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="#constantexprs">Constant Expressions</a></li>
87 <li><a href="#metadata">Embedded Metadata</a></li>
90 <li><a href="#othervalues">Other Values</a>
92 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
95 <li><a href="#intrinsic_globals">Intrinsic Global Variables</a>
97 <li><a href="#intg_used">The '<tt>llvm.used</tt>' Global Variable</a></li>
98 <li><a href="#intg_compiler_used">The '<tt>llvm.compiler.used</tt>'
99 Global Variable</a></li>
100 <li><a href="#intg_global_ctors">The '<tt>llvm.global_ctors</tt>'
101 Global Variable</a></li>
102 <li><a href="#intg_global_dtors">The '<tt>llvm.global_dtors</tt>'
103 Global Variable</a></li>
106 <li><a href="#instref">Instruction Reference</a>
108 <li><a href="#terminators">Terminator Instructions</a>
110 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
111 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
112 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
113 <li><a href="#i_indbr">'<tt>indbr</tt>' Instruction</a></li>
114 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
115 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
116 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
119 <li><a href="#binaryops">Binary Operations</a>
121 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
122 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
123 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
124 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
125 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
126 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
127 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
128 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
129 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
130 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
131 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
132 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
135 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
137 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
138 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
139 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
140 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
141 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
142 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
145 <li><a href="#vectorops">Vector Operations</a>
147 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
148 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
149 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
152 <li><a href="#aggregateops">Aggregate Operations</a>
154 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
155 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
158 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
160 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
161 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
162 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
163 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
166 <li><a href="#convertops">Conversion Operations</a>
168 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
169 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
170 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
171 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
172 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
173 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
174 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
175 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
176 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
177 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
178 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
179 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
182 <li><a href="#otherops">Other Operations</a>
184 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
185 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
186 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
187 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
188 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
189 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
194 <li><a href="#intrinsics">Intrinsic Functions</a>
196 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
198 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
199 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
200 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
203 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
205 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
206 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
207 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
210 <li><a href="#int_codegen">Code Generator Intrinsics</a>
212 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
213 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
214 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
215 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
216 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
217 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
218 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
221 <li><a href="#int_libc">Standard C Library Intrinsics</a>
223 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
224 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
225 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
226 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
227 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
228 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
229 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
230 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
233 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
235 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
236 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
237 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
238 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
241 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
243 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
244 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
245 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
246 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
247 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
248 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
251 <li><a href="#int_debugger">Debugger intrinsics</a></li>
252 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
253 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
255 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
258 <li><a href="#int_atomics">Atomic intrinsics</a>
260 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
261 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
262 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
263 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
264 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
265 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
266 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
267 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
268 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
269 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
270 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
271 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
272 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
275 <li><a href="#int_memorymarkers">Memory Use Markers</a>
277 <li><a href="#int_lifetime_start"><tt>llvm.lifetime.start</tt></a></li>
278 <li><a href="#int_lifetime_end"><tt>llvm.lifetime.end</tt></a></li>
279 <li><a href="#int_invariant_start"><tt>llvm.invariant.start</tt></a></li>
280 <li><a href="#int_invariant_end"><tt>llvm.invariant.end</tt></a></li>
283 <li><a href="#int_general">General intrinsics</a>
285 <li><a href="#int_var_annotation">
286 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
287 <li><a href="#int_annotation">
288 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
289 <li><a href="#int_trap">
290 '<tt>llvm.trap</tt>' Intrinsic</a></li>
291 <li><a href="#int_stackprotector">
292 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
299 <div class="doc_author">
300 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
301 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
304 <!-- *********************************************************************** -->
305 <div class="doc_section"> <a name="abstract">Abstract </a></div>
306 <!-- *********************************************************************** -->
308 <div class="doc_text">
310 <p>This document is a reference manual for the LLVM assembly language. LLVM is
311 a Static Single Assignment (SSA) based representation that provides type
312 safety, low-level operations, flexibility, and the capability of representing
313 'all' high-level languages cleanly. It is the common code representation
314 used throughout all phases of the LLVM compilation strategy.</p>
318 <!-- *********************************************************************** -->
319 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
320 <!-- *********************************************************************** -->
322 <div class="doc_text">
324 <p>The LLVM code representation is designed to be used in three different forms:
325 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
326 for fast loading by a Just-In-Time compiler), and as a human readable
327 assembly language representation. This allows LLVM to provide a powerful
328 intermediate representation for efficient compiler transformations and
329 analysis, while providing a natural means to debug and visualize the
330 transformations. The three different forms of LLVM are all equivalent. This
331 document describes the human readable representation and notation.</p>
333 <p>The LLVM representation aims to be light-weight and low-level while being
334 expressive, typed, and extensible at the same time. It aims to be a
335 "universal IR" of sorts, by being at a low enough level that high-level ideas
336 may be cleanly mapped to it (similar to how microprocessors are "universal
337 IR's", allowing many source languages to be mapped to them). By providing
338 type information, LLVM can be used as the target of optimizations: for
339 example, through pointer analysis, it can be proven that a C automatic
340 variable is never accessed outside of the current function... allowing it to
341 be promoted to a simple SSA value instead of a memory location.</p>
345 <!-- _______________________________________________________________________ -->
346 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
348 <div class="doc_text">
350 <p>It is important to note that this document describes 'well formed' LLVM
351 assembly language. There is a difference between what the parser accepts and
352 what is considered 'well formed'. For example, the following instruction is
353 syntactically okay, but not well formed:</p>
355 <div class="doc_code">
357 %x = <a href="#i_add">add</a> i32 1, %x
361 <p>...because the definition of <tt>%x</tt> does not dominate all of its
362 uses. The LLVM infrastructure provides a verification pass that may be used
363 to verify that an LLVM module is well formed. This pass is automatically run
364 by the parser after parsing input assembly and by the optimizer before it
365 outputs bitcode. The violations pointed out by the verifier pass indicate
366 bugs in transformation passes or input to the parser.</p>
370 <!-- Describe the typesetting conventions here. -->
372 <!-- *********************************************************************** -->
373 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
374 <!-- *********************************************************************** -->
376 <div class="doc_text">
378 <p>LLVM identifiers come in two basic types: global and local. Global
379 identifiers (functions, global variables) begin with the <tt>'@'</tt>
380 character. Local identifiers (register names, types) begin with
381 the <tt>'%'</tt> character. Additionally, there are three different formats
382 for identifiers, for different purposes:</p>
385 <li>Named values are represented as a string of characters with their prefix.
386 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
387 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
388 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
389 other characters in their names can be surrounded with quotes. Special
390 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
391 ASCII code for the character in hexadecimal. In this way, any character
392 can be used in a name value, even quotes themselves.</li>
394 <li>Unnamed values are represented as an unsigned numeric value with their
395 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
397 <li>Constants, which are described in a <a href="#constants">section about
398 constants</a>, below.</li>
401 <p>LLVM requires that values start with a prefix for two reasons: Compilers
402 don't need to worry about name clashes with reserved words, and the set of
403 reserved words may be expanded in the future without penalty. Additionally,
404 unnamed identifiers allow a compiler to quickly come up with a temporary
405 variable without having to avoid symbol table conflicts.</p>
407 <p>Reserved words in LLVM are very similar to reserved words in other
408 languages. There are keywords for different opcodes
409 ('<tt><a href="#i_add">add</a></tt>',
410 '<tt><a href="#i_bitcast">bitcast</a></tt>',
411 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
412 ('<tt><a href="#t_void">void</a></tt>',
413 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
414 reserved words cannot conflict with variable names, because none of them
415 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
417 <p>Here is an example of LLVM code to multiply the integer variable
418 '<tt>%X</tt>' by 8:</p>
422 <div class="doc_code">
424 %result = <a href="#i_mul">mul</a> i32 %X, 8
428 <p>After strength reduction:</p>
430 <div class="doc_code">
432 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
436 <p>And the hard way:</p>
438 <div class="doc_code">
440 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
441 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
442 %result = <a href="#i_add">add</a> i32 %1, %1
446 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
447 lexical features of LLVM:</p>
450 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
453 <li>Unnamed temporaries are created when the result of a computation is not
454 assigned to a named value.</li>
456 <li>Unnamed temporaries are numbered sequentially</li>
459 <p>...and it also shows a convention that we follow in this document. When
460 demonstrating instructions, we will follow an instruction with a comment that
461 defines the type and name of value produced. Comments are shown in italic
466 <!-- *********************************************************************** -->
467 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
468 <!-- *********************************************************************** -->
470 <!-- ======================================================================= -->
471 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
474 <div class="doc_text">
476 <p>LLVM programs are composed of "Module"s, each of which is a translation unit
477 of the input programs. Each module consists of functions, global variables,
478 and symbol table entries. Modules may be combined together with the LLVM
479 linker, which merges function (and global variable) definitions, resolves
480 forward declarations, and merges symbol table entries. Here is an example of
481 the "hello world" module:</p>
483 <div class="doc_code">
484 <pre><i>; Declare the string constant as a global constant...</i>
485 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
486 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
488 <i>; External declaration of the puts function</i>
489 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
491 <i>; Definition of main function</i>
492 define i32 @main() { <i>; i32()* </i>
493 <i>; Convert [13 x i8]* to i8 *...</i>
495 href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
497 <i>; Call puts function to write out the string to stdout...</i>
499 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
501 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 is typically used to implement
562 inline functions, templates, or other code which must be generated in each
563 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
564 allowed to be discarded.</dd>
566 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
567 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
568 <tt>linkonce</tt> linkage, except that unreferenced globals with
569 <tt>weak</tt> linkage may not be discarded. This is used for globals that
570 are declared "weak" in C source code.</dd>
572 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
573 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
574 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
576 Symbols with "<tt>common</tt>" linkage are merged in the same way as
577 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
578 <tt>common</tt> symbols may not have an explicit section,
579 must have a zero initializer, and may not be marked '<a
580 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
581 have common linkage.</dd>
584 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
585 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
586 pointer to array type. When two global variables with appending linkage
587 are linked together, the two global arrays are appended together. This is
588 the LLVM, typesafe, equivalent of having the system linker append together
589 "sections" with identical names when .o files are linked.</dd>
591 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
592 <dd>The semantics of this linkage follow the ELF object file model: the symbol
593 is weak until linked, if not linked, the symbol becomes null instead of
594 being an undefined reference.</dd>
596 <dt><tt><b><a name="linkage_linkonce_odr">linkonce_odr</a></b></tt>: </dt>
597 <dt><tt><b><a name="linkage_weak_odr">weak_odr</a></b></tt>: </dt>
598 <dd>Some languages allow differing globals to be merged, such as two functions
599 with different semantics. Other languages, such as <tt>C++</tt>, ensure
600 that only equivalent globals are ever merged (the "one definition rule" -
601 "ODR"). Such languages can use the <tt>linkonce_odr</tt>
602 and <tt>weak_odr</tt> linkage types to indicate that the global will only
603 be merged with equivalent globals. These linkage types are otherwise the
604 same as their non-<tt>odr</tt> versions.</dd>
606 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
607 <dd>If none of the above identifiers are used, the global is externally
608 visible, meaning that it participates in linkage and can be used to
609 resolve external symbol references.</dd>
612 <p>The next two types of linkage are targeted for Microsoft Windows platform
613 only. They are designed to support importing (exporting) symbols from (to)
614 DLLs (Dynamic Link Libraries).</p>
617 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
618 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
619 or variable via a global pointer to a pointer that is set up by the DLL
620 exporting the symbol. On Microsoft Windows targets, the pointer name is
621 formed by combining <code>__imp_</code> and the function or variable
624 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
625 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
626 pointer to a pointer in a DLL, so that it can be referenced with the
627 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
628 name is formed by combining <code>__imp_</code> and the function or
632 <p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
633 another module defined a "<tt>.LC0</tt>" variable and was linked with this
634 one, one of the two would be renamed, preventing a collision. Since
635 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
636 declarations), they are accessible outside of the current module.</p>
638 <p>It is illegal for a function <i>declaration</i> to have any linkage type
639 other than "externally visible", <tt>dllimport</tt>
640 or <tt>extern_weak</tt>.</p>
642 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
643 or <tt>weak_odr</tt> linkages.</p>
647 <!-- ======================================================================= -->
648 <div class="doc_subsection">
649 <a name="callingconv">Calling Conventions</a>
652 <div class="doc_text">
654 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
655 and <a href="#i_invoke">invokes</a> can all have an optional calling
656 convention specified for the call. The calling convention of any pair of
657 dynamic caller/callee must match, or the behavior of the program is
658 undefined. The following calling conventions are supported by LLVM, and more
659 may be added in the future:</p>
662 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
663 <dd>This calling convention (the default if no other calling convention is
664 specified) matches the target C calling conventions. This calling
665 convention supports varargs function calls and tolerates some mismatch in
666 the declared prototype and implemented declaration of the function (as
669 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
670 <dd>This calling convention attempts to make calls as fast as possible
671 (e.g. by passing things in registers). This calling convention allows the
672 target to use whatever tricks it wants to produce fast code for the
673 target, without having to conform to an externally specified ABI
674 (Application Binary Interface). Implementations of this convention should
675 allow arbitrary <a href="CodeGenerator.html#tailcallopt">tail call
676 optimization</a> to be supported. This calling convention does not
677 support varargs and requires the prototype of all callees to exactly match
678 the prototype of the function definition.</dd>
680 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
681 <dd>This calling convention attempts to make code in the caller as efficient
682 as possible under the assumption that the call is not commonly executed.
683 As such, these calls often preserve all registers so that the call does
684 not break any live ranges in the caller side. This calling convention
685 does not support varargs and requires the prototype of all callees to
686 exactly match the prototype of the function definition.</dd>
688 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
689 <dd>Any calling convention may be specified by number, allowing
690 target-specific calling conventions to be used. Target specific calling
691 conventions start at 64.</dd>
694 <p>More calling conventions can be added/defined on an as-needed basis, to
695 support Pascal conventions or any other well-known target-independent
700 <!-- ======================================================================= -->
701 <div class="doc_subsection">
702 <a name="visibility">Visibility Styles</a>
705 <div class="doc_text">
707 <p>All Global Variables and Functions have one of the following visibility
711 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
712 <dd>On targets that use the ELF object file format, default visibility means
713 that the declaration is visible to other modules and, in shared libraries,
714 means that the declared entity may be overridden. On Darwin, default
715 visibility means that the declaration is visible to other modules. Default
716 visibility corresponds to "external linkage" in the language.</dd>
718 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
719 <dd>Two declarations of an object with hidden visibility refer to the same
720 object if they are in the same shared object. Usually, hidden visibility
721 indicates that the symbol will not be placed into the dynamic symbol
722 table, so no other module (executable or shared library) can reference it
725 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
726 <dd>On ELF, protected visibility indicates that the symbol will be placed in
727 the dynamic symbol table, but that references within the defining module
728 will bind to the local symbol. That is, the symbol cannot be overridden by
734 <!-- ======================================================================= -->
735 <div class="doc_subsection">
736 <a name="namedtypes">Named Types</a>
739 <div class="doc_text">
741 <p>LLVM IR allows you to specify name aliases for certain types. This can make
742 it easier to read the IR and make the IR more condensed (particularly when
743 recursive types are involved). An example of a name specification is:</p>
745 <div class="doc_code">
747 %mytype = type { %mytype*, i32 }
751 <p>You may give a name to any <a href="#typesystem">type</a> except
752 "<a href="t_void">void</a>". Type name aliases may be used anywhere a type
753 is expected with the syntax "%mytype".</p>
755 <p>Note that type names are aliases for the structural type that they indicate,
756 and that you can therefore specify multiple names for the same type. This
757 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
758 uses structural typing, the name is not part of the type. When printing out
759 LLVM IR, the printer will pick <em>one name</em> to render all types of a
760 particular shape. This means that if you have code where two different
761 source types end up having the same LLVM type, that the dumper will sometimes
762 print the "wrong" or unexpected type. This is an important design point and
763 isn't going to change.</p>
767 <!-- ======================================================================= -->
768 <div class="doc_subsection">
769 <a name="globalvars">Global Variables</a>
772 <div class="doc_text">
774 <p>Global variables define regions of memory allocated at compilation time
775 instead of run-time. Global variables may optionally be initialized, may
776 have an explicit section to be placed in, and may have an optional explicit
777 alignment specified. A variable may be defined as "thread_local", which
778 means that it will not be shared by threads (each thread will have a
779 separated copy of the variable). A variable may be defined as a global
780 "constant," which indicates that the contents of the variable
781 will <b>never</b> be modified (enabling better optimization, allowing the
782 global data to be placed in the read-only section of an executable, etc).
783 Note that variables that need runtime initialization cannot be marked
784 "constant" as there is a store to the variable.</p>
786 <p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
787 constant, even if the final definition of the global is not. This capability
788 can be used to enable slightly better optimization of the program, but
789 requires the language definition to guarantee that optimizations based on the
790 'constantness' are valid for the translation units that do not include the
793 <p>As SSA values, global variables define pointer values that are in scope
794 (i.e. they dominate) all basic blocks in the program. Global variables
795 always define a pointer to their "content" type because they describe a
796 region of memory, and all memory objects in LLVM are accessed through
799 <p>A global variable may be declared to reside in a target-specific numbered
800 address space. For targets that support them, address spaces may affect how
801 optimizations are performed and/or what target instructions are used to
802 access the variable. The default address space is zero. The address space
803 qualifier must precede any other attributes.</p>
805 <p>LLVM allows an explicit section to be specified for globals. If the target
806 supports it, it will emit globals to the section specified.</p>
808 <p>An explicit alignment may be specified for a global. If not present, or if
809 the alignment is set to zero, the alignment of the global is set by the
810 target to whatever it feels convenient. If an explicit alignment is
811 specified, the global is forced to have at least that much alignment. All
812 alignments must be a power of 2.</p>
814 <p>For example, the following defines a global in a numbered address space with
815 an initializer, section, and alignment:</p>
817 <div class="doc_code">
819 @G = addrspace(5) constant float 1.0, section "foo", align 4
826 <!-- ======================================================================= -->
827 <div class="doc_subsection">
828 <a name="functionstructure">Functions</a>
831 <div class="doc_text">
833 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord, an
834 optional <a href="#linkage">linkage type</a>, an optional
835 <a href="#visibility">visibility style</a>, an optional
836 <a href="#callingconv">calling convention</a>, a return type, an optional
837 <a href="#paramattrs">parameter attribute</a> for the return type, a function
838 name, a (possibly empty) argument list (each with optional
839 <a href="#paramattrs">parameter attributes</a>), optional
840 <a href="#fnattrs">function attributes</a>, an optional section, an optional
841 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
842 curly brace, a list of basic blocks, and a closing curly brace.</p>
844 <p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
845 optional <a href="#linkage">linkage type</a>, an optional
846 <a href="#visibility">visibility style</a>, an optional
847 <a href="#callingconv">calling convention</a>, a return type, an optional
848 <a href="#paramattrs">parameter attribute</a> for the return type, a function
849 name, a possibly empty list of arguments, an optional alignment, and an
850 optional <a href="#gc">garbage collector name</a>.</p>
852 <p>A function definition contains a list of basic blocks, forming the CFG
853 (Control Flow Graph) for the function. Each basic block may optionally start
854 with a label (giving the basic block a symbol table entry), contains a list
855 of instructions, and ends with a <a href="#terminators">terminator</a>
856 instruction (such as a branch or function return).</p>
858 <p>The first basic block in a function is special in two ways: it is immediately
859 executed on entrance to the function, and it is not allowed to have
860 predecessor basic blocks (i.e. there can not be any branches to the entry
861 block of a function). Because the block can have no predecessors, it also
862 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
864 <p>LLVM allows an explicit section to be specified for functions. If the target
865 supports it, it will emit functions to the section specified.</p>
867 <p>An explicit alignment may be specified for a function. If not present, or if
868 the alignment is set to zero, the alignment of the function is set by the
869 target to whatever it feels convenient. If an explicit alignment is
870 specified, the function is forced to have at least that much alignment. All
871 alignments must be a power of 2.</p>
874 <div class="doc_code">
876 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
877 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
878 <ResultType> @<FunctionName> ([argument list])
879 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
880 [<a href="#gc">gc</a>] { ... }
886 <!-- ======================================================================= -->
887 <div class="doc_subsection">
888 <a name="aliasstructure">Aliases</a>
891 <div class="doc_text">
893 <p>Aliases act as "second name" for the aliasee value (which can be either
894 function, global variable, another alias or bitcast of global value). Aliases
895 may have an optional <a href="#linkage">linkage type</a>, and an
896 optional <a href="#visibility">visibility style</a>.</p>
899 <div class="doc_code">
901 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
907 <!-- ======================================================================= -->
908 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
910 <div class="doc_text">
912 <p>The return type and each parameter of a function type may have a set of
913 <i>parameter attributes</i> associated with them. Parameter attributes are
914 used to communicate additional information about the result or parameters of
915 a function. Parameter attributes are considered to be part of the function,
916 not of the function type, so functions with different parameter attributes
917 can have the same function type.</p>
919 <p>Parameter attributes are simple keywords that follow the type specified. If
920 multiple parameter attributes are needed, they are space separated. For
923 <div class="doc_code">
925 declare i32 @printf(i8* noalias nocapture, ...)
926 declare i32 @atoi(i8 zeroext)
927 declare signext i8 @returns_signed_char()
931 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
932 <tt>readonly</tt>) come immediately after the argument list.</p>
934 <p>Currently, only the following parameter attributes are defined:</p>
937 <dt><tt>zeroext</tt></dt>
938 <dd>This indicates to the code generator that the parameter or return value
939 should be zero-extended to a 32-bit value by the caller (for a parameter)
940 or the callee (for a return value).</dd>
942 <dt><tt>signext</tt></dt>
943 <dd>This indicates to the code generator that the parameter or return value
944 should be sign-extended to a 32-bit value by the caller (for a parameter)
945 or the callee (for a return value).</dd>
947 <dt><tt>inreg</tt></dt>
948 <dd>This indicates that this parameter or return value should be treated in a
949 special target-dependent fashion during while emitting code for a function
950 call or return (usually, by putting it in a register as opposed to memory,
951 though some targets use it to distinguish between two different kinds of
952 registers). Use of this attribute is target-specific.</dd>
954 <dt><tt><a name="byval">byval</a></tt></dt>
955 <dd>This indicates that the pointer parameter should really be passed by value
956 to the function. The attribute implies that a hidden copy of the pointee
957 is made between the caller and the callee, so the callee is unable to
958 modify the value in the callee. This attribute is only valid on LLVM
959 pointer arguments. It is generally used to pass structs and arrays by
960 value, but is also valid on pointers to scalars. The copy is considered
961 to belong to the caller not the callee (for example,
962 <tt><a href="#readonly">readonly</a></tt> functions should not write to
963 <tt>byval</tt> parameters). This is not a valid attribute for return
964 values. The byval attribute also supports specifying an alignment with
965 the align attribute. This has a target-specific effect on the code
966 generator that usually indicates a desired alignment for the synthesized
969 <dt><tt>sret</tt></dt>
970 <dd>This indicates that the pointer parameter specifies the address of a
971 structure that is the return value of the function in the source program.
972 This pointer must be guaranteed by the caller to be valid: loads and
973 stores to the structure may be assumed by the callee to not to trap. This
974 may only be applied to the first parameter. This is not a valid attribute
975 for return values. </dd>
977 <dt><tt>noalias</tt></dt>
978 <dd>This indicates that the pointer does not alias any global or any other
979 parameter. The caller is responsible for ensuring that this is the
980 case. On a function return value, <tt>noalias</tt> additionally indicates
981 that the pointer does not alias any other pointers visible to the
982 caller. For further details, please see the discussion of the NoAlias
984 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
987 <dt><tt>nocapture</tt></dt>
988 <dd>This indicates that the callee does not make any copies of the pointer
989 that outlive the callee itself. This is not a valid attribute for return
992 <dt><tt>nest</tt></dt>
993 <dd>This indicates that the pointer parameter can be excised using the
994 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
995 attribute for return values.</dd>
1000 <!-- ======================================================================= -->
1001 <div class="doc_subsection">
1002 <a name="gc">Garbage Collector Names</a>
1005 <div class="doc_text">
1007 <p>Each function may specify a garbage collector name, which is simply a
1010 <div class="doc_code">
1012 define void @f() gc "name" { ...
1016 <p>The compiler declares the supported values of <i>name</i>. Specifying a
1017 collector which will cause the compiler to alter its output in order to
1018 support the named garbage collection algorithm.</p>
1022 <!-- ======================================================================= -->
1023 <div class="doc_subsection">
1024 <a name="fnattrs">Function Attributes</a>
1027 <div class="doc_text">
1029 <p>Function attributes are set to communicate additional information about a
1030 function. Function attributes are considered to be part of the function, not
1031 of the function type, so functions with different parameter attributes can
1032 have the same function type.</p>
1034 <p>Function attributes are simple keywords that follow the type specified. If
1035 multiple attributes are needed, they are space separated. For example:</p>
1037 <div class="doc_code">
1039 define void @f() noinline { ... }
1040 define void @f() alwaysinline { ... }
1041 define void @f() alwaysinline optsize { ... }
1042 define void @f() optsize
1047 <dt><tt>alwaysinline</tt></dt>
1048 <dd>This attribute indicates that the inliner should attempt to inline this
1049 function into callers whenever possible, ignoring any active inlining size
1050 threshold for this caller.</dd>
1052 <dt><tt>inlinehint</tt></dt>
1053 <dd>This attribute indicates that the source code contained a hint that inlining
1054 this function is desirable (such as the "inline" keyword in C/C++). It
1055 is just a hint; it imposes no requirements on the inliner.</dd>
1057 <dt><tt>noinline</tt></dt>
1058 <dd>This attribute indicates that the inliner should never inline this
1059 function in any situation. This attribute may not be used together with
1060 the <tt>alwaysinline</tt> attribute.</dd>
1062 <dt><tt>optsize</tt></dt>
1063 <dd>This attribute suggests that optimization passes and code generator passes
1064 make choices that keep the code size of this function low, and otherwise
1065 do optimizations specifically to reduce code size.</dd>
1067 <dt><tt>noreturn</tt></dt>
1068 <dd>This function attribute indicates that the function never returns
1069 normally. This produces undefined behavior at runtime if the function
1070 ever does dynamically return.</dd>
1072 <dt><tt>nounwind</tt></dt>
1073 <dd>This function attribute indicates that the function never returns with an
1074 unwind or exceptional control flow. If the function does unwind, its
1075 runtime behavior is undefined.</dd>
1077 <dt><tt>readnone</tt></dt>
1078 <dd>This attribute indicates that the function computes its result (or decides
1079 to unwind an exception) based strictly on its arguments, without
1080 dereferencing any pointer arguments or otherwise accessing any mutable
1081 state (e.g. memory, control registers, etc) visible to caller functions.
1082 It does not write through any pointer arguments
1083 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1084 changes any state visible to callers. This means that it cannot unwind
1085 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1086 could use the <tt>unwind</tt> instruction.</dd>
1088 <dt><tt><a name="readonly">readonly</a></tt></dt>
1089 <dd>This attribute indicates that the function does not write through any
1090 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1091 arguments) or otherwise modify any state (e.g. memory, control registers,
1092 etc) visible to caller functions. It may dereference pointer arguments
1093 and read state that may be set in the caller. A readonly function always
1094 returns the same value (or unwinds an exception identically) when called
1095 with the same set of arguments and global state. It cannot unwind an
1096 exception by calling the <tt>C++</tt> exception throwing methods, but may
1097 use the <tt>unwind</tt> instruction.</dd>
1099 <dt><tt><a name="ssp">ssp</a></tt></dt>
1100 <dd>This attribute indicates that the function should emit a stack smashing
1101 protector. It is in the form of a "canary"—a random value placed on
1102 the stack before the local variables that's checked upon return from the
1103 function to see if it has been overwritten. A heuristic is used to
1104 determine if a function needs stack protectors or not.<br>
1106 If a function that has an <tt>ssp</tt> attribute is inlined into a
1107 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1108 function will have an <tt>ssp</tt> attribute.</dd>
1110 <dt><tt>sspreq</tt></dt>
1111 <dd>This attribute indicates that the function should <em>always</em> emit a
1112 stack smashing protector. This overrides
1113 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1115 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1116 function that doesn't have an <tt>sspreq</tt> attribute or which has
1117 an <tt>ssp</tt> attribute, then the resulting function will have
1118 an <tt>sspreq</tt> attribute.</dd>
1120 <dt><tt>noredzone</tt></dt>
1121 <dd>This attribute indicates that the code generator should not use a red
1122 zone, even if the target-specific ABI normally permits it.</dd>
1124 <dt><tt>noimplicitfloat</tt></dt>
1125 <dd>This attributes disables implicit floating point instructions.</dd>
1127 <dt><tt>naked</tt></dt>
1128 <dd>This attribute disables prologue / epilogue emission for the function.
1129 This can have very system-specific consequences.</dd>
1134 <!-- ======================================================================= -->
1135 <div class="doc_subsection">
1136 <a name="moduleasm">Module-Level Inline Assembly</a>
1139 <div class="doc_text">
1141 <p>Modules may contain "module-level inline asm" blocks, which corresponds to
1142 the GCC "file scope inline asm" blocks. These blocks are internally
1143 concatenated by LLVM and treated as a single unit, but may be separated in
1144 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
1146 <div class="doc_code">
1148 module asm "inline asm code goes here"
1149 module asm "more can go here"
1153 <p>The strings can contain any character by escaping non-printable characters.
1154 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1157 <p>The inline asm code is simply printed to the machine code .s file when
1158 assembly code is generated.</p>
1162 <!-- ======================================================================= -->
1163 <div class="doc_subsection">
1164 <a name="datalayout">Data Layout</a>
1167 <div class="doc_text">
1169 <p>A module may specify a target specific data layout string that specifies how
1170 data is to be laid out in memory. The syntax for the data layout is
1173 <div class="doc_code">
1175 target datalayout = "<i>layout specification</i>"
1179 <p>The <i>layout specification</i> consists of a list of specifications
1180 separated by the minus sign character ('-'). Each specification starts with
1181 a letter and may include other information after the letter to define some
1182 aspect of the data layout. The specifications accepted are as follows:</p>
1186 <dd>Specifies that the target lays out data in big-endian form. That is, the
1187 bits with the most significance have the lowest address location.</dd>
1190 <dd>Specifies that the target lays out data in little-endian form. That is,
1191 the bits with the least significance have the lowest address
1194 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1195 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1196 <i>preferred</i> alignments. All sizes are in bits. Specifying
1197 the <i>pref</i> alignment is optional. If omitted, the
1198 preceding <tt>:</tt> should be omitted too.</dd>
1200 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1201 <dd>This specifies the alignment for an integer type of a given bit
1202 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1204 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1205 <dd>This specifies the alignment for a vector type of a given bit
1208 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1209 <dd>This specifies the alignment for a floating point type of a given bit
1210 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1213 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1214 <dd>This specifies the alignment for an aggregate type of a given bit
1217 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1218 <dd>This specifies the alignment for a stack object of a given bit
1222 <p>When constructing the data layout for a given target, LLVM starts with a
1223 default set of specifications which are then (possibly) overriden by the
1224 specifications in the <tt>datalayout</tt> keyword. The default specifications
1225 are given in this list:</p>
1228 <li><tt>E</tt> - big endian</li>
1229 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1230 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1231 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1232 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1233 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1234 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1235 alignment of 64-bits</li>
1236 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1237 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1238 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1239 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1240 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1241 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
1244 <p>When LLVM is determining the alignment for a given type, it uses the
1245 following rules:</p>
1248 <li>If the type sought is an exact match for one of the specifications, that
1249 specification is used.</li>
1251 <li>If no match is found, and the type sought is an integer type, then the
1252 smallest integer type that is larger than the bitwidth of the sought type
1253 is used. If none of the specifications are larger than the bitwidth then
1254 the the largest integer type is used. For example, given the default
1255 specifications above, the i7 type will use the alignment of i8 (next
1256 largest) while both i65 and i256 will use the alignment of i64 (largest
1259 <li>If no match is found, and the type sought is a vector type, then the
1260 largest vector type that is smaller than the sought vector type will be
1261 used as a fall back. This happens because <128 x double> can be
1262 implemented in terms of 64 <2 x double>, for example.</li>
1267 <!-- ======================================================================= -->
1268 <div class="doc_subsection">
1269 <a name="pointeraliasing">Pointer Aliasing Rules</a>
1272 <div class="doc_text">
1274 <p>Any memory access must be done through a pointer value associated
1275 with an address range of the memory access, otherwise the behavior
1276 is undefined. Pointer values are associated with address ranges
1277 according to the following rules:</p>
1280 <li>A pointer value formed from a
1281 <tt><a href="#i_getelementptr">getelementptr</a></tt> instruction
1282 is associated with the addresses associated with the first operand
1283 of the <tt>getelementptr</tt>.</li>
1284 <li>An address of a global variable is associated with the address
1285 range of the variable's storage.</li>
1286 <li>The result value of an allocation instruction is associated with
1287 the address range of the allocated storage.</li>
1288 <li>A null pointer in the default address-space is associated with
1290 <li>A pointer value formed by an
1291 <tt><a href="#i_inttoptr">inttoptr</a></tt> is associated with all
1292 address ranges of all pointer values that contribute (directly or
1293 indirectly) to the computation of the pointer's value.</li>
1294 <li>The result value of a
1295 <tt><a href="#i_bitcast">bitcast</a></tt> is associated with all
1296 addresses associated with the operand of the <tt>bitcast</tt>.</li>
1297 <li>An integer constant other than zero or a pointer value returned
1298 from a function not defined within LLVM may be associated with address
1299 ranges allocated through mechanisms other than those provided by
1300 LLVM. Such ranges shall not overlap with any ranges of addresses
1301 allocated by mechanisms provided by LLVM.</li>
1304 <p>LLVM IR does not associate types with memory. The result type of a
1305 <tt><a href="#i_load">load</a></tt> merely indicates the size and
1306 alignment of the memory from which to load, as well as the
1307 interpretation of the value. The first operand of a
1308 <tt><a href="#i_store">store</a></tt> similarly only indicates the size
1309 and alignment of the store.</p>
1311 <p>Consequently, type-based alias analysis, aka TBAA, aka
1312 <tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1313 LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1314 additional information which specialized optimization passes may use
1315 to implement type-based alias analysis.</p>
1319 <!-- *********************************************************************** -->
1320 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1321 <!-- *********************************************************************** -->
1323 <div class="doc_text">
1325 <p>The LLVM type system is one of the most important features of the
1326 intermediate representation. Being typed enables a number of optimizations
1327 to be performed on the intermediate representation directly, without having
1328 to do extra analyses on the side before the transformation. A strong type
1329 system makes it easier to read the generated code and enables novel analyses
1330 and transformations that are not feasible to perform on normal three address
1331 code representations.</p>
1335 <!-- ======================================================================= -->
1336 <div class="doc_subsection"> <a name="t_classifications">Type
1337 Classifications</a> </div>
1339 <div class="doc_text">
1341 <p>The types fall into a few useful classifications:</p>
1343 <table border="1" cellspacing="0" cellpadding="4">
1345 <tr><th>Classification</th><th>Types</th></tr>
1347 <td><a href="#t_integer">integer</a></td>
1348 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1351 <td><a href="#t_floating">floating point</a></td>
1352 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1355 <td><a name="t_firstclass">first class</a></td>
1356 <td><a href="#t_integer">integer</a>,
1357 <a href="#t_floating">floating point</a>,
1358 <a href="#t_pointer">pointer</a>,
1359 <a href="#t_vector">vector</a>,
1360 <a href="#t_struct">structure</a>,
1361 <a href="#t_array">array</a>,
1362 <a href="#t_label">label</a>,
1363 <a href="#t_metadata">metadata</a>.
1367 <td><a href="#t_primitive">primitive</a></td>
1368 <td><a href="#t_label">label</a>,
1369 <a href="#t_void">void</a>,
1370 <a href="#t_floating">floating point</a>,
1371 <a href="#t_metadata">metadata</a>.</td>
1374 <td><a href="#t_derived">derived</a></td>
1375 <td><a href="#t_integer">integer</a>,
1376 <a href="#t_array">array</a>,
1377 <a href="#t_function">function</a>,
1378 <a href="#t_pointer">pointer</a>,
1379 <a href="#t_struct">structure</a>,
1380 <a href="#t_pstruct">packed structure</a>,
1381 <a href="#t_vector">vector</a>,
1382 <a href="#t_opaque">opaque</a>.
1388 <p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1389 important. Values of these types are the only ones which can be produced by
1394 <!-- ======================================================================= -->
1395 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1397 <div class="doc_text">
1399 <p>The primitive types are the fundamental building blocks of the LLVM
1404 <!-- _______________________________________________________________________ -->
1405 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1407 <div class="doc_text">
1410 <p>The integer type is a very simple type that simply specifies an arbitrary
1411 bit width for the integer type desired. Any bit width from 1 bit to
1412 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
1419 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1423 <table class="layout">
1425 <td class="left"><tt>i1</tt></td>
1426 <td class="left">a single-bit integer.</td>
1429 <td class="left"><tt>i32</tt></td>
1430 <td class="left">a 32-bit integer.</td>
1433 <td class="left"><tt>i1942652</tt></td>
1434 <td class="left">a really big integer of over 1 million bits.</td>
1438 <p>Note that the code generator does not yet support large integer types to be
1439 used as function return types. The specific limit on how large a return type
1440 the code generator can currently handle is target-dependent; currently it's
1441 often 64 bits for 32-bit targets and 128 bits for 64-bit targets.</p>
1445 <!-- _______________________________________________________________________ -->
1446 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1448 <div class="doc_text">
1452 <tr><th>Type</th><th>Description</th></tr>
1453 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1454 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1455 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1456 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1457 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1463 <!-- _______________________________________________________________________ -->
1464 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1466 <div class="doc_text">
1469 <p>The void type does not represent any value and has no size.</p>
1478 <!-- _______________________________________________________________________ -->
1479 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1481 <div class="doc_text">
1484 <p>The label type represents code labels.</p>
1493 <!-- _______________________________________________________________________ -->
1494 <div class="doc_subsubsection"> <a name="t_metadata">Metadata Type</a> </div>
1496 <div class="doc_text">
1499 <p>The metadata type represents embedded metadata. No derived types may be
1500 created from metadata except for <a href="#t_function">function</a>
1511 <!-- ======================================================================= -->
1512 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1514 <div class="doc_text">
1516 <p>The real power in LLVM comes from the derived types in the system. This is
1517 what allows a programmer to represent arrays, functions, pointers, and other
1518 useful types. Each of these types contain one or more element types which
1519 may be a primitive type, or another derived type. For example, it is
1520 possible to have a two dimensional array, using an array as the element type
1521 of another array.</p>
1525 <!-- _______________________________________________________________________ -->
1526 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1528 <div class="doc_text">
1531 <p>The array type is a very simple derived type that arranges elements
1532 sequentially in memory. The array type requires a size (number of elements)
1533 and an underlying data type.</p>
1537 [<# elements> x <elementtype>]
1540 <p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1541 be any type with a size.</p>
1544 <table class="layout">
1546 <td class="left"><tt>[40 x i32]</tt></td>
1547 <td class="left">Array of 40 32-bit integer values.</td>
1550 <td class="left"><tt>[41 x i32]</tt></td>
1551 <td class="left">Array of 41 32-bit integer values.</td>
1554 <td class="left"><tt>[4 x i8]</tt></td>
1555 <td class="left">Array of 4 8-bit integer values.</td>
1558 <p>Here are some examples of multidimensional arrays:</p>
1559 <table class="layout">
1561 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1562 <td class="left">3x4 array of 32-bit integer values.</td>
1565 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1566 <td class="left">12x10 array of single precision floating point values.</td>
1569 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1570 <td class="left">2x3x4 array of 16-bit integer values.</td>
1574 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1575 length array. Normally, accesses past the end of an array are undefined in
1576 LLVM (e.g. it is illegal to access the 5th element of a 3 element array). As
1577 a special case, however, zero length arrays are recognized to be variable
1578 length. This allows implementation of 'pascal style arrays' with the LLVM
1579 type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
1581 <p>Note that the code generator does not yet support large aggregate types to be
1582 used as function return types. The specific limit on how large an aggregate
1583 return type the code generator can currently handle is target-dependent, and
1584 also dependent on the aggregate element types.</p>
1588 <!-- _______________________________________________________________________ -->
1589 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1591 <div class="doc_text">
1594 <p>The function type can be thought of as a function signature. It consists of
1595 a return type and a list of formal parameter types. The return type of a
1596 function type is a scalar type, a void type, or a struct type. If the return
1597 type is a struct type then all struct elements must be of first class types,
1598 and the struct must have at least one element.</p>
1602 <returntype> (<parameter list>)
1605 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1606 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1607 which indicates that the function takes a variable number of arguments.
1608 Variable argument functions can access their arguments with
1609 the <a href="#int_varargs">variable argument handling intrinsic</a>
1610 functions. '<tt><returntype></tt>' is a any type except
1611 <a href="#t_label">label</a>.</p>
1614 <table class="layout">
1616 <td class="left"><tt>i32 (i32)</tt></td>
1617 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1619 </tr><tr class="layout">
1620 <td class="left"><tt>float (i16 signext, i32 *) *
1622 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1623 an <tt>i16</tt> that should be sign extended and a
1624 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1627 </tr><tr class="layout">
1628 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1629 <td class="left">A vararg function that takes at least one
1630 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1631 which returns an integer. This is the signature for <tt>printf</tt> in
1634 </tr><tr class="layout">
1635 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1636 <td class="left">A function taking an <tt>i32</tt>, returning a
1637 <a href="#t_struct">structure</a> containing two <tt>i32</tt> values
1644 <!-- _______________________________________________________________________ -->
1645 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1647 <div class="doc_text">
1650 <p>The structure type is used to represent a collection of data members together
1651 in memory. The packing of the field types is defined to match the ABI of the
1652 underlying processor. The elements of a structure may be any type that has a
1655 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1656 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1657 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1661 { <type list> }
1665 <table class="layout">
1667 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1668 <td class="left">A triple of three <tt>i32</tt> values</td>
1669 </tr><tr class="layout">
1670 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1671 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1672 second element is a <a href="#t_pointer">pointer</a> to a
1673 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1674 an <tt>i32</tt>.</td>
1678 <p>Note that the code generator does not yet support large aggregate types to be
1679 used as function return types. The specific limit on how large an aggregate
1680 return type the code generator can currently handle is target-dependent, and
1681 also dependent on the aggregate element types.</p>
1685 <!-- _______________________________________________________________________ -->
1686 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1689 <div class="doc_text">
1692 <p>The packed structure type is used to represent a collection of data members
1693 together in memory. There is no padding between fields. Further, the
1694 alignment of a packed structure is 1 byte. The elements of a packed
1695 structure may be any type that has a size.</p>
1697 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1698 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1699 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1703 < { <type list> } >
1707 <table class="layout">
1709 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1710 <td class="left">A triple of three <tt>i32</tt> values</td>
1711 </tr><tr class="layout">
1713 <tt>< { float, i32 (i32)* } ></tt></td>
1714 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1715 second element is a <a href="#t_pointer">pointer</a> to a
1716 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1717 an <tt>i32</tt>.</td>
1723 <!-- _______________________________________________________________________ -->
1724 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1726 <div class="doc_text">
1729 <p>As in many languages, the pointer type represents a pointer or reference to
1730 another object, which must live in memory. Pointer types may have an optional
1731 address space attribute defining the target-specific numbered address space
1732 where the pointed-to object resides. The default address space is zero.</p>
1734 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
1735 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
1743 <table class="layout">
1745 <td class="left"><tt>[4 x i32]*</tt></td>
1746 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1747 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1750 <td class="left"><tt>i32 (i32 *) *</tt></td>
1751 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1752 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1756 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1757 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1758 that resides in address space #5.</td>
1764 <!-- _______________________________________________________________________ -->
1765 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1767 <div class="doc_text">
1770 <p>A vector type is a simple derived type that represents a vector of elements.
1771 Vector types are used when multiple primitive data are operated in parallel
1772 using a single instruction (SIMD). A vector type requires a size (number of
1773 elements) and an underlying primitive data type. Vectors must have a power
1774 of two length (1, 2, 4, 8, 16 ...). Vector types are considered
1775 <a href="#t_firstclass">first class</a>.</p>
1779 < <# elements> x <elementtype> >
1782 <p>The number of elements is a constant integer value; elementtype may be any
1783 integer or floating point type.</p>
1786 <table class="layout">
1788 <td class="left"><tt><4 x i32></tt></td>
1789 <td class="left">Vector of 4 32-bit integer values.</td>
1792 <td class="left"><tt><8 x float></tt></td>
1793 <td class="left">Vector of 8 32-bit floating-point values.</td>
1796 <td class="left"><tt><2 x i64></tt></td>
1797 <td class="left">Vector of 2 64-bit integer values.</td>
1801 <p>Note that the code generator does not yet support large vector types to be
1802 used as function return types. The specific limit on how large a vector
1803 return type codegen can currently handle is target-dependent; currently it's
1804 often a few times longer than a hardware vector register.</p>
1808 <!-- _______________________________________________________________________ -->
1809 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1810 <div class="doc_text">
1813 <p>Opaque types are used to represent unknown types in the system. This
1814 corresponds (for example) to the C notion of a forward declared structure
1815 type. In LLVM, opaque types can eventually be resolved to any type (not just
1816 a structure type).</p>
1824 <table class="layout">
1826 <td class="left"><tt>opaque</tt></td>
1827 <td class="left">An opaque type.</td>
1833 <!-- ======================================================================= -->
1834 <div class="doc_subsection">
1835 <a name="t_uprefs">Type Up-references</a>
1838 <div class="doc_text">
1841 <p>An "up reference" allows you to refer to a lexically enclosing type without
1842 requiring it to have a name. For instance, a structure declaration may
1843 contain a pointer to any of the types it is lexically a member of. Example
1844 of up references (with their equivalent as named type declarations)
1848 { \2 * } %x = type { %x* }
1849 { \2 }* %y = type { %y }*
1853 <p>An up reference is needed by the asmprinter for printing out cyclic types
1854 when there is no declared name for a type in the cycle. Because the
1855 asmprinter does not want to print out an infinite type string, it needs a
1856 syntax to handle recursive types that have no names (all names are optional
1864 <p>The level is the count of the lexical type that is being referred to.</p>
1867 <table class="layout">
1869 <td class="left"><tt>\1*</tt></td>
1870 <td class="left">Self-referential pointer.</td>
1873 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
1874 <td class="left">Recursive structure where the upref refers to the out-most
1881 <!-- *********************************************************************** -->
1882 <div class="doc_section"> <a name="constants">Constants</a> </div>
1883 <!-- *********************************************************************** -->
1885 <div class="doc_text">
1887 <p>LLVM has several different basic types of constants. This section describes
1888 them all and their syntax.</p>
1892 <!-- ======================================================================= -->
1893 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1895 <div class="doc_text">
1898 <dt><b>Boolean constants</b></dt>
1899 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1900 constants of the <tt><a href="#t_integer">i1</a></tt> type.</dd>
1902 <dt><b>Integer constants</b></dt>
1903 <dd>Standard integers (such as '4') are constants of
1904 the <a href="#t_integer">integer</a> type. Negative numbers may be used
1905 with integer types.</dd>
1907 <dt><b>Floating point constants</b></dt>
1908 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1909 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1910 notation (see below). The assembler requires the exact decimal value of a
1911 floating-point constant. For example, the assembler accepts 1.25 but
1912 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1913 constants must have a <a href="#t_floating">floating point</a> type. </dd>
1915 <dt><b>Null pointer constants</b></dt>
1916 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1917 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1920 <p>The one non-intuitive notation for constants is the hexadecimal form of
1921 floating point constants. For example, the form '<tt>double
1922 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
1923 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
1924 constants are required (and the only time that they are generated by the
1925 disassembler) is when a floating point constant must be emitted but it cannot
1926 be represented as a decimal floating point number in a reasonable number of
1927 digits. For example, NaN's, infinities, and other special values are
1928 represented in their IEEE hexadecimal format so that assembly and disassembly
1929 do not cause any bits to change in the constants.</p>
1931 <p>When using the hexadecimal form, constants of types float and double are
1932 represented using the 16-digit form shown above (which matches the IEEE754
1933 representation for double); float values must, however, be exactly
1934 representable as IEE754 single precision. Hexadecimal format is always used
1935 for long double, and there are three forms of long double. The 80-bit format
1936 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
1937 The 128-bit format used by PowerPC (two adjacent doubles) is represented
1938 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
1939 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
1940 currently supported target uses this format. Long doubles will only work if
1941 they match the long double format on your target. All hexadecimal formats
1942 are big-endian (sign bit at the left).</p>
1946 <!-- ======================================================================= -->
1947 <div class="doc_subsection">
1948 <a name="aggregateconstants"></a> <!-- old anchor -->
1949 <a name="complexconstants">Complex Constants</a>
1952 <div class="doc_text">
1954 <p>Complex constants are a (potentially recursive) combination of simple
1955 constants and smaller complex constants.</p>
1958 <dt><b>Structure constants</b></dt>
1959 <dd>Structure constants are represented with notation similar to structure
1960 type definitions (a comma separated list of elements, surrounded by braces
1961 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1962 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
1963 Structure constants must have <a href="#t_struct">structure type</a>, and
1964 the number and types of elements must match those specified by the
1967 <dt><b>Array constants</b></dt>
1968 <dd>Array constants are represented with notation similar to array type
1969 definitions (a comma separated list of elements, surrounded by square
1970 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
1971 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
1972 the number and types of elements must match those specified by the
1975 <dt><b>Vector constants</b></dt>
1976 <dd>Vector constants are represented with notation similar to vector type
1977 definitions (a comma separated list of elements, surrounded by
1978 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32
1979 42, i32 11, i32 74, i32 100 ></tt>". Vector constants must
1980 have <a href="#t_vector">vector type</a>, and the number and types of
1981 elements must match those specified by the type.</dd>
1983 <dt><b>Zero initialization</b></dt>
1984 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1985 value to zero of <em>any</em> type, including scalar and aggregate types.
1986 This is often used to avoid having to print large zero initializers
1987 (e.g. for large arrays) and is always exactly equivalent to using explicit
1988 zero initializers.</dd>
1990 <dt><b>Metadata node</b></dt>
1991 <dd>A metadata node is a structure-like constant with
1992 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
1993 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
1994 be interpreted as part of the instruction stream, metadata is a place to
1995 attach additional information such as debug info.</dd>
2000 <!-- ======================================================================= -->
2001 <div class="doc_subsection">
2002 <a name="globalconstants">Global Variable and Function Addresses</a>
2005 <div class="doc_text">
2007 <p>The addresses of <a href="#globalvars">global variables</a>
2008 and <a href="#functionstructure">functions</a> are always implicitly valid
2009 (link-time) constants. These constants are explicitly referenced when
2010 the <a href="#identifiers">identifier for the global</a> is used and always
2011 have <a href="#t_pointer">pointer</a> type. For example, the following is a
2012 legal LLVM file:</p>
2014 <div class="doc_code">
2018 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2024 <!-- ======================================================================= -->
2025 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
2026 <div class="doc_text">
2028 <p>The string '<tt>undef</tt>' can be used anywhere a constant is expected, and
2029 indicates that the user of the value may receive an unspecified bit-pattern.
2030 Undefined values may be of any type (other than label or void) and be used
2031 anywhere a constant is permitted.</p>
2033 <p>Undefined values are useful because they indicate to the compiler that the
2034 program is well defined no matter what value is used. This gives the
2035 compiler more freedom to optimize. Here are some examples of (potentially
2036 surprising) transformations that are valid (in pseudo IR):</p>
2039 <div class="doc_code">
2051 <p>This is safe because all of the output bits are affected by the undef bits.
2052 Any output bit can have a zero or one depending on the input bits.</p>
2054 <div class="doc_code">
2067 <p>These logical operations have bits that are not always affected by the input.
2068 For example, if "%X" has a zero bit, then the output of the 'and' operation will
2069 always be a zero, no matter what the corresponding bit from the undef is. As
2070 such, it is unsafe to optimize or assume that the result of the and is undef.
2071 However, it is safe to assume that all bits of the undef could be 0, and
2072 optimize the and to 0. Likewise, it is safe to assume that all the bits of
2073 the undef operand to the or could be set, allowing the or to be folded to
2076 <div class="doc_code">
2078 %A = select undef, %X, %Y
2079 %B = select undef, 42, %Y
2080 %C = select %X, %Y, undef
2092 <p>This set of examples show that undefined select (and conditional branch)
2093 conditions can go "either way" but they have to come from one of the two
2094 operands. In the %A example, if %X and %Y were both known to have a clear low
2095 bit, then %A would have to have a cleared low bit. However, in the %C example,
2096 the optimizer is allowed to assume that the undef operand could be the same as
2097 %Y, allowing the whole select to be eliminated.</p>
2100 <div class="doc_code">
2102 %A = xor undef, undef
2121 <p>This example points out that two undef operands are not necessarily the same.
2122 This can be surprising to people (and also matches C semantics) where they
2123 assume that "X^X" is always zero, even if X is undef. This isn't true for a
2124 number of reasons, but the short answer is that an undef "variable" can
2125 arbitrarily change its value over its "live range". This is true because the
2126 "variable" doesn't actually <em>have a live range</em>. Instead, the value is
2127 logically read from arbitrary registers that happen to be around when needed,
2128 so the value is not necessarily consistent over time. In fact, %A and %C need
2129 to have the same semantics or the core LLVM "replace all uses with" concept
2132 <div class="doc_code">
2142 <p>These examples show the crucial difference between an <em>undefined
2143 value</em> and <em>undefined behavior</em>. An undefined value (like undef) is
2144 allowed to have an arbitrary bit-pattern. This means that the %A operation
2145 can be constant folded to undef because the undef could be an SNaN, and fdiv is
2146 not (currently) defined on SNaN's. However, in the second example, we can make
2147 a more aggressive assumption: because the undef is allowed to be an arbitrary
2148 value, we are allowed to assume that it could be zero. Since a divide by zero
2149 has <em>undefined behavior</em>, we are allowed to assume that the operation
2150 does not execute at all. This allows us to delete the divide and all code after
2151 it: since the undefined operation "can't happen", the optimizer can assume that
2152 it occurs in dead code.
2155 <div class="doc_code">
2157 a: store undef -> %X
2158 b: store %X -> undef
2165 <p>These examples reiterate the fdiv example: a store "of" an undefined value
2166 can be assumed to not have any effect: we can assume that the value is
2167 overwritten with bits that happen to match what was already there. However, a
2168 store "to" an undefined location could clobber arbitrary memory, therefore, it
2169 has undefined behavior.</p>
2173 <!-- ======================================================================= -->
2174 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
2177 <div class="doc_text">
2179 <p>Constant expressions are used to allow expressions involving other constants
2180 to be used as constants. Constant expressions may be of
2181 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2182 operation that does not have side effects (e.g. load and call are not
2183 supported). The following is the syntax for constant expressions:</p>
2186 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
2187 <dd>Truncate a constant to another type. The bit size of CST must be larger
2188 than the bit size of TYPE. Both types must be integers.</dd>
2190 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
2191 <dd>Zero extend a constant to another type. The bit size of CST must be
2192 smaller or equal to the bit size of TYPE. Both types must be
2195 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
2196 <dd>Sign extend a constant to another type. The bit size of CST must be
2197 smaller or equal to the bit size of TYPE. Both types must be
2200 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
2201 <dd>Truncate a floating point constant to another floating point type. The
2202 size of CST must be larger than the size of TYPE. Both types must be
2203 floating point.</dd>
2205 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
2206 <dd>Floating point extend a constant to another type. The size of CST must be
2207 smaller or equal to the size of TYPE. Both types must be floating
2210 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
2211 <dd>Convert a floating point constant to the corresponding unsigned integer
2212 constant. TYPE must be a scalar or vector integer type. CST must be of
2213 scalar or vector floating point type. Both CST and TYPE must be scalars,
2214 or vectors of the same number of elements. If the value won't fit in the
2215 integer type, the results are undefined.</dd>
2217 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
2218 <dd>Convert a floating point constant to the corresponding signed integer
2219 constant. TYPE must be a scalar or vector integer type. CST must be of
2220 scalar or vector floating point type. Both CST and TYPE must be scalars,
2221 or vectors of the same number of elements. If the value won't fit in the
2222 integer type, the results are undefined.</dd>
2224 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
2225 <dd>Convert an unsigned integer constant to the corresponding floating point
2226 constant. TYPE must be a scalar or vector floating point type. CST must be
2227 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2228 vectors of the same number of elements. If the value won't fit in the
2229 floating point type, the results are undefined.</dd>
2231 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
2232 <dd>Convert a signed integer constant to the corresponding floating point
2233 constant. TYPE must be a scalar or vector floating point type. CST must be
2234 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2235 vectors of the same number of elements. If the value won't fit in the
2236 floating point type, the results are undefined.</dd>
2238 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
2239 <dd>Convert a pointer typed constant to the corresponding integer constant
2240 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2241 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2242 make it fit in <tt>TYPE</tt>.</dd>
2244 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
2245 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2246 type. CST must be of integer type. The CST value is zero extended,
2247 truncated, or unchanged to make it fit in a pointer size. This one is
2248 <i>really</i> dangerous!</dd>
2250 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
2251 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2252 are the same as those for the <a href="#i_bitcast">bitcast
2253 instruction</a>.</dd>
2255 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
2256 <dt><b><tt>getelementptr inbounds ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
2257 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2258 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2259 instruction, the index list may have zero or more indexes, which are
2260 required to make sense for the type of "CSTPTR".</dd>
2262 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
2263 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
2265 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
2266 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2268 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
2269 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2271 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
2272 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2275 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
2276 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2279 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
2280 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2283 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
2284 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2285 be any of the <a href="#binaryops">binary</a>
2286 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2287 on operands are the same as those for the corresponding instruction
2288 (e.g. no bitwise operations on floating point values are allowed).</dd>
2293 <!-- ======================================================================= -->
2294 <div class="doc_subsection"><a name="metadata">Embedded Metadata</a>
2297 <div class="doc_text">
2299 <p>Embedded metadata provides a way to attach arbitrary data to the instruction
2300 stream without affecting the behaviour of the program. There are two
2301 metadata primitives, strings and nodes. All metadata has the
2302 <tt>metadata</tt> type and is identified in syntax by a preceding exclamation
2303 point ('<tt>!</tt>').</p>
2305 <p>A metadata string is a string surrounded by double quotes. It can contain
2306 any character by escaping non-printable characters with "\xx" where "xx" is
2307 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
2309 <p>Metadata nodes are represented with notation similar to structure constants
2310 (a comma separated list of elements, surrounded by braces and preceded by an
2311 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2314 <p>A metadata node will attempt to track changes to the values it holds. In the
2315 event that a value is deleted, it will be replaced with a typeless
2316 "<tt>null</tt>", such as "<tt>metadata !{null, i32 10}</tt>".</p>
2318 <p>Optimizations may rely on metadata to provide additional information about
2319 the program that isn't available in the instructions, or that isn't easily
2320 computable. Similarly, the code generator may expect a certain metadata
2321 format to be used to express debugging information.</p>
2325 <!-- *********************************************************************** -->
2326 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2327 <!-- *********************************************************************** -->
2329 <!-- ======================================================================= -->
2330 <div class="doc_subsection">
2331 <a name="inlineasm">Inline Assembler Expressions</a>
2334 <div class="doc_text">
2336 <p>LLVM supports inline assembler expressions (as opposed
2337 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2338 a special value. This value represents the inline assembler as a string
2339 (containing the instructions to emit), a list of operand constraints (stored
2340 as a string), a flag that indicates whether or not the inline asm
2341 expression has side effects, and a flag indicating whether the function
2342 containing the asm needs to align its stack conservatively. An example
2343 inline assembler expression is:</p>
2345 <div class="doc_code">
2347 i32 (i32) asm "bswap $0", "=r,r"
2351 <p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2352 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2355 <div class="doc_code">
2357 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2361 <p>Inline asms with side effects not visible in the constraint list must be
2362 marked as having side effects. This is done through the use of the
2363 '<tt>sideeffect</tt>' keyword, like so:</p>
2365 <div class="doc_code">
2367 call void asm sideeffect "eieio", ""()
2371 <p>In some cases inline asms will contain code that will not work unless the
2372 stack is aligned in some way, such as calls or SSE instructions on x86,
2373 yet will not contain code that does that alignment within the asm.
2374 The compiler should make conservative assumptions about what the asm might
2375 contain and should generate its usual stack alignment code in the prologue
2376 if the '<tt>alignstack</tt>' keyword is present:</p>
2378 <div class="doc_code">
2380 call void asm alignstack "eieio", ""()
2384 <p>If both keywords appear the '<tt>sideeffect</tt>' keyword must come
2387 <p>TODO: The format of the asm and constraints string still need to be
2388 documented here. Constraints on what can be done (e.g. duplication, moving,
2389 etc need to be documented). This is probably best done by reference to
2390 another document that covers inline asm from a holistic perspective.</p>
2395 <!-- *********************************************************************** -->
2396 <div class="doc_section">
2397 <a name="intrinsic_globals">Intrinsic Global Variables</a>
2399 <!-- *********************************************************************** -->
2401 <p>LLVM has a number of "magic" global variables that contain data that affect
2402 code generation or other IR semantics. These are documented here. All globals
2403 of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2404 section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2407 <!-- ======================================================================= -->
2408 <div class="doc_subsection">
2409 <a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
2412 <div class="doc_text">
2414 <p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2415 href="#linkage_appending">appending linkage</a>. This array contains a list of
2416 pointers to global variables and functions which may optionally have a pointer
2417 cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2423 @llvm.used = appending global [2 x i8*] [
2425 i8* bitcast (i32* @Y to i8*)
2426 ], section "llvm.metadata"
2429 <p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2430 compiler, assembler, and linker are required to treat the symbol as if there is
2431 a reference to the global that it cannot see. For example, if a variable has
2432 internal linkage and no references other than that from the <tt>@llvm.used</tt>
2433 list, it cannot be deleted. This is commonly used to represent references from
2434 inline asms and other things the compiler cannot "see", and corresponds to
2435 "attribute((used))" in GNU C.</p>
2437 <p>On some targets, the code generator must emit a directive to the assembler or
2438 object file to prevent the assembler and linker from molesting the symbol.</p>
2442 <!-- ======================================================================= -->
2443 <div class="doc_subsection">
2444 <a name="intg_compiler_used">The '<tt>llvm.compiler.used</tt>' Global Variable</a>
2447 <div class="doc_text">
2449 <p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2450 <tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2451 touching the symbol. On targets that support it, this allows an intelligent
2452 linker to optimize references to the symbol without being impeded as it would be
2453 by <tt>@llvm.used</tt>.</p>
2455 <p>This is a rare construct that should only be used in rare circumstances, and
2456 should not be exposed to source languages.</p>
2460 <!-- ======================================================================= -->
2461 <div class="doc_subsection">
2462 <a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
2465 <div class="doc_text">
2467 <p>TODO: Describe this.</p>
2471 <!-- ======================================================================= -->
2472 <div class="doc_subsection">
2473 <a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
2476 <div class="doc_text">
2478 <p>TODO: Describe this.</p>
2483 <!-- *********************************************************************** -->
2484 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2485 <!-- *********************************************************************** -->
2487 <div class="doc_text">
2489 <p>The LLVM instruction set consists of several different classifications of
2490 instructions: <a href="#terminators">terminator
2491 instructions</a>, <a href="#binaryops">binary instructions</a>,
2492 <a href="#bitwiseops">bitwise binary instructions</a>,
2493 <a href="#memoryops">memory instructions</a>, and
2494 <a href="#otherops">other instructions</a>.</p>
2498 <!-- ======================================================================= -->
2499 <div class="doc_subsection"> <a name="terminators">Terminator
2500 Instructions</a> </div>
2502 <div class="doc_text">
2504 <p>As mentioned <a href="#functionstructure">previously</a>, every basic block
2505 in a program ends with a "Terminator" instruction, which indicates which
2506 block should be executed after the current block is finished. These
2507 terminator instructions typically yield a '<tt>void</tt>' value: they produce
2508 control flow, not values (the one exception being the
2509 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2511 <p>There are six different terminator instructions: the
2512 '<a href="#i_ret"><tt>ret</tt></a>' instruction, the
2513 '<a href="#i_br"><tt>br</tt></a>' instruction, the
2514 '<a href="#i_switch"><tt>switch</tt></a>' instruction, the
2515 '<a href="#i_indbr">'<tt>indbr</tt>' Instruction, the
2516 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the
2517 '<a href="#i_unwind"><tt>unwind</tt></a>' instruction, and the
2518 '<a href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2522 <!-- _______________________________________________________________________ -->
2523 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2524 Instruction</a> </div>
2526 <div class="doc_text">
2530 ret <type> <value> <i>; Return a value from a non-void function</i>
2531 ret void <i>; Return from void function</i>
2535 <p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
2536 a value) from a function back to the caller.</p>
2538 <p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
2539 value and then causes control flow, and one that just causes control flow to
2543 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
2544 return value. The type of the return value must be a
2545 '<a href="#t_firstclass">first class</a>' type.</p>
2547 <p>A function is not <a href="#wellformed">well formed</a> if it it has a
2548 non-void return type and contains a '<tt>ret</tt>' instruction with no return
2549 value or a return value with a type that does not match its type, or if it
2550 has a void return type and contains a '<tt>ret</tt>' instruction with a
2554 <p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
2555 the calling function's context. If the caller is a
2556 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
2557 instruction after the call. If the caller was an
2558 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
2559 the beginning of the "normal" destination block. If the instruction returns
2560 a value, that value shall set the call or invoke instruction's return
2565 ret i32 5 <i>; Return an integer value of 5</i>
2566 ret void <i>; Return from a void function</i>
2567 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
2570 <p>Note that the code generator does not yet fully support large
2571 return values. The specific sizes that are currently supported are
2572 dependent on the target. For integers, on 32-bit targets the limit
2573 is often 64 bits, and on 64-bit targets the limit is often 128 bits.
2574 For aggregate types, the current limits are dependent on the element
2575 types; for example targets are often limited to 2 total integer
2576 elements and 2 total floating-point elements.</p>
2579 <!-- _______________________________________________________________________ -->
2580 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2582 <div class="doc_text">
2586 br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2590 <p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
2591 different basic block in the current function. There are two forms of this
2592 instruction, corresponding to a conditional branch and an unconditional
2596 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
2597 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
2598 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
2602 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2603 argument is evaluated. If the value is <tt>true</tt>, control flows to the
2604 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2605 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2610 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
2611 br i1 %cond, label %IfEqual, label %IfUnequal
2613 <a href="#i_ret">ret</a> i32 1
2615 <a href="#i_ret">ret</a> i32 0
2620 <!-- _______________________________________________________________________ -->
2621 <div class="doc_subsubsection">
2622 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2625 <div class="doc_text">
2629 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2633 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2634 several different places. It is a generalization of the '<tt>br</tt>'
2635 instruction, allowing a branch to occur to one of many possible
2639 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2640 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
2641 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
2642 The table is not allowed to contain duplicate constant entries.</p>
2645 <p>The <tt>switch</tt> instruction specifies a table of values and
2646 destinations. When the '<tt>switch</tt>' instruction is executed, this table
2647 is searched for the given value. If the value is found, control flow is
2648 transferred to the corresponding destination; otherwise, control flow is
2649 transferred to the default destination.</p>
2651 <h5>Implementation:</h5>
2652 <p>Depending on properties of the target machine and the particular
2653 <tt>switch</tt> instruction, this instruction may be code generated in
2654 different ways. For example, it could be generated as a series of chained
2655 conditional branches or with a lookup table.</p>
2659 <i>; Emulate a conditional br instruction</i>
2660 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2661 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
2663 <i>; Emulate an unconditional br instruction</i>
2664 switch i32 0, label %dest [ ]
2666 <i>; Implement a jump table:</i>
2667 switch i32 %val, label %otherwise [ i32 0, label %onzero
2669 i32 2, label %ontwo ]
2675 <!-- _______________________________________________________________________ -->
2676 <div class="doc_subsubsection">
2677 <a name="i_indbr">'<tt>indbr</tt>' Instruction</a>
2680 <div class="doc_text">
2684 indbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
2689 <p>The '<tt>indbr</tt>' instruction implements an indirect branch to a label
2690 within the current function, whose address is specified by
2691 "<tt>address</tt>".</p>
2695 <p>The '<tt>address</tt>' argument is the address of the label to jump to. The
2696 rest of the arguments indicate the full set of possible destinations that the
2697 address may point to. Blocks are allowed to occur multiple times in the
2698 destination list, though this isn't particularly useful.</p>
2700 <p>This destination list is required so that dataflow analysis has an accurate
2701 understanding of the CFG.</p>
2705 <p>Control transfers to the block specified in the address argument. All
2706 possible destination blocks must be listed in the label list, otherwise this
2707 instruction has undefined behavior. This implies that jumps to labels
2708 defined in other functions have undefined behavior as well.</p>
2710 <h5>Implementation:</h5>
2712 <p>This is typically implemented with a jump through a register.</p>
2716 indbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
2722 <!-- _______________________________________________________________________ -->
2723 <div class="doc_subsubsection">
2724 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2727 <div class="doc_text">
2731 <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>]
2732 to label <normal label> unwind label <exception label>
2736 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2737 function, with the possibility of control flow transfer to either the
2738 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
2739 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
2740 control flow will return to the "normal" label. If the callee (or any
2741 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
2742 instruction, control is interrupted and continued at the dynamically nearest
2743 "exception" label.</p>
2746 <p>This instruction requires several arguments:</p>
2749 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
2750 convention</a> the call should use. If none is specified, the call
2751 defaults to using C calling conventions.</li>
2753 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2754 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
2755 '<tt>inreg</tt>' attributes are valid here.</li>
2757 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2758 function value being invoked. In most cases, this is a direct function
2759 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
2760 off an arbitrary pointer to function value.</li>
2762 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2763 function to be invoked. </li>
2765 <li>'<tt>function args</tt>': argument list whose types match the function
2766 signature argument types. If the function signature indicates the
2767 function accepts a variable number of arguments, the extra arguments can
2770 <li>'<tt>normal label</tt>': the label reached when the called function
2771 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2773 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2774 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2776 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
2777 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2778 '<tt>readnone</tt>' attributes are valid here.</li>
2782 <p>This instruction is designed to operate as a standard
2783 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
2784 primary difference is that it establishes an association with a label, which
2785 is used by the runtime library to unwind the stack.</p>
2787 <p>This instruction is used in languages with destructors to ensure that proper
2788 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2789 exception. Additionally, this is important for implementation of
2790 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
2792 <p>For the purposes of the SSA form, the definition of the value returned by the
2793 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
2794 block to the "normal" label. If the callee unwinds then no return value is
2799 %retval = invoke i32 @Test(i32 15) to label %Continue
2800 unwind label %TestCleanup <i>; {i32}:retval set</i>
2801 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
2802 unwind label %TestCleanup <i>; {i32}:retval set</i>
2807 <!-- _______________________________________________________________________ -->
2809 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2810 Instruction</a> </div>
2812 <div class="doc_text">
2820 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2821 at the first callee in the dynamic call stack which used
2822 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
2823 This is primarily used to implement exception handling.</p>
2826 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
2827 immediately halt. The dynamic call stack is then searched for the
2828 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
2829 Once found, execution continues at the "exceptional" destination block
2830 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
2831 instruction in the dynamic call chain, undefined behavior results.</p>
2835 <!-- _______________________________________________________________________ -->
2837 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2838 Instruction</a> </div>
2840 <div class="doc_text">
2848 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2849 instruction is used to inform the optimizer that a particular portion of the
2850 code is not reachable. This can be used to indicate that the code after a
2851 no-return function cannot be reached, and other facts.</p>
2854 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2858 <!-- ======================================================================= -->
2859 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2861 <div class="doc_text">
2863 <p>Binary operators are used to do most of the computation in a program. They
2864 require two operands of the same type, execute an operation on them, and
2865 produce a single value. The operands might represent multiple data, as is
2866 the case with the <a href="#t_vector">vector</a> data type. The result value
2867 has the same type as its operands.</p>
2869 <p>There are several different binary operators:</p>
2873 <!-- _______________________________________________________________________ -->
2874 <div class="doc_subsubsection">
2875 <a name="i_add">'<tt>add</tt>' Instruction</a>
2878 <div class="doc_text">
2882 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2883 <result> = add nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2884 <result> = add nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2885 <result> = add nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2889 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2892 <p>The two arguments to the '<tt>add</tt>' instruction must
2893 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
2894 integer values. Both arguments must have identical types.</p>
2897 <p>The value produced is the integer sum of the two operands.</p>
2899 <p>If the sum has unsigned overflow, the result returned is the mathematical
2900 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
2902 <p>Because LLVM integers use a two's complement representation, this instruction
2903 is appropriate for both signed and unsigned integers.</p>
2905 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
2906 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
2907 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
2908 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
2912 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2917 <!-- _______________________________________________________________________ -->
2918 <div class="doc_subsubsection">
2919 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
2922 <div class="doc_text">
2926 <result> = fadd <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2930 <p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
2933 <p>The two arguments to the '<tt>fadd</tt>' instruction must be
2934 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2935 floating point values. Both arguments must have identical types.</p>
2938 <p>The value produced is the floating point sum of the two operands.</p>
2942 <result> = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
2947 <!-- _______________________________________________________________________ -->
2948 <div class="doc_subsubsection">
2949 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2952 <div class="doc_text">
2956 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2957 <result> = sub nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2958 <result> = sub nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2959 <result> = sub nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2963 <p>The '<tt>sub</tt>' instruction returns the difference of its two
2966 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
2967 '<tt>neg</tt>' instruction present in most other intermediate
2968 representations.</p>
2971 <p>The two arguments to the '<tt>sub</tt>' instruction must
2972 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
2973 integer values. Both arguments must have identical types.</p>
2976 <p>The value produced is the integer difference of the two operands.</p>
2978 <p>If the difference has unsigned overflow, the result returned is the
2979 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
2982 <p>Because LLVM integers use a two's complement representation, this instruction
2983 is appropriate for both signed and unsigned integers.</p>
2985 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
2986 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
2987 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
2988 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
2992 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2993 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2998 <!-- _______________________________________________________________________ -->
2999 <div class="doc_subsubsection">
3000 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
3003 <div class="doc_text">
3007 <result> = fsub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3011 <p>The '<tt>fsub</tt>' instruction returns the difference of its two
3014 <p>Note that the '<tt>fsub</tt>' instruction is used to represent the
3015 '<tt>fneg</tt>' instruction present in most other intermediate
3016 representations.</p>
3019 <p>The two arguments to the '<tt>fsub</tt>' instruction must be
3020 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3021 floating point values. Both arguments must have identical types.</p>
3024 <p>The value produced is the floating point difference of the two operands.</p>
3028 <result> = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
3029 <result> = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
3034 <!-- _______________________________________________________________________ -->
3035 <div class="doc_subsubsection">
3036 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
3039 <div class="doc_text">
3043 <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3044 <result> = mul nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3045 <result> = mul nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3046 <result> = mul nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3050 <p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
3053 <p>The two arguments to the '<tt>mul</tt>' instruction must
3054 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3055 integer values. Both arguments must have identical types.</p>
3058 <p>The value produced is the integer product of the two operands.</p>
3060 <p>If the result of the multiplication has unsigned overflow, the result
3061 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
3062 width of the result.</p>
3064 <p>Because LLVM integers use a two's complement representation, and the result
3065 is the same width as the operands, this instruction returns the correct
3066 result for both signed and unsigned integers. If a full product
3067 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
3068 be sign-extended or zero-extended as appropriate to the width of the full
3071 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3072 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3073 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
3074 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
3078 <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
3083 <!-- _______________________________________________________________________ -->
3084 <div class="doc_subsubsection">
3085 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
3088 <div class="doc_text">
3092 <result> = fmul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3096 <p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
3099 <p>The two arguments to the '<tt>fmul</tt>' instruction must be
3100 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3101 floating point values. Both arguments must have identical types.</p>
3104 <p>The value produced is the floating point product of the two operands.</p>
3108 <result> = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
3113 <!-- _______________________________________________________________________ -->
3114 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
3117 <div class="doc_text">
3121 <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3125 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
3128 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
3129 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3130 values. Both arguments must have identical types.</p>
3133 <p>The value produced is the unsigned integer quotient of the two operands.</p>
3135 <p>Note that unsigned integer division and signed integer division are distinct
3136 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
3138 <p>Division by zero leads to undefined behavior.</p>
3142 <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3147 <!-- _______________________________________________________________________ -->
3148 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
3151 <div class="doc_text">
3155 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3156 <result> = sdiv exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3160 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
3163 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
3164 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3165 values. Both arguments must have identical types.</p>
3168 <p>The value produced is the signed integer quotient of the two operands rounded
3171 <p>Note that signed integer division and unsigned integer division are distinct
3172 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
3174 <p>Division by zero leads to undefined behavior. Overflow also leads to
3175 undefined behavior; this is a rare case, but can occur, for example, by doing
3176 a 32-bit division of -2147483648 by -1.</p>
3178 <p>If the <tt>exact</tt> keyword is present, the result value of the
3179 <tt>sdiv</tt> is undefined if the result would be rounded or if overflow
3184 <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3189 <!-- _______________________________________________________________________ -->
3190 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
3191 Instruction</a> </div>
3193 <div class="doc_text">
3197 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3201 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
3204 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
3205 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3206 floating point values. Both arguments must have identical types.</p>
3209 <p>The value produced is the floating point quotient of the two operands.</p>
3213 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
3218 <!-- _______________________________________________________________________ -->
3219 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
3222 <div class="doc_text">
3226 <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3230 <p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3231 division of its two arguments.</p>
3234 <p>The two arguments to the '<tt>urem</tt>' instruction must be
3235 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3236 values. Both arguments must have identical types.</p>
3239 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
3240 This instruction always performs an unsigned division to get the
3243 <p>Note that unsigned integer remainder and signed integer remainder are
3244 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3246 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
3250 <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3255 <!-- _______________________________________________________________________ -->
3256 <div class="doc_subsubsection">
3257 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
3260 <div class="doc_text">
3264 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3268 <p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3269 division of its two operands. This instruction can also take
3270 <a href="#t_vector">vector</a> versions of the values in which case the
3271 elements must be integers.</p>
3274 <p>The two arguments to the '<tt>srem</tt>' instruction must be
3275 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3276 values. Both arguments must have identical types.</p>
3279 <p>This instruction returns the <i>remainder</i> of a division (where the result
3280 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
3281 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
3282 a value. For more information about the difference,
3283 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3284 Math Forum</a>. For a table of how this is implemented in various languages,
3285 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3286 Wikipedia: modulo operation</a>.</p>
3288 <p>Note that signed integer remainder and unsigned integer remainder are
3289 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3291 <p>Taking the remainder of a division by zero leads to undefined behavior.
3292 Overflow also leads to undefined behavior; this is a rare case, but can
3293 occur, for example, by taking the remainder of a 32-bit division of
3294 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3295 lets srem be implemented using instructions that return both the result of
3296 the division and the remainder.)</p>
3300 <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3305 <!-- _______________________________________________________________________ -->
3306 <div class="doc_subsubsection">
3307 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
3309 <div class="doc_text">
3313 <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3317 <p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3318 its two operands.</p>
3321 <p>The two arguments to the '<tt>frem</tt>' instruction must be
3322 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3323 floating point values. Both arguments must have identical types.</p>
3326 <p>This instruction returns the <i>remainder</i> of a division. The remainder
3327 has the same sign as the dividend.</p>
3331 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
3336 <!-- ======================================================================= -->
3337 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
3338 Operations</a> </div>
3340 <div class="doc_text">
3342 <p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3343 program. They are generally very efficient instructions and can commonly be
3344 strength reduced from other instructions. They require two operands of the
3345 same type, execute an operation on them, and produce a single value. The
3346 resulting value is the same type as its operands.</p>
3350 <!-- _______________________________________________________________________ -->
3351 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
3352 Instruction</a> </div>
3354 <div class="doc_text">
3358 <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3362 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3363 a specified number of bits.</p>
3366 <p>Both arguments to the '<tt>shl</tt>' instruction must be the
3367 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3368 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3371 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3372 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3373 is (statically or dynamically) negative or equal to or larger than the number
3374 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3375 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3376 shift amount in <tt>op2</tt>.</p>
3380 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
3381 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
3382 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
3383 <result> = shl i32 1, 32 <i>; undefined</i>
3384 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
3389 <!-- _______________________________________________________________________ -->
3390 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
3391 Instruction</a> </div>
3393 <div class="doc_text">
3397 <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3401 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
3402 operand shifted to the right a specified number of bits with zero fill.</p>
3405 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
3406 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3407 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3410 <p>This instruction always performs a logical shift right operation. The most
3411 significant bits of the result will be filled with zero bits after the shift.
3412 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
3413 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3414 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3415 shift amount in <tt>op2</tt>.</p>
3419 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
3420 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
3421 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
3422 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
3423 <result> = lshr i32 1, 32 <i>; undefined</i>
3424 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
3429 <!-- _______________________________________________________________________ -->
3430 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
3431 Instruction</a> </div>
3432 <div class="doc_text">
3436 <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3440 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3441 operand shifted to the right a specified number of bits with sign
3445 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
3446 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3447 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3450 <p>This instruction always performs an arithmetic shift right operation, The
3451 most significant bits of the result will be filled with the sign bit
3452 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3453 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
3454 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
3455 the corresponding shift amount in <tt>op2</tt>.</p>
3459 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3460 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3461 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3462 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
3463 <result> = ashr i32 1, 32 <i>; undefined</i>
3464 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
3469 <!-- _______________________________________________________________________ -->
3470 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
3471 Instruction</a> </div>
3473 <div class="doc_text">
3477 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3481 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
3485 <p>The two arguments to the '<tt>and</tt>' instruction must be
3486 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3487 values. Both arguments must have identical types.</p>
3490 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
3492 <table border="1" cellspacing="0" cellpadding="4">
3524 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
3525 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
3526 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
3529 <!-- _______________________________________________________________________ -->
3530 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
3532 <div class="doc_text">
3536 <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3540 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
3544 <p>The two arguments to the '<tt>or</tt>' instruction must be
3545 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3546 values. Both arguments must have identical types.</p>
3549 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
3551 <table border="1" cellspacing="0" cellpadding="4">
3583 <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
3584 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
3585 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
3590 <!-- _______________________________________________________________________ -->
3591 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3592 Instruction</a> </div>
3594 <div class="doc_text">
3598 <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3602 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
3603 its two operands. The <tt>xor</tt> is used to implement the "one's
3604 complement" operation, which is the "~" operator in C.</p>
3607 <p>The two arguments to the '<tt>xor</tt>' instruction must be
3608 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3609 values. Both arguments must have identical types.</p>
3612 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
3614 <table border="1" cellspacing="0" cellpadding="4">
3646 <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
3647 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3648 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3649 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
3654 <!-- ======================================================================= -->
3655 <div class="doc_subsection">
3656 <a name="vectorops">Vector Operations</a>
3659 <div class="doc_text">
3661 <p>LLVM supports several instructions to represent vector operations in a
3662 target-independent manner. These instructions cover the element-access and
3663 vector-specific operations needed to process vectors effectively. While LLVM
3664 does directly support these vector operations, many sophisticated algorithms
3665 will want to use target-specific intrinsics to take full advantage of a
3666 specific target.</p>
3670 <!-- _______________________________________________________________________ -->
3671 <div class="doc_subsubsection">
3672 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3675 <div class="doc_text">
3679 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
3683 <p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
3684 from a vector at a specified index.</p>
3688 <p>The first operand of an '<tt>extractelement</tt>' instruction is a value
3689 of <a href="#t_vector">vector</a> type. The second operand is an index
3690 indicating the position from which to extract the element. The index may be
3694 <p>The result is a scalar of the same type as the element type of
3695 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3696 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3697 results are undefined.</p>
3701 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
3706 <!-- _______________________________________________________________________ -->
3707 <div class="doc_subsubsection">
3708 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
3711 <div class="doc_text">
3715 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
3719 <p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
3720 vector at a specified index.</p>
3723 <p>The first operand of an '<tt>insertelement</tt>' instruction is a value
3724 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
3725 whose type must equal the element type of the first operand. The third
3726 operand is an index indicating the position at which to insert the value.
3727 The index may be a variable.</p>
3730 <p>The result is a vector of the same type as <tt>val</tt>. Its element values
3731 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
3732 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3733 results are undefined.</p>
3737 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
3742 <!-- _______________________________________________________________________ -->
3743 <div class="doc_subsubsection">
3744 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3747 <div class="doc_text">
3751 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
3755 <p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
3756 from two input vectors, returning a vector with the same element type as the
3757 input and length that is the same as the shuffle mask.</p>
3760 <p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3761 with types that match each other. The third argument is a shuffle mask whose
3762 element type is always 'i32'. The result of the instruction is a vector
3763 whose length is the same as the shuffle mask and whose element type is the
3764 same as the element type of the first two operands.</p>
3766 <p>The shuffle mask operand is required to be a constant vector with either
3767 constant integer or undef values.</p>
3770 <p>The elements of the two input vectors are numbered from left to right across
3771 both of the vectors. The shuffle mask operand specifies, for each element of
3772 the result vector, which element of the two input vectors the result element
3773 gets. The element selector may be undef (meaning "don't care") and the
3774 second operand may be undef if performing a shuffle from only one vector.</p>
3778 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3779 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
3780 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
3781 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
3782 %result = shufflevector <8 x i32> %v1, <8 x i32> undef,
3783 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
3784 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3785 <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>
3790 <!-- ======================================================================= -->
3791 <div class="doc_subsection">
3792 <a name="aggregateops">Aggregate Operations</a>
3795 <div class="doc_text">
3797 <p>LLVM supports several instructions for working with aggregate values.</p>
3801 <!-- _______________________________________________________________________ -->
3802 <div class="doc_subsubsection">
3803 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3806 <div class="doc_text">
3810 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
3814 <p>The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3815 or array element from an aggregate value.</p>
3818 <p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
3819 of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type. The
3820 operands are constant indices to specify which value to extract in a similar
3821 manner as indices in a
3822 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
3825 <p>The result is the value at the position in the aggregate specified by the
3830 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
3835 <!-- _______________________________________________________________________ -->
3836 <div class="doc_subsubsection">
3837 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3840 <div class="doc_text">
3844 <result> = insertvalue <aggregate type> <val>, <ty> <val>, <idx> <i>; yields <n x <ty>></i>
3848 <p>The '<tt>insertvalue</tt>' instruction inserts a value into a struct field or
3849 array element in an aggregate.</p>
3853 <p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
3854 of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type. The
3855 second operand is a first-class value to insert. The following operands are
3856 constant indices indicating the position at which to insert the value in a
3857 similar manner as indices in a
3858 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction. The
3859 value to insert must have the same type as the value identified by the
3863 <p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
3864 that of <tt>val</tt> except that the value at the position specified by the
3865 indices is that of <tt>elt</tt>.</p>
3869 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
3875 <!-- ======================================================================= -->
3876 <div class="doc_subsection">
3877 <a name="memoryops">Memory Access and Addressing Operations</a>
3880 <div class="doc_text">
3882 <p>A key design point of an SSA-based representation is how it represents
3883 memory. In LLVM, no memory locations are in SSA form, which makes things
3884 very simple. This section describes how to read, write, and allocate
3889 <!-- _______________________________________________________________________ -->
3890 <div class="doc_subsubsection">
3891 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3894 <div class="doc_text">
3898 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3902 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3903 currently executing function, to be automatically released when this function
3904 returns to its caller. The object is always allocated in the generic address
3905 space (address space zero).</p>
3908 <p>The '<tt>alloca</tt>' instruction
3909 allocates <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the
3910 runtime stack, returning a pointer of the appropriate type to the program.
3911 If "NumElements" is specified, it is the number of elements allocated,
3912 otherwise "NumElements" is defaulted to be one. If a constant alignment is
3913 specified, the value result of the allocation is guaranteed to be aligned to
3914 at least that boundary. If not specified, or if zero, the target can choose
3915 to align the allocation on any convenient boundary compatible with the
3918 <p>'<tt>type</tt>' may be any sized type.</p>
3921 <p>Memory is allocated; a pointer is returned. The operation is undefined if
3922 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
3923 memory is automatically released when the function returns. The
3924 '<tt>alloca</tt>' instruction is commonly used to represent automatic
3925 variables that must have an address available. When the function returns
3926 (either with the <tt><a href="#i_ret">ret</a></tt>
3927 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
3928 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
3932 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3933 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3934 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3935 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3940 <!-- _______________________________________________________________________ -->
3941 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3942 Instruction</a> </div>
3944 <div class="doc_text">
3948 <result> = load <ty>* <pointer>[, align <alignment>]
3949 <result> = volatile load <ty>* <pointer>[, align <alignment>]
3953 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3956 <p>The argument to the '<tt>load</tt>' instruction specifies the memory address
3957 from which to load. The pointer must point to
3958 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3959 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
3960 number or order of execution of this <tt>load</tt> with other
3961 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3964 <p>The optional constant "align" argument specifies the alignment of the
3965 operation (that is, the alignment of the memory address). A value of 0 or an
3966 omitted "align" argument means that the operation has the preferential
3967 alignment for the target. It is the responsibility of the code emitter to
3968 ensure that the alignment information is correct. Overestimating the
3969 alignment results in an undefined behavior. Underestimating the alignment may
3970 produce less efficient code. An alignment of 1 is always safe.</p>
3973 <p>The location of memory pointed to is loaded. If the value being loaded is of
3974 scalar type then the number of bytes read does not exceed the minimum number
3975 of bytes needed to hold all bits of the type. For example, loading an
3976 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
3977 <tt>i20</tt> with a size that is not an integral number of bytes, the result
3978 is undefined if the value was not originally written using a store of the
3983 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3984 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3985 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3990 <!-- _______________________________________________________________________ -->
3991 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3992 Instruction</a> </div>
3994 <div class="doc_text">
3998 store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3999 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
4003 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
4006 <p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
4007 and an address at which to store it. The type of the
4008 '<tt><pointer></tt>' operand must be a pointer to
4009 the <a href="#t_firstclass">first class</a> type of the
4010 '<tt><value></tt>' operand. If the <tt>store</tt> is marked
4011 as <tt>volatile</tt>, then the optimizer is not allowed to modify the number
4012 or order of execution of this <tt>store</tt> with other
4013 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
4016 <p>The optional constant "align" argument specifies the alignment of the
4017 operation (that is, the alignment of the memory address). A value of 0 or an
4018 omitted "align" argument means that the operation has the preferential
4019 alignment for the target. It is the responsibility of the code emitter to
4020 ensure that the alignment information is correct. Overestimating the
4021 alignment results in an undefined behavior. Underestimating the alignment may
4022 produce less efficient code. An alignment of 1 is always safe.</p>
4025 <p>The contents of memory are updated to contain '<tt><value></tt>' at the
4026 location specified by the '<tt><pointer></tt>' operand. If
4027 '<tt><value></tt>' is of scalar type then the number of bytes written
4028 does not exceed the minimum number of bytes needed to hold all bits of the
4029 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
4030 writing a value of a type like <tt>i20</tt> with a size that is not an
4031 integral number of bytes, it is unspecified what happens to the extra bits
4032 that do not belong to the type, but they will typically be overwritten.</p>
4036 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4037 store i32 3, i32* %ptr <i>; yields {void}</i>
4038 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
4043 <!-- _______________________________________________________________________ -->
4044 <div class="doc_subsubsection">
4045 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
4048 <div class="doc_text">
4052 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
4053 <result> = getelementptr inbounds <pty>* <ptrval>{, <ty> <idx>}*
4057 <p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
4058 subelement of an aggregate data structure. It performs address calculation
4059 only and does not access memory.</p>
4062 <p>The first argument is always a pointer, and forms the basis of the
4063 calculation. The remaining arguments are indices that indicate which of the
4064 elements of the aggregate object are indexed. The interpretation of each
4065 index is dependent on the type being indexed into. The first index always
4066 indexes the pointer value given as the first argument, the second index
4067 indexes a value of the type pointed to (not necessarily the value directly
4068 pointed to, since the first index can be non-zero), etc. The first type
4069 indexed into must be a pointer value, subsequent types can be arrays, vectors
4070 and structs. Note that subsequent types being indexed into can never be
4071 pointers, since that would require loading the pointer before continuing
4074 <p>The type of each index argument depends on the type it is indexing into.
4075 When indexing into a (optionally packed) structure, only <tt>i32</tt> integer
4076 <b>constants</b> are allowed. When indexing into an array, pointer or
4077 vector, integers of any width are allowed, and they are not required to be
4080 <p>For example, let's consider a C code fragment and how it gets compiled to
4083 <div class="doc_code">
4096 int *foo(struct ST *s) {
4097 return &s[1].Z.B[5][13];
4102 <p>The LLVM code generated by the GCC frontend is:</p>
4104 <div class="doc_code">
4106 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
4107 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
4109 define i32* @foo(%ST* %s) {
4111 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
4118 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
4119 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
4120 }</tt>' type, a structure. The second index indexes into the third element
4121 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
4122 i8 }</tt>' type, another structure. The third index indexes into the second
4123 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
4124 array. The two dimensions of the array are subscripted into, yielding an
4125 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
4126 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
4128 <p>Note that it is perfectly legal to index partially through a structure,
4129 returning a pointer to an inner element. Because of this, the LLVM code for
4130 the given testcase is equivalent to:</p>
4133 define i32* @foo(%ST* %s) {
4134 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
4135 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
4136 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
4137 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
4138 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
4143 <p>If the <tt>inbounds</tt> keyword is present, the result value of the
4144 <tt>getelementptr</tt> is undefined if the base pointer is not an
4145 <i>in bounds</i> address of an allocated object, or if any of the addresses
4146 that would be formed by successive addition of the offsets implied by the
4147 indices to the base address with infinitely precise arithmetic are not an
4148 <i>in bounds</i> address of that allocated object.
4149 The <i>in bounds</i> addresses for an allocated object are all the addresses
4150 that point into the object, plus the address one byte past the end.</p>
4152 <p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
4153 the base address with silently-wrapping two's complement arithmetic, and
4154 the result value of the <tt>getelementptr</tt> may be outside the object
4155 pointed to by the base pointer. The result value may not necessarily be
4156 used to access memory though, even if it happens to point into allocated
4157 storage. See the <a href="#pointeraliasing">Pointer Aliasing Rules</a>
4158 section for more information.</p>
4160 <p>The getelementptr instruction is often confusing. For some more insight into
4161 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
4165 <i>; yields [12 x i8]*:aptr</i>
4166 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
4167 <i>; yields i8*:vptr</i>
4168 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
4169 <i>; yields i8*:eptr</i>
4170 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
4171 <i>; yields i32*:iptr</i>
4172 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
4177 <!-- ======================================================================= -->
4178 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
4181 <div class="doc_text">
4183 <p>The instructions in this category are the conversion instructions (casting)
4184 which all take a single operand and a type. They perform various bit
4185 conversions on the operand.</p>
4189 <!-- _______________________________________________________________________ -->
4190 <div class="doc_subsubsection">
4191 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
4193 <div class="doc_text">
4197 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
4201 <p>The '<tt>trunc</tt>' instruction truncates its operand to the
4202 type <tt>ty2</tt>.</p>
4205 <p>The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
4206 be an <a href="#t_integer">integer</a> type, and a type that specifies the
4207 size and type of the result, which must be
4208 an <a href="#t_integer">integer</a> type. The bit size of <tt>value</tt> must
4209 be larger than the bit size of <tt>ty2</tt>. Equal sized types are not
4213 <p>The '<tt>trunc</tt>' instruction truncates the high order bits
4214 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
4215 source size must be larger than the destination size, <tt>trunc</tt> cannot
4216 be a <i>no-op cast</i>. It will always truncate bits.</p>
4220 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
4221 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
4222 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
4227 <!-- _______________________________________________________________________ -->
4228 <div class="doc_subsubsection">
4229 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
4231 <div class="doc_text">
4235 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
4239 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
4244 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
4245 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4246 also be of <a href="#t_integer">integer</a> type. The bit size of the
4247 <tt>value</tt> must be smaller than the bit size of the destination type,
4251 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
4252 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
4254 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
4258 %X = zext i32 257 to i64 <i>; yields i64:257</i>
4259 %Y = zext i1 true to i32 <i>; yields i32:1</i>
4264 <!-- _______________________________________________________________________ -->
4265 <div class="doc_subsubsection">
4266 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
4268 <div class="doc_text">
4272 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
4276 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
4279 <p>The '<tt>sext</tt>' instruction takes a value to cast, which must be of
4280 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4281 also be of <a href="#t_integer">integer</a> type. The bit size of the
4282 <tt>value</tt> must be smaller than the bit size of the destination type,
4286 <p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
4287 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
4288 of the type <tt>ty2</tt>.</p>
4290 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
4294 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
4295 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
4300 <!-- _______________________________________________________________________ -->
4301 <div class="doc_subsubsection">
4302 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
4305 <div class="doc_text">
4309 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
4313 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
4317 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
4318 point</a> value to cast and a <a href="#t_floating">floating point</a> type
4319 to cast it to. The size of <tt>value</tt> must be larger than the size of
4320 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
4321 <i>no-op cast</i>.</p>
4324 <p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
4325 <a href="#t_floating">floating point</a> type to a smaller
4326 <a href="#t_floating">floating point</a> type. If the value cannot fit
4327 within the destination type, <tt>ty2</tt>, then the results are
4332 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4333 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4338 <!-- _______________________________________________________________________ -->
4339 <div class="doc_subsubsection">
4340 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
4342 <div class="doc_text">
4346 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
4350 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
4351 floating point value.</p>
4354 <p>The '<tt>fpext</tt>' instruction takes a
4355 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
4356 a <a href="#t_floating">floating point</a> type to cast it to. The source
4357 type must be smaller than the destination type.</p>
4360 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
4361 <a href="#t_floating">floating point</a> type to a larger
4362 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4363 used to make a <i>no-op cast</i> because it always changes bits. Use
4364 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
4368 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
4369 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
4374 <!-- _______________________________________________________________________ -->
4375 <div class="doc_subsubsection">
4376 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
4378 <div class="doc_text">
4382 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
4386 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
4387 unsigned integer equivalent of type <tt>ty2</tt>.</p>
4390 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
4391 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4392 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4393 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4394 vector integer type with the same number of elements as <tt>ty</tt></p>
4397 <p>The '<tt>fptoui</tt>' instruction converts its
4398 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4399 towards zero) unsigned integer value. If the value cannot fit
4400 in <tt>ty2</tt>, the results are undefined.</p>
4404 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
4405 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
4406 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
4411 <!-- _______________________________________________________________________ -->
4412 <div class="doc_subsubsection">
4413 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
4415 <div class="doc_text">
4419 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
4423 <p>The '<tt>fptosi</tt>' instruction converts
4424 <a href="#t_floating">floating point</a> <tt>value</tt> to
4425 type <tt>ty2</tt>.</p>
4428 <p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
4429 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4430 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4431 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4432 vector integer type with the same number of elements as <tt>ty</tt></p>
4435 <p>The '<tt>fptosi</tt>' instruction converts its
4436 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4437 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4438 the results are undefined.</p>
4442 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
4443 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
4444 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
4449 <!-- _______________________________________________________________________ -->
4450 <div class="doc_subsubsection">
4451 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
4453 <div class="doc_text">
4457 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4461 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
4462 integer and converts that value to the <tt>ty2</tt> type.</p>
4465 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
4466 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4467 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4468 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4469 floating point type with the same number of elements as <tt>ty</tt></p>
4472 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
4473 integer quantity and converts it to the corresponding floating point
4474 value. If the value cannot fit in the floating point value, the results are
4479 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
4480 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
4485 <!-- _______________________________________________________________________ -->
4486 <div class="doc_subsubsection">
4487 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
4489 <div class="doc_text">
4493 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4497 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
4498 and converts that value to the <tt>ty2</tt> type.</p>
4501 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
4502 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4503 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4504 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4505 floating point type with the same number of elements as <tt>ty</tt></p>
4508 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
4509 quantity and converts it to the corresponding floating point value. If the
4510 value cannot fit in the floating point value, the results are undefined.</p>
4514 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
4515 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
4520 <!-- _______________________________________________________________________ -->
4521 <div class="doc_subsubsection">
4522 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4524 <div class="doc_text">
4528 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
4532 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4533 the integer type <tt>ty2</tt>.</p>
4536 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4537 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
4538 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
4541 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4542 <tt>ty2</tt> by interpreting the pointer value as an integer and either
4543 truncating or zero extending that value to the size of the integer type. If
4544 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4545 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4546 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4551 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4552 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4557 <!-- _______________________________________________________________________ -->
4558 <div class="doc_subsubsection">
4559 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4561 <div class="doc_text">
4565 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
4569 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
4570 pointer type, <tt>ty2</tt>.</p>
4573 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4574 value to cast, and a type to cast it to, which must be a
4575 <a href="#t_pointer">pointer</a> type.</p>
4578 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4579 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
4580 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4581 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
4582 than the size of a pointer then a zero extension is done. If they are the
4583 same size, nothing is done (<i>no-op cast</i>).</p>
4587 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4588 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4589 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4594 <!-- _______________________________________________________________________ -->
4595 <div class="doc_subsubsection">
4596 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4598 <div class="doc_text">
4602 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
4606 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4607 <tt>ty2</tt> without changing any bits.</p>
4610 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
4611 non-aggregate first class value, and a type to cast it to, which must also be
4612 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
4613 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
4614 identical. If the source type is a pointer, the destination type must also be
4615 a pointer. This instruction supports bitwise conversion of vectors to
4616 integers and to vectors of other types (as long as they have the same
4620 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4621 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4622 this conversion. The conversion is done as if the <tt>value</tt> had been
4623 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
4624 be converted to other pointer types with this instruction. To convert
4625 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
4626 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4630 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4631 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
4632 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
4637 <!-- ======================================================================= -->
4638 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4640 <div class="doc_text">
4642 <p>The instructions in this category are the "miscellaneous" instructions, which
4643 defy better classification.</p>
4647 <!-- _______________________________________________________________________ -->
4648 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4651 <div class="doc_text">
4655 <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4659 <p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
4660 boolean values based on comparison of its two integer, integer vector, or
4661 pointer operands.</p>
4664 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
4665 the condition code indicating the kind of comparison to perform. It is not a
4666 value, just a keyword. The possible condition code are:</p>
4669 <li><tt>eq</tt>: equal</li>
4670 <li><tt>ne</tt>: not equal </li>
4671 <li><tt>ugt</tt>: unsigned greater than</li>
4672 <li><tt>uge</tt>: unsigned greater or equal</li>
4673 <li><tt>ult</tt>: unsigned less than</li>
4674 <li><tt>ule</tt>: unsigned less or equal</li>
4675 <li><tt>sgt</tt>: signed greater than</li>
4676 <li><tt>sge</tt>: signed greater or equal</li>
4677 <li><tt>slt</tt>: signed less than</li>
4678 <li><tt>sle</tt>: signed less or equal</li>
4681 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
4682 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
4683 typed. They must also be identical types.</p>
4686 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
4687 condition code given as <tt>cond</tt>. The comparison performed always yields
4688 either an <a href="#t_integer"><tt>i1</tt></a> or vector of <tt>i1</tt>
4689 result, as follows:</p>
4692 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
4693 <tt>false</tt> otherwise. No sign interpretation is necessary or
4696 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
4697 <tt>false</tt> otherwise. No sign interpretation is necessary or
4700 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
4701 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4703 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
4704 <tt>true</tt> if <tt>op1</tt> is greater than or equal
4705 to <tt>op2</tt>.</li>
4707 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
4708 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4710 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
4711 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4713 <li><tt>sgt</tt>: interprets the operands as signed values and yields
4714 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4716 <li><tt>sge</tt>: interprets the operands as signed values and yields
4717 <tt>true</tt> if <tt>op1</tt> is greater than or equal
4718 to <tt>op2</tt>.</li>
4720 <li><tt>slt</tt>: interprets the operands as signed values and yields
4721 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4723 <li><tt>sle</tt>: interprets the operands as signed values and yields
4724 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4727 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
4728 values are compared as if they were integers.</p>
4730 <p>If the operands are integer vectors, then they are compared element by
4731 element. The result is an <tt>i1</tt> vector with the same number of elements
4732 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
4736 <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
4737 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
4738 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
4739 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4740 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4741 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
4744 <p>Note that the code generator does not yet support vector types with
4745 the <tt>icmp</tt> instruction.</p>
4749 <!-- _______________________________________________________________________ -->
4750 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4753 <div class="doc_text">
4757 <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4761 <p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
4762 values based on comparison of its operands.</p>
4764 <p>If the operands are floating point scalars, then the result type is a boolean
4765 (<a href="#t_integer"><tt>i1</tt></a>).</p>
4767 <p>If the operands are floating point vectors, then the result type is a vector
4768 of boolean with the same number of elements as the operands being
4772 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4773 the condition code indicating the kind of comparison to perform. It is not a
4774 value, just a keyword. The possible condition code are:</p>
4777 <li><tt>false</tt>: no comparison, always returns false</li>
4778 <li><tt>oeq</tt>: ordered and equal</li>
4779 <li><tt>ogt</tt>: ordered and greater than </li>
4780 <li><tt>oge</tt>: ordered and greater than or equal</li>
4781 <li><tt>olt</tt>: ordered and less than </li>
4782 <li><tt>ole</tt>: ordered and less than or equal</li>
4783 <li><tt>one</tt>: ordered and not equal</li>
4784 <li><tt>ord</tt>: ordered (no nans)</li>
4785 <li><tt>ueq</tt>: unordered or equal</li>
4786 <li><tt>ugt</tt>: unordered or greater than </li>
4787 <li><tt>uge</tt>: unordered or greater than or equal</li>
4788 <li><tt>ult</tt>: unordered or less than </li>
4789 <li><tt>ule</tt>: unordered or less than or equal</li>
4790 <li><tt>une</tt>: unordered or not equal</li>
4791 <li><tt>uno</tt>: unordered (either nans)</li>
4792 <li><tt>true</tt>: no comparison, always returns true</li>
4795 <p><i>Ordered</i> means that neither operand is a QNAN while
4796 <i>unordered</i> means that either operand may be a QNAN.</p>
4798 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
4799 a <a href="#t_floating">floating point</a> type or
4800 a <a href="#t_vector">vector</a> of floating point type. They must have
4801 identical types.</p>
4804 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4805 according to the condition code given as <tt>cond</tt>. If the operands are
4806 vectors, then the vectors are compared element by element. Each comparison
4807 performed always yields an <a href="#t_integer">i1</a> result, as
4811 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4813 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4814 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4816 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4817 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
4819 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4820 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4822 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4823 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4825 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4826 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4828 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4829 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4831 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4833 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
4834 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4836 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
4837 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4839 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
4840 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4842 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
4843 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4845 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
4846 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4848 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
4849 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4851 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4853 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4858 <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
4859 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4860 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4861 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
4864 <p>Note that the code generator does not yet support vector types with
4865 the <tt>fcmp</tt> instruction.</p>
4869 <!-- _______________________________________________________________________ -->
4870 <div class="doc_subsubsection">
4871 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4874 <div class="doc_text">
4878 <result> = phi <ty> [ <val0>, <label0>], ...
4882 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in the
4883 SSA graph representing the function.</p>
4886 <p>The type of the incoming values is specified with the first type field. After
4887 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
4888 one pair for each predecessor basic block of the current block. Only values
4889 of <a href="#t_firstclass">first class</a> type may be used as the value
4890 arguments to the PHI node. Only labels may be used as the label
4893 <p>There must be no non-phi instructions between the start of a basic block and
4894 the PHI instructions: i.e. PHI instructions must be first in a basic
4897 <p>For the purposes of the SSA form, the use of each incoming value is deemed to
4898 occur on the edge from the corresponding predecessor block to the current
4899 block (but after any definition of an '<tt>invoke</tt>' instruction's return
4900 value on the same edge).</p>
4903 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4904 specified by the pair corresponding to the predecessor basic block that
4905 executed just prior to the current block.</p>
4909 Loop: ; Infinite loop that counts from 0 on up...
4910 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4911 %nextindvar = add i32 %indvar, 1
4917 <!-- _______________________________________________________________________ -->
4918 <div class="doc_subsubsection">
4919 <a name="i_select">'<tt>select</tt>' Instruction</a>
4922 <div class="doc_text">
4926 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
4928 <i>selty</i> is either i1 or {<N x i1>}
4932 <p>The '<tt>select</tt>' instruction is used to choose one value based on a
4933 condition, without branching.</p>
4937 <p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
4938 values indicating the condition, and two values of the
4939 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
4940 vectors and the condition is a scalar, then entire vectors are selected, not
4941 individual elements.</p>
4944 <p>If the condition is an i1 and it evaluates to 1, the instruction returns the
4945 first value argument; otherwise, it returns the second value argument.</p>
4947 <p>If the condition is a vector of i1, then the value arguments must be vectors
4948 of the same size, and the selection is done element by element.</p>
4952 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4955 <p>Note that the code generator does not yet support conditions
4956 with vector type.</p>
4960 <!-- _______________________________________________________________________ -->
4961 <div class="doc_subsubsection">
4962 <a name="i_call">'<tt>call</tt>' Instruction</a>
4965 <div class="doc_text">
4969 <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>]
4973 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4976 <p>This instruction requires several arguments:</p>
4979 <li>The optional "tail" marker indicates whether the callee function accesses
4980 any allocas or varargs in the caller. If the "tail" marker is present,
4981 the function call is eligible for tail call optimization. Note that calls
4982 may be marked "tail" even if they do not occur before
4983 a <a href="#i_ret"><tt>ret</tt></a> instruction.</li>
4985 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
4986 convention</a> the call should use. If none is specified, the call
4987 defaults to using C calling conventions.</li>
4989 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4990 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
4991 '<tt>inreg</tt>' attributes are valid here.</li>
4993 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
4994 type of the return value. Functions that return no value are marked
4995 <tt><a href="#t_void">void</a></tt>.</li>
4997 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
4998 being invoked. The argument types must match the types implied by this
4999 signature. This type can be omitted if the function is not varargs and if
5000 the function type does not return a pointer to a function.</li>
5002 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
5003 be invoked. In most cases, this is a direct function invocation, but
5004 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
5005 to function value.</li>
5007 <li>'<tt>function args</tt>': argument list whose types match the function
5008 signature argument types. All arguments must be of
5009 <a href="#t_firstclass">first class</a> type. If the function signature
5010 indicates the function accepts a variable number of arguments, the extra
5011 arguments can be specified.</li>
5013 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
5014 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
5015 '<tt>readnone</tt>' attributes are valid here.</li>
5019 <p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
5020 a specified function, with its incoming arguments bound to the specified
5021 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
5022 function, control flow continues with the instruction after the function
5023 call, and the return value of the function is bound to the result
5028 %retval = call i32 @test(i32 %argc)
5029 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
5030 %X = tail call i32 @foo() <i>; yields i32</i>
5031 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
5032 call void %foo(i8 97 signext)
5034 %struct.A = type { i32, i8 }
5035 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
5036 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
5037 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
5038 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
5039 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
5042 <p>llvm treats calls to some functions with names and arguments that match the
5043 standard C99 library as being the C99 library functions, and may perform
5044 optimizations or generate code for them under that assumption. This is
5045 something we'd like to change in the future to provide better support for
5046 freestanding environments and non-C-based langauges.</p>
5050 <!-- _______________________________________________________________________ -->
5051 <div class="doc_subsubsection">
5052 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
5055 <div class="doc_text">
5059 <resultval> = va_arg <va_list*> <arglist>, <argty>
5063 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
5064 the "variable argument" area of a function call. It is used to implement the
5065 <tt>va_arg</tt> macro in C.</p>
5068 <p>This instruction takes a <tt>va_list*</tt> value and the type of the
5069 argument. It returns a value of the specified argument type and increments
5070 the <tt>va_list</tt> to point to the next argument. The actual type
5071 of <tt>va_list</tt> is target specific.</p>
5074 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
5075 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
5076 to the next argument. For more information, see the variable argument
5077 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
5079 <p>It is legal for this instruction to be called in a function which does not
5080 take a variable number of arguments, for example, the <tt>vfprintf</tt>
5083 <p><tt>va_arg</tt> is an LLVM instruction instead of
5084 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
5088 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
5090 <p>Note that the code generator does not yet fully support va_arg on many
5091 targets. Also, it does not currently support va_arg with aggregate types on
5096 <!-- *********************************************************************** -->
5097 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
5098 <!-- *********************************************************************** -->
5100 <div class="doc_text">
5102 <p>LLVM supports the notion of an "intrinsic function". These functions have
5103 well known names and semantics and are required to follow certain
5104 restrictions. Overall, these intrinsics represent an extension mechanism for
5105 the LLVM language that does not require changing all of the transformations
5106 in LLVM when adding to the language (or the bitcode reader/writer, the
5107 parser, etc...).</p>
5109 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
5110 prefix is reserved in LLVM for intrinsic names; thus, function names may not
5111 begin with this prefix. Intrinsic functions must always be external
5112 functions: you cannot define the body of intrinsic functions. Intrinsic
5113 functions may only be used in call or invoke instructions: it is illegal to
5114 take the address of an intrinsic function. Additionally, because intrinsic
5115 functions are part of the LLVM language, it is required if any are added that
5116 they be documented here.</p>
5118 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
5119 family of functions that perform the same operation but on different data
5120 types. Because LLVM can represent over 8 million different integer types,
5121 overloading is used commonly to allow an intrinsic function to operate on any
5122 integer type. One or more of the argument types or the result type can be
5123 overloaded to accept any integer type. Argument types may also be defined as
5124 exactly matching a previous argument's type or the result type. This allows
5125 an intrinsic function which accepts multiple arguments, but needs all of them
5126 to be of the same type, to only be overloaded with respect to a single
5127 argument or the result.</p>
5129 <p>Overloaded intrinsics will have the names of its overloaded argument types
5130 encoded into its function name, each preceded by a period. Only those types
5131 which are overloaded result in a name suffix. Arguments whose type is matched
5132 against another type do not. For example, the <tt>llvm.ctpop</tt> function
5133 can take an integer of any width and returns an integer of exactly the same
5134 integer width. This leads to a family of functions such as
5135 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
5136 %val)</tt>. Only one type, the return type, is overloaded, and only one type
5137 suffix is required. Because the argument's type is matched against the return
5138 type, it does not require its own name suffix.</p>
5140 <p>To learn how to add an intrinsic function, please see the
5141 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
5145 <!-- ======================================================================= -->
5146 <div class="doc_subsection">
5147 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
5150 <div class="doc_text">
5152 <p>Variable argument support is defined in LLVM with
5153 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
5154 intrinsic functions. These functions are related to the similarly named
5155 macros defined in the <tt><stdarg.h></tt> header file.</p>
5157 <p>All of these functions operate on arguments that use a target-specific value
5158 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
5159 not define what this type is, so all transformations should be prepared to
5160 handle these functions regardless of the type used.</p>
5162 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
5163 instruction and the variable argument handling intrinsic functions are
5166 <div class="doc_code">
5168 define i32 @test(i32 %X, ...) {
5169 ; Initialize variable argument processing
5171 %ap2 = bitcast i8** %ap to i8*
5172 call void @llvm.va_start(i8* %ap2)
5174 ; Read a single integer argument
5175 %tmp = va_arg i8** %ap, i32
5177 ; Demonstrate usage of llvm.va_copy and llvm.va_end
5179 %aq2 = bitcast i8** %aq to i8*
5180 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
5181 call void @llvm.va_end(i8* %aq2)
5183 ; Stop processing of arguments.
5184 call void @llvm.va_end(i8* %ap2)
5188 declare void @llvm.va_start(i8*)
5189 declare void @llvm.va_copy(i8*, i8*)
5190 declare void @llvm.va_end(i8*)
5196 <!-- _______________________________________________________________________ -->
5197 <div class="doc_subsubsection">
5198 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
5202 <div class="doc_text">
5206 declare void %llvm.va_start(i8* <arglist>)
5210 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*<arglist></tt>
5211 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
5214 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
5217 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
5218 macro available in C. In a target-dependent way, it initializes
5219 the <tt>va_list</tt> element to which the argument points, so that the next
5220 call to <tt>va_arg</tt> will produce the first variable argument passed to
5221 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
5222 need to know the last argument of the function as the compiler can figure
5227 <!-- _______________________________________________________________________ -->
5228 <div class="doc_subsubsection">
5229 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
5232 <div class="doc_text">
5236 declare void @llvm.va_end(i8* <arglist>)
5240 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
5241 which has been initialized previously
5242 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
5243 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
5246 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
5249 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
5250 macro available in C. In a target-dependent way, it destroys
5251 the <tt>va_list</tt> element to which the argument points. Calls
5252 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
5253 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
5254 with calls to <tt>llvm.va_end</tt>.</p>
5258 <!-- _______________________________________________________________________ -->
5259 <div class="doc_subsubsection">
5260 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
5263 <div class="doc_text">
5267 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
5271 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
5272 from the source argument list to the destination argument list.</p>
5275 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
5276 The second argument is a pointer to a <tt>va_list</tt> element to copy
5280 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
5281 macro available in C. In a target-dependent way, it copies the
5282 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
5283 element. This intrinsic is necessary because
5284 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
5285 arbitrarily complex and require, for example, memory allocation.</p>
5289 <!-- ======================================================================= -->
5290 <div class="doc_subsection">
5291 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
5294 <div class="doc_text">
5296 <p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
5297 Collection</a> (GC) requires the implementation and generation of these
5298 intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
5299 roots on the stack</a>, as well as garbage collector implementations that
5300 require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
5301 barriers. Front-ends for type-safe garbage collected languages should generate
5302 these intrinsics to make use of the LLVM garbage collectors. For more details,
5303 see <a href="GarbageCollection.html">Accurate Garbage Collection with
5306 <p>The garbage collection intrinsics only operate on objects in the generic
5307 address space (address space zero).</p>
5311 <!-- _______________________________________________________________________ -->
5312 <div class="doc_subsubsection">
5313 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
5316 <div class="doc_text">
5320 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
5324 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
5325 the code generator, and allows some metadata to be associated with it.</p>
5328 <p>The first argument specifies the address of a stack object that contains the
5329 root pointer. The second pointer (which must be either a constant or a
5330 global value address) contains the meta-data to be associated with the
5334 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
5335 location. At compile-time, the code generator generates information to allow
5336 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5337 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5342 <!-- _______________________________________________________________________ -->
5343 <div class="doc_subsubsection">
5344 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
5347 <div class="doc_text">
5351 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
5355 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
5356 locations, allowing garbage collector implementations that require read
5360 <p>The second argument is the address to read from, which should be an address
5361 allocated from the garbage collector. The first object is a pointer to the
5362 start of the referenced object, if needed by the language runtime (otherwise
5366 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
5367 instruction, but may be replaced with substantially more complex code by the
5368 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5369 may only be used in a function which <a href="#gc">specifies a GC
5374 <!-- _______________________________________________________________________ -->
5375 <div class="doc_subsubsection">
5376 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
5379 <div class="doc_text">
5383 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
5387 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
5388 locations, allowing garbage collector implementations that require write
5389 barriers (such as generational or reference counting collectors).</p>
5392 <p>The first argument is the reference to store, the second is the start of the
5393 object to store it to, and the third is the address of the field of Obj to
5394 store to. If the runtime does not require a pointer to the object, Obj may
5398 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
5399 instruction, but may be replaced with substantially more complex code by the
5400 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5401 may only be used in a function which <a href="#gc">specifies a GC
5406 <!-- ======================================================================= -->
5407 <div class="doc_subsection">
5408 <a name="int_codegen">Code Generator Intrinsics</a>
5411 <div class="doc_text">
5413 <p>These intrinsics are provided by LLVM to expose special features that may
5414 only be implemented with code generator support.</p>
5418 <!-- _______________________________________________________________________ -->
5419 <div class="doc_subsubsection">
5420 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
5423 <div class="doc_text">
5427 declare i8 *@llvm.returnaddress(i32 <level>)
5431 <p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5432 target-specific value indicating the return address of the current function
5433 or one of its callers.</p>
5436 <p>The argument to this intrinsic indicates which function to return the address
5437 for. Zero indicates the calling function, one indicates its caller, etc.
5438 The argument is <b>required</b> to be a constant integer value.</p>
5441 <p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
5442 indicating the return address of the specified call frame, or zero if it
5443 cannot be identified. The value returned by this intrinsic is likely to be
5444 incorrect or 0 for arguments other than zero, so it should only be used for
5445 debugging purposes.</p>
5447 <p>Note that calling this intrinsic does not prevent function inlining or other
5448 aggressive transformations, so the value returned may not be that of the
5449 obvious source-language caller.</p>
5453 <!-- _______________________________________________________________________ -->
5454 <div class="doc_subsubsection">
5455 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5458 <div class="doc_text">
5462 declare i8 *@llvm.frameaddress(i32 <level>)
5466 <p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5467 target-specific frame pointer value for the specified stack frame.</p>
5470 <p>The argument to this intrinsic indicates which function to return the frame
5471 pointer for. Zero indicates the calling function, one indicates its caller,
5472 etc. The argument is <b>required</b> to be a constant integer value.</p>
5475 <p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
5476 indicating the frame address of the specified call frame, or zero if it
5477 cannot be identified. The value returned by this intrinsic is likely to be
5478 incorrect or 0 for arguments other than zero, so it should only be used for
5479 debugging purposes.</p>
5481 <p>Note that calling this intrinsic does not prevent function inlining or other
5482 aggressive transformations, so the value returned may not be that of the
5483 obvious source-language caller.</p>
5487 <!-- _______________________________________________________________________ -->
5488 <div class="doc_subsubsection">
5489 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5492 <div class="doc_text">
5496 declare i8 *@llvm.stacksave()
5500 <p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
5501 of the function stack, for use
5502 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
5503 useful for implementing language features like scoped automatic variable
5504 sized arrays in C99.</p>
5507 <p>This intrinsic returns a opaque pointer value that can be passed
5508 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
5509 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
5510 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
5511 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
5512 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
5513 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
5517 <!-- _______________________________________________________________________ -->
5518 <div class="doc_subsubsection">
5519 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5522 <div class="doc_text">
5526 declare void @llvm.stackrestore(i8 * %ptr)
5530 <p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5531 the function stack to the state it was in when the
5532 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
5533 executed. This is useful for implementing language features like scoped
5534 automatic variable sized arrays in C99.</p>
5537 <p>See the description
5538 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
5542 <!-- _______________________________________________________________________ -->
5543 <div class="doc_subsubsection">
5544 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5547 <div class="doc_text">
5551 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
5555 <p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
5556 insert a prefetch instruction if supported; otherwise, it is a noop.
5557 Prefetches have no effect on the behavior of the program but can change its
5558 performance characteristics.</p>
5561 <p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
5562 specifier determining if the fetch should be for a read (0) or write (1),
5563 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5564 locality, to (3) - extremely local keep in cache. The <tt>rw</tt>
5565 and <tt>locality</tt> arguments must be constant integers.</p>
5568 <p>This intrinsic does not modify the behavior of the program. In particular,
5569 prefetches cannot trap and do not produce a value. On targets that support
5570 this intrinsic, the prefetch can provide hints to the processor cache for
5571 better performance.</p>
5575 <!-- _______________________________________________________________________ -->
5576 <div class="doc_subsubsection">
5577 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5580 <div class="doc_text">
5584 declare void @llvm.pcmarker(i32 <id>)
5588 <p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
5589 Counter (PC) in a region of code to simulators and other tools. The method
5590 is target specific, but it is expected that the marker will use exported
5591 symbols to transmit the PC of the marker. The marker makes no guarantees
5592 that it will remain with any specific instruction after optimizations. It is
5593 possible that the presence of a marker will inhibit optimizations. The
5594 intended use is to be inserted after optimizations to allow correlations of
5595 simulation runs.</p>
5598 <p><tt>id</tt> is a numerical id identifying the marker.</p>
5601 <p>This intrinsic does not modify the behavior of the program. Backends that do
5602 not support this intrinisic may ignore it.</p>
5606 <!-- _______________________________________________________________________ -->
5607 <div class="doc_subsubsection">
5608 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5611 <div class="doc_text">
5615 declare i64 @llvm.readcyclecounter( )
5619 <p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5620 counter register (or similar low latency, high accuracy clocks) on those
5621 targets that support it. On X86, it should map to RDTSC. On Alpha, it
5622 should map to RPCC. As the backing counters overflow quickly (on the order
5623 of 9 seconds on alpha), this should only be used for small timings.</p>
5626 <p>When directly supported, reading the cycle counter should not modify any
5627 memory. Implementations are allowed to either return a application specific
5628 value or a system wide value. On backends without support, this is lowered
5629 to a constant 0.</p>
5633 <!-- ======================================================================= -->
5634 <div class="doc_subsection">
5635 <a name="int_libc">Standard C Library Intrinsics</a>
5638 <div class="doc_text">
5640 <p>LLVM provides intrinsics for a few important standard C library functions.
5641 These intrinsics allow source-language front-ends to pass information about
5642 the alignment of the pointer arguments to the code generator, providing
5643 opportunity for more efficient code generation.</p>
5647 <!-- _______________________________________________________________________ -->
5648 <div class="doc_subsubsection">
5649 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5652 <div class="doc_text">
5655 <p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
5656 integer bit width. Not all targets support all bit widths however.</p>
5659 declare void @llvm.memcpy.i8(i8 * <dest>, i8 * <src>,
5660 i8 <len>, i32 <align>)
5661 declare void @llvm.memcpy.i16(i8 * <dest>, i8 * <src>,
5662 i16 <len>, i32 <align>)
5663 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
5664 i32 <len>, i32 <align>)
5665 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
5666 i64 <len>, i32 <align>)
5670 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
5671 source location to the destination location.</p>
5673 <p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5674 intrinsics do not return a value, and takes an extra alignment argument.</p>
5677 <p>The first argument is a pointer to the destination, the second is a pointer
5678 to the source. The third argument is an integer argument specifying the
5679 number of bytes to copy, and the fourth argument is the alignment of the
5680 source and destination locations.</p>
5682 <p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5683 then the caller guarantees that both the source and destination pointers are
5684 aligned to that boundary.</p>
5687 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
5688 source location to the destination location, which are not allowed to
5689 overlap. It copies "len" bytes of memory over. If the argument is known to
5690 be aligned to some boundary, this can be specified as the fourth argument,
5691 otherwise it should be set to 0 or 1.</p>
5695 <!-- _______________________________________________________________________ -->
5696 <div class="doc_subsubsection">
5697 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5700 <div class="doc_text">
5703 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5704 width. Not all targets support all bit widths however.</p>
5707 declare void @llvm.memmove.i8(i8 * <dest>, i8 * <src>,
5708 i8 <len>, i32 <align>)
5709 declare void @llvm.memmove.i16(i8 * <dest>, i8 * <src>,
5710 i16 <len>, i32 <align>)
5711 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
5712 i32 <len>, i32 <align>)
5713 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
5714 i64 <len>, i32 <align>)
5718 <p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
5719 source location to the destination location. It is similar to the
5720 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
5723 <p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5724 intrinsics do not return a value, and takes an extra alignment argument.</p>
5727 <p>The first argument is a pointer to the destination, the second is a pointer
5728 to the source. The third argument is an integer argument specifying the
5729 number of bytes to copy, and the fourth argument is the alignment of the
5730 source and destination locations.</p>
5732 <p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5733 then the caller guarantees that the source and destination pointers are
5734 aligned to that boundary.</p>
5737 <p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
5738 source location to the destination location, which may overlap. It copies
5739 "len" bytes of memory over. If the argument is known to be aligned to some
5740 boundary, this can be specified as the fourth argument, otherwise it should
5741 be set to 0 or 1.</p>
5745 <!-- _______________________________________________________________________ -->
5746 <div class="doc_subsubsection">
5747 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5750 <div class="doc_text">
5753 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5754 width. Not all targets support all bit widths however.</p>
5757 declare void @llvm.memset.i8(i8 * <dest>, i8 <val>,
5758 i8 <len>, i32 <align>)
5759 declare void @llvm.memset.i16(i8 * <dest>, i8 <val>,
5760 i16 <len>, i32 <align>)
5761 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
5762 i32 <len>, i32 <align>)
5763 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
5764 i64 <len>, i32 <align>)
5768 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
5769 particular byte value.</p>
5771 <p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
5772 intrinsic does not return a value, and takes an extra alignment argument.</p>
5775 <p>The first argument is a pointer to the destination to fill, the second is the
5776 byte value to fill it with, the third argument is an integer argument
5777 specifying the number of bytes to fill, and the fourth argument is the known
5778 alignment of destination location.</p>
5780 <p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5781 then the caller guarantees that the destination pointer is aligned to that
5785 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
5786 at the destination location. If the argument is known to be aligned to some
5787 boundary, this can be specified as the fourth argument, otherwise it should
5788 be set to 0 or 1.</p>
5792 <!-- _______________________________________________________________________ -->
5793 <div class="doc_subsubsection">
5794 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5797 <div class="doc_text">
5800 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
5801 floating point or vector of floating point type. Not all targets support all
5805 declare float @llvm.sqrt.f32(float %Val)
5806 declare double @llvm.sqrt.f64(double %Val)
5807 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5808 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5809 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
5813 <p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
5814 returning the same value as the libm '<tt>sqrt</tt>' functions would.
5815 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
5816 behavior for negative numbers other than -0.0 (which allows for better
5817 optimization, because there is no need to worry about errno being
5818 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
5821 <p>The argument and return value are floating point numbers of the same
5825 <p>This function returns the sqrt of the specified operand if it is a
5826 nonnegative floating point number.</p>
5830 <!-- _______________________________________________________________________ -->
5831 <div class="doc_subsubsection">
5832 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5835 <div class="doc_text">
5838 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
5839 floating point or vector of floating point type. Not all targets support all
5843 declare float @llvm.powi.f32(float %Val, i32 %power)
5844 declare double @llvm.powi.f64(double %Val, i32 %power)
5845 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5846 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5847 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
5851 <p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5852 specified (positive or negative) power. The order of evaluation of
5853 multiplications is not defined. When a vector of floating point type is
5854 used, the second argument remains a scalar integer value.</p>
5857 <p>The second argument is an integer power, and the first is a value to raise to
5861 <p>This function returns the first value raised to the second power with an
5862 unspecified sequence of rounding operations.</p>
5866 <!-- _______________________________________________________________________ -->
5867 <div class="doc_subsubsection">
5868 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5871 <div class="doc_text">
5874 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5875 floating point or vector of floating point type. Not all targets support all
5879 declare float @llvm.sin.f32(float %Val)
5880 declare double @llvm.sin.f64(double %Val)
5881 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5882 declare fp128 @llvm.sin.f128(fp128 %Val)
5883 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5887 <p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
5890 <p>The argument and return value are floating point numbers of the same
5894 <p>This function returns the sine of the specified operand, returning the same
5895 values as the libm <tt>sin</tt> functions would, and handles error conditions
5896 in the same way.</p>
5900 <!-- _______________________________________________________________________ -->
5901 <div class="doc_subsubsection">
5902 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5905 <div class="doc_text">
5908 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5909 floating point or vector of floating point type. Not all targets support all
5913 declare float @llvm.cos.f32(float %Val)
5914 declare double @llvm.cos.f64(double %Val)
5915 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5916 declare fp128 @llvm.cos.f128(fp128 %Val)
5917 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5921 <p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
5924 <p>The argument and return value are floating point numbers of the same
5928 <p>This function returns the cosine of the specified operand, returning the same
5929 values as the libm <tt>cos</tt> functions would, and handles error conditions
5930 in the same way.</p>
5934 <!-- _______________________________________________________________________ -->
5935 <div class="doc_subsubsection">
5936 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5939 <div class="doc_text">
5942 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5943 floating point or vector of floating point type. Not all targets support all
5947 declare float @llvm.pow.f32(float %Val, float %Power)
5948 declare double @llvm.pow.f64(double %Val, double %Power)
5949 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5950 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5951 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5955 <p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5956 specified (positive or negative) power.</p>
5959 <p>The second argument is a floating point power, and the first is a value to
5960 raise to that power.</p>
5963 <p>This function returns the first value raised to the second power, returning
5964 the same values as the libm <tt>pow</tt> functions would, and handles error
5965 conditions in the same way.</p>
5969 <!-- ======================================================================= -->
5970 <div class="doc_subsection">
5971 <a name="int_manip">Bit Manipulation Intrinsics</a>
5974 <div class="doc_text">
5976 <p>LLVM provides intrinsics for a few important bit manipulation operations.
5977 These allow efficient code generation for some algorithms.</p>
5981 <!-- _______________________________________________________________________ -->
5982 <div class="doc_subsubsection">
5983 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5986 <div class="doc_text">
5989 <p>This is an overloaded intrinsic function. You can use bswap on any integer
5990 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
5993 declare i16 @llvm.bswap.i16(i16 <id>)
5994 declare i32 @llvm.bswap.i32(i32 <id>)
5995 declare i64 @llvm.bswap.i64(i64 <id>)
5999 <p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
6000 values with an even number of bytes (positive multiple of 16 bits). These
6001 are useful for performing operations on data that is not in the target's
6002 native byte order.</p>
6005 <p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
6006 and low byte of the input i16 swapped. Similarly,
6007 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
6008 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
6009 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
6010 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
6011 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
6012 more, respectively).</p>
6016 <!-- _______________________________________________________________________ -->
6017 <div class="doc_subsubsection">
6018 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
6021 <div class="doc_text">
6024 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
6025 width. Not all targets support all bit widths however.</p>
6028 declare i8 @llvm.ctpop.i8(i8 <src>)
6029 declare i16 @llvm.ctpop.i16(i16 <src>)
6030 declare i32 @llvm.ctpop.i32(i32 <src>)
6031 declare i64 @llvm.ctpop.i64(i64 <src>)
6032 declare i256 @llvm.ctpop.i256(i256 <src>)
6036 <p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
6040 <p>The only argument is the value to be counted. The argument may be of any
6041 integer type. The return type must match the argument type.</p>
6044 <p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.</p>
6048 <!-- _______________________________________________________________________ -->
6049 <div class="doc_subsubsection">
6050 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
6053 <div class="doc_text">
6056 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
6057 integer bit width. Not all targets support all bit widths however.</p>
6060 declare i8 @llvm.ctlz.i8 (i8 <src>)
6061 declare i16 @llvm.ctlz.i16(i16 <src>)
6062 declare i32 @llvm.ctlz.i32(i32 <src>)
6063 declare i64 @llvm.ctlz.i64(i64 <src>)
6064 declare i256 @llvm.ctlz.i256(i256 <src>)
6068 <p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
6069 leading zeros in a variable.</p>
6072 <p>The only argument is the value to be counted. The argument may be of any
6073 integer type. The return type must match the argument type.</p>
6076 <p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
6077 zeros in a variable. If the src == 0 then the result is the size in bits of
6078 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
6082 <!-- _______________________________________________________________________ -->
6083 <div class="doc_subsubsection">
6084 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
6087 <div class="doc_text">
6090 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
6091 integer bit width. Not all targets support all bit widths however.</p>
6094 declare i8 @llvm.cttz.i8 (i8 <src>)
6095 declare i16 @llvm.cttz.i16(i16 <src>)
6096 declare i32 @llvm.cttz.i32(i32 <src>)
6097 declare i64 @llvm.cttz.i64(i64 <src>)
6098 declare i256 @llvm.cttz.i256(i256 <src>)
6102 <p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
6106 <p>The only argument is the value to be counted. The argument may be of any
6107 integer type. The return type must match the argument type.</p>
6110 <p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
6111 zeros in a variable. If the src == 0 then the result is the size in bits of
6112 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
6116 <!-- ======================================================================= -->
6117 <div class="doc_subsection">
6118 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
6121 <div class="doc_text">
6123 <p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
6127 <!-- _______________________________________________________________________ -->
6128 <div class="doc_subsubsection">
6129 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
6132 <div class="doc_text">
6135 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
6136 on any integer bit width.</p>
6139 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
6140 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6141 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6145 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6146 a signed addition of the two arguments, and indicate whether an overflow
6147 occurred during the signed summation.</p>
6150 <p>The arguments (%a and %b) and the first element of the result structure may
6151 be of integer types of any bit width, but they must have the same bit
6152 width. The second element of the result structure must be of
6153 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6154 undergo signed addition.</p>
6157 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6158 a signed addition of the two variables. They return a structure — the
6159 first element of which is the signed summation, and the second element of
6160 which is a bit specifying if the signed summation resulted in an
6165 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6166 %sum = extractvalue {i32, i1} %res, 0
6167 %obit = extractvalue {i32, i1} %res, 1
6168 br i1 %obit, label %overflow, label %normal
6173 <!-- _______________________________________________________________________ -->
6174 <div class="doc_subsubsection">
6175 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
6178 <div class="doc_text">
6181 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
6182 on any integer bit width.</p>
6185 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6186 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6187 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6191 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6192 an unsigned addition of the two arguments, and indicate whether a carry
6193 occurred during the unsigned summation.</p>
6196 <p>The arguments (%a and %b) and the first element of the result structure may
6197 be of integer types of any bit width, but they must have the same bit
6198 width. The second element of the result structure must be of
6199 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6200 undergo unsigned addition.</p>
6203 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6204 an unsigned addition of the two arguments. They return a structure —
6205 the first element of which is the sum, and the second element of which is a
6206 bit specifying if the unsigned summation resulted in a carry.</p>
6210 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6211 %sum = extractvalue {i32, i1} %res, 0
6212 %obit = extractvalue {i32, i1} %res, 1
6213 br i1 %obit, label %carry, label %normal
6218 <!-- _______________________________________________________________________ -->
6219 <div class="doc_subsubsection">
6220 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
6223 <div class="doc_text">
6226 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
6227 on any integer bit width.</p>
6230 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6231 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6232 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6236 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6237 a signed subtraction of the two arguments, and indicate whether an overflow
6238 occurred during the signed subtraction.</p>
6241 <p>The arguments (%a and %b) and the first element of the result structure may
6242 be of integer types of any bit width, but they must have the same bit
6243 width. The second element of the result structure must be of
6244 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6245 undergo signed subtraction.</p>
6248 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6249 a signed subtraction of the two arguments. They return a structure —
6250 the first element of which is the subtraction, and the second element of
6251 which is a bit specifying if the signed subtraction resulted in an
6256 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6257 %sum = extractvalue {i32, i1} %res, 0
6258 %obit = extractvalue {i32, i1} %res, 1
6259 br i1 %obit, label %overflow, label %normal
6264 <!-- _______________________________________________________________________ -->
6265 <div class="doc_subsubsection">
6266 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
6269 <div class="doc_text">
6272 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
6273 on any integer bit width.</p>
6276 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6277 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6278 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6282 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6283 an unsigned subtraction of the two arguments, and indicate whether an
6284 overflow occurred during the unsigned subtraction.</p>
6287 <p>The arguments (%a and %b) and the first element of the result structure may
6288 be of integer types of any bit width, but they must have the same bit
6289 width. The second element of the result structure must be of
6290 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6291 undergo unsigned subtraction.</p>
6294 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6295 an unsigned subtraction of the two arguments. They return a structure —
6296 the first element of which is the subtraction, and the second element of
6297 which is a bit specifying if the unsigned subtraction resulted in an
6302 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6303 %sum = extractvalue {i32, i1} %res, 0
6304 %obit = extractvalue {i32, i1} %res, 1
6305 br i1 %obit, label %overflow, label %normal
6310 <!-- _______________________________________________________________________ -->
6311 <div class="doc_subsubsection">
6312 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
6315 <div class="doc_text">
6318 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
6319 on any integer bit width.</p>
6322 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6323 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6324 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6329 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6330 a signed multiplication of the two arguments, and indicate whether an
6331 overflow occurred during the signed multiplication.</p>
6334 <p>The arguments (%a and %b) and the first element of the result structure may
6335 be of integer types of any bit width, but they must have the same bit
6336 width. The second element of the result structure must be of
6337 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6338 undergo signed multiplication.</p>
6341 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6342 a signed multiplication of the two arguments. They return a structure —
6343 the first element of which is the multiplication, and the second element of
6344 which is a bit specifying if the signed multiplication resulted in an
6349 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6350 %sum = extractvalue {i32, i1} %res, 0
6351 %obit = extractvalue {i32, i1} %res, 1
6352 br i1 %obit, label %overflow, label %normal
6357 <!-- _______________________________________________________________________ -->
6358 <div class="doc_subsubsection">
6359 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6362 <div class="doc_text">
6365 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6366 on any integer bit width.</p>
6369 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6370 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6371 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6375 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6376 a unsigned multiplication of the two arguments, and indicate whether an
6377 overflow occurred during the unsigned multiplication.</p>
6380 <p>The arguments (%a and %b) and the first element of the result structure may
6381 be of integer types of any bit width, but they must have the same bit
6382 width. The second element of the result structure must be of
6383 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6384 undergo unsigned multiplication.</p>
6387 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6388 an unsigned multiplication of the two arguments. They return a structure
6389 — the first element of which is the multiplication, and the second
6390 element of which is a bit specifying if the unsigned multiplication resulted
6395 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6396 %sum = extractvalue {i32, i1} %res, 0
6397 %obit = extractvalue {i32, i1} %res, 1
6398 br i1 %obit, label %overflow, label %normal
6403 <!-- ======================================================================= -->
6404 <div class="doc_subsection">
6405 <a name="int_debugger">Debugger Intrinsics</a>
6408 <div class="doc_text">
6410 <p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
6411 prefix), are described in
6412 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
6413 Level Debugging</a> document.</p>
6417 <!-- ======================================================================= -->
6418 <div class="doc_subsection">
6419 <a name="int_eh">Exception Handling Intrinsics</a>
6422 <div class="doc_text">
6424 <p>The LLVM exception handling intrinsics (which all start with
6425 <tt>llvm.eh.</tt> prefix), are described in
6426 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6427 Handling</a> document.</p>
6431 <!-- ======================================================================= -->
6432 <div class="doc_subsection">
6433 <a name="int_trampoline">Trampoline Intrinsic</a>
6436 <div class="doc_text">
6438 <p>This intrinsic makes it possible to excise one parameter, marked with
6439 the <tt>nest</tt> attribute, from a function. The result is a callable
6440 function pointer lacking the nest parameter - the caller does not need to
6441 provide a value for it. Instead, the value to use is stored in advance in a
6442 "trampoline", a block of memory usually allocated on the stack, which also
6443 contains code to splice the nest value into the argument list. This is used
6444 to implement the GCC nested function address extension.</p>
6446 <p>For example, if the function is
6447 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
6448 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
6451 <div class="doc_code">
6453 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6454 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6455 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
6456 %fp = bitcast i8* %p to i32 (i32, i32)*
6460 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
6461 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
6465 <!-- _______________________________________________________________________ -->
6466 <div class="doc_subsubsection">
6467 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6470 <div class="doc_text">
6474 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
6478 <p>This fills the memory pointed to by <tt>tramp</tt> with code and returns a
6479 function pointer suitable for executing it.</p>
6482 <p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6483 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
6484 sufficiently aligned block of memory; this memory is written to by the
6485 intrinsic. Note that the size and the alignment are target-specific - LLVM
6486 currently provides no portable way of determining them, so a front-end that
6487 generates this intrinsic needs to have some target-specific knowledge.
6488 The <tt>func</tt> argument must hold a function bitcast to
6489 an <tt>i8*</tt>.</p>
6492 <p>The block of memory pointed to by <tt>tramp</tt> is filled with target
6493 dependent code, turning it into a function. A pointer to this function is
6494 returned, but needs to be bitcast to an <a href="#int_trampoline">appropriate
6495 function pointer type</a> before being called. The new function's signature
6496 is the same as that of <tt>func</tt> with any arguments marked with
6497 the <tt>nest</tt> attribute removed. At most one such <tt>nest</tt> argument
6498 is allowed, and it must be of pointer type. Calling the new function is
6499 equivalent to calling <tt>func</tt> with the same argument list, but
6500 with <tt>nval</tt> used for the missing <tt>nest</tt> argument. If, after
6501 calling <tt>llvm.init.trampoline</tt>, the memory pointed to
6502 by <tt>tramp</tt> is modified, then the effect of any later call to the
6503 returned function pointer is undefined.</p>
6507 <!-- ======================================================================= -->
6508 <div class="doc_subsection">
6509 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6512 <div class="doc_text">
6514 <p>These intrinsic functions expand the "universal IR" of LLVM to represent
6515 hardware constructs for atomic operations and memory synchronization. This
6516 provides an interface to the hardware, not an interface to the programmer. It
6517 is aimed at a low enough level to allow any programming models or APIs
6518 (Application Programming Interfaces) which need atomic behaviors to map
6519 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
6520 hardware provides a "universal IR" for source languages, it also provides a
6521 starting point for developing a "universal" atomic operation and
6522 synchronization IR.</p>
6524 <p>These do <em>not</em> form an API such as high-level threading libraries,
6525 software transaction memory systems, atomic primitives, and intrinsic
6526 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6527 application libraries. The hardware interface provided by LLVM should allow
6528 a clean implementation of all of these APIs and parallel programming models.
6529 No one model or paradigm should be selected above others unless the hardware
6530 itself ubiquitously does so.</p>
6534 <!-- _______________________________________________________________________ -->
6535 <div class="doc_subsubsection">
6536 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6538 <div class="doc_text">
6541 declare void @llvm.memory.barrier( i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>, i1 <device> )
6545 <p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6546 specific pairs of memory access types.</p>
6549 <p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6550 The first four arguments enables a specific barrier as listed below. The
6551 fith argument specifies that the barrier applies to io or device or uncached
6555 <li><tt>ll</tt>: load-load barrier</li>
6556 <li><tt>ls</tt>: load-store barrier</li>
6557 <li><tt>sl</tt>: store-load barrier</li>
6558 <li><tt>ss</tt>: store-store barrier</li>
6559 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
6563 <p>This intrinsic causes the system to enforce some ordering constraints upon
6564 the loads and stores of the program. This barrier does not
6565 indicate <em>when</em> any events will occur, it only enforces
6566 an <em>order</em> in which they occur. For any of the specified pairs of load
6567 and store operations (f.ex. load-load, or store-load), all of the first
6568 operations preceding the barrier will complete before any of the second
6569 operations succeeding the barrier begin. Specifically the semantics for each
6570 pairing is as follows:</p>
6573 <li><tt>ll</tt>: All loads before the barrier must complete before any load
6574 after the barrier begins.</li>
6575 <li><tt>ls</tt>: All loads before the barrier must complete before any
6576 store after the barrier begins.</li>
6577 <li><tt>ss</tt>: All stores before the barrier must complete before any
6578 store after the barrier begins.</li>
6579 <li><tt>sl</tt>: All stores before the barrier must complete before any
6580 load after the barrier begins.</li>
6583 <p>These semantics are applied with a logical "and" behavior when more than one
6584 is enabled in a single memory barrier intrinsic.</p>
6586 <p>Backends may implement stronger barriers than those requested when they do
6587 not support as fine grained a barrier as requested. Some architectures do
6588 not need all types of barriers and on such architectures, these become
6593 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6594 %ptr = bitcast i8* %mallocP to i32*
6597 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
6598 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
6599 <i>; guarantee the above finishes</i>
6600 store i32 8, %ptr <i>; before this begins</i>
6605 <!-- _______________________________________________________________________ -->
6606 <div class="doc_subsubsection">
6607 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
6610 <div class="doc_text">
6613 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
6614 any integer bit width and for different address spaces. Not all targets
6615 support all bit widths however.</p>
6618 declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* <ptr>, i8 <cmp>, i8 <val> )
6619 declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* <ptr>, i16 <cmp>, i16 <val> )
6620 declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* <ptr>, i32 <cmp>, i32 <val> )
6621 declare i64 @llvm.atomic.cmp.swap.i64.p0i64( i64* <ptr>, i64 <cmp>, i64 <val> )
6625 <p>This loads a value in memory and compares it to a given value. If they are
6626 equal, it stores a new value into the memory.</p>
6629 <p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
6630 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6631 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6632 this integer type. While any bit width integer may be used, targets may only
6633 lower representations they support in hardware.</p>
6636 <p>This entire intrinsic must be executed atomically. It first loads the value
6637 in memory pointed to by <tt>ptr</tt> and compares it with the
6638 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
6639 memory. The loaded value is yielded in all cases. This provides the
6640 equivalent of an atomic compare-and-swap operation within the SSA
6645 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6646 %ptr = bitcast i8* %mallocP to i32*
6649 %val1 = add i32 4, 4
6650 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
6651 <i>; yields {i32}:result1 = 4</i>
6652 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6653 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6655 %val2 = add i32 1, 1
6656 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
6657 <i>; yields {i32}:result2 = 8</i>
6658 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6660 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6665 <!-- _______________________________________________________________________ -->
6666 <div class="doc_subsubsection">
6667 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6669 <div class="doc_text">
6672 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6673 integer bit width. Not all targets support all bit widths however.</p>
6676 declare i8 @llvm.atomic.swap.i8.p0i8( i8* <ptr>, i8 <val> )
6677 declare i16 @llvm.atomic.swap.i16.p0i16( i16* <ptr>, i16 <val> )
6678 declare i32 @llvm.atomic.swap.i32.p0i32( i32* <ptr>, i32 <val> )
6679 declare i64 @llvm.atomic.swap.i64.p0i64( i64* <ptr>, i64 <val> )
6683 <p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6684 the value from memory. It then stores the value in <tt>val</tt> in the memory
6685 at <tt>ptr</tt>.</p>
6688 <p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
6689 the <tt>val</tt> argument and the result must be integers of the same bit
6690 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6691 integer type. The targets may only lower integer representations they
6695 <p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6696 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6697 equivalent of an atomic swap operation within the SSA framework.</p>
6701 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6702 %ptr = bitcast i8* %mallocP to i32*
6705 %val1 = add i32 4, 4
6706 %result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
6707 <i>; yields {i32}:result1 = 4</i>
6708 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6709 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6711 %val2 = add i32 1, 1
6712 %result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
6713 <i>; yields {i32}:result2 = 8</i>
6715 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6716 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6721 <!-- _______________________________________________________________________ -->
6722 <div class="doc_subsubsection">
6723 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
6727 <div class="doc_text">
6730 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
6731 any integer bit width. Not all targets support all bit widths however.</p>
6734 declare i8 @llvm.atomic.load.add.i8..p0i8( i8* <ptr>, i8 <delta> )
6735 declare i16 @llvm.atomic.load.add.i16..p0i16( i16* <ptr>, i16 <delta> )
6736 declare i32 @llvm.atomic.load.add.i32..p0i32( i32* <ptr>, i32 <delta> )
6737 declare i64 @llvm.atomic.load.add.i64..p0i64( i64* <ptr>, i64 <delta> )
6741 <p>This intrinsic adds <tt>delta</tt> to the value stored in memory
6742 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
6745 <p>The intrinsic takes two arguments, the first a pointer to an integer value
6746 and the second an integer value. The result is also an integer value. These
6747 integer types can have any bit width, but they must all have the same bit
6748 width. The targets may only lower integer representations they support.</p>
6751 <p>This intrinsic does a series of operations atomically. It first loads the
6752 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6753 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
6757 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6758 %ptr = bitcast i8* %mallocP to i32*
6760 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
6761 <i>; yields {i32}:result1 = 4</i>
6762 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
6763 <i>; yields {i32}:result2 = 8</i>
6764 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
6765 <i>; yields {i32}:result3 = 10</i>
6766 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
6771 <!-- _______________________________________________________________________ -->
6772 <div class="doc_subsubsection">
6773 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6777 <div class="doc_text">
6780 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
6781 any integer bit width and for different address spaces. Not all targets
6782 support all bit widths however.</p>
6785 declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* <ptr>, i8 <delta> )
6786 declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* <ptr>, i16 <delta> )
6787 declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* <ptr>, i32 <delta> )
6788 declare i64 @llvm.atomic.load.sub.i64.p0i32( i64* <ptr>, i64 <delta> )
6792 <p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6793 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
6796 <p>The intrinsic takes two arguments, the first a pointer to an integer value
6797 and the second an integer value. The result is also an integer value. These
6798 integer types can have any bit width, but they must all have the same bit
6799 width. The targets may only lower integer representations they support.</p>
6802 <p>This intrinsic does a series of operations atomically. It first loads the
6803 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6804 result to <tt>ptr</tt>. It yields the original value stored
6805 at <tt>ptr</tt>.</p>
6809 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6810 %ptr = bitcast i8* %mallocP to i32*
6812 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
6813 <i>; yields {i32}:result1 = 8</i>
6814 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
6815 <i>; yields {i32}:result2 = 4</i>
6816 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
6817 <i>; yields {i32}:result3 = 2</i>
6818 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6823 <!-- _______________________________________________________________________ -->
6824 <div class="doc_subsubsection">
6825 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6826 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6827 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6828 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6831 <div class="doc_text">
6834 <p>These are overloaded intrinsics. You can
6835 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
6836 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
6837 bit width and for different address spaces. Not all targets support all bit
6841 declare i8 @llvm.atomic.load.and.i8.p0i8( i8* <ptr>, i8 <delta> )
6842 declare i16 @llvm.atomic.load.and.i16.p0i16( i16* <ptr>, i16 <delta> )
6843 declare i32 @llvm.atomic.load.and.i32.p0i32( i32* <ptr>, i32 <delta> )
6844 declare i64 @llvm.atomic.load.and.i64.p0i64( i64* <ptr>, i64 <delta> )
6848 declare i8 @llvm.atomic.load.or.i8.p0i8( i8* <ptr>, i8 <delta> )
6849 declare i16 @llvm.atomic.load.or.i16.p0i16( i16* <ptr>, i16 <delta> )
6850 declare i32 @llvm.atomic.load.or.i32.p0i32( i32* <ptr>, i32 <delta> )
6851 declare i64 @llvm.atomic.load.or.i64.p0i64( i64* <ptr>, i64 <delta> )
6855 declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* <ptr>, i8 <delta> )
6856 declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* <ptr>, i16 <delta> )
6857 declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* <ptr>, i32 <delta> )
6858 declare i64 @llvm.atomic.load.nand.i64.p0i32( i64* <ptr>, i64 <delta> )
6862 declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* <ptr>, i8 <delta> )
6863 declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* <ptr>, i16 <delta> )
6864 declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* <ptr>, i32 <delta> )
6865 declare i64 @llvm.atomic.load.xor.i64.p0i32( i64* <ptr>, i64 <delta> )
6869 <p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6870 the value stored in memory at <tt>ptr</tt>. It yields the original value
6871 at <tt>ptr</tt>.</p>
6874 <p>These intrinsics take two arguments, the first a pointer to an integer value
6875 and the second an integer value. The result is also an integer value. These
6876 integer types can have any bit width, but they must all have the same bit
6877 width. The targets may only lower integer representations they support.</p>
6880 <p>These intrinsics does a series of operations atomically. They first load the
6881 value stored at <tt>ptr</tt>. They then do the bitwise
6882 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
6883 original value stored at <tt>ptr</tt>.</p>
6887 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6888 %ptr = bitcast i8* %mallocP to i32*
6889 store i32 0x0F0F, %ptr
6890 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
6891 <i>; yields {i32}:result0 = 0x0F0F</i>
6892 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
6893 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
6894 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
6895 <i>; yields {i32}:result2 = 0xF0</i>
6896 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
6897 <i>; yields {i32}:result3 = FF</i>
6898 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6903 <!-- _______________________________________________________________________ -->
6904 <div class="doc_subsubsection">
6905 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6906 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6907 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6908 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6911 <div class="doc_text">
6914 <p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6915 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
6916 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6917 address spaces. Not all targets support all bit widths however.</p>
6920 declare i8 @llvm.atomic.load.max.i8.p0i8( i8* <ptr>, i8 <delta> )
6921 declare i16 @llvm.atomic.load.max.i16.p0i16( i16* <ptr>, i16 <delta> )
6922 declare i32 @llvm.atomic.load.max.i32.p0i32( i32* <ptr>, i32 <delta> )
6923 declare i64 @llvm.atomic.load.max.i64.p0i64( i64* <ptr>, i64 <delta> )
6927 declare i8 @llvm.atomic.load.min.i8.p0i8( i8* <ptr>, i8 <delta> )
6928 declare i16 @llvm.atomic.load.min.i16.p0i16( i16* <ptr>, i16 <delta> )
6929 declare i32 @llvm.atomic.load.min.i32..p0i32( i32* <ptr>, i32 <delta> )
6930 declare i64 @llvm.atomic.load.min.i64..p0i64( i64* <ptr>, i64 <delta> )
6934 declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* <ptr>, i8 <delta> )
6935 declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* <ptr>, i16 <delta> )
6936 declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* <ptr>, i32 <delta> )
6937 declare i64 @llvm.atomic.load.umax.i64.p0i64( i64* <ptr>, i64 <delta> )
6941 declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* <ptr>, i8 <delta> )
6942 declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* <ptr>, i16 <delta> )
6943 declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* <ptr>, i32 <delta> )
6944 declare i64 @llvm.atomic.load.umin.i64..p0i64( i64* <ptr>, i64 <delta> )
6948 <p>These intrinsics takes the signed or unsigned minimum or maximum of
6949 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6950 original value at <tt>ptr</tt>.</p>
6953 <p>These intrinsics take two arguments, the first a pointer to an integer value
6954 and the second an integer value. The result is also an integer value. These
6955 integer types can have any bit width, but they must all have the same bit
6956 width. The targets may only lower integer representations they support.</p>
6959 <p>These intrinsics does a series of operations atomically. They first load the
6960 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
6961 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
6962 yield the original value stored at <tt>ptr</tt>.</p>
6966 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6967 %ptr = bitcast i8* %mallocP to i32*
6969 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
6970 <i>; yields {i32}:result0 = 7</i>
6971 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
6972 <i>; yields {i32}:result1 = -2</i>
6973 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
6974 <i>; yields {i32}:result2 = 8</i>
6975 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
6976 <i>; yields {i32}:result3 = 8</i>
6977 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6983 <!-- ======================================================================= -->
6984 <div class="doc_subsection">
6985 <a name="int_memorymarkers">Memory Use Markers</a>
6988 <div class="doc_text">
6990 <p>This class of intrinsics exists to information about the lifetime of memory
6991 objects and ranges where variables are immutable.</p>
6995 <!-- _______________________________________________________________________ -->
6996 <div class="doc_subsubsection">
6997 <a name="int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a>
7000 <div class="doc_text">
7004 declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
7008 <p>The '<tt>llvm.lifetime.start</tt>' intrinsic specifies the start of a memory
7009 object's lifetime.</p>
7012 <p>The first argument is a constant integer representing the size of the
7013 object, or -1 if it is variable sized. The second argument is a pointer to
7017 <p>This intrinsic indicates that before this point in the code, the value of the
7018 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7019 never be used and has an undefined value. A load from the pointer that
7020 precedes this intrinsic can be replaced with
7021 <tt>'<a href="#undefvalues">undef</a>'</tt>.</p>
7025 <!-- _______________________________________________________________________ -->
7026 <div class="doc_subsubsection">
7027 <a name="int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a>
7030 <div class="doc_text">
7034 declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
7038 <p>The '<tt>llvm.lifetime.end</tt>' intrinsic specifies the end of a memory
7039 object's lifetime.</p>
7042 <p>The first argument is a constant integer representing the size of the
7043 object, or -1 if it is variable sized. The second argument is a pointer to
7047 <p>This intrinsic indicates that after this point in the code, the value of the
7048 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7049 never be used and has an undefined value. Any stores into the memory object
7050 following this intrinsic may be removed as dead.
7054 <!-- _______________________________________________________________________ -->
7055 <div class="doc_subsubsection">
7056 <a name="int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a>
7059 <div class="doc_text">
7063 declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>) readonly
7067 <p>The '<tt>llvm.invariant.start</tt>' intrinsic specifies that the contents of
7068 a memory object will not change.</p>
7071 <p>The first argument is a constant integer representing the size of the
7072 object, or -1 if it is variable sized. The second argument is a pointer to
7076 <p>This intrinsic indicates that until an <tt>llvm.invariant.end</tt> that uses
7077 the return value, the referenced memory location is constant and
7082 <!-- _______________________________________________________________________ -->
7083 <div class="doc_subsubsection">
7084 <a name="int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a>
7087 <div class="doc_text">
7091 declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
7095 <p>The '<tt>llvm.invariant.end</tt>' intrinsic specifies that the contents of
7096 a memory object are mutable.</p>
7099 <p>The first argument is the matching <tt>llvm.invariant.start</tt> intrinsic.
7100 The second argument is a constant integer representing the size of the
7101 object, or -1 if it is variable sized and the third argument is a pointer
7105 <p>This intrinsic indicates that the memory is mutable again.</p>
7109 <!-- ======================================================================= -->
7110 <div class="doc_subsection">
7111 <a name="int_general">General Intrinsics</a>
7114 <div class="doc_text">
7116 <p>This class of intrinsics is designed to be generic and has no specific
7121 <!-- _______________________________________________________________________ -->
7122 <div class="doc_subsubsection">
7123 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
7126 <div class="doc_text">
7130 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
7134 <p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
7137 <p>The first argument is a pointer to a value, the second is a pointer to a
7138 global string, the third is a pointer to a global string which is the source
7139 file name, and the last argument is the line number.</p>
7142 <p>This intrinsic allows annotation of local variables with arbitrary strings.
7143 This can be useful for special purpose optimizations that want to look for
7144 these annotations. These have no other defined use, they are ignored by code
7145 generation and optimization.</p>
7149 <!-- _______________________________________________________________________ -->
7150 <div class="doc_subsubsection">
7151 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
7154 <div class="doc_text">
7157 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
7158 any integer bit width.</p>
7161 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
7162 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
7163 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
7164 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
7165 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
7169 <p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
7172 <p>The first argument is an integer value (result of some expression), the
7173 second is a pointer to a global string, the third is a pointer to a global
7174 string which is the source file name, and the last argument is the line
7175 number. It returns the value of the first argument.</p>
7178 <p>This intrinsic allows annotations to be put on arbitrary expressions with
7179 arbitrary strings. This can be useful for special purpose optimizations that
7180 want to look for these annotations. These have no other defined use, they
7181 are ignored by code generation and optimization.</p>
7185 <!-- _______________________________________________________________________ -->
7186 <div class="doc_subsubsection">
7187 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
7190 <div class="doc_text">
7194 declare void @llvm.trap()
7198 <p>The '<tt>llvm.trap</tt>' intrinsic.</p>
7204 <p>This intrinsics is lowered to the target dependent trap instruction. If the
7205 target does not have a trap instruction, this intrinsic will be lowered to
7206 the call of the <tt>abort()</tt> function.</p>
7210 <!-- _______________________________________________________________________ -->
7211 <div class="doc_subsubsection">
7212 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
7215 <div class="doc_text">
7219 declare void @llvm.stackprotector( i8* <guard>, i8** <slot> )
7223 <p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
7224 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
7225 ensure that it is placed on the stack before local variables.</p>
7228 <p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
7229 arguments. The first argument is the value loaded from the stack
7230 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
7231 that has enough space to hold the value of the guard.</p>
7234 <p>This intrinsic causes the prologue/epilogue inserter to force the position of
7235 the <tt>AllocaInst</tt> stack slot to be before local variables on the
7236 stack. This is to ensure that if a local variable on the stack is
7237 overwritten, it will destroy the value of the guard. When the function exits,
7238 the guard on the stack is checked against the original guard. If they're
7239 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
7244 <!-- *********************************************************************** -->
7247 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
7248 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
7249 <a href="http://validator.w3.org/check/referer"><img
7250 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
7252 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
7253 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
7254 Last modified: $Date$