1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>LLVM Assembly Language Reference Manual</title>
6 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
8 <meta name="description"
9 content="LLVM Assembly Language Reference Manual.">
10 <link rel="stylesheet" href="llvm.css" type="text/css">
15 <div class="doc_title"> LLVM Language Reference Manual </div>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
20 <li><a href="#highlevel">High Level Structure</a>
22 <li><a href="#modulestructure">Module Structure</a></li>
23 <li><a href="#linkage">Linkage Types</a>
25 <li><a href="#linkage_private">'<tt>private</tt>' Linkage</a></li>
26 <li><a href="#linkage_linker_private">'<tt>linker_private</tt>' Linkage</a></li>
27 <li><a href="#linkage_internal">'<tt>internal</tt>' Linkage</a></li>
28 <li><a href="#linkage_available_externally">'<tt>available_externally</tt>' Linkage</a></li>
29 <li><a href="#linkage_linkonce">'<tt>linkonce</tt>' Linkage</a></li>
30 <li><a href="#linkage_common">'<tt>common</tt>' Linkage</a></li>
31 <li><a href="#linkage_weak">'<tt>weak</tt>' Linkage</a></li>
32 <li><a href="#linkage_appending">'<tt>appending</tt>' Linkage</a></li>
33 <li><a href="#linkage_externweak">'<tt>extern_weak</tt>' Linkage</a></li>
34 <li><a href="#linkage_linkonce_odr">'<tt>linkonce_odr</tt>' Linkage</a></li>
35 <li><a href="#linkage_weak">'<tt>weak_odr</tt>' Linkage</a></li>
36 <li><a href="#linkage_external">'<tt>externally visible</tt>' Linkage</a></li>
37 <li><a href="#linkage_dllimport">'<tt>dllimport</tt>' Linkage</a></li>
38 <li><a href="#linkage_dllexport">'<tt>dllexport</tt>' Linkage</a></li>
41 <li><a href="#callingconv">Calling Conventions</a></li>
42 <li><a href="#namedtypes">Named Types</a></li>
43 <li><a href="#globalvars">Global Variables</a></li>
44 <li><a href="#functionstructure">Functions</a></li>
45 <li><a href="#aliasstructure">Aliases</a></li>
46 <li><a href="#paramattrs">Parameter Attributes</a></li>
47 <li><a href="#fnattrs">Function Attributes</a></li>
48 <li><a href="#gc">Garbage Collector Names</a></li>
49 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
50 <li><a href="#datalayout">Data Layout</a></li>
51 <li><a href="#pointeraliasing">Pointer Aliasing Rules</a></li>
54 <li><a href="#typesystem">Type System</a>
56 <li><a href="#t_classifications">Type Classifications</a></li>
57 <li><a href="#t_primitive">Primitive Types</a>
59 <li><a href="#t_integer">Integer Type</a></li>
60 <li><a href="#t_floating">Floating Point Types</a></li>
61 <li><a href="#t_void">Void Type</a></li>
62 <li><a href="#t_label">Label Type</a></li>
63 <li><a href="#t_metadata">Metadata Type</a></li>
66 <li><a href="#t_derived">Derived Types</a>
68 <li><a href="#t_array">Array Type</a></li>
69 <li><a href="#t_function">Function Type</a></li>
70 <li><a href="#t_pointer">Pointer Type</a></li>
71 <li><a href="#t_struct">Structure Type</a></li>
72 <li><a href="#t_pstruct">Packed Structure Type</a></li>
73 <li><a href="#t_vector">Vector Type</a></li>
74 <li><a href="#t_opaque">Opaque Type</a></li>
77 <li><a href="#t_uprefs">Type Up-references</a></li>
80 <li><a href="#constants">Constants</a>
82 <li><a href="#simpleconstants">Simple Constants</a></li>
83 <li><a href="#complexconstants">Complex Constants</a></li>
84 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
85 <li><a href="#undefvalues">Undefined Values</a></li>
86 <li><a href="#blockaddress">Addresses of Basic Blocks</a></li>
87 <li><a href="#constantexprs">Constant Expressions</a></li>
88 <li><a href="#metadata">Embedded Metadata</a></li>
91 <li><a href="#othervalues">Other Values</a>
93 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
96 <li><a href="#intrinsic_globals">Intrinsic Global Variables</a>
98 <li><a href="#intg_used">The '<tt>llvm.used</tt>' Global Variable</a></li>
99 <li><a href="#intg_compiler_used">The '<tt>llvm.compiler.used</tt>'
100 Global Variable</a></li>
101 <li><a href="#intg_global_ctors">The '<tt>llvm.global_ctors</tt>'
102 Global Variable</a></li>
103 <li><a href="#intg_global_dtors">The '<tt>llvm.global_dtors</tt>'
104 Global Variable</a></li>
107 <li><a href="#instref">Instruction Reference</a>
109 <li><a href="#terminators">Terminator Instructions</a>
111 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
112 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
113 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
114 <li><a href="#i_indirectbr">'<tt>indirectbr</tt>' Instruction</a></li>
115 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
116 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
117 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
120 <li><a href="#binaryops">Binary Operations</a>
122 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
123 <li><a href="#i_fadd">'<tt>fadd</tt>' Instruction</a></li>
124 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
125 <li><a href="#i_fsub">'<tt>fsub</tt>' Instruction</a></li>
126 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
127 <li><a href="#i_fmul">'<tt>fmul</tt>' Instruction</a></li>
128 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
129 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
130 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
131 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
132 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
133 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
136 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
138 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
139 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
140 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
141 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
142 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
143 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
146 <li><a href="#vectorops">Vector Operations</a>
148 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
149 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
150 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
153 <li><a href="#aggregateops">Aggregate Operations</a>
155 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
156 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
159 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
161 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
162 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
163 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
164 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
167 <li><a href="#convertops">Conversion Operations</a>
169 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
170 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
171 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
172 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
173 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
174 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
175 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
176 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
177 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
178 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
179 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
180 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
183 <li><a href="#otherops">Other Operations</a>
185 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
186 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
187 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
188 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
189 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
190 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
195 <li><a href="#intrinsics">Intrinsic Functions</a>
197 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
199 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
200 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
201 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
204 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
206 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
207 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
208 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
211 <li><a href="#int_codegen">Code Generator Intrinsics</a>
213 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
214 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
215 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
216 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
217 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
218 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
219 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
222 <li><a href="#int_libc">Standard C Library Intrinsics</a>
224 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
225 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
226 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
227 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
228 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
229 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
230 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
231 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
234 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
236 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
237 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
238 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
239 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
242 <li><a href="#int_overflow">Arithmetic with Overflow Intrinsics</a>
244 <li><a href="#int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt> Intrinsics</a></li>
245 <li><a href="#int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt> Intrinsics</a></li>
246 <li><a href="#int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt> Intrinsics</a></li>
247 <li><a href="#int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt> Intrinsics</a></li>
248 <li><a href="#int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt> Intrinsics</a></li>
249 <li><a href="#int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt> Intrinsics</a></li>
252 <li><a href="#int_debugger">Debugger intrinsics</a></li>
253 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
254 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
256 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
259 <li><a href="#int_atomics">Atomic intrinsics</a>
261 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
262 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
263 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
264 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
265 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
266 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
267 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
268 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
269 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
270 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
271 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
272 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
273 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
276 <li><a href="#int_memorymarkers">Memory Use Markers</a>
278 <li><a href="#int_lifetime_start"><tt>llvm.lifetime.start</tt></a></li>
279 <li><a href="#int_lifetime_end"><tt>llvm.lifetime.end</tt></a></li>
280 <li><a href="#int_invariant_start"><tt>llvm.invariant.start</tt></a></li>
281 <li><a href="#int_invariant_end"><tt>llvm.invariant.end</tt></a></li>
284 <li><a href="#int_general">General intrinsics</a>
286 <li><a href="#int_var_annotation">
287 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
288 <li><a href="#int_annotation">
289 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
290 <li><a href="#int_trap">
291 '<tt>llvm.trap</tt>' Intrinsic</a></li>
292 <li><a href="#int_stackprotector">
293 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
300 <div class="doc_author">
301 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
302 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
305 <!-- *********************************************************************** -->
306 <div class="doc_section"> <a name="abstract">Abstract </a></div>
307 <!-- *********************************************************************** -->
309 <div class="doc_text">
311 <p>This document is a reference manual for the LLVM assembly language. LLVM is
312 a Static Single Assignment (SSA) based representation that provides type
313 safety, low-level operations, flexibility, and the capability of representing
314 'all' high-level languages cleanly. It is the common code representation
315 used throughout all phases of the LLVM compilation strategy.</p>
319 <!-- *********************************************************************** -->
320 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
321 <!-- *********************************************************************** -->
323 <div class="doc_text">
325 <p>The LLVM code representation is designed to be used in three different forms:
326 as an in-memory compiler IR, as an on-disk bitcode representation (suitable
327 for fast loading by a Just-In-Time compiler), and as a human readable
328 assembly language representation. This allows LLVM to provide a powerful
329 intermediate representation for efficient compiler transformations and
330 analysis, while providing a natural means to debug and visualize the
331 transformations. The three different forms of LLVM are all equivalent. This
332 document describes the human readable representation and notation.</p>
334 <p>The LLVM representation aims to be light-weight and low-level while being
335 expressive, typed, and extensible at the same time. It aims to be a
336 "universal IR" of sorts, by being at a low enough level that high-level ideas
337 may be cleanly mapped to it (similar to how microprocessors are "universal
338 IR's", allowing many source languages to be mapped to them). By providing
339 type information, LLVM can be used as the target of optimizations: for
340 example, through pointer analysis, it can be proven that a C automatic
341 variable is never accessed outside of the current function... allowing it to
342 be promoted to a simple SSA value instead of a memory location.</p>
346 <!-- _______________________________________________________________________ -->
347 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
349 <div class="doc_text">
351 <p>It is important to note that this document describes 'well formed' LLVM
352 assembly language. There is a difference between what the parser accepts and
353 what is considered 'well formed'. For example, the following instruction is
354 syntactically okay, but not well formed:</p>
356 <div class="doc_code">
358 %x = <a href="#i_add">add</a> i32 1, %x
362 <p>...because the definition of <tt>%x</tt> does not dominate all of its
363 uses. The LLVM infrastructure provides a verification pass that may be used
364 to verify that an LLVM module is well formed. This pass is automatically run
365 by the parser after parsing input assembly and by the optimizer before it
366 outputs bitcode. The violations pointed out by the verifier pass indicate
367 bugs in transformation passes or input to the parser.</p>
371 <!-- Describe the typesetting conventions here. -->
373 <!-- *********************************************************************** -->
374 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
375 <!-- *********************************************************************** -->
377 <div class="doc_text">
379 <p>LLVM identifiers come in two basic types: global and local. Global
380 identifiers (functions, global variables) begin with the <tt>'@'</tt>
381 character. Local identifiers (register names, types) begin with
382 the <tt>'%'</tt> character. Additionally, there are three different formats
383 for identifiers, for different purposes:</p>
386 <li>Named values are represented as a string of characters with their prefix.
387 For example, <tt>%foo</tt>, <tt>@DivisionByZero</tt>,
388 <tt>%a.really.long.identifier</tt>. The actual regular expression used is
389 '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers which require
390 other characters in their names can be surrounded with quotes. Special
391 characters may be escaped using <tt>"\xx"</tt> where <tt>xx</tt> is the
392 ASCII code for the character in hexadecimal. In this way, any character
393 can be used in a name value, even quotes themselves.</li>
395 <li>Unnamed values are represented as an unsigned numeric value with their
396 prefix. For example, <tt>%12</tt>, <tt>@2</tt>, <tt>%44</tt>.</li>
398 <li>Constants, which are described in a <a href="#constants">section about
399 constants</a>, below.</li>
402 <p>LLVM requires that values start with a prefix for two reasons: Compilers
403 don't need to worry about name clashes with reserved words, and the set of
404 reserved words may be expanded in the future without penalty. Additionally,
405 unnamed identifiers allow a compiler to quickly come up with a temporary
406 variable without having to avoid symbol table conflicts.</p>
408 <p>Reserved words in LLVM are very similar to reserved words in other
409 languages. There are keywords for different opcodes
410 ('<tt><a href="#i_add">add</a></tt>',
411 '<tt><a href="#i_bitcast">bitcast</a></tt>',
412 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names
413 ('<tt><a href="#t_void">void</a></tt>',
414 '<tt><a href="#t_primitive">i32</a></tt>', etc...), and others. These
415 reserved words cannot conflict with variable names, because none of them
416 start with a prefix character (<tt>'%'</tt> or <tt>'@'</tt>).</p>
418 <p>Here is an example of LLVM code to multiply the integer variable
419 '<tt>%X</tt>' by 8:</p>
423 <div class="doc_code">
425 %result = <a href="#i_mul">mul</a> i32 %X, 8
429 <p>After strength reduction:</p>
431 <div class="doc_code">
433 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
437 <p>And the hard way:</p>
439 <div class="doc_code">
441 %0 = <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
442 %1 = <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
443 %result = <a href="#i_add">add</a> i32 %1, %1
447 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several important
448 lexical features of LLVM:</p>
451 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
454 <li>Unnamed temporaries are created when the result of a computation is not
455 assigned to a named value.</li>
457 <li>Unnamed temporaries are numbered sequentially</li>
460 <p>...and it also shows a convention that we follow in this document. When
461 demonstrating instructions, we will follow an instruction with a comment that
462 defines the type and name of value produced. Comments are shown in italic
467 <!-- *********************************************************************** -->
468 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
469 <!-- *********************************************************************** -->
471 <!-- ======================================================================= -->
472 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
475 <div class="doc_text">
477 <p>LLVM programs are composed of "Module"s, each of which is a translation unit
478 of the input programs. Each module consists of functions, global variables,
479 and symbol table entries. Modules may be combined together with the LLVM
480 linker, which merges function (and global variable) definitions, resolves
481 forward declarations, and merges symbol table entries. Here is an example of
482 the "hello world" module:</p>
484 <div class="doc_code">
485 <pre><i>; Declare the string constant as a global constant...</i>
486 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
487 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
489 <i>; External declaration of the puts function</i>
490 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
492 <i>; Definition of main function</i>
493 define i32 @main() { <i>; i32()* </i>
494 <i>; Convert [13 x i8]* to i8 *...</i>
496 href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
498 <i>; Call puts function to write out the string to stdout...</i>
500 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
502 href="#i_ret">ret</a> i32 0<br>}<br>
506 <p>This example is made up of a <a href="#globalvars">global variable</a> named
507 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function, and
508 a <a href="#functionstructure">function definition</a> for
511 <p>In general, a module is made up of a list of global values, where both
512 functions and global variables are global values. Global values are
513 represented by a pointer to a memory location (in this case, a pointer to an
514 array of char, and a pointer to a function), and have one of the
515 following <a href="#linkage">linkage types</a>.</p>
519 <!-- ======================================================================= -->
520 <div class="doc_subsection">
521 <a name="linkage">Linkage Types</a>
524 <div class="doc_text">
526 <p>All Global Variables and Functions have one of the following types of
530 <dt><tt><b><a name="linkage_private">private</a></b></tt>: </dt>
531 <dd>Global values with private linkage are only directly accessible by objects
532 in the current module. In particular, linking code into a module with an
533 private global value may cause the private to be renamed as necessary to
534 avoid collisions. Because the symbol is private to the module, all
535 references can be updated. This doesn't show up in any symbol table in the
538 <dt><tt><b><a name="linkage_linker_private">linker_private</a></b></tt>: </dt>
539 <dd>Similar to private, but the symbol is passed through the assembler and
540 removed by the linker after evaluation. Note that (unlike private
541 symbols) linker_private symbols are subject to coalescing by the linker:
542 weak symbols get merged and redefinitions are rejected. However, unlike
543 normal strong symbols, they are removed by the linker from the final
544 linked image (executable or dynamic library).</dd>
546 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
547 <dd>Similar to private, but the value shows as a local symbol
548 (<tt>STB_LOCAL</tt> in the case of ELF) in the object file. This
549 corresponds to the notion of the '<tt>static</tt>' keyword in C.</dd>
551 <dt><tt><b><a name="linkage_available_externally">available_externally</a></b></tt>: </dt>
552 <dd>Globals with "<tt>available_externally</tt>" linkage are never emitted
553 into the object file corresponding to the LLVM module. They exist to
554 allow inlining and other optimizations to take place given knowledge of
555 the definition of the global, which is known to be somewhere outside the
556 module. Globals with <tt>available_externally</tt> linkage are allowed to
557 be discarded at will, and are otherwise the same as <tt>linkonce_odr</tt>.
558 This linkage type is only allowed on definitions, not declarations.</dd>
560 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
561 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
562 the same name when linkage occurs. This is typically used to implement
563 inline functions, templates, or other code which must be generated in each
564 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
565 allowed to be discarded.</dd>
567 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
568 <dd>"<tt>weak</tt>" linkage has the same merging semantics as
569 <tt>linkonce</tt> linkage, except that unreferenced globals with
570 <tt>weak</tt> linkage may not be discarded. This is used for globals that
571 are declared "weak" in C source code.</dd>
573 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
574 <dd>"<tt>common</tt>" linkage is most similar to "<tt>weak</tt>" linkage, but
575 they are used for tentative definitions in C, such as "<tt>int X;</tt>" at
577 Symbols with "<tt>common</tt>" linkage are merged in the same way as
578 <tt>weak symbols</tt>, and they may not be deleted if unreferenced.
579 <tt>common</tt> symbols may not have an explicit section,
580 must have a zero initializer, and may not be marked '<a
581 href="#globalvars"><tt>constant</tt></a>'. Functions and aliases may not
582 have common linkage.</dd>
585 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
586 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
587 pointer to array type. When two global variables with appending linkage
588 are linked together, the two global arrays are appended together. This is
589 the LLVM, typesafe, equivalent of having the system linker append together
590 "sections" with identical names when .o files are linked.</dd>
592 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
593 <dd>The semantics of this linkage follow the ELF object file model: the symbol
594 is weak until linked, if not linked, the symbol becomes null instead of
595 being an undefined reference.</dd>
597 <dt><tt><b><a name="linkage_linkonce_odr">linkonce_odr</a></b></tt>: </dt>
598 <dt><tt><b><a name="linkage_weak_odr">weak_odr</a></b></tt>: </dt>
599 <dd>Some languages allow differing globals to be merged, such as two functions
600 with different semantics. Other languages, such as <tt>C++</tt>, ensure
601 that only equivalent globals are ever merged (the "one definition rule" -
602 "ODR"). Such languages can use the <tt>linkonce_odr</tt>
603 and <tt>weak_odr</tt> linkage types to indicate that the global will only
604 be merged with equivalent globals. These linkage types are otherwise the
605 same as their non-<tt>odr</tt> versions.</dd>
607 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
608 <dd>If none of the above identifiers are used, the global is externally
609 visible, meaning that it participates in linkage and can be used to
610 resolve external symbol references.</dd>
613 <p>The next two types of linkage are targeted for Microsoft Windows platform
614 only. They are designed to support importing (exporting) symbols from (to)
615 DLLs (Dynamic Link Libraries).</p>
618 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
619 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
620 or variable via a global pointer to a pointer that is set up by the DLL
621 exporting the symbol. On Microsoft Windows targets, the pointer name is
622 formed by combining <code>__imp_</code> and the function or variable
625 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
626 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
627 pointer to a pointer in a DLL, so that it can be referenced with the
628 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
629 name is formed by combining <code>__imp_</code> and the function or
633 <p>For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
634 another module defined a "<tt>.LC0</tt>" variable and was linked with this
635 one, one of the two would be renamed, preventing a collision. Since
636 "<tt>main</tt>" and "<tt>puts</tt>" are external (i.e., lacking any linkage
637 declarations), they are accessible outside of the current module.</p>
639 <p>It is illegal for a function <i>declaration</i> to have any linkage type
640 other than "externally visible", <tt>dllimport</tt>
641 or <tt>extern_weak</tt>.</p>
643 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt>, <tt>weak</tt>
644 or <tt>weak_odr</tt> linkages.</p>
648 <!-- ======================================================================= -->
649 <div class="doc_subsection">
650 <a name="callingconv">Calling Conventions</a>
653 <div class="doc_text">
655 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
656 and <a href="#i_invoke">invokes</a> can all have an optional calling
657 convention specified for the call. The calling convention of any pair of
658 dynamic caller/callee must match, or the behavior of the program is
659 undefined. The following calling conventions are supported by LLVM, and more
660 may be added in the future:</p>
663 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
664 <dd>This calling convention (the default if no other calling convention is
665 specified) matches the target C calling conventions. This calling
666 convention supports varargs function calls and tolerates some mismatch in
667 the declared prototype and implemented declaration of the function (as
670 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
671 <dd>This calling convention attempts to make calls as fast as possible
672 (e.g. by passing things in registers). This calling convention allows the
673 target to use whatever tricks it wants to produce fast code for the
674 target, without having to conform to an externally specified ABI
675 (Application Binary Interface). Implementations of this convention should
676 allow arbitrary <a href="CodeGenerator.html#tailcallopt">tail call
677 optimization</a> to be supported. This calling convention does not
678 support varargs and requires the prototype of all callees to exactly match
679 the prototype of the function definition.</dd>
681 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
682 <dd>This calling convention attempts to make code in the caller as efficient
683 as possible under the assumption that the call is not commonly executed.
684 As such, these calls often preserve all registers so that the call does
685 not break any live ranges in the caller side. This calling convention
686 does not support varargs and requires the prototype of all callees to
687 exactly match the prototype of the function definition.</dd>
689 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
690 <dd>Any calling convention may be specified by number, allowing
691 target-specific calling conventions to be used. Target specific calling
692 conventions start at 64.</dd>
695 <p>More calling conventions can be added/defined on an as-needed basis, to
696 support Pascal conventions or any other well-known target-independent
701 <!-- ======================================================================= -->
702 <div class="doc_subsection">
703 <a name="visibility">Visibility Styles</a>
706 <div class="doc_text">
708 <p>All Global Variables and Functions have one of the following visibility
712 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
713 <dd>On targets that use the ELF object file format, default visibility means
714 that the declaration is visible to other modules and, in shared libraries,
715 means that the declared entity may be overridden. On Darwin, default
716 visibility means that the declaration is visible to other modules. Default
717 visibility corresponds to "external linkage" in the language.</dd>
719 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
720 <dd>Two declarations of an object with hidden visibility refer to the same
721 object if they are in the same shared object. Usually, hidden visibility
722 indicates that the symbol will not be placed into the dynamic symbol
723 table, so no other module (executable or shared library) can reference it
726 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
727 <dd>On ELF, protected visibility indicates that the symbol will be placed in
728 the dynamic symbol table, but that references within the defining module
729 will bind to the local symbol. That is, the symbol cannot be overridden by
735 <!-- ======================================================================= -->
736 <div class="doc_subsection">
737 <a name="namedtypes">Named Types</a>
740 <div class="doc_text">
742 <p>LLVM IR allows you to specify name aliases for certain types. This can make
743 it easier to read the IR and make the IR more condensed (particularly when
744 recursive types are involved). An example of a name specification is:</p>
746 <div class="doc_code">
748 %mytype = type { %mytype*, i32 }
752 <p>You may give a name to any <a href="#typesystem">type</a> except
753 "<a href="t_void">void</a>". Type name aliases may be used anywhere a type
754 is expected with the syntax "%mytype".</p>
756 <p>Note that type names are aliases for the structural type that they indicate,
757 and that you can therefore specify multiple names for the same type. This
758 often leads to confusing behavior when dumping out a .ll file. Since LLVM IR
759 uses structural typing, the name is not part of the type. When printing out
760 LLVM IR, the printer will pick <em>one name</em> to render all types of a
761 particular shape. This means that if you have code where two different
762 source types end up having the same LLVM type, that the dumper will sometimes
763 print the "wrong" or unexpected type. This is an important design point and
764 isn't going to change.</p>
768 <!-- ======================================================================= -->
769 <div class="doc_subsection">
770 <a name="globalvars">Global Variables</a>
773 <div class="doc_text">
775 <p>Global variables define regions of memory allocated at compilation time
776 instead of run-time. Global variables may optionally be initialized, may
777 have an explicit section to be placed in, and may have an optional explicit
778 alignment specified. A variable may be defined as "thread_local", which
779 means that it will not be shared by threads (each thread will have a
780 separated copy of the variable). A variable may be defined as a global
781 "constant," which indicates that the contents of the variable
782 will <b>never</b> be modified (enabling better optimization, allowing the
783 global data to be placed in the read-only section of an executable, etc).
784 Note that variables that need runtime initialization cannot be marked
785 "constant" as there is a store to the variable.</p>
787 <p>LLVM explicitly allows <em>declarations</em> of global variables to be marked
788 constant, even if the final definition of the global is not. This capability
789 can be used to enable slightly better optimization of the program, but
790 requires the language definition to guarantee that optimizations based on the
791 'constantness' are valid for the translation units that do not include the
794 <p>As SSA values, global variables define pointer values that are in scope
795 (i.e. they dominate) all basic blocks in the program. Global variables
796 always define a pointer to their "content" type because they describe a
797 region of memory, and all memory objects in LLVM are accessed through
800 <p>A global variable may be declared to reside in a target-specific numbered
801 address space. For targets that support them, address spaces may affect how
802 optimizations are performed and/or what target instructions are used to
803 access the variable. The default address space is zero. The address space
804 qualifier must precede any other attributes.</p>
806 <p>LLVM allows an explicit section to be specified for globals. If the target
807 supports it, it will emit globals to the section specified.</p>
809 <p>An explicit alignment may be specified for a global. If not present, or if
810 the alignment is set to zero, the alignment of the global is set by the
811 target to whatever it feels convenient. If an explicit alignment is
812 specified, the global is forced to have at least that much alignment. All
813 alignments must be a power of 2.</p>
815 <p>For example, the following defines a global in a numbered address space with
816 an initializer, section, and alignment:</p>
818 <div class="doc_code">
820 @G = addrspace(5) constant float 1.0, section "foo", align 4
827 <!-- ======================================================================= -->
828 <div class="doc_subsection">
829 <a name="functionstructure">Functions</a>
832 <div class="doc_text">
834 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord, an
835 optional <a href="#linkage">linkage type</a>, an optional
836 <a href="#visibility">visibility style</a>, an optional
837 <a href="#callingconv">calling convention</a>, a return type, an optional
838 <a href="#paramattrs">parameter attribute</a> for the return type, a function
839 name, a (possibly empty) argument list (each with optional
840 <a href="#paramattrs">parameter attributes</a>), optional
841 <a href="#fnattrs">function attributes</a>, an optional section, an optional
842 alignment, an optional <a href="#gc">garbage collector name</a>, an opening
843 curly brace, a list of basic blocks, and a closing curly brace.</p>
845 <p>LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
846 optional <a href="#linkage">linkage type</a>, an optional
847 <a href="#visibility">visibility style</a>, an optional
848 <a href="#callingconv">calling convention</a>, a return type, an optional
849 <a href="#paramattrs">parameter attribute</a> for the return type, a function
850 name, a possibly empty list of arguments, an optional alignment, and an
851 optional <a href="#gc">garbage collector name</a>.</p>
853 <p>A function definition contains a list of basic blocks, forming the CFG
854 (Control Flow Graph) for the function. Each basic block may optionally start
855 with a label (giving the basic block a symbol table entry), contains a list
856 of instructions, and ends with a <a href="#terminators">terminator</a>
857 instruction (such as a branch or function return).</p>
859 <p>The first basic block in a function is special in two ways: it is immediately
860 executed on entrance to the function, and it is not allowed to have
861 predecessor basic blocks (i.e. there can not be any branches to the entry
862 block of a function). Because the block can have no predecessors, it also
863 cannot have any <a href="#i_phi">PHI nodes</a>.</p>
865 <p>LLVM allows an explicit section to be specified for functions. If the target
866 supports it, it will emit functions to the section specified.</p>
868 <p>An explicit alignment may be specified for a function. If not present, or if
869 the alignment is set to zero, the alignment of the function is set by the
870 target to whatever it feels convenient. If an explicit alignment is
871 specified, the function is forced to have at least that much alignment. All
872 alignments must be a power of 2.</p>
875 <div class="doc_code">
877 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
878 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
879 <ResultType> @<FunctionName> ([argument list])
880 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
881 [<a href="#gc">gc</a>] { ... }
887 <!-- ======================================================================= -->
888 <div class="doc_subsection">
889 <a name="aliasstructure">Aliases</a>
892 <div class="doc_text">
894 <p>Aliases act as "second name" for the aliasee value (which can be either
895 function, global variable, another alias or bitcast of global value). Aliases
896 may have an optional <a href="#linkage">linkage type</a>, and an
897 optional <a href="#visibility">visibility style</a>.</p>
900 <div class="doc_code">
902 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
908 <!-- ======================================================================= -->
909 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
911 <div class="doc_text">
913 <p>The return type and each parameter of a function type may have a set of
914 <i>parameter attributes</i> associated with them. Parameter attributes are
915 used to communicate additional information about the result or parameters of
916 a function. Parameter attributes are considered to be part of the function,
917 not of the function type, so functions with different parameter attributes
918 can have the same function type.</p>
920 <p>Parameter attributes are simple keywords that follow the type specified. If
921 multiple parameter attributes are needed, they are space separated. For
924 <div class="doc_code">
926 declare i32 @printf(i8* noalias nocapture, ...)
927 declare i32 @atoi(i8 zeroext)
928 declare signext i8 @returns_signed_char()
932 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
933 <tt>readonly</tt>) come immediately after the argument list.</p>
935 <p>Currently, only the following parameter attributes are defined:</p>
938 <dt><tt>zeroext</tt></dt>
939 <dd>This indicates to the code generator that the parameter or return value
940 should be zero-extended to a 32-bit value by the caller (for a parameter)
941 or the callee (for a return value).</dd>
943 <dt><tt>signext</tt></dt>
944 <dd>This indicates to the code generator that the parameter or return value
945 should be sign-extended to a 32-bit value by the caller (for a parameter)
946 or the callee (for a return value).</dd>
948 <dt><tt>inreg</tt></dt>
949 <dd>This indicates that this parameter or return value should be treated in a
950 special target-dependent fashion during while emitting code for a function
951 call or return (usually, by putting it in a register as opposed to memory,
952 though some targets use it to distinguish between two different kinds of
953 registers). Use of this attribute is target-specific.</dd>
955 <dt><tt><a name="byval">byval</a></tt></dt>
956 <dd>This indicates that the pointer parameter should really be passed by value
957 to the function. The attribute implies that a hidden copy of the pointee
958 is made between the caller and the callee, so the callee is unable to
959 modify the value in the callee. This attribute is only valid on LLVM
960 pointer arguments. It is generally used to pass structs and arrays by
961 value, but is also valid on pointers to scalars. The copy is considered
962 to belong to the caller not the callee (for example,
963 <tt><a href="#readonly">readonly</a></tt> functions should not write to
964 <tt>byval</tt> parameters). This is not a valid attribute for return
965 values. The byval attribute also supports specifying an alignment with
966 the align attribute. This has a target-specific effect on the code
967 generator that usually indicates a desired alignment for the synthesized
970 <dt><tt>sret</tt></dt>
971 <dd>This indicates that the pointer parameter specifies the address of a
972 structure that is the return value of the function in the source program.
973 This pointer must be guaranteed by the caller to be valid: loads and
974 stores to the structure may be assumed by the callee to not to trap. This
975 may only be applied to the first parameter. This is not a valid attribute
976 for return values. </dd>
978 <dt><tt>noalias</tt></dt>
979 <dd>This indicates that the pointer does not alias any global or any other
980 parameter. The caller is responsible for ensuring that this is the
981 case. On a function return value, <tt>noalias</tt> additionally indicates
982 that the pointer does not alias any other pointers visible to the
983 caller. For further details, please see the discussion of the NoAlias
985 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
988 <dt><tt>nocapture</tt></dt>
989 <dd>This indicates that the callee does not make any copies of the pointer
990 that outlive the callee itself. This is not a valid attribute for return
993 <dt><tt>nest</tt></dt>
994 <dd>This indicates that the pointer parameter can be excised using the
995 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
996 attribute for return values.</dd>
1001 <!-- ======================================================================= -->
1002 <div class="doc_subsection">
1003 <a name="gc">Garbage Collector Names</a>
1006 <div class="doc_text">
1008 <p>Each function may specify a garbage collector name, which is simply a
1011 <div class="doc_code">
1013 define void @f() gc "name" { ...
1017 <p>The compiler declares the supported values of <i>name</i>. Specifying a
1018 collector which will cause the compiler to alter its output in order to
1019 support the named garbage collection algorithm.</p>
1023 <!-- ======================================================================= -->
1024 <div class="doc_subsection">
1025 <a name="fnattrs">Function Attributes</a>
1028 <div class="doc_text">
1030 <p>Function attributes are set to communicate additional information about a
1031 function. Function attributes are considered to be part of the function, not
1032 of the function type, so functions with different parameter attributes can
1033 have the same function type.</p>
1035 <p>Function attributes are simple keywords that follow the type specified. If
1036 multiple attributes are needed, they are space separated. For example:</p>
1038 <div class="doc_code">
1040 define void @f() noinline { ... }
1041 define void @f() alwaysinline { ... }
1042 define void @f() alwaysinline optsize { ... }
1043 define void @f() optsize
1048 <dt><tt>alwaysinline</tt></dt>
1049 <dd>This attribute indicates that the inliner should attempt to inline this
1050 function into callers whenever possible, ignoring any active inlining size
1051 threshold for this caller.</dd>
1053 <dt><tt>inlinehint</tt></dt>
1054 <dd>This attribute indicates that the source code contained a hint that inlining
1055 this function is desirable (such as the "inline" keyword in C/C++). It
1056 is just a hint; it imposes no requirements on the inliner.</dd>
1058 <dt><tt>noinline</tt></dt>
1059 <dd>This attribute indicates that the inliner should never inline this
1060 function in any situation. This attribute may not be used together with
1061 the <tt>alwaysinline</tt> attribute.</dd>
1063 <dt><tt>optsize</tt></dt>
1064 <dd>This attribute suggests that optimization passes and code generator passes
1065 make choices that keep the code size of this function low, and otherwise
1066 do optimizations specifically to reduce code size.</dd>
1068 <dt><tt>noreturn</tt></dt>
1069 <dd>This function attribute indicates that the function never returns
1070 normally. This produces undefined behavior at runtime if the function
1071 ever does dynamically return.</dd>
1073 <dt><tt>nounwind</tt></dt>
1074 <dd>This function attribute indicates that the function never returns with an
1075 unwind or exceptional control flow. If the function does unwind, its
1076 runtime behavior is undefined.</dd>
1078 <dt><tt>readnone</tt></dt>
1079 <dd>This attribute indicates that the function computes its result (or decides
1080 to unwind an exception) based strictly on its arguments, without
1081 dereferencing any pointer arguments or otherwise accessing any mutable
1082 state (e.g. memory, control registers, etc) visible to caller functions.
1083 It does not write through any pointer arguments
1084 (including <tt><a href="#byval">byval</a></tt> arguments) and never
1085 changes any state visible to callers. This means that it cannot unwind
1086 exceptions by calling the <tt>C++</tt> exception throwing methods, but
1087 could use the <tt>unwind</tt> instruction.</dd>
1089 <dt><tt><a name="readonly">readonly</a></tt></dt>
1090 <dd>This attribute indicates that the function does not write through any
1091 pointer arguments (including <tt><a href="#byval">byval</a></tt>
1092 arguments) or otherwise modify any state (e.g. memory, control registers,
1093 etc) visible to caller functions. It may dereference pointer arguments
1094 and read state that may be set in the caller. A readonly function always
1095 returns the same value (or unwinds an exception identically) when called
1096 with the same set of arguments and global state. It cannot unwind an
1097 exception by calling the <tt>C++</tt> exception throwing methods, but may
1098 use the <tt>unwind</tt> instruction.</dd>
1100 <dt><tt><a name="ssp">ssp</a></tt></dt>
1101 <dd>This attribute indicates that the function should emit a stack smashing
1102 protector. It is in the form of a "canary"—a random value placed on
1103 the stack before the local variables that's checked upon return from the
1104 function to see if it has been overwritten. A heuristic is used to
1105 determine if a function needs stack protectors or not.<br>
1107 If a function that has an <tt>ssp</tt> attribute is inlined into a
1108 function that doesn't have an <tt>ssp</tt> attribute, then the resulting
1109 function will have an <tt>ssp</tt> attribute.</dd>
1111 <dt><tt>sspreq</tt></dt>
1112 <dd>This attribute indicates that the function should <em>always</em> emit a
1113 stack smashing protector. This overrides
1114 the <tt><a href="#ssp">ssp</a></tt> function attribute.<br>
1116 If a function that has an <tt>sspreq</tt> attribute is inlined into a
1117 function that doesn't have an <tt>sspreq</tt> attribute or which has
1118 an <tt>ssp</tt> attribute, then the resulting function will have
1119 an <tt>sspreq</tt> attribute.</dd>
1121 <dt><tt>noredzone</tt></dt>
1122 <dd>This attribute indicates that the code generator should not use a red
1123 zone, even if the target-specific ABI normally permits it.</dd>
1125 <dt><tt>noimplicitfloat</tt></dt>
1126 <dd>This attributes disables implicit floating point instructions.</dd>
1128 <dt><tt>naked</tt></dt>
1129 <dd>This attribute disables prologue / epilogue emission for the function.
1130 This can have very system-specific consequences.</dd>
1135 <!-- ======================================================================= -->
1136 <div class="doc_subsection">
1137 <a name="moduleasm">Module-Level Inline Assembly</a>
1140 <div class="doc_text">
1142 <p>Modules may contain "module-level inline asm" blocks, which corresponds to
1143 the GCC "file scope inline asm" blocks. These blocks are internally
1144 concatenated by LLVM and treated as a single unit, but may be separated in
1145 the <tt>.ll</tt> file if desired. The syntax is very simple:</p>
1147 <div class="doc_code">
1149 module asm "inline asm code goes here"
1150 module asm "more can go here"
1154 <p>The strings can contain any character by escaping non-printable characters.
1155 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1158 <p>The inline asm code is simply printed to the machine code .s file when
1159 assembly code is generated.</p>
1163 <!-- ======================================================================= -->
1164 <div class="doc_subsection">
1165 <a name="datalayout">Data Layout</a>
1168 <div class="doc_text">
1170 <p>A module may specify a target specific data layout string that specifies how
1171 data is to be laid out in memory. The syntax for the data layout is
1174 <div class="doc_code">
1176 target datalayout = "<i>layout specification</i>"
1180 <p>The <i>layout specification</i> consists of a list of specifications
1181 separated by the minus sign character ('-'). Each specification starts with
1182 a letter and may include other information after the letter to define some
1183 aspect of the data layout. The specifications accepted are as follows:</p>
1187 <dd>Specifies that the target lays out data in big-endian form. That is, the
1188 bits with the most significance have the lowest address location.</dd>
1191 <dd>Specifies that the target lays out data in little-endian form. That is,
1192 the bits with the least significance have the lowest address
1195 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1196 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1197 <i>preferred</i> alignments. All sizes are in bits. Specifying
1198 the <i>pref</i> alignment is optional. If omitted, the
1199 preceding <tt>:</tt> should be omitted too.</dd>
1201 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1202 <dd>This specifies the alignment for an integer type of a given bit
1203 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1205 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1206 <dd>This specifies the alignment for a vector type of a given bit
1209 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1210 <dd>This specifies the alignment for a floating point type of a given bit
1211 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1214 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1215 <dd>This specifies the alignment for an aggregate type of a given bit
1218 <dt><tt>s<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1219 <dd>This specifies the alignment for a stack object of a given bit
1223 <p>When constructing the data layout for a given target, LLVM starts with a
1224 default set of specifications which are then (possibly) overriden by the
1225 specifications in the <tt>datalayout</tt> keyword. The default specifications
1226 are given in this list:</p>
1229 <li><tt>E</tt> - big endian</li>
1230 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1231 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1232 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1233 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1234 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1235 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1236 alignment of 64-bits</li>
1237 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1238 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1239 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1240 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1241 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1242 <li><tt>s0:64:64</tt> - stack objects are 64-bit aligned</li>
1245 <p>When LLVM is determining the alignment for a given type, it uses the
1246 following rules:</p>
1249 <li>If the type sought is an exact match for one of the specifications, that
1250 specification is used.</li>
1252 <li>If no match is found, and the type sought is an integer type, then the
1253 smallest integer type that is larger than the bitwidth of the sought type
1254 is used. If none of the specifications are larger than the bitwidth then
1255 the the largest integer type is used. For example, given the default
1256 specifications above, the i7 type will use the alignment of i8 (next
1257 largest) while both i65 and i256 will use the alignment of i64 (largest
1260 <li>If no match is found, and the type sought is a vector type, then the
1261 largest vector type that is smaller than the sought vector type will be
1262 used as a fall back. This happens because <128 x double> can be
1263 implemented in terms of 64 <2 x double>, for example.</li>
1268 <!-- ======================================================================= -->
1269 <div class="doc_subsection">
1270 <a name="pointeraliasing">Pointer Aliasing Rules</a>
1273 <div class="doc_text">
1275 <p>Any memory access must be done through a pointer value associated
1276 with an address range of the memory access, otherwise the behavior
1277 is undefined. Pointer values are associated with address ranges
1278 according to the following rules:</p>
1281 <li>A pointer value formed from a
1282 <tt><a href="#i_getelementptr">getelementptr</a></tt> instruction
1283 is associated with the addresses associated with the first operand
1284 of the <tt>getelementptr</tt>.</li>
1285 <li>An address of a global variable is associated with the address
1286 range of the variable's storage.</li>
1287 <li>The result value of an allocation instruction is associated with
1288 the address range of the allocated storage.</li>
1289 <li>A null pointer in the default address-space is associated with
1291 <li>A pointer value formed by an
1292 <tt><a href="#i_inttoptr">inttoptr</a></tt> is associated with all
1293 address ranges of all pointer values that contribute (directly or
1294 indirectly) to the computation of the pointer's value.</li>
1295 <li>The result value of a
1296 <tt><a href="#i_bitcast">bitcast</a></tt> is associated with all
1297 addresses associated with the operand of the <tt>bitcast</tt>.</li>
1298 <li>An integer constant other than zero or a pointer value returned
1299 from a function not defined within LLVM may be associated with address
1300 ranges allocated through mechanisms other than those provided by
1301 LLVM. Such ranges shall not overlap with any ranges of addresses
1302 allocated by mechanisms provided by LLVM.</li>
1305 <p>LLVM IR does not associate types with memory. The result type of a
1306 <tt><a href="#i_load">load</a></tt> merely indicates the size and
1307 alignment of the memory from which to load, as well as the
1308 interpretation of the value. The first operand of a
1309 <tt><a href="#i_store">store</a></tt> similarly only indicates the size
1310 and alignment of the store.</p>
1312 <p>Consequently, type-based alias analysis, aka TBAA, aka
1313 <tt>-fstrict-aliasing</tt>, is not applicable to general unadorned
1314 LLVM IR. <a href="#metadata">Metadata</a> may be used to encode
1315 additional information which specialized optimization passes may use
1316 to implement type-based alias analysis.</p>
1320 <!-- *********************************************************************** -->
1321 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1322 <!-- *********************************************************************** -->
1324 <div class="doc_text">
1326 <p>The LLVM type system is one of the most important features of the
1327 intermediate representation. Being typed enables a number of optimizations
1328 to be performed on the intermediate representation directly, without having
1329 to do extra analyses on the side before the transformation. A strong type
1330 system makes it easier to read the generated code and enables novel analyses
1331 and transformations that are not feasible to perform on normal three address
1332 code representations.</p>
1336 <!-- ======================================================================= -->
1337 <div class="doc_subsection"> <a name="t_classifications">Type
1338 Classifications</a> </div>
1340 <div class="doc_text">
1342 <p>The types fall into a few useful classifications:</p>
1344 <table border="1" cellspacing="0" cellpadding="4">
1346 <tr><th>Classification</th><th>Types</th></tr>
1348 <td><a href="#t_integer">integer</a></td>
1349 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1352 <td><a href="#t_floating">floating point</a></td>
1353 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1356 <td><a name="t_firstclass">first class</a></td>
1357 <td><a href="#t_integer">integer</a>,
1358 <a href="#t_floating">floating point</a>,
1359 <a href="#t_pointer">pointer</a>,
1360 <a href="#t_vector">vector</a>,
1361 <a href="#t_struct">structure</a>,
1362 <a href="#t_array">array</a>,
1363 <a href="#t_label">label</a>,
1364 <a href="#t_metadata">metadata</a>.
1368 <td><a href="#t_primitive">primitive</a></td>
1369 <td><a href="#t_label">label</a>,
1370 <a href="#t_void">void</a>,
1371 <a href="#t_floating">floating point</a>,
1372 <a href="#t_metadata">metadata</a>.</td>
1375 <td><a href="#t_derived">derived</a></td>
1376 <td><a href="#t_integer">integer</a>,
1377 <a href="#t_array">array</a>,
1378 <a href="#t_function">function</a>,
1379 <a href="#t_pointer">pointer</a>,
1380 <a href="#t_struct">structure</a>,
1381 <a href="#t_pstruct">packed structure</a>,
1382 <a href="#t_vector">vector</a>,
1383 <a href="#t_opaque">opaque</a>.
1389 <p>The <a href="#t_firstclass">first class</a> types are perhaps the most
1390 important. Values of these types are the only ones which can be produced by
1395 <!-- ======================================================================= -->
1396 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1398 <div class="doc_text">
1400 <p>The primitive types are the fundamental building blocks of the LLVM
1405 <!-- _______________________________________________________________________ -->
1406 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1408 <div class="doc_text">
1411 <p>The integer type is a very simple type that simply specifies an arbitrary
1412 bit width for the integer type desired. Any bit width from 1 bit to
1413 2<sup>23</sup>-1 (about 8 million) can be specified.</p>
1420 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1424 <table class="layout">
1426 <td class="left"><tt>i1</tt></td>
1427 <td class="left">a single-bit integer.</td>
1430 <td class="left"><tt>i32</tt></td>
1431 <td class="left">a 32-bit integer.</td>
1434 <td class="left"><tt>i1942652</tt></td>
1435 <td class="left">a really big integer of over 1 million bits.</td>
1439 <p>Note that the code generator does not yet support large integer types to be
1440 used as function return types. The specific limit on how large a return type
1441 the code generator can currently handle is target-dependent; currently it's
1442 often 64 bits for 32-bit targets and 128 bits for 64-bit targets.</p>
1446 <!-- _______________________________________________________________________ -->
1447 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1449 <div class="doc_text">
1453 <tr><th>Type</th><th>Description</th></tr>
1454 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1455 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1456 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1457 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1458 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1464 <!-- _______________________________________________________________________ -->
1465 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1467 <div class="doc_text">
1470 <p>The void type does not represent any value and has no size.</p>
1479 <!-- _______________________________________________________________________ -->
1480 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1482 <div class="doc_text">
1485 <p>The label type represents code labels.</p>
1494 <!-- _______________________________________________________________________ -->
1495 <div class="doc_subsubsection"> <a name="t_metadata">Metadata Type</a> </div>
1497 <div class="doc_text">
1500 <p>The metadata type represents embedded metadata. No derived types may be
1501 created from metadata except for <a href="#t_function">function</a>
1512 <!-- ======================================================================= -->
1513 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1515 <div class="doc_text">
1517 <p>The real power in LLVM comes from the derived types in the system. This is
1518 what allows a programmer to represent arrays, functions, pointers, and other
1519 useful types. Each of these types contain one or more element types which
1520 may be a primitive type, or another derived type. For example, it is
1521 possible to have a two dimensional array, using an array as the element type
1522 of another array.</p>
1526 <!-- _______________________________________________________________________ -->
1527 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1529 <div class="doc_text">
1532 <p>The array type is a very simple derived type that arranges elements
1533 sequentially in memory. The array type requires a size (number of elements)
1534 and an underlying data type.</p>
1538 [<# elements> x <elementtype>]
1541 <p>The number of elements is a constant integer value; <tt>elementtype</tt> may
1542 be any type with a size.</p>
1545 <table class="layout">
1547 <td class="left"><tt>[40 x i32]</tt></td>
1548 <td class="left">Array of 40 32-bit integer values.</td>
1551 <td class="left"><tt>[41 x i32]</tt></td>
1552 <td class="left">Array of 41 32-bit integer values.</td>
1555 <td class="left"><tt>[4 x i8]</tt></td>
1556 <td class="left">Array of 4 8-bit integer values.</td>
1559 <p>Here are some examples of multidimensional arrays:</p>
1560 <table class="layout">
1562 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1563 <td class="left">3x4 array of 32-bit integer values.</td>
1566 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1567 <td class="left">12x10 array of single precision floating point values.</td>
1570 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1571 <td class="left">2x3x4 array of 16-bit integer values.</td>
1575 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1576 length array. Normally, accesses past the end of an array are undefined in
1577 LLVM (e.g. it is illegal to access the 5th element of a 3 element array). As
1578 a special case, however, zero length arrays are recognized to be variable
1579 length. This allows implementation of 'pascal style arrays' with the LLVM
1580 type "<tt>{ i32, [0 x float]}</tt>", for example.</p>
1582 <p>Note that the code generator does not yet support large aggregate types to be
1583 used as function return types. The specific limit on how large an aggregate
1584 return type the code generator can currently handle is target-dependent, and
1585 also dependent on the aggregate element types.</p>
1589 <!-- _______________________________________________________________________ -->
1590 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1592 <div class="doc_text">
1595 <p>The function type can be thought of as a function signature. It consists of
1596 a return type and a list of formal parameter types. The return type of a
1597 function type is a scalar type, a void type, or a struct type. If the return
1598 type is a struct type then all struct elements must be of first class types,
1599 and the struct must have at least one element.</p>
1603 <returntype> (<parameter list>)
1606 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1607 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1608 which indicates that the function takes a variable number of arguments.
1609 Variable argument functions can access their arguments with
1610 the <a href="#int_varargs">variable argument handling intrinsic</a>
1611 functions. '<tt><returntype></tt>' is a any type except
1612 <a href="#t_label">label</a>.</p>
1615 <table class="layout">
1617 <td class="left"><tt>i32 (i32)</tt></td>
1618 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1620 </tr><tr class="layout">
1621 <td class="left"><tt>float (i16 signext, i32 *) *
1623 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1624 an <tt>i16</tt> that should be sign extended and a
1625 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1628 </tr><tr class="layout">
1629 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1630 <td class="left">A vararg function that takes at least one
1631 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1632 which returns an integer. This is the signature for <tt>printf</tt> in
1635 </tr><tr class="layout">
1636 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1637 <td class="left">A function taking an <tt>i32</tt>, returning a
1638 <a href="#t_struct">structure</a> containing two <tt>i32</tt> values
1645 <!-- _______________________________________________________________________ -->
1646 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1648 <div class="doc_text">
1651 <p>The structure type is used to represent a collection of data members together
1652 in memory. The packing of the field types is defined to match the ABI of the
1653 underlying processor. The elements of a structure may be any type that has a
1656 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1657 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1658 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1662 { <type list> }
1666 <table class="layout">
1668 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1669 <td class="left">A triple of three <tt>i32</tt> values</td>
1670 </tr><tr class="layout">
1671 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1672 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1673 second element is a <a href="#t_pointer">pointer</a> to a
1674 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1675 an <tt>i32</tt>.</td>
1679 <p>Note that the code generator does not yet support large aggregate types to be
1680 used as function return types. The specific limit on how large an aggregate
1681 return type the code generator can currently handle is target-dependent, and
1682 also dependent on the aggregate element types.</p>
1686 <!-- _______________________________________________________________________ -->
1687 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1690 <div class="doc_text">
1693 <p>The packed structure type is used to represent a collection of data members
1694 together in memory. There is no padding between fields. Further, the
1695 alignment of a packed structure is 1 byte. The elements of a packed
1696 structure may be any type that has a size.</p>
1698 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt> and
1699 '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a field with
1700 the '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
1704 < { <type list> } >
1708 <table class="layout">
1710 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1711 <td class="left">A triple of three <tt>i32</tt> values</td>
1712 </tr><tr class="layout">
1714 <tt>< { float, i32 (i32)* } ></tt></td>
1715 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1716 second element is a <a href="#t_pointer">pointer</a> to a
1717 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1718 an <tt>i32</tt>.</td>
1724 <!-- _______________________________________________________________________ -->
1725 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1727 <div class="doc_text">
1730 <p>As in many languages, the pointer type represents a pointer or reference to
1731 another object, which must live in memory. Pointer types may have an optional
1732 address space attribute defining the target-specific numbered address space
1733 where the pointed-to object resides. The default address space is zero.</p>
1735 <p>Note that LLVM does not permit pointers to void (<tt>void*</tt>) nor does it
1736 permit pointers to labels (<tt>label*</tt>). Use <tt>i8*</tt> instead.</p>
1744 <table class="layout">
1746 <td class="left"><tt>[4 x i32]*</tt></td>
1747 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1748 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1751 <td class="left"><tt>i32 (i32 *) *</tt></td>
1752 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1753 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1757 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1758 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1759 that resides in address space #5.</td>
1765 <!-- _______________________________________________________________________ -->
1766 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1768 <div class="doc_text">
1771 <p>A vector type is a simple derived type that represents a vector of elements.
1772 Vector types are used when multiple primitive data are operated in parallel
1773 using a single instruction (SIMD). A vector type requires a size (number of
1774 elements) and an underlying primitive data type. Vectors must have a power
1775 of two length (1, 2, 4, 8, 16 ...). Vector types are considered
1776 <a href="#t_firstclass">first class</a>.</p>
1780 < <# elements> x <elementtype> >
1783 <p>The number of elements is a constant integer value; elementtype may be any
1784 integer or floating point type.</p>
1787 <table class="layout">
1789 <td class="left"><tt><4 x i32></tt></td>
1790 <td class="left">Vector of 4 32-bit integer values.</td>
1793 <td class="left"><tt><8 x float></tt></td>
1794 <td class="left">Vector of 8 32-bit floating-point values.</td>
1797 <td class="left"><tt><2 x i64></tt></td>
1798 <td class="left">Vector of 2 64-bit integer values.</td>
1802 <p>Note that the code generator does not yet support large vector types to be
1803 used as function return types. The specific limit on how large a vector
1804 return type codegen can currently handle is target-dependent; currently it's
1805 often a few times longer than a hardware vector register.</p>
1809 <!-- _______________________________________________________________________ -->
1810 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1811 <div class="doc_text">
1814 <p>Opaque types are used to represent unknown types in the system. This
1815 corresponds (for example) to the C notion of a forward declared structure
1816 type. In LLVM, opaque types can eventually be resolved to any type (not just
1817 a structure type).</p>
1825 <table class="layout">
1827 <td class="left"><tt>opaque</tt></td>
1828 <td class="left">An opaque type.</td>
1834 <!-- ======================================================================= -->
1835 <div class="doc_subsection">
1836 <a name="t_uprefs">Type Up-references</a>
1839 <div class="doc_text">
1842 <p>An "up reference" allows you to refer to a lexically enclosing type without
1843 requiring it to have a name. For instance, a structure declaration may
1844 contain a pointer to any of the types it is lexically a member of. Example
1845 of up references (with their equivalent as named type declarations)
1849 { \2 * } %x = type { %x* }
1850 { \2 }* %y = type { %y }*
1854 <p>An up reference is needed by the asmprinter for printing out cyclic types
1855 when there is no declared name for a type in the cycle. Because the
1856 asmprinter does not want to print out an infinite type string, it needs a
1857 syntax to handle recursive types that have no names (all names are optional
1865 <p>The level is the count of the lexical type that is being referred to.</p>
1868 <table class="layout">
1870 <td class="left"><tt>\1*</tt></td>
1871 <td class="left">Self-referential pointer.</td>
1874 <td class="left"><tt>{ { \3*, i8 }, i32 }</tt></td>
1875 <td class="left">Recursive structure where the upref refers to the out-most
1882 <!-- *********************************************************************** -->
1883 <div class="doc_section"> <a name="constants">Constants</a> </div>
1884 <!-- *********************************************************************** -->
1886 <div class="doc_text">
1888 <p>LLVM has several different basic types of constants. This section describes
1889 them all and their syntax.</p>
1893 <!-- ======================================================================= -->
1894 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1896 <div class="doc_text">
1899 <dt><b>Boolean constants</b></dt>
1900 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1901 constants of the <tt><a href="#t_integer">i1</a></tt> type.</dd>
1903 <dt><b>Integer constants</b></dt>
1904 <dd>Standard integers (such as '4') are constants of
1905 the <a href="#t_integer">integer</a> type. Negative numbers may be used
1906 with integer types.</dd>
1908 <dt><b>Floating point constants</b></dt>
1909 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1910 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1911 notation (see below). The assembler requires the exact decimal value of a
1912 floating-point constant. For example, the assembler accepts 1.25 but
1913 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1914 constants must have a <a href="#t_floating">floating point</a> type. </dd>
1916 <dt><b>Null pointer constants</b></dt>
1917 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1918 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1921 <p>The one non-intuitive notation for constants is the hexadecimal form of
1922 floating point constants. For example, the form '<tt>double
1923 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than)
1924 '<tt>double 4.5e+15</tt>'. The only time hexadecimal floating point
1925 constants are required (and the only time that they are generated by the
1926 disassembler) is when a floating point constant must be emitted but it cannot
1927 be represented as a decimal floating point number in a reasonable number of
1928 digits. For example, NaN's, infinities, and other special values are
1929 represented in their IEEE hexadecimal format so that assembly and disassembly
1930 do not cause any bits to change in the constants.</p>
1932 <p>When using the hexadecimal form, constants of types float and double are
1933 represented using the 16-digit form shown above (which matches the IEEE754
1934 representation for double); float values must, however, be exactly
1935 representable as IEE754 single precision. Hexadecimal format is always used
1936 for long double, and there are three forms of long double. The 80-bit format
1937 used by x86 is represented as <tt>0xK</tt> followed by 20 hexadecimal digits.
1938 The 128-bit format used by PowerPC (two adjacent doubles) is represented
1939 by <tt>0xM</tt> followed by 32 hexadecimal digits. The IEEE 128-bit format
1940 is represented by <tt>0xL</tt> followed by 32 hexadecimal digits; no
1941 currently supported target uses this format. Long doubles will only work if
1942 they match the long double format on your target. All hexadecimal formats
1943 are big-endian (sign bit at the left).</p>
1947 <!-- ======================================================================= -->
1948 <div class="doc_subsection">
1949 <a name="aggregateconstants"></a> <!-- old anchor -->
1950 <a name="complexconstants">Complex Constants</a>
1953 <div class="doc_text">
1955 <p>Complex constants are a (potentially recursive) combination of simple
1956 constants and smaller complex constants.</p>
1959 <dt><b>Structure constants</b></dt>
1960 <dd>Structure constants are represented with notation similar to structure
1961 type definitions (a comma separated list of elements, surrounded by braces
1962 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1963 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>".
1964 Structure constants must have <a href="#t_struct">structure type</a>, and
1965 the number and types of elements must match those specified by the
1968 <dt><b>Array constants</b></dt>
1969 <dd>Array constants are represented with notation similar to array type
1970 definitions (a comma separated list of elements, surrounded by square
1971 brackets (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74
1972 ]</tt>". Array constants must have <a href="#t_array">array type</a>, and
1973 the number and types of elements must match those specified by the
1976 <dt><b>Vector constants</b></dt>
1977 <dd>Vector constants are represented with notation similar to vector type
1978 definitions (a comma separated list of elements, surrounded by
1979 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32
1980 42, i32 11, i32 74, i32 100 ></tt>". Vector constants must
1981 have <a href="#t_vector">vector type</a>, and the number and types of
1982 elements must match those specified by the type.</dd>
1984 <dt><b>Zero initialization</b></dt>
1985 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1986 value to zero of <em>any</em> type, including scalar and aggregate types.
1987 This is often used to avoid having to print large zero initializers
1988 (e.g. for large arrays) and is always exactly equivalent to using explicit
1989 zero initializers.</dd>
1991 <dt><b>Metadata node</b></dt>
1992 <dd>A metadata node is a structure-like constant with
1993 <a href="#t_metadata">metadata type</a>. For example: "<tt>metadata !{
1994 i32 0, metadata !"test" }</tt>". Unlike other constants that are meant to
1995 be interpreted as part of the instruction stream, metadata is a place to
1996 attach additional information such as debug info.</dd>
2001 <!-- ======================================================================= -->
2002 <div class="doc_subsection">
2003 <a name="globalconstants">Global Variable and Function Addresses</a>
2006 <div class="doc_text">
2008 <p>The addresses of <a href="#globalvars">global variables</a>
2009 and <a href="#functionstructure">functions</a> are always implicitly valid
2010 (link-time) constants. These constants are explicitly referenced when
2011 the <a href="#identifiers">identifier for the global</a> is used and always
2012 have <a href="#t_pointer">pointer</a> type. For example, the following is a
2013 legal LLVM file:</p>
2015 <div class="doc_code">
2019 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
2025 <!-- ======================================================================= -->
2026 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
2027 <div class="doc_text">
2029 <p>The string '<tt>undef</tt>' can be used anywhere a constant is expected, and
2030 indicates that the user of the value may receive an unspecified bit-pattern.
2031 Undefined values may be of any type (other than label or void) and be used
2032 anywhere a constant is permitted.</p>
2034 <p>Undefined values are useful because they indicate to the compiler that the
2035 program is well defined no matter what value is used. This gives the
2036 compiler more freedom to optimize. Here are some examples of (potentially
2037 surprising) transformations that are valid (in pseudo IR):</p>
2040 <div class="doc_code">
2052 <p>This is safe because all of the output bits are affected by the undef bits.
2053 Any output bit can have a zero or one depending on the input bits.</p>
2055 <div class="doc_code">
2068 <p>These logical operations have bits that are not always affected by the input.
2069 For example, if "%X" has a zero bit, then the output of the 'and' operation will
2070 always be a zero, no matter what the corresponding bit from the undef is. As
2071 such, it is unsafe to optimize or assume that the result of the and is undef.
2072 However, it is safe to assume that all bits of the undef could be 0, and
2073 optimize the and to 0. Likewise, it is safe to assume that all the bits of
2074 the undef operand to the or could be set, allowing the or to be folded to
2077 <div class="doc_code">
2079 %A = select undef, %X, %Y
2080 %B = select undef, 42, %Y
2081 %C = select %X, %Y, undef
2093 <p>This set of examples show that undefined select (and conditional branch)
2094 conditions can go "either way" but they have to come from one of the two
2095 operands. In the %A example, if %X and %Y were both known to have a clear low
2096 bit, then %A would have to have a cleared low bit. However, in the %C example,
2097 the optimizer is allowed to assume that the undef operand could be the same as
2098 %Y, allowing the whole select to be eliminated.</p>
2101 <div class="doc_code">
2103 %A = xor undef, undef
2122 <p>This example points out that two undef operands are not necessarily the same.
2123 This can be surprising to people (and also matches C semantics) where they
2124 assume that "X^X" is always zero, even if X is undef. This isn't true for a
2125 number of reasons, but the short answer is that an undef "variable" can
2126 arbitrarily change its value over its "live range". This is true because the
2127 "variable" doesn't actually <em>have a live range</em>. Instead, the value is
2128 logically read from arbitrary registers that happen to be around when needed,
2129 so the value is not necessarily consistent over time. In fact, %A and %C need
2130 to have the same semantics or the core LLVM "replace all uses with" concept
2133 <div class="doc_code">
2143 <p>These examples show the crucial difference between an <em>undefined
2144 value</em> and <em>undefined behavior</em>. An undefined value (like undef) is
2145 allowed to have an arbitrary bit-pattern. This means that the %A operation
2146 can be constant folded to undef because the undef could be an SNaN, and fdiv is
2147 not (currently) defined on SNaN's. However, in the second example, we can make
2148 a more aggressive assumption: because the undef is allowed to be an arbitrary
2149 value, we are allowed to assume that it could be zero. Since a divide by zero
2150 has <em>undefined behavior</em>, we are allowed to assume that the operation
2151 does not execute at all. This allows us to delete the divide and all code after
2152 it: since the undefined operation "can't happen", the optimizer can assume that
2153 it occurs in dead code.
2156 <div class="doc_code">
2158 a: store undef -> %X
2159 b: store %X -> undef
2166 <p>These examples reiterate the fdiv example: a store "of" an undefined value
2167 can be assumed to not have any effect: we can assume that the value is
2168 overwritten with bits that happen to match what was already there. However, a
2169 store "to" an undefined location could clobber arbitrary memory, therefore, it
2170 has undefined behavior.</p>
2174 <!-- ======================================================================= -->
2175 <div class="doc_subsection"><a name="blockaddress">Addresses of Basic
2177 <div class="doc_text">
2179 <p><b><tt>blockaddress(@function, %block)</tt></b></p>
2181 <p>The '<tt>blockaddress</tt>' constant computes the address of the specified
2182 basic block in the specified function, and always has an i8* type. Taking
2183 the address of the entry block is illegal.</p>
2185 <p>This value only has defined behavior when used as an operand to the
2186 '<a href="#i_indirectbr"><tt>indirectbr</tt></a>' instruction or for comparisons
2187 against null. Pointer equality tests between labels addresses is undefined
2188 behavior - though, again, comparison against null is ok, and no label is
2189 equal to the null pointer. This may also be passed around as an opaque
2190 pointer sized value as long as the bits are not inspected. This allows
2191 <tt>ptrtoint</tt> and arithmetic to be performed on these values so long as
2192 the original value is reconstituted before the <tt>indirectbr</tt>.</p>
2194 <p>Finally, some targets may provide defined semantics when
2195 using the value as the operand to an inline assembly, but that is target
2202 <!-- ======================================================================= -->
2203 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
2206 <div class="doc_text">
2208 <p>Constant expressions are used to allow expressions involving other constants
2209 to be used as constants. Constant expressions may be of
2210 any <a href="#t_firstclass">first class</a> type and may involve any LLVM
2211 operation that does not have side effects (e.g. load and call are not
2212 supported). The following is the syntax for constant expressions:</p>
2215 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
2216 <dd>Truncate a constant to another type. The bit size of CST must be larger
2217 than the bit size of TYPE. Both types must be integers.</dd>
2219 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
2220 <dd>Zero extend a constant to another type. The bit size of CST must be
2221 smaller or equal to the bit size of TYPE. Both types must be
2224 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
2225 <dd>Sign extend a constant to another type. The bit size of CST must be
2226 smaller or equal to the bit size of TYPE. Both types must be
2229 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
2230 <dd>Truncate a floating point constant to another floating point type. The
2231 size of CST must be larger than the size of TYPE. Both types must be
2232 floating point.</dd>
2234 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
2235 <dd>Floating point extend a constant to another type. The size of CST must be
2236 smaller or equal to the size of TYPE. Both types must be floating
2239 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
2240 <dd>Convert a floating point constant to the corresponding unsigned integer
2241 constant. TYPE must be a scalar or vector integer type. CST must be of
2242 scalar or vector floating point type. Both CST and TYPE must be scalars,
2243 or vectors of the same number of elements. If the value won't fit in the
2244 integer type, the results are undefined.</dd>
2246 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
2247 <dd>Convert a floating point constant to the corresponding signed integer
2248 constant. TYPE must be a scalar or vector integer type. CST must be of
2249 scalar or vector floating point type. Both CST and TYPE must be scalars,
2250 or vectors of the same number of elements. If the value won't fit in the
2251 integer type, the results are undefined.</dd>
2253 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
2254 <dd>Convert an unsigned integer constant to the corresponding floating point
2255 constant. TYPE must be a scalar or vector floating point type. CST must be
2256 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2257 vectors of the same number of elements. If the value won't fit in the
2258 floating point type, the results are undefined.</dd>
2260 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
2261 <dd>Convert a signed integer constant to the corresponding floating point
2262 constant. TYPE must be a scalar or vector floating point type. CST must be
2263 of scalar or vector integer type. Both CST and TYPE must be scalars, or
2264 vectors of the same number of elements. If the value won't fit in the
2265 floating point type, the results are undefined.</dd>
2267 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
2268 <dd>Convert a pointer typed constant to the corresponding integer constant
2269 <tt>TYPE</tt> must be an integer type. <tt>CST</tt> must be of pointer
2270 type. The <tt>CST</tt> value is zero extended, truncated, or unchanged to
2271 make it fit in <tt>TYPE</tt>.</dd>
2273 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
2274 <dd>Convert a integer constant to a pointer constant. TYPE must be a pointer
2275 type. CST must be of integer type. The CST value is zero extended,
2276 truncated, or unchanged to make it fit in a pointer size. This one is
2277 <i>really</i> dangerous!</dd>
2279 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
2280 <dd>Convert a constant, CST, to another TYPE. The constraints of the operands
2281 are the same as those for the <a href="#i_bitcast">bitcast
2282 instruction</a>.</dd>
2284 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
2285 <dt><b><tt>getelementptr inbounds ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
2286 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
2287 constants. As with the <a href="#i_getelementptr">getelementptr</a>
2288 instruction, the index list may have zero or more indexes, which are
2289 required to make sense for the type of "CSTPTR".</dd>
2291 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
2292 <dd>Perform the <a href="#i_select">select operation</a> on constants.</dd>
2294 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
2295 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
2297 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
2298 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
2300 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
2301 <dd>Perform the <a href="#i_extractelement">extractelement operation</a> on
2304 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
2305 <dd>Perform the <a href="#i_insertelement">insertelement operation</a> on
2308 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
2309 <dd>Perform the <a href="#i_shufflevector">shufflevector operation</a> on
2312 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
2313 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
2314 be any of the <a href="#binaryops">binary</a>
2315 or <a href="#bitwiseops">bitwise binary</a> operations. The constraints
2316 on operands are the same as those for the corresponding instruction
2317 (e.g. no bitwise operations on floating point values are allowed).</dd>
2322 <!-- ======================================================================= -->
2323 <div class="doc_subsection"><a name="metadata">Embedded Metadata</a>
2326 <div class="doc_text">
2328 <p>Embedded metadata provides a way to attach arbitrary data to the instruction
2329 stream without affecting the behaviour of the program. There are two
2330 metadata primitives, strings and nodes. All metadata has the
2331 <tt>metadata</tt> type and is identified in syntax by a preceding exclamation
2332 point ('<tt>!</tt>').</p>
2334 <p>A metadata string is a string surrounded by double quotes. It can contain
2335 any character by escaping non-printable characters with "\xx" where "xx" is
2336 the two digit hex code. For example: "<tt>!"test\00"</tt>".</p>
2338 <p>Metadata nodes are represented with notation similar to structure constants
2339 (a comma separated list of elements, surrounded by braces and preceded by an
2340 exclamation point). For example: "<tt>!{ metadata !"test\00", i32
2343 <p>A metadata node will attempt to track changes to the values it holds. In the
2344 event that a value is deleted, it will be replaced with a typeless
2345 "<tt>null</tt>", such as "<tt>metadata !{null, i32 10}</tt>".</p>
2347 <p>Optimizations may rely on metadata to provide additional information about
2348 the program that isn't available in the instructions, or that isn't easily
2349 computable. Similarly, the code generator may expect a certain metadata
2350 format to be used to express debugging information.</p>
2354 <!-- *********************************************************************** -->
2355 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
2356 <!-- *********************************************************************** -->
2358 <!-- ======================================================================= -->
2359 <div class="doc_subsection">
2360 <a name="inlineasm">Inline Assembler Expressions</a>
2363 <div class="doc_text">
2365 <p>LLVM supports inline assembler expressions (as opposed
2366 to <a href="#moduleasm"> Module-Level Inline Assembly</a>) through the use of
2367 a special value. This value represents the inline assembler as a string
2368 (containing the instructions to emit), a list of operand constraints (stored
2369 as a string), a flag that indicates whether or not the inline asm
2370 expression has side effects, and a flag indicating whether the function
2371 containing the asm needs to align its stack conservatively. An example
2372 inline assembler expression is:</p>
2374 <div class="doc_code">
2376 i32 (i32) asm "bswap $0", "=r,r"
2380 <p>Inline assembler expressions may <b>only</b> be used as the callee operand of
2381 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we
2384 <div class="doc_code">
2386 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
2390 <p>Inline asms with side effects not visible in the constraint list must be
2391 marked as having side effects. This is done through the use of the
2392 '<tt>sideeffect</tt>' keyword, like so:</p>
2394 <div class="doc_code">
2396 call void asm sideeffect "eieio", ""()
2400 <p>In some cases inline asms will contain code that will not work unless the
2401 stack is aligned in some way, such as calls or SSE instructions on x86,
2402 yet will not contain code that does that alignment within the asm.
2403 The compiler should make conservative assumptions about what the asm might
2404 contain and should generate its usual stack alignment code in the prologue
2405 if the '<tt>alignstack</tt>' keyword is present:</p>
2407 <div class="doc_code">
2409 call void asm alignstack "eieio", ""()
2413 <p>If both keywords appear the '<tt>sideeffect</tt>' keyword must come
2416 <p>TODO: The format of the asm and constraints string still need to be
2417 documented here. Constraints on what can be done (e.g. duplication, moving,
2418 etc need to be documented). This is probably best done by reference to
2419 another document that covers inline asm from a holistic perspective.</p>
2424 <!-- *********************************************************************** -->
2425 <div class="doc_section">
2426 <a name="intrinsic_globals">Intrinsic Global Variables</a>
2428 <!-- *********************************************************************** -->
2430 <p>LLVM has a number of "magic" global variables that contain data that affect
2431 code generation or other IR semantics. These are documented here. All globals
2432 of this sort should have a section specified as "<tt>llvm.metadata</tt>". This
2433 section and all globals that start with "<tt>llvm.</tt>" are reserved for use
2436 <!-- ======================================================================= -->
2437 <div class="doc_subsection">
2438 <a name="intg_used">The '<tt>llvm.used</tt>' Global Variable</a>
2441 <div class="doc_text">
2443 <p>The <tt>@llvm.used</tt> global is an array with i8* element type which has <a
2444 href="#linkage_appending">appending linkage</a>. This array contains a list of
2445 pointers to global variables and functions which may optionally have a pointer
2446 cast formed of bitcast or getelementptr. For example, a legal use of it is:</p>
2452 @llvm.used = appending global [2 x i8*] [
2454 i8* bitcast (i32* @Y to i8*)
2455 ], section "llvm.metadata"
2458 <p>If a global variable appears in the <tt>@llvm.used</tt> list, then the
2459 compiler, assembler, and linker are required to treat the symbol as if there is
2460 a reference to the global that it cannot see. For example, if a variable has
2461 internal linkage and no references other than that from the <tt>@llvm.used</tt>
2462 list, it cannot be deleted. This is commonly used to represent references from
2463 inline asms and other things the compiler cannot "see", and corresponds to
2464 "attribute((used))" in GNU C.</p>
2466 <p>On some targets, the code generator must emit a directive to the assembler or
2467 object file to prevent the assembler and linker from molesting the symbol.</p>
2471 <!-- ======================================================================= -->
2472 <div class="doc_subsection">
2473 <a name="intg_compiler_used">The '<tt>llvm.compiler.used</tt>' Global Variable</a>
2476 <div class="doc_text">
2478 <p>The <tt>@llvm.compiler.used</tt> directive is the same as the
2479 <tt>@llvm.used</tt> directive, except that it only prevents the compiler from
2480 touching the symbol. On targets that support it, this allows an intelligent
2481 linker to optimize references to the symbol without being impeded as it would be
2482 by <tt>@llvm.used</tt>.</p>
2484 <p>This is a rare construct that should only be used in rare circumstances, and
2485 should not be exposed to source languages.</p>
2489 <!-- ======================================================================= -->
2490 <div class="doc_subsection">
2491 <a name="intg_global_ctors">The '<tt>llvm.global_ctors</tt>' Global Variable</a>
2494 <div class="doc_text">
2496 <p>TODO: Describe this.</p>
2500 <!-- ======================================================================= -->
2501 <div class="doc_subsection">
2502 <a name="intg_global_dtors">The '<tt>llvm.global_dtors</tt>' Global Variable</a>
2505 <div class="doc_text">
2507 <p>TODO: Describe this.</p>
2512 <!-- *********************************************************************** -->
2513 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
2514 <!-- *********************************************************************** -->
2516 <div class="doc_text">
2518 <p>The LLVM instruction set consists of several different classifications of
2519 instructions: <a href="#terminators">terminator
2520 instructions</a>, <a href="#binaryops">binary instructions</a>,
2521 <a href="#bitwiseops">bitwise binary instructions</a>,
2522 <a href="#memoryops">memory instructions</a>, and
2523 <a href="#otherops">other instructions</a>.</p>
2527 <!-- ======================================================================= -->
2528 <div class="doc_subsection"> <a name="terminators">Terminator
2529 Instructions</a> </div>
2531 <div class="doc_text">
2533 <p>As mentioned <a href="#functionstructure">previously</a>, every basic block
2534 in a program ends with a "Terminator" instruction, which indicates which
2535 block should be executed after the current block is finished. These
2536 terminator instructions typically yield a '<tt>void</tt>' value: they produce
2537 control flow, not values (the one exception being the
2538 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
2540 <p>There are six different terminator instructions: the
2541 '<a href="#i_ret"><tt>ret</tt></a>' instruction, the
2542 '<a href="#i_br"><tt>br</tt></a>' instruction, the
2543 '<a href="#i_switch"><tt>switch</tt></a>' instruction, the
2544 '<a href="#i_indirectbr">'<tt>indirectbr</tt>' Instruction, the
2545 '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the
2546 '<a href="#i_unwind"><tt>unwind</tt></a>' instruction, and the
2547 '<a href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
2551 <!-- _______________________________________________________________________ -->
2552 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
2553 Instruction</a> </div>
2555 <div class="doc_text">
2559 ret <type> <value> <i>; Return a value from a non-void function</i>
2560 ret void <i>; Return from void function</i>
2564 <p>The '<tt>ret</tt>' instruction is used to return control flow (and optionally
2565 a value) from a function back to the caller.</p>
2567 <p>There are two forms of the '<tt>ret</tt>' instruction: one that returns a
2568 value and then causes control flow, and one that just causes control flow to
2572 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument, the
2573 return value. The type of the return value must be a
2574 '<a href="#t_firstclass">first class</a>' type.</p>
2576 <p>A function is not <a href="#wellformed">well formed</a> if it it has a
2577 non-void return type and contains a '<tt>ret</tt>' instruction with no return
2578 value or a return value with a type that does not match its type, or if it
2579 has a void return type and contains a '<tt>ret</tt>' instruction with a
2583 <p>When the '<tt>ret</tt>' instruction is executed, control flow returns back to
2584 the calling function's context. If the caller is a
2585 "<a href="#i_call"><tt>call</tt></a>" instruction, execution continues at the
2586 instruction after the call. If the caller was an
2587 "<a href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at
2588 the beginning of the "normal" destination block. If the instruction returns
2589 a value, that value shall set the call or invoke instruction's return
2594 ret i32 5 <i>; Return an integer value of 5</i>
2595 ret void <i>; Return from a void function</i>
2596 ret { i32, i8 } { i32 4, i8 2 } <i>; Return a struct of values 4 and 2</i>
2599 <p>Note that the code generator does not yet fully support large
2600 return values. The specific sizes that are currently supported are
2601 dependent on the target. For integers, on 32-bit targets the limit
2602 is often 64 bits, and on 64-bit targets the limit is often 128 bits.
2603 For aggregate types, the current limits are dependent on the element
2604 types; for example targets are often limited to 2 total integer
2605 elements and 2 total floating-point elements.</p>
2608 <!-- _______________________________________________________________________ -->
2609 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2611 <div class="doc_text">
2615 br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2619 <p>The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
2620 different basic block in the current function. There are two forms of this
2621 instruction, corresponding to a conditional branch and an unconditional
2625 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a single
2626 '<tt>i1</tt>' value and two '<tt>label</tt>' values. The unconditional form
2627 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
2631 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2632 argument is evaluated. If the value is <tt>true</tt>, control flows to the
2633 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2634 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2639 %cond = <a href="#i_icmp">icmp</a> eq i32 %a, %b
2640 br i1 %cond, label %IfEqual, label %IfUnequal
2642 <a href="#i_ret">ret</a> i32 1
2644 <a href="#i_ret">ret</a> i32 0
2649 <!-- _______________________________________________________________________ -->
2650 <div class="doc_subsubsection">
2651 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2654 <div class="doc_text">
2658 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2662 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2663 several different places. It is a generalization of the '<tt>br</tt>'
2664 instruction, allowing a branch to occur to one of many possible
2668 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2669 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination,
2670 and an array of pairs of comparison value constants and '<tt>label</tt>'s.
2671 The table is not allowed to contain duplicate constant entries.</p>
2674 <p>The <tt>switch</tt> instruction specifies a table of values and
2675 destinations. When the '<tt>switch</tt>' instruction is executed, this table
2676 is searched for the given value. If the value is found, control flow is
2677 transferred to the corresponding destination; otherwise, control flow is
2678 transferred to the default destination.</p>
2680 <h5>Implementation:</h5>
2681 <p>Depending on properties of the target machine and the particular
2682 <tt>switch</tt> instruction, this instruction may be code generated in
2683 different ways. For example, it could be generated as a series of chained
2684 conditional branches or with a lookup table.</p>
2688 <i>; Emulate a conditional br instruction</i>
2689 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2690 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
2692 <i>; Emulate an unconditional br instruction</i>
2693 switch i32 0, label %dest [ ]
2695 <i>; Implement a jump table:</i>
2696 switch i32 %val, label %otherwise [ i32 0, label %onzero
2698 i32 2, label %ontwo ]
2704 <!-- _______________________________________________________________________ -->
2705 <div class="doc_subsubsection">
2706 <a name="i_indirectbr">'<tt>indirectbr</tt>' Instruction</a>
2709 <div class="doc_text">
2713 indirectbr <somety>* <address>, [ label <dest1>, label <dest2>, ... ]
2718 <p>The '<tt>indirectbr</tt>' instruction implements an indirect branch to a label
2719 within the current function, whose address is specified by
2720 "<tt>address</tt>". Address must be derived from a <a
2721 href="#blockaddress">blockaddress</a> constant.</p>
2725 <p>The '<tt>address</tt>' argument is the address of the label to jump to. The
2726 rest of the arguments indicate the full set of possible destinations that the
2727 address may point to. Blocks are allowed to occur multiple times in the
2728 destination list, though this isn't particularly useful.</p>
2730 <p>This destination list is required so that dataflow analysis has an accurate
2731 understanding of the CFG.</p>
2735 <p>Control transfers to the block specified in the address argument. All
2736 possible destination blocks must be listed in the label list, otherwise this
2737 instruction has undefined behavior. This implies that jumps to labels
2738 defined in other functions have undefined behavior as well.</p>
2740 <h5>Implementation:</h5>
2742 <p>This is typically implemented with a jump through a register.</p>
2746 indirectbr i8* %Addr, [ label %bb1, label %bb2, label %bb3 ]
2752 <!-- _______________________________________________________________________ -->
2753 <div class="doc_subsubsection">
2754 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2757 <div class="doc_text">
2761 <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>]
2762 to label <normal label> unwind label <exception label>
2766 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2767 function, with the possibility of control flow transfer to either the
2768 '<tt>normal</tt>' label or the '<tt>exception</tt>' label. If the callee
2769 function returns with the "<tt><a href="#i_ret">ret</a></tt>" instruction,
2770 control flow will return to the "normal" label. If the callee (or any
2771 indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
2772 instruction, control is interrupted and continued at the dynamically nearest
2773 "exception" label.</p>
2776 <p>This instruction requires several arguments:</p>
2779 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
2780 convention</a> the call should use. If none is specified, the call
2781 defaults to using C calling conventions.</li>
2783 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2784 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
2785 '<tt>inreg</tt>' attributes are valid here.</li>
2787 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2788 function value being invoked. In most cases, this is a direct function
2789 invocation, but indirect <tt>invoke</tt>s are just as possible, branching
2790 off an arbitrary pointer to function value.</li>
2792 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2793 function to be invoked. </li>
2795 <li>'<tt>function args</tt>': argument list whose types match the function
2796 signature argument types. If the function signature indicates the
2797 function accepts a variable number of arguments, the extra arguments can
2800 <li>'<tt>normal label</tt>': the label reached when the called function
2801 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2803 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2804 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2806 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
2807 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2808 '<tt>readnone</tt>' attributes are valid here.</li>
2812 <p>This instruction is designed to operate as a standard
2813 '<tt><a href="#i_call">call</a></tt>' instruction in most regards. The
2814 primary difference is that it establishes an association with a label, which
2815 is used by the runtime library to unwind the stack.</p>
2817 <p>This instruction is used in languages with destructors to ensure that proper
2818 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2819 exception. Additionally, this is important for implementation of
2820 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
2822 <p>For the purposes of the SSA form, the definition of the value returned by the
2823 '<tt>invoke</tt>' instruction is deemed to occur on the edge from the current
2824 block to the "normal" label. If the callee unwinds then no return value is
2829 %retval = invoke i32 @Test(i32 15) to label %Continue
2830 unwind label %TestCleanup <i>; {i32}:retval set</i>
2831 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
2832 unwind label %TestCleanup <i>; {i32}:retval set</i>
2837 <!-- _______________________________________________________________________ -->
2839 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2840 Instruction</a> </div>
2842 <div class="doc_text">
2850 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2851 at the first callee in the dynamic call stack which used
2852 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call.
2853 This is primarily used to implement exception handling.</p>
2856 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
2857 immediately halt. The dynamic call stack is then searched for the
2858 first <a href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack.
2859 Once found, execution continues at the "exceptional" destination block
2860 specified by the <tt>invoke</tt> instruction. If there is no <tt>invoke</tt>
2861 instruction in the dynamic call chain, undefined behavior results.</p>
2865 <!-- _______________________________________________________________________ -->
2867 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2868 Instruction</a> </div>
2870 <div class="doc_text">
2878 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2879 instruction is used to inform the optimizer that a particular portion of the
2880 code is not reachable. This can be used to indicate that the code after a
2881 no-return function cannot be reached, and other facts.</p>
2884 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2888 <!-- ======================================================================= -->
2889 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2891 <div class="doc_text">
2893 <p>Binary operators are used to do most of the computation in a program. They
2894 require two operands of the same type, execute an operation on them, and
2895 produce a single value. The operands might represent multiple data, as is
2896 the case with the <a href="#t_vector">vector</a> data type. The result value
2897 has the same type as its operands.</p>
2899 <p>There are several different binary operators:</p>
2903 <!-- _______________________________________________________________________ -->
2904 <div class="doc_subsubsection">
2905 <a name="i_add">'<tt>add</tt>' Instruction</a>
2908 <div class="doc_text">
2912 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2913 <result> = add nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2914 <result> = add nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2915 <result> = add nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2919 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2922 <p>The two arguments to the '<tt>add</tt>' instruction must
2923 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
2924 integer values. Both arguments must have identical types.</p>
2927 <p>The value produced is the integer sum of the two operands.</p>
2929 <p>If the sum has unsigned overflow, the result returned is the mathematical
2930 result modulo 2<sup>n</sup>, where n is the bit width of the result.</p>
2932 <p>Because LLVM integers use a two's complement representation, this instruction
2933 is appropriate for both signed and unsigned integers.</p>
2935 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
2936 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
2937 <tt>nsw</tt> keywords are present, the result value of the <tt>add</tt>
2938 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
2942 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2947 <!-- _______________________________________________________________________ -->
2948 <div class="doc_subsubsection">
2949 <a name="i_fadd">'<tt>fadd</tt>' Instruction</a>
2952 <div class="doc_text">
2956 <result> = fadd <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2960 <p>The '<tt>fadd</tt>' instruction returns the sum of its two operands.</p>
2963 <p>The two arguments to the '<tt>fadd</tt>' instruction must be
2964 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
2965 floating point values. Both arguments must have identical types.</p>
2968 <p>The value produced is the floating point sum of the two operands.</p>
2972 <result> = fadd float 4.0, %var <i>; yields {float}:result = 4.0 + %var</i>
2977 <!-- _______________________________________________________________________ -->
2978 <div class="doc_subsubsection">
2979 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2982 <div class="doc_text">
2986 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2987 <result> = sub nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2988 <result> = sub nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2989 <result> = sub nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2993 <p>The '<tt>sub</tt>' instruction returns the difference of its two
2996 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
2997 '<tt>neg</tt>' instruction present in most other intermediate
2998 representations.</p>
3001 <p>The two arguments to the '<tt>sub</tt>' instruction must
3002 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3003 integer values. Both arguments must have identical types.</p>
3006 <p>The value produced is the integer difference of the two operands.</p>
3008 <p>If the difference has unsigned overflow, the result returned is the
3009 mathematical result modulo 2<sup>n</sup>, where n is the bit width of the
3012 <p>Because LLVM integers use a two's complement representation, this instruction
3013 is appropriate for both signed and unsigned integers.</p>
3015 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3016 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3017 <tt>nsw</tt> keywords are present, the result value of the <tt>sub</tt>
3018 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
3022 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
3023 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
3028 <!-- _______________________________________________________________________ -->
3029 <div class="doc_subsubsection">
3030 <a name="i_fsub">'<tt>fsub</tt>' Instruction</a>
3033 <div class="doc_text">
3037 <result> = fsub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3041 <p>The '<tt>fsub</tt>' instruction returns the difference of its two
3044 <p>Note that the '<tt>fsub</tt>' instruction is used to represent the
3045 '<tt>fneg</tt>' instruction present in most other intermediate
3046 representations.</p>
3049 <p>The two arguments to the '<tt>fsub</tt>' instruction must be
3050 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3051 floating point values. Both arguments must have identical types.</p>
3054 <p>The value produced is the floating point difference of the two operands.</p>
3058 <result> = fsub float 4.0, %var <i>; yields {float}:result = 4.0 - %var</i>
3059 <result> = fsub float -0.0, %val <i>; yields {float}:result = -%var</i>
3064 <!-- _______________________________________________________________________ -->
3065 <div class="doc_subsubsection">
3066 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
3069 <div class="doc_text">
3073 <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3074 <result> = mul nuw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3075 <result> = mul nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3076 <result> = mul nuw nsw <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3080 <p>The '<tt>mul</tt>' instruction returns the product of its two operands.</p>
3083 <p>The two arguments to the '<tt>mul</tt>' instruction must
3084 be <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3085 integer values. Both arguments must have identical types.</p>
3088 <p>The value produced is the integer product of the two operands.</p>
3090 <p>If the result of the multiplication has unsigned overflow, the result
3091 returned is the mathematical result modulo 2<sup>n</sup>, where n is the bit
3092 width of the result.</p>
3094 <p>Because LLVM integers use a two's complement representation, and the result
3095 is the same width as the operands, this instruction returns the correct
3096 result for both signed and unsigned integers. If a full product
3097 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands should
3098 be sign-extended or zero-extended as appropriate to the width of the full
3101 <p><tt>nuw</tt> and <tt>nsw</tt> stand for "No Unsigned Wrap"
3102 and "No Signed Wrap", respectively. If the <tt>nuw</tt> and/or
3103 <tt>nsw</tt> keywords are present, the result value of the <tt>mul</tt>
3104 is undefined if unsigned and/or signed overflow, respectively, occurs.</p>
3108 <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
3113 <!-- _______________________________________________________________________ -->
3114 <div class="doc_subsubsection">
3115 <a name="i_fmul">'<tt>fmul</tt>' Instruction</a>
3118 <div class="doc_text">
3122 <result> = fmul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3126 <p>The '<tt>fmul</tt>' instruction returns the product of its two operands.</p>
3129 <p>The two arguments to the '<tt>fmul</tt>' instruction must be
3130 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3131 floating point values. Both arguments must have identical types.</p>
3134 <p>The value produced is the floating point product of the two operands.</p>
3138 <result> = fmul float 4.0, %var <i>; yields {float}:result = 4.0 * %var</i>
3143 <!-- _______________________________________________________________________ -->
3144 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
3147 <div class="doc_text">
3151 <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3155 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two operands.</p>
3158 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
3159 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3160 values. Both arguments must have identical types.</p>
3163 <p>The value produced is the unsigned integer quotient of the two operands.</p>
3165 <p>Note that unsigned integer division and signed integer division are distinct
3166 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
3168 <p>Division by zero leads to undefined behavior.</p>
3172 <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3177 <!-- _______________________________________________________________________ -->
3178 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
3181 <div class="doc_text">
3185 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3186 <result> = sdiv exact <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3190 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two operands.</p>
3193 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
3194 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3195 values. Both arguments must have identical types.</p>
3198 <p>The value produced is the signed integer quotient of the two operands rounded
3201 <p>Note that signed integer division and unsigned integer division are distinct
3202 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
3204 <p>Division by zero leads to undefined behavior. Overflow also leads to
3205 undefined behavior; this is a rare case, but can occur, for example, by doing
3206 a 32-bit division of -2147483648 by -1.</p>
3208 <p>If the <tt>exact</tt> keyword is present, the result value of the
3209 <tt>sdiv</tt> is undefined if the result would be rounded or if overflow
3214 <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
3219 <!-- _______________________________________________________________________ -->
3220 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
3221 Instruction</a> </div>
3223 <div class="doc_text">
3227 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3231 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two operands.</p>
3234 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
3235 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3236 floating point values. Both arguments must have identical types.</p>
3239 <p>The value produced is the floating point quotient of the two operands.</p>
3243 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
3248 <!-- _______________________________________________________________________ -->
3249 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
3252 <div class="doc_text">
3256 <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3260 <p>The '<tt>urem</tt>' instruction returns the remainder from the unsigned
3261 division of its two arguments.</p>
3264 <p>The two arguments to the '<tt>urem</tt>' instruction must be
3265 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3266 values. Both arguments must have identical types.</p>
3269 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
3270 This instruction always performs an unsigned division to get the
3273 <p>Note that unsigned integer remainder and signed integer remainder are
3274 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
3276 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
3280 <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3285 <!-- _______________________________________________________________________ -->
3286 <div class="doc_subsubsection">
3287 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
3290 <div class="doc_text">
3294 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3298 <p>The '<tt>srem</tt>' instruction returns the remainder from the signed
3299 division of its two operands. This instruction can also take
3300 <a href="#t_vector">vector</a> versions of the values in which case the
3301 elements must be integers.</p>
3304 <p>The two arguments to the '<tt>srem</tt>' instruction must be
3305 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3306 values. Both arguments must have identical types.</p>
3309 <p>This instruction returns the <i>remainder</i> of a division (where the result
3310 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
3311 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
3312 a value. For more information about the difference,
3313 see <a href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
3314 Math Forum</a>. For a table of how this is implemented in various languages,
3315 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
3316 Wikipedia: modulo operation</a>.</p>
3318 <p>Note that signed integer remainder and unsigned integer remainder are
3319 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
3321 <p>Taking the remainder of a division by zero leads to undefined behavior.
3322 Overflow also leads to undefined behavior; this is a rare case, but can
3323 occur, for example, by taking the remainder of a 32-bit division of
3324 -2147483648 by -1. (The remainder doesn't actually overflow, but this rule
3325 lets srem be implemented using instructions that return both the result of
3326 the division and the remainder.)</p>
3330 <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
3335 <!-- _______________________________________________________________________ -->
3336 <div class="doc_subsubsection">
3337 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
3339 <div class="doc_text">
3343 <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3347 <p>The '<tt>frem</tt>' instruction returns the remainder from the division of
3348 its two operands.</p>
3351 <p>The two arguments to the '<tt>frem</tt>' instruction must be
3352 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a> of
3353 floating point values. Both arguments must have identical types.</p>
3356 <p>This instruction returns the <i>remainder</i> of a division. The remainder
3357 has the same sign as the dividend.</p>
3361 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
3366 <!-- ======================================================================= -->
3367 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
3368 Operations</a> </div>
3370 <div class="doc_text">
3372 <p>Bitwise binary operators are used to do various forms of bit-twiddling in a
3373 program. They are generally very efficient instructions and can commonly be
3374 strength reduced from other instructions. They require two operands of the
3375 same type, execute an operation on them, and produce a single value. The
3376 resulting value is the same type as its operands.</p>
3380 <!-- _______________________________________________________________________ -->
3381 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
3382 Instruction</a> </div>
3384 <div class="doc_text">
3388 <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3392 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to the left
3393 a specified number of bits.</p>
3396 <p>Both arguments to the '<tt>shl</tt>' instruction must be the
3397 same <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of
3398 integer type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3401 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod
3402 2<sup>n</sup>, where <tt>n</tt> is the width of the result. If <tt>op2</tt>
3403 is (statically or dynamically) negative or equal to or larger than the number
3404 of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3405 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3406 shift amount in <tt>op2</tt>.</p>
3410 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
3411 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
3412 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
3413 <result> = shl i32 1, 32 <i>; undefined</i>
3414 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
3419 <!-- _______________________________________________________________________ -->
3420 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
3421 Instruction</a> </div>
3423 <div class="doc_text">
3427 <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3431 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
3432 operand shifted to the right a specified number of bits with zero fill.</p>
3435 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
3436 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3437 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3440 <p>This instruction always performs a logical shift right operation. The most
3441 significant bits of the result will be filled with zero bits after the shift.
3442 If <tt>op2</tt> is (statically or dynamically) equal to or larger than the
3443 number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
3444 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding
3445 shift amount in <tt>op2</tt>.</p>
3449 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
3450 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
3451 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
3452 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
3453 <result> = lshr i32 1, 32 <i>; undefined</i>
3454 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
3459 <!-- _______________________________________________________________________ -->
3460 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
3461 Instruction</a> </div>
3462 <div class="doc_text">
3466 <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3470 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
3471 operand shifted to the right a specified number of bits with sign
3475 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
3476 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3477 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
3480 <p>This instruction always performs an arithmetic shift right operation, The
3481 most significant bits of the result will be filled with the sign bit
3482 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
3483 larger than the number of bits in <tt>op1</tt>, the result is undefined. If
3484 the arguments are vectors, each vector element of <tt>op1</tt> is shifted by
3485 the corresponding shift amount in <tt>op2</tt>.</p>
3489 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
3490 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
3491 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
3492 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
3493 <result> = ashr i32 1, 32 <i>; undefined</i>
3494 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
3499 <!-- _______________________________________________________________________ -->
3500 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
3501 Instruction</a> </div>
3503 <div class="doc_text">
3507 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3511 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of its two
3515 <p>The two arguments to the '<tt>and</tt>' instruction must be
3516 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3517 values. Both arguments must have identical types.</p>
3520 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
3522 <table border="1" cellspacing="0" cellpadding="4">
3554 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
3555 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
3556 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
3559 <!-- _______________________________________________________________________ -->
3560 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
3562 <div class="doc_text">
3566 <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3570 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive or of its
3574 <p>The two arguments to the '<tt>or</tt>' instruction must be
3575 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3576 values. Both arguments must have identical types.</p>
3579 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
3581 <table border="1" cellspacing="0" cellpadding="4">
3613 <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
3614 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
3615 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
3620 <!-- _______________________________________________________________________ -->
3621 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
3622 Instruction</a> </div>
3624 <div class="doc_text">
3628 <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
3632 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of
3633 its two operands. The <tt>xor</tt> is used to implement the "one's
3634 complement" operation, which is the "~" operator in C.</p>
3637 <p>The two arguments to the '<tt>xor</tt>' instruction must be
3638 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
3639 values. Both arguments must have identical types.</p>
3642 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
3644 <table border="1" cellspacing="0" cellpadding="4">
3676 <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
3677 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
3678 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
3679 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
3684 <!-- ======================================================================= -->
3685 <div class="doc_subsection">
3686 <a name="vectorops">Vector Operations</a>
3689 <div class="doc_text">
3691 <p>LLVM supports several instructions to represent vector operations in a
3692 target-independent manner. These instructions cover the element-access and
3693 vector-specific operations needed to process vectors effectively. While LLVM
3694 does directly support these vector operations, many sophisticated algorithms
3695 will want to use target-specific intrinsics to take full advantage of a
3696 specific target.</p>
3700 <!-- _______________________________________________________________________ -->
3701 <div class="doc_subsubsection">
3702 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
3705 <div class="doc_text">
3709 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
3713 <p>The '<tt>extractelement</tt>' instruction extracts a single scalar element
3714 from a vector at a specified index.</p>
3718 <p>The first operand of an '<tt>extractelement</tt>' instruction is a value
3719 of <a href="#t_vector">vector</a> type. The second operand is an index
3720 indicating the position from which to extract the element. The index may be
3724 <p>The result is a scalar of the same type as the element type of
3725 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
3726 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3727 results are undefined.</p>
3731 <result> = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
3736 <!-- _______________________________________________________________________ -->
3737 <div class="doc_subsubsection">
3738 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
3741 <div class="doc_text">
3745 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
3749 <p>The '<tt>insertelement</tt>' instruction inserts a scalar element into a
3750 vector at a specified index.</p>
3753 <p>The first operand of an '<tt>insertelement</tt>' instruction is a value
3754 of <a href="#t_vector">vector</a> type. The second operand is a scalar value
3755 whose type must equal the element type of the first operand. The third
3756 operand is an index indicating the position at which to insert the value.
3757 The index may be a variable.</p>
3760 <p>The result is a vector of the same type as <tt>val</tt>. Its element values
3761 are those of <tt>val</tt> except at position <tt>idx</tt>, where it gets the
3762 value <tt>elt</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
3763 results are undefined.</p>
3767 <result> = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
3772 <!-- _______________________________________________________________________ -->
3773 <div class="doc_subsubsection">
3774 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3777 <div class="doc_text">
3781 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
3785 <p>The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
3786 from two input vectors, returning a vector with the same element type as the
3787 input and length that is the same as the shuffle mask.</p>
3790 <p>The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3791 with types that match each other. The third argument is a shuffle mask whose
3792 element type is always 'i32'. The result of the instruction is a vector
3793 whose length is the same as the shuffle mask and whose element type is the
3794 same as the element type of the first two operands.</p>
3796 <p>The shuffle mask operand is required to be a constant vector with either
3797 constant integer or undef values.</p>
3800 <p>The elements of the two input vectors are numbered from left to right across
3801 both of the vectors. The shuffle mask operand specifies, for each element of
3802 the result vector, which element of the two input vectors the result element
3803 gets. The element selector may be undef (meaning "don't care") and the
3804 second operand may be undef if performing a shuffle from only one vector.</p>
3808 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3809 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
3810 <result> = shufflevector <4 x i32> %v1, <4 x i32> undef,
3811 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
3812 <result> = shufflevector <8 x i32> %v1, <8 x i32> undef,
3813 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
3814 <result> = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3815 <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>
3820 <!-- ======================================================================= -->
3821 <div class="doc_subsection">
3822 <a name="aggregateops">Aggregate Operations</a>
3825 <div class="doc_text">
3827 <p>LLVM supports several instructions for working with aggregate values.</p>
3831 <!-- _______________________________________________________________________ -->
3832 <div class="doc_subsubsection">
3833 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3836 <div class="doc_text">
3840 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
3844 <p>The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3845 or array element from an aggregate value.</p>
3848 <p>The first operand of an '<tt>extractvalue</tt>' instruction is a value
3849 of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type. The
3850 operands are constant indices to specify which value to extract in a similar
3851 manner as indices in a
3852 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.</p>
3855 <p>The result is the value at the position in the aggregate specified by the
3860 <result> = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
3865 <!-- _______________________________________________________________________ -->
3866 <div class="doc_subsubsection">
3867 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3870 <div class="doc_text">
3874 <result> = insertvalue <aggregate type> <val>, <ty> <val>, <idx> <i>; yields <n x <ty>></i>
3878 <p>The '<tt>insertvalue</tt>' instruction inserts a value into a struct field or
3879 array element in an aggregate.</p>
3883 <p>The first operand of an '<tt>insertvalue</tt>' instruction is a value
3884 of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type. The
3885 second operand is a first-class value to insert. The following operands are
3886 constant indices indicating the position at which to insert the value in a
3887 similar manner as indices in a
3888 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction. The
3889 value to insert must have the same type as the value identified by the
3893 <p>The result is an aggregate of the same type as <tt>val</tt>. Its value is
3894 that of <tt>val</tt> except that the value at the position specified by the
3895 indices is that of <tt>elt</tt>.</p>
3899 <result> = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
3905 <!-- ======================================================================= -->
3906 <div class="doc_subsection">
3907 <a name="memoryops">Memory Access and Addressing Operations</a>
3910 <div class="doc_text">
3912 <p>A key design point of an SSA-based representation is how it represents
3913 memory. In LLVM, no memory locations are in SSA form, which makes things
3914 very simple. This section describes how to read, write, and allocate
3919 <!-- _______________________________________________________________________ -->
3920 <div class="doc_subsubsection">
3921 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3924 <div class="doc_text">
3928 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3932 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3933 currently executing function, to be automatically released when this function
3934 returns to its caller. The object is always allocated in the generic address
3935 space (address space zero).</p>
3938 <p>The '<tt>alloca</tt>' instruction
3939 allocates <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the
3940 runtime stack, returning a pointer of the appropriate type to the program.
3941 If "NumElements" is specified, it is the number of elements allocated,
3942 otherwise "NumElements" is defaulted to be one. If a constant alignment is
3943 specified, the value result of the allocation is guaranteed to be aligned to
3944 at least that boundary. If not specified, or if zero, the target can choose
3945 to align the allocation on any convenient boundary compatible with the
3948 <p>'<tt>type</tt>' may be any sized type.</p>
3951 <p>Memory is allocated; a pointer is returned. The operation is undefined if
3952 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
3953 memory is automatically released when the function returns. The
3954 '<tt>alloca</tt>' instruction is commonly used to represent automatic
3955 variables that must have an address available. When the function returns
3956 (either with the <tt><a href="#i_ret">ret</a></tt>
3957 or <tt><a href="#i_unwind">unwind</a></tt> instructions), the memory is
3958 reclaimed. Allocating zero bytes is legal, but the result is undefined.</p>
3962 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3963 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3964 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3965 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3970 <!-- _______________________________________________________________________ -->
3971 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3972 Instruction</a> </div>
3974 <div class="doc_text">
3978 <result> = load <ty>* <pointer>[, align <alignment>]
3979 <result> = volatile load <ty>* <pointer>[, align <alignment>]
3983 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3986 <p>The argument to the '<tt>load</tt>' instruction specifies the memory address
3987 from which to load. The pointer must point to
3988 a <a href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3989 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify the
3990 number or order of execution of this <tt>load</tt> with other
3991 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3994 <p>The optional constant "align" argument specifies the alignment of the
3995 operation (that is, the alignment of the memory address). A value of 0 or an
3996 omitted "align" argument means that the operation has the preferential
3997 alignment for the target. It is the responsibility of the code emitter to
3998 ensure that the alignment information is correct. Overestimating the
3999 alignment results in an undefined behavior. Underestimating the alignment may
4000 produce less efficient code. An alignment of 1 is always safe.</p>
4003 <p>The location of memory pointed to is loaded. If the value being loaded is of
4004 scalar type then the number of bytes read does not exceed the minimum number
4005 of bytes needed to hold all bits of the type. For example, loading an
4006 <tt>i24</tt> reads at most three bytes. When loading a value of a type like
4007 <tt>i20</tt> with a size that is not an integral number of bytes, the result
4008 is undefined if the value was not originally written using a store of the
4013 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4014 <a href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
4015 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
4020 <!-- _______________________________________________________________________ -->
4021 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
4022 Instruction</a> </div>
4024 <div class="doc_text">
4028 store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
4029 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
4033 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
4036 <p>There are two arguments to the '<tt>store</tt>' instruction: a value to store
4037 and an address at which to store it. The type of the
4038 '<tt><pointer></tt>' operand must be a pointer to
4039 the <a href="#t_firstclass">first class</a> type of the
4040 '<tt><value></tt>' operand. If the <tt>store</tt> is marked
4041 as <tt>volatile</tt>, then the optimizer is not allowed to modify the number
4042 or order of execution of this <tt>store</tt> with other
4043 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
4046 <p>The optional constant "align" argument specifies the alignment of the
4047 operation (that is, the alignment of the memory address). A value of 0 or an
4048 omitted "align" argument means that the operation has the preferential
4049 alignment for the target. It is the responsibility of the code emitter to
4050 ensure that the alignment information is correct. Overestimating the
4051 alignment results in an undefined behavior. Underestimating the alignment may
4052 produce less efficient code. An alignment of 1 is always safe.</p>
4055 <p>The contents of memory are updated to contain '<tt><value></tt>' at the
4056 location specified by the '<tt><pointer></tt>' operand. If
4057 '<tt><value></tt>' is of scalar type then the number of bytes written
4058 does not exceed the minimum number of bytes needed to hold all bits of the
4059 type. For example, storing an <tt>i24</tt> writes at most three bytes. When
4060 writing a value of a type like <tt>i20</tt> with a size that is not an
4061 integral number of bytes, it is unspecified what happens to the extra bits
4062 that do not belong to the type, but they will typically be overwritten.</p>
4066 %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
4067 store i32 3, i32* %ptr <i>; yields {void}</i>
4068 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
4073 <!-- _______________________________________________________________________ -->
4074 <div class="doc_subsubsection">
4075 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
4078 <div class="doc_text">
4082 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
4083 <result> = getelementptr inbounds <pty>* <ptrval>{, <ty> <idx>}*
4087 <p>The '<tt>getelementptr</tt>' instruction is used to get the address of a
4088 subelement of an aggregate data structure. It performs address calculation
4089 only and does not access memory.</p>
4092 <p>The first argument is always a pointer, and forms the basis of the
4093 calculation. The remaining arguments are indices that indicate which of the
4094 elements of the aggregate object are indexed. The interpretation of each
4095 index is dependent on the type being indexed into. The first index always
4096 indexes the pointer value given as the first argument, the second index
4097 indexes a value of the type pointed to (not necessarily the value directly
4098 pointed to, since the first index can be non-zero), etc. The first type
4099 indexed into must be a pointer value, subsequent types can be arrays, vectors
4100 and structs. Note that subsequent types being indexed into can never be
4101 pointers, since that would require loading the pointer before continuing
4104 <p>The type of each index argument depends on the type it is indexing into.
4105 When indexing into a (optionally packed) structure, only <tt>i32</tt> integer
4106 <b>constants</b> are allowed. When indexing into an array, pointer or
4107 vector, integers of any width are allowed, and they are not required to be
4110 <p>For example, let's consider a C code fragment and how it gets compiled to
4113 <div class="doc_code">
4126 int *foo(struct ST *s) {
4127 return &s[1].Z.B[5][13];
4132 <p>The LLVM code generated by the GCC frontend is:</p>
4134 <div class="doc_code">
4136 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
4137 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
4139 define i32* @foo(%ST* %s) {
4141 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
4148 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
4149 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
4150 }</tt>' type, a structure. The second index indexes into the third element
4151 of the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
4152 i8 }</tt>' type, another structure. The third index indexes into the second
4153 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
4154 array. The two dimensions of the array are subscripted into, yielding an
4155 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a
4156 pointer to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
4158 <p>Note that it is perfectly legal to index partially through a structure,
4159 returning a pointer to an inner element. Because of this, the LLVM code for
4160 the given testcase is equivalent to:</p>
4163 define i32* @foo(%ST* %s) {
4164 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
4165 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
4166 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
4167 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
4168 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
4173 <p>If the <tt>inbounds</tt> keyword is present, the result value of the
4174 <tt>getelementptr</tt> is undefined if the base pointer is not an
4175 <i>in bounds</i> address of an allocated object, or if any of the addresses
4176 that would be formed by successive addition of the offsets implied by the
4177 indices to the base address with infinitely precise arithmetic are not an
4178 <i>in bounds</i> address of that allocated object.
4179 The <i>in bounds</i> addresses for an allocated object are all the addresses
4180 that point into the object, plus the address one byte past the end.</p>
4182 <p>If the <tt>inbounds</tt> keyword is not present, the offsets are added to
4183 the base address with silently-wrapping two's complement arithmetic, and
4184 the result value of the <tt>getelementptr</tt> may be outside the object
4185 pointed to by the base pointer. The result value may not necessarily be
4186 used to access memory though, even if it happens to point into allocated
4187 storage. See the <a href="#pointeraliasing">Pointer Aliasing Rules</a>
4188 section for more information.</p>
4190 <p>The getelementptr instruction is often confusing. For some more insight into
4191 how it works, see <a href="GetElementPtr.html">the getelementptr FAQ</a>.</p>
4195 <i>; yields [12 x i8]*:aptr</i>
4196 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
4197 <i>; yields i8*:vptr</i>
4198 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
4199 <i>; yields i8*:eptr</i>
4200 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
4201 <i>; yields i32*:iptr</i>
4202 %iptr = getelementptr [10 x i32]* @arr, i16 0, i16 0
4207 <!-- ======================================================================= -->
4208 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
4211 <div class="doc_text">
4213 <p>The instructions in this category are the conversion instructions (casting)
4214 which all take a single operand and a type. They perform various bit
4215 conversions on the operand.</p>
4219 <!-- _______________________________________________________________________ -->
4220 <div class="doc_subsubsection">
4221 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
4223 <div class="doc_text">
4227 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
4231 <p>The '<tt>trunc</tt>' instruction truncates its operand to the
4232 type <tt>ty2</tt>.</p>
4235 <p>The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
4236 be an <a href="#t_integer">integer</a> type, and a type that specifies the
4237 size and type of the result, which must be
4238 an <a href="#t_integer">integer</a> type. The bit size of <tt>value</tt> must
4239 be larger than the bit size of <tt>ty2</tt>. Equal sized types are not
4243 <p>The '<tt>trunc</tt>' instruction truncates the high order bits
4244 in <tt>value</tt> and converts the remaining bits to <tt>ty2</tt>. Since the
4245 source size must be larger than the destination size, <tt>trunc</tt> cannot
4246 be a <i>no-op cast</i>. It will always truncate bits.</p>
4250 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
4251 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
4252 %Z = trunc i32 122 to i1 <i>; yields i1:false</i>
4257 <!-- _______________________________________________________________________ -->
4258 <div class="doc_subsubsection">
4259 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
4261 <div class="doc_text">
4265 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
4269 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
4274 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
4275 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4276 also be of <a href="#t_integer">integer</a> type. The bit size of the
4277 <tt>value</tt> must be smaller than the bit size of the destination type,
4281 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
4282 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
4284 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
4288 %X = zext i32 257 to i64 <i>; yields i64:257</i>
4289 %Y = zext i1 true to i32 <i>; yields i32:1</i>
4294 <!-- _______________________________________________________________________ -->
4295 <div class="doc_subsubsection">
4296 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
4298 <div class="doc_text">
4302 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
4306 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
4309 <p>The '<tt>sext</tt>' instruction takes a value to cast, which must be of
4310 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
4311 also be of <a href="#t_integer">integer</a> type. The bit size of the
4312 <tt>value</tt> must be smaller than the bit size of the destination type,
4316 <p>The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
4317 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size
4318 of the type <tt>ty2</tt>.</p>
4320 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
4324 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
4325 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
4330 <!-- _______________________________________________________________________ -->
4331 <div class="doc_subsubsection">
4332 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
4335 <div class="doc_text">
4339 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
4343 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
4347 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
4348 point</a> value to cast and a <a href="#t_floating">floating point</a> type
4349 to cast it to. The size of <tt>value</tt> must be larger than the size of
4350 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
4351 <i>no-op cast</i>.</p>
4354 <p>The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
4355 <a href="#t_floating">floating point</a> type to a smaller
4356 <a href="#t_floating">floating point</a> type. If the value cannot fit
4357 within the destination type, <tt>ty2</tt>, then the results are
4362 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
4363 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
4368 <!-- _______________________________________________________________________ -->
4369 <div class="doc_subsubsection">
4370 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
4372 <div class="doc_text">
4376 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
4380 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
4381 floating point value.</p>
4384 <p>The '<tt>fpext</tt>' instruction takes a
4385 <a href="#t_floating">floating point</a> <tt>value</tt> to cast, and
4386 a <a href="#t_floating">floating point</a> type to cast it to. The source
4387 type must be smaller than the destination type.</p>
4390 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
4391 <a href="#t_floating">floating point</a> type to a larger
4392 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
4393 used to make a <i>no-op cast</i> because it always changes bits. Use
4394 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
4398 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
4399 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
4404 <!-- _______________________________________________________________________ -->
4405 <div class="doc_subsubsection">
4406 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
4408 <div class="doc_text">
4412 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
4416 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
4417 unsigned integer equivalent of type <tt>ty2</tt>.</p>
4420 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
4421 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4422 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4423 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4424 vector integer type with the same number of elements as <tt>ty</tt></p>
4427 <p>The '<tt>fptoui</tt>' instruction converts its
4428 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4429 towards zero) unsigned integer value. If the value cannot fit
4430 in <tt>ty2</tt>, the results are undefined.</p>
4434 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
4435 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
4436 %Z = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
4441 <!-- _______________________________________________________________________ -->
4442 <div class="doc_subsubsection">
4443 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
4445 <div class="doc_text">
4449 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
4453 <p>The '<tt>fptosi</tt>' instruction converts
4454 <a href="#t_floating">floating point</a> <tt>value</tt> to
4455 type <tt>ty2</tt>.</p>
4458 <p>The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
4459 scalar or vector <a href="#t_floating">floating point</a> value, and a type
4460 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
4461 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
4462 vector integer type with the same number of elements as <tt>ty</tt></p>
4465 <p>The '<tt>fptosi</tt>' instruction converts its
4466 <a href="#t_floating">floating point</a> operand into the nearest (rounding
4467 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
4468 the results are undefined.</p>
4472 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
4473 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
4474 %Z = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
4479 <!-- _______________________________________________________________________ -->
4480 <div class="doc_subsubsection">
4481 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
4483 <div class="doc_text">
4487 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4491 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
4492 integer and converts that value to the <tt>ty2</tt> type.</p>
4495 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
4496 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4497 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4498 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4499 floating point type with the same number of elements as <tt>ty</tt></p>
4502 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
4503 integer quantity and converts it to the corresponding floating point
4504 value. If the value cannot fit in the floating point value, the results are
4509 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
4510 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
4515 <!-- _______________________________________________________________________ -->
4516 <div class="doc_subsubsection">
4517 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
4519 <div class="doc_text">
4523 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
4527 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed integer
4528 and converts that value to the <tt>ty2</tt> type.</p>
4531 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
4532 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast
4533 it to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
4534 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
4535 floating point type with the same number of elements as <tt>ty</tt></p>
4538 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed integer
4539 quantity and converts it to the corresponding floating point value. If the
4540 value cannot fit in the floating point value, the results are undefined.</p>
4544 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
4545 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
4550 <!-- _______________________________________________________________________ -->
4551 <div class="doc_subsubsection">
4552 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
4554 <div class="doc_text">
4558 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
4562 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
4563 the integer type <tt>ty2</tt>.</p>
4566 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
4567 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
4568 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
4571 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
4572 <tt>ty2</tt> by interpreting the pointer value as an integer and either
4573 truncating or zero extending that value to the size of the integer type. If
4574 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
4575 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
4576 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
4581 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
4582 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
4587 <!-- _______________________________________________________________________ -->
4588 <div class="doc_subsubsection">
4589 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
4591 <div class="doc_text">
4595 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
4599 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to a
4600 pointer type, <tt>ty2</tt>.</p>
4603 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
4604 value to cast, and a type to cast it to, which must be a
4605 <a href="#t_pointer">pointer</a> type.</p>
4608 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
4609 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
4610 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
4611 size of a pointer then a truncation is done. If <tt>value</tt> is smaller
4612 than the size of a pointer then a zero extension is done. If they are the
4613 same size, nothing is done (<i>no-op cast</i>).</p>
4617 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
4618 %Y = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
4619 %Z = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
4624 <!-- _______________________________________________________________________ -->
4625 <div class="doc_subsubsection">
4626 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
4628 <div class="doc_text">
4632 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
4636 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4637 <tt>ty2</tt> without changing any bits.</p>
4640 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be a
4641 non-aggregate first class value, and a type to cast it to, which must also be
4642 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes
4643 of <tt>value</tt> and the destination type, <tt>ty2</tt>, must be
4644 identical. If the source type is a pointer, the destination type must also be
4645 a pointer. This instruction supports bitwise conversion of vectors to
4646 integers and to vectors of other types (as long as they have the same
4650 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
4651 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
4652 this conversion. The conversion is done as if the <tt>value</tt> had been
4653 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only
4654 be converted to other pointer types with this instruction. To convert
4655 pointers to other types, use the <a href="#i_inttoptr">inttoptr</a> or
4656 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
4660 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
4661 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
4662 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
4667 <!-- ======================================================================= -->
4668 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
4670 <div class="doc_text">
4672 <p>The instructions in this category are the "miscellaneous" instructions, which
4673 defy better classification.</p>
4677 <!-- _______________________________________________________________________ -->
4678 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
4681 <div class="doc_text">
4685 <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4689 <p>The '<tt>icmp</tt>' instruction returns a boolean value or a vector of
4690 boolean values based on comparison of its two integer, integer vector, or
4691 pointer operands.</p>
4694 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
4695 the condition code indicating the kind of comparison to perform. It is not a
4696 value, just a keyword. The possible condition code are:</p>
4699 <li><tt>eq</tt>: equal</li>
4700 <li><tt>ne</tt>: not equal </li>
4701 <li><tt>ugt</tt>: unsigned greater than</li>
4702 <li><tt>uge</tt>: unsigned greater or equal</li>
4703 <li><tt>ult</tt>: unsigned less than</li>
4704 <li><tt>ule</tt>: unsigned less or equal</li>
4705 <li><tt>sgt</tt>: signed greater than</li>
4706 <li><tt>sge</tt>: signed greater or equal</li>
4707 <li><tt>slt</tt>: signed less than</li>
4708 <li><tt>sle</tt>: signed less or equal</li>
4711 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
4712 <a href="#t_pointer">pointer</a> or integer <a href="#t_vector">vector</a>
4713 typed. They must also be identical types.</p>
4716 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to the
4717 condition code given as <tt>cond</tt>. The comparison performed always yields
4718 either an <a href="#t_integer"><tt>i1</tt></a> or vector of <tt>i1</tt>
4719 result, as follows:</p>
4722 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
4723 <tt>false</tt> otherwise. No sign interpretation is necessary or
4726 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
4727 <tt>false</tt> otherwise. No sign interpretation is necessary or
4730 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
4731 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4733 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
4734 <tt>true</tt> if <tt>op1</tt> is greater than or equal
4735 to <tt>op2</tt>.</li>
4737 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
4738 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4740 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
4741 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4743 <li><tt>sgt</tt>: interprets the operands as signed values and yields
4744 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4746 <li><tt>sge</tt>: interprets the operands as signed values and yields
4747 <tt>true</tt> if <tt>op1</tt> is greater than or equal
4748 to <tt>op2</tt>.</li>
4750 <li><tt>slt</tt>: interprets the operands as signed values and yields
4751 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4753 <li><tt>sle</tt>: interprets the operands as signed values and yields
4754 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4757 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
4758 values are compared as if they were integers.</p>
4760 <p>If the operands are integer vectors, then they are compared element by
4761 element. The result is an <tt>i1</tt> vector with the same number of elements
4762 as the values being compared. Otherwise, the result is an <tt>i1</tt>.</p>
4766 <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
4767 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
4768 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
4769 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4770 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4771 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
4774 <p>Note that the code generator does not yet support vector types with
4775 the <tt>icmp</tt> instruction.</p>
4779 <!-- _______________________________________________________________________ -->
4780 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4783 <div class="doc_text">
4787 <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4791 <p>The '<tt>fcmp</tt>' instruction returns a boolean value or vector of boolean
4792 values based on comparison of its operands.</p>
4794 <p>If the operands are floating point scalars, then the result type is a boolean
4795 (<a href="#t_integer"><tt>i1</tt></a>).</p>
4797 <p>If the operands are floating point vectors, then the result type is a vector
4798 of boolean with the same number of elements as the operands being
4802 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4803 the condition code indicating the kind of comparison to perform. It is not a
4804 value, just a keyword. The possible condition code are:</p>
4807 <li><tt>false</tt>: no comparison, always returns false</li>
4808 <li><tt>oeq</tt>: ordered and equal</li>
4809 <li><tt>ogt</tt>: ordered and greater than </li>
4810 <li><tt>oge</tt>: ordered and greater than or equal</li>
4811 <li><tt>olt</tt>: ordered and less than </li>
4812 <li><tt>ole</tt>: ordered and less than or equal</li>
4813 <li><tt>one</tt>: ordered and not equal</li>
4814 <li><tt>ord</tt>: ordered (no nans)</li>
4815 <li><tt>ueq</tt>: unordered or equal</li>
4816 <li><tt>ugt</tt>: unordered or greater than </li>
4817 <li><tt>uge</tt>: unordered or greater than or equal</li>
4818 <li><tt>ult</tt>: unordered or less than </li>
4819 <li><tt>ule</tt>: unordered or less than or equal</li>
4820 <li><tt>une</tt>: unordered or not equal</li>
4821 <li><tt>uno</tt>: unordered (either nans)</li>
4822 <li><tt>true</tt>: no comparison, always returns true</li>
4825 <p><i>Ordered</i> means that neither operand is a QNAN while
4826 <i>unordered</i> means that either operand may be a QNAN.</p>
4828 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be either
4829 a <a href="#t_floating">floating point</a> type or
4830 a <a href="#t_vector">vector</a> of floating point type. They must have
4831 identical types.</p>
4834 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4835 according to the condition code given as <tt>cond</tt>. If the operands are
4836 vectors, then the vectors are compared element by element. Each comparison
4837 performed always yields an <a href="#t_integer">i1</a> result, as
4841 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4843 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4844 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4846 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4847 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
4849 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4850 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4852 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4853 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4855 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4856 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4858 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4859 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4861 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4863 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
4864 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4866 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
4867 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4869 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
4870 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4872 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
4873 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4875 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
4876 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4878 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
4879 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4881 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4883 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4888 <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
4889 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4890 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4891 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
4894 <p>Note that the code generator does not yet support vector types with
4895 the <tt>fcmp</tt> instruction.</p>
4899 <!-- _______________________________________________________________________ -->
4900 <div class="doc_subsubsection">
4901 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4904 <div class="doc_text">
4908 <result> = phi <ty> [ <val0>, <label0>], ...
4912 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in the
4913 SSA graph representing the function.</p>
4916 <p>The type of the incoming values is specified with the first type field. After
4917 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
4918 one pair for each predecessor basic block of the current block. Only values
4919 of <a href="#t_firstclass">first class</a> type may be used as the value
4920 arguments to the PHI node. Only labels may be used as the label
4923 <p>There must be no non-phi instructions between the start of a basic block and
4924 the PHI instructions: i.e. PHI instructions must be first in a basic
4927 <p>For the purposes of the SSA form, the use of each incoming value is deemed to
4928 occur on the edge from the corresponding predecessor block to the current
4929 block (but after any definition of an '<tt>invoke</tt>' instruction's return
4930 value on the same edge).</p>
4933 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4934 specified by the pair corresponding to the predecessor basic block that
4935 executed just prior to the current block.</p>
4939 Loop: ; Infinite loop that counts from 0 on up...
4940 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4941 %nextindvar = add i32 %indvar, 1
4947 <!-- _______________________________________________________________________ -->
4948 <div class="doc_subsubsection">
4949 <a name="i_select">'<tt>select</tt>' Instruction</a>
4952 <div class="doc_text">
4956 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
4958 <i>selty</i> is either i1 or {<N x i1>}
4962 <p>The '<tt>select</tt>' instruction is used to choose one value based on a
4963 condition, without branching.</p>
4967 <p>The '<tt>select</tt>' instruction requires an 'i1' value or a vector of 'i1'
4968 values indicating the condition, and two values of the
4969 same <a href="#t_firstclass">first class</a> type. If the val1/val2 are
4970 vectors and the condition is a scalar, then entire vectors are selected, not
4971 individual elements.</p>
4974 <p>If the condition is an i1 and it evaluates to 1, the instruction returns the
4975 first value argument; otherwise, it returns the second value argument.</p>
4977 <p>If the condition is a vector of i1, then the value arguments must be vectors
4978 of the same size, and the selection is done element by element.</p>
4982 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4985 <p>Note that the code generator does not yet support conditions
4986 with vector type.</p>
4990 <!-- _______________________________________________________________________ -->
4991 <div class="doc_subsubsection">
4992 <a name="i_call">'<tt>call</tt>' Instruction</a>
4995 <div class="doc_text">
4999 <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>]
5003 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
5006 <p>This instruction requires several arguments:</p>
5009 <li>The optional "tail" marker indicates whether the callee function accesses
5010 any allocas or varargs in the caller. If the "tail" marker is present,
5011 the function call is eligible for tail call optimization. Note that calls
5012 may be marked "tail" even if they do not occur before
5013 a <a href="#i_ret"><tt>ret</tt></a> instruction.</li>
5015 <li>The optional "cconv" marker indicates which <a href="#callingconv">calling
5016 convention</a> the call should use. If none is specified, the call
5017 defaults to using C calling conventions.</li>
5019 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
5020 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>', and
5021 '<tt>inreg</tt>' attributes are valid here.</li>
5023 <li>'<tt>ty</tt>': the type of the call instruction itself which is also the
5024 type of the return value. Functions that return no value are marked
5025 <tt><a href="#t_void">void</a></tt>.</li>
5027 <li>'<tt>fnty</tt>': shall be the signature of the pointer to function value
5028 being invoked. The argument types must match the types implied by this
5029 signature. This type can be omitted if the function is not varargs and if
5030 the function type does not return a pointer to a function.</li>
5032 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
5033 be invoked. In most cases, this is a direct function invocation, but
5034 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
5035 to function value.</li>
5037 <li>'<tt>function args</tt>': argument list whose types match the function
5038 signature argument types. All arguments must be of
5039 <a href="#t_firstclass">first class</a> type. If the function signature
5040 indicates the function accepts a variable number of arguments, the extra
5041 arguments can be specified.</li>
5043 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
5044 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
5045 '<tt>readnone</tt>' attributes are valid here.</li>
5049 <p>The '<tt>call</tt>' instruction is used to cause control flow to transfer to
5050 a specified function, with its incoming arguments bound to the specified
5051 values. Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called
5052 function, control flow continues with the instruction after the function
5053 call, and the return value of the function is bound to the result
5058 %retval = call i32 @test(i32 %argc)
5059 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
5060 %X = tail call i32 @foo() <i>; yields i32</i>
5061 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
5062 call void %foo(i8 97 signext)
5064 %struct.A = type { i32, i8 }
5065 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
5066 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
5067 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
5068 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
5069 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
5072 <p>llvm treats calls to some functions with names and arguments that match the
5073 standard C99 library as being the C99 library functions, and may perform
5074 optimizations or generate code for them under that assumption. This is
5075 something we'd like to change in the future to provide better support for
5076 freestanding environments and non-C-based langauges.</p>
5080 <!-- _______________________________________________________________________ -->
5081 <div class="doc_subsubsection">
5082 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
5085 <div class="doc_text">
5089 <resultval> = va_arg <va_list*> <arglist>, <argty>
5093 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
5094 the "variable argument" area of a function call. It is used to implement the
5095 <tt>va_arg</tt> macro in C.</p>
5098 <p>This instruction takes a <tt>va_list*</tt> value and the type of the
5099 argument. It returns a value of the specified argument type and increments
5100 the <tt>va_list</tt> to point to the next argument. The actual type
5101 of <tt>va_list</tt> is target specific.</p>
5104 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified type
5105 from the specified <tt>va_list</tt> and causes the <tt>va_list</tt> to point
5106 to the next argument. For more information, see the variable argument
5107 handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
5109 <p>It is legal for this instruction to be called in a function which does not
5110 take a variable number of arguments, for example, the <tt>vfprintf</tt>
5113 <p><tt>va_arg</tt> is an LLVM instruction instead of
5114 an <a href="#intrinsics">intrinsic function</a> because it takes a type as an
5118 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
5120 <p>Note that the code generator does not yet fully support va_arg on many
5121 targets. Also, it does not currently support va_arg with aggregate types on
5126 <!-- *********************************************************************** -->
5127 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
5128 <!-- *********************************************************************** -->
5130 <div class="doc_text">
5132 <p>LLVM supports the notion of an "intrinsic function". These functions have
5133 well known names and semantics and are required to follow certain
5134 restrictions. Overall, these intrinsics represent an extension mechanism for
5135 the LLVM language that does not require changing all of the transformations
5136 in LLVM when adding to the language (or the bitcode reader/writer, the
5137 parser, etc...).</p>
5139 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
5140 prefix is reserved in LLVM for intrinsic names; thus, function names may not
5141 begin with this prefix. Intrinsic functions must always be external
5142 functions: you cannot define the body of intrinsic functions. Intrinsic
5143 functions may only be used in call or invoke instructions: it is illegal to
5144 take the address of an intrinsic function. Additionally, because intrinsic
5145 functions are part of the LLVM language, it is required if any are added that
5146 they be documented here.</p>
5148 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents a
5149 family of functions that perform the same operation but on different data
5150 types. Because LLVM can represent over 8 million different integer types,
5151 overloading is used commonly to allow an intrinsic function to operate on any
5152 integer type. One or more of the argument types or the result type can be
5153 overloaded to accept any integer type. Argument types may also be defined as
5154 exactly matching a previous argument's type or the result type. This allows
5155 an intrinsic function which accepts multiple arguments, but needs all of them
5156 to be of the same type, to only be overloaded with respect to a single
5157 argument or the result.</p>
5159 <p>Overloaded intrinsics will have the names of its overloaded argument types
5160 encoded into its function name, each preceded by a period. Only those types
5161 which are overloaded result in a name suffix. Arguments whose type is matched
5162 against another type do not. For example, the <tt>llvm.ctpop</tt> function
5163 can take an integer of any width and returns an integer of exactly the same
5164 integer width. This leads to a family of functions such as
5165 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29
5166 %val)</tt>. Only one type, the return type, is overloaded, and only one type
5167 suffix is required. Because the argument's type is matched against the return
5168 type, it does not require its own name suffix.</p>
5170 <p>To learn how to add an intrinsic function, please see the
5171 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.</p>
5175 <!-- ======================================================================= -->
5176 <div class="doc_subsection">
5177 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
5180 <div class="doc_text">
5182 <p>Variable argument support is defined in LLVM with
5183 the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
5184 intrinsic functions. These functions are related to the similarly named
5185 macros defined in the <tt><stdarg.h></tt> header file.</p>
5187 <p>All of these functions operate on arguments that use a target-specific value
5188 type "<tt>va_list</tt>". The LLVM assembly language reference manual does
5189 not define what this type is, so all transformations should be prepared to
5190 handle these functions regardless of the type used.</p>
5192 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
5193 instruction and the variable argument handling intrinsic functions are
5196 <div class="doc_code">
5198 define i32 @test(i32 %X, ...) {
5199 ; Initialize variable argument processing
5201 %ap2 = bitcast i8** %ap to i8*
5202 call void @llvm.va_start(i8* %ap2)
5204 ; Read a single integer argument
5205 %tmp = va_arg i8** %ap, i32
5207 ; Demonstrate usage of llvm.va_copy and llvm.va_end
5209 %aq2 = bitcast i8** %aq to i8*
5210 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
5211 call void @llvm.va_end(i8* %aq2)
5213 ; Stop processing of arguments.
5214 call void @llvm.va_end(i8* %ap2)
5218 declare void @llvm.va_start(i8*)
5219 declare void @llvm.va_copy(i8*, i8*)
5220 declare void @llvm.va_end(i8*)
5226 <!-- _______________________________________________________________________ -->
5227 <div class="doc_subsubsection">
5228 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
5232 <div class="doc_text">
5236 declare void %llvm.va_start(i8* <arglist>)
5240 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*<arglist></tt>
5241 for subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt>.</p>
5244 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
5247 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
5248 macro available in C. In a target-dependent way, it initializes
5249 the <tt>va_list</tt> element to which the argument points, so that the next
5250 call to <tt>va_arg</tt> will produce the first variable argument passed to
5251 the function. Unlike the C <tt>va_start</tt> macro, this intrinsic does not
5252 need to know the last argument of the function as the compiler can figure
5257 <!-- _______________________________________________________________________ -->
5258 <div class="doc_subsubsection">
5259 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
5262 <div class="doc_text">
5266 declare void @llvm.va_end(i8* <arglist>)
5270 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
5271 which has been initialized previously
5272 with <tt><a href="#int_va_start">llvm.va_start</a></tt>
5273 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
5276 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
5279 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
5280 macro available in C. In a target-dependent way, it destroys
5281 the <tt>va_list</tt> element to which the argument points. Calls
5282 to <a href="#int_va_start"><tt>llvm.va_start</tt></a>
5283 and <a href="#int_va_copy"> <tt>llvm.va_copy</tt></a> must be matched exactly
5284 with calls to <tt>llvm.va_end</tt>.</p>
5288 <!-- _______________________________________________________________________ -->
5289 <div class="doc_subsubsection">
5290 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
5293 <div class="doc_text">
5297 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
5301 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
5302 from the source argument list to the destination argument list.</p>
5305 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
5306 The second argument is a pointer to a <tt>va_list</tt> element to copy
5310 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
5311 macro available in C. In a target-dependent way, it copies the
5312 source <tt>va_list</tt> element into the destination <tt>va_list</tt>
5313 element. This intrinsic is necessary because
5314 the <tt><a href="#int_va_start"> llvm.va_start</a></tt> intrinsic may be
5315 arbitrarily complex and require, for example, memory allocation.</p>
5319 <!-- ======================================================================= -->
5320 <div class="doc_subsection">
5321 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
5324 <div class="doc_text">
5326 <p>LLVM support for <a href="GarbageCollection.html">Accurate Garbage
5327 Collection</a> (GC) requires the implementation and generation of these
5328 intrinsics. These intrinsics allow identification of <a href="#int_gcroot">GC
5329 roots on the stack</a>, as well as garbage collector implementations that
5330 require <a href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a>
5331 barriers. Front-ends for type-safe garbage collected languages should generate
5332 these intrinsics to make use of the LLVM garbage collectors. For more details,
5333 see <a href="GarbageCollection.html">Accurate Garbage Collection with
5336 <p>The garbage collection intrinsics only operate on objects in the generic
5337 address space (address space zero).</p>
5341 <!-- _______________________________________________________________________ -->
5342 <div class="doc_subsubsection">
5343 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
5346 <div class="doc_text">
5350 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
5354 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
5355 the code generator, and allows some metadata to be associated with it.</p>
5358 <p>The first argument specifies the address of a stack object that contains the
5359 root pointer. The second pointer (which must be either a constant or a
5360 global value address) contains the meta-data to be associated with the
5364 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
5365 location. At compile-time, the code generator generates information to allow
5366 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
5367 intrinsic may only be used in a function which <a href="#gc">specifies a GC
5372 <!-- _______________________________________________________________________ -->
5373 <div class="doc_subsubsection">
5374 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
5377 <div class="doc_text">
5381 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
5385 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
5386 locations, allowing garbage collector implementations that require read
5390 <p>The second argument is the address to read from, which should be an address
5391 allocated from the garbage collector. The first object is a pointer to the
5392 start of the referenced object, if needed by the language runtime (otherwise
5396 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
5397 instruction, but may be replaced with substantially more complex code by the
5398 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
5399 may only be used in a function which <a href="#gc">specifies a GC
5404 <!-- _______________________________________________________________________ -->
5405 <div class="doc_subsubsection">
5406 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
5409 <div class="doc_text">
5413 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
5417 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
5418 locations, allowing garbage collector implementations that require write
5419 barriers (such as generational or reference counting collectors).</p>
5422 <p>The first argument is the reference to store, the second is the start of the
5423 object to store it to, and the third is the address of the field of Obj to
5424 store to. If the runtime does not require a pointer to the object, Obj may
5428 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
5429 instruction, but may be replaced with substantially more complex code by the
5430 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
5431 may only be used in a function which <a href="#gc">specifies a GC
5436 <!-- ======================================================================= -->
5437 <div class="doc_subsection">
5438 <a name="int_codegen">Code Generator Intrinsics</a>
5441 <div class="doc_text">
5443 <p>These intrinsics are provided by LLVM to expose special features that may
5444 only be implemented with code generator support.</p>
5448 <!-- _______________________________________________________________________ -->
5449 <div class="doc_subsubsection">
5450 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
5453 <div class="doc_text">
5457 declare i8 *@llvm.returnaddress(i32 <level>)
5461 <p>The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
5462 target-specific value indicating the return address of the current function
5463 or one of its callers.</p>
5466 <p>The argument to this intrinsic indicates which function to return the address
5467 for. Zero indicates the calling function, one indicates its caller, etc.
5468 The argument is <b>required</b> to be a constant integer value.</p>
5471 <p>The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer
5472 indicating the return address of the specified call frame, or zero if it
5473 cannot be identified. The value returned by this intrinsic is likely to be
5474 incorrect or 0 for arguments other than zero, so it should only be used for
5475 debugging purposes.</p>
5477 <p>Note that calling this intrinsic does not prevent function inlining or other
5478 aggressive transformations, so the value returned may not be that of the
5479 obvious source-language caller.</p>
5483 <!-- _______________________________________________________________________ -->
5484 <div class="doc_subsubsection">
5485 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
5488 <div class="doc_text">
5492 declare i8 *@llvm.frameaddress(i32 <level>)
5496 <p>The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
5497 target-specific frame pointer value for the specified stack frame.</p>
5500 <p>The argument to this intrinsic indicates which function to return the frame
5501 pointer for. Zero indicates the calling function, one indicates its caller,
5502 etc. The argument is <b>required</b> to be a constant integer value.</p>
5505 <p>The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer
5506 indicating the frame address of the specified call frame, or zero if it
5507 cannot be identified. The value returned by this intrinsic is likely to be
5508 incorrect or 0 for arguments other than zero, so it should only be used for
5509 debugging purposes.</p>
5511 <p>Note that calling this intrinsic does not prevent function inlining or other
5512 aggressive transformations, so the value returned may not be that of the
5513 obvious source-language caller.</p>
5517 <!-- _______________________________________________________________________ -->
5518 <div class="doc_subsubsection">
5519 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
5522 <div class="doc_text">
5526 declare i8 *@llvm.stacksave()
5530 <p>The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state
5531 of the function stack, for use
5532 with <a href="#int_stackrestore"> <tt>llvm.stackrestore</tt></a>. This is
5533 useful for implementing language features like scoped automatic variable
5534 sized arrays in C99.</p>
5537 <p>This intrinsic returns a opaque pointer value that can be passed
5538 to <a href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When
5539 an <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved
5540 from <tt>llvm.stacksave</tt>, it effectively restores the state of the stack
5541 to the state it was in when the <tt>llvm.stacksave</tt> intrinsic executed.
5542 In practice, this pops any <a href="#i_alloca">alloca</a> blocks from the
5543 stack that were allocated after the <tt>llvm.stacksave</tt> was executed.</p>
5547 <!-- _______________________________________________________________________ -->
5548 <div class="doc_subsubsection">
5549 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
5552 <div class="doc_text">
5556 declare void @llvm.stackrestore(i8 * %ptr)
5560 <p>The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
5561 the function stack to the state it was in when the
5562 corresponding <a href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic
5563 executed. This is useful for implementing language features like scoped
5564 automatic variable sized arrays in C99.</p>
5567 <p>See the description
5568 for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.</p>
5572 <!-- _______________________________________________________________________ -->
5573 <div class="doc_subsubsection">
5574 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
5577 <div class="doc_text">
5581 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
5585 <p>The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to
5586 insert a prefetch instruction if supported; otherwise, it is a noop.
5587 Prefetches have no effect on the behavior of the program but can change its
5588 performance characteristics.</p>
5591 <p><tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the
5592 specifier determining if the fetch should be for a read (0) or write (1),
5593 and <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5594 locality, to (3) - extremely local keep in cache. The <tt>rw</tt>
5595 and <tt>locality</tt> arguments must be constant integers.</p>
5598 <p>This intrinsic does not modify the behavior of the program. In particular,
5599 prefetches cannot trap and do not produce a value. On targets that support
5600 this intrinsic, the prefetch can provide hints to the processor cache for
5601 better performance.</p>
5605 <!-- _______________________________________________________________________ -->
5606 <div class="doc_subsubsection">
5607 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5610 <div class="doc_text">
5614 declare void @llvm.pcmarker(i32 <id>)
5618 <p>The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program
5619 Counter (PC) in a region of code to simulators and other tools. The method
5620 is target specific, but it is expected that the marker will use exported
5621 symbols to transmit the PC of the marker. The marker makes no guarantees
5622 that it will remain with any specific instruction after optimizations. It is
5623 possible that the presence of a marker will inhibit optimizations. The
5624 intended use is to be inserted after optimizations to allow correlations of
5625 simulation runs.</p>
5628 <p><tt>id</tt> is a numerical id identifying the marker.</p>
5631 <p>This intrinsic does not modify the behavior of the program. Backends that do
5632 not support this intrinisic may ignore it.</p>
5636 <!-- _______________________________________________________________________ -->
5637 <div class="doc_subsubsection">
5638 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5641 <div class="doc_text">
5645 declare i64 @llvm.readcyclecounter( )
5649 <p>The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5650 counter register (or similar low latency, high accuracy clocks) on those
5651 targets that support it. On X86, it should map to RDTSC. On Alpha, it
5652 should map to RPCC. As the backing counters overflow quickly (on the order
5653 of 9 seconds on alpha), this should only be used for small timings.</p>
5656 <p>When directly supported, reading the cycle counter should not modify any
5657 memory. Implementations are allowed to either return a application specific
5658 value or a system wide value. On backends without support, this is lowered
5659 to a constant 0.</p>
5663 <!-- ======================================================================= -->
5664 <div class="doc_subsection">
5665 <a name="int_libc">Standard C Library Intrinsics</a>
5668 <div class="doc_text">
5670 <p>LLVM provides intrinsics for a few important standard C library functions.
5671 These intrinsics allow source-language front-ends to pass information about
5672 the alignment of the pointer arguments to the code generator, providing
5673 opportunity for more efficient code generation.</p>
5677 <!-- _______________________________________________________________________ -->
5678 <div class="doc_subsubsection">
5679 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5682 <div class="doc_text">
5685 <p>This is an overloaded intrinsic. You can use <tt>llvm.memcpy</tt> on any
5686 integer bit width. Not all targets support all bit widths however.</p>
5689 declare void @llvm.memcpy.i8(i8 * <dest>, i8 * <src>,
5690 i8 <len>, i32 <align>)
5691 declare void @llvm.memcpy.i16(i8 * <dest>, i8 * <src>,
5692 i16 <len>, i32 <align>)
5693 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
5694 i32 <len>, i32 <align>)
5695 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
5696 i64 <len>, i32 <align>)
5700 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
5701 source location to the destination location.</p>
5703 <p>Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5704 intrinsics do not return a value, and takes an extra alignment argument.</p>
5707 <p>The first argument is a pointer to the destination, the second is a pointer
5708 to the source. The third argument is an integer argument specifying the
5709 number of bytes to copy, and the fourth argument is the alignment of the
5710 source and destination locations.</p>
5712 <p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5713 then the caller guarantees that both the source and destination pointers are
5714 aligned to that boundary.</p>
5717 <p>The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the
5718 source location to the destination location, which are not allowed to
5719 overlap. It copies "len" bytes of memory over. If the argument is known to
5720 be aligned to some boundary, this can be specified as the fourth argument,
5721 otherwise it should be set to 0 or 1.</p>
5725 <!-- _______________________________________________________________________ -->
5726 <div class="doc_subsubsection">
5727 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5730 <div class="doc_text">
5733 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5734 width. Not all targets support all bit widths however.</p>
5737 declare void @llvm.memmove.i8(i8 * <dest>, i8 * <src>,
5738 i8 <len>, i32 <align>)
5739 declare void @llvm.memmove.i16(i8 * <dest>, i8 * <src>,
5740 i16 <len>, i32 <align>)
5741 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
5742 i32 <len>, i32 <align>)
5743 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
5744 i64 <len>, i32 <align>)
5748 <p>The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the
5749 source location to the destination location. It is similar to the
5750 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to
5753 <p>Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5754 intrinsics do not return a value, and takes an extra alignment argument.</p>
5757 <p>The first argument is a pointer to the destination, the second is a pointer
5758 to the source. The third argument is an integer argument specifying the
5759 number of bytes to copy, and the fourth argument is the alignment of the
5760 source and destination locations.</p>
5762 <p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5763 then the caller guarantees that the source and destination pointers are
5764 aligned to that boundary.</p>
5767 <p>The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the
5768 source location to the destination location, which may overlap. It copies
5769 "len" bytes of memory over. If the argument is known to be aligned to some
5770 boundary, this can be specified as the fourth argument, otherwise it should
5771 be set to 0 or 1.</p>
5775 <!-- _______________________________________________________________________ -->
5776 <div class="doc_subsubsection">
5777 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5780 <div class="doc_text">
5783 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5784 width. Not all targets support all bit widths however.</p>
5787 declare void @llvm.memset.i8(i8 * <dest>, i8 <val>,
5788 i8 <len>, i32 <align>)
5789 declare void @llvm.memset.i16(i8 * <dest>, i8 <val>,
5790 i16 <len>, i32 <align>)
5791 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
5792 i32 <len>, i32 <align>)
5793 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
5794 i64 <len>, i32 <align>)
5798 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a
5799 particular byte value.</p>
5801 <p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
5802 intrinsic does not return a value, and takes an extra alignment argument.</p>
5805 <p>The first argument is a pointer to the destination to fill, the second is the
5806 byte value to fill it with, the third argument is an integer argument
5807 specifying the number of bytes to fill, and the fourth argument is the known
5808 alignment of destination location.</p>
5810 <p>If the call to this intrinisic has an alignment value that is not 0 or 1,
5811 then the caller guarantees that the destination pointer is aligned to that
5815 <p>The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting
5816 at the destination location. If the argument is known to be aligned to some
5817 boundary, this can be specified as the fourth argument, otherwise it should
5818 be set to 0 or 1.</p>
5822 <!-- _______________________________________________________________________ -->
5823 <div class="doc_subsubsection">
5824 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5827 <div class="doc_text">
5830 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
5831 floating point or vector of floating point type. Not all targets support all
5835 declare float @llvm.sqrt.f32(float %Val)
5836 declare double @llvm.sqrt.f64(double %Val)
5837 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5838 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5839 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
5843 <p>The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
5844 returning the same value as the libm '<tt>sqrt</tt>' functions would.
5845 Unlike <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined
5846 behavior for negative numbers other than -0.0 (which allows for better
5847 optimization, because there is no need to worry about errno being
5848 set). <tt>llvm.sqrt(-0.0)</tt> is defined to return -0.0 like IEEE sqrt.</p>
5851 <p>The argument and return value are floating point numbers of the same
5855 <p>This function returns the sqrt of the specified operand if it is a
5856 nonnegative floating point number.</p>
5860 <!-- _______________________________________________________________________ -->
5861 <div class="doc_subsubsection">
5862 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5865 <div class="doc_text">
5868 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
5869 floating point or vector of floating point type. Not all targets support all
5873 declare float @llvm.powi.f32(float %Val, i32 %power)
5874 declare double @llvm.powi.f64(double %Val, i32 %power)
5875 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5876 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5877 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
5881 <p>The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5882 specified (positive or negative) power. The order of evaluation of
5883 multiplications is not defined. When a vector of floating point type is
5884 used, the second argument remains a scalar integer value.</p>
5887 <p>The second argument is an integer power, and the first is a value to raise to
5891 <p>This function returns the first value raised to the second power with an
5892 unspecified sequence of rounding operations.</p>
5896 <!-- _______________________________________________________________________ -->
5897 <div class="doc_subsubsection">
5898 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5901 <div class="doc_text">
5904 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5905 floating point or vector of floating point type. Not all targets support all
5909 declare float @llvm.sin.f32(float %Val)
5910 declare double @llvm.sin.f64(double %Val)
5911 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5912 declare fp128 @llvm.sin.f128(fp128 %Val)
5913 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5917 <p>The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.</p>
5920 <p>The argument and return value are floating point numbers of the same
5924 <p>This function returns the sine of the specified operand, returning the same
5925 values as the libm <tt>sin</tt> functions would, and handles error conditions
5926 in the same way.</p>
5930 <!-- _______________________________________________________________________ -->
5931 <div class="doc_subsubsection">
5932 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5935 <div class="doc_text">
5938 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5939 floating point or vector of floating point type. Not all targets support all
5943 declare float @llvm.cos.f32(float %Val)
5944 declare double @llvm.cos.f64(double %Val)
5945 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5946 declare fp128 @llvm.cos.f128(fp128 %Val)
5947 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5951 <p>The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.</p>
5954 <p>The argument and return value are floating point numbers of the same
5958 <p>This function returns the cosine of the specified operand, returning the same
5959 values as the libm <tt>cos</tt> functions would, and handles error conditions
5960 in the same way.</p>
5964 <!-- _______________________________________________________________________ -->
5965 <div class="doc_subsubsection">
5966 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5969 <div class="doc_text">
5972 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5973 floating point or vector of floating point type. Not all targets support all
5977 declare float @llvm.pow.f32(float %Val, float %Power)
5978 declare double @llvm.pow.f64(double %Val, double %Power)
5979 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5980 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5981 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5985 <p>The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5986 specified (positive or negative) power.</p>
5989 <p>The second argument is a floating point power, and the first is a value to
5990 raise to that power.</p>
5993 <p>This function returns the first value raised to the second power, returning
5994 the same values as the libm <tt>pow</tt> functions would, and handles error
5995 conditions in the same way.</p>
5999 <!-- ======================================================================= -->
6000 <div class="doc_subsection">
6001 <a name="int_manip">Bit Manipulation Intrinsics</a>
6004 <div class="doc_text">
6006 <p>LLVM provides intrinsics for a few important bit manipulation operations.
6007 These allow efficient code generation for some algorithms.</p>
6011 <!-- _______________________________________________________________________ -->
6012 <div class="doc_subsubsection">
6013 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
6016 <div class="doc_text">
6019 <p>This is an overloaded intrinsic function. You can use bswap on any integer
6020 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
6023 declare i16 @llvm.bswap.i16(i16 <id>)
6024 declare i32 @llvm.bswap.i32(i32 <id>)
6025 declare i64 @llvm.bswap.i64(i64 <id>)
6029 <p>The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
6030 values with an even number of bytes (positive multiple of 16 bits). These
6031 are useful for performing operations on data that is not in the target's
6032 native byte order.</p>
6035 <p>The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
6036 and low byte of the input i16 swapped. Similarly,
6037 the <tt>llvm.bswap.i32</tt> intrinsic returns an i32 value that has the four
6038 bytes of the input i32 swapped, so that if the input bytes are numbered 0, 1,
6039 2, 3 then the returned i32 will have its bytes in 3, 2, 1, 0 order.
6040 The <tt>llvm.bswap.i48</tt>, <tt>llvm.bswap.i64</tt> and other intrinsics
6041 extend this concept to additional even-byte lengths (6 bytes, 8 bytes and
6042 more, respectively).</p>
6046 <!-- _______________________________________________________________________ -->
6047 <div class="doc_subsubsection">
6048 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
6051 <div class="doc_text">
6054 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
6055 width. Not all targets support all bit widths however.</p>
6058 declare i8 @llvm.ctpop.i8(i8 <src>)
6059 declare i16 @llvm.ctpop.i16(i16 <src>)
6060 declare i32 @llvm.ctpop.i32(i32 <src>)
6061 declare i64 @llvm.ctpop.i64(i64 <src>)
6062 declare i256 @llvm.ctpop.i256(i256 <src>)
6066 <p>The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set
6070 <p>The only argument is the value to be counted. The argument may be of any
6071 integer type. The return type must match the argument type.</p>
6074 <p>The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.</p>
6078 <!-- _______________________________________________________________________ -->
6079 <div class="doc_subsubsection">
6080 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
6083 <div class="doc_text">
6086 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
6087 integer bit width. Not all targets support all bit widths however.</p>
6090 declare i8 @llvm.ctlz.i8 (i8 <src>)
6091 declare i16 @llvm.ctlz.i16(i16 <src>)
6092 declare i32 @llvm.ctlz.i32(i32 <src>)
6093 declare i64 @llvm.ctlz.i64(i64 <src>)
6094 declare i256 @llvm.ctlz.i256(i256 <src>)
6098 <p>The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
6099 leading zeros in a variable.</p>
6102 <p>The only argument is the value to be counted. The argument may be of any
6103 integer type. The return type must match the argument type.</p>
6106 <p>The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant)
6107 zeros in a variable. If the src == 0 then the result is the size in bits of
6108 the type of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.</p>
6112 <!-- _______________________________________________________________________ -->
6113 <div class="doc_subsubsection">
6114 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
6117 <div class="doc_text">
6120 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
6121 integer bit width. Not all targets support all bit widths however.</p>
6124 declare i8 @llvm.cttz.i8 (i8 <src>)
6125 declare i16 @llvm.cttz.i16(i16 <src>)
6126 declare i32 @llvm.cttz.i32(i32 <src>)
6127 declare i64 @llvm.cttz.i64(i64 <src>)
6128 declare i256 @llvm.cttz.i256(i256 <src>)
6132 <p>The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
6136 <p>The only argument is the value to be counted. The argument may be of any
6137 integer type. The return type must match the argument type.</p>
6140 <p>The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant)
6141 zeros in a variable. If the src == 0 then the result is the size in bits of
6142 the type of src. For example, <tt>llvm.cttz(2) = 1</tt>.</p>
6146 <!-- ======================================================================= -->
6147 <div class="doc_subsection">
6148 <a name="int_overflow">Arithmetic with Overflow Intrinsics</a>
6151 <div class="doc_text">
6153 <p>LLVM provides intrinsics for some arithmetic with overflow operations.</p>
6157 <!-- _______________________________________________________________________ -->
6158 <div class="doc_subsubsection">
6159 <a name="int_sadd_overflow">'<tt>llvm.sadd.with.overflow.*</tt>' Intrinsics</a>
6162 <div class="doc_text">
6165 <p>This is an overloaded intrinsic. You can use <tt>llvm.sadd.with.overflow</tt>
6166 on any integer bit width.</p>
6169 declare {i16, i1} @llvm.sadd.with.overflow.i16(i16 %a, i16 %b)
6170 declare {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6171 declare {i64, i1} @llvm.sadd.with.overflow.i64(i64 %a, i64 %b)
6175 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6176 a signed addition of the two arguments, and indicate whether an overflow
6177 occurred during the signed summation.</p>
6180 <p>The arguments (%a and %b) and the first element of the result structure may
6181 be of integer types of any bit width, but they must have the same bit
6182 width. The second element of the result structure must be of
6183 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6184 undergo signed addition.</p>
6187 <p>The '<tt>llvm.sadd.with.overflow</tt>' family of intrinsic functions perform
6188 a signed addition of the two variables. They return a structure — the
6189 first element of which is the signed summation, and the second element of
6190 which is a bit specifying if the signed summation resulted in an
6195 %res = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %b)
6196 %sum = extractvalue {i32, i1} %res, 0
6197 %obit = extractvalue {i32, i1} %res, 1
6198 br i1 %obit, label %overflow, label %normal
6203 <!-- _______________________________________________________________________ -->
6204 <div class="doc_subsubsection">
6205 <a name="int_uadd_overflow">'<tt>llvm.uadd.with.overflow.*</tt>' Intrinsics</a>
6208 <div class="doc_text">
6211 <p>This is an overloaded intrinsic. You can use <tt>llvm.uadd.with.overflow</tt>
6212 on any integer bit width.</p>
6215 declare {i16, i1} @llvm.uadd.with.overflow.i16(i16 %a, i16 %b)
6216 declare {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6217 declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b)
6221 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6222 an unsigned addition of the two arguments, and indicate whether a carry
6223 occurred during the unsigned summation.</p>
6226 <p>The arguments (%a and %b) and the first element of the result structure may
6227 be of integer types of any bit width, but they must have the same bit
6228 width. The second element of the result structure must be of
6229 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6230 undergo unsigned addition.</p>
6233 <p>The '<tt>llvm.uadd.with.overflow</tt>' family of intrinsic functions perform
6234 an unsigned addition of the two arguments. They return a structure —
6235 the first element of which is the sum, and the second element of which is a
6236 bit specifying if the unsigned summation resulted in a carry.</p>
6240 %res = call {i32, i1} @llvm.uadd.with.overflow.i32(i32 %a, i32 %b)
6241 %sum = extractvalue {i32, i1} %res, 0
6242 %obit = extractvalue {i32, i1} %res, 1
6243 br i1 %obit, label %carry, label %normal
6248 <!-- _______________________________________________________________________ -->
6249 <div class="doc_subsubsection">
6250 <a name="int_ssub_overflow">'<tt>llvm.ssub.with.overflow.*</tt>' Intrinsics</a>
6253 <div class="doc_text">
6256 <p>This is an overloaded intrinsic. You can use <tt>llvm.ssub.with.overflow</tt>
6257 on any integer bit width.</p>
6260 declare {i16, i1} @llvm.ssub.with.overflow.i16(i16 %a, i16 %b)
6261 declare {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6262 declare {i64, i1} @llvm.ssub.with.overflow.i64(i64 %a, i64 %b)
6266 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6267 a signed subtraction of the two arguments, and indicate whether an overflow
6268 occurred during the signed subtraction.</p>
6271 <p>The arguments (%a and %b) and the first element of the result structure may
6272 be of integer types of any bit width, but they must have the same bit
6273 width. The second element of the result structure must be of
6274 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6275 undergo signed subtraction.</p>
6278 <p>The '<tt>llvm.ssub.with.overflow</tt>' family of intrinsic functions perform
6279 a signed subtraction of the two arguments. They return a structure —
6280 the first element of which is the subtraction, and the second element of
6281 which is a bit specifying if the signed subtraction resulted in an
6286 %res = call {i32, i1} @llvm.ssub.with.overflow.i32(i32 %a, i32 %b)
6287 %sum = extractvalue {i32, i1} %res, 0
6288 %obit = extractvalue {i32, i1} %res, 1
6289 br i1 %obit, label %overflow, label %normal
6294 <!-- _______________________________________________________________________ -->
6295 <div class="doc_subsubsection">
6296 <a name="int_usub_overflow">'<tt>llvm.usub.with.overflow.*</tt>' Intrinsics</a>
6299 <div class="doc_text">
6302 <p>This is an overloaded intrinsic. You can use <tt>llvm.usub.with.overflow</tt>
6303 on any integer bit width.</p>
6306 declare {i16, i1} @llvm.usub.with.overflow.i16(i16 %a, i16 %b)
6307 declare {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6308 declare {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b)
6312 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6313 an unsigned subtraction of the two arguments, and indicate whether an
6314 overflow occurred during the unsigned subtraction.</p>
6317 <p>The arguments (%a and %b) and the first element of the result structure may
6318 be of integer types of any bit width, but they must have the same bit
6319 width. The second element of the result structure must be of
6320 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6321 undergo unsigned subtraction.</p>
6324 <p>The '<tt>llvm.usub.with.overflow</tt>' family of intrinsic functions perform
6325 an unsigned subtraction of the two arguments. They return a structure —
6326 the first element of which is the subtraction, and the second element of
6327 which is a bit specifying if the unsigned subtraction resulted in an
6332 %res = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
6333 %sum = extractvalue {i32, i1} %res, 0
6334 %obit = extractvalue {i32, i1} %res, 1
6335 br i1 %obit, label %overflow, label %normal
6340 <!-- _______________________________________________________________________ -->
6341 <div class="doc_subsubsection">
6342 <a name="int_smul_overflow">'<tt>llvm.smul.with.overflow.*</tt>' Intrinsics</a>
6345 <div class="doc_text">
6348 <p>This is an overloaded intrinsic. You can use <tt>llvm.smul.with.overflow</tt>
6349 on any integer bit width.</p>
6352 declare {i16, i1} @llvm.smul.with.overflow.i16(i16 %a, i16 %b)
6353 declare {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6354 declare {i64, i1} @llvm.smul.with.overflow.i64(i64 %a, i64 %b)
6359 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6360 a signed multiplication of the two arguments, and indicate whether an
6361 overflow occurred during the signed multiplication.</p>
6364 <p>The arguments (%a and %b) and the first element of the result structure may
6365 be of integer types of any bit width, but they must have the same bit
6366 width. The second element of the result structure must be of
6367 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6368 undergo signed multiplication.</p>
6371 <p>The '<tt>llvm.smul.with.overflow</tt>' family of intrinsic functions perform
6372 a signed multiplication of the two arguments. They return a structure —
6373 the first element of which is the multiplication, and the second element of
6374 which is a bit specifying if the signed multiplication resulted in an
6379 %res = call {i32, i1} @llvm.smul.with.overflow.i32(i32 %a, i32 %b)
6380 %sum = extractvalue {i32, i1} %res, 0
6381 %obit = extractvalue {i32, i1} %res, 1
6382 br i1 %obit, label %overflow, label %normal
6387 <!-- _______________________________________________________________________ -->
6388 <div class="doc_subsubsection">
6389 <a name="int_umul_overflow">'<tt>llvm.umul.with.overflow.*</tt>' Intrinsics</a>
6392 <div class="doc_text">
6395 <p>This is an overloaded intrinsic. You can use <tt>llvm.umul.with.overflow</tt>
6396 on any integer bit width.</p>
6399 declare {i16, i1} @llvm.umul.with.overflow.i16(i16 %a, i16 %b)
6400 declare {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6401 declare {i64, i1} @llvm.umul.with.overflow.i64(i64 %a, i64 %b)
6405 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6406 a unsigned multiplication of the two arguments, and indicate whether an
6407 overflow occurred during the unsigned multiplication.</p>
6410 <p>The arguments (%a and %b) and the first element of the result structure may
6411 be of integer types of any bit width, but they must have the same bit
6412 width. The second element of the result structure must be of
6413 type <tt>i1</tt>. <tt>%a</tt> and <tt>%b</tt> are the two values that will
6414 undergo unsigned multiplication.</p>
6417 <p>The '<tt>llvm.umul.with.overflow</tt>' family of intrinsic functions perform
6418 an unsigned multiplication of the two arguments. They return a structure
6419 — the first element of which is the multiplication, and the second
6420 element of which is a bit specifying if the unsigned multiplication resulted
6425 %res = call {i32, i1} @llvm.umul.with.overflow.i32(i32 %a, i32 %b)
6426 %sum = extractvalue {i32, i1} %res, 0
6427 %obit = extractvalue {i32, i1} %res, 1
6428 br i1 %obit, label %overflow, label %normal
6433 <!-- ======================================================================= -->
6434 <div class="doc_subsection">
6435 <a name="int_debugger">Debugger Intrinsics</a>
6438 <div class="doc_text">
6440 <p>The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt>
6441 prefix), are described in
6442 the <a href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source
6443 Level Debugging</a> document.</p>
6447 <!-- ======================================================================= -->
6448 <div class="doc_subsection">
6449 <a name="int_eh">Exception Handling Intrinsics</a>
6452 <div class="doc_text">
6454 <p>The LLVM exception handling intrinsics (which all start with
6455 <tt>llvm.eh.</tt> prefix), are described in
6456 the <a href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
6457 Handling</a> document.</p>
6461 <!-- ======================================================================= -->
6462 <div class="doc_subsection">
6463 <a name="int_trampoline">Trampoline Intrinsic</a>
6466 <div class="doc_text">
6468 <p>This intrinsic makes it possible to excise one parameter, marked with
6469 the <tt>nest</tt> attribute, from a function. The result is a callable
6470 function pointer lacking the nest parameter - the caller does not need to
6471 provide a value for it. Instead, the value to use is stored in advance in a
6472 "trampoline", a block of memory usually allocated on the stack, which also
6473 contains code to splice the nest value into the argument list. This is used
6474 to implement the GCC nested function address extension.</p>
6476 <p>For example, if the function is
6477 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
6478 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as
6481 <div class="doc_code">
6483 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
6484 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
6485 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
6486 %fp = bitcast i8* %p to i32 (i32, i32)*
6490 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
6491 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
6495 <!-- _______________________________________________________________________ -->
6496 <div class="doc_subsubsection">
6497 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
6500 <div class="doc_text">
6504 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
6508 <p>This fills the memory pointed to by <tt>tramp</tt> with code and returns a
6509 function pointer suitable for executing it.</p>
6512 <p>The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
6513 pointers. The <tt>tramp</tt> argument must point to a sufficiently large and
6514 sufficiently aligned block of memory; this memory is written to by the
6515 intrinsic. Note that the size and the alignment are target-specific - LLVM
6516 currently provides no portable way of determining them, so a front-end that
6517 generates this intrinsic needs to have some target-specific knowledge.
6518 The <tt>func</tt> argument must hold a function bitcast to
6519 an <tt>i8*</tt>.</p>
6522 <p>The block of memory pointed to by <tt>tramp</tt> is filled with target
6523 dependent code, turning it into a function. A pointer to this function is
6524 returned, but needs to be bitcast to an <a href="#int_trampoline">appropriate
6525 function pointer type</a> before being called. The new function's signature
6526 is the same as that of <tt>func</tt> with any arguments marked with
6527 the <tt>nest</tt> attribute removed. At most one such <tt>nest</tt> argument
6528 is allowed, and it must be of pointer type. Calling the new function is
6529 equivalent to calling <tt>func</tt> with the same argument list, but
6530 with <tt>nval</tt> used for the missing <tt>nest</tt> argument. If, after
6531 calling <tt>llvm.init.trampoline</tt>, the memory pointed to
6532 by <tt>tramp</tt> is modified, then the effect of any later call to the
6533 returned function pointer is undefined.</p>
6537 <!-- ======================================================================= -->
6538 <div class="doc_subsection">
6539 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
6542 <div class="doc_text">
6544 <p>These intrinsic functions expand the "universal IR" of LLVM to represent
6545 hardware constructs for atomic operations and memory synchronization. This
6546 provides an interface to the hardware, not an interface to the programmer. It
6547 is aimed at a low enough level to allow any programming models or APIs
6548 (Application Programming Interfaces) which need atomic behaviors to map
6549 cleanly onto it. It is also modeled primarily on hardware behavior. Just as
6550 hardware provides a "universal IR" for source languages, it also provides a
6551 starting point for developing a "universal" atomic operation and
6552 synchronization IR.</p>
6554 <p>These do <em>not</em> form an API such as high-level threading libraries,
6555 software transaction memory systems, atomic primitives, and intrinsic
6556 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
6557 application libraries. The hardware interface provided by LLVM should allow
6558 a clean implementation of all of these APIs and parallel programming models.
6559 No one model or paradigm should be selected above others unless the hardware
6560 itself ubiquitously does so.</p>
6564 <!-- _______________________________________________________________________ -->
6565 <div class="doc_subsubsection">
6566 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
6568 <div class="doc_text">
6571 declare void @llvm.memory.barrier( i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>, i1 <device> )
6575 <p>The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
6576 specific pairs of memory access types.</p>
6579 <p>The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
6580 The first four arguments enables a specific barrier as listed below. The
6581 fith argument specifies that the barrier applies to io or device or uncached
6585 <li><tt>ll</tt>: load-load barrier</li>
6586 <li><tt>ls</tt>: load-store barrier</li>
6587 <li><tt>sl</tt>: store-load barrier</li>
6588 <li><tt>ss</tt>: store-store barrier</li>
6589 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
6593 <p>This intrinsic causes the system to enforce some ordering constraints upon
6594 the loads and stores of the program. This barrier does not
6595 indicate <em>when</em> any events will occur, it only enforces
6596 an <em>order</em> in which they occur. For any of the specified pairs of load
6597 and store operations (f.ex. load-load, or store-load), all of the first
6598 operations preceding the barrier will complete before any of the second
6599 operations succeeding the barrier begin. Specifically the semantics for each
6600 pairing is as follows:</p>
6603 <li><tt>ll</tt>: All loads before the barrier must complete before any load
6604 after the barrier begins.</li>
6605 <li><tt>ls</tt>: All loads before the barrier must complete before any
6606 store after the barrier begins.</li>
6607 <li><tt>ss</tt>: All stores before the barrier must complete before any
6608 store after the barrier begins.</li>
6609 <li><tt>sl</tt>: All stores before the barrier must complete before any
6610 load after the barrier begins.</li>
6613 <p>These semantics are applied with a logical "and" behavior when more than one
6614 is enabled in a single memory barrier intrinsic.</p>
6616 <p>Backends may implement stronger barriers than those requested when they do
6617 not support as fine grained a barrier as requested. Some architectures do
6618 not need all types of barriers and on such architectures, these become
6623 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6624 %ptr = bitcast i8* %mallocP to i32*
6627 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
6628 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
6629 <i>; guarantee the above finishes</i>
6630 store i32 8, %ptr <i>; before this begins</i>
6635 <!-- _______________________________________________________________________ -->
6636 <div class="doc_subsubsection">
6637 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
6640 <div class="doc_text">
6643 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
6644 any integer bit width and for different address spaces. Not all targets
6645 support all bit widths however.</p>
6648 declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* <ptr>, i8 <cmp>, i8 <val> )
6649 declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* <ptr>, i16 <cmp>, i16 <val> )
6650 declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* <ptr>, i32 <cmp>, i32 <val> )
6651 declare i64 @llvm.atomic.cmp.swap.i64.p0i64( i64* <ptr>, i64 <cmp>, i64 <val> )
6655 <p>This loads a value in memory and compares it to a given value. If they are
6656 equal, it stores a new value into the memory.</p>
6659 <p>The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result
6660 as well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6661 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6662 this integer type. While any bit width integer may be used, targets may only
6663 lower representations they support in hardware.</p>
6666 <p>This entire intrinsic must be executed atomically. It first loads the value
6667 in memory pointed to by <tt>ptr</tt> and compares it with the
6668 value <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the
6669 memory. The loaded value is yielded in all cases. This provides the
6670 equivalent of an atomic compare-and-swap operation within the SSA
6675 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6676 %ptr = bitcast i8* %mallocP to i32*
6679 %val1 = add i32 4, 4
6680 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
6681 <i>; yields {i32}:result1 = 4</i>
6682 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6683 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6685 %val2 = add i32 1, 1
6686 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
6687 <i>; yields {i32}:result2 = 8</i>
6688 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6690 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6695 <!-- _______________________________________________________________________ -->
6696 <div class="doc_subsubsection">
6697 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6699 <div class="doc_text">
6702 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6703 integer bit width. Not all targets support all bit widths however.</p>
6706 declare i8 @llvm.atomic.swap.i8.p0i8( i8* <ptr>, i8 <val> )
6707 declare i16 @llvm.atomic.swap.i16.p0i16( i16* <ptr>, i16 <val> )
6708 declare i32 @llvm.atomic.swap.i32.p0i32( i32* <ptr>, i32 <val> )
6709 declare i64 @llvm.atomic.swap.i64.p0i64( i64* <ptr>, i64 <val> )
6713 <p>This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6714 the value from memory. It then stores the value in <tt>val</tt> in the memory
6715 at <tt>ptr</tt>.</p>
6718 <p>The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both
6719 the <tt>val</tt> argument and the result must be integers of the same bit
6720 width. The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6721 integer type. The targets may only lower integer representations they
6725 <p>This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6726 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6727 equivalent of an atomic swap operation within the SSA framework.</p>
6731 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6732 %ptr = bitcast i8* %mallocP to i32*
6735 %val1 = add i32 4, 4
6736 %result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
6737 <i>; yields {i32}:result1 = 4</i>
6738 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6739 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6741 %val2 = add i32 1, 1
6742 %result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
6743 <i>; yields {i32}:result2 = 8</i>
6745 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6746 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6751 <!-- _______________________________________________________________________ -->
6752 <div class="doc_subsubsection">
6753 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
6757 <div class="doc_text">
6760 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on
6761 any integer bit width. Not all targets support all bit widths however.</p>
6764 declare i8 @llvm.atomic.load.add.i8..p0i8( i8* <ptr>, i8 <delta> )
6765 declare i16 @llvm.atomic.load.add.i16..p0i16( i16* <ptr>, i16 <delta> )
6766 declare i32 @llvm.atomic.load.add.i32..p0i32( i32* <ptr>, i32 <delta> )
6767 declare i64 @llvm.atomic.load.add.i64..p0i64( i64* <ptr>, i64 <delta> )
6771 <p>This intrinsic adds <tt>delta</tt> to the value stored in memory
6772 at <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
6775 <p>The intrinsic takes two arguments, the first a pointer to an integer value
6776 and the second an integer value. The result is also an integer value. These
6777 integer types can have any bit width, but they must all have the same bit
6778 width. The targets may only lower integer representations they support.</p>
6781 <p>This intrinsic does a series of operations atomically. It first loads the
6782 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6783 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.</p>
6787 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6788 %ptr = bitcast i8* %mallocP to i32*
6790 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
6791 <i>; yields {i32}:result1 = 4</i>
6792 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
6793 <i>; yields {i32}:result2 = 8</i>
6794 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
6795 <i>; yields {i32}:result3 = 10</i>
6796 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
6801 <!-- _______________________________________________________________________ -->
6802 <div class="doc_subsubsection">
6803 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6807 <div class="doc_text">
6810 <p>This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
6811 any integer bit width and for different address spaces. Not all targets
6812 support all bit widths however.</p>
6815 declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* <ptr>, i8 <delta> )
6816 declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* <ptr>, i16 <delta> )
6817 declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* <ptr>, i32 <delta> )
6818 declare i64 @llvm.atomic.load.sub.i64.p0i32( i64* <ptr>, i64 <delta> )
6822 <p>This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6823 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.</p>
6826 <p>The intrinsic takes two arguments, the first a pointer to an integer value
6827 and the second an integer value. The result is also an integer value. These
6828 integer types can have any bit width, but they must all have the same bit
6829 width. The targets may only lower integer representations they support.</p>
6832 <p>This intrinsic does a series of operations atomically. It first loads the
6833 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6834 result to <tt>ptr</tt>. It yields the original value stored
6835 at <tt>ptr</tt>.</p>
6839 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6840 %ptr = bitcast i8* %mallocP to i32*
6842 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
6843 <i>; yields {i32}:result1 = 8</i>
6844 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
6845 <i>; yields {i32}:result2 = 4</i>
6846 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
6847 <i>; yields {i32}:result3 = 2</i>
6848 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6853 <!-- _______________________________________________________________________ -->
6854 <div class="doc_subsubsection">
6855 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6856 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6857 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6858 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6861 <div class="doc_text">
6864 <p>These are overloaded intrinsics. You can
6865 use <tt>llvm.atomic.load_and</tt>, <tt>llvm.atomic.load_nand</tt>,
6866 <tt>llvm.atomic.load_or</tt>, and <tt>llvm.atomic.load_xor</tt> on any integer
6867 bit width and for different address spaces. Not all targets support all bit
6871 declare i8 @llvm.atomic.load.and.i8.p0i8( i8* <ptr>, i8 <delta> )
6872 declare i16 @llvm.atomic.load.and.i16.p0i16( i16* <ptr>, i16 <delta> )
6873 declare i32 @llvm.atomic.load.and.i32.p0i32( i32* <ptr>, i32 <delta> )
6874 declare i64 @llvm.atomic.load.and.i64.p0i64( i64* <ptr>, i64 <delta> )
6878 declare i8 @llvm.atomic.load.or.i8.p0i8( i8* <ptr>, i8 <delta> )
6879 declare i16 @llvm.atomic.load.or.i16.p0i16( i16* <ptr>, i16 <delta> )
6880 declare i32 @llvm.atomic.load.or.i32.p0i32( i32* <ptr>, i32 <delta> )
6881 declare i64 @llvm.atomic.load.or.i64.p0i64( i64* <ptr>, i64 <delta> )
6885 declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* <ptr>, i8 <delta> )
6886 declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* <ptr>, i16 <delta> )
6887 declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* <ptr>, i32 <delta> )
6888 declare i64 @llvm.atomic.load.nand.i64.p0i32( i64* <ptr>, i64 <delta> )
6892 declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* <ptr>, i8 <delta> )
6893 declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* <ptr>, i16 <delta> )
6894 declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* <ptr>, i32 <delta> )
6895 declare i64 @llvm.atomic.load.xor.i64.p0i32( i64* <ptr>, i64 <delta> )
6899 <p>These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6900 the value stored in memory at <tt>ptr</tt>. It yields the original value
6901 at <tt>ptr</tt>.</p>
6904 <p>These intrinsics take two arguments, the first a pointer to an integer value
6905 and the second an integer value. The result is also an integer value. These
6906 integer types can have any bit width, but they must all have the same bit
6907 width. The targets may only lower integer representations they support.</p>
6910 <p>These intrinsics does a series of operations atomically. They first load the
6911 value stored at <tt>ptr</tt>. They then do the bitwise
6912 operation <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the
6913 original value stored at <tt>ptr</tt>.</p>
6917 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6918 %ptr = bitcast i8* %mallocP to i32*
6919 store i32 0x0F0F, %ptr
6920 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
6921 <i>; yields {i32}:result0 = 0x0F0F</i>
6922 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
6923 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
6924 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
6925 <i>; yields {i32}:result2 = 0xF0</i>
6926 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
6927 <i>; yields {i32}:result3 = FF</i>
6928 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6933 <!-- _______________________________________________________________________ -->
6934 <div class="doc_subsubsection">
6935 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6936 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6937 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6938 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6941 <div class="doc_text">
6944 <p>These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6945 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
6946 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6947 address spaces. Not all targets support all bit widths however.</p>
6950 declare i8 @llvm.atomic.load.max.i8.p0i8( i8* <ptr>, i8 <delta> )
6951 declare i16 @llvm.atomic.load.max.i16.p0i16( i16* <ptr>, i16 <delta> )
6952 declare i32 @llvm.atomic.load.max.i32.p0i32( i32* <ptr>, i32 <delta> )
6953 declare i64 @llvm.atomic.load.max.i64.p0i64( i64* <ptr>, i64 <delta> )
6957 declare i8 @llvm.atomic.load.min.i8.p0i8( i8* <ptr>, i8 <delta> )
6958 declare i16 @llvm.atomic.load.min.i16.p0i16( i16* <ptr>, i16 <delta> )
6959 declare i32 @llvm.atomic.load.min.i32..p0i32( i32* <ptr>, i32 <delta> )
6960 declare i64 @llvm.atomic.load.min.i64..p0i64( i64* <ptr>, i64 <delta> )
6964 declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* <ptr>, i8 <delta> )
6965 declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* <ptr>, i16 <delta> )
6966 declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* <ptr>, i32 <delta> )
6967 declare i64 @llvm.atomic.load.umax.i64.p0i64( i64* <ptr>, i64 <delta> )
6971 declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* <ptr>, i8 <delta> )
6972 declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* <ptr>, i16 <delta> )
6973 declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* <ptr>, i32 <delta> )
6974 declare i64 @llvm.atomic.load.umin.i64..p0i64( i64* <ptr>, i64 <delta> )
6978 <p>These intrinsics takes the signed or unsigned minimum or maximum of
6979 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6980 original value at <tt>ptr</tt>.</p>
6983 <p>These intrinsics take two arguments, the first a pointer to an integer value
6984 and the second an integer value. The result is also an integer value. These
6985 integer types can have any bit width, but they must all have the same bit
6986 width. The targets may only lower integer representations they support.</p>
6989 <p>These intrinsics does a series of operations atomically. They first load the
6990 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or
6991 max <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They
6992 yield the original value stored at <tt>ptr</tt>.</p>
6996 %mallocP = tail call i8* @malloc(i32 ptrtoint (i32* getelementptr (i32* null, i32 1) to i32))
6997 %ptr = bitcast i8* %mallocP to i32*
6999 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
7000 <i>; yields {i32}:result0 = 7</i>
7001 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
7002 <i>; yields {i32}:result1 = -2</i>
7003 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
7004 <i>; yields {i32}:result2 = 8</i>
7005 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
7006 <i>; yields {i32}:result3 = 8</i>
7007 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
7013 <!-- ======================================================================= -->
7014 <div class="doc_subsection">
7015 <a name="int_memorymarkers">Memory Use Markers</a>
7018 <div class="doc_text">
7020 <p>This class of intrinsics exists to information about the lifetime of memory
7021 objects and ranges where variables are immutable.</p>
7025 <!-- _______________________________________________________________________ -->
7026 <div class="doc_subsubsection">
7027 <a name="int_lifetime_start">'<tt>llvm.lifetime.start</tt>' Intrinsic</a>
7030 <div class="doc_text">
7034 declare void @llvm.lifetime.start(i64 <size>, i8* nocapture <ptr>)
7038 <p>The '<tt>llvm.lifetime.start</tt>' intrinsic specifies the start 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 before 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. A load from the pointer that
7050 precedes this intrinsic can be replaced with
7051 <tt>'<a href="#undefvalues">undef</a>'</tt>.</p>
7055 <!-- _______________________________________________________________________ -->
7056 <div class="doc_subsubsection">
7057 <a name="int_lifetime_end">'<tt>llvm.lifetime.end</tt>' Intrinsic</a>
7060 <div class="doc_text">
7064 declare void @llvm.lifetime.end(i64 <size>, i8* nocapture <ptr>)
7068 <p>The '<tt>llvm.lifetime.end</tt>' intrinsic specifies the end of a memory
7069 object's lifetime.</p>
7072 <p>The first argument is a constant integer representing the size of the
7073 object, or -1 if it is variable sized. The second argument is a pointer to
7077 <p>This intrinsic indicates that after this point in the code, the value of the
7078 memory pointed to by <tt>ptr</tt> is dead. This means that it is known to
7079 never be used and has an undefined value. Any stores into the memory object
7080 following this intrinsic may be removed as dead.
7084 <!-- _______________________________________________________________________ -->
7085 <div class="doc_subsubsection">
7086 <a name="int_invariant_start">'<tt>llvm.invariant.start</tt>' Intrinsic</a>
7089 <div class="doc_text">
7093 declare {}* @llvm.invariant.start(i64 <size>, i8* nocapture <ptr>) readonly
7097 <p>The '<tt>llvm.invariant.start</tt>' intrinsic specifies that the contents of
7098 a memory object will not change.</p>
7101 <p>The first argument is a constant integer representing the size of the
7102 object, or -1 if it is variable sized. The second argument is a pointer to
7106 <p>This intrinsic indicates that until an <tt>llvm.invariant.end</tt> that uses
7107 the return value, the referenced memory location is constant and
7112 <!-- _______________________________________________________________________ -->
7113 <div class="doc_subsubsection">
7114 <a name="int_invariant_end">'<tt>llvm.invariant.end</tt>' Intrinsic</a>
7117 <div class="doc_text">
7121 declare void @llvm.invariant.end({}* <start>, i64 <size>, i8* nocapture <ptr>)
7125 <p>The '<tt>llvm.invariant.end</tt>' intrinsic specifies that the contents of
7126 a memory object are mutable.</p>
7129 <p>The first argument is the matching <tt>llvm.invariant.start</tt> intrinsic.
7130 The second argument is a constant integer representing the size of the
7131 object, or -1 if it is variable sized and the third argument is a pointer
7135 <p>This intrinsic indicates that the memory is mutable again.</p>
7139 <!-- ======================================================================= -->
7140 <div class="doc_subsection">
7141 <a name="int_general">General Intrinsics</a>
7144 <div class="doc_text">
7146 <p>This class of intrinsics is designed to be generic and has no specific
7151 <!-- _______________________________________________________________________ -->
7152 <div class="doc_subsubsection">
7153 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
7156 <div class="doc_text">
7160 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
7164 <p>The '<tt>llvm.var.annotation</tt>' intrinsic.</p>
7167 <p>The first argument is a pointer to a value, the second is a pointer to a
7168 global string, the third is a pointer to a global string which is the source
7169 file name, and the last argument is the line number.</p>
7172 <p>This intrinsic allows annotation of local variables with arbitrary strings.
7173 This can be useful for special purpose optimizations that want to look for
7174 these annotations. These have no other defined use, they are ignored by code
7175 generation and optimization.</p>
7179 <!-- _______________________________________________________________________ -->
7180 <div class="doc_subsubsection">
7181 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
7184 <div class="doc_text">
7187 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
7188 any integer bit width.</p>
7191 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
7192 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
7193 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
7194 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
7195 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
7199 <p>The '<tt>llvm.annotation</tt>' intrinsic.</p>
7202 <p>The first argument is an integer value (result of some expression), the
7203 second is a pointer to a global string, the third is a pointer to a global
7204 string which is the source file name, and the last argument is the line
7205 number. It returns the value of the first argument.</p>
7208 <p>This intrinsic allows annotations to be put on arbitrary expressions with
7209 arbitrary strings. This can be useful for special purpose optimizations that
7210 want to look for these annotations. These have no other defined use, they
7211 are ignored by code generation and optimization.</p>
7215 <!-- _______________________________________________________________________ -->
7216 <div class="doc_subsubsection">
7217 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
7220 <div class="doc_text">
7224 declare void @llvm.trap()
7228 <p>The '<tt>llvm.trap</tt>' intrinsic.</p>
7234 <p>This intrinsics is lowered to the target dependent trap instruction. If the
7235 target does not have a trap instruction, this intrinsic will be lowered to
7236 the call of the <tt>abort()</tt> function.</p>
7240 <!-- _______________________________________________________________________ -->
7241 <div class="doc_subsubsection">
7242 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
7245 <div class="doc_text">
7249 declare void @llvm.stackprotector( i8* <guard>, i8** <slot> )
7253 <p>The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and
7254 stores it onto the stack at <tt>slot</tt>. The stack slot is adjusted to
7255 ensure that it is placed on the stack before local variables.</p>
7258 <p>The <tt>llvm.stackprotector</tt> intrinsic requires two pointer
7259 arguments. The first argument is the value loaded from the stack
7260 guard <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt>
7261 that has enough space to hold the value of the guard.</p>
7264 <p>This intrinsic causes the prologue/epilogue inserter to force the position of
7265 the <tt>AllocaInst</tt> stack slot to be before local variables on the
7266 stack. This is to ensure that if a local variable on the stack is
7267 overwritten, it will destroy the value of the guard. When the function exits,
7268 the guard on the stack is checked against the original guard. If they're
7269 different, then the program aborts by calling the <tt>__stack_chk_fail()</tt>
7274 <!-- *********************************************************************** -->
7277 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
7278 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
7279 <a href="http://validator.w3.org/check/referer"><img
7280 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
7282 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
7283 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
7284 Last modified: $Date$