow
6607 occurred during the signed subtraction.</p>
6610 <p>The arguments (%a and %b) and the first element of the result structure may
6611 be of integer types of any bit width, but they must have the same bit
6612 width. The second element of the result structure must be of
6613 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6614 undergo signed subtraction.</p>
6617 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6618 a signed subtraction of the two arguments. They return a structure —
6619 the first element of which is the subtraction, and the second element of
6620 which is a bit specifying if the signed subtraction resulted in an
6625 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6626 %sum = extractvalue {i32, i1} %res, 0
6627 %obit = extractvalue {i32, i1} %res, 1
6628 br i1 %obit, label %overflow, label %normal
6633 <!-- _______________________________________________________________________ -->
6634 <div class="doc_subsubsection">
6635 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
6638 <div class="doc_text">
6641 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
6642 on any integer bit width.</p>
6645 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6646 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6647 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6651 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6652 an unsigned subtraction of the two arguments, and indicate whether an
6653 overflow occurred during the unsigned subtraction.</p>
6656 <p>The arguments (%a and %b) and the first element of the result structure may
6657 be of integer types of any bit width, but they must have the same bit
6658 width. The second element of the result structure must be of
6659 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6660 undergo unsigned subtraction.</p>
6663 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6664 an unsigned subtraction of the two arguments. They return a structure —
6665 the first element of which is the subtraction, and the second element of
6666 which is a bit specifying if the unsigned subtraction resulted in an
6671 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6672 %sum = extractvalue {i32, i1} %res, 0
6673 %obit = extractvalue {i32, i1} %res, 1
6674 br i1 %obit, label %overflow, label %normal
6679 <!-- _______________________________________________________________________ -->
6680 <div class="doc_subsubsection">
6681 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
6684 <div class="doc_text">
6687 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
6688 on any integer bit width.</p>
6691 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6692 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6693 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6698 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6699 a signed multiplication of the two arguments, and indicate whether an
6700 overflow occurred during the signed multiplication.</p>
6703 <p>The arguments (%a and %b) and the first element of the result structure may
6704 be of integer types of any bit width, but they must have the same bit
6705 width. The second element of the result structure must be of
6706 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6707 undergo signed multiplication.</p>
6710 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6711 a signed multiplication of the two arguments. They return a structure —
6712 the first element of which is the multiplication, and the second element of
6713 which is a bit specifying if the signed multiplication resulted in an
6718 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6719 %sum = extractvalue {i32, i1} %res, 0
6720 %obit = extractvalue {i32, i1} %res, 1
6721 br i1 %obit, label %overflow, label %normal
6726 <!-- _______________________________________________________________________ -->
6727 <div class="doc_subsubsection">
6728 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6731 <div class="doc_text">
6734 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6735 on any integer bit width.</p>
6738 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6739 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6740 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6744 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6745 a unsigned multiplication of the two arguments, and indicate whether an
6746 overflow occurred during the unsigned multiplication.</p>
6749 <p>The arguments (%a and %b) and the first element of the result structure may
6750 be of integer types of any bit width, but they must have the same bit
6751 width. The second element of the result structure must be of
6752 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6753 undergo unsigned multiplication.</p>
6756 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6757 an unsigned multiplication of the two arguments. They return a structure
6758 — the first element of which is the multiplication, and the second
6759 element of which is a bit specifying if the unsigned multiplication resulted
6764 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6765 %sum = extractvalue {i32, i1} %res, 0
6766 %obit = extractvalue {i32, i1} %res, 1
6767 br i1 %obit, label %overflow, label %normal
6772 <!-- ======================================================================= -->
6773 <div class="doc_subsection">
6774 <a name="int_fp16">Half Precision Floating Point Intrinsics</a>
6777 <div class="doc_text">
6779 <p>Half precision floating point is a storage-only format. This means that it is
6780 a dense encoding (in memory) but does not support computation in the
6783 <p>This means that code must first load the half-precision floating point
6784 value as an i16, then convert it to float with <a
6785 href="#int_convert_from_fp16"><tt>llvm.convert.from.fp16</tt></a>.
6786 Computation can then be performed on the float value (including extending to
6787 double etc). To store the value back to memory, it is first converted to
6788 float if needed, then converted to i16 with
6789 <a href="#int_convert_to_fp16"><tt>llvm.convert.to.fp16</tt></a>, then
6790 storing as an i16 value.</p>
6793 <!-- _______________________________________________________________________ -->
6794 <div class="doc_subsubsection">
6795 <a name="int_convert_to_fp16">'<tt>llvm.convert.to.fp16</tt>' Intrinsic</a>
6798 <div class="doc_text">
6802 declare i16 @llvm.convert.to.fp16(f32 %a)
6806 <p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
6807 a conversion from single precision floating point format to half precision
6808 floating point format.</p>
6811 <p>The intrinsic function contains single argument - the value to be
6815 <p>The '<tt>llvm.convert.to.fp16</tt>' intrinsic function performs
6816 a conversion from single precision floating point format to half precision
6817 floating point format. The return value is an <tt>i16</tt> which
6818 contains the converted number.</p>
6822 %res = call i16 @llvm.convert.to.fp16(f32 %a)
6823 store i16 %res, i16* @x, align 2
6828 <!-- _______________________________________________________________________ -->
6829 <div class="doc_subsubsection">
6830 <a name="int_convert_from_fp16">'<tt>llvm.convert.from.fp16</tt>' Intrinsic</a>
6833 <div class="doc_text">
6837 declare f32 @llvm.convert.from.fp16(i16 %a)
6841 <p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs
6842 a conversion from half precision floating point format to single precision
6843 floating point format.</p>
6846 <p>The intrinsic function contains single argument - the value to be
6850 <p>The '<tt>llvm.convert.from.fp16</tt>' intrinsic function performs a
6851 conversion from half single precision floating point format to single
6852 precision floating point format. The input half-float value is represented by
6853 an <tt>i16</tt> value.</p>
6857 %a = load i16* @x, align 2
6858 %res = call f32 @llvm.convert.from.fp16(i16 %a)
6863 <!-- ======================================================================= -->
6864 <div class="doc_subsection">
6865 <a name="int_debugger">Debugger Intrinsics</a>
6868 <div class="doc_text">
6870 <p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
6871 prefix), are described in
6872 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
6873 Level Debugging</a> document.</p>
6877 <!-- ======================================================================= -->
6878 <div class="doc_subsection">
6879 <a name="int_eh">Exception Handling Intrinsics</a>
6882 <div class="doc_text">
6884 <p>The LLVM exception handling intrinsics (which all start with
6885 <tt>llvm.eh.</tt> prefix), are described in
6886 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6887 Handling</a> document.</p>
6891 <!-- ======================================================================= -->
6892 <div class="doc_subsection">
6893 <a name="int_trampoline">Trampoline Intrinsic</a>
6896 <div class="doc_text">
6898 <p>This intrinsic makes it possible to excise one parameter, marked with
6899 the <a href="#nest"><tt>nest</tt></a> attribute, from a function.
6900 The result is a callable
6901 function pointer lacking the nest parameter - the caller does not need to
6902 provide a value for it. Instead, the value to use is stored in advance in a
6903 "trampoline", a block of memory usually allocated on the stack, which also
6904 contains code to splice the nest value into the argument list. This is used
6905 to implement the GCC nested function address extension.</p>
6907 <p>For example, if the function is
6908 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
6909 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
6912 <pre class="doc_code">
6913 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6914 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6915 %p = call i8* @llvm.init.trampoline(i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval)
6916 %fp = bitcast i8* %p to i32 (i32, i32)*
6919 <p>The call <tt>%val = call i32 %fp(i32 %x, i32 %y)</tt> is then equivalent
6920 to <tt>%val = call i32 %f(i8* %nval, i32 %x, i32 %y)</tt>.</p>
6924 <!-- _______________________________________________________________________ -->
6925 <div class="doc_subsubsection">
6926 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6929 <div class="doc_text">
6933 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
6937 <p>This fills the memory pointed to by <tt>tramp</tt> with code and returns a
6938 function pointer suitable for executing it.</p>
6941 <p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6942 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
6943 sufficiently aligned block of memory; this memory is written to by the
6944 intrinsic. Note that the size and the alignment are target-specific - LLVM
6945 currently provides no portable way of determining them, so a front-end that
6946 generates this intrinsic needs to have some target-specific knowledge.
6947 The <tt>func</tt> argument must hold a function bitcast to
6948 an <tt>i8*</tt>.</p>
6951 <p>The block of memory pointed to by <tt>tramp</tt> is filled with target
6952 dependent code, turning it into a function. A pointer to this function is
6953 returned, but needs to be bitcast to an <a href="#int_trampoline">appropriate
6954 function pointer type</a> before being called. The new function's signature
6955 is the same as that of <tt>func</tt> with any arguments marked with
6956 the <tt>nest</tt> attribute removed. At most one such <tt>nest</tt> argument
6957 is allowed, and it must be of pointer type. Calling the new function is
6958 equivalent to calling <tt>func</tt> with the same argument list, but
6959 with <tt>nval</tt> used for the missing <tt>nest</tt> argument. If, after
6960 calling <tt>llvm.init.trampoline</tt>, the memory pointed to
6961 by <tt>tramp</tt> is modified, then the effect of any later call to the
6962 returned function pointer is undefined.</p>
6966 <!-- ======================================================================= -->
6967 <div class="doc_subsection">
6968 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6971 <div class="doc_text">
6973 <p>These intrinsic functions expand the "universal IR" of LLVM to represent
6974 hardware constructs for atomic operations and memory synchronization. This
6975 provides an interface to the hardware, not an interface to the programmer. It
6976 is aimed at a low enough level to allow any programming models or APIs
6977 (Application Programming Interfaces) which need atomic behaviors to map
6978 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
6979 hardware provides a "universal IR" for source languages, it also provides a
6980 starting point for developing a "universal" atomic operation and
6981 synchronization IR.</p>
6983 <p>These do <em>not</em> form an API such as high-level threading libraries,
6984 software transaction memory systems, atomic primitives, and intrinsic
6985 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6986 application libraries. The hardware interface provided by LLVM should allow
6987 a clean implementation of all of these APIs and parallel programming models.
6988 No one model or paradigm should be selected above others unless the hardware
6989 itself ubiquitously does so.</p>
6993 <!-- _______________________________________________________________________ -->
6994 <div class="doc_subsubsection">
6995 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6997 <div class="doc_text">
7000 declare void @llvm.memory.barrier(i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>, i1 <device>)
7004 <p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
7005 specific pairs of memory access types.</p>
7008 <p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
7009 The first four arguments enables a specific barrier as listed below. The
7010 fifth argument specifies that the barrier applies to io or device or uncached
7014 <li><tt>ll</tt>: load-load barrier</li>
7015 <li><tt>ls</tt>: load-store barrier</li>
7016 <li><tt>sl</tt>: store-load barrier</li>
7017 <li><tt>ss</tt>: store-store barrier</li>
7018 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
7022 <p>This intrinsic causes the system to enforce some ordering constraints upon
7023 the loads and stores of the program. This barrier does not
7024 indicate <em>when</em> any events will occur, it only enforces
7025 an <em>order</em> in which they occur. For any of the specified pairs of load
7026 and store operations (f.ex. load-load, or store-load), all of the first
7027 operations preceding the barrier will complete before any of the second
7028 operations succeeding the barrier begin. Specifically the semantics for each
7029 pairing is as follows:</p>
7032 <li><tt>ll</tt>: All loads before the barrier must complete before any load
7033 after the barrier begins.</li>
7034 <li><tt>ls</tt>: All loads before the barrier must complete before any
7035 store after the barrier begins.</li>
7036 <li><tt>ss</tt>: All stores before the barrier must complete before any
7037 store after the barrier begins.</li>
7038 <li><tt>sl</tt>: All stores before the barrier must complete before any
7039 load after the barrier begins.</li>
7042 <p>These semantics are applied with a logical "and" behavior when more than one
7043 is enabled in a single memory barrier intrinsic.</p>
7045 <p>Backends may implement stronger barriers than those requested when they do
7046 not support as fine grained a barrier as requested. Some architectures do
7047 not need all types of barriers and on such architectures, these become
7052 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7053 %ptr = bitcast i8* %mallocP to i32*
7056 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
7057 call void @llvm.memory.barrier(i1 false, i1 true, i1 false, i1 false)
7058 <i>; guarantee the above finishes</i>
7059 store i32 8, %ptr <i>; before this begins</i>
7064 <!-- _______________________________________________________________________ -->
7065 <div class="doc_subsubsection">
7066 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
7069 <div class="doc_text">
7072 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
7073 any integer bit width and for different address spaces. Not all targets
7074 support all bit widths however.</p>
7077 declare i8 @llvm.atomic.cmp.swap.i8.p0i8(i8* <ptr>, i8 <cmp>, i8 <val>)
7078 declare i16 @llvm.atomic.cmp.swap.i16.p0i16(i16* <ptr>, i16 <cmp>, i16 <val>)
7079 declare i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* <ptr>, i32 <cmp>, i32 <val>)
7080 declare i64 @llvm.atomic.cmp.swap.i64.p0i64(i64* <ptr>, i64 <cmp>, i64 <val>)
7084 <p>This loads a value in memory and compares it to a given value. If they are
7085 equal, it stores a new value into the memory.</p>
7088 <p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
7089 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
7090 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
7091 this integer type. While any bit width integer may be used, targets may only
7092 lower representations they support in hardware.</p>
7095 <p>This entire intrinsic must be executed atomically. It first loads the value
7096 in memory pointed to by <tt>ptr</tt> and compares it with the
7097 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
7098 memory. The loaded value is yielded in all cases. This provides the
7099 equivalent of an atomic compare-and-swap operation within the SSA
7104 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7105 %ptr = bitcast i8* %mallocP to i32*
7108 %val1 = add i32 4, 4
7109 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 4, %val1)
7110 <i>; yields {i32}:result1 = 4</i>
7111 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7112 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7114 %val2 = add i32 1, 1
7115 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32(i32* %ptr, i32 5, %val2)
7116 <i>; yields {i32}:result2 = 8</i>
7117 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
7119 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
7124 <!-- _______________________________________________________________________ -->
7125 <div class="doc_subsubsection">
7126 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
7128 <div class="doc_text">
7131 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
7132 integer bit width. Not all targets support all bit widths however.</p>
7135 declare i8 @llvm.atomic.swap.i8.p0i8(i8* <ptr>, i8 <val>)
7136 declare i16 @llvm.atomic.swap.i16.p0i16(i16* <ptr>, i16 <val>)
7137 declare i32 @llvm.atomic.swap.i32.p0i32(i32* <ptr>, i32 <val>)
7138 declare i64 @llvm.atomic.swap.i64.p0i64(i64* <ptr>, i64 <val>)
7142 <p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
7143 the value from memory. It then stores the value in <tt>val</tt> in the memory
7144 at <tt>ptr</tt>.</p>
7147 <p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
7148 the <tt>val</tt> argument and the result must be integers of the same bit
7149 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
7150 integer type. The targets may only lower integer representations they
7154 <p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
7155 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
7156 equivalent of an atomic swap operation within the SSA framework.</p>
7160 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7161 %ptr = bitcast i8* %mallocP to i32*
7164 %val1 = add i32 4, 4
7165 %result1 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val1)
7166 <i>; yields {i32}:result1 = 4</i>
7167 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
7168 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
7170 %val2 = add i32 1, 1
7171 %result2 = call i32 @llvm.atomic.swap.i32.p0i32(i32* %ptr, i32 %val2)
7172 <i>; yields {i32}:result2 = 8</i>
7174 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
7175 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
7180 <!-- _______________________________________________________________________ -->
7181 <div class="doc_subsubsection">
7182 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
7186 <div class="doc_text">
7189 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
7190 any integer bit width. Not all targets support all bit widths however.</p>
7193 declare i8 @llvm.atomic.load.add.i8.p0i8(i8* <ptr>, i8 <delta>)
7194 declare i16 @llvm.atomic.load.add.i16.p0i16(i16* <ptr>, i16 <delta>)
7195 declare i32 @llvm.atomic.load.add.i32.p0i32(i32* <ptr>, i32 <delta>)
7196 declare i64 @llvm.atomic.load.add.i64.p0i64(i64* <ptr>, i64 <delta>)
7200 <p>This intrinsic adds <tt>delta</tt> to the value stored in memory
7201 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7204 <p>The intrinsic takes two arguments, the first a pointer to an integer value
7205 and the second an integer value. The result is also an integer value. These
7206 integer types can have any bit width, but they must all have the same bit
7207 width. The targets may only lower integer representations they support.</p>
7210 <p>This intrinsic does a series of operations atomically. It first loads the
7211 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
7212 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
7216 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7217 %ptr = bitcast i8* %mallocP to i32*
7219 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 4)
7220 <i>; yields {i32}:result1 = 4</i>
7221 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 2)
7222 <i>; yields {i32}:result2 = 8</i>
7223 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32(i32* %ptr, i32 5)
7224 <i>; yields {i32}:result3 = 10</i>
7225 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
7230 <!-- _______________________________________________________________________ -->
7231 <div class="doc_subsubsection">
7232 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
7236 <div class="doc_text">
7239 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
7240 any integer bit width and for different address spaces. Not all targets
7241 support all bit widths however.</p>
7244 declare i8 @llvm.atomic.load.sub.i8.p0i32(i8* <ptr>, i8 <delta>)
7245 declare i16 @llvm.atomic.load.sub.i16.p0i32(i16* <ptr>, i16 <delta>)
7246 declare i32 @llvm.atomic.load.sub.i32.p0i32(i32* <ptr>, i32 <delta>)
7247 declare i64 @llvm.atomic.load.sub.i64.p0i32(i64* <ptr>, i64 <delta>)
7251 <p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
7252 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
7255 <p>The intrinsic takes two arguments, the first a pointer to an integer value
7256 and the second an integer value. The result is also an integer value. These
7257 integer types can have any bit width, but they must all have the same bit
7258 width. The targets may only lower integer representations they support.</p>
7261 <p>This intrinsic does a series of operations atomically. It first loads the
7262 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
7263 result to <tt>ptr</tt>. It yields the original value stored
7264 at <tt>ptr</tt>.</p>
7268 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7269 %ptr = bitcast i8* %mallocP to i32*
7271 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 4)
7272 <i>; yields {i32}:result1 = 8</i>
7273 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 2)
7274 <i>; yields {i32}:result2 = 4</i>
7275 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %ptr, i32 5)
7276 <i>; yields {i32}:result3 = 2</i>
7277 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
7282 <!-- _______________________________________________________________________ -->
7283 <div class="doc_subsubsection">
7284 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
7285 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
7286 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
7287 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
7290 <div class="doc_text">
7293 <p>These are overloaded intrinsics. You can
7294 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
7295 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
7296 bit width and for different address spaces. Not all targets support all bit
7300 declare i8 @llvm.atomic.load.and.i8.p0i8(i8* <ptr>, i8 <delta>)
7301 declare i16 @llvm.atomic.load.and.i16.p0i16(i16* <ptr>, i16 <delta>)
7302 declare i32 @llvm.atomic.load.and.i32.p0i32(i32* <ptr>, i32 <delta>)
7303 declare i64 @llvm.atomic.load.and.i64.p0i64(i64* <ptr>, i64 <delta>)
7307 declare i8 @llvm.atomic.load.or.i8.p0i8(i8* <ptr>, i8 <delta>)
7308 declare i16 @llvm.atomic.load.or.i16.p0i16(i16* <ptr>, i16 <delta>)
7309 declare i32 @llvm.atomic.load.or.i32.p0i32(i32* <ptr>, i32 <delta>)
7310 declare i64 @llvm.atomic.load.or.i64.p0i64(i64* <ptr>, i64 <delta>)
7314 declare i8 @llvm.atomic.load.nand.i8.p0i32(i8* <ptr>, i8 <delta>)
7315 declare i16 @llvm.atomic.load.nand.i16.p0i32(i16* <ptr>, i16 <delta>)
7316 declare i32 @llvm.atomic.load.nand.i32.p0i32(i32* <ptr>, i32 <delta>)
7317 declare i64 @llvm.atomic.load.nand.i64.p0i32(i64* <ptr>, i64 <delta>)
7321 declare i8 @llvm.atomic.load.xor.i8.p0i32(i8* <ptr>, i8 <delta>)
7322 declare i16 @llvm.atomic.load.xor.i16.p0i32(i16* <ptr>, i16 <delta>)
7323 declare i32 @llvm.atomic.load.xor.i32.p0i32(i32* <ptr>, i32 <delta>)
7324 declare i64 @llvm.atomic.load.xor.i64.p0i32(i64* <ptr>, i64 <delta>)
7328 <p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
7329 the value stored in memory at <tt>ptr</tt>. It yields the original value
7330 at <tt>ptr</tt>.</p>
7333 <p>These intrinsics take two arguments, the first a pointer to an integer value
7334 and the second an integer value. The result is also an integer value. These
7335 integer types can have any bit width, but they must all have the same bit
7336 width. The targets may only lower integer representations they support.</p>
7339 <p>These intrinsics does a series of operations atomically. They first load the
7340 value stored at <tt>ptr</tt>. They then do the bitwise
7341 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
7342 original value stored at <tt>ptr</tt>.</p>
7346 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7347 %ptr = bitcast i8* %mallocP to i32*
7348 store i32 0x0F0F, %ptr
7349 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32(i32* %ptr, i32 0xFF)
7350 <i>; yields {i32}:result0 = 0x0F0F</i>
7351 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32(i32* %ptr, i32 0xFF)
7352 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
7353 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32(i32* %ptr, i32 0F)
7354 <i>; yields {i32}:result2 = 0xF0</i>
7355 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32(i32* %ptr, i32 0F)
7356 <i>; yields {i32}:result3 = FF</i>
7357 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
7362 <!-- _______________________________________________________________________ -->
7363 <div class="doc_subsubsection">
7364 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
7365 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
7366 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
7367 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
7370 <div class="doc_text">
7373 <p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
7374 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
7375 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
7376 address spaces. Not all targets support all bit widths however.</p>
7379 declare i8 @llvm.atomic.load.max.i8.p0i8(i8* <ptr>, i8 <delta>)
7380 declare i16 @llvm.atomic.load.max.i16.p0i16(i16* <ptr>, i16 <delta>)
7381 declare i32 @llvm.atomic.load.max.i32.p0i32(i32* <ptr>, i32 <delta>)
7382 declare i64 @llvm.atomic.load.max.i64.p0i64(i64* <ptr>, i64 <delta>)
7386 declare i8 @llvm.atomic.load.min.i8.p0i8(i8* <ptr>, i8 <delta>)
7387 declare i16 @llvm.atomic.load.min.i16.p0i16(i16* <ptr>, i16 <delta>)
7388 declare i32 @llvm.atomic.load.min.i32.p0i32(i32* <ptr>, i32 <delta>)
7389 declare i64 @llvm.atomic.load.min.i64.p0i64(i64* <ptr>, i64 <delta>)
7393 declare i8 @llvm.atomic.load.umax.i8.p0i8(i8* <ptr>, i8 <delta>)
7394 declare i16 @llvm.atomic.load.umax.i16.p0i16(i16* <ptr>, i16 <delta>)
7395 declare i32 @llvm.atomic.load.umax.i32.p0i32(i32* <ptr>, i32 <delta>)
7396 declare i64 @llvm.atomic.load.umax.i64.p0i64(i64* <ptr>, i64 <delta>)
7400 declare i8 @llvm.atomic.load.umin.i8.p0i8(i8* <ptr>, i8 <delta>)
7401 declare i16 @llvm.atomic.load.umin.i16.p0i16(i16* <ptr>, i16 <delta>)
7402 declare i32 @llvm.atomic.load.umin.i32.p0i32(i32* <ptr>, i32 <delta>)
7403 declare i64 @llvm.atomic.load.umin.i64.p0i64(i64* <ptr>, i64 <delta>)
7407 <p>These intrinsics takes the signed or unsigned minimum or maximum of
7408 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
7409 original value at <tt>ptr</tt>.</p>
7412 <p>These intrinsics take two arguments, the first a pointer to an integer value
7413 and the second an integer value. The result is also an integer value. These
7414 integer types can have any bit width, but they must all have the same bit
7415 width. The targets may only lower integer representations they support.</p>
7418 <p>These intrinsics does a series of operations atomically. They first load the
7419 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
7420 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
7421 yield the original value stored at <tt>ptr</tt>.</p>
7425 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
7426 %ptr = bitcast i8* %mallocP to i32*
7428 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32(i32* %ptr, i32 -2)
7429 <i>; yields {i32}:result0 = 7</i>
7430 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32(i32* %ptr, i32 8)
7431 <i>; yields {i32}:result1 = -2</i>
7432 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32(i32* %ptr, i32 10)
7433 <i>; yields {i32}:result2 = 8</i>
7434 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32(i32* %ptr, i32 30)
7435 <i>; yields {i32}:result3 = 8</i>
7436 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
7442 <!-- ======================================================================= -->
7443 <div class="doc_subsection">
7444 <a name="int_memorymarkers">Memory Use Markers</a>
7447 <div class="doc_text">
7449 <p>This class of intrinsics exists to information about the lifetime of memory
7450 objects and ranges where variables are immutable.</p>
7454 <!-- _______________________________________________________________________ -->
7455 <div class="doc_subsubsection">
7456 <a name="int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a>
7459 <div class="doc_text">
7463 declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
7467 <p>The '<tt>llvm.lifetime.start</tt>' intrinsic specifies the start of a memory
7468 object's lifetime.</p>
7471 <p>The first argument is a constant integer representing the size of the
7472 object, or -1 if it is variable sized. The second argument is a pointer to
7476 <p>This intrinsic indicates that before this point in the code, the value of the
7477 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7478 never be used and has an undefined value. A load from the pointer that
7479 precedes this intrinsic can be replaced with
7480 <tt>'<a href="#undefvalues">undef</a>'</tt>.</p>
7484 <!-- _______________________________________________________________________ -->
7485 <div class="doc_subsubsection">
7486 <a name="int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a>
7489 <div class="doc_text">
7493 declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
7497 <p>The '<tt>llvm.lifetime.end</tt>' intrinsic specifies the end of a memory
7498 object's lifetime.</p>
7501 <p>The first argument is a constant integer representing the size of the
7502 object, or -1 if it is variable sized. The second argument is a pointer to
7506 <p>This intrinsic indicates that after this point in the code, the value of the
7507 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7508 never be used and has an undefined value. Any stores into the memory object
7509 following this intrinsic may be removed as dead.
7513 <!-- _______________________________________________________________________ -->
7514 <div class="doc_subsubsection">
7515 <a name="int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a>
7518 <div class="doc_text">
7522 declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>) readonly
7526 <p>The '<tt>llvm.invariant.start</tt>' intrinsic specifies that the contents of
7527 a memory object will not change.</p>
7530 <p>The first argument is a constant integer representing the size of the
7531 object, or -1 if it is variable sized. The second argument is a pointer to
7535 <p>This intrinsic indicates that until an <tt>llvm.invariant.end</tt> that uses
7536 the return value, the referenced memory location is constant and
7541 <!-- _______________________________________________________________________ -->
7542 <div class="doc_subsubsection">
7543 <a name="int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a>
7546 <div class="doc_text">
7550 declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
7554 <p>The '<tt>llvm.invariant.end</tt>' intrinsic specifies that the contents of
7555 a memory object are mutable.</p>
7558 <p>The first argument is the matching <tt>llvm.invariant.start</tt> intrinsic.
7559 The second argument is a constant integer representing the size of the
7560 object, or -1 if it is variable sized and the third argument is a pointer
7564 <p>This intrinsic indicates that the memory is mutable again.</p>
7568 <!-- ======================================================================= -->
7569 <div class="doc_subsection">
7570 <a name="int_general">General Intrinsics</a>
7573 <div class="doc_text">
7575 <p>This class of intrinsics is designed to be generic and has no specific
7580 <!-- _______________________________________________________________________ -->
7581 <div class="doc_subsubsection">
7582 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
7585 <div class="doc_text">
7589 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int>)
7593 <p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
7596 <p>The first argument is a pointer to a value, the second is a pointer to a
7597 global string, the third is a pointer to a global string which is the source
7598 file name, and the last argument is the line number.</p>
7601 <p>This intrinsic allows annotation of local variables with arbitrary strings.
7602 This can be useful for special purpose optimizations that want to look for
7603 these annotations. These have no other defined use, they are ignored by code
7604 generation and optimization.</p>
7608 <!-- _______________________________________________________________________ -->
7609 <div class="doc_subsubsection">
7610 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
7613 <div class="doc_text">
7616 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
7617 any integer bit width.</p>
7620 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int>)
7621 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int>)
7622 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int>)
7623 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int>)
7624 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int>)
7628 <p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
7631 <p>The first argument is an integer value (result of some expression), the
7632 second is a pointer to a global string, the third is a pointer to a global
7633 string which is the source file name, and the last argument is the line
7634 number. It returns the value of the first argument.</p>
7637 <p>This intrinsic allows annotations to be put on arbitrary expressions with
7638 arbitrary strings. This can be useful for special purpose optimizations that
7639 want to look for these annotations. These have no other defined use, they
7640 are ignored by code generation and optimization.</p>
7644 <!-- _______________________________________________________________________ -->
7645 <div class="doc_subsubsection">
7646 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
7649 <div class="doc_text">
7653 declare void @llvm.trap()
7657 <p>The '<tt>llvm.trap</tt>' intrinsic.</p>
7663 <p>This intrinsics is lowered to the target dependent trap instruction. If the
7664 target does not have a trap instruction, this intrinsic will be lowered to
7665 the call of the <tt>abort()</tt> function.</p>
7669 <!-- _______________________________________________________________________ -->
7670 <div class="doc_subsubsection">
7671 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
7674 <div class="doc_text">
7678 declare void @llvm.stackprotector(i8* <guard>, i8** <slot>)
7682 <p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
7683 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
7684 ensure that it is placed on the stack before local variables.</p>
7687 <p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
7688 arguments. The first argument is the value loaded from the stack
7689 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
7690 that has enough space to hold the value of the guard.</p>
7693 <p>This intrinsic causes the prologue/epilogue inserter to force the position of
7694 the <tt>AllocaInst</tt> stack slot to be before local variables on the
7695 stack. This is to ensure that if a local variable on the stack is
7696 overwritten, it will destroy the value of the guard. When the function exits,
7697 the guard on the stack is checked against the original guard. If they're
7698 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
7703 <!-- _______________________________________________________________________ -->
7704 <div class="doc_subsubsection">
7705 <a name="int_objectsize">'<tt>llvm.objectsize</tt>' Intrinsic</a>
7708 <div class="doc_text">
7712 declare i32 @llvm.objectsize.i32(i8* <object>, i1 <type>)
7713 declare i64 @llvm.objectsize.i64(i8* <object>, i1 <type>)
7717 <p>The <tt>llvm.objectsize</tt> intrinsic is designed to provide information
7718 to the optimizers to discover at compile time either a) when an
7719 operation like memcpy will either overflow a buffer that corresponds to
7720 an object, or b) to determine that a runtime check for overflow isn't
7721 necessary. An object in this context means an allocation of a
7722 specific class, structure, array, or other object.</p>
7725 <p>The <tt>llvm.objectsize</tt> intrinsic takes two arguments. The first
7726 argument is a pointer to or into the <tt>object</tt>. The second argument
7727 is a boolean 0 or 1. This argument determines whether you want the
7728 maximum (0) or minimum (1) bytes remaining. This needs to be a literal 0 or
7729 1, variables are not allowed.</p>
7732 <p>The <tt>llvm.objectsize</tt> intrinsic is lowered to either a constant
7733 representing the size of the object concerned or <tt>i32/i64 -1 or 0</tt>
7734 (depending on the <tt>type</tt> argument if the size cannot be determined
7735 at compile time.</p>
7739 <!-- *********************************************************************** -->
7742 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
7743 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
7744 <a href="http://validator.w3.org/check/referer"><img
7745 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
7747 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
7748 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
7749 Last modified: $Date$