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_internal">'<tt>internal</tt>' Linkage</a></li>
28 <li><a href="#linkage_available_externally">'<tt>available_externally</tt>' Linkage</a></li>
29 <li><a href="#linkage_linkonce">'<tt>linkonce</tt>' Linkage</a></li>
30 <li><a href="#linkage_common">'<tt>common</tt>' Linkage</a></li>
31 <li><a href="#linkage_weak">'<tt>weak</tt>' Linkage</a></li>
32 <li><a href="#linkage_appending">'<tt>appending</tt>' Linkage</a></li>
33 <li><a href="#linkage_externweak">'<tt>extern_weak</tt>' Linkage</a></li>
34 <li><a href="#linkage_linkonce_odr">'<tt>linkonce_odr</tt>' Linkage</a></li>
35 <li><a href="#linkage_weak">'<tt>weak_odr</tt>' Linkage</a></li>
36 <li><a href="#linkage_external">'<tt>externally visible</tt>' Linkage</a></li>
37 <li><a href="#linkage_dllimport">'<tt>dllimport</tt>' Linkage</a></li>
38 <li><a href="#linkage_dllexport">'<tt>dllexport</tt>' Linkage</a></li>
41 <li><a href="#callingconv">Calling Conventions</a></li>
42 <li><a href="#namedtypes">Named Types</a></li>
43 <li><a href="#globalvars">Global Variables</a></li>
44 <li><a href="#functionstructure">Functions</a></li>
45 <li><a href="#aliasstructure">Aliases</a></li>
46 <li><a href="#paramattrs">Parameter Attributes</a></li>
47 <li><a href="#fnattrs">Function Attributes</a></li>
48 <li><a href="#gc">Garbage Collector Names</a></li>
49 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
50 <li><a href="#datalayout">Data Layout</a></li>
51 <li><a href="#pointeraliasing">Pointer Aliasing Rules</a></li>
54 <li><a href="#typesystem">Type System</a>
56 <li><a href="#t_classifications">Type Classifications</a></li>
57 <li><a href="#t_primitive">Primitive Types</a>
59 <li><a href="#t_integer">Integer Type</a></li>
60 <li><a href="#t_floating">Floating Point Types</a></li>
61 <li><a href="#t_void">Void Type</a></li>
62 <li><a href="#t_label">Label Type</a></li>
63 <li><a href="#t_metadata">Metadata Type</a></li>
66 <li><a href="#t_derived">Derived Types</a>
68 <li><a href="#t_array">Array Type</a></li>
69 <li><a href="#t_function">Function Type</a></li>
70 <li><a href="#t_pointer">Pointer Type</a></li>
71 <li><a href="#t_struct">Structure Type</a></li>
72 <li><a href="#t_pstruct">Packed Structure Type</a></li>
73 <li><a href="#t_vector">Vector Type</a></li>
74 <li><a href="#t_opaque">Opaque Type</a></li>
77 <li><a href="#t_uprefs">Type Up-references</a></li>
80 <li><a href="#constants">Constants</a>
82 <li><a href="#simpleconstants">Simple Constants</a></li>
83 <li><a href="#complexconstants">Complex Constants</a></li>
84 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
85 <li><a href="#undefvalues">Undefined Values</a></li>
86 <li><a href="#blockaddress">Addresses of Basic Blocks</a></li>
87 <li><a href="#constantexprs">Constant Expressions</a></li>
88 <li><a href="#metadata">Embedded Metadata</a></li>
91 <li><a href="#othervalues">Other Values</a>
93 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
96 <li><a href="#intrinsic_globals">Intrinsic Global Variables</a>
98 <li><a href="#intg_used">The '<tt>llvm.used</tt>' Global Variable</a></li>
99 <li><a href="#intg_compiler_used">The '<tt>llvm.compiler.used</tt>'
100 Global Variable</a></li>
101 <li><a href="#intg_global_ctors">The '<tt>llvm.global_ctors</tt>'
102 Global Variable</a></li>
103 <li><a href="#intg_global_dtors">The '<tt>llvm.global_dtors</tt>'
104 Global Variable</a></li>
107 <li><a href="#instref">Instruction Reference</a>
109 <li><a href="#terminators">Terminator Instructions</a>
111 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
112 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
113 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
114 <li><a href="#i_indirectbr">'<tt>indirectbr</tt>' Instruction</a></li>
115 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
116 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
117 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
120 <li><a href="#binaryops">Binary Operations</a>
122 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
123 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
124 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
125 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
126 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
127 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
128 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
129 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
130 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
131 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
132 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
133 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
136 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
138 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
139 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
140 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
141 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
142 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
143 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
146 <li><a href="#vectorops">Vector Operations</a>
148 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
149 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
150 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
153 <li><a href="#aggregateops">Aggregate Operations</a>
155 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
156 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
159 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
161 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
162 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
163 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
164 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
167 <li><a href="#convertops">Conversion Operations</a>
169 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
170 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
171 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
172 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
173 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
174 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
175 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
176 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
177 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
178 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
179 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
180 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
183 <li><a href="#otherops">Other Operations</a>
185 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
186 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
187 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
188 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
189 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
190 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
195 <li><a href="#intrinsics">Intrinsic Functions</a>
197 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
199 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
200 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
201 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
204 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
206 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
207 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
208 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
211 <li><a href="#int_codegen">Code Generator Intrinsics</a>
213 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
214 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
215 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
216 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
217 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
218 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
219 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
222 <li><a href="#int_libc">Standard C Library Intrinsics</a>
224 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
225 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
226 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
227 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
228 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
229 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
230 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
231 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
234 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
236 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
237 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
238 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
239 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
242 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
244 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
245 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
246 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
247 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
248 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
249 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
252 <li><a href="#int_debugger">Debugger intrinsics</a></li>
253 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
254 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
256 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
259 <li><a href="#int_atomics">Atomic intrinsics</a>
261 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
262 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
263 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
264 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
265 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
266 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
267 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
268 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
269 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
270 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
271 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
272 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
273 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
276 <li><a href="#int_memorymarkers">Memory Use Markers</a>
278 <li><a href="#int_lifetime_start"><tt>llvm.lifetime.start</tt></a></li>
279 <li><a href="#int_lifetime_end"><tt>llvm.lifetime.end</tt></a></li>
280 <li><a href="#int_invariant_start"><tt>llvm.invariant.start</tt></a></li>
281 <li><a href="#int_invariant_end"><tt>llvm.invariant.end</tt></a></li>
284 <li><a href="#int_general">General intrinsics</a>
286 <li><a href="#int_var_annotation">
287 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
288 <li><a href="#int_annotation">
289 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
290 <li><a href="#int_trap">
291 '<tt>llvm.trap</tt>' Intrinsic</a></li>
292 <li><a href="#int_stackprotector">
293 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
294 <li><a href="#int_objectsize">
295 '<tt>llvm.objectsize</tt>' Intrinsic</a></li>
302 <div class="doc_author">
303 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
304 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
307 <!-- *********************************************************************** -->
308 <div class="doc_section"> <a name="abstract">Abstract </a></div>
309 <!-- *********************************************************************** -->
311 <div class="doc_text">
313 <p>This document is a reference manual for the LLVM assembly language. LLVM is
314 a Static Single Assignment (SSA) based representation that provides type
315 safety, low-level operations, flexibility, and the capability of representing
316 'all' high-level languages cleanly. It is the common code representation
317 used throughout all phases of the LLVM compilation strategy.</p>
321 <!-- *********************************************************************** -->
322 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
323 <!-- *********************************************************************** -->
325 <div class="doc_text">
327 <p>The LLVM code representation is designed to be used in three different forms:
328 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
329 for fast loading by a Just-In-Time compiler), and as a human readable
330 assembly language representation. This allows LLVM to provide a powerful
331 intermediate representation for efficient compiler transformations and
332 analysis, while providing a natural means to debug and visualize the
333 transformations. The three different forms of LLVM are all equivalent. This
334 document describes the human readable representation and notation.</p>
336 <p>The LLVM representation aims to be light-weight and low-level while being
337 expressive, typed, and extensible at the same time. It aims to be a
338 "universal IR" of sorts, by being at a low enough level that high-level ideas
339 may be cleanly mapped to it (similar to how microprocessors are "universal
340 IR's", allowing many source languages to be mapped to them). By providing
341 type information, LLVM can be used as the target of optimizations: for
342 example, through pointer analysis, it can be proven that a C automatic
343 variable is never accessed outside of the current function, allowing it to
344 be promoted to a simple SSA value instead of a memory location.</p>
348 <!-- _______________________________________________________________________ -->
349 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
351 <div class="doc_text">
353 <p>It is important to note that this document describes 'well formed' LLVM
354 assembly language. There is a difference between what the parser accepts and
355 what is considered 'well formed'. For example, the following instruction is
356 syntactically okay, but not well formed:</p>
358 <div class="doc_code">
360 %x = <a href="#i_add">add</a> i32 1, %x
364 <p>because the definition of <tt>%x</tt> does not dominate all of its uses. The
365 LLVM infrastructure provides a verification pass that may be used to verify
366 that an LLVM module is well formed. This pass is automatically run by the
367 parser after parsing input assembly and by the optimizer before it outputs
368 bitcode. The violations pointed out by the verifier pass indicate bugs in
369 transformation passes or input to the parser.</p>
373 <!-- Describe the typesetting conventions here. -->
375 <!-- *********************************************************************** -->
376 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
377 <!-- *********************************************************************** -->
379 <div class="doc_text">
381 <p>LLVM identifiers come in two basic types: global and local. Global
382 identifiers (functions, global variables) begin with the <tt>'@'</tt>
383 character. Local identifiers (register names, types) begin with
384 the <tt>'%'</tt> character. Additionally, there are three different formats
385 for identifiers, for different purposes:</p>
388 <li>Named values are represented as a string of characters with their prefix.
389 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
390 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
391 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
392 other characters in their names can be surrounded with quotes. Special
393 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
394 ASCII code for the character in hexadecimal. In this way, any character
395 can be used in a name value, even quotes themselves.</li>
397 <li>Unnamed values are represented as an unsigned numeric value with their
398 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
400 <li>Constants, which are described in a <a href="#constants">section about
401 constants</a>, below.</li>
404 <p>LLVM requires that values start with a prefix for two reasons: Compilers
405 don't need to worry about name clashes with reserved words, and the set of
406 reserved words may be expanded in the future without penalty. Additionally,
407 unnamed identifiers allow a compiler to quickly come up with a temporary
408 variable without having to avoid symbol table conflicts.</p>
410 <p>Reserved words in LLVM are very similar to reserved words in other
411 languages. There are keywords for different opcodes
412 ('<tt><a href="#i_add">add</a></tt>',
413 '<tt><a href="#i_bitcast">bitcast</a></tt>',
414 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
415 ('<tt><a href="#t_void">void</a></tt>',
416 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
417 reserved words cannot conflict with variable names, because none of them
418 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
420 <p>Here is an example of LLVM code to multiply the integer variable
421 '<tt>%X</tt>' by 8:</p>
425 <div class="doc_code">
427 %result = <a href="#i_mul">mul</a> i32 %X, 8
431 <p>After strength reduction:</p>
433 <div class="doc_code">
435 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
439 <p>And the hard way:</p>
441 <div class="doc_code">
443 %0 = <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
444 %1 = <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
445 %result = <a href="#i_add">add</a> i32 %1, %1
449 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
450 lexical features of LLVM:</p>
453 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
456 <li>Unnamed temporaries are created when the result of a computation is not
457 assigned to a named value.</li>
459 <li>Unnamed temporaries are numbered sequentially</li>
462 <p>It also shows a convention that we follow in this document. When
463 demonstrating instructions, we will follow an instruction with a comment that
464 defines the type and name of value produced. Comments are shown in italic
469 <!-- *********************************************************************** -->
470 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
471 <!-- *********************************************************************** -->
473 <!-- ======================================================================= -->
474 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
477 <div class="doc_text">
479 <p>LLVM programs are composed of "Module"s, each of which is a translation unit
480 of the input programs. Each module consists of functions, global variables,
481 and symbol table entries. Modules may be combined together with the LLVM
482 linker, which merges function (and global variable) definitions, resolves
483 forward declarations, and merges symbol table entries. Here is an example of
484 the "hello world" module:</p>
486 <div class="doc_code">
488 <i>; Declare the string constant as a global constant.</i>
489 <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>
491 <i>; External declaration of the puts function</i>
492 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
494 <i>; Definition of main function</i>
495 define i32 @main() { <i>; i32()* </i>
496 <i>; Convert [13 x i8]* to i8 *...</i>
497 %cast210 = <a href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
499 <i>; Call puts function to write out the string to stdout.</i>
500 <a href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
501 <a href="#i_ret">ret</a> i32 0<br>}<br>
505 <p>This example is made up of a <a href="#globalvars">global variable</a> named
506 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function, and
507 a <a href="#functionstructure">function definition</a> for
510 <p>In general, a module is made up of a list of global values, where both
511 functions and global variables are global values. Global values are
512 represented by a pointer to a memory location (in this case, a pointer to an
513 array of char, and a pointer to a function), and have one of the
514 following <a href="#linkage">linkage types</a>.</p>
518 <!-- ======================================================================= -->
519 <div class="doc_subsection">
520 <a name="linkage">Linkage Types</a>
523 <div class="doc_text">
525 <p>All Global Variables and Functions have one of the following types of
529 <dt><tt><b><a name="linkage_private">private</a></b></tt></dt>
530 <dd>Global values with private linkage are only directly accessible by objects
531 in the current module. In particular, linking code into a module with an
532 private global value may cause the private to be renamed as necessary to
533 avoid collisions. Because the symbol is private to the module, all
534 references can be updated. This doesn't show up in any symbol table in the
537 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt></dt>
538 <dd>Similar to private, but the symbol is passed through the assembler and
539 removed by the linker after evaluation. Note that (unlike private
540 symbols) linker_private symbols are subject to coalescing by the linker:
541 weak symbols get merged and redefinitions are rejected. However, unlike
542 normal strong symbols, they are removed by the linker from the final
543 linked image (executable or dynamic library).</dd>
545 <dt><tt><b><a name="linkage_internal">internal</a></b></tt></dt>
546 <dd>Similar to private, but the value shows as a local symbol
547 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
548 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
550 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt></dt>
551 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
552 into the object file corresponding to the LLVM module. They exist to
553 allow inlining and other optimizations to take place given knowledge of
554 the definition of the global, which is known to be somewhere outside the
555 module. Globals with <tt>available_externally</tt> linkage are allowed to
556 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
557 This linkage type is only allowed on definitions, not declarations.</dd>
559 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt></dt>
560 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
561 the same name when linkage occurs. This can be used to implement
562 some forms of inline functions, templates, or other code which must be
563 generated in each translation unit that uses it, but where the body may
564 be overridden with a more definitive definition later. Unreferenced
565 <tt>linkonce</tt> globals are allowed to be discarded. Note that
566 <tt>linkonce</tt> linkage does not actually allow the optimizer to
567 inline the body of this function into callers because it doesn't know if
568 this definition of the function is the definitive definition within the
569 program or whether it will be overridden by a stronger definition.
570 To enable inlining and other optimizations, use "<tt>linkonce_odr</tt>"
573 <dt><tt><b><a name="linkage_weak">weak</a></b></tt></dt>
574 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
575 <tt>linkonce</tt> linkage, except that unreferenced globals with
576 <tt>weak</tt> linkage may not be discarded. This is used for globals that
577 are declared "weak" in C source code.</dd>
579 <dt><tt><b><a name="linkage_common">common</a></b></tt></dt>
580 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
581 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
583 Symbols with "<tt>common</tt>" linkage are merged in the same way as
584 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
585 <tt>common</tt> symbols may not have an explicit section,
586 must have a zero initializer, and may not be marked '<a
587 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
588 have common linkage.</dd>
591 <dt><tt><b><a name="linkage_appending">appending</a></b></tt></dt>
592 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
593 pointer to array type. When two global variables with appending linkage
594 are linked together, the two global arrays are appended together. This is
595 the LLVM, typesafe, equivalent of having the system linker append together
596 "sections" with identical names when .o files are linked.</dd>
598 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt></dt>
599 <dd>The semantics of this linkage follow the ELF object file model: the symbol
600 is weak until linked, if not linked, the symbol becomes null instead of
601 being an undefined reference.</dd>
603 <dt><tt><b><a name="linkage_linkonce_odr">linkonce_odr</a></b></tt></dt>
604 <dt><tt><b><a name="linkage_weak_odr">weak_odr</a></b></tt></dt>
605 <dd>Some languages allow differing globals to be merged, such as two functions
606 with different semantics. Other languages, such as <tt>C++</tt>, ensure
607 that only equivalent globals are ever merged (the "one definition rule" -
608 "ODR"). Such languages can use the <tt>linkonce_odr</tt>
609 and <tt>weak_odr</tt> linkage types to indicate that the global will only
610 be merged with equivalent globals. These linkage types are otherwise the
611 same as their non-<tt>odr</tt> versions.</dd>
613 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
614 <dd>If none of the above identifiers are used, the global is externally
615 visible, meaning that it participates in linkage and can be used to
616 resolve external symbol references.</dd>
619 <p>The next two types of linkage are targeted for Microsoft Windows platform
620 only. They are designed to support importing (exporting) symbols from (to)
621 DLLs (Dynamic Link Libraries).</p>
624 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt></dt>
625 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
626 or variable via a global pointer to a pointer that is set up by the DLL
627 exporting the symbol. On Microsoft Windows targets, the pointer name is
628 formed by combining <code>__imp_</code> and the function or variable
631 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt></dt>
632 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
633 pointer to a pointer in a DLL, so that it can be referenced with the
634 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
635 name is formed by combining <code>__imp_</code> and the function or
639 <p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
640 another module defined a "<tt>.LC0</tt>" variable and was linked with this
641 one, one of the two would be renamed, preventing a collision. Since
642 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
643 declarations), they are accessible outside of the current module.</p>
645 <p>It is illegal for a function <i>declaration</i> to have any linkage type
646 other than "externally visible", <tt>dllimport</tt>
647 or <tt>extern_weak</tt>.</p>
649 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
650 or <tt>weak_odr</tt> linkages.</p>
654 <!-- ======================================================================= -->
655 <div class="doc_subsection">
656 <a name="callingconv">Calling Conventions</a>
659 <div class="doc_text">
661 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
662 and <a href="#i_invoke">invokes</a> can all have an optional calling
663 convention specified for the call. The calling convention of any pair of
664 dynamic caller/callee must match, or the behavior of the program is
665 undefined. The following calling conventions are supported by LLVM, and more
666 may be added in the future:</p>
669 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
670 <dd>This calling convention (the default if no other calling convention is
671 specified) matches the target C calling conventions. This calling
672 convention supports varargs function calls and tolerates some mismatch in
673 the declared prototype and implemented declaration of the function (as
676 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
677 <dd>This calling convention attempts to make calls as fast as possible
678 (e.g. by passing things in registers). This calling convention allows the
679 target to use whatever tricks it wants to produce fast code for the
680 target, without having to conform to an externally specified ABI
681 (Application Binary Interface). Implementations of this convention should
682 allow arbitrary <a href="CodeGenerator.html#tailcallopt">tail call
683 optimization</a> to be supported. This calling convention does not
684 support varargs and requires the prototype of all callees to exactly match
685 the prototype of the function definition.</dd>
687 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
688 <dd>This calling convention attempts to make code in the caller as efficient
689 as possible under the assumption that the call is not commonly executed.
690 As such, these calls often preserve all registers so that the call does
691 not break any live ranges in the caller side. This calling convention
692 does not support varargs and requires the prototype of all callees to
693 exactly match the prototype of the function definition.</dd>
695 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
696 <dd>Any calling convention may be specified by number, allowing
697 target-specific calling conventions to be used. Target specific calling
698 conventions start at 64.</dd>
701 <p>More calling conventions can be added/defined on an as-needed basis, to
702 support Pascal conventions or any other well-known target-independent
707 <!-- ======================================================================= -->
708 <div class="doc_subsection">
709 <a name="visibility">Visibility Styles</a>
712 <div class="doc_text">
714 <p>All Global Variables and Functions have one of the following visibility
718 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
719 <dd>On targets that use the ELF object file format, default visibility means
720 that the declaration is visible to other modules and, in shared libraries,
721 means that the declared entity may be overridden. On Darwin, default
722 visibility means that the declaration is visible to other modules. Default
723 visibility corresponds to "external linkage" in the language.</dd>
725 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
726 <dd>Two declarations of an object with hidden visibility refer to the same
727 object if they are in the same shared object. Usually, hidden visibility
728 indicates that the symbol will not be placed into the dynamic symbol
729 table, so no other module (executable or shared library) can reference it
732 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
733 <dd>On ELF, protected visibility indicates that the symbol will be placed in
734 the dynamic symbol table, but that references within the defining module
735 will bind to the local symbol. That is, the symbol cannot be overridden by
741 <!-- ======================================================================= -->
742 <div class="doc_subsection">
743 <a name="namedtypes">Named Types</a>
746 <div class="doc_text">
748 <p>LLVM IR allows you to specify name aliases for certain types. This can make
749 it easier to read the IR and make the IR more condensed (particularly when
750 recursive types are involved). An example of a name specification is:</p>
752 <div class="doc_code">
754 %mytype = type { %mytype*, i32 }
758 <p>You may give a name to any <a href="#typesystem">type</a> except
759 "<a href="t_void">void</a>". Type name aliases may be used anywhere a type
760 is expected with the syntax "%mytype".</p>
762 <p>Note that type names are aliases for the structural type that they indicate,
763 and that you can therefore specify multiple names for the same type. This
764 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
765 uses structural typing, the name is not part of the type. When printing out
766 LLVM IR, the printer will pick <em>one name</em> to render all types of a
767 particular shape. This means that if you have code where two different
768 source types end up having the same LLVM type, that the dumper will sometimes
769 print the "wrong" or unexpected type. This is an important design point and
770 isn't going to change.</p>
774 <!-- ======================================================================= -->
775 <div class="doc_subsection">
776 <a name="globalvars">Global Variables</a>
779 <div class="doc_text">
781 <p>Global variables define regions of memory allocated at compilation time
782 instead of run-time. Global variables may optionally be initialized, may
783 have an explicit section to be placed in, and may have an optional explicit
784 alignment specified. A variable may be defined as "thread_local", which
785 means that it will not be shared by threads (each thread will have a
786 separated copy of the variable). A variable may be defined as a global
787 "constant," which indicates that the contents of the variable
788 will <b>never</b> be modified (enabling better optimization, allowing the
789 global data to be placed in the read-only section of an executable, etc).
790 Note that variables that need runtime initialization cannot be marked
791 "constant" as there is a store to the variable.</p>
793 <p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
794 constant, even if the final definition of the global is not. This capability
795 can be used to enable slightly better optimization of the program, but
796 requires the language definition to guarantee that optimizations based on the
797 'constantness' are valid for the translation units that do not include the
800 <p>As SSA values, global variables define pointer values that are in scope
801 (i.e. they dominate) all basic blocks in the program. Global variables
802 always define a pointer to their "content" type because they describe a
803 region of memory, and all memory objects in LLVM are accessed through
806 <p>A global variable may be declared to reside in a target-specific numbered
807 address space. For targets that support them, address spaces may affect how
808 optimizations are performed and/or what target instructions are used to
809 access the variable. The default address space is zero. The address space
810 qualifier must precede any other attributes.</p>
812 <p>LLVM allows an explicit section to be specified for globals. If the target
813 supports it, it will emit globals to the section specified.</p>
815 <p>An explicit alignment may be specified for a global. If not present, or if
816 the alignment is set to zero, the alignment of the global is set by the
817 target to whatever it feels convenient. If an explicit alignment is
818 specified, the global is forced to have at least that much alignment. All
819 alignments must be a power of 2.</p>
821 <p>For example, the following defines a global in a numbered address space with
822 an initializer, section, and alignment:</p>
824 <div class="doc_code">
826 @G = addrspace(5) constant float 1.0, section "foo", align 4
833 <!-- ======================================================================= -->
834 <div class="doc_subsection">
835 <a name="functionstructure">Functions</a>
838 <div class="doc_text">
840 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord, an
841 optional <a href="#linkage">linkage type</a>, an optional
842 <a href="#visibility">visibility style</a>, an optional
843 <a href="#callingconv">calling convention</a>, a return type, an optional
844 <a href="#paramattrs">parameter attribute</a> for the return type, a function
845 name, a (possibly empty) argument list (each with optional
846 <a href="#paramattrs">parameter attributes</a>), optional
847 <a href="#fnattrs">function attributes</a>, an optional section, an optional
848 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
849 curly brace, a list of basic blocks, and a closing curly brace.</p>
851 <p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
852 optional <a href="#linkage">linkage type</a>, an optional
853 <a href="#visibility">visibility style</a>, an optional
854 <a href="#callingconv">calling convention</a>, a return type, an optional
855 <a href="#paramattrs">parameter attribute</a> for the return type, a function
856 name, a possibly empty list of arguments, an optional alignment, and an
857 optional <a href="#gc">garbage collector name</a>.</p>
859 <p>A function definition contains a list of basic blocks, forming the CFG
860 (Control Flow Graph) for the function. Each basic block may optionally start
861 with a label (giving the basic block a symbol table entry), contains a list
862 of instructions, and ends with a <a href="#terminators">terminator</a>
863 instruction (such as a branch or function return).</p>
865 <p>The first basic block in a function is special in two ways: it is immediately
866 executed on entrance to the function, and it is not allowed to have
867 predecessor basic blocks (i.e. there can not be any branches to the entry
868 block of a function). Because the block can have no predecessors, it also
869 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
871 <p>LLVM allows an explicit section to be specified for functions. If the target
872 supports it, it will emit functions to the section specified.</p>
874 <p>An explicit alignment may be specified for a function. If not present, or if
875 the alignment is set to zero, the alignment of the function is set by the
876 target to whatever it feels convenient. If an explicit alignment is
877 specified, the function is forced to have at least that much alignment. All
878 alignments must be a power of 2.</p>
881 <div class="doc_code">
883 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
884 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
885 <ResultType> @<FunctionName> ([argument list])
886 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
887 [<a href="#gc">gc</a>] { ... }
893 <!-- ======================================================================= -->
894 <div class="doc_subsection">
895 <a name="aliasstructure">Aliases</a>
898 <div class="doc_text">
900 <p>Aliases act as "second name" for the aliasee value (which can be either
901 function, global variable, another alias or bitcast of global value). Aliases
902 may have an optional <a href="#linkage">linkage type</a>, and an
903 optional <a href="#visibility">visibility style</a>.</p>
906 <div class="doc_code">
908 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
914 <!-- ======================================================================= -->
915 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
917 <div class="doc_text">
919 <p>The return type and each parameter of a function type may have a set of
920 <i>parameter attributes</i> associated with them. Parameter attributes are
921 used to communicate additional information about the result or parameters of
922 a function. Parameter attributes are considered to be part of the function,
923 not of the function type, so functions with different parameter attributes
924 can have the same function type.</p>
926 <p>Parameter attributes are simple keywords that follow the type specified. If
927 multiple parameter attributes are needed, they are space separated. For
930 <div class="doc_code">
932 declare i32 @printf(i8* noalias nocapture, ...)
933 declare i32 @atoi(i8 zeroext)
934 declare signext i8 @returns_signed_char()
938 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
939 <tt>readonly</tt>) come immediately after the argument list.</p>
941 <p>Currently, only the following parameter attributes are defined:</p>
944 <dt><tt><b>zeroext</b></tt></dt>
945 <dd>This indicates to the code generator that the parameter or return value
946 should be zero-extended to a 32-bit value by the caller (for a parameter)
947 or the callee (for a return value).</dd>
949 <dt><tt><b>signext</b></tt></dt>
950 <dd>This indicates to the code generator that the parameter or return value
951 should be sign-extended to a 32-bit value by the caller (for a parameter)
952 or the callee (for a return value).</dd>
954 <dt><tt><b>inreg</b></tt></dt>
955 <dd>This indicates that this parameter or return value should be treated in a
956 special target-dependent fashion during while emitting code for a function
957 call or return (usually, by putting it in a register as opposed to memory,
958 though some targets use it to distinguish between two different kinds of
959 registers). Use of this attribute is target-specific.</dd>
961 <dt><tt><b><a name="byval">byval</a></b></tt></dt>
962 <dd>This indicates that the pointer parameter should really be passed by value
963 to the function. The attribute implies that a hidden copy of the pointee
964 is made between the caller and the callee, so the callee is unable to
965 modify the value in the callee. This attribute is only valid on LLVM
966 pointer arguments. It is generally used to pass structs and arrays by
967 value, but is also valid on pointers to scalars. The copy is considered
968 to belong to the caller not the callee (for example,
969 <tt><a href="#readonly">readonly</a></tt> functions should not write to
970 <tt>byval</tt> parameters). This is not a valid attribute for return
971 values. The byval attribute also supports specifying an alignment with
972 the align attribute. This has a target-specific effect on the code
973 generator that usually indicates a desired alignment for the synthesized
976 <dt><tt><b>sret</b></tt></dt>
977 <dd>This indicates that the pointer parameter specifies the address of a
978 structure that is the return value of the function in the source program.
979 This pointer must be guaranteed by the caller to be valid: loads and
980 stores to the structure may be assumed by the callee to not to trap. This
981 may only be applied to the first parameter. This is not a valid attribute
982 for return values. </dd>
984 <dt><tt><b>noalias</b></tt></dt>
985 <dd>This indicates that the pointer does not alias any global or any other
986 parameter. The caller is responsible for ensuring that this is the
987 case. On a function return value, <tt>noalias</tt> additionally indicates
988 that the pointer does not alias any other pointers visible to the
989 caller. For further details, please see the discussion of the NoAlias
991 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
994 <dt><tt><b>nocapture</b></tt></dt>
995 <dd>This indicates that the callee does not make any copies of the pointer
996 that outlive the callee itself. This is not a valid attribute for return
999 <dt><tt><b>nest</b></tt></dt>
1000 <dd>This indicates that the pointer parameter can be excised using the
1001 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
1002 attribute for return values.</dd>
1007 <!-- ======================================================================= -->
1008 <div class="doc_subsection">
1009 <a name="gc">Garbage Collector Names</a>
1012 <div class="doc_text">
1014 <p>Each function may specify a garbage collector name, which is simply a
1017 <div class="doc_code">
1019 define void @f() gc "name" { ... }
1023 <p>The compiler declares the supported values of <i>name</i>. Specifying a
1024 collector which will cause the compiler to alter its output in order to
1025 support the named garbage collection algorithm.</p>
1029 <!-- ======================================================================= -->
1030 <div class="doc_subsection">
1031 <a name="fnattrs">Function Attributes</a>
1034 <div class="doc_text">
1036 <p>Function attributes are set to communicate additional information about a
1037 function. Function attributes are considered to be part of the function, not
1038 of the function type, so functions with different parameter attributes can
1039 have the same function type.</p>
1041 <p>Function attributes are simple keywords that follow the type specified. If
1042 multiple attributes are needed, they are space separated. For example:</p>
1044 <div class="doc_code">
1046 define void @f() noinline { ... }
1047 define void @f() alwaysinline { ... }
1048 define void @f() alwaysinline optsize { ... }
1049 define void @f() optsize { ... }
1054 <dt><tt><b>alwaysinline</b></tt></dt>
1055 <dd>This attribute indicates that the inliner should attempt to inline this
1056 function into callers whenever possible, ignoring any active inlining size
1057 threshold for this caller.</dd>
1059 <dt><tt><b>inlinehint</b></tt></dt>
1060 <dd>This attribute indicates that the source code contained a hint that inlining
1061 this function is desirable (such as the "inline" keyword in C/C++). It
1062 is just a hint; it imposes no requirements on the inliner.</dd>
1064 <dt><tt><b>noinline</b></tt></dt>
1065 <dd>This attribute indicates that the inliner should never inline this
1066 function in any situation. This attribute may not be used together with
1067 the <tt>alwaysinline</tt> attribute.</dd>
1069 <dt><tt><b>optsize</b></tt></dt>
1070 <dd>This attribute suggests that optimization passes and code generator passes
1071 make choices that keep the code size of this function low, and otherwise
1072 do optimizations specifically to reduce code size.</dd>
1074 <dt><tt><b>noreturn</b></tt></dt>
1075 <dd>This function attribute indicates that the function never returns
1076 normally. This produces undefined behavior at runtime if the function
1077 ever does dynamically return.</dd>
1079 <dt><tt><b>nounwind</b></tt></dt>
1080 <dd>This function attribute indicates that the function never returns with an
1081 unwind or exceptional control flow. If the function does unwind, its
1082 runtime behavior is undefined.</dd>
1084 <dt><tt><b>readnone</b></tt></dt>
1085 <dd>This attribute indicates that the function computes its result (or decides
1086 to unwind an exception) based strictly on its arguments, without
1087 dereferencing any pointer arguments or otherwise accessing any mutable
1088 state (e.g. memory, control registers, etc) visible to caller functions.
1089 It does not write through any pointer arguments
1090 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1091 changes any state visible to callers. This means that it cannot unwind
1092 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1093 could use the <tt>unwind</tt> instruction.</dd>
1095 <dt><tt><b><a name="readonly">readonly</a></b></tt></dt>
1096 <dd>This attribute indicates that the function does not write through any
1097 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1098 arguments) or otherwise modify any state (e.g. memory, control registers,
1099 etc) visible to caller functions. It may dereference pointer arguments
1100 and read state that may be set in the caller. A readonly function always
1101 returns the same value (or unwinds an exception identically) when called
1102 with the same set of arguments and global state. It cannot unwind an
1103 exception by calling the <tt>C++</tt> exception throwing methods, but may
1104 use the <tt>unwind</tt> instruction.</dd>
1106 <dt><tt><b><a name="ssp">ssp</a></b></tt></dt>
1107 <dd>This attribute indicates that the function should emit a stack smashing
1108 protector. It is in the form of a "canary"—a random value placed on
1109 the stack before the local variables that's checked upon return from the
1110 function to see if it has been overwritten. A heuristic is used to
1111 determine if a function needs stack protectors or not.<br>
1113 If a function that has an <tt>ssp</tt> attribute is inlined into a
1114 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1115 function will have an <tt>ssp</tt> attribute.</dd>
1117 <dt><tt><b>sspreq</b></tt></dt>
1118 <dd>This attribute indicates that the function should <em>always</em> emit a
1119 stack smashing protector. This overrides
1120 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1122 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1123 function that doesn't have an <tt>sspreq</tt> attribute or which has
1124 an <tt>ssp</tt> attribute, then the resulting function will have
1125 an <tt>sspreq</tt> attribute.</dd>
1127 <dt><tt><b>noredzone</b></tt></dt>
1128 <dd>This attribute indicates that the code generator should not use a red
1129 zone, even if the target-specific ABI normally permits it.</dd>
1131 <dt><tt><b>noimplicitfloat</b></tt></dt>
1132 <dd>This attributes disables implicit floating point instructions.</dd>
1134 <dt><tt><b>naked</b></tt></dt>
1135 <dd>This attribute disables prologue / epilogue emission for the function.
1136 This can have very system-specific consequences.</dd>
1141 <!-- ======================================================================= -->
1142 <div class="doc_subsection">
1143 <a name="moduleasm">Module-Level Inline Assembly</a>
1146 <div class="doc_text">
1148 <p>Modules may contain "module-level inline asm" blocks, which corresponds to
1149 the GCC "file scope inline asm" blocks. These blocks are internally
1150 concatenated by LLVM and treated as a single unit, but may be separated in
1151 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
1153 <div class="doc_code">
1155 module asm "inline asm code goes here"
1156 module asm "more can go here"
1160 <p>The strings can contain any character by escaping non-printable characters.
1161 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1164 <p>The inline asm code is simply printed to the machine code .s file when
1165 assembly code is generated.</p>
1169 <!-- ======================================================================= -->
1170 <div class="doc_subsection">
1171 <a name="datalayout">Data Layout</a>
1174 <div class="doc_text">
1176 <p>A module may specify a target specific data layout string that specifies how
1177 data is to be laid out in memory. The syntax for the data layout is
1180 <div class="doc_code">
1182 target datalayout = "<i>layout specification</i>"
1186 <p>The <i>layout specification</i> consists of a list of specifications
1187 separated by the minus sign character ('-'). Each specification starts with
1188 a letter and may include other information after the letter to define some
1189 aspect of the data layout. The specifications accepted are as follows:</p>
1193 <dd>Specifies that the target lays out data in big-endian form. That is, the
1194 bits with the most significance have the lowest address location.</dd>
1197 <dd>Specifies that the target lays out data in little-endian form. That is,
1198 the bits with the least significance have the lowest address
1201 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1202 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1203 <i>preferred</i> alignments. All sizes are in bits. Specifying
1204 the <i>pref</i> alignment is optional. If omitted, the
1205 preceding <tt>:</tt> should be omitted too.</dd>
1207 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1208 <dd>This specifies the alignment for an integer type of a given bit
1209 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1211 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1212 <dd>This specifies the alignment for a vector type of a given bit
1215 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1216 <dd>This specifies the alignment for a floating point type of a given bit
1217 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1220 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1221 <dd>This specifies the alignment for an aggregate type of a given bit
1224 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1225 <dd>This specifies the alignment for a stack object of a given bit
1228 <dt><tt>n<i>size1</i>:<i>size2</i>:<i>size3</i>...</tt></dt>
1229 <dd>This specifies a set of native integer widths for the target CPU
1230 in bits. For example, it might contain "n32" for 32-bit PowerPC,
1231 "n32:64" for PowerPC 64, or "n8:16:32:64" for X86-64. Elements of
1232 this set are considered to support most general arithmetic
1233 operations efficiently.</dd>
1236 <p>When constructing the data layout for a given target, LLVM starts with a
1237 default set of specifications which are then (possibly) overriden by the
1238 specifications in the <tt>datalayout</tt> keyword. The default specifications
1239 are given in this list:</p>
1242 <li><tt>E</tt> - big endian</li>
1243 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1244 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1245 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1246 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1247 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1248 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1249 alignment of 64-bits</li>
1250 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1251 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1252 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1253 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1254 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1255 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
1258 <p>When LLVM is determining the alignment for a given type, it uses the
1259 following rules:</p>
1262 <li>If the type sought is an exact match for one of the specifications, that
1263 specification is used.</li>
1265 <li>If no match is found, and the type sought is an integer type, then the
1266 smallest integer type that is larger than the bitwidth of the sought type
1267 is used. If none of the specifications are larger than the bitwidth then
1268 the the largest integer type is used. For example, given the default
1269 specifications above, the i7 type will use the alignment of i8 (next
1270 largest) while both i65 and i256 will use the alignment of i64 (largest
1273 <li>If no match is found, and the type sought is a vector type, then the
1274 largest vector type that is smaller than the sought vector type will be
1275 used as a fall back. This happens because <128 x double> can be
1276 implemented in terms of 64 <2 x double>, for example.</li>
1281 <!-- ======================================================================= -->
1282 <div class="doc_subsection">
1283 <a name="pointeraliasing">Pointer Aliasing Rules</a>
1286 <div class="doc_text">
1288 <p>Any memory access must be done through a pointer value associated
1289 with an address range of the memory access, otherwise the behavior
1290 is undefined. Pointer values are associated with address ranges
1291 according to the following rules:</p>
1294 <li>A pointer value formed from a
1295 <tt><a href="#i_getelementptr">getelementptr</a></tt> instruction
1296 is associated with the addresses associated with the first operand
1297 of the <tt>getelementptr</tt>.</li>
1298 <li>An address of a global variable is associated with the address
1299 range of the variable's storage.</li>
1300 <li>The result value of an allocation instruction is associated with
1301 the address range of the allocated storage.</li>
1302 <li>A null pointer in the default address-space is associated with
1304 <li>A pointer value formed by an
1305 <tt><a href="#i_inttoptr">inttoptr</a></tt> is associated with all
1306 address ranges of all pointer values that contribute (directly or
1307 indirectly) to the computation of the pointer's value.</li>
1308 <li>The result value of a
1309 <tt><a href="#i_bitcast">bitcast</a></tt> is associated with all
1310 addresses associated with the operand of the <tt>bitcast</tt>.</li>
1311 <li>An integer constant other than zero or a pointer value returned
1312 from a function not defined within LLVM may be associated with address
1313 ranges allocated through mechanisms other than those provided by
1314 LLVM. Such ranges shall not overlap with any ranges of addresses
1315 allocated by mechanisms provided by LLVM.</li>
1318 <p>LLVM IR does not associate types with memory. The result type of a
1319 <tt><a href="#i_load">load</a></tt> merely indicates the size and
1320 alignment of the memory from which to load, as well as the
1321 interpretation of the value. The first operand of a
1322 <tt><a href="#i_store">store</a></tt> similarly only indicates the size
1323 and alignment of the store.</p>
1325 <p>Consequently, type-based alias analysis, aka TBAA, aka
1326 <tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1327 LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1328 additional information which specialized optimization passes may use
1329 to implement type-based alias analysis.</p>
1333 <!-- *********************************************************************** -->
1334 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1335 <!-- *********************************************************************** -->
1337 <div class="doc_text">
1339 <p>The LLVM type system is one of the most important features of the
1340 intermediate representation. Being typed enables a number of optimizations
1341 to be performed on the intermediate representation directly, without having
1342 to do extra analyses on the side before the transformation. A strong type
1343 system makes it easier to read the generated code and enables novel analyses
1344 and transformations that are not feasible to perform on normal three address
1345 code representations.</p>
1349 <!-- ======================================================================= -->
1350 <div class="doc_subsection"> <a name="t_classifications">Type
1351 Classifications</a> </div>
1353 <div class="doc_text">
1355 <p>The types fall into a few useful classifications:</p>
1357 <table border="1" cellspacing="0" cellpadding="4">
1359 <tr><th>Classification</th><th>Types</th></tr>
1361 <td><a href="#t_integer">integer</a></td>
1362 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1365 <td><a href="#t_floating">floating point</a></td>
1366 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1369 <td><a name="t_firstclass">first class</a></td>
1370 <td><a href="#t_integer">integer</a>,
1371 <a href="#t_floating">floating point</a>,
1372 <a href="#t_pointer">pointer</a>,
1373 <a href="#t_vector">vector</a>,
1374 <a href="#t_struct">structure</a>,
1375 <a href="#t_array">array</a>,
1376 <a href="#t_label">label</a>,
1377 <a href="#t_metadata">metadata</a>.
1381 <td><a href="#t_primitive">primitive</a></td>
1382 <td><a href="#t_label">label</a>,
1383 <a href="#t_void">void</a>,
1384 <a href="#t_floating">floating point</a>,
1385 <a href="#t_metadata">metadata</a>.</td>
1388 <td><a href="#t_derived">derived</a></td>
1389 <td><a href="#t_integer">integer</a>,
1390 <a href="#t_array">array</a>,
1391 <a href="#t_function">function</a>,
1392 <a href="#t_pointer">pointer</a>,
1393 <a href="#t_struct">structure</a>,
1394 <a href="#t_pstruct">packed structure</a>,
1395 <a href="#t_vector">vector</a>,
1396 <a href="#t_opaque">opaque</a>.
1402 <p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1403 important. Values of these types are the only ones which can be produced by
1408 <!-- ======================================================================= -->
1409 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1411 <div class="doc_text">
1413 <p>The primitive types are the fundamental building blocks of the LLVM
1418 <!-- _______________________________________________________________________ -->
1419 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1421 <div class="doc_text">
1424 <p>The integer type is a very simple type that simply specifies an arbitrary
1425 bit width for the integer type desired. Any bit width from 1 bit to
1426 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
1433 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1437 <table class="layout">
1439 <td class="left"><tt>i1</tt></td>
1440 <td class="left">a single-bit integer.</td>
1443 <td class="left"><tt>i32</tt></td>
1444 <td class="left">a 32-bit integer.</td>
1447 <td class="left"><tt>i1942652</tt></td>
1448 <td class="left">a really big integer of over 1 million bits.</td>
1454 <!-- _______________________________________________________________________ -->
1455 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1457 <div class="doc_text">
1461 <tr><th>Type</th><th>Description</th></tr>
1462 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1463 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1464 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1465 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1466 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1472 <!-- _______________________________________________________________________ -->
1473 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1475 <div class="doc_text">
1478 <p>The void type does not represent any value and has no size.</p>
1487 <!-- _______________________________________________________________________ -->
1488 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1490 <div class="doc_text">
1493 <p>The label type represents code labels.</p>
1502 <!-- _______________________________________________________________________ -->
1503 <div class="doc_subsubsection"> <a name="t_metadata">Metadata Type</a> </div>
1505 <div class="doc_text">
1508 <p>The metadata type represents embedded metadata. No derived types may be
1509 created from metadata except for <a href="#t_function">function</a>
1520 <!-- ======================================================================= -->
1521 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1523 <div class="doc_text">
1525 <p>The real power in LLVM comes from the derived types in the system. This is
1526 what allows a programmer to represent arrays, functions, pointers, and other
1527 useful types. Each of these types contain one or more element types which
1528 may be a primitive type, or another derived type. For example, it is
1529 possible to have a two dimensional array, using an array as the element type
1530 of another array.</p>
1534 <!-- _______________________________________________________________________ -->
1535 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1537 <div class="doc_text">
1540 <p>The array type is a very simple derived type that arranges elements
1541 sequentially in memory. The array type requires a size (number of elements)
1542 and an underlying data type.</p>
1546 [<# elements> x <elementtype>]
1549 <p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1550 be any type with a size.</p>
1553 <table class="layout">
1555 <td class="left"><tt>[40 x i32]</tt></td>
1556 <td class="left">Array of 40 32-bit integer values.</td>
1559 <td class="left"><tt>[41 x i32]</tt></td>
1560 <td class="left">Array of 41 32-bit integer values.</td>
1563 <td class="left"><tt>[4 x i8]</tt></td>
1564 <td class="left">Array of 4 8-bit integer values.</td>
1567 <p>Here are some examples of multidimensional arrays:</p>
1568 <table class="layout">
1570 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1571 <td class="left">3x4 array of 32-bit integer values.</td>
1574 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1575 <td class="left">12x10 array of single precision floating point values.</td>
1578 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1579 <td class="left">2x3x4 array of 16-bit integer values.</td>
1583 <p>There is no restriction on indexing beyond the end of the array implied by
1584 a static type (though there are restrictions on indexing beyond the bounds
1585 of an allocated object in some cases). This means that single-dimension
1586 'variable sized array' addressing can be implemented in LLVM with a zero
1587 length array type. An implementation of 'pascal style arrays' in LLVM could
1588 use the type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
1592 <!-- _______________________________________________________________________ -->
1593 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1595 <div class="doc_text">
1598 <p>The function type can be thought of as a function signature. It consists of
1599 a return type and a list of formal parameter types. The return type of a
1600 function type is a scalar type, a void type, or a struct type. If the return
1601 type is a struct type then all struct elements must be of first class types,
1602 and the struct must have at least one element.</p>
1606 <returntype> (<parameter list>)
1609 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1610 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1611 which indicates that the function takes a variable number of arguments.
1612 Variable argument functions can access their arguments with
1613 the <a href="#int_varargs">variable argument handling intrinsic</a>
1614 functions. '<tt><returntype></tt>' is a any type except
1615 <a href="#t_label">label</a>.</p>
1618 <table class="layout">
1620 <td class="left"><tt>i32 (i32)</tt></td>
1621 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1623 </tr><tr class="layout">
1624 <td class="left"><tt>float (i16 signext, i32 *) *
1626 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1627 an <tt>i16</tt> that should be sign extended and a
1628 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1631 </tr><tr class="layout">
1632 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1633 <td class="left">A vararg function that takes at least one
1634 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1635 which returns an integer. This is the signature for <tt>printf</tt> in
1638 </tr><tr class="layout">
1639 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1640 <td class="left">A function taking an <tt>i32</tt>, returning a
1641 <a href="#t_struct">structure</a> containing two <tt>i32</tt> values
1648 <!-- _______________________________________________________________________ -->
1649 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1651 <div class="doc_text">
1654 <p>The structure type is used to represent a collection of data members together
1655 in memory. The packing of the field types is defined to match the ABI of the
1656 underlying processor. The elements of a structure may be any type that has a
1659 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1660 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1661 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1665 { <type list> }
1669 <table class="layout">
1671 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1672 <td class="left">A triple of three <tt>i32</tt> values</td>
1673 </tr><tr class="layout">
1674 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1675 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1676 second element is a <a href="#t_pointer">pointer</a> to a
1677 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1678 an <tt>i32</tt>.</td>
1684 <!-- _______________________________________________________________________ -->
1685 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1688 <div class="doc_text">
1691 <p>The packed structure type is used to represent a collection of data members
1692 together in memory. There is no padding between fields. Further, the
1693 alignment of a packed structure is 1 byte. The elements of a packed
1694 structure may be any type that has a size.</p>
1696 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1697 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1698 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1702 < { <type list> } >
1706 <table class="layout">
1708 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1709 <td class="left">A triple of three <tt>i32</tt> values</td>
1710 </tr><tr class="layout">
1712 <tt>< { float, i32 (i32)* } ></tt></td>
1713 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1714 second element is a <a href="#t_pointer">pointer</a> to a
1715 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1716 an <tt>i32</tt>.</td>
1722 <!-- _______________________________________________________________________ -->
1723 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1725 <div class="doc_text">
1728 <p>As in many languages, the pointer type represents a pointer or reference to
1729 another object, which must live in memory. Pointer types may have an optional
1730 address space attribute defining the target-specific numbered address space
1731 where the pointed-to object resides. The default address space is zero.</p>
1733 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
1734 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
1742 <table class="layout">
1744 <td class="left"><tt>[4 x i32]*</tt></td>
1745 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1746 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1749 <td class="left"><tt>i32 (i32 *) *</tt></td>
1750 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1751 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1755 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1756 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1757 that resides in address space #5.</td>
1763 <!-- _______________________________________________________________________ -->
1764 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1766 <div class="doc_text">
1769 <p>A vector type is a simple derived type that represents a vector of elements.
1770 Vector types are used when multiple primitive data are operated in parallel
1771 using a single instruction (SIMD). A vector type requires a size (number of
1772 elements) and an underlying primitive data type. Vector types are considered
1773 <a href="#t_firstclass">first class</a>.</p>
1777 < <# elements> x <elementtype> >
1780 <p>The number of elements is a constant integer value; elementtype may be any
1781 integer or floating point type.</p>
1784 <table class="layout">
1786 <td class="left"><tt><4 x i32></tt></td>
1787 <td class="left">Vector of 4 32-bit integer values.</td>
1790 <td class="left"><tt><8 x float></tt></td>
1791 <td class="left">Vector of 8 32-bit floating-point values.</td>
1794 <td class="left"><tt><2 x i64></tt></td>
1795 <td class="left">Vector of 2 64-bit integer values.</td>
1801 <!-- _______________________________________________________________________ -->
1802 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1803 <div class="doc_text">
1806 <p>Opaque types are used to represent unknown types in the system. This
1807 corresponds (for example) to the C notion of a forward declared structure
1808 type. In LLVM, opaque types can eventually be resolved to any type (not just
1809 a structure type).</p>
1817 <table class="layout">
1819 <td class="left"><tt>opaque</tt></td>
1820 <td class="left">An opaque type.</td>
1826 <!-- ======================================================================= -->
1827 <div class="doc_subsection">
1828 <a name="t_uprefs">Type Up-references</a>
1831 <div class="doc_text">
1834 <p>An "up reference" allows you to refer to a lexically enclosing type without
1835 requiring it to have a name. For instance, a structure declaration may
1836 contain a pointer to any of the types it is lexically a member of. Example
1837 of up references (with their equivalent as named type declarations)
1841 { \2 * } %x = type { %x* }
1842 { \2 }* %y = type { %y }*
1846 <p>An up reference is needed by the asmprinter for printing out cyclic types
1847 when there is no declared name for a type in the cycle. Because the
1848 asmprinter does not want to print out an infinite type string, it needs a
1849 syntax to handle recursive types that have no names (all names are optional
1857 <p>The level is the count of the lexical type that is being referred to.</p>
1860 <table class="layout">
1862 <td class="left"><tt>\1*</tt></td>
1863 <td class="left">Self-referential pointer.</td>
1866 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
1867 <td class="left">Recursive structure where the upref refers to the out-most
1874 <!-- *********************************************************************** -->
1875 <div class="doc_section"> <a name="constants">Constants</a> </div>
1876 <!-- *********************************************************************** -->
1878 <div class="doc_text">
1880 <p>LLVM has several different basic types of constants. This section describes
1881 them all and their syntax.</p>
1885 <!-- ======================================================================= -->
1886 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1888 <div class="doc_text">
1891 <dt><b>Boolean constants</b></dt>
1892 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1893 constants of the <tt><a href="#t_integer">i1</a></tt> type.</dd>
1895 <dt><b>Integer constants</b></dt>
1896 <dd>Standard integers (such as '4') are constants of
1897 the <a href="#t_integer">integer</a> type. Negative numbers may be used
1898 with integer types.</dd>
1900 <dt><b>Floating point constants</b></dt>
1901 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1902 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1903 notation (see below). The assembler requires the exact decimal value of a
1904 floating-point constant. For example, the assembler accepts 1.25 but
1905 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1906 constants must have a <a href="#t_floating">floating point</a> type. </dd>
1908 <dt><b>Null pointer constants</b></dt>
1909 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1910 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1913 <p>The one non-intuitive notation for constants is the hexadecimal form of
1914 floating point constants. For example, the form '<tt>double
1915 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
1916 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
1917 constants are required (and the only time that they are generated by the
1918 disassembler) is when a floating point constant must be emitted but it cannot
1919 be represented as a decimal floating point number in a reasonable number of
1920 digits. For example, NaN's, infinities, and other special values are
1921 represented in their IEEE hexadecimal format so that assembly and disassembly
1922 do not cause any bits to change in the constants.</p>
1924 <p>When using the hexadecimal form, constants of types float and double are
1925 represented using the 16-digit form shown above (which matches the IEEE754
1926 representation for double); float values must, however, be exactly
1927 representable as IEE754 single precision. Hexadecimal format is always used
1928 for long double, and there are three forms of long double. The 80-bit format
1929 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
1930 The 128-bit format used by PowerPC (two adjacent doubles) is represented
1931 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
1932 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
1933 currently supported target uses this format. Long doubles will only work if
1934 they match the long double format on your target. All hexadecimal formats
1935 are big-endian (sign bit at the left).</p>
1939 <!-- ======================================================================= -->
1940 <div class="doc_subsection">
1941 <a name="aggregateconstants"></a> <!-- old anchor -->
1942 <a name="complexconstants">Complex Constants</a>
1945 <div class="doc_text">
1947 <p>Complex constants are a (potentially recursive) combination of simple
1948 constants and smaller complex constants.</p>
1951 <dt><b>Structure constants</b></dt>
1952 <dd>Structure constants are represented with notation similar to structure
1953 type definitions (a comma separated list of elements, surrounded by braces
1954 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1955 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
1956 Structure constants must have <a href="#t_struct">structure type</a>, and
1957 the number and types of elements must match those specified by the
1960 <dt><b>Array constants</b></dt>
1961 <dd>Array constants are represented with notation similar to array type
1962 definitions (a comma separated list of elements, surrounded by square
1963 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
1964 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
1965 the number and types of elements must match those specified by the
1968 <dt><b>Vector constants</b></dt>
1969 <dd>Vector constants are represented with notation similar to vector type
1970 definitions (a comma separated list of elements, surrounded by
1971 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32
1972 42, i32 11, i32 74, i32 100 ></tt>". Vector constants must
1973 have <a href="#t_vector">vector type</a>, and the number and types of
1974 elements must match those specified by the type.</dd>
1976 <dt><b>Zero initialization</b></dt>
1977 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1978 value to zero of <em>any</em> type, including scalar and aggregate types.
1979 This is often used to avoid having to print large zero initializers
1980 (e.g. for large arrays) and is always exactly equivalent to using explicit
1981 zero initializers.</dd>
1983 <dt><b>Metadata node</b></dt>
1984 <dd>A metadata node is a structure-like constant with
1985 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
1986 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
1987 be interpreted as part of the instruction stream, metadata is a place to
1988 attach additional information such as debug info.</dd>
1993 <!-- ======================================================================= -->
1994 <div class="doc_subsection">
1995 <a name="globalconstants">Global Variable and Function Addresses</a>
1998 <div class="doc_text">
2000 <p>The addresses of <a href="#globalvars">global variables</a>
2001 and <a href="#functionstructure">functions</a> are always implicitly valid
2002 (link-time) constants. These constants are explicitly referenced when
2003 the <a href="#identifiers">identifier for the global</a> is used and always
2004 have <a href="#t_pointer">pointer</a> type. For example, the following is a
2005 legal LLVM file:</p>
2007 <div class="doc_code">
2011 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2017 <!-- ======================================================================= -->
2018 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
2019 <div class="doc_text">
2021 <p>The string '<tt>undef</tt>' can be used anywhere a constant is expected, and
2022 indicates that the user of the value may receive an unspecified bit-pattern.
2023 Undefined values may be of any type (other than label or void) and be used
2024 anywhere a constant is permitted.</p>
2026 <p>Undefined values are useful because they indicate to the compiler that the
2027 program is well defined no matter what value is used. This gives the
2028 compiler more freedom to optimize. Here are some examples of (potentially
2029 surprising) transformations that are valid (in pseudo IR):</p>
2032 <div class="doc_code">
2044 <p>This is safe because all of the output bits are affected by the undef bits.
2045 Any output bit can have a zero or one depending on the input bits.</p>
2047 <div class="doc_code">
2060 <p>These logical operations have bits that are not always affected by the input.
2061 For example, if "%X" has a zero bit, then the output of the 'and' operation will
2062 always be a zero, no matter what the corresponding bit from the undef is. As
2063 such, it is unsafe to optimize or assume that the result of the and is undef.
2064 However, it is safe to assume that all bits of the undef could be 0, and
2065 optimize the and to 0. Likewise, it is safe to assume that all the bits of
2066 the undef operand to the or could be set, allowing the or to be folded to
2069 <div class="doc_code">
2071 %A = select undef, %X, %Y
2072 %B = select undef, 42, %Y
2073 %C = select %X, %Y, undef
2085 <p>This set of examples show that undefined select (and conditional branch)
2086 conditions can go "either way" but they have to come from one of the two
2087 operands. In the %A example, if %X and %Y were both known to have a clear low
2088 bit, then %A would have to have a cleared low bit. However, in the %C example,
2089 the optimizer is allowed to assume that the undef operand could be the same as
2090 %Y, allowing the whole select to be eliminated.</p>
2093 <div class="doc_code">
2095 %A = xor undef, undef
2114 <p>This example points out that two undef operands are not necessarily the same.
2115 This can be surprising to people (and also matches C semantics) where they
2116 assume that "X^X" is always zero, even if X is undef. This isn't true for a
2117 number of reasons, but the short answer is that an undef "variable" can
2118 arbitrarily change its value over its "live range". This is true because the
2119 "variable" doesn't actually <em>have a live range</em>. Instead, the value is
2120 logically read from arbitrary registers that happen to be around when needed,
2121 so the value is not necessarily consistent over time. In fact, %A and %C need
2122 to have the same semantics or the core LLVM "replace all uses with" concept
2125 <div class="doc_code">
2135 <p>These examples show the crucial difference between an <em>undefined
2136 value</em> and <em>undefined behavior</em>. An undefined value (like undef) is
2137 allowed to have an arbitrary bit-pattern. This means that the %A operation
2138 can be constant folded to undef because the undef could be an SNaN, and fdiv is
2139 not (currently) defined on SNaN's. However, in the second example, we can make
2140 a more aggressive assumption: because the undef is allowed to be an arbitrary
2141 value, we are allowed to assume that it could be zero. Since a divide by zero
2142 has <em>undefined behavior</em>, we are allowed to assume that the operation
2143 does not execute at all. This allows us to delete the divide and all code after
2144 it: since the undefined operation "can't happen", the optimizer can assume that
2145 it occurs in dead code.
2148 <div class="doc_code">
2150 a: store undef -> %X
2151 b: store %X -> undef
2158 <p>These examples reiterate the fdiv example: a store "of" an undefined value
2159 can be assumed to not have any effect: we can assume that the value is
2160 overwritten with bits that happen to match what was already there. However, a
2161 store "to" an undefined location could clobber arbitrary memory, therefore, it
2162 has undefined behavior.</p>
2166 <!-- ======================================================================= -->
2167 <div class="doc_subsection"><a name="blockaddress">Addresses of Basic
2169 <div class="doc_text">
2171 <p><b><tt>blockaddress(@function, %block)</tt></b></p>
2173 <p>The '<tt>blockaddress</tt>' constant computes the address of the specified
2174 basic block in the specified function, and always has an i8* type. Taking
2175 the address of the entry block is illegal.</p>
2177 <p>This value only has defined behavior when used as an operand to the
2178 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>' instruction or for comparisons
2179 against null. Pointer equality tests between labels addresses is undefined
2180 behavior - though, again, comparison against null is ok, and no label is
2181 equal to the null pointer. This may also be passed around as an opaque
2182 pointer sized value as long as the bits are not inspected. This allows
2183 <tt>ptrtoint</tt> and arithmetic to be performed on these values so long as
2184 the original value is reconstituted before the <tt>indirectbr</tt>.</p>
2186 <p>Finally, some targets may provide defined semantics when
2187 using the value as the operand to an inline assembly, but that is target
2194 <!-- ======================================================================= -->
2195 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
2198 <div class="doc_text">
2200 <p>Constant expressions are used to allow expressions involving other constants
2201 to be used as constants. Constant expressions may be of
2202 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2203 operation that does not have side effects (e.g. load and call are not
2204 supported). The following is the syntax for constant expressions:</p>
2207 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
2208 <dd>Truncate a constant to another type. The bit size of CST must be larger
2209 than the bit size of TYPE. Both types must be integers.</dd>
2211 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
2212 <dd>Zero extend a constant to another type. The bit size of CST must be
2213 smaller or equal to the bit size of TYPE. Both types must be
2216 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
2217 <dd>Sign extend a constant to another type. The bit size of CST must be
2218 smaller or equal to the bit size of TYPE. Both types must be
2221 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
2222 <dd>Truncate a floating point constant to another floating point type. The
2223 size of CST must be larger than the size of TYPE. Both types must be
2224 floating point.</dd>
2226 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
2227 <dd>Floating point extend a constant to another type. The size of CST must be
2228 smaller or equal to the size of TYPE. Both types must be floating
2231 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
2232 <dd>Convert a floating point constant to the corresponding unsigned integer
2233 constant. TYPE must be a scalar or vector integer type. CST must be of
2234 scalar or vector floating point type. Both CST and TYPE must be scalars,
2235 or vectors of the same number of elements. If the value won't fit in the
2236 integer type, the results are undefined.</dd>
2238 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
2239 <dd>Convert a floating point constant to the corresponding signed integer
2240 constant. TYPE must be a scalar or vector integer type. CST must be of
2241 scalar or vector floating point type. Both CST and TYPE must be scalars,
2242 or vectors of the same number of elements. If the value won't fit in the
2243 integer type, the results are undefined.</dd>
2245 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
2246 <dd>Convert an unsigned integer constant to the corresponding floating point
2247 constant. TYPE must be a scalar or vector floating point type. CST must be
2248 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2249 vectors of the same number of elements. If the value won't fit in the
2250 floating point type, the results are undefined.</dd>
2252 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
2253 <dd>Convert a signed integer constant to the corresponding floating point
2254 constant. TYPE must be a scalar or vector floating point type. CST must be
2255 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2256 vectors of the same number of elements. If the value won't fit in the
2257 floating point type, the results are undefined.</dd>
2259 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
2260 <dd>Convert a pointer typed constant to the corresponding integer constant
2261 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2262 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2263 make it fit in <tt>TYPE</tt>.</dd>
2265 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
2266 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2267 type. CST must be of integer type. The CST value is zero extended,
2268 truncated, or unchanged to make it fit in a pointer size. This one is
2269 <i>really</i> dangerous!</dd>
2271 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
2272 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2273 are the same as those for the <a href="#i_bitcast">bitcast
2274 instruction</a>.</dd>
2276 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
2277 <dt><b><tt>getelementptr inbounds ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
2278 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2279 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2280 instruction, the index list may have zero or more indexes, which are
2281 required to make sense for the type of "CSTPTR".</dd>
2283 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
2284 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
2286 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
2287 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2289 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
2290 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2292 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
2293 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2296 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
2297 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2300 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
2301 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2304 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
2305 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2306 be any of the <a href="#binaryops">binary</a>
2307 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2308 on operands are the same as those for the corresponding instruction
2309 (e.g. no bitwise operations on floating point values are allowed).</dd>
2314 <!-- ======================================================================= -->
2315 <div class="doc_subsection"><a name="metadata">Embedded Metadata</a>
2318 <div class="doc_text">
2320 <p>Embedded metadata provides a way to attach arbitrary data to the instruction
2321 stream without affecting the behaviour of the program. There are two
2322 metadata primitives, strings and nodes. All metadata has the
2323 <tt>metadata</tt> type and is identified in syntax by a preceding exclamation
2324 point ('<tt>!</tt>').</p>
2326 <p>A metadata string is a string surrounded by double quotes. It can contain
2327 any character by escaping non-printable characters with "\xx" where "xx" is
2328 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
2330 <p>Metadata nodes are represented with notation similar to structure constants
2331 (a comma separated list of elements, surrounded by braces and preceded by an
2332 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2335 <p>A metadata node will attempt to track changes to the values it holds. In the
2336 event that a value is deleted, it will be replaced with a typeless
2337 "<tt>null</tt>", such as "<tt>metadata !{null, i32 10}</tt>".</p>
2339 <p>A named metadata is a collection of metadata nodes. For example: "<tt>!foo =
2340 metadata !{!4, !3}</tt>".
2342 <p>Optimizations may rely on metadata to provide additional information about
2343 the program that isn't available in the instructions, or that isn't easily
2344 computable. Similarly, the code generator may expect a certain metadata
2345 format to be used to express debugging information.</p>
2349 <!-- *********************************************************************** -->
2350 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2351 <!-- *********************************************************************** -->
2353 <!-- ======================================================================= -->
2354 <div class="doc_subsection">
2355 <a name="inlineasm">Inline Assembler Expressions</a>
2358 <div class="doc_text">
2360 <p>LLVM supports inline assembler expressions (as opposed
2361 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2362 a special value. This value represents the inline assembler as a string
2363 (containing the instructions to emit), a list of operand constraints (stored
2364 as a string), a flag that indicates whether or not the inline asm
2365 expression has side effects, and a flag indicating whether the function
2366 containing the asm needs to align its stack conservatively. An example
2367 inline assembler expression is:</p>
2369 <div class="doc_code">
2371 i32 (i32) asm "bswap $0", "=r,r"
2375 <p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2376 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2379 <div class="doc_code">
2381 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2385 <p>Inline asms with side effects not visible in the constraint list must be
2386 marked as having side effects. This is done through the use of the
2387 '<tt>sideeffect</tt>' keyword, like so:</p>
2389 <div class="doc_code">
2391 call void asm sideeffect "eieio", ""()
2395 <p>In some cases inline asms will contain code that will not work unless the
2396 stack is aligned in some way, such as calls or SSE instructions on x86,
2397 yet will not contain code that does that alignment within the asm.
2398 The compiler should make conservative assumptions about what the asm might
2399 contain and should generate its usual stack alignment code in the prologue
2400 if the '<tt>alignstack</tt>' keyword is present:</p>
2402 <div class="doc_code">
2404 call void asm alignstack "eieio", ""()
2408 <p>If both keywords appear the '<tt>sideeffect</tt>' keyword must come
2411 <p>TODO: The format of the asm and constraints string still need to be
2412 documented here. Constraints on what can be done (e.g. duplication, moving,
2413 etc need to be documented). This is probably best done by reference to
2414 another document that covers inline asm from a holistic perspective.</p>
2419 <!-- *********************************************************************** -->
2420 <div class="doc_section">
2421 <a name="intrinsic_globals">Intrinsic Global Variables</a>
2423 <!-- *********************************************************************** -->
2425 <p>LLVM has a number of "magic" global variables that contain data that affect
2426 code generation or other IR semantics. These are documented here. All globals
2427 of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2428 section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2431 <!-- ======================================================================= -->
2432 <div class="doc_subsection">
2433 <a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
2436 <div class="doc_text">
2438 <p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2439 href="#linkage_appending">appending linkage</a>. This array contains a list of
2440 pointers to global variables and functions which may optionally have a pointer
2441 cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2447 @llvm.used = appending global [2 x i8*] [
2449 i8* bitcast (i32* @Y to i8*)
2450 ], section "llvm.metadata"
2453 <p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2454 compiler, assembler, and linker are required to treat the symbol as if there is
2455 a reference to the global that it cannot see. For example, if a variable has
2456 internal linkage and no references other than that from the <tt>@llvm.used</tt>
2457 list, it cannot be deleted. This is commonly used to represent references from
2458 inline asms and other things the compiler cannot "see", and corresponds to
2459 "attribute((used))" in GNU C.</p>
2461 <p>On some targets, the code generator must emit a directive to the assembler or
2462 object file to prevent the assembler and linker from molesting the symbol.</p>
2466 <!-- ======================================================================= -->
2467 <div class="doc_subsection">
2468 <a name="intg_compiler_used">The '<tt>llvm.compiler.used</tt>' Global Variable</a>
2471 <div class="doc_text">
2473 <p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2474 <tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2475 touching the symbol. On targets that support it, this allows an intelligent
2476 linker to optimize references to the symbol without being impeded as it would be
2477 by <tt>@llvm.used</tt>.</p>
2479 <p>This is a rare construct that should only be used in rare circumstances, and
2480 should not be exposed to source languages.</p>
2484 <!-- ======================================================================= -->
2485 <div class="doc_subsection">
2486 <a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
2489 <div class="doc_text">
2491 <p>TODO: Describe this.</p>
2495 <!-- ======================================================================= -->
2496 <div class="doc_subsection">
2497 <a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
2500 <div class="doc_text">
2502 <p>TODO: Describe this.</p>
2507 <!-- *********************************************************************** -->
2508 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2509 <!-- *********************************************************************** -->
2511 <div class="doc_text">
2513 <p>The LLVM instruction set consists of several different classifications of
2514 instructions: <a href="#terminators">terminator
2515 instructions</a>, <a href="#binaryops">binary instructions</a>,
2516 <a href="#bitwiseops">bitwise binary instructions</a>,
2517 <a href="#memoryops">memory instructions</a>, and
2518 <a href="#otherops">other instructions</a>.</p>
2522 <!-- ======================================================================= -->
2523 <div class="doc_subsection"> <a name="terminators">Terminator
2524 Instructions</a> </div>
2526 <div class="doc_text">
2528 <p>As mentioned <a href="#functionstructure">previously</a>, every basic block
2529 in a program ends with a "Terminator" instruction, which indicates which
2530 block should be executed after the current block is finished. These
2531 terminator instructions typically yield a '<tt>void</tt>' value: they produce
2532 control flow, not values (the one exception being the
2533 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2535 <p>There are six different terminator instructions: the
2536 '<a href="#i_ret"><tt>ret</tt></a>' instruction, the
2537 '<a href="#i_br"><tt>br</tt></a>' instruction, the
2538 '<a href="#i_switch"><tt>switch</tt></a>' instruction, the
2539 '<a href="#i_indirectbr">'<tt>indirectbr</tt></a>' Instruction, the
2540 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the
2541 '<a href="#i_unwind"><tt>unwind</tt></a>' instruction, and the
2542 '<a href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2546 <!-- _______________________________________________________________________ -->
2547 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2548 Instruction</a> </div>
2550 <div class="doc_text">
2554 ret <type> <value> <i>; Return a value from a non-void function</i>
2555 ret void <i>; Return from void function</i>
2559 <p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
2560 a value) from a function back to the caller.</p>
2562 <p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
2563 value and then causes control flow, and one that just causes control flow to
2567 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
2568 return value. The type of the return value must be a
2569 '<a href="#t_firstclass">first class</a>' type.</p>
2571 <p>A function is not <a href="#wellformed">well formed</a> if it it has a
2572 non-void return type and contains a '<tt>ret</tt>' instruction with no return
2573 value or a return value with a type that does not match its type, or if it
2574 has a void return type and contains a '<tt>ret</tt>' instruction with a
2578 <p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
2579 the calling function's context. If the caller is a
2580 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
2581 instruction after the call. If the caller was an
2582 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
2583 the beginning of the "normal" destination block. If the instruction returns
2584 a value, that value shall set the call or invoke instruction's return
2589 ret i32 5 <i>; Return an integer value of 5</i>
2590 ret void <i>; Return from a void function</i>
2591 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
2595 <!-- _______________________________________________________________________ -->
2596 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2598 <div class="doc_text">
2602 br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2606 <p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
2607 different basic block in the current function. There are two forms of this
2608 instruction, corresponding to a conditional branch and an unconditional
2612 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
2613 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
2614 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
2618 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2619 argument is evaluated. If the value is <tt>true</tt>, control flows to the
2620 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2621 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2626 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
2627 br i1 %cond, label %IfEqual, label %IfUnequal
2629 <a href="#i_ret">ret</a> i32 1
2631 <a href="#i_ret">ret</a> i32 0
2636 <!-- _______________________________________________________________________ -->
2637 <div class="doc_subsubsection">
2638 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2641 <div class="doc_text">
2645 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2649 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2650 several different places. It is a generalization of the '<tt>br</tt>'
2651 instruction, allowing a branch to occur to one of many possible
2655 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2656 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
2657 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
2658 The table is not allowed to contain duplicate constant entries.</p>
2661 <p>The <tt>switch</tt> instruction specifies a table of values and
2662 destinations. When the '<tt>switch</tt>' instruction is executed, this table
2663 is searched for the given value. If the value is found, control flow is
2664 transferred to the corresponding destination; otherwise, control flow is
2665 transferred to the default destination.</p>
2667 <h5>Implementation:</h5>
2668 <p>Depending on properties of the target machine and the particular
2669 <tt>switch</tt> instruction, this instruction may be code generated in
2670 different ways. For example, it could be generated as a series of chained
2671 conditional branches or with a lookup table.</p>
2675 <i>; Emulate a conditional br instruction</i>
2676 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2677 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
2679 <i>; Emulate an unconditional br instruction</i>
2680 switch i32 0, label %dest [ ]
2682 <i>; Implement a jump table:</i>
2683 switch i32 %val, label %otherwise [ i32 0, label %onzero
2685 i32 2, label %ontwo ]
2691 <!-- _______________________________________________________________________ -->
2692 <div class="doc_subsubsection">
2693 <a name="i_indirectbr">'<tt>indirectbr</tt>' Instruction</a>
2696 <div class="doc_text">
2700 indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
2705 <p>The '<tt>indirectbr</tt>' instruction implements an indirect branch to a label
2706 within the current function, whose address is specified by
2707 "<tt>address</tt>". Address must be derived from a <a
2708 href="#blockaddress">blockaddress</a> constant.</p>
2712 <p>The '<tt>address</tt>' argument is the address of the label to jump to. The
2713 rest of the arguments indicate the full set of possible destinations that the
2714 address may point to. Blocks are allowed to occur multiple times in the
2715 destination list, though this isn't particularly useful.</p>
2717 <p>This destination list is required so that dataflow analysis has an accurate
2718 understanding of the CFG.</p>
2722 <p>Control transfers to the block specified in the address argument. All
2723 possible destination blocks must be listed in the label list, otherwise this
2724 instruction has undefined behavior. This implies that jumps to labels
2725 defined in other functions have undefined behavior as well.</p>
2727 <h5>Implementation:</h5>
2729 <p>This is typically implemented with a jump through a register.</p>
2733 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
2739 <!-- _______________________________________________________________________ -->
2740 <div class="doc_subsubsection">
2741 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2744 <div class="doc_text">
2748 <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>]
2749 to label <normal label> unwind label <exception label>
2753 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2754 function, with the possibility of control flow transfer to either the
2755 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
2756 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
2757 control flow will return to the "normal" label. If the callee (or any
2758 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
2759 instruction, control is interrupted and continued at the dynamically nearest
2760 "exception" label.</p>
2763 <p>This instruction requires several arguments:</p>
2766 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
2767 convention</a> the call should use. If none is specified, the call
2768 defaults to using C calling conventions.</li>
2770 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2771 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
2772 '<tt>inreg</tt>' attributes are valid here.</li>
2774 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2775 function value being invoked. In most cases, this is a direct function
2776 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
2777 off an arbitrary pointer to function value.</li>
2779 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2780 function to be invoked. </li>
2782 <li>'<tt>function args</tt>': argument list whose types match the function
2783 signature argument types. If the function signature indicates the
2784 function accepts a variable number of arguments, the extra arguments can
2787 <li>'<tt>normal label</tt>': the label reached when the called function
2788 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2790 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2791 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2793 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
2794 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2795 '<tt>readnone</tt>' attributes are valid here.</li>
2799 <p>This instruction is designed to operate as a standard
2800 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
2801 primary difference is that it establishes an association with a label, which
2802 is used by the runtime library to unwind the stack.</p>
2804 <p>This instruction is used in languages with destructors to ensure that proper
2805 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2806 exception. Additionally, this is important for implementation of
2807 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
2809 <p>For the purposes of the SSA form, the definition of the value returned by the
2810 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
2811 block to the "normal" label. If the callee unwinds then no return value is
2816 %retval = invoke i32 @Test(i32 15) to label %Continue
2817 unwind label %TestCleanup <i>; {i32}:retval set</i>
2818 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
2819 unwind label %TestCleanup <i>; {i32}:retval set</i>
2824 <!-- _______________________________________________________________________ -->
2826 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2827 Instruction</a> </div>
2829 <div class="doc_text">
2837 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2838 at the first callee in the dynamic call stack which used
2839 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
2840 This is primarily used to implement exception handling.</p>
2843 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
2844 immediately halt. The dynamic call stack is then searched for the
2845 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
2846 Once found, execution continues at the "exceptional" destination block
2847 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
2848 instruction in the dynamic call chain, undefined behavior results.</p>
2852 <!-- _______________________________________________________________________ -->
2854 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2855 Instruction</a> </div>
2857 <div class="doc_text">
2865 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2866 instruction is used to inform the optimizer that a particular portion of the
2867 code is not reachable. This can be used to indicate that the code after a
2868 no-return function cannot be reached, and other facts.</p>
2871 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2875 <!-- ======================================================================= -->
2876 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2878 <div class="doc_text">
2880 <p>Binary operators are used to do most of the computation in a program. They
2881 require two operands of the same type, execute an operation on them, and
2882 produce a single value. The operands might represent multiple data, as is
2883 the case with the <a href="#t_vector">vector</a> data type. The result value
2884 has the same type as its operands.</p>
2886 <p>There are several different binary operators:</p>
2890 <!-- _______________________________________________________________________ -->
2891 <div class="doc_subsubsection">
2892 <a name="i_add">'<tt>add</tt>' Instruction</a>
2895 <div class="doc_text">
2899 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2900 <result> = add nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2901 <result> = add nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2902 <result> = add nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2906 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2909 <p>The two arguments to the '<tt>add</tt>' instruction must
2910 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
2911 integer values. Both arguments must have identical types.</p>
2914 <p>The value produced is the integer sum of the two operands.</p>
2916 <p>If the sum has unsigned overflow, the result returned is the mathematical
2917 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
2919 <p>Because LLVM integers use a two's complement representation, this instruction
2920 is appropriate for both signed and unsigned integers.</p>
2922 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
2923 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
2924 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
2925 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
2929 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2934 <!-- _______________________________________________________________________ -->
2935 <div class="doc_subsubsection">
2936 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
2939 <div class="doc_text">
2943 <result> = fadd <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2947 <p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
2950 <p>The two arguments to the '<tt>fadd</tt>' instruction must be
2951 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2952 floating point values. Both arguments must have identical types.</p>
2955 <p>The value produced is the floating point sum of the two operands.</p>
2959 <result> = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
2964 <!-- _______________________________________________________________________ -->
2965 <div class="doc_subsubsection">
2966 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2969 <div class="doc_text">
2973 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2974 <result> = sub nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2975 <result> = sub nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2976 <result> = sub nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2980 <p>The '<tt>sub</tt>' instruction returns the difference of its two
2983 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
2984 '<tt>neg</tt>' instruction present in most other intermediate
2985 representations.</p>
2988 <p>The two arguments to the '<tt>sub</tt>' instruction must
2989 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
2990 integer values. Both arguments must have identical types.</p>
2993 <p>The value produced is the integer difference of the two operands.</p>
2995 <p>If the difference has unsigned overflow, the result returned is the
2996 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
2999 <p>Because LLVM integers use a two's complement representation, this instruction
3000 is appropriate for both signed and unsigned integers.</p>
3002 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3003 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3004 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
3005 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
3009 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
3010 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
3015 <!-- _______________________________________________________________________ -->
3016 <div class="doc_subsubsection">
3017 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
3020 <div class="doc_text">
3024 <result> = fsub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3028 <p>The '<tt>fsub</tt>' instruction returns the difference of its two
3031 <p>Note that the '<tt>fsub</tt>' instruction is used to represent the
3032 '<tt>fneg</tt>' instruction present in most other intermediate
3033 representations.</p>
3036 <p>The two arguments to the '<tt>fsub</tt>' instruction must be
3037 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3038 floating point values. Both arguments must have identical types.</p>
3041 <p>The value produced is the floating point difference of the two operands.</p>
3045 <result> = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
3046 <result> = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
3051 <!-- _______________________________________________________________________ -->
3052 <div class="doc_subsubsection">
3053 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
3056 <div class="doc_text">
3060 <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3061 <result> = mul nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3062 <result> = mul nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3063 <result> = mul nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3067 <p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
3070 <p>The two arguments to the '<tt>mul</tt>' instruction must
3071 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3072 integer values. Both arguments must have identical types.</p>
3075 <p>The value produced is the integer product of the two operands.</p>
3077 <p>If the result of the multiplication has unsigned overflow, the result
3078 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
3079 width of the result.</p>
3081 <p>Because LLVM integers use a two's complement representation, and the result
3082 is the same width as the operands, this instruction returns the correct
3083 result for both signed and unsigned integers. If a full product
3084 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
3085 be sign-extended or zero-extended as appropriate to the width of the full
3088 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3089 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3090 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
3091 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
3095 <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
3100 <!-- _______________________________________________________________________ -->
3101 <div class="doc_subsubsection">
3102 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
3105 <div class="doc_text">
3109 <result> = fmul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3113 <p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
3116 <p>The two arguments to the '<tt>fmul</tt>' instruction must be
3117 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3118 floating point values. Both arguments must have identical types.</p>
3121 <p>The value produced is the floating point product of the two operands.</p>
3125 <result> = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
3130 <!-- _______________________________________________________________________ -->
3131 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
3134 <div class="doc_text">
3138 <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3142 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
3145 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
3146 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3147 values. Both arguments must have identical types.</p>
3150 <p>The value produced is the unsigned integer quotient of the two operands.</p>
3152 <p>Note that unsigned integer division and signed integer division are distinct
3153 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
3155 <p>Division by zero leads to undefined behavior.</p>
3159 <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3164 <!-- _______________________________________________________________________ -->
3165 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
3168 <div class="doc_text">
3172 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3173 <result> = sdiv exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3177 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
3180 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
3181 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3182 values. Both arguments must have identical types.</p>
3185 <p>The value produced is the signed integer quotient of the two operands rounded
3188 <p>Note that signed integer division and unsigned integer division are distinct
3189 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
3191 <p>Division by zero leads to undefined behavior. Overflow also leads to
3192 undefined behavior; this is a rare case, but can occur, for example, by doing
3193 a 32-bit division of -2147483648 by -1.</p>
3195 <p>If the <tt>exact</tt> keyword is present, the result value of the
3196 <tt>sdiv</tt> is undefined if the result would be rounded or if overflow
3201 <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3206 <!-- _______________________________________________________________________ -->
3207 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
3208 Instruction</a> </div>
3210 <div class="doc_text">
3214 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3218 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
3221 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
3222 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3223 floating point values. Both arguments must have identical types.</p>
3226 <p>The value produced is the floating point quotient of the two operands.</p>
3230 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
3235 <!-- _______________________________________________________________________ -->
3236 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
3239 <div class="doc_text">
3243 <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3247 <p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3248 division of its two arguments.</p>
3251 <p>The two arguments to the '<tt>urem</tt>' instruction must be
3252 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3253 values. Both arguments must have identical types.</p>
3256 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
3257 This instruction always performs an unsigned division to get the
3260 <p>Note that unsigned integer remainder and signed integer remainder are
3261 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3263 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
3267 <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3272 <!-- _______________________________________________________________________ -->
3273 <div class="doc_subsubsection">
3274 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
3277 <div class="doc_text">
3281 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3285 <p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3286 division of its two operands. This instruction can also take
3287 <a href="#t_vector">vector</a> versions of the values in which case the
3288 elements must be integers.</p>
3291 <p>The two arguments to the '<tt>srem</tt>' instruction must be
3292 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3293 values. Both arguments must have identical types.</p>
3296 <p>This instruction returns the <i>remainder</i> of a division (where the result
3297 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
3298 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
3299 a value. For more information about the difference,
3300 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3301 Math Forum</a>. For a table of how this is implemented in various languages,
3302 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3303 Wikipedia: modulo operation</a>.</p>
3305 <p>Note that signed integer remainder and unsigned integer remainder are
3306 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3308 <p>Taking the remainder of a division by zero leads to undefined behavior.
3309 Overflow also leads to undefined behavior; this is a rare case, but can
3310 occur, for example, by taking the remainder of a 32-bit division of
3311 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3312 lets srem be implemented using instructions that return both the result of
3313 the division and the remainder.)</p>
3317 <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3322 <!-- _______________________________________________________________________ -->
3323 <div class="doc_subsubsection">
3324 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
3326 <div class="doc_text">
3330 <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3334 <p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3335 its two operands.</p>
3338 <p>The two arguments to the '<tt>frem</tt>' instruction must be
3339 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3340 floating point values. Both arguments must have identical types.</p>
3343 <p>This instruction returns the <i>remainder</i> of a division. The remainder
3344 has the same sign as the dividend.</p>
3348 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
3353 <!-- ======================================================================= -->
3354 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
3355 Operations</a> </div>
3357 <div class="doc_text">
3359 <p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3360 program. They are generally very efficient instructions and can commonly be
3361 strength reduced from other instructions. They require two operands of the
3362 same type, execute an operation on them, and produce a single value. The
3363 resulting value is the same type as its operands.</p>
3367 <!-- _______________________________________________________________________ -->
3368 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
3369 Instruction</a> </div>
3371 <div class="doc_text">
3375 <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3379 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3380 a specified number of bits.</p>
3383 <p>Both arguments to the '<tt>shl</tt>' instruction must be the
3384 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3385 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3388 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3389 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3390 is (statically or dynamically) negative or equal to or larger than the number
3391 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3392 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3393 shift amount in <tt>op2</tt>.</p>
3397 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
3398 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
3399 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
3400 <result> = shl i32 1, 32 <i>; undefined</i>
3401 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
3406 <!-- _______________________________________________________________________ -->
3407 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
3408 Instruction</a> </div>
3410 <div class="doc_text">
3414 <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3418 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
3419 operand shifted to the right a specified number of bits with zero fill.</p>
3422 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
3423 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3424 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3427 <p>This instruction always performs a logical shift right operation. The most
3428 significant bits of the result will be filled with zero bits after the shift.
3429 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
3430 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3431 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3432 shift amount in <tt>op2</tt>.</p>
3436 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
3437 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
3438 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
3439 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
3440 <result> = lshr i32 1, 32 <i>; undefined</i>
3441 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
3446 <!-- _______________________________________________________________________ -->
3447 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
3448 Instruction</a> </div>
3449 <div class="doc_text">
3453 <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3457 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3458 operand shifted to the right a specified number of bits with sign
3462 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
3463 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3464 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3467 <p>This instruction always performs an arithmetic shift right operation, The
3468 most significant bits of the result will be filled with the sign bit
3469 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3470 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
3471 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
3472 the corresponding shift amount in <tt>op2</tt>.</p>
3476 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3477 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3478 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3479 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
3480 <result> = ashr i32 1, 32 <i>; undefined</i>
3481 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
3486 <!-- _______________________________________________________________________ -->
3487 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
3488 Instruction</a> </div>
3490 <div class="doc_text">
3494 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3498 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
3502 <p>The two arguments to the '<tt>and</tt>' instruction must be
3503 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3504 values. Both arguments must have identical types.</p>
3507 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
3509 <table border="1" cellspacing="0" cellpadding="4">
3541 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
3542 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
3543 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
3546 <!-- _______________________________________________________________________ -->
3547 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
3549 <div class="doc_text">
3553 <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3557 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
3561 <p>The two arguments to the '<tt>or</tt>' instruction must be
3562 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3563 values. Both arguments must have identical types.</p>
3566 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
3568 <table border="1" cellspacing="0" cellpadding="4">
3600 <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
3601 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
3602 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
3607 <!-- _______________________________________________________________________ -->
3608 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3609 Instruction</a> </div>
3611 <div class="doc_text">
3615 <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3619 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
3620 its two operands. The <tt>xor</tt> is used to implement the "one's
3621 complement" operation, which is the "~" operator in C.</p>
3624 <p>The two arguments to the '<tt>xor</tt>' instruction must be
3625 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3626 values. Both arguments must have identical types.</p>
3629 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
3631 <table border="1" cellspacing="0" cellpadding="4">
3663 <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
3664 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3665 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3666 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
3671 <!-- ======================================================================= -->
3672 <div class="doc_subsection">
3673 <a name="vectorops">Vector Operations</a>
3676 <div class="doc_text">
3678 <p>LLVM supports several instructions to represent vector operations in a
3679 target-independent manner. These instructions cover the element-access and
3680 vector-specific operations needed to process vectors effectively. While LLVM
3681 does directly support these vector operations, many sophisticated algorithms
3682 will want to use target-specific intrinsics to take full advantage of a
3683 specific target.</p>
3687 <!-- _______________________________________________________________________ -->
3688 <div class="doc_subsubsection">
3689 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3692 <div class="doc_text">
3696 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
3700 <p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
3701 from a vector at a specified index.</p>
3705 <p>The first operand of an '<tt>extractelement</tt>' instruction is a value
3706 of <a href="#t_vector">vector</a> type. The second operand is an index
3707 indicating the position from which to extract the element. The index may be
3711 <p>The result is a scalar of the same type as the element type of
3712 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3713 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3714 results are undefined.</p>
3718 <result> = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
3723 <!-- _______________________________________________________________________ -->
3724 <div class="doc_subsubsection">
3725 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
3728 <div class="doc_text">
3732 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
3736 <p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
3737 vector at a specified index.</p>
3740 <p>The first operand of an '<tt>insertelement</tt>' instruction is a value
3741 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
3742 whose type must equal the element type of the first operand. The third
3743 operand is an index indicating the position at which to insert the value.
3744 The index may be a variable.</p>
3747 <p>The result is a vector of the same type as <tt>val</tt>. Its element values
3748 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
3749 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3750 results are undefined.</p>
3754 <result> = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
3759 <!-- _______________________________________________________________________ -->
3760 <div class="doc_subsubsection">
3761 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3764 <div class="doc_text">
3768 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
3772 <p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
3773 from two input vectors, returning a vector with the same element type as the
3774 input and length that is the same as the shuffle mask.</p>
3777 <p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3778 with types that match each other. The third argument is a shuffle mask whose
3779 element type is always 'i32'. The result of the instruction is a vector
3780 whose length is the same as the shuffle mask and whose element type is the
3781 same as the element type of the first two operands.</p>
3783 <p>The shuffle mask operand is required to be a constant vector with either
3784 constant integer or undef values.</p>
3787 <p>The elements of the two input vectors are numbered from left to right across
3788 both of the vectors. The shuffle mask operand specifies, for each element of
3789 the result vector, which element of the two input vectors the result element
3790 gets. The element selector may be undef (meaning "don't care") and the
3791 second operand may be undef if performing a shuffle from only one vector.</p>
3795 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3796 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
3797 <result> = shufflevector <4 x i32> %v1, <4 x i32> undef,
3798 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
3799 <result> = shufflevector <8 x i32> %v1, <8 x i32> undef,
3800 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
3801 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3802 <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>
3807 <!-- ======================================================================= -->
3808 <div class="doc_subsection">
3809 <a name="aggregateops">Aggregate Operations</a>
3812 <div class="doc_text">
3814 <p>LLVM supports several instructions for working with aggregate values.</p>
3818 <!-- _______________________________________________________________________ -->
3819 <div class="doc_subsubsection">
3820 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3823 <div class="doc_text">
3827 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
3831 <p>The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3832 or array element from an aggregate value.</p>
3835 <p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
3836 of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type. The
3837 operands are constant indices to specify which value to extract in a similar
3838 manner as indices in a
3839 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
3842 <p>The result is the value at the position in the aggregate specified by the
3847 <result> = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
3852 <!-- _______________________________________________________________________ -->
3853 <div class="doc_subsubsection">
3854 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3857 <div class="doc_text">
3861 <result> = insertvalue <aggregate type> <val>, <ty> <val>, <idx> <i>; yields <n x <ty>></i>
3865 <p>The '<tt>insertvalue</tt>' instruction inserts a value into a struct field or
3866 array element in an aggregate.</p>
3870 <p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
3871 of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type. The
3872 second operand is a first-class value to insert. The following operands are
3873 constant indices indicating the position at which to insert the value in a
3874 similar manner as indices in a
3875 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction. The
3876 value to insert must have the same type as the value identified by the
3880 <p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
3881 that of <tt>val</tt> except that the value at the position specified by the
3882 indices is that of <tt>elt</tt>.</p>
3886 <result> = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
3892 <!-- ======================================================================= -->
3893 <div class="doc_subsection">
3894 <a name="memoryops">Memory Access and Addressing Operations</a>
3897 <div class="doc_text">
3899 <p>A key design point of an SSA-based representation is how it represents
3900 memory. In LLVM, no memory locations are in SSA form, which makes things
3901 very simple. This section describes how to read, write, and allocate
3906 <!-- _______________________________________________________________________ -->
3907 <div class="doc_subsubsection">
3908 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3911 <div class="doc_text">
3915 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3919 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3920 currently executing function, to be automatically released when this function
3921 returns to its caller. The object is always allocated in the generic address
3922 space (address space zero).</p>
3925 <p>The '<tt>alloca</tt>' instruction
3926 allocates <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the
3927 runtime stack, returning a pointer of the appropriate type to the program.
3928 If "NumElements" is specified, it is the number of elements allocated,
3929 otherwise "NumElements" is defaulted to be one. If a constant alignment is
3930 specified, the value result of the allocation is guaranteed to be aligned to
3931 at least that boundary. If not specified, or if zero, the target can choose
3932 to align the allocation on any convenient boundary compatible with the
3935 <p>'<tt>type</tt>' may be any sized type.</p>
3938 <p>Memory is allocated; a pointer is returned. The operation is undefined if
3939 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
3940 memory is automatically released when the function returns. The
3941 '<tt>alloca</tt>' instruction is commonly used to represent automatic
3942 variables that must have an address available. When the function returns
3943 (either with the <tt><a href="#i_ret">ret</a></tt>
3944 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
3945 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
3949 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3950 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3951 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3952 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3957 <!-- _______________________________________________________________________ -->
3958 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3959 Instruction</a> </div>
3961 <div class="doc_text">
3965 <result> = load <ty>* <pointer>[, align <alignment>]
3966 <result> = volatile load <ty>* <pointer>[, align <alignment>]
3970 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3973 <p>The argument to the '<tt>load</tt>' instruction specifies the memory address
3974 from which to load. The pointer must point to
3975 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3976 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
3977 number or order of execution of this <tt>load</tt> with other
3978 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3981 <p>The optional constant "align" argument specifies the alignment of the
3982 operation (that is, the alignment of the memory address). A value of 0 or an
3983 omitted "align" argument means that the operation has the preferential
3984 alignment for the target. It is the responsibility of the code emitter to
3985 ensure that the alignment information is correct. Overestimating the
3986 alignment results in an undefined behavior. Underestimating the alignment may
3987 produce less efficient code. An alignment of 1 is always safe.</p>
3990 <p>The location of memory pointed to is loaded. If the value being loaded is of
3991 scalar type then the number of bytes read does not exceed the minimum number
3992 of bytes needed to hold all bits of the type. For example, loading an
3993 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
3994 <tt>i20</tt> with a size that is not an integral number of bytes, the result
3995 is undefined if the value was not originally written using a store of the
4000 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4001 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
4002 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
4007 <!-- _______________________________________________________________________ -->
4008 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
4009 Instruction</a> </div>
4011 <div class="doc_text">
4015 store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
4016 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
4020 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
4023 <p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
4024 and an address at which to store it. The type of the
4025 '<tt><pointer></tt>' operand must be a pointer to
4026 the <a href="#t_firstclass">first class</a> type of the
4027 '<tt><value></tt>' operand. If the <tt>store</tt> is marked
4028 as <tt>volatile</tt>, then the optimizer is not allowed to modify the number
4029 or order of execution of this <tt>store</tt> with other
4030 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
4033 <p>The optional constant "align" argument specifies the alignment of the
4034 operation (that is, the alignment of the memory address). A value of 0 or an
4035 omitted "align" argument means that the operation has the preferential
4036 alignment for the target. It is the responsibility of the code emitter to
4037 ensure that the alignment information is correct. Overestimating the
4038 alignment results in an undefined behavior. Underestimating the alignment may
4039 produce less efficient code. An alignment of 1 is always safe.</p>
4042 <p>The contents of memory are updated to contain '<tt><value></tt>' at the
4043 location specified by the '<tt><pointer></tt>' operand. If
4044 '<tt><value></tt>' is of scalar type then the number of bytes written
4045 does not exceed the minimum number of bytes needed to hold all bits of the
4046 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
4047 writing a value of a type like <tt>i20</tt> with a size that is not an
4048 integral number of bytes, it is unspecified what happens to the extra bits
4049 that do not belong to the type, but they will typically be overwritten.</p>
4053 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4054 store i32 3, i32* %ptr <i>; yields {void}</i>
4055 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
4060 <!-- _______________________________________________________________________ -->
4061 <div class="doc_subsubsection">
4062 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
4065 <div class="doc_text">
4069 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
4070 <result> = getelementptr inbounds <pty>* <ptrval>{, <ty> <idx>}*
4074 <p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
4075 subelement of an aggregate data structure. It performs address calculation
4076 only and does not access memory.</p>
4079 <p>The first argument is always a pointer, and forms the basis of the
4080 calculation. The remaining arguments are indices that indicate which of the
4081 elements of the aggregate object are indexed. The interpretation of each
4082 index is dependent on the type being indexed into. The first index always
4083 indexes the pointer value given as the first argument, the second index
4084 indexes a value of the type pointed to (not necessarily the value directly
4085 pointed to, since the first index can be non-zero), etc. The first type
4086 indexed into must be a pointer value, subsequent types can be arrays, vectors
4087 and structs. Note that subsequent types being indexed into can never be
4088 pointers, since that would require loading the pointer before continuing
4091 <p>The type of each index argument depends on the type it is indexing into.
4092 When indexing into a (optionally packed) structure, only <tt>i32</tt> integer
4093 <b>constants</b> are allowed. When indexing into an array, pointer or
4094 vector, integers of any width are allowed, and they are not required to be
4097 <p>For example, let's consider a C code fragment and how it gets compiled to
4100 <div class="doc_code">
4113 int *foo(struct ST *s) {
4114 return &s[1].Z.B[5][13];
4119 <p>The LLVM code generated by the GCC frontend is:</p>
4121 <div class="doc_code">
4123 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
4124 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
4126 define i32* @foo(%ST* %s) {
4128 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
4135 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
4136 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
4137 }</tt>' type, a structure. The second index indexes into the third element
4138 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
4139 i8 }</tt>' type, another structure. The third index indexes into the second
4140 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
4141 array. The two dimensions of the array are subscripted into, yielding an
4142 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
4143 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
4145 <p>Note that it is perfectly legal to index partially through a structure,
4146 returning a pointer to an inner element. Because of this, the LLVM code for
4147 the given testcase is equivalent to:</p>
4150 define i32* @foo(%ST* %s) {
4151 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
4152 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
4153 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
4154 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
4155 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
4160 <p>If the <tt>inbounds</tt> keyword is present, the result value of the
4161 <tt>getelementptr</tt> is undefined if the base pointer is not an
4162 <i>in bounds</i> address of an allocated object, or if any of the addresses
4163 that would be formed by successive addition of the offsets implied by the
4164 indices to the base address with infinitely precise arithmetic are not an
4165 <i>in bounds</i> address of that allocated object.
4166 The <i>in bounds</i> addresses for an allocated object are all the addresses
4167 that point into the object, plus the address one byte past the end.</p>
4169 <p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
4170 the base address with silently-wrapping two's complement arithmetic, and
4171 the result value of the <tt>getelementptr</tt> may be outside the object
4172 pointed to by the base pointer. The result value may not necessarily be
4173 used to access memory though, even if it happens to point into allocated
4174 storage. See the <a href="#pointeraliasing">Pointer Aliasing Rules</a>
4175 section for more information.</p>
4177 <p>The getelementptr instruction is often confusing. For some more insight into
4178 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
4182 <i>; yields [12 x i8]*:aptr</i>
4183 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
4184 <i>; yields i8*:vptr</i>
4185 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
4186 <i>; yields i8*:eptr</i>
4187 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
4188 <i>; yields i32*:iptr</i>
4189 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
4194 <!-- ======================================================================= -->
4195 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
4198 <div class="doc_text">
4200 <p>The instructions in this category are the conversion instructions (casting)
4201 which all take a single operand and a type. They perform various bit
4202 conversions on the operand.</p>
4206 <!-- _______________________________________________________________________ -->
4207 <div class="doc_subsubsection">
4208 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
4210 <div class="doc_text">
4214 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
4218 <p>The '<tt>trunc</tt>' instruction truncates its operand to the
4219 type <tt>ty2</tt>.</p>
4222 <p>The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
4223 be an <a href="#t_integer">integer</a> type, and a type that specifies the
4224 size and type of the result, which must be
4225 an <a href="#t_integer">integer</a> type. The bit size of <tt>value</tt> must
4226 be larger than the bit size of <tt>ty2</tt>. Equal sized types are not
4230 <p>The '<tt>trunc</tt>' instruction truncates the high order bits
4231 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
4232 source size must be larger than the destination size, <tt>trunc</tt> cannot
4233 be a <i>no-op cast</i>. It will always truncate bits.</p>
4237 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
4238 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
4239 %Z = trunc i32 122 to i1 <i>; yields i1:false</i>
4244 <!-- _______________________________________________________________________ -->
4245 <div class="doc_subsubsection">
4246 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
4248 <div class="doc_text">
4252 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
4256 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
4261 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
4262 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4263 also be of <a href="#t_integer">integer</a> type. The bit size of the
4264 <tt>value</tt> must be smaller than the bit size of the destination type,
4268 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
4269 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
4271 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
4275 %X = zext i32 257 to i64 <i>; yields i64:257</i>
4276 %Y = zext i1 true to i32 <i>; yields i32:1</i>
4281 <!-- _______________________________________________________________________ -->
4282 <div class="doc_subsubsection">
4283 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
4285 <div class="doc_text">
4289 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
4293 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
4296 <p>The '<tt>sext</tt>' instruction takes a value to cast, which must be of
4297 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4298 also be of <a href="#t_integer">integer</a> type. The bit size of the
4299 <tt>value</tt> must be smaller than the bit size of the destination type,
4303 <p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
4304 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
4305 of the type <tt>ty2</tt>.</p>
4307 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
4311 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
4312 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
4317 <!-- _______________________________________________________________________ -->
4318 <div class="doc_subsubsection">
4319 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
4322 <div class="doc_text">
4326 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
4330 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
4334 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
4335 point</a> value to cast and a <a href="#t_floating">floating point</a> type
4336 to cast it to. The size of <tt>value</tt> must be larger than the size of
4337 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
4338 <i>no-op cast</i>.</p>
4341 <p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
4342 <a href="#t_floating">floating point</a> type to a smaller
4343 <a href="#t_floating">floating point</a> type. If the value cannot fit
4344 within the destination type, <tt>ty2</tt>, then the results are
4349 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4350 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4355 <!-- _______________________________________________________________________ -->
4356 <div class="doc_subsubsection">
4357 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
4359 <div class="doc_text">
4363 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
4367 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
4368 floating point value.</p>
4371 <p>The '<tt>fpext</tt>' instruction takes a
4372 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
4373 a <a href="#t_floating">floating point</a> type to cast it to. The source
4374 type must be smaller than the destination type.</p>
4377 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
4378 <a href="#t_floating">floating point</a> type to a larger
4379 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4380 used to make a <i>no-op cast</i> because it always changes bits. Use
4381 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
4385 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
4386 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
4391 <!-- _______________________________________________________________________ -->
4392 <div class="doc_subsubsection">
4393 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
4395 <div class="doc_text">
4399 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
4403 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
4404 unsigned integer equivalent of type <tt>ty2</tt>.</p>
4407 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
4408 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4409 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4410 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4411 vector integer type with the same number of elements as <tt>ty</tt></p>
4414 <p>The '<tt>fptoui</tt>' instruction converts its
4415 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4416 towards zero) unsigned integer value. If the value cannot fit
4417 in <tt>ty2</tt>, the results are undefined.</p>
4421 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
4422 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
4423 %Z = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
4428 <!-- _______________________________________________________________________ -->
4429 <div class="doc_subsubsection">
4430 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
4432 <div class="doc_text">
4436 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
4440 <p>The '<tt>fptosi</tt>' instruction converts
4441 <a href="#t_floating">floating point</a> <tt>value</tt> to
4442 type <tt>ty2</tt>.</p>
4445 <p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
4446 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4447 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4448 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4449 vector integer type with the same number of elements as <tt>ty</tt></p>
4452 <p>The '<tt>fptosi</tt>' instruction converts its
4453 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4454 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4455 the results are undefined.</p>
4459 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
4460 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
4461 %Z = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
4466 <!-- _______________________________________________________________________ -->
4467 <div class="doc_subsubsection">
4468 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
4470 <div class="doc_text">
4474 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4478 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
4479 integer and converts that value to the <tt>ty2</tt> type.</p>
4482 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
4483 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4484 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4485 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4486 floating point type with the same number of elements as <tt>ty</tt></p>
4489 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
4490 integer quantity and converts it to the corresponding floating point
4491 value. If the value cannot fit in the floating point value, the results are
4496 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
4497 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
4502 <!-- _______________________________________________________________________ -->
4503 <div class="doc_subsubsection">
4504 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
4506 <div class="doc_text">
4510 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4514 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
4515 and converts that value to the <tt>ty2</tt> type.</p>
4518 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
4519 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4520 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4521 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4522 floating point type with the same number of elements as <tt>ty</tt></p>
4525 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
4526 quantity and converts it to the corresponding floating point value. If the
4527 value cannot fit in the floating point value, the results are undefined.</p>
4531 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
4532 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
4537 <!-- _______________________________________________________________________ -->
4538 <div class="doc_subsubsection">
4539 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4541 <div class="doc_text">
4545 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
4549 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4550 the integer type <tt>ty2</tt>.</p>
4553 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4554 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
4555 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
4558 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4559 <tt>ty2</tt> by interpreting the pointer value as an integer and either
4560 truncating or zero extending that value to the size of the integer type. If
4561 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4562 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4563 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4568 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4569 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4574 <!-- _______________________________________________________________________ -->
4575 <div class="doc_subsubsection">
4576 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4578 <div class="doc_text">
4582 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
4586 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
4587 pointer type, <tt>ty2</tt>.</p>
4590 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4591 value to cast, and a type to cast it to, which must be a
4592 <a href="#t_pointer">pointer</a> type.</p>
4595 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4596 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
4597 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4598 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
4599 than the size of a pointer then a zero extension is done. If they are the
4600 same size, nothing is done (<i>no-op cast</i>).</p>
4604 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4605 %Y = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4606 %Z = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4611 <!-- _______________________________________________________________________ -->
4612 <div class="doc_subsubsection">
4613 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4615 <div class="doc_text">
4619 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
4623 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4624 <tt>ty2</tt> without changing any bits.</p>
4627 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
4628 non-aggregate first class value, and a type to cast it to, which must also be
4629 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
4630 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
4631 identical. If the source type is a pointer, the destination type must also be
4632 a pointer. This instruction supports bitwise conversion of vectors to
4633 integers and to vectors of other types (as long as they have the same
4637 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4638 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4639 this conversion. The conversion is done as if the <tt>value</tt> had been
4640 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
4641 be converted to other pointer types with this instruction. To convert
4642 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
4643 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4647 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4648 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
4649 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
4654 <!-- ======================================================================= -->
4655 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4657 <div class="doc_text">
4659 <p>The instructions in this category are the "miscellaneous" instructions, which
4660 defy better classification.</p>
4664 <!-- _______________________________________________________________________ -->
4665 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4668 <div class="doc_text">
4672 <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4676 <p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
4677 boolean values based on comparison of its two integer, integer vector, or
4678 pointer operands.</p>
4681 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
4682 the condition code indicating the kind of comparison to perform. It is not a
4683 value, just a keyword. The possible condition code are:</p>
4686 <li><tt>eq</tt>: equal</li>
4687 <li><tt>ne</tt>: not equal </li>
4688 <li><tt>ugt</tt>: unsigned greater than</li>
4689 <li><tt>uge</tt>: unsigned greater or equal</li>
4690 <li><tt>ult</tt>: unsigned less than</li>
4691 <li><tt>ule</tt>: unsigned less or equal</li>
4692 <li><tt>sgt</tt>: signed greater than</li>
4693 <li><tt>sge</tt>: signed greater or equal</li>
4694 <li><tt>slt</tt>: signed less than</li>
4695 <li><tt>sle</tt>: signed less or equal</li>
4698 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
4699 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
4700 typed. They must also be identical types.</p>
4703 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
4704 condition code given as <tt>cond</tt>. The comparison performed always yields
4705 either an <a href="#t_integer"><tt>i1</tt></a> or vector of <tt>i1</tt>
4706 result, as follows:</p>
4709 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
4710 <tt>false</tt> otherwise. No sign interpretation is necessary or
4713 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
4714 <tt>false</tt> otherwise. No sign interpretation is necessary or
4717 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
4718 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4720 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
4721 <tt>true</tt> if <tt>op1</tt> is greater than or equal
4722 to <tt>op2</tt>.</li>
4724 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
4725 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4727 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
4728 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4730 <li><tt>sgt</tt>: interprets the operands as signed values and yields
4731 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4733 <li><tt>sge</tt>: interprets the operands as signed values and yields
4734 <tt>true</tt> if <tt>op1</tt> is greater than or equal
4735 to <tt>op2</tt>.</li>
4737 <li><tt>slt</tt>: interprets the operands as signed values and yields
4738 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4740 <li><tt>sle</tt>: interprets the operands as signed values and yields
4741 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4744 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
4745 values are compared as if they were integers.</p>
4747 <p>If the operands are integer vectors, then they are compared element by
4748 element. The result is an <tt>i1</tt> vector with the same number of elements
4749 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
4753 <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
4754 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
4755 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
4756 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4757 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4758 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
4761 <p>Note that the code generator does not yet support vector types with
4762 the <tt>icmp</tt> instruction.</p>
4766 <!-- _______________________________________________________________________ -->
4767 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4770 <div class="doc_text">
4774 <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4778 <p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
4779 values based on comparison of its operands.</p>
4781 <p>If the operands are floating point scalars, then the result type is a boolean
4782 (<a href="#t_integer"><tt>i1</tt></a>).</p>
4784 <p>If the operands are floating point vectors, then the result type is a vector
4785 of boolean with the same number of elements as the operands being
4789 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4790 the condition code indicating the kind of comparison to perform. It is not a
4791 value, just a keyword. The possible condition code are:</p>
4794 <li><tt>false</tt>: no comparison, always returns false</li>
4795 <li><tt>oeq</tt>: ordered and equal</li>
4796 <li><tt>ogt</tt>: ordered and greater than </li>
4797 <li><tt>oge</tt>: ordered and greater than or equal</li>
4798 <li><tt>olt</tt>: ordered and less than </li>
4799 <li><tt>ole</tt>: ordered and less than or equal</li>
4800 <li><tt>one</tt>: ordered and not equal</li>
4801 <li><tt>ord</tt>: ordered (no nans)</li>
4802 <li><tt>ueq</tt>: unordered or equal</li>
4803 <li><tt>ugt</tt>: unordered or greater than </li>
4804 <li><tt>uge</tt>: unordered or greater than or equal</li>
4805 <li><tt>ult</tt>: unordered or less than </li>
4806 <li><tt>ule</tt>: unordered or less than or equal</li>
4807 <li><tt>une</tt>: unordered or not equal</li>
4808 <li><tt>uno</tt>: unordered (either nans)</li>
4809 <li><tt>true</tt>: no comparison, always returns true</li>
4812 <p><i>Ordered</i> means that neither operand is a QNAN while
4813 <i>unordered</i> means that either operand may be a QNAN.</p>
4815 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
4816 a <a href="#t_floating">floating point</a> type or
4817 a <a href="#t_vector">vector</a> of floating point type. They must have
4818 identical types.</p>
4821 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4822 according to the condition code given as <tt>cond</tt>. If the operands are
4823 vectors, then the vectors are compared element by element. Each comparison
4824 performed always yields an <a href="#t_integer">i1</a> result, as
4828 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4830 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4831 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4833 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4834 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
4836 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4837 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4839 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4840 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4842 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4843 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4845 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4846 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4848 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4850 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
4851 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4853 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
4854 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4856 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
4857 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4859 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
4860 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4862 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
4863 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4865 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
4866 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4868 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4870 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4875 <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
4876 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4877 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4878 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
4881 <p>Note that the code generator does not yet support vector types with
4882 the <tt>fcmp</tt> instruction.</p>
4886 <!-- _______________________________________________________________________ -->
4887 <div class="doc_subsubsection">
4888 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4891 <div class="doc_text">
4895 <result> = phi <ty> [ <val0>, <label0>], ...
4899 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in the
4900 SSA graph representing the function.</p>
4903 <p>The type of the incoming values is specified with the first type field. After
4904 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
4905 one pair for each predecessor basic block of the current block. Only values
4906 of <a href="#t_firstclass">first class</a> type may be used as the value
4907 arguments to the PHI node. Only labels may be used as the label
4910 <p>There must be no non-phi instructions between the start of a basic block and
4911 the PHI instructions: i.e. PHI instructions must be first in a basic
4914 <p>For the purposes of the SSA form, the use of each incoming value is deemed to
4915 occur on the edge from the corresponding predecessor block to the current
4916 block (but after any definition of an '<tt>invoke</tt>' instruction's return
4917 value on the same edge).</p>
4920 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4921 specified by the pair corresponding to the predecessor basic block that
4922 executed just prior to the current block.</p>
4926 Loop: ; Infinite loop that counts from 0 on up...
4927 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4928 %nextindvar = add i32 %indvar, 1
4934 <!-- _______________________________________________________________________ -->
4935 <div class="doc_subsubsection">
4936 <a name="i_select">'<tt>select</tt>' Instruction</a>
4939 <div class="doc_text">
4943 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
4945 <i>selty</i> is either i1 or {<N x i1>}
4949 <p>The '<tt>select</tt>' instruction is used to choose one value based on a
4950 condition, without branching.</p>
4954 <p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
4955 values indicating the condition, and two values of the
4956 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
4957 vectors and the condition is a scalar, then entire vectors are selected, not
4958 individual elements.</p>
4961 <p>If the condition is an i1 and it evaluates to 1, the instruction returns the
4962 first value argument; otherwise, it returns the second value argument.</p>
4964 <p>If the condition is a vector of i1, then the value arguments must be vectors
4965 of the same size, and the selection is done element by element.</p>
4969 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4972 <p>Note that the code generator does not yet support conditions
4973 with vector type.</p>
4977 <!-- _______________________________________________________________________ -->
4978 <div class="doc_subsubsection">
4979 <a name="i_call">'<tt>call</tt>' Instruction</a>
4982 <div class="doc_text">
4986 <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>]
4990 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4993 <p>This instruction requires several arguments:</p>
4996 <li>The optional "tail" marker indicates whether the callee function accesses
4997 any allocas or varargs in the caller. If the "tail" marker is present,
4998 the function call is eligible for tail call optimization. Note that calls
4999 may be marked "tail" even if they do not occur before
5000 a <a href="#i_ret"><tt>ret</tt></a> instruction.</li>
5002 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
5003 convention</a> the call should use. If none is specified, the call
5004 defaults to using C calling conventions.</li>
5006 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
5007 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
5008 '<tt>inreg</tt>' attributes are valid here.</li>
5010 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
5011 type of the return value. Functions that return no value are marked
5012 <tt><a href="#t_void">void</a></tt>.</li>
5014 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
5015 being invoked. The argument types must match the types implied by this
5016 signature. This type can be omitted if the function is not varargs and if
5017 the function type does not return a pointer to a function.</li>
5019 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
5020 be invoked. In most cases, this is a direct function invocation, but
5021 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
5022 to function value.</li>
5024 <li>'<tt>function args</tt>': argument list whose types match the function
5025 signature argument types. All arguments must be of
5026 <a href="#t_firstclass">first class</a> type. If the function signature
5027 indicates the function accepts a variable number of arguments, the extra
5028 arguments can be specified.</li>
5030 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
5031 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
5032 '<tt>readnone</tt>' attributes are valid here.</li>
5036 <p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
5037 a specified function, with its incoming arguments bound to the specified
5038 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
5039 function, control flow continues with the instruction after the function
5040 call, and the return value of the function is bound to the result
5045 %retval = call i32 @test(i32 %argc)
5046 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
5047 %X = tail call i32 @foo() <i>; yields i32</i>
5048 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
5049 call void %foo(i8 97 signext)
5051 %struct.A = type { i32, i8 }
5052 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
5053 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
5054 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
5055 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
5056 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
5059 <p>llvm treats calls to some functions with names and arguments that match the
5060 standard C99 library as being the C99 library functions, and may perform
5061 optimizations or generate code for them under that assumption. This is
5062 something we'd like to change in the future to provide better support for
5063 freestanding environments and non-C-based langauges.</p>
5067 <!-- _______________________________________________________________________ -->
5068 <div class="doc_subsubsection">
5069 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
5072 <div class="doc_text">
5076 <resultval> = va_arg <va_list*> <arglist>, <argty>
5080 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
5081 the "variable argument" area of a function call. It is used to implement the
5082 <tt>va_arg</tt> macro in C.</p>
5085 <p>This instruction takes a <tt>va_list*</tt> value and the type of the
5086 argument. It returns a value of the specified argument type and increments
5087 the <tt>va_list</tt> to point to the next argument. The actual type
5088 of <tt>va_list</tt> is target specific.</p>
5091 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
5092 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
5093 to the next argument. For more information, see the variable argument
5094 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
5096 <p>It is legal for this instruction to be called in a function which does not
5097 take a variable number of arguments, for example, the <tt>vfprintf</tt>
5100 <p><tt>va_arg</tt> is an LLVM instruction instead of
5101 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
5105 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
5107 <p>Note that the code generator does not yet fully support va_arg on many
5108 targets. Also, it does not currently support va_arg with aggregate types on
5113 <!-- *********************************************************************** -->
5114 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
5115 <!-- *********************************************************************** -->
5117 <div class="doc_text">
5119 <p>LLVM supports the notion of an "intrinsic function". These functions have
5120 well known names and semantics and are required to follow certain
5121 restrictions. Overall, these intrinsics represent an extension mechanism for
5122 the LLVM language that does not require changing all of the transformations
5123 in LLVM when adding to the language (or the bitcode reader/writer, the
5124 parser, etc...).</p>
5126 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
5127 prefix is reserved in LLVM for intrinsic names; thus, function names may not
5128 begin with this prefix. Intrinsic functions must always be external
5129 functions: you cannot define the body of intrinsic functions. Intrinsic
5130 functions may only be used in call or invoke instructions: it is illegal to
5131 take the address of an intrinsic function. Additionally, because intrinsic
5132 functions are part of the LLVM language, it is required if any are added that
5133 they be documented here.</p>
5135 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
5136 family of functions that perform the same operation but on different data
5137 types. Because LLVM can represent over 8 million different integer types,
5138 overloading is used commonly to allow an intrinsic function to operate on any
5139 integer type. One or more of the argument types or the result type can be
5140 overloaded to accept any integer type. Argument types may also be defined as
5141 exactly matching a previous argument's type or the result type. This allows
5142 an intrinsic function which accepts multiple arguments, but needs all of them
5143 to be of the same type, to only be overloaded with respect to a single
5144 argument or the result.</p>
5146 <p>Overloaded intrinsics will have the names of its overloaded argument types
5147 encoded into its function name, each preceded by a period. Only those types
5148 which are overloaded result in a name suffix. Arguments whose type is matched
5149 against another type do not. For example, the <tt>llvm.ctpop</tt> function
5150 can take an integer of any width and returns an integer of exactly the same
5151 integer width. This leads to a family of functions such as
5152 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
5153 %val)</tt>. Only one type, the return type, is overloaded, and only one type
5154 suffix is required. Because the argument's type is matched against the return
5155 type, it does not require its own name suffix.</p>
5157 <p>To learn how to add an intrinsic function, please see the
5158 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
5162 <!-- ======================================================================= -->
5163 <div class="doc_subsection">
5164 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
5167 <div class="doc_text">
5169 <p>Variable argument support is defined in LLVM with
5170 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
5171 intrinsic functions. These functions are related to the similarly named
5172 macros defined in the <tt><stdarg.h></tt> header file.</p>
5174 <p>All of these functions operate on arguments that use a target-specific value
5175 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
5176 not define what this type is, so all transformations should be prepared to
5177 handle these functions regardless of the type used.</p>
5179 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
5180 instruction and the variable argument handling intrinsic functions are
5183 <div class="doc_code">
5185 define i32 @test(i32 %X, ...) {
5186 ; Initialize variable argument processing
5188 %ap2 = bitcast i8** %ap to i8*
5189 call void @llvm.va_start(i8* %ap2)
5191 ; Read a single integer argument
5192 %tmp = va_arg i8** %ap, i32
5194 ; Demonstrate usage of llvm.va_copy and llvm.va_end
5196 %aq2 = bitcast i8** %aq to i8*
5197 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
5198 call void @llvm.va_end(i8* %aq2)
5200 ; Stop processing of arguments.
5201 call void @llvm.va_end(i8* %ap2)
5205 declare void @llvm.va_start(i8*)
5206 declare void @llvm.va_copy(i8*, i8*)
5207 declare void @llvm.va_end(i8*)
5213 <!-- _______________________________________________________________________ -->
5214 <div class="doc_subsubsection">
5215 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
5219 <div class="doc_text">
5223 declare void %llvm.va_start(i8* <arglist>)
5227 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*<arglist></tt>
5228 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
5231 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
5234 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
5235 macro available in C. In a target-dependent way, it initializes
5236 the <tt>va_list</tt> element to which the argument points, so that the next
5237 call to <tt>va_arg</tt> will produce the first variable argument passed to
5238 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
5239 need to know the last argument of the function as the compiler can figure
5244 <!-- _______________________________________________________________________ -->
5245 <div class="doc_subsubsection">
5246 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
5249 <div class="doc_text">
5253 declare void @llvm.va_end(i8* <arglist>)
5257 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
5258 which has been initialized previously
5259 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
5260 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
5263 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
5266 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
5267 macro available in C. In a target-dependent way, it destroys
5268 the <tt>va_list</tt> element to which the argument points. Calls
5269 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
5270 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
5271 with calls to <tt>llvm.va_end</tt>.</p>
5275 <!-- _______________________________________________________________________ -->
5276 <div class="doc_subsubsection">
5277 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
5280 <div class="doc_text">
5284 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
5288 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
5289 from the source argument list to the destination argument list.</p>
5292 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
5293 The second argument is a pointer to a <tt>va_list</tt> element to copy
5297 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
5298 macro available in C. In a target-dependent way, it copies the
5299 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
5300 element. This intrinsic is necessary because
5301 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
5302 arbitrarily complex and require, for example, memory allocation.</p>
5306 <!-- ======================================================================= -->
5307 <div class="doc_subsection">
5308 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
5311 <div class="doc_text">
5313 <p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
5314 Collection</a> (GC) requires the implementation and generation of these
5315 intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
5316 roots on the stack</a>, as well as garbage collector implementations that
5317 require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
5318 barriers. Front-ends for type-safe garbage collected languages should generate
5319 these intrinsics to make use of the LLVM garbage collectors. For more details,
5320 see <a href="GarbageCollection.html">Accurate Garbage Collection with
5323 <p>The garbage collection intrinsics only operate on objects in the generic
5324 address space (address space zero).</p>
5328 <!-- _______________________________________________________________________ -->
5329 <div class="doc_subsubsection">
5330 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
5333 <div class="doc_text">
5337 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
5341 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
5342 the code generator, and allows some metadata to be associated with it.</p>
5345 <p>The first argument specifies the address of a stack object that contains the
5346 root pointer. The second pointer (which must be either a constant or a
5347 global value address) contains the meta-data to be associated with the
5351 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
5352 location. At compile-time, the code generator generates information to allow
5353 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5354 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5359 <!-- _______________________________________________________________________ -->
5360 <div class="doc_subsubsection">
5361 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
5364 <div class="doc_text">
5368 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
5372 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
5373 locations, allowing garbage collector implementations that require read
5377 <p>The second argument is the address to read from, which should be an address
5378 allocated from the garbage collector. The first object is a pointer to the
5379 start of the referenced object, if needed by the language runtime (otherwise
5383 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
5384 instruction, but may be replaced with substantially more complex code by the
5385 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5386 may only be used in a function which <a href="#gc">specifies a GC
5391 <!-- _______________________________________________________________________ -->
5392 <div class="doc_subsubsection">
5393 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
5396 <div class="doc_text">
5400 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
5404 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
5405 locations, allowing garbage collector implementations that require write
5406 barriers (such as generational or reference counting collectors).</p>
5409 <p>The first argument is the reference to store, the second is the start of the
5410 object to store it to, and the third is the address of the field of Obj to
5411 store to. If the runtime does not require a pointer to the object, Obj may
5415 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
5416 instruction, but may be replaced with substantially more complex code by the
5417 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5418 may only be used in a function which <a href="#gc">specifies a GC
5423 <!-- ======================================================================= -->
5424 <div class="doc_subsection">
5425 <a name="int_codegen">Code Generator Intrinsics</a>
5428 <div class="doc_text">
5430 <p>These intrinsics are provided by LLVM to expose special features that may
5431 only be implemented with code generator support.</p>
5435 <!-- _______________________________________________________________________ -->
5436 <div class="doc_subsubsection">
5437 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
5440 <div class="doc_text">
5444 declare i8 *@llvm.returnaddress(i32 <level>)
5448 <p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5449 target-specific value indicating the return address of the current function
5450 or one of its callers.</p>
5453 <p>The argument to this intrinsic indicates which function to return the address
5454 for. Zero indicates the calling function, one indicates its caller, etc.
5455 The argument is <b>required</b> to be a constant integer value.</p>
5458 <p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
5459 indicating the return address of the specified call frame, or zero if it
5460 cannot be identified. The value returned by this intrinsic is likely to be
5461 incorrect or 0 for arguments other than zero, so it should only be used for
5462 debugging purposes.</p>
5464 <p>Note that calling this intrinsic does not prevent function inlining or other
5465 aggressive transformations, so the value returned may not be that of the
5466 obvious source-language caller.</p>
5470 <!-- _______________________________________________________________________ -->
5471 <div class="doc_subsubsection">
5472 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5475 <div class="doc_text">
5479 declare i8 *@llvm.frameaddress(i32 <level>)
5483 <p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5484 target-specific frame pointer value for the specified stack frame.</p>
5487 <p>The argument to this intrinsic indicates which function to return the frame
5488 pointer for. Zero indicates the calling function, one indicates its caller,
5489 etc. The argument is <b>required</b> to be a constant integer value.</p>
5492 <p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
5493 indicating the frame address of the specified call frame, or zero if it
5494 cannot be identified. The value returned by this intrinsic is likely to be
5495 incorrect or 0 for arguments other than zero, so it should only be used for
5496 debugging purposes.</p>
5498 <p>Note that calling this intrinsic does not prevent function inlining or other
5499 aggressive transformations, so the value returned may not be that of the
5500 obvious source-language caller.</p>
5504 <!-- _______________________________________________________________________ -->
5505 <div class="doc_subsubsection">
5506 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5509 <div class="doc_text">
5513 declare i8 *@llvm.stacksave()
5517 <p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
5518 of the function stack, for use
5519 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
5520 useful for implementing language features like scoped automatic variable
5521 sized arrays in C99.</p>
5524 <p>This intrinsic returns a opaque pointer value that can be passed
5525 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
5526 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
5527 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
5528 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
5529 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
5530 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
5534 <!-- _______________________________________________________________________ -->
5535 <div class="doc_subsubsection">
5536 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5539 <div class="doc_text">
5543 declare void @llvm.stackrestore(i8 * %ptr)
5547 <p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5548 the function stack to the state it was in when the
5549 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
5550 executed. This is useful for implementing language features like scoped
5551 automatic variable sized arrays in C99.</p>
5554 <p>See the description
5555 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
5559 <!-- _______________________________________________________________________ -->
5560 <div class="doc_subsubsection">
5561 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5564 <div class="doc_text">
5568 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
5572 <p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
5573 insert a prefetch instruction if supported; otherwise, it is a noop.
5574 Prefetches have no effect on the behavior of the program but can change its
5575 performance characteristics.</p>
5578 <p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
5579 specifier determining if the fetch should be for a read (0) or write (1),
5580 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5581 locality, to (3) - extremely local keep in cache. The <tt>rw</tt>
5582 and <tt>locality</tt> arguments must be constant integers.</p>
5585 <p>This intrinsic does not modify the behavior of the program. In particular,
5586 prefetches cannot trap and do not produce a value. On targets that support
5587 this intrinsic, the prefetch can provide hints to the processor cache for
5588 better performance.</p>
5592 <!-- _______________________________________________________________________ -->
5593 <div class="doc_subsubsection">
5594 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5597 <div class="doc_text">
5601 declare void @llvm.pcmarker(i32 <id>)
5605 <p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
5606 Counter (PC) in a region of code to simulators and other tools. The method
5607 is target specific, but it is expected that the marker will use exported
5608 symbols to transmit the PC of the marker. The marker makes no guarantees
5609 that it will remain with any specific instruction after optimizations. It is
5610 possible that the presence of a marker will inhibit optimizations. The
5611 intended use is to be inserted after optimizations to allow correlations of
5612 simulation runs.</p>
5615 <p><tt>id</tt> is a numerical id identifying the marker.</p>
5618 <p>This intrinsic does not modify the behavior of the program. Backends that do
5619 not support this intrinisic may ignore it.</p>
5623 <!-- _______________________________________________________________________ -->
5624 <div class="doc_subsubsection">
5625 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5628 <div class="doc_text">
5632 declare i64 @llvm.readcyclecounter( )
5636 <p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5637 counter register (or similar low latency, high accuracy clocks) on those
5638 targets that support it. On X86, it should map to RDTSC. On Alpha, it
5639 should map to RPCC. As the backing counters overflow quickly (on the order
5640 of 9 seconds on alpha), this should only be used for small timings.</p>
5643 <p>When directly supported, reading the cycle counter should not modify any
5644 memory. Implementations are allowed to either return a application specific
5645 value or a system wide value. On backends without support, this is lowered
5646 to a constant 0.</p>
5650 <!-- ======================================================================= -->
5651 <div class="doc_subsection">
5652 <a name="int_libc">Standard C Library Intrinsics</a>
5655 <div class="doc_text">
5657 <p>LLVM provides intrinsics for a few important standard C library functions.
5658 These intrinsics allow source-language front-ends to pass information about
5659 the alignment of the pointer arguments to the code generator, providing
5660 opportunity for more efficient code generation.</p>
5664 <!-- _______________________________________________________________________ -->
5665 <div class="doc_subsubsection">
5666 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5669 <div class="doc_text">
5672 <p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
5673 integer bit width. Not all targets support all bit widths however.</p>
5676 declare void @llvm.memcpy.i8(i8 * <dest>, i8 * <src>,
5677 i8 <len>, i32 <align>)
5678 declare void @llvm.memcpy.i16(i8 * <dest>, i8 * <src>,
5679 i16 <len>, i32 <align>)
5680 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
5681 i32 <len>, i32 <align>)
5682 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
5683 i64 <len>, i32 <align>)
5687 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
5688 source location to the destination location.</p>
5690 <p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5691 intrinsics do not return a value, and takes an extra alignment argument.</p>
5694 <p>The first argument is a pointer to the destination, the second is a pointer
5695 to the source. The third argument is an integer argument specifying the
5696 number of bytes to copy, and the fourth argument is the alignment of the
5697 source and destination locations.</p>
5699 <p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5700 then the caller guarantees that both the source and destination pointers are
5701 aligned to that boundary.</p>
5704 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
5705 source location to the destination location, which are not allowed to
5706 overlap. It copies "len" bytes of memory over. If the argument is known to
5707 be aligned to some boundary, this can be specified as the fourth argument,
5708 otherwise it should be set to 0 or 1.</p>
5712 <!-- _______________________________________________________________________ -->
5713 <div class="doc_subsubsection">
5714 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5717 <div class="doc_text">
5720 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5721 width. Not all targets support all bit widths however.</p>
5724 declare void @llvm.memmove.i8(i8 * <dest>, i8 * <src>,
5725 i8 <len>, i32 <align>)
5726 declare void @llvm.memmove.i16(i8 * <dest>, i8 * <src>,
5727 i16 <len>, i32 <align>)
5728 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
5729 i32 <len>, i32 <align>)
5730 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
5731 i64 <len>, i32 <align>)
5735 <p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
5736 source location to the destination location. It is similar to the
5737 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
5740 <p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5741 intrinsics do not return a value, and takes an extra alignment argument.</p>
5744 <p>The first argument is a pointer to the destination, the second is a pointer
5745 to the source. The third argument is an integer argument specifying the
5746 number of bytes to copy, and the fourth argument is the alignment of the
5747 source and destination locations.</p>
5749 <p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5750 then the caller guarantees that the source and destination pointers are
5751 aligned to that boundary.</p>
5754 <p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
5755 source location to the destination location, which may overlap. It copies
5756 "len" bytes of memory over. If the argument is known to be aligned to some
5757 boundary, this can be specified as the fourth argument, otherwise it should
5758 be set to 0 or 1.</p>
5762 <!-- _______________________________________________________________________ -->
5763 <div class="doc_subsubsection">
5764 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5767 <div class="doc_text">
5770 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5771 width. Not all targets support all bit widths however.</p>
5774 declare void @llvm.memset.i8(i8 * <dest>, i8 <val>,
5775 i8 <len>, i32 <align>)
5776 declare void @llvm.memset.i16(i8 * <dest>, i8 <val>,
5777 i16 <len>, i32 <align>)
5778 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
5779 i32 <len>, i32 <align>)
5780 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
5781 i64 <len>, i32 <align>)
5785 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
5786 particular byte value.</p>
5788 <p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
5789 intrinsic does not return a value, and takes an extra alignment argument.</p>
5792 <p>The first argument is a pointer to the destination to fill, the second is the
5793 byte value to fill it with, the third argument is an integer argument
5794 specifying the number of bytes to fill, and the fourth argument is the known
5795 alignment of destination location.</p>
5797 <p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5798 then the caller guarantees that the destination pointer is aligned to that
5802 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
5803 at the destination location. If the argument is known to be aligned to some
5804 boundary, this can be specified as the fourth argument, otherwise it should
5805 be set to 0 or 1.</p>
5809 <!-- _______________________________________________________________________ -->
5810 <div class="doc_subsubsection">
5811 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5814 <div class="doc_text">
5817 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
5818 floating point or vector of floating point type. Not all targets support all
5822 declare float @llvm.sqrt.f32(float %Val)
5823 declare double @llvm.sqrt.f64(double %Val)
5824 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5825 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5826 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
5830 <p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
5831 returning the same value as the libm '<tt>sqrt</tt>' functions would.
5832 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
5833 behavior for negative numbers other than -0.0 (which allows for better
5834 optimization, because there is no need to worry about errno being
5835 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
5838 <p>The argument and return value are floating point numbers of the same
5842 <p>This function returns the sqrt of the specified operand if it is a
5843 nonnegative floating point number.</p>
5847 <!-- _______________________________________________________________________ -->
5848 <div class="doc_subsubsection">
5849 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5852 <div class="doc_text">
5855 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
5856 floating point or vector of floating point type. Not all targets support all
5860 declare float @llvm.powi.f32(float %Val, i32 %power)
5861 declare double @llvm.powi.f64(double %Val, i32 %power)
5862 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5863 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5864 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
5868 <p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5869 specified (positive or negative) power. The order of evaluation of
5870 multiplications is not defined. When a vector of floating point type is
5871 used, the second argument remains a scalar integer value.</p>
5874 <p>The second argument is an integer power, and the first is a value to raise to
5878 <p>This function returns the first value raised to the second power with an
5879 unspecified sequence of rounding operations.</p>
5883 <!-- _______________________________________________________________________ -->
5884 <div class="doc_subsubsection">
5885 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5888 <div class="doc_text">
5891 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5892 floating point or vector of floating point type. Not all targets support all
5896 declare float @llvm.sin.f32(float %Val)
5897 declare double @llvm.sin.f64(double %Val)
5898 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5899 declare fp128 @llvm.sin.f128(fp128 %Val)
5900 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5904 <p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
5907 <p>The argument and return value are floating point numbers of the same
5911 <p>This function returns the sine of the specified operand, returning the same
5912 values as the libm <tt>sin</tt> functions would, and handles error conditions
5913 in the same way.</p>
5917 <!-- _______________________________________________________________________ -->
5918 <div class="doc_subsubsection">
5919 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5922 <div class="doc_text">
5925 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5926 floating point or vector of floating point type. Not all targets support all
5930 declare float @llvm.cos.f32(float %Val)
5931 declare double @llvm.cos.f64(double %Val)
5932 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5933 declare fp128 @llvm.cos.f128(fp128 %Val)
5934 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5938 <p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
5941 <p>The argument and return value are floating point numbers of the same
5945 <p>This function returns the cosine of the specified operand, returning the same
5946 values as the libm <tt>cos</tt> functions would, and handles error conditions
5947 in the same way.</p>
5951 <!-- _______________________________________________________________________ -->
5952 <div class="doc_subsubsection">
5953 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5956 <div class="doc_text">
5959 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5960 floating point or vector of floating point type. Not all targets support all
5964 declare float @llvm.pow.f32(float %Val, float %Power)
5965 declare double @llvm.pow.f64(double %Val, double %Power)
5966 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5967 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5968 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5972 <p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5973 specified (positive or negative) power.</p>
5976 <p>The second argument is a floating point power, and the first is a value to
5977 raise to that power.</p>
5980 <p>This function returns the first value raised to the second power, returning
5981 the same values as the libm <tt>pow</tt> functions would, and handles error
5982 conditions in the same way.</p>
5986 <!-- ======================================================================= -->
5987 <div class="doc_subsection">
5988 <a name="int_manip">Bit Manipulation Intrinsics</a>
5991 <div class="doc_text">
5993 <p>LLVM provides intrinsics for a few important bit manipulation operations.
5994 These allow efficient code generation for some algorithms.</p>
5998 <!-- _______________________________________________________________________ -->
5999 <div class="doc_subsubsection">
6000 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
6003 <div class="doc_text">
6006 <p>This is an overloaded intrinsic function. You can use bswap on any integer
6007 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
6010 declare i16 @llvm.bswap.i16(i16 <id>)
6011 declare i32 @llvm.bswap.i32(i32 <id>)
6012 declare i64 @llvm.bswap.i64(i64 <id>)
6016 <p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
6017 values with an even number of bytes (positive multiple of 16 bits). These
6018 are useful for performing operations on data that is not in the target's
6019 native byte order.</p>
6022 <p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
6023 and low byte of the input i16 swapped. Similarly,
6024 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
6025 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
6026 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
6027 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
6028 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
6029 more, respectively).</p>
6033 <!-- _______________________________________________________________________ -->
6034 <div class="doc_subsubsection">
6035 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
6038 <div class="doc_text">
6041 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
6042 width. Not all targets support all bit widths however.</p>
6045 declare i8 @llvm.ctpop.i8(i8 <src>)
6046 declare i16 @llvm.ctpop.i16(i16 <src>)
6047 declare i32 @llvm.ctpop.i32(i32 <src>)
6048 declare i64 @llvm.ctpop.i64(i64 <src>)
6049 declare i256 @llvm.ctpop.i256(i256 <src>)
6053 <p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
6057 <p>The only argument is the value to be counted. The argument may be of any
6058 integer type. The return type must match the argument type.</p>
6061 <p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.</p>
6065 <!-- _______________________________________________________________________ -->
6066 <div class="doc_subsubsection">
6067 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
6070 <div class="doc_text">
6073 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
6074 integer bit width. Not all targets support all bit widths however.</p>
6077 declare i8 @llvm.ctlz.i8 (i8 <src>)
6078 declare i16 @llvm.ctlz.i16(i16 <src>)
6079 declare i32 @llvm.ctlz.i32(i32 <src>)
6080 declare i64 @llvm.ctlz.i64(i64 <src>)
6081 declare i256 @llvm.ctlz.i256(i256 <src>)
6085 <p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
6086 leading zeros in a variable.</p>
6089 <p>The only argument is the value to be counted. The argument may be of any
6090 integer type. The return type must match the argument type.</p>
6093 <p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
6094 zeros in a variable. If the src == 0 then the result is the size in bits of
6095 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
6099 <!-- _______________________________________________________________________ -->
6100 <div class="doc_subsubsection">
6101 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
6104 <div class="doc_text">
6107 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
6108 integer bit width. Not all targets support all bit widths however.</p>
6111 declare i8 @llvm.cttz.i8 (i8 <src>)
6112 declare i16 @llvm.cttz.i16(i16 <src>)
6113 declare i32 @llvm.cttz.i32(i32 <src>)
6114 declare i64 @llvm.cttz.i64(i64 <src>)
6115 declare i256 @llvm.cttz.i256(i256 <src>)
6119 <p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
6123 <p>The only argument is the value to be counted. The argument may be of any
6124 integer type. The return type must match the argument type.</p>
6127 <p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
6128 zeros in a variable. If the src == 0 then the result is the size in bits of
6129 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
6133 <!-- ======================================================================= -->
6134 <div class="doc_subsection">
6135 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
6138 <div class="doc_text">
6140 <p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
6144 <!-- _______________________________________________________________________ -->
6145 <div class="doc_subsubsection">
6146 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
6149 <div class="doc_text">
6152 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
6153 on any integer bit width.</p>
6156 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
6157 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6158 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6162 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6163 a signed addition of the two arguments, and indicate whether an overflow
6164 occurred during the signed summation.</p>
6167 <p>The arguments (%a and %b) and the first element of the result structure may
6168 be of integer types of any bit width, but they must have the same bit
6169 width. The second element of the result structure must be of
6170 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6171 undergo signed addition.</p>
6174 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6175 a signed addition of the two variables. They return a structure — the
6176 first element of which is the signed summation, and the second element of
6177 which is a bit specifying if the signed summation resulted in an
6182 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6183 %sum = extractvalue {i32, i1} %res, 0
6184 %obit = extractvalue {i32, i1} %res, 1
6185 br i1 %obit, label %overflow, label %normal
6190 <!-- _______________________________________________________________________ -->
6191 <div class="doc_subsubsection">
6192 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
6195 <div class="doc_text">
6198 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
6199 on any integer bit width.</p>
6202 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6203 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6204 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6208 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6209 an unsigned addition of the two arguments, and indicate whether a carry
6210 occurred during the unsigned summation.</p>
6213 <p>The arguments (%a and %b) and the first element of the result structure may
6214 be of integer types of any bit width, but they must have the same bit
6215 width. The second element of the result structure must be of
6216 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6217 undergo unsigned addition.</p>
6220 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6221 an unsigned addition of the two arguments. They return a structure —
6222 the first element of which is the sum, and the second element of which is a
6223 bit specifying if the unsigned summation resulted in a carry.</p>
6227 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6228 %sum = extractvalue {i32, i1} %res, 0
6229 %obit = extractvalue {i32, i1} %res, 1
6230 br i1 %obit, label %carry, label %normal
6235 <!-- _______________________________________________________________________ -->
6236 <div class="doc_subsubsection">
6237 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
6240 <div class="doc_text">
6243 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
6244 on any integer bit width.</p>
6247 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6248 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6249 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6253 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6254 a signed subtraction of the two arguments, and indicate whether an overflow
6255 occurred during the signed subtraction.</p>
6258 <p>The arguments (%a and %b) and the first element of the result structure may
6259 be of integer types of any bit width, but they must have the same bit
6260 width. The second element of the result structure must be of
6261 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6262 undergo signed subtraction.</p>
6265 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6266 a signed subtraction of the two arguments. They return a structure —
6267 the first element of which is the subtraction, and the second element of
6268 which is a bit specifying if the signed subtraction resulted in an
6273 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6274 %sum = extractvalue {i32, i1} %res, 0
6275 %obit = extractvalue {i32, i1} %res, 1
6276 br i1 %obit, label %overflow, label %normal
6281 <!-- _______________________________________________________________________ -->
6282 <div class="doc_subsubsection">
6283 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
6286 <div class="doc_text">
6289 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
6290 on any integer bit width.</p>
6293 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6294 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6295 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6299 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6300 an unsigned subtraction of the two arguments, and indicate whether an
6301 overflow occurred during the unsigned subtraction.</p>
6304 <p>The arguments (%a and %b) and the first element of the result structure may
6305 be of integer types of any bit width, but they must have the same bit
6306 width. The second element of the result structure must be of
6307 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6308 undergo unsigned subtraction.</p>
6311 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6312 an unsigned subtraction of the two arguments. They return a structure —
6313 the first element of which is the subtraction, and the second element of
6314 which is a bit specifying if the unsigned subtraction resulted in an
6319 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6320 %sum = extractvalue {i32, i1} %res, 0
6321 %obit = extractvalue {i32, i1} %res, 1
6322 br i1 %obit, label %overflow, label %normal
6327 <!-- _______________________________________________________________________ -->
6328 <div class="doc_subsubsection">
6329 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
6332 <div class="doc_text">
6335 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
6336 on any integer bit width.</p>
6339 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6340 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6341 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6346 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6347 a signed multiplication of the two arguments, and indicate whether an
6348 overflow occurred during the signed multiplication.</p>
6351 <p>The arguments (%a and %b) and the first element of the result structure may
6352 be of integer types of any bit width, but they must have the same bit
6353 width. The second element of the result structure must be of
6354 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6355 undergo signed multiplication.</p>
6358 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6359 a signed multiplication of the two arguments. They return a structure —
6360 the first element of which is the multiplication, and the second element of
6361 which is a bit specifying if the signed multiplication resulted in an
6366 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6367 %sum = extractvalue {i32, i1} %res, 0
6368 %obit = extractvalue {i32, i1} %res, 1
6369 br i1 %obit, label %overflow, label %normal
6374 <!-- _______________________________________________________________________ -->
6375 <div class="doc_subsubsection">
6376 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6379 <div class="doc_text">
6382 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6383 on any integer bit width.</p>
6386 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6387 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6388 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6392 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6393 a unsigned multiplication of the two arguments, and indicate whether an
6394 overflow occurred during the unsigned multiplication.</p>
6397 <p>The arguments (%a and %b) and the first element of the result structure may
6398 be of integer types of any bit width, but they must have the same bit
6399 width. The second element of the result structure must be of
6400 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6401 undergo unsigned multiplication.</p>
6404 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6405 an unsigned multiplication of the two arguments. They return a structure
6406 — the first element of which is the multiplication, and the second
6407 element of which is a bit specifying if the unsigned multiplication resulted
6412 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6413 %sum = extractvalue {i32, i1} %res, 0
6414 %obit = extractvalue {i32, i1} %res, 1
6415 br i1 %obit, label %overflow, label %normal
6420 <!-- ======================================================================= -->
6421 <div class="doc_subsection">
6422 <a name="int_debugger">Debugger Intrinsics</a>
6425 <div class="doc_text">
6427 <p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
6428 prefix), are described in
6429 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
6430 Level Debugging</a> document.</p>
6434 <!-- ======================================================================= -->
6435 <div class="doc_subsection">
6436 <a name="int_eh">Exception Handling Intrinsics</a>
6439 <div class="doc_text">
6441 <p>The LLVM exception handling intrinsics (which all start with
6442 <tt>llvm.eh.</tt> prefix), are described in
6443 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6444 Handling</a> document.</p>
6448 <!-- ======================================================================= -->
6449 <div class="doc_subsection">
6450 <a name="int_trampoline">Trampoline Intrinsic</a>
6453 <div class="doc_text">
6455 <p>This intrinsic makes it possible to excise one parameter, marked with
6456 the <tt>nest</tt> attribute, from a function. The result is a callable
6457 function pointer lacking the nest parameter - the caller does not need to
6458 provide a value for it. Instead, the value to use is stored in advance in a
6459 "trampoline", a block of memory usually allocated on the stack, which also
6460 contains code to splice the nest value into the argument list. This is used
6461 to implement the GCC nested function address extension.</p>
6463 <p>For example, if the function is
6464 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
6465 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
6468 <div class="doc_code">
6470 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6471 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6472 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
6473 %fp = bitcast i8* %p to i32 (i32, i32)*
6477 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
6478 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
6482 <!-- _______________________________________________________________________ -->
6483 <div class="doc_subsubsection">
6484 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6487 <div class="doc_text">
6491 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
6495 <p>This fills the memory pointed to by <tt>tramp</tt> with code and returns a
6496 function pointer suitable for executing it.</p>
6499 <p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6500 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
6501 sufficiently aligned block of memory; this memory is written to by the
6502 intrinsic. Note that the size and the alignment are target-specific - LLVM
6503 currently provides no portable way of determining them, so a front-end that
6504 generates this intrinsic needs to have some target-specific knowledge.
6505 The <tt>func</tt> argument must hold a function bitcast to
6506 an <tt>i8*</tt>.</p>
6509 <p>The block of memory pointed to by <tt>tramp</tt> is filled with target
6510 dependent code, turning it into a function. A pointer to this function is
6511 returned, but needs to be bitcast to an <a href="#int_trampoline">appropriate
6512 function pointer type</a> before being called. The new function's signature
6513 is the same as that of <tt>func</tt> with any arguments marked with
6514 the <tt>nest</tt> attribute removed. At most one such <tt>nest</tt> argument
6515 is allowed, and it must be of pointer type. Calling the new function is
6516 equivalent to calling <tt>func</tt> with the same argument list, but
6517 with <tt>nval</tt> used for the missing <tt>nest</tt> argument. If, after
6518 calling <tt>llvm.init.trampoline</tt>, the memory pointed to
6519 by <tt>tramp</tt> is modified, then the effect of any later call to the
6520 returned function pointer is undefined.</p>
6524 <!-- ======================================================================= -->
6525 <div class="doc_subsection">
6526 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6529 <div class="doc_text">
6531 <p>These intrinsic functions expand the "universal IR" of LLVM to represent
6532 hardware constructs for atomic operations and memory synchronization. This
6533 provides an interface to the hardware, not an interface to the programmer. It
6534 is aimed at a low enough level to allow any programming models or APIs
6535 (Application Programming Interfaces) which need atomic behaviors to map
6536 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
6537 hardware provides a "universal IR" for source languages, it also provides a
6538 starting point for developing a "universal" atomic operation and
6539 synchronization IR.</p>
6541 <p>These do <em>not</em> form an API such as high-level threading libraries,
6542 software transaction memory systems, atomic primitives, and intrinsic
6543 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6544 application libraries. The hardware interface provided by LLVM should allow
6545 a clean implementation of all of these APIs and parallel programming models.
6546 No one model or paradigm should be selected above others unless the hardware
6547 itself ubiquitously does so.</p>
6551 <!-- _______________________________________________________________________ -->
6552 <div class="doc_subsubsection">
6553 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6555 <div class="doc_text">
6558 declare void @llvm.memory.barrier( i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>, i1 <device> )
6562 <p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6563 specific pairs of memory access types.</p>
6566 <p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6567 The first four arguments enables a specific barrier as listed below. The
6568 fith argument specifies that the barrier applies to io or device or uncached
6572 <li><tt>ll</tt>: load-load barrier</li>
6573 <li><tt>ls</tt>: load-store barrier</li>
6574 <li><tt>sl</tt>: store-load barrier</li>
6575 <li><tt>ss</tt>: store-store barrier</li>
6576 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
6580 <p>This intrinsic causes the system to enforce some ordering constraints upon
6581 the loads and stores of the program. This barrier does not
6582 indicate <em>when</em> any events will occur, it only enforces
6583 an <em>order</em> in which they occur. For any of the specified pairs of load
6584 and store operations (f.ex. load-load, or store-load), all of the first
6585 operations preceding the barrier will complete before any of the second
6586 operations succeeding the barrier begin. Specifically the semantics for each
6587 pairing is as follows:</p>
6590 <li><tt>ll</tt>: All loads before the barrier must complete before any load
6591 after the barrier begins.</li>
6592 <li><tt>ls</tt>: All loads before the barrier must complete before any
6593 store after the barrier begins.</li>
6594 <li><tt>ss</tt>: All stores before the barrier must complete before any
6595 store after the barrier begins.</li>
6596 <li><tt>sl</tt>: All stores before the barrier must complete before any
6597 load after the barrier begins.</li>
6600 <p>These semantics are applied with a logical "and" behavior when more than one
6601 is enabled in a single memory barrier intrinsic.</p>
6603 <p>Backends may implement stronger barriers than those requested when they do
6604 not support as fine grained a barrier as requested. Some architectures do
6605 not need all types of barriers and on such architectures, these become
6610 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6611 %ptr = bitcast i8* %mallocP to i32*
6614 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
6615 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
6616 <i>; guarantee the above finishes</i>
6617 store i32 8, %ptr <i>; before this begins</i>
6622 <!-- _______________________________________________________________________ -->
6623 <div class="doc_subsubsection">
6624 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
6627 <div class="doc_text">
6630 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
6631 any integer bit width and for different address spaces. Not all targets
6632 support all bit widths however.</p>
6635 declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* <ptr>, i8 <cmp>, i8 <val> )
6636 declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* <ptr>, i16 <cmp>, i16 <val> )
6637 declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* <ptr>, i32 <cmp>, i32 <val> )
6638 declare i64 @llvm.atomic.cmp.swap.i64.p0i64( i64* <ptr>, i64 <cmp>, i64 <val> )
6642 <p>This loads a value in memory and compares it to a given value. If they are
6643 equal, it stores a new value into the memory.</p>
6646 <p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
6647 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6648 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6649 this integer type. While any bit width integer may be used, targets may only
6650 lower representations they support in hardware.</p>
6653 <p>This entire intrinsic must be executed atomically. It first loads the value
6654 in memory pointed to by <tt>ptr</tt> and compares it with the
6655 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
6656 memory. The loaded value is yielded in all cases. This provides the
6657 equivalent of an atomic compare-and-swap operation within the SSA
6662 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6663 %ptr = bitcast i8* %mallocP to i32*
6666 %val1 = add i32 4, 4
6667 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
6668 <i>; yields {i32}:result1 = 4</i>
6669 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6670 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6672 %val2 = add i32 1, 1
6673 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
6674 <i>; yields {i32}:result2 = 8</i>
6675 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6677 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6682 <!-- _______________________________________________________________________ -->
6683 <div class="doc_subsubsection">
6684 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6686 <div class="doc_text">
6689 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6690 integer bit width. Not all targets support all bit widths however.</p>
6693 declare i8 @llvm.atomic.swap.i8.p0i8( i8* <ptr>, i8 <val> )
6694 declare i16 @llvm.atomic.swap.i16.p0i16( i16* <ptr>, i16 <val> )
6695 declare i32 @llvm.atomic.swap.i32.p0i32( i32* <ptr>, i32 <val> )
6696 declare i64 @llvm.atomic.swap.i64.p0i64( i64* <ptr>, i64 <val> )
6700 <p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6701 the value from memory. It then stores the value in <tt>val</tt> in the memory
6702 at <tt>ptr</tt>.</p>
6705 <p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
6706 the <tt>val</tt> argument and the result must be integers of the same bit
6707 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6708 integer type. The targets may only lower integer representations they
6712 <p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6713 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6714 equivalent of an atomic swap operation within the SSA framework.</p>
6718 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6719 %ptr = bitcast i8* %mallocP to i32*
6722 %val1 = add i32 4, 4
6723 %result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
6724 <i>; yields {i32}:result1 = 4</i>
6725 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6726 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6728 %val2 = add i32 1, 1
6729 %result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
6730 <i>; yields {i32}:result2 = 8</i>
6732 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6733 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6738 <!-- _______________________________________________________________________ -->
6739 <div class="doc_subsubsection">
6740 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
6744 <div class="doc_text">
6747 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
6748 any integer bit width. Not all targets support all bit widths however.</p>
6751 declare i8 @llvm.atomic.load.add.i8..p0i8( i8* <ptr>, i8 <delta> )
6752 declare i16 @llvm.atomic.load.add.i16..p0i16( i16* <ptr>, i16 <delta> )
6753 declare i32 @llvm.atomic.load.add.i32..p0i32( i32* <ptr>, i32 <delta> )
6754 declare i64 @llvm.atomic.load.add.i64..p0i64( i64* <ptr>, i64 <delta> )
6758 <p>This intrinsic adds <tt>delta</tt> to the value stored in memory
6759 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
6762 <p>The intrinsic takes two arguments, the first a pointer to an integer value
6763 and the second an integer value. The result is also an integer value. These
6764 integer types can have any bit width, but they must all have the same bit
6765 width. The targets may only lower integer representations they support.</p>
6768 <p>This intrinsic does a series of operations atomically. It first loads the
6769 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6770 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
6774 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6775 %ptr = bitcast i8* %mallocP to i32*
6777 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
6778 <i>; yields {i32}:result1 = 4</i>
6779 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
6780 <i>; yields {i32}:result2 = 8</i>
6781 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
6782 <i>; yields {i32}:result3 = 10</i>
6783 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
6788 <!-- _______________________________________________________________________ -->
6789 <div class="doc_subsubsection">
6790 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6794 <div class="doc_text">
6797 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
6798 any integer bit width and for different address spaces. Not all targets
6799 support all bit widths however.</p>
6802 declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* <ptr>, i8 <delta> )
6803 declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* <ptr>, i16 <delta> )
6804 declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* <ptr>, i32 <delta> )
6805 declare i64 @llvm.atomic.load.sub.i64.p0i32( i64* <ptr>, i64 <delta> )
6809 <p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6810 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
6813 <p>The intrinsic takes two arguments, the first a pointer to an integer value
6814 and the second an integer value. The result is also an integer value. These
6815 integer types can have any bit width, but they must all have the same bit
6816 width. The targets may only lower integer representations they support.</p>
6819 <p>This intrinsic does a series of operations atomically. It first loads the
6820 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6821 result to <tt>ptr</tt>. It yields the original value stored
6822 at <tt>ptr</tt>.</p>
6826 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6827 %ptr = bitcast i8* %mallocP to i32*
6829 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
6830 <i>; yields {i32}:result1 = 8</i>
6831 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
6832 <i>; yields {i32}:result2 = 4</i>
6833 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
6834 <i>; yields {i32}:result3 = 2</i>
6835 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6840 <!-- _______________________________________________________________________ -->
6841 <div class="doc_subsubsection">
6842 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6843 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6844 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6845 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6848 <div class="doc_text">
6851 <p>These are overloaded intrinsics. You can
6852 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
6853 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
6854 bit width and for different address spaces. Not all targets support all bit
6858 declare i8 @llvm.atomic.load.and.i8.p0i8( i8* <ptr>, i8 <delta> )
6859 declare i16 @llvm.atomic.load.and.i16.p0i16( i16* <ptr>, i16 <delta> )
6860 declare i32 @llvm.atomic.load.and.i32.p0i32( i32* <ptr>, i32 <delta> )
6861 declare i64 @llvm.atomic.load.and.i64.p0i64( i64* <ptr>, i64 <delta> )
6865 declare i8 @llvm.atomic.load.or.i8.p0i8( i8* <ptr>, i8 <delta> )
6866 declare i16 @llvm.atomic.load.or.i16.p0i16( i16* <ptr>, i16 <delta> )
6867 declare i32 @llvm.atomic.load.or.i32.p0i32( i32* <ptr>, i32 <delta> )
6868 declare i64 @llvm.atomic.load.or.i64.p0i64( i64* <ptr>, i64 <delta> )
6872 declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* <ptr>, i8 <delta> )
6873 declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* <ptr>, i16 <delta> )
6874 declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* <ptr>, i32 <delta> )
6875 declare i64 @llvm.atomic.load.nand.i64.p0i32( i64* <ptr>, i64 <delta> )
6879 declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* <ptr>, i8 <delta> )
6880 declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* <ptr>, i16 <delta> )
6881 declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* <ptr>, i32 <delta> )
6882 declare i64 @llvm.atomic.load.xor.i64.p0i32( i64* <ptr>, i64 <delta> )
6886 <p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6887 the value stored in memory at <tt>ptr</tt>. It yields the original value
6888 at <tt>ptr</tt>.</p>
6891 <p>These intrinsics take two arguments, the first a pointer to an integer value
6892 and the second an integer value. The result is also an integer value. These
6893 integer types can have any bit width, but they must all have the same bit
6894 width. The targets may only lower integer representations they support.</p>
6897 <p>These intrinsics does a series of operations atomically. They first load the
6898 value stored at <tt>ptr</tt>. They then do the bitwise
6899 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
6900 original value stored at <tt>ptr</tt>.</p>
6904 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6905 %ptr = bitcast i8* %mallocP to i32*
6906 store i32 0x0F0F, %ptr
6907 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
6908 <i>; yields {i32}:result0 = 0x0F0F</i>
6909 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
6910 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
6911 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
6912 <i>; yields {i32}:result2 = 0xF0</i>
6913 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
6914 <i>; yields {i32}:result3 = FF</i>
6915 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6920 <!-- _______________________________________________________________________ -->
6921 <div class="doc_subsubsection">
6922 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6923 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6924 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6925 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6928 <div class="doc_text">
6931 <p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6932 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
6933 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6934 address spaces. Not all targets support all bit widths however.</p>
6937 declare i8 @llvm.atomic.load.max.i8.p0i8( i8* <ptr>, i8 <delta> )
6938 declare i16 @llvm.atomic.load.max.i16.p0i16( i16* <ptr>, i16 <delta> )
6939 declare i32 @llvm.atomic.load.max.i32.p0i32( i32* <ptr>, i32 <delta> )
6940 declare i64 @llvm.atomic.load.max.i64.p0i64( i64* <ptr>, i64 <delta> )
6944 declare i8 @llvm.atomic.load.min.i8.p0i8( i8* <ptr>, i8 <delta> )
6945 declare i16 @llvm.atomic.load.min.i16.p0i16( i16* <ptr>, i16 <delta> )
6946 declare i32 @llvm.atomic.load.min.i32..p0i32( i32* <ptr>, i32 <delta> )
6947 declare i64 @llvm.atomic.load.min.i64..p0i64( i64* <ptr>, i64 <delta> )
6951 declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* <ptr>, i8 <delta> )
6952 declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* <ptr>, i16 <delta> )
6953 declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* <ptr>, i32 <delta> )
6954 declare i64 @llvm.atomic.load.umax.i64.p0i64( i64* <ptr>, i64 <delta> )
6958 declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* <ptr>, i8 <delta> )
6959 declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* <ptr>, i16 <delta> )
6960 declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* <ptr>, i32 <delta> )
6961 declare i64 @llvm.atomic.load.umin.i64..p0i64( i64* <ptr>, i64 <delta> )
6965 <p>These intrinsics takes the signed or unsigned minimum or maximum of
6966 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6967 original value at <tt>ptr</tt>.</p>
6970 <p>These intrinsics take two arguments, the first a pointer to an integer value
6971 and the second an integer value. The result is also an integer value. These
6972 integer types can have any bit width, but they must all have the same bit
6973 width. The targets may only lower integer representations they support.</p>
6976 <p>These intrinsics does a series of operations atomically. They first load the
6977 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
6978 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
6979 yield the original value stored at <tt>ptr</tt>.</p>
6983 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6984 %ptr = bitcast i8* %mallocP to i32*
6986 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
6987 <i>; yields {i32}:result0 = 7</i>
6988 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
6989 <i>; yields {i32}:result1 = -2</i>
6990 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
6991 <i>; yields {i32}:result2 = 8</i>
6992 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
6993 <i>; yields {i32}:result3 = 8</i>
6994 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
7000 <!-- ======================================================================= -->
7001 <div class="doc_subsection">
7002 <a name="int_memorymarkers">Memory Use Markers</a>
7005 <div class="doc_text">
7007 <p>This class of intrinsics exists to information about the lifetime of memory
7008 objects and ranges where variables are immutable.</p>
7012 <!-- _______________________________________________________________________ -->
7013 <div class="doc_subsubsection">
7014 <a name="int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a>
7017 <div class="doc_text">
7021 declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
7025 <p>The '<tt>llvm.lifetime.start</tt>' intrinsic specifies the start of a memory
7026 object's lifetime.</p>
7029 <p>The first argument is a constant integer representing the size of the
7030 object, or -1 if it is variable sized. The second argument is a pointer to
7034 <p>This intrinsic indicates that before this point in the code, the value of the
7035 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7036 never be used and has an undefined value. A load from the pointer that
7037 precedes this intrinsic can be replaced with
7038 <tt>'<a href="#undefvalues">undef</a>'</tt>.</p>
7042 <!-- _______________________________________________________________________ -->
7043 <div class="doc_subsubsection">
7044 <a name="int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a>
7047 <div class="doc_text">
7051 declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
7055 <p>The '<tt>llvm.lifetime.end</tt>' intrinsic specifies the end of a memory
7056 object's lifetime.</p>
7059 <p>The first argument is a constant integer representing the size of the
7060 object, or -1 if it is variable sized. The second argument is a pointer to
7064 <p>This intrinsic indicates that after this point in the code, the value of the
7065 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7066 never be used and has an undefined value. Any stores into the memory object
7067 following this intrinsic may be removed as dead.
7071 <!-- _______________________________________________________________________ -->
7072 <div class="doc_subsubsection">
7073 <a name="int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a>
7076 <div class="doc_text">
7080 declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>) readonly
7084 <p>The '<tt>llvm.invariant.start</tt>' intrinsic specifies that the contents of
7085 a memory object will not change.</p>
7088 <p>The first argument is a constant integer representing the size of the
7089 object, or -1 if it is variable sized. The second argument is a pointer to
7093 <p>This intrinsic indicates that until an <tt>llvm.invariant.end</tt> that uses
7094 the return value, the referenced memory location is constant and
7099 <!-- _______________________________________________________________________ -->
7100 <div class="doc_subsubsection">
7101 <a name="int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a>
7104 <div class="doc_text">
7108 declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
7112 <p>The '<tt>llvm.invariant.end</tt>' intrinsic specifies that the contents of
7113 a memory object are mutable.</p>
7116 <p>The first argument is the matching <tt>llvm.invariant.start</tt> intrinsic.
7117 The second argument is a constant integer representing the size of the
7118 object, or -1 if it is variable sized and the third argument is a pointer
7122 <p>This intrinsic indicates that the memory is mutable again.</p>
7126 <!-- ======================================================================= -->
7127 <div class="doc_subsection">
7128 <a name="int_general">General Intrinsics</a>
7131 <div class="doc_text">
7133 <p>This class of intrinsics is designed to be generic and has no specific
7138 <!-- _______________________________________________________________________ -->
7139 <div class="doc_subsubsection">
7140 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
7143 <div class="doc_text">
7147 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
7151 <p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
7154 <p>The first argument is a pointer to a value, the second is a pointer to a
7155 global string, the third is a pointer to a global string which is the source
7156 file name, and the last argument is the line number.</p>
7159 <p>This intrinsic allows annotation of local variables with arbitrary strings.
7160 This can be useful for special purpose optimizations that want to look for
7161 these annotations. These have no other defined use, they are ignored by code
7162 generation and optimization.</p>
7166 <!-- _______________________________________________________________________ -->
7167 <div class="doc_subsubsection">
7168 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
7171 <div class="doc_text">
7174 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
7175 any integer bit width.</p>
7178 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
7179 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
7180 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
7181 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
7182 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
7186 <p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
7189 <p>The first argument is an integer value (result of some expression), the
7190 second is a pointer to a global string, the third is a pointer to a global
7191 string which is the source file name, and the last argument is the line
7192 number. It returns the value of the first argument.</p>
7195 <p>This intrinsic allows annotations to be put on arbitrary expressions with
7196 arbitrary strings. This can be useful for special purpose optimizations that
7197 want to look for these annotations. These have no other defined use, they
7198 are ignored by code generation and optimization.</p>
7202 <!-- _______________________________________________________________________ -->
7203 <div class="doc_subsubsection">
7204 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
7207 <div class="doc_text">
7211 declare void @llvm.trap()
7215 <p>The '<tt>llvm.trap</tt>' intrinsic.</p>
7221 <p>This intrinsics is lowered to the target dependent trap instruction. If the
7222 target does not have a trap instruction, this intrinsic will be lowered to
7223 the call of the <tt>abort()</tt> function.</p>
7227 <!-- _______________________________________________________________________ -->
7228 <div class="doc_subsubsection">
7229 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
7232 <div class="doc_text">
7236 declare void @llvm.stackprotector( i8* <guard>, i8** <slot> )
7240 <p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
7241 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
7242 ensure that it is placed on the stack before local variables.</p>
7245 <p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
7246 arguments. The first argument is the value loaded from the stack
7247 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
7248 that has enough space to hold the value of the guard.</p>
7251 <p>This intrinsic causes the prologue/epilogue inserter to force the position of
7252 the <tt>AllocaInst</tt> stack slot to be before local variables on the
7253 stack. This is to ensure that if a local variable on the stack is
7254 overwritten, it will destroy the value of the guard. When the function exits,
7255 the guard on the stack is checked against the original guard. If they're
7256 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
7261 <!-- _______________________________________________________________________ -->
7262 <div class="doc_subsubsection">
7263 <a name="int_objectsize">'<tt>llvm.objectsize</tt>' Intrinsic</a>
7266 <div class="doc_text">
7270 declare i32 @llvm.objectsize.i32( i8* <object>, i1 <type> )
7271 declare i64 @llvm.objectsize.i64( i8* <object>, i1 <type> )
7275 <p>The <tt>llvm.objectsize</tt> intrinsic is designed to provide information
7276 to the optimizers to discover at compile time either a) when an
7277 operation like memcpy will either overflow a buffer that corresponds to
7278 an object, or b) to determine that a runtime check for overflow isn't
7279 necessary. An object in this context means an allocation of a
7280 specific class, structure, array, or other object.</p>
7283 <p>The <tt>llvm.objectsize</tt> intrinsic takes two arguments. The first
7284 argument is a pointer to or into the <tt>object</tt>. The second argument
7285 is a boolean 0 or 1. This argument determines whether you want the
7286 maximum (0) or minimum (1) bytes remaining. This needs to be a literal 0 or
7287 1, variables are not allowed.</p>
7290 <p>The <tt>llvm.objectsize</tt> intrinsic is lowered to either a constant
7291 representing the size of the object concerned or <tt>i32/i64 -1 or 0</tt>
7292 (depending on the <tt>type</tt> argument if the size cannot be determined
7293 at compile time.</p>
7297 <!-- *********************************************************************** -->
7300 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
7301 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
7302 <a href="http://validator.w3.org/check/referer"><img
7303 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
7305 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
7306 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
7307 Last modified: $Date$