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></li>
24 <li><a href="#callingconv">Calling Conventions</a></li>
25 <li><a href="#namedtypes">Named Types</a></li>
26 <li><a href="#globalvars">Global Variables</a></li>
27 <li><a href="#functionstructure">Functions</a></li>
28 <li><a href="#aliasstructure">Aliases</a></li>
29 <li><a href="#paramattrs">Parameter Attributes</a></li>
30 <li><a href="#fnattrs">Function Attributes</a></li>
31 <li><a href="#gc">Garbage Collector Names</a></li>
32 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
33 <li><a href="#datalayout">Data Layout</a></li>
36 <li><a href="#typesystem">Type System</a>
38 <li><a href="#t_classifications">Type Classifications</a></li>
39 <li><a href="#t_primitive">Primitive Types</a>
41 <li><a href="#t_floating">Floating Point Types</a></li>
42 <li><a href="#t_void">Void Type</a></li>
43 <li><a href="#t_label">Label Type</a></li>
46 <li><a href="#t_derived">Derived Types</a>
48 <li><a href="#t_integer">Integer Type</a></li>
49 <li><a href="#t_array">Array Type</a></li>
50 <li><a href="#t_function">Function Type</a></li>
51 <li><a href="#t_pointer">Pointer Type</a></li>
52 <li><a href="#t_struct">Structure Type</a></li>
53 <li><a href="#t_pstruct">Packed Structure Type</a></li>
54 <li><a href="#t_vector">Vector Type</a></li>
55 <li><a href="#t_opaque">Opaque Type</a></li>
60 <li><a href="#constants">Constants</a>
62 <li><a href="#simpleconstants">Simple Constants</a></li>
63 <li><a href="#aggregateconstants">Aggregate Constants</a></li>
64 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
65 <li><a href="#undefvalues">Undefined Values</a></li>
66 <li><a href="#constantexprs">Constant Expressions</a></li>
69 <li><a href="#othervalues">Other Values</a>
71 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
74 <li><a href="#instref">Instruction Reference</a>
76 <li><a href="#terminators">Terminator Instructions</a>
78 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
79 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
80 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
81 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
82 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
83 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
86 <li><a href="#binaryops">Binary Operations</a>
88 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
89 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
90 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
91 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
92 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
93 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
94 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
95 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
96 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
99 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
101 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
102 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
103 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
104 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
105 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
106 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
109 <li><a href="#vectorops">Vector Operations</a>
111 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
112 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
113 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
116 <li><a href="#aggregateops">Aggregate Operations</a>
118 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
119 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
122 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
124 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
125 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
126 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
127 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
128 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
129 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
132 <li><a href="#convertops">Conversion Operations</a>
134 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
135 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
136 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
137 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
138 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
139 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
140 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
141 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
142 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
143 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
144 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
145 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
148 <li><a href="#otherops">Other Operations</a>
150 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
151 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
152 <li><a href="#i_vicmp">'<tt>vicmp</tt>' Instruction</a></li>
153 <li><a href="#i_vfcmp">'<tt>vfcmp</tt>' Instruction</a></li>
154 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
155 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
156 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
157 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
162 <li><a href="#intrinsics">Intrinsic Functions</a>
164 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
166 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
167 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
168 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
171 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
173 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
174 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
175 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
178 <li><a href="#int_codegen">Code Generator Intrinsics</a>
180 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
181 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
182 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
183 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
184 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
185 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
186 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
189 <li><a href="#int_libc">Standard C Library Intrinsics</a>
191 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
192 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
193 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
194 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
195 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
196 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
197 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
198 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
201 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
203 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
204 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
205 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
206 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
207 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
208 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
211 <li><a href="#int_debugger">Debugger intrinsics</a></li>
212 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
213 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
215 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
218 <li><a href="#int_atomics">Atomic intrinsics</a>
220 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
221 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
222 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
223 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
224 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
225 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
226 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
227 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
228 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
229 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
230 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
231 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
232 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
235 <li><a href="#int_general">General intrinsics</a>
237 <li><a href="#int_var_annotation">
238 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
239 <li><a href="#int_annotation">
240 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
241 <li><a href="#int_trap">
242 '<tt>llvm.trap</tt>' Intrinsic</a></li>
243 <li><a href="#int_stackprotector">
244 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
251 <div class="doc_author">
252 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
253 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
256 <!-- *********************************************************************** -->
257 <div class="doc_section"> <a name="abstract">Abstract </a></div>
258 <!-- *********************************************************************** -->
260 <div class="doc_text">
261 <p>This document is a reference manual for the LLVM assembly language.
262 LLVM is a Static Single Assignment (SSA) based representation that provides
263 type safety, low-level operations, flexibility, and the capability of
264 representing 'all' high-level languages cleanly. It is the common code
265 representation used throughout all phases of the LLVM compilation
269 <!-- *********************************************************************** -->
270 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
271 <!-- *********************************************************************** -->
273 <div class="doc_text">
275 <p>The LLVM code representation is designed to be used in three
276 different forms: as an in-memory compiler IR, as an on-disk bitcode
277 representation (suitable for fast loading by a Just-In-Time compiler),
278 and as a human readable assembly language representation. This allows
279 LLVM to provide a powerful intermediate representation for efficient
280 compiler transformations and analysis, while providing a natural means
281 to debug and visualize the transformations. The three different forms
282 of LLVM are all equivalent. This document describes the human readable
283 representation and notation.</p>
285 <p>The LLVM representation aims to be light-weight and low-level
286 while being expressive, typed, and extensible at the same time. It
287 aims to be a "universal IR" of sorts, by being at a low enough level
288 that high-level ideas may be cleanly mapped to it (similar to how
289 microprocessors are "universal IR's", allowing many source languages to
290 be mapped to them). By providing type information, LLVM can be used as
291 the target of optimizations: for example, through pointer analysis, it
292 can be proven that a C automatic variable is never accessed outside of
293 the current function... allowing it to be promoted to a simple SSA
294 value instead of a memory location.</p>
298 <!-- _______________________________________________________________________ -->
299 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
301 <div class="doc_text">
303 <p>It is important to note that this document describes 'well formed'
304 LLVM assembly language. There is a difference between what the parser
305 accepts and what is considered 'well formed'. For example, the
306 following instruction is syntactically okay, but not well formed:</p>
308 <div class="doc_code">
310 %x = <a href="#i_add">add</a> i32 1, %x
314 <p>...because the definition of <tt>%x</tt> does not dominate all of
315 its uses. The LLVM infrastructure provides a verification pass that may
316 be used to verify that an LLVM module is well formed. This pass is
317 automatically run by the parser after parsing input assembly and by
318 the optimizer before it outputs bitcode. The violations pointed out
319 by the verifier pass indicate bugs in transformation passes or input to
323 <!-- Describe the typesetting conventions here. -->
325 <!-- *********************************************************************** -->
326 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
327 <!-- *********************************************************************** -->
329 <div class="doc_text">
331 <p>LLVM identifiers come in two basic types: global and local. Global
332 identifiers (functions, global variables) begin with the @ character. Local
333 identifiers (register names, types) begin with the % character. Additionally,
334 there are three different formats for identifiers, for different purposes:</p>
337 <li>Named values are represented as a string of characters with their prefix.
338 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
339 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
340 Identifiers which require other characters in their names can be surrounded
341 with quotes. Special characters may be escaped using "\xx" where xx is the
342 ASCII code for the character in hexadecimal. In this way, any character can
343 be used in a name value, even quotes themselves.
345 <li>Unnamed values are represented as an unsigned numeric value with their
346 prefix. For example, %12, @2, %44.</li>
348 <li>Constants, which are described in a <a href="#constants">section about
349 constants</a>, below.</li>
352 <p>LLVM requires that values start with a prefix for two reasons: Compilers
353 don't need to worry about name clashes with reserved words, and the set of
354 reserved words may be expanded in the future without penalty. Additionally,
355 unnamed identifiers allow a compiler to quickly come up with a temporary
356 variable without having to avoid symbol table conflicts.</p>
358 <p>Reserved words in LLVM are very similar to reserved words in other
359 languages. There are keywords for different opcodes
360 ('<tt><a href="#i_add">add</a></tt>',
361 '<tt><a href="#i_bitcast">bitcast</a></tt>',
362 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
363 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
364 and others. These reserved words cannot conflict with variable names, because
365 none of them start with a prefix character ('%' or '@').</p>
367 <p>Here is an example of LLVM code to multiply the integer variable
368 '<tt>%X</tt>' by 8:</p>
372 <div class="doc_code">
374 %result = <a href="#i_mul">mul</a> i32 %X, 8
378 <p>After strength reduction:</p>
380 <div class="doc_code">
382 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
386 <p>And the hard way:</p>
388 <div class="doc_code">
390 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
391 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
392 %result = <a href="#i_add">add</a> i32 %1, %1
396 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
397 important lexical features of LLVM:</p>
401 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
404 <li>Unnamed temporaries are created when the result of a computation is not
405 assigned to a named value.</li>
407 <li>Unnamed temporaries are numbered sequentially</li>
411 <p>...and it also shows a convention that we follow in this document. When
412 demonstrating instructions, we will follow an instruction with a comment that
413 defines the type and name of value produced. Comments are shown in italic
418 <!-- *********************************************************************** -->
419 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
420 <!-- *********************************************************************** -->
422 <!-- ======================================================================= -->
423 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
426 <div class="doc_text">
428 <p>LLVM programs are composed of "Module"s, each of which is a
429 translation unit of the input programs. Each module consists of
430 functions, global variables, and symbol table entries. Modules may be
431 combined together with the LLVM linker, which merges function (and
432 global variable) definitions, resolves forward declarations, and merges
433 symbol table entries. Here is an example of the "hello world" module:</p>
435 <div class="doc_code">
436 <pre><i>; Declare the string constant as a global constant...</i>
437 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
438 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
440 <i>; External declaration of the puts function</i>
441 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
443 <i>; Definition of main function</i>
444 define i32 @main() { <i>; i32()* </i>
445 <i>; Convert [13 x i8]* to i8 *...</i>
447 href="#i_getelementptr">getelementptr</a> [13 x i8]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
449 <i>; Call puts function to write out the string to stdout...</i>
451 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
453 href="#i_ret">ret</a> i32 0<br>}<br>
457 <p>This example is made up of a <a href="#globalvars">global variable</a>
458 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
459 function, and a <a href="#functionstructure">function definition</a>
460 for "<tt>main</tt>".</p>
462 <p>In general, a module is made up of a list of global values,
463 where both functions and global variables are global values. Global values are
464 represented by a pointer to a memory location (in this case, a pointer to an
465 array of char, and a pointer to a function), and have one of the following <a
466 href="#linkage">linkage types</a>.</p>
470 <!-- ======================================================================= -->
471 <div class="doc_subsection">
472 <a name="linkage">Linkage Types</a>
475 <div class="doc_text">
478 All Global Variables and Functions have one of the following types of linkage:
483 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
485 <dd>Global values with internal linkage are only directly accessible by
486 objects in the current module. In particular, linking code into a module with
487 an internal global value may cause the internal to be renamed as necessary to
488 avoid collisions. Because the symbol is internal to the module, all
489 references can be updated. This corresponds to the notion of the
490 '<tt>static</tt>' keyword in C.
493 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
495 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
496 the same name when linkage occurs. This is typically used to implement
497 inline functions, templates, or other code which must be generated in each
498 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
499 allowed to be discarded.
502 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
504 <dd>"<tt>common</tt>" linkage is exactly the same as <tt>linkonce</tt>
505 linkage, except that unreferenced <tt>common</tt> globals may not be
506 discarded. This is used for globals that may be emitted in multiple
507 translation units, but that are not guaranteed to be emitted into every
508 translation unit that uses them. One example of this is tentative
509 definitions in C, such as "<tt>int X;</tt>" at global scope.
512 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
514 <dd>"<tt>weak</tt>" linkage is the same as <tt>common</tt> linkage, except
515 that some targets may choose to emit different assembly sequences for them
516 for target-dependent reasons. This is used for globals that are declared
517 "weak" in C source code.
520 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
522 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
523 pointer to array type. When two global variables with appending linkage are
524 linked together, the two global arrays are appended together. This is the
525 LLVM, typesafe, equivalent of having the system linker append together
526 "sections" with identical names when .o files are linked.
529 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
530 <dd>The semantics of this linkage follow the ELF object file model: the
531 symbol is weak until linked, if not linked, the symbol becomes null instead
532 of being an undefined reference.
535 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
537 <dd>If none of the above identifiers are used, the global is externally
538 visible, meaning that it participates in linkage and can be used to resolve
539 external symbol references.
544 The next two types of linkage are targeted for Microsoft Windows platform
545 only. They are designed to support importing (exporting) symbols from (to)
546 DLLs (Dynamic Link Libraries).
550 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
552 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
553 or variable via a global pointer to a pointer that is set up by the DLL
554 exporting the symbol. On Microsoft Windows targets, the pointer name is
555 formed by combining <code>_imp__</code> and the function or variable name.
558 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
560 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
561 pointer to a pointer in a DLL, so that it can be referenced with the
562 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
563 name is formed by combining <code>_imp__</code> and the function or variable
569 <p>For example, since the "<tt>.LC0</tt>"
570 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
571 variable and was linked with this one, one of the two would be renamed,
572 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
573 external (i.e., lacking any linkage declarations), they are accessible
574 outside of the current module.</p>
575 <p>It is illegal for a function <i>declaration</i>
576 to have any linkage type other than "externally visible", <tt>dllimport</tt>,
577 or <tt>extern_weak</tt>.</p>
578 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
582 <!-- ======================================================================= -->
583 <div class="doc_subsection">
584 <a name="callingconv">Calling Conventions</a>
587 <div class="doc_text">
589 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
590 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
591 specified for the call. The calling convention of any pair of dynamic
592 caller/callee must match, or the behavior of the program is undefined. The
593 following calling conventions are supported by LLVM, and more may be added in
597 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
599 <dd>This calling convention (the default if no other calling convention is
600 specified) matches the target C calling conventions. This calling convention
601 supports varargs function calls and tolerates some mismatch in the declared
602 prototype and implemented declaration of the function (as does normal C).
605 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
607 <dd>This calling convention attempts to make calls as fast as possible
608 (e.g. by passing things in registers). This calling convention allows the
609 target to use whatever tricks it wants to produce fast code for the target,
610 without having to conform to an externally specified ABI (Application Binary
611 Interface). Implementations of this convention should allow arbitrary
612 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> to be
613 supported. This calling convention does not support varargs and requires the
614 prototype of all callees to exactly match the prototype of the function
618 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
620 <dd>This calling convention attempts to make code in the caller as efficient
621 as possible under the assumption that the call is not commonly executed. As
622 such, these calls often preserve all registers so that the call does not break
623 any live ranges in the caller side. This calling convention does not support
624 varargs and requires the prototype of all callees to exactly match the
625 prototype of the function definition.
628 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
630 <dd>Any calling convention may be specified by number, allowing
631 target-specific calling conventions to be used. Target specific calling
632 conventions start at 64.
636 <p>More calling conventions can be added/defined on an as-needed basis, to
637 support pascal conventions or any other well-known target-independent
642 <!-- ======================================================================= -->
643 <div class="doc_subsection">
644 <a name="visibility">Visibility Styles</a>
647 <div class="doc_text">
650 All Global Variables and Functions have one of the following visibility styles:
654 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
656 <dd>On targets that use the ELF object file format, default visibility means
657 that the declaration is visible to other
658 modules and, in shared libraries, means that the declared entity may be
659 overridden. On Darwin, default visibility means that the declaration is
660 visible to other modules. Default visibility corresponds to "external
661 linkage" in the language.
664 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
666 <dd>Two declarations of an object with hidden visibility refer to the same
667 object if they are in the same shared object. Usually, hidden visibility
668 indicates that the symbol will not be placed into the dynamic symbol table,
669 so no other module (executable or shared library) can reference it
673 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
675 <dd>On ELF, protected visibility indicates that the symbol will be placed in
676 the dynamic symbol table, but that references within the defining module will
677 bind to the local symbol. That is, the symbol cannot be overridden by another
684 <!-- ======================================================================= -->
685 <div class="doc_subsection">
686 <a name="namedtypes">Named Types</a>
689 <div class="doc_text">
691 <p>LLVM IR allows you to specify name aliases for certain types. This can make
692 it easier to read the IR and make the IR more condensed (particularly when
693 recursive types are involved). An example of a name specification is:
696 <div class="doc_code">
698 %mytype = type { %mytype*, i32 }
702 <p>You may give a name to any <a href="#typesystem">type</a> except "<a
703 href="t_void">void</a>". Type name aliases may be used anywhere a type is
704 expected with the syntax "%mytype".</p>
706 <p>Note that type names are aliases for the structural type that they indicate,
707 and that you can therefore specify multiple names for the same type. This often
708 leads to confusing behavior when dumping out a .ll file. Since LLVM IR uses
709 structural typing, the name is not part of the type. When printing out LLVM IR,
710 the printer will pick <em>one name</em> to render all types of a particular
711 shape. This means that if you have code where two different source types end up
712 having the same LLVM type, that the dumper will sometimes print the "wrong" or
713 unexpected type. This is an important design point and isn't going to
719 <!-- ======================================================================= -->
720 <div class="doc_subsection">
721 <a name="globalvars">Global Variables</a>
724 <div class="doc_text">
726 <p>Global variables define regions of memory allocated at compilation time
727 instead of run-time. Global variables may optionally be initialized, may have
728 an explicit section to be placed in, and may have an optional explicit alignment
729 specified. A variable may be defined as "thread_local", which means that it
730 will not be shared by threads (each thread will have a separated copy of the
731 variable). A variable may be defined as a global "constant," which indicates
732 that the contents of the variable will <b>never</b> be modified (enabling better
733 optimization, allowing the global data to be placed in the read-only section of
734 an executable, etc). Note that variables that need runtime initialization
735 cannot be marked "constant" as there is a store to the variable.</p>
738 LLVM explicitly allows <em>declarations</em> of global variables to be marked
739 constant, even if the final definition of the global is not. This capability
740 can be used to enable slightly better optimization of the program, but requires
741 the language definition to guarantee that optimizations based on the
742 'constantness' are valid for the translation units that do not include the
746 <p>As SSA values, global variables define pointer values that are in
747 scope (i.e. they dominate) all basic blocks in the program. Global
748 variables always define a pointer to their "content" type because they
749 describe a region of memory, and all memory objects in LLVM are
750 accessed through pointers.</p>
752 <p>A global variable may be declared to reside in a target-specifc numbered
753 address space. For targets that support them, address spaces may affect how
754 optimizations are performed and/or what target instructions are used to access
755 the variable. The default address space is zero. The address space qualifier
756 must precede any other attributes.</p>
758 <p>LLVM allows an explicit section to be specified for globals. If the target
759 supports it, it will emit globals to the section specified.</p>
761 <p>An explicit alignment may be specified for a global. If not present, or if
762 the alignment is set to zero, the alignment of the global is set by the target
763 to whatever it feels convenient. If an explicit alignment is specified, the
764 global is forced to have at least that much alignment. All alignments must be
767 <p>For example, the following defines a global in a numbered address space with
768 an initializer, section, and alignment:</p>
770 <div class="doc_code">
772 @G = addrspace(5) constant float 1.0, section "foo", align 4
779 <!-- ======================================================================= -->
780 <div class="doc_subsection">
781 <a name="functionstructure">Functions</a>
784 <div class="doc_text">
786 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
787 an optional <a href="#linkage">linkage type</a>, an optional
788 <a href="#visibility">visibility style</a>, an optional
789 <a href="#callingconv">calling convention</a>, a return type, an optional
790 <a href="#paramattrs">parameter attribute</a> for the return type, a function
791 name, a (possibly empty) argument list (each with optional
792 <a href="#paramattrs">parameter attributes</a>), optional
793 <a href="#fnattrs">function attributes</a>, an optional section,
794 an optional alignment, an optional <a href="#gc">garbage collector name</a>,
795 an opening curly brace, a list of basic blocks, and a closing curly brace.
797 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
798 optional <a href="#linkage">linkage type</a>, an optional
799 <a href="#visibility">visibility style</a>, an optional
800 <a href="#callingconv">calling convention</a>, a return type, an optional
801 <a href="#paramattrs">parameter attribute</a> for the return type, a function
802 name, a possibly empty list of arguments, an optional alignment, and an optional
803 <a href="#gc">garbage collector name</a>.</p>
805 <p>A function definition contains a list of basic blocks, forming the CFG
806 (Control Flow Graph) for
807 the function. Each basic block may optionally start with a label (giving the
808 basic block a symbol table entry), contains a list of instructions, and ends
809 with a <a href="#terminators">terminator</a> instruction (such as a branch or
810 function return).</p>
812 <p>The first basic block in a function is special in two ways: it is immediately
813 executed on entrance to the function, and it is not allowed to have predecessor
814 basic blocks (i.e. there can not be any branches to the entry block of a
815 function). Because the block can have no predecessors, it also cannot have any
816 <a href="#i_phi">PHI nodes</a>.</p>
818 <p>LLVM allows an explicit section to be specified for functions. If the target
819 supports it, it will emit functions to the section specified.</p>
821 <p>An explicit alignment may be specified for a function. If not present, or if
822 the alignment is set to zero, the alignment of the function is set by the target
823 to whatever it feels convenient. If an explicit alignment is specified, the
824 function is forced to have at least that much alignment. All alignments must be
829 <div class="doc_code">
831 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
832 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
833 <ResultType> @<FunctionName> ([argument list])
834 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
835 [<a href="#gc">gc</a>] { ... }
842 <!-- ======================================================================= -->
843 <div class="doc_subsection">
844 <a name="aliasstructure">Aliases</a>
846 <div class="doc_text">
847 <p>Aliases act as "second name" for the aliasee value (which can be either
848 function, global variable, another alias or bitcast of global value). Aliases
849 may have an optional <a href="#linkage">linkage type</a>, and an
850 optional <a href="#visibility">visibility style</a>.</p>
854 <div class="doc_code">
856 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
864 <!-- ======================================================================= -->
865 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
866 <div class="doc_text">
867 <p>The return type and each parameter of a function type may have a set of
868 <i>parameter attributes</i> associated with them. Parameter attributes are
869 used to communicate additional information about the result or parameters of
870 a function. Parameter attributes are considered to be part of the function,
871 not of the function type, so functions with different parameter attributes
872 can have the same function type.</p>
874 <p>Parameter attributes are simple keywords that follow the type specified. If
875 multiple parameter attributes are needed, they are space separated. For
878 <div class="doc_code">
880 declare i32 @printf(i8* noalias , ...)
881 declare i32 @atoi(i8 zeroext)
882 declare signext i8 @returns_signed_char()
886 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
887 <tt>readonly</tt>) come immediately after the argument list.</p>
889 <p>Currently, only the following parameter attributes are defined:</p>
891 <dt><tt>zeroext</tt></dt>
892 <dd>This indicates to the code generator that the parameter or return value
893 should be zero-extended to a 32-bit value by the caller (for a parameter)
894 or the callee (for a return value).</dd>
896 <dt><tt>signext</tt></dt>
897 <dd>This indicates to the code generator that the parameter or return value
898 should be sign-extended to a 32-bit value by the caller (for a parameter)
899 or the callee (for a return value).</dd>
901 <dt><tt>inreg</tt></dt>
902 <dd>This indicates that this parameter or return value should be treated
903 in a special target-dependent fashion during while emitting code for a
904 function call or return (usually, by putting it in a register as opposed
905 to memory, though some targets use it to distinguish between two different
906 kinds of registers). Use of this attribute is target-specific.</dd>
908 <dt><tt><a name="byval">byval</a></tt></dt>
909 <dd>This indicates that the pointer parameter should really be passed by
910 value to the function. The attribute implies that a hidden copy of the
911 pointee is made between the caller and the callee, so the callee is unable
912 to modify the value in the callee. This attribute is only valid on LLVM
913 pointer arguments. It is generally used to pass structs and arrays by
914 value, but is also valid on pointers to scalars. The copy is considered to
915 belong to the caller not the callee (for example,
916 <tt><a href="#readonly">readonly</a></tt> functions should not write to
917 <tt>byval</tt> parameters). This is not a valid attribute for return
920 <dt><tt>sret</tt></dt>
921 <dd>This indicates that the pointer parameter specifies the address of a
922 structure that is the return value of the function in the source program.
923 This pointer must be guaranteed by the caller to be valid: loads and stores
924 to the structure may be assumed by the callee to not to trap. This may only
925 be applied to the first parameter. This is not a valid attribute for
928 <dt><tt>noalias</tt></dt>
929 <dd>This indicates that the pointer does not alias any global or any other
930 parameter. The caller is responsible for ensuring that this is the
931 case. On a function return value, <tt>noalias</tt> additionally indicates
932 that the pointer does not alias any other pointers visible to the
933 caller. For further details, please see the discussion of the NoAlias
935 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
938 <dt><tt>nocapture</tt></dt>
939 <dd>This indicates that the callee does not make any copies of the pointer
940 that outlive the callee itself. This is not a valid attribute for return
943 <dt><tt>nest</tt></dt>
944 <dd>This indicates that the pointer parameter can be excised using the
945 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
946 attribute for return values.</dd>
951 <!-- ======================================================================= -->
952 <div class="doc_subsection">
953 <a name="gc">Garbage Collector Names</a>
956 <div class="doc_text">
957 <p>Each function may specify a garbage collector name, which is simply a
960 <div class="doc_code"><pre
961 >define void @f() gc "name" { ...</pre></div>
963 <p>The compiler declares the supported values of <i>name</i>. Specifying a
964 collector which will cause the compiler to alter its output in order to support
965 the named garbage collection algorithm.</p>
968 <!-- ======================================================================= -->
969 <div class="doc_subsection">
970 <a name="fnattrs">Function Attributes</a>
973 <div class="doc_text">
975 <p>Function attributes are set to communicate additional information about
976 a function. Function attributes are considered to be part of the function,
977 not of the function type, so functions with different parameter attributes
978 can have the same function type.</p>
980 <p>Function attributes are simple keywords that follow the type specified. If
981 multiple attributes are needed, they are space separated. For
984 <div class="doc_code">
986 define void @f() noinline { ... }
987 define void @f() alwaysinline { ... }
988 define void @f() alwaysinline optsize { ... }
989 define void @f() optsize
994 <dt><tt>alwaysinline</tt></dt>
995 <dd>This attribute indicates that the inliner should attempt to inline this
996 function into callers whenever possible, ignoring any active inlining size
997 threshold for this caller.</dd>
999 <dt><tt>noinline</tt></dt>
1000 <dd>This attribute indicates that the inliner should never inline this function
1001 in any situation. This attribute may not be used together with the
1002 <tt>alwaysinline</tt> attribute.</dd>
1004 <dt><tt>optsize</tt></dt>
1005 <dd>This attribute suggests that optimization passes and code generator passes
1006 make choices that keep the code size of this function low, and otherwise do
1007 optimizations specifically to reduce code size.</dd>
1009 <dt><tt>noreturn</tt></dt>
1010 <dd>This function attribute indicates that the function never returns normally.
1011 This produces undefined behavior at runtime if the function ever does
1012 dynamically return.</dd>
1014 <dt><tt>nounwind</tt></dt>
1015 <dd>This function attribute indicates that the function never returns with an
1016 unwind or exceptional control flow. If the function does unwind, its runtime
1017 behavior is undefined.</dd>
1019 <dt><tt>readnone</tt></dt>
1020 <dd>This attribute indicates that the function computes its result (or the
1021 exception it throws) based strictly on its arguments, without dereferencing any
1022 pointer arguments or otherwise accessing any mutable state (e.g. memory, control
1023 registers, etc) visible to caller functions. It does not write through any
1024 pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments) and
1025 never changes any state visible to callers.</dd>
1027 <dt><tt><a name="readonly">readonly</a></tt></dt>
1028 <dd>This attribute indicates that the function does not write through any
1029 pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments)
1030 or otherwise modify any state (e.g. memory, control registers, etc) visible to
1031 caller functions. It may dereference pointer arguments and read state that may
1032 be set in the caller. A readonly function always returns the same value (or
1033 throws the same exception) when called with the same set of arguments and global
1036 <dt><tt><a name="ssp">ssp</a></tt></dt>
1037 <dd>This attribute indicates that the function should emit a stack smashing
1038 protector. It is in the form of a "canary"—a random value placed on the
1039 stack before the local variables that's checked upon return from the function to
1040 see if it has been overwritten. A heuristic is used to determine if a function
1041 needs stack protectors or not.
1043 <p>If a function that has an <tt>ssp</tt> attribute is inlined into a function
1044 that doesn't have an <tt>ssp</tt> attribute, then the resulting function will
1045 have an <tt>ssp</tt> attribute.</p></dd>
1047 <dt><tt>sspreq</tt></dt>
1048 <dd>This attribute indicates that the function should <em>always</em> emit a
1049 stack smashing protector. This overrides the <tt><a href="#ssp">ssp</a></tt>
1052 <p>If a function that has an <tt>sspreq</tt> attribute is inlined into a
1053 function that doesn't have an <tt>sspreq</tt> attribute or which has
1054 an <tt>ssp</tt> attribute, then the resulting function will have
1055 an <tt>sspreq</tt> attribute.</p></dd>
1060 <!-- ======================================================================= -->
1061 <div class="doc_subsection">
1062 <a name="moduleasm">Module-Level Inline Assembly</a>
1065 <div class="doc_text">
1067 Modules may contain "module-level inline asm" blocks, which corresponds to the
1068 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
1069 LLVM and treated as a single unit, but may be separated in the .ll file if
1070 desired. The syntax is very simple:
1073 <div class="doc_code">
1075 module asm "inline asm code goes here"
1076 module asm "more can go here"
1080 <p>The strings can contain any character by escaping non-printable characters.
1081 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1086 The inline asm code is simply printed to the machine code .s file when
1087 assembly code is generated.
1091 <!-- ======================================================================= -->
1092 <div class="doc_subsection">
1093 <a name="datalayout">Data Layout</a>
1096 <div class="doc_text">
1097 <p>A module may specify a target specific data layout string that specifies how
1098 data is to be laid out in memory. The syntax for the data layout is simply:</p>
1099 <pre> target datalayout = "<i>layout specification</i>"</pre>
1100 <p>The <i>layout specification</i> consists of a list of specifications
1101 separated by the minus sign character ('-'). Each specification starts with a
1102 letter and may include other information after the letter to define some
1103 aspect of the data layout. The specifications accepted are as follows: </p>
1106 <dd>Specifies that the target lays out data in big-endian form. That is, the
1107 bits with the most significance have the lowest address location.</dd>
1109 <dd>Specifies that the target lays out data in little-endian form. That is,
1110 the bits with the least significance have the lowest address location.</dd>
1111 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1112 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1113 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1114 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1116 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1117 <dd>This specifies the alignment for an integer type of a given bit
1118 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1119 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1120 <dd>This specifies the alignment for a vector type of a given bit
1122 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1123 <dd>This specifies the alignment for a floating point type of a given bit
1124 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1126 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1127 <dd>This specifies the alignment for an aggregate type of a given bit
1130 <p>When constructing the data layout for a given target, LLVM starts with a
1131 default set of specifications which are then (possibly) overriden by the
1132 specifications in the <tt>datalayout</tt> keyword. The default specifications
1133 are given in this list:</p>
1135 <li><tt>E</tt> - big endian</li>
1136 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1137 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1138 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1139 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1140 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1141 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1142 alignment of 64-bits</li>
1143 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1144 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1145 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1146 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1147 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1149 <p>When LLVM is determining the alignment for a given type, it uses the
1150 following rules:</p>
1152 <li>If the type sought is an exact match for one of the specifications, that
1153 specification is used.</li>
1154 <li>If no match is found, and the type sought is an integer type, then the
1155 smallest integer type that is larger than the bitwidth of the sought type is
1156 used. If none of the specifications are larger than the bitwidth then the the
1157 largest integer type is used. For example, given the default specifications
1158 above, the i7 type will use the alignment of i8 (next largest) while both
1159 i65 and i256 will use the alignment of i64 (largest specified).</li>
1160 <li>If no match is found, and the type sought is a vector type, then the
1161 largest vector type that is smaller than the sought vector type will be used
1162 as a fall back. This happens because <128 x double> can be implemented
1163 in terms of 64 <2 x double>, for example.</li>
1167 <!-- *********************************************************************** -->
1168 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1169 <!-- *********************************************************************** -->
1171 <div class="doc_text">
1173 <p>The LLVM type system is one of the most important features of the
1174 intermediate representation. Being typed enables a number of
1175 optimizations to be performed on the intermediate representation directly,
1176 without having to do
1177 extra analyses on the side before the transformation. A strong type
1178 system makes it easier to read the generated code and enables novel
1179 analyses and transformations that are not feasible to perform on normal
1180 three address code representations.</p>
1184 <!-- ======================================================================= -->
1185 <div class="doc_subsection"> <a name="t_classifications">Type
1186 Classifications</a> </div>
1187 <div class="doc_text">
1188 <p>The types fall into a few useful
1189 classifications:</p>
1191 <table border="1" cellspacing="0" cellpadding="4">
1193 <tr><th>Classification</th><th>Types</th></tr>
1195 <td><a href="#t_integer">integer</a></td>
1196 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1199 <td><a href="#t_floating">floating point</a></td>
1200 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1203 <td><a name="t_firstclass">first class</a></td>
1204 <td><a href="#t_integer">integer</a>,
1205 <a href="#t_floating">floating point</a>,
1206 <a href="#t_pointer">pointer</a>,
1207 <a href="#t_vector">vector</a>,
1208 <a href="#t_struct">structure</a>,
1209 <a href="#t_array">array</a>,
1210 <a href="#t_label">label</a>.
1214 <td><a href="#t_primitive">primitive</a></td>
1215 <td><a href="#t_label">label</a>,
1216 <a href="#t_void">void</a>,
1217 <a href="#t_floating">floating point</a>.</td>
1220 <td><a href="#t_derived">derived</a></td>
1221 <td><a href="#t_integer">integer</a>,
1222 <a href="#t_array">array</a>,
1223 <a href="#t_function">function</a>,
1224 <a href="#t_pointer">pointer</a>,
1225 <a href="#t_struct">structure</a>,
1226 <a href="#t_pstruct">packed structure</a>,
1227 <a href="#t_vector">vector</a>,
1228 <a href="#t_opaque">opaque</a>.
1234 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
1235 most important. Values of these types are the only ones which can be
1236 produced by instructions, passed as arguments, or used as operands to
1240 <!-- ======================================================================= -->
1241 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1243 <div class="doc_text">
1244 <p>The primitive types are the fundamental building blocks of the LLVM
1249 <!-- _______________________________________________________________________ -->
1250 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1252 <div class="doc_text">
1255 <tr><th>Type</th><th>Description</th></tr>
1256 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1257 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1258 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1259 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1260 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1265 <!-- _______________________________________________________________________ -->
1266 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1268 <div class="doc_text">
1270 <p>The void type does not represent any value and has no size.</p>
1279 <!-- _______________________________________________________________________ -->
1280 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1282 <div class="doc_text">
1284 <p>The label type represents code labels.</p>
1294 <!-- ======================================================================= -->
1295 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1297 <div class="doc_text">
1299 <p>The real power in LLVM comes from the derived types in the system.
1300 This is what allows a programmer to represent arrays, functions,
1301 pointers, and other useful types. Note that these derived types may be
1302 recursive: For example, it is possible to have a two dimensional array.</p>
1306 <!-- _______________________________________________________________________ -->
1307 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1309 <div class="doc_text">
1312 <p>The integer type is a very simple derived type that simply specifies an
1313 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1314 2^23-1 (about 8 million) can be specified.</p>
1322 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1326 <table class="layout">
1329 <td><tt>i1</tt></td>
1330 <td>a single-bit integer.</td>
1332 <td><tt>i32</tt></td>
1333 <td>a 32-bit integer.</td>
1335 <td><tt>i1942652</tt></td>
1336 <td>a really big integer of over 1 million bits.</td>
1342 <!-- _______________________________________________________________________ -->
1343 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1345 <div class="doc_text">
1349 <p>The array type is a very simple derived type that arranges elements
1350 sequentially in memory. The array type requires a size (number of
1351 elements) and an underlying data type.</p>
1356 [<# elements> x <elementtype>]
1359 <p>The number of elements is a constant integer value; elementtype may
1360 be any type with a size.</p>
1363 <table class="layout">
1365 <td class="left"><tt>[40 x i32]</tt></td>
1366 <td class="left">Array of 40 32-bit integer values.</td>
1369 <td class="left"><tt>[41 x i32]</tt></td>
1370 <td class="left">Array of 41 32-bit integer values.</td>
1373 <td class="left"><tt>[4 x i8]</tt></td>
1374 <td class="left">Array of 4 8-bit integer values.</td>
1377 <p>Here are some examples of multidimensional arrays:</p>
1378 <table class="layout">
1380 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1381 <td class="left">3x4 array of 32-bit integer values.</td>
1384 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1385 <td class="left">12x10 array of single precision floating point values.</td>
1388 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1389 <td class="left">2x3x4 array of 16-bit integer values.</td>
1393 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1394 length array. Normally, accesses past the end of an array are undefined in
1395 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1396 As a special case, however, zero length arrays are recognized to be variable
1397 length. This allows implementation of 'pascal style arrays' with the LLVM
1398 type "{ i32, [0 x float]}", for example.</p>
1402 <!-- _______________________________________________________________________ -->
1403 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1404 <div class="doc_text">
1408 <p>The function type can be thought of as a function signature. It
1409 consists of a return type and a list of formal parameter types. The
1410 return type of a function type is a scalar type, a void type, or a struct type.
1411 If the return type is a struct type then all struct elements must be of first
1412 class types, and the struct must have at least one element.</p>
1417 <returntype list> (<parameter list>)
1420 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1421 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1422 which indicates that the function takes a variable number of arguments.
1423 Variable argument functions can access their arguments with the <a
1424 href="#int_varargs">variable argument handling intrinsic</a> functions.
1425 '<tt><returntype list></tt>' is a comma-separated list of
1426 <a href="#t_firstclass">first class</a> type specifiers.</p>
1429 <table class="layout">
1431 <td class="left"><tt>i32 (i32)</tt></td>
1432 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1434 </tr><tr class="layout">
1435 <td class="left"><tt>float (i16 signext, i32 *) *
1437 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1438 an <tt>i16</tt> that should be sign extended and a
1439 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1442 </tr><tr class="layout">
1443 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1444 <td class="left">A vararg function that takes at least one
1445 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1446 which returns an integer. This is the signature for <tt>printf</tt> in
1449 </tr><tr class="layout">
1450 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1451 <td class="left">A function taking an <tt>i32</tt>, returning two
1452 <tt>i32</tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
1458 <!-- _______________________________________________________________________ -->
1459 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1460 <div class="doc_text">
1462 <p>The structure type is used to represent a collection of data members
1463 together in memory. The packing of the field types is defined to match
1464 the ABI of the underlying processor. The elements of a structure may
1465 be any type that has a size.</p>
1466 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1467 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1468 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1471 <pre> { <type list> }<br></pre>
1473 <table class="layout">
1475 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1476 <td class="left">A triple of three <tt>i32</tt> values</td>
1477 </tr><tr class="layout">
1478 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1479 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1480 second element is a <a href="#t_pointer">pointer</a> to a
1481 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1482 an <tt>i32</tt>.</td>
1487 <!-- _______________________________________________________________________ -->
1488 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1490 <div class="doc_text">
1492 <p>The packed structure type is used to represent a collection of data members
1493 together in memory. There is no padding between fields. Further, the alignment
1494 of a packed structure is 1 byte. The elements of a packed structure may
1495 be any type that has a size.</p>
1496 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1497 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1498 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1501 <pre> < { <type list> } > <br></pre>
1503 <table class="layout">
1505 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1506 <td class="left">A triple of three <tt>i32</tt> values</td>
1507 </tr><tr class="layout">
1509 <tt>< { float, i32 (i32)* } ></tt></td>
1510 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1511 second element is a <a href="#t_pointer">pointer</a> to a
1512 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1513 an <tt>i32</tt>.</td>
1518 <!-- _______________________________________________________________________ -->
1519 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1520 <div class="doc_text">
1522 <p>As in many languages, the pointer type represents a pointer or
1523 reference to another object, which must live in memory. Pointer types may have
1524 an optional address space attribute defining the target-specific numbered
1525 address space where the pointed-to object resides. The default address space is
1528 <pre> <type> *<br></pre>
1530 <table class="layout">
1532 <td class="left"><tt>[4 x i32]*</tt></td>
1533 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1534 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1537 <td class="left"><tt>i32 (i32 *) *</tt></td>
1538 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1539 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1543 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1544 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1545 that resides in address space #5.</td>
1550 <!-- _______________________________________________________________________ -->
1551 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1552 <div class="doc_text">
1556 <p>A vector type is a simple derived type that represents a vector
1557 of elements. Vector types are used when multiple primitive data
1558 are operated in parallel using a single instruction (SIMD).
1559 A vector type requires a size (number of
1560 elements) and an underlying primitive data type. Vectors must have a power
1561 of two length (1, 2, 4, 8, 16 ...). Vector types are
1562 considered <a href="#t_firstclass">first class</a>.</p>
1567 < <# elements> x <elementtype> >
1570 <p>The number of elements is a constant integer value; elementtype may
1571 be any integer or floating point type.</p>
1575 <table class="layout">
1577 <td class="left"><tt><4 x i32></tt></td>
1578 <td class="left">Vector of 4 32-bit integer values.</td>
1581 <td class="left"><tt><8 x float></tt></td>
1582 <td class="left">Vector of 8 32-bit floating-point values.</td>
1585 <td class="left"><tt><2 x i64></tt></td>
1586 <td class="left">Vector of 2 64-bit integer values.</td>
1591 <!-- _______________________________________________________________________ -->
1592 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1593 <div class="doc_text">
1597 <p>Opaque types are used to represent unknown types in the system. This
1598 corresponds (for example) to the C notion of a forward declared structure type.
1599 In LLVM, opaque types can eventually be resolved to any type (not just a
1600 structure type).</p>
1610 <table class="layout">
1612 <td class="left"><tt>opaque</tt></td>
1613 <td class="left">An opaque type.</td>
1619 <!-- *********************************************************************** -->
1620 <div class="doc_section"> <a name="constants">Constants</a> </div>
1621 <!-- *********************************************************************** -->
1623 <div class="doc_text">
1625 <p>LLVM has several different basic types of constants. This section describes
1626 them all and their syntax.</p>
1630 <!-- ======================================================================= -->
1631 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1633 <div class="doc_text">
1636 <dt><b>Boolean constants</b></dt>
1638 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1639 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1642 <dt><b>Integer constants</b></dt>
1644 <dd>Standard integers (such as '4') are constants of the <a
1645 href="#t_integer">integer</a> type. Negative numbers may be used with
1649 <dt><b>Floating point constants</b></dt>
1651 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1652 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1653 notation (see below). The assembler requires the exact decimal value of
1654 a floating-point constant. For example, the assembler accepts 1.25 but
1655 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1656 constants must have a <a href="#t_floating">floating point</a> type. </dd>
1658 <dt><b>Null pointer constants</b></dt>
1660 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1661 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1665 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1666 of floating point constants. For example, the form '<tt>double
1667 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1668 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1669 (and the only time that they are generated by the disassembler) is when a
1670 floating point constant must be emitted but it cannot be represented as a
1671 decimal floating point number. For example, NaN's, infinities, and other
1672 special values are represented in their IEEE hexadecimal format so that
1673 assembly and disassembly do not cause any bits to change in the constants.</p>
1677 <!-- ======================================================================= -->
1678 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1681 <div class="doc_text">
1682 <p>Aggregate constants arise from aggregation of simple constants
1683 and smaller aggregate constants.</p>
1686 <dt><b>Structure constants</b></dt>
1688 <dd>Structure constants are represented with notation similar to structure
1689 type definitions (a comma separated list of elements, surrounded by braces
1690 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1691 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1692 must have <a href="#t_struct">structure type</a>, and the number and
1693 types of elements must match those specified by the type.
1696 <dt><b>Array constants</b></dt>
1698 <dd>Array constants are represented with notation similar to array type
1699 definitions (a comma separated list of elements, surrounded by square brackets
1700 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1701 constants must have <a href="#t_array">array type</a>, and the number and
1702 types of elements must match those specified by the type.
1705 <dt><b>Vector constants</b></dt>
1707 <dd>Vector constants are represented with notation similar to vector type
1708 definitions (a comma separated list of elements, surrounded by
1709 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1710 i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
1711 href="#t_vector">vector type</a>, and the number and types of elements must
1712 match those specified by the type.
1715 <dt><b>Zero initialization</b></dt>
1717 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1718 value to zero of <em>any</em> type, including scalar and aggregate types.
1719 This is often used to avoid having to print large zero initializers (e.g. for
1720 large arrays) and is always exactly equivalent to using explicit zero
1727 <!-- ======================================================================= -->
1728 <div class="doc_subsection">
1729 <a name="globalconstants">Global Variable and Function Addresses</a>
1732 <div class="doc_text">
1734 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1735 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1736 constants. These constants are explicitly referenced when the <a
1737 href="#identifiers">identifier for the global</a> is used and always have <a
1738 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1741 <div class="doc_code">
1745 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1751 <!-- ======================================================================= -->
1752 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1753 <div class="doc_text">
1754 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1755 no specific value. Undefined values may be of any type and be used anywhere
1756 a constant is permitted.</p>
1758 <p>Undefined values indicate to the compiler that the program is well defined
1759 no matter what value is used, giving the compiler more freedom to optimize.
1763 <!-- ======================================================================= -->
1764 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1767 <div class="doc_text">
1769 <p>Constant expressions are used to allow expressions involving other constants
1770 to be used as constants. Constant expressions may be of any <a
1771 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1772 that does not have side effects (e.g. load and call are not supported). The
1773 following is the syntax for constant expressions:</p>
1776 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1777 <dd>Truncate a constant to another type. The bit size of CST must be larger
1778 than the bit size of TYPE. Both types must be integers.</dd>
1780 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1781 <dd>Zero extend a constant to another type. The bit size of CST must be
1782 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1784 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1785 <dd>Sign extend a constant to another type. The bit size of CST must be
1786 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1788 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1789 <dd>Truncate a floating point constant to another floating point type. The
1790 size of CST must be larger than the size of TYPE. Both types must be
1791 floating point.</dd>
1793 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1794 <dd>Floating point extend a constant to another type. The size of CST must be
1795 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1797 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
1798 <dd>Convert a floating point constant to the corresponding unsigned integer
1799 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1800 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1801 of the same number of elements. If the value won't fit in the integer type,
1802 the results are undefined.</dd>
1804 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1805 <dd>Convert a floating point constant to the corresponding signed integer
1806 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1807 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1808 of the same number of elements. If the value won't fit in the integer type,
1809 the results are undefined.</dd>
1811 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1812 <dd>Convert an unsigned integer constant to the corresponding floating point
1813 constant. TYPE must be a scalar or vector floating point type. CST must be of
1814 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1815 of the same number of elements. If the value won't fit in the floating point
1816 type, the results are undefined.</dd>
1818 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1819 <dd>Convert a signed integer constant to the corresponding floating point
1820 constant. TYPE must be a scalar or vector floating point type. CST must be of
1821 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1822 of the same number of elements. If the value won't fit in the floating point
1823 type, the results are undefined.</dd>
1825 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1826 <dd>Convert a pointer typed constant to the corresponding integer constant
1827 TYPE must be an integer type. CST must be of pointer type. The CST value is
1828 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1830 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1831 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1832 pointer type. CST must be of integer type. The CST value is zero extended,
1833 truncated, or unchanged to make it fit in a pointer size. This one is
1834 <i>really</i> dangerous!</dd>
1836 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1837 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1838 identical (same number of bits). The conversion is done as if the CST value
1839 was stored to memory and read back as TYPE. In other words, no bits change
1840 with this operator, just the type. This can be used for conversion of
1841 vector types to any other type, as long as they have the same bit width. For
1842 pointers it is only valid to cast to another pointer type. It is not valid
1843 to bitcast to or from an aggregate type.
1846 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1848 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1849 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1850 instruction, the index list may have zero or more indexes, which are required
1851 to make sense for the type of "CSTPTR".</dd>
1853 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1855 <dd>Perform the <a href="#i_select">select operation</a> on
1858 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1859 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1861 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1862 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1864 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
1865 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
1867 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
1868 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
1870 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1872 <dd>Perform the <a href="#i_extractelement">extractelement
1873 operation</a> on constants.</dd>
1875 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1877 <dd>Perform the <a href="#i_insertelement">insertelement
1878 operation</a> on constants.</dd>
1881 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1883 <dd>Perform the <a href="#i_shufflevector">shufflevector
1884 operation</a> on constants.</dd>
1886 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1888 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1889 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1890 binary</a> operations. The constraints on operands are the same as those for
1891 the corresponding instruction (e.g. no bitwise operations on floating point
1892 values are allowed).</dd>
1896 <!-- *********************************************************************** -->
1897 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1898 <!-- *********************************************************************** -->
1900 <!-- ======================================================================= -->
1901 <div class="doc_subsection">
1902 <a name="inlineasm">Inline Assembler Expressions</a>
1905 <div class="doc_text">
1908 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1909 Module-Level Inline Assembly</a>) through the use of a special value. This
1910 value represents the inline assembler as a string (containing the instructions
1911 to emit), a list of operand constraints (stored as a string), and a flag that
1912 indicates whether or not the inline asm expression has side effects. An example
1913 inline assembler expression is:
1916 <div class="doc_code">
1918 i32 (i32) asm "bswap $0", "=r,r"
1923 Inline assembler expressions may <b>only</b> be used as the callee operand of
1924 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1927 <div class="doc_code">
1929 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1934 Inline asms with side effects not visible in the constraint list must be marked
1935 as having side effects. This is done through the use of the
1936 '<tt>sideeffect</tt>' keyword, like so:
1939 <div class="doc_code">
1941 call void asm sideeffect "eieio", ""()
1945 <p>TODO: The format of the asm and constraints string still need to be
1946 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1947 need to be documented). This is probably best done by reference to another
1948 document that covers inline asm from a holistic perspective.
1953 <!-- *********************************************************************** -->
1954 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1955 <!-- *********************************************************************** -->
1957 <div class="doc_text">
1959 <p>The LLVM instruction set consists of several different
1960 classifications of instructions: <a href="#terminators">terminator
1961 instructions</a>, <a href="#binaryops">binary instructions</a>,
1962 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1963 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1964 instructions</a>.</p>
1968 <!-- ======================================================================= -->
1969 <div class="doc_subsection"> <a name="terminators">Terminator
1970 Instructions</a> </div>
1972 <div class="doc_text">
1974 <p>As mentioned <a href="#functionstructure">previously</a>, every
1975 basic block in a program ends with a "Terminator" instruction, which
1976 indicates which block should be executed after the current block is
1977 finished. These terminator instructions typically yield a '<tt>void</tt>'
1978 value: they produce control flow, not values (the one exception being
1979 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1980 <p>There are six different terminator instructions: the '<a
1981 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1982 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1983 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1984 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1985 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1989 <!-- _______________________________________________________________________ -->
1990 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1991 Instruction</a> </div>
1992 <div class="doc_text">
1995 ret <type> <value> <i>; Return a value from a non-void function</i>
1996 ret void <i>; Return from void function</i>
2001 <p>The '<tt>ret</tt>' instruction is used to return control flow (and
2002 optionally a value) from a function back to the caller.</p>
2003 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
2004 returns a value and then causes control flow, and one that just causes
2005 control flow to occur.</p>
2009 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument,
2010 the return value. The type of the return value must be a
2011 '<a href="#t_firstclass">first class</a>' type.</p>
2013 <p>A function is not <a href="#wellformed">well formed</a> if
2014 it it has a non-void return type and contains a '<tt>ret</tt>'
2015 instruction with no return value or a return value with a type that
2016 does not match its type, or if it has a void return type and contains
2017 a '<tt>ret</tt>' instruction with a return value.</p>
2021 <p>When the '<tt>ret</tt>' instruction is executed, control flow
2022 returns back to the calling function's context. If the caller is a "<a
2023 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
2024 the instruction after the call. If the caller was an "<a
2025 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
2026 at the beginning of the "normal" destination block. If the instruction
2027 returns a value, that value shall set the call or invoke instruction's
2033 ret i32 5 <i>; Return an integer value of 5</i>
2034 ret void <i>; Return from a void function</i>
2035 ret { i32, i8 } { i32 4, i8 2 } <i>; Return an aggregate of values 4 and 2</i>
2038 <!-- _______________________________________________________________________ -->
2039 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2040 <div class="doc_text">
2042 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2045 <p>The '<tt>br</tt>' instruction is used to cause control flow to
2046 transfer to a different basic block in the current function. There are
2047 two forms of this instruction, corresponding to a conditional branch
2048 and an unconditional branch.</p>
2050 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
2051 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
2052 unconditional form of the '<tt>br</tt>' instruction takes a single
2053 '<tt>label</tt>' value as a target.</p>
2055 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2056 argument is evaluated. If the value is <tt>true</tt>, control flows
2057 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2058 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2060 <pre>Test:<br> %cond = <a href="#i_icmp">icmp</a> eq, i32 %a, %b<br> br i1 %cond, label %IfEqual, label %IfUnequal<br>IfEqual:<br> <a
2061 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
2063 <!-- _______________________________________________________________________ -->
2064 <div class="doc_subsubsection">
2065 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2068 <div class="doc_text">
2072 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2077 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2078 several different places. It is a generalization of the '<tt>br</tt>'
2079 instruction, allowing a branch to occur to one of many possible
2085 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2086 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
2087 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
2088 table is not allowed to contain duplicate constant entries.</p>
2092 <p>The <tt>switch</tt> instruction specifies a table of values and
2093 destinations. When the '<tt>switch</tt>' instruction is executed, this
2094 table is searched for the given value. If the value is found, control flow is
2095 transfered to the corresponding destination; otherwise, control flow is
2096 transfered to the default destination.</p>
2098 <h5>Implementation:</h5>
2100 <p>Depending on properties of the target machine and the particular
2101 <tt>switch</tt> instruction, this instruction may be code generated in different
2102 ways. For example, it could be generated as a series of chained conditional
2103 branches or with a lookup table.</p>
2108 <i>; Emulate a conditional br instruction</i>
2109 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2110 switch i32 %Val, label %truedest [ i32 0, label %falsedest ]
2112 <i>; Emulate an unconditional br instruction</i>
2113 switch i32 0, label %dest [ ]
2115 <i>; Implement a jump table:</i>
2116 switch i32 %val, label %otherwise [ i32 0, label %onzero
2118 i32 2, label %ontwo ]
2122 <!-- _______________________________________________________________________ -->
2123 <div class="doc_subsubsection">
2124 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2127 <div class="doc_text">
2132 <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>]
2133 to label <normal label> unwind label <exception label>
2138 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2139 function, with the possibility of control flow transfer to either the
2140 '<tt>normal</tt>' label or the
2141 '<tt>exception</tt>' label. If the callee function returns with the
2142 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2143 "normal" label. If the callee (or any indirect callees) returns with the "<a
2144 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
2145 continued at the dynamically nearest "exception" label.</p>
2149 <p>This instruction requires several arguments:</p>
2153 The optional "cconv" marker indicates which <a href="#callingconv">calling
2154 convention</a> the call should use. If none is specified, the call defaults
2155 to using C calling conventions.
2158 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2159 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
2160 and '<tt>inreg</tt>' attributes are valid here.</li>
2162 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2163 function value being invoked. In most cases, this is a direct function
2164 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2165 an arbitrary pointer to function value.
2168 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2169 function to be invoked. </li>
2171 <li>'<tt>function args</tt>': argument list whose types match the function
2172 signature argument types. If the function signature indicates the function
2173 accepts a variable number of arguments, the extra arguments can be
2176 <li>'<tt>normal label</tt>': the label reached when the called function
2177 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2179 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2180 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2182 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
2183 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2184 '<tt>readnone</tt>' attributes are valid here.</li>
2189 <p>This instruction is designed to operate as a standard '<tt><a
2190 href="#i_call">call</a></tt>' instruction in most regards. The primary
2191 difference is that it establishes an association with a label, which is used by
2192 the runtime library to unwind the stack.</p>
2194 <p>This instruction is used in languages with destructors to ensure that proper
2195 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2196 exception. Additionally, this is important for implementation of
2197 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
2201 %retval = invoke i32 @Test(i32 15) to label %Continue
2202 unwind label %TestCleanup <i>; {i32}:retval set</i>
2203 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
2204 unwind label %TestCleanup <i>; {i32}:retval set</i>
2209 <!-- _______________________________________________________________________ -->
2211 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2212 Instruction</a> </div>
2214 <div class="doc_text">
2223 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2224 at the first callee in the dynamic call stack which used an <a
2225 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2226 primarily used to implement exception handling.</p>
2230 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
2231 immediately halt. The dynamic call stack is then searched for the first <a
2232 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2233 execution continues at the "exceptional" destination block specified by the
2234 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2235 dynamic call chain, undefined behavior results.</p>
2238 <!-- _______________________________________________________________________ -->
2240 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2241 Instruction</a> </div>
2243 <div class="doc_text">
2252 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2253 instruction is used to inform the optimizer that a particular portion of the
2254 code is not reachable. This can be used to indicate that the code after a
2255 no-return function cannot be reached, and other facts.</p>
2259 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2264 <!-- ======================================================================= -->
2265 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2266 <div class="doc_text">
2267 <p>Binary operators are used to do most of the computation in a
2268 program. They require two operands of the same type, execute an operation on them, and
2269 produce a single value. The operands might represent
2270 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
2271 The result value has the same type as its operands.</p>
2272 <p>There are several different binary operators:</p>
2274 <!-- _______________________________________________________________________ -->
2275 <div class="doc_subsubsection">
2276 <a name="i_add">'<tt>add</tt>' Instruction</a>
2279 <div class="doc_text">
2284 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2289 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2293 <p>The two arguments to the '<tt>add</tt>' instruction must be <a
2294 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>, or
2295 <a href="#t_vector">vector</a> values. Both arguments must have identical
2300 <p>The value produced is the integer or floating point sum of the two
2303 <p>If an integer sum has unsigned overflow, the result returned is the
2304 mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2307 <p>Because LLVM integers use a two's complement representation, this
2308 instruction is appropriate for both signed and unsigned integers.</p>
2313 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2316 <!-- _______________________________________________________________________ -->
2317 <div class="doc_subsubsection">
2318 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2321 <div class="doc_text">
2326 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2331 <p>The '<tt>sub</tt>' instruction returns the difference of its two
2334 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
2335 '<tt>neg</tt>' instruction present in most other intermediate
2336 representations.</p>
2340 <p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2341 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2342 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2347 <p>The value produced is the integer or floating point difference of
2348 the two operands.</p>
2350 <p>If an integer difference has unsigned overflow, the result returned is the
2351 mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2354 <p>Because LLVM integers use a two's complement representation, this
2355 instruction is appropriate for both signed and unsigned integers.</p>
2359 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2360 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2364 <!-- _______________________________________________________________________ -->
2365 <div class="doc_subsubsection">
2366 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2369 <div class="doc_text">
2372 <pre> <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2375 <p>The '<tt>mul</tt>' instruction returns the product of its two
2380 <p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2381 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2382 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2387 <p>The value produced is the integer or floating point product of the
2390 <p>If the result of an integer multiplication has unsigned overflow,
2391 the result returned is the mathematical result modulo
2392 2<sup>n</sup>, where n is the bit width of the result.</p>
2393 <p>Because LLVM integers use a two's complement representation, and the
2394 result is the same width as the operands, this instruction returns the
2395 correct result for both signed and unsigned integers. If a full product
2396 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2397 should be sign-extended or zero-extended as appropriate to the
2398 width of the full product.</p>
2400 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2404 <!-- _______________________________________________________________________ -->
2405 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2407 <div class="doc_text">
2409 <pre> <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2412 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2417 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2418 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2419 values. Both arguments must have identical types.</p>
2423 <p>The value produced is the unsigned integer quotient of the two operands.</p>
2424 <p>Note that unsigned integer division and signed integer division are distinct
2425 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2426 <p>Division by zero leads to undefined behavior.</p>
2428 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2431 <!-- _______________________________________________________________________ -->
2432 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2434 <div class="doc_text">
2437 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2442 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2447 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2448 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2449 values. Both arguments must have identical types.</p>
2452 <p>The value produced is the signed integer quotient of the two operands rounded towards zero.</p>
2453 <p>Note that signed integer division and unsigned integer division are distinct
2454 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2455 <p>Division by zero leads to undefined behavior. Overflow also leads to
2456 undefined behavior; this is a rare case, but can occur, for example,
2457 by doing a 32-bit division of -2147483648 by -1.</p>
2459 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2462 <!-- _______________________________________________________________________ -->
2463 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2464 Instruction</a> </div>
2465 <div class="doc_text">
2468 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2472 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2477 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2478 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2479 of floating point values. Both arguments must have identical types.</p>
2483 <p>The value produced is the floating point quotient of the two operands.</p>
2488 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2492 <!-- _______________________________________________________________________ -->
2493 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2495 <div class="doc_text">
2497 <pre> <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2500 <p>The '<tt>urem</tt>' instruction returns the remainder from the
2501 unsigned division of its two arguments.</p>
2503 <p>The two arguments to the '<tt>urem</tt>' instruction must be
2504 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2505 values. Both arguments must have identical types.</p>
2507 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2508 This instruction always performs an unsigned division to get the remainder.</p>
2509 <p>Note that unsigned integer remainder and signed integer remainder are
2510 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2511 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
2513 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2517 <!-- _______________________________________________________________________ -->
2518 <div class="doc_subsubsection">
2519 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2522 <div class="doc_text">
2527 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2532 <p>The '<tt>srem</tt>' instruction returns the remainder from the
2533 signed division of its two operands. This instruction can also take
2534 <a href="#t_vector">vector</a> versions of the values in which case
2535 the elements must be integers.</p>
2539 <p>The two arguments to the '<tt>srem</tt>' instruction must be
2540 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2541 values. Both arguments must have identical types.</p>
2545 <p>This instruction returns the <i>remainder</i> of a division (where the result
2546 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2547 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
2548 a value. For more information about the difference, see <a
2549 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2550 Math Forum</a>. For a table of how this is implemented in various languages,
2551 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2552 Wikipedia: modulo operation</a>.</p>
2553 <p>Note that signed integer remainder and unsigned integer remainder are
2554 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2555 <p>Taking the remainder of a division by zero leads to undefined behavior.
2556 Overflow also leads to undefined behavior; this is a rare case, but can occur,
2557 for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2558 (The remainder doesn't actually overflow, but this rule lets srem be
2559 implemented using instructions that return both the result of the division
2560 and the remainder.)</p>
2562 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2566 <!-- _______________________________________________________________________ -->
2567 <div class="doc_subsubsection">
2568 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2570 <div class="doc_text">
2573 <pre> <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2576 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2577 division of its two operands.</p>
2579 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2580 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2581 of floating point values. Both arguments must have identical types.</p>
2585 <p>This instruction returns the <i>remainder</i> of a division.
2586 The remainder has the same sign as the dividend.</p>
2591 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2595 <!-- ======================================================================= -->
2596 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2597 Operations</a> </div>
2598 <div class="doc_text">
2599 <p>Bitwise binary operators are used to do various forms of
2600 bit-twiddling in a program. They are generally very efficient
2601 instructions and can commonly be strength reduced from other
2602 instructions. They require two operands of the same type, execute an operation on them,
2603 and produce a single value. The resulting value is the same type as its operands.</p>
2606 <!-- _______________________________________________________________________ -->
2607 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2608 Instruction</a> </div>
2609 <div class="doc_text">
2611 <pre> <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2616 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2617 the left a specified number of bits.</p>
2621 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2622 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2623 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2627 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2628 where n is the width of the result. If <tt>op2</tt> is (statically or dynamically) negative or
2629 equal to or larger than the number of bits in <tt>op1</tt>, the result is undefined.
2630 If the arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2631 corresponding shift amount in <tt>op2</tt>.</p>
2633 <h5>Example:</h5><pre>
2634 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2635 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2636 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2637 <result> = shl i32 1, 32 <i>; undefined</i>
2638 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
2641 <!-- _______________________________________________________________________ -->
2642 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2643 Instruction</a> </div>
2644 <div class="doc_text">
2646 <pre> <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2650 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2651 operand shifted to the right a specified number of bits with zero fill.</p>
2654 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2655 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2656 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2660 <p>This instruction always performs a logical shift right operation. The most
2661 significant bits of the result will be filled with zero bits after the
2662 shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
2663 the number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
2664 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding shift
2665 amount in <tt>op2</tt>.</p>
2669 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2670 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2671 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2672 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2673 <result> = lshr i32 1, 32 <i>; undefined</i>
2674 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
2678 <!-- _______________________________________________________________________ -->
2679 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2680 Instruction</a> </div>
2681 <div class="doc_text">
2684 <pre> <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2688 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2689 operand shifted to the right a specified number of bits with sign extension.</p>
2692 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2693 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2694 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2697 <p>This instruction always performs an arithmetic shift right operation,
2698 The most significant bits of the result will be filled with the sign bit
2699 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
2700 larger than the number of bits in <tt>op1</tt>, the result is undefined. If the
2701 arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2702 corresponding shift amount in <tt>op2</tt>.</p>
2706 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2707 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2708 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2709 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
2710 <result> = ashr i32 1, 32 <i>; undefined</i>
2711 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
2715 <!-- _______________________________________________________________________ -->
2716 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2717 Instruction</a> </div>
2719 <div class="doc_text">
2724 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2729 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2730 its two operands.</p>
2734 <p>The two arguments to the '<tt>and</tt>' instruction must be
2735 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2736 values. Both arguments must have identical types.</p>
2739 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2742 <table border="1" cellspacing="0" cellpadding="4">
2774 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
2775 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2776 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2779 <!-- _______________________________________________________________________ -->
2780 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2781 <div class="doc_text">
2783 <pre> <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2786 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2787 or of its two operands.</p>
2790 <p>The two arguments to the '<tt>or</tt>' instruction must be
2791 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2792 values. Both arguments must have identical types.</p>
2794 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2797 <table border="1" cellspacing="0" cellpadding="4">
2828 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2829 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2830 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2833 <!-- _______________________________________________________________________ -->
2834 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2835 Instruction</a> </div>
2836 <div class="doc_text">
2838 <pre> <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2841 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2842 or of its two operands. The <tt>xor</tt> is used to implement the
2843 "one's complement" operation, which is the "~" operator in C.</p>
2845 <p>The two arguments to the '<tt>xor</tt>' instruction must be
2846 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2847 values. Both arguments must have identical types.</p>
2851 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2854 <table border="1" cellspacing="0" cellpadding="4">
2886 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2887 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2888 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2889 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2893 <!-- ======================================================================= -->
2894 <div class="doc_subsection">
2895 <a name="vectorops">Vector Operations</a>
2898 <div class="doc_text">
2900 <p>LLVM supports several instructions to represent vector operations in a
2901 target-independent manner. These instructions cover the element-access and
2902 vector-specific operations needed to process vectors effectively. While LLVM
2903 does directly support these vector operations, many sophisticated algorithms
2904 will want to use target-specific intrinsics to take full advantage of a specific
2909 <!-- _______________________________________________________________________ -->
2910 <div class="doc_subsubsection">
2911 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2914 <div class="doc_text">
2919 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
2925 The '<tt>extractelement</tt>' instruction extracts a single scalar
2926 element from a vector at a specified index.
2933 The first operand of an '<tt>extractelement</tt>' instruction is a
2934 value of <a href="#t_vector">vector</a> type. The second operand is
2935 an index indicating the position from which to extract the element.
2936 The index may be a variable.</p>
2941 The result is a scalar of the same type as the element type of
2942 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2943 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2944 results are undefined.
2950 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
2955 <!-- _______________________________________________________________________ -->
2956 <div class="doc_subsubsection">
2957 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2960 <div class="doc_text">
2965 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
2971 The '<tt>insertelement</tt>' instruction inserts a scalar
2972 element into a vector at a specified index.
2979 The first operand of an '<tt>insertelement</tt>' instruction is a
2980 value of <a href="#t_vector">vector</a> type. The second operand is a
2981 scalar value whose type must equal the element type of the first
2982 operand. The third operand is an index indicating the position at
2983 which to insert the value. The index may be a variable.</p>
2988 The result is a vector of the same type as <tt>val</tt>. Its
2989 element values are those of <tt>val</tt> except at position
2990 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2991 exceeds the length of <tt>val</tt>, the results are undefined.
2997 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
3001 <!-- _______________________________________________________________________ -->
3002 <div class="doc_subsubsection">
3003 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
3006 <div class="doc_text">
3011 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
3017 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
3018 from two input vectors, returning a vector with the same element type as
3019 the input and length that is the same as the shuffle mask.
3025 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
3026 with types that match each other. The third argument is a shuffle mask whose
3027 element type is always 'i32'. The result of the instruction is a vector whose
3028 length is the same as the shuffle mask and whose element type is the same as
3029 the element type of the first two operands.
3033 The shuffle mask operand is required to be a constant vector with either
3034 constant integer or undef values.
3040 The elements of the two input vectors are numbered from left to right across
3041 both of the vectors. The shuffle mask operand specifies, for each element of
3042 the result vector, which element of the two input vectors the result element
3043 gets. The element selector may be undef (meaning "don't care") and the second
3044 operand may be undef if performing a shuffle from only one vector.
3050 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3051 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
3052 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
3053 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
3054 %result = shufflevector <8 x i32> %v1, <8 x i32> undef,
3055 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
3056 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3057 <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>
3062 <!-- ======================================================================= -->
3063 <div class="doc_subsection">
3064 <a name="aggregateops">Aggregate Operations</a>
3067 <div class="doc_text">
3069 <p>LLVM supports several instructions for working with aggregate values.
3074 <!-- _______________________________________________________________________ -->
3075 <div class="doc_subsubsection">
3076 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3079 <div class="doc_text">
3084 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
3090 The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3091 or array element from an aggregate value.
3098 The first operand of an '<tt>extractvalue</tt>' instruction is a
3099 value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
3100 type. The operands are constant indices to specify which value to extract
3101 in a similar manner as indices in a
3102 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3108 The result is the value at the position in the aggregate specified by
3115 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
3120 <!-- _______________________________________________________________________ -->
3121 <div class="doc_subsubsection">
3122 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3125 <div class="doc_text">
3130 <result> = insertvalue <aggregate type> <val>, <ty> <val>, <idx> <i>; yields <n x <ty>></i>
3136 The '<tt>insertvalue</tt>' instruction inserts a value
3137 into a struct field or array element in an aggregate.
3144 The first operand of an '<tt>insertvalue</tt>' instruction is a
3145 value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3146 The second operand is a first-class value to insert.
3147 The following operands are constant indices
3148 indicating the position at which to insert the value in a similar manner as
3150 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3151 The value to insert must have the same type as the value identified
3158 The result is an aggregate of the same type as <tt>val</tt>. Its
3159 value is that of <tt>val</tt> except that the value at the position
3160 specified by the indices is that of <tt>elt</tt>.
3166 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
3171 <!-- ======================================================================= -->
3172 <div class="doc_subsection">
3173 <a name="memoryops">Memory Access and Addressing Operations</a>
3176 <div class="doc_text">
3178 <p>A key design point of an SSA-based representation is how it
3179 represents memory. In LLVM, no memory locations are in SSA form, which
3180 makes things very simple. This section describes how to read, write,
3181 allocate, and free memory in LLVM.</p>
3185 <!-- _______________________________________________________________________ -->
3186 <div class="doc_subsubsection">
3187 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3190 <div class="doc_text">
3195 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3200 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
3201 heap and returns a pointer to it. The object is always allocated in the generic
3202 address space (address space zero).</p>
3206 <p>The '<tt>malloc</tt>' instruction allocates
3207 <tt>sizeof(<type>)*NumElements</tt>
3208 bytes of memory from the operating system and returns a pointer of the
3209 appropriate type to the program. If "NumElements" is specified, it is the
3210 number of elements allocated, otherwise "NumElements" is defaulted to be one.
3211 If a constant alignment is specified, the value result of the allocation is guaranteed to
3212 be aligned to at least that boundary. If not specified, or if zero, the target can
3213 choose to align the allocation on any convenient boundary.</p>
3215 <p>'<tt>type</tt>' must be a sized type.</p>
3219 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
3220 a pointer is returned. The result of a zero byte allocation is undefined. The
3221 result is null if there is insufficient memory available.</p>
3226 %array = malloc [4 x i8] <i>; yields {[%4 x i8]*}:array</i>
3228 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3229 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3230 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3231 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3232 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3236 <!-- _______________________________________________________________________ -->
3237 <div class="doc_subsubsection">
3238 <a name="i_free">'<tt>free</tt>' Instruction</a>
3241 <div class="doc_text">
3246 free <type> <value> <i>; yields {void}</i>
3251 <p>The '<tt>free</tt>' instruction returns memory back to the unused
3252 memory heap to be reallocated in the future.</p>
3256 <p>'<tt>value</tt>' shall be a pointer value that points to a value
3257 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3262 <p>Access to the memory pointed to by the pointer is no longer defined
3263 after this instruction executes. If the pointer is null, the operation
3269 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
3270 free [4 x i8]* %array
3274 <!-- _______________________________________________________________________ -->
3275 <div class="doc_subsubsection">
3276 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3279 <div class="doc_text">
3284 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3289 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3290 currently executing function, to be automatically released when this function
3291 returns to its caller. The object is always allocated in the generic address
3292 space (address space zero).</p>
3296 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
3297 bytes of memory on the runtime stack, returning a pointer of the
3298 appropriate type to the program. If "NumElements" is specified, it is the
3299 number of elements allocated, otherwise "NumElements" is defaulted to be one.
3300 If a constant alignment is specified, the value result of the allocation is guaranteed
3301 to be aligned to at least that boundary. If not specified, or if zero, the target
3302 can choose to align the allocation on any convenient boundary.</p>
3304 <p>'<tt>type</tt>' may be any sized type.</p>
3308 <p>Memory is allocated; a pointer is returned. The operation is undefiend if
3309 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
3310 memory is automatically released when the function returns. The '<tt>alloca</tt>'
3311 instruction is commonly used to represent automatic variables that must
3312 have an address available. When the function returns (either with the <tt><a
3313 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
3314 instructions), the memory is reclaimed. Allocating zero bytes
3315 is legal, but the result is undefined.</p>
3320 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3321 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3322 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3323 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3327 <!-- _______________________________________________________________________ -->
3328 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3329 Instruction</a> </div>
3330 <div class="doc_text">
3332 <pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
3334 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3336 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
3337 address from which to load. The pointer must point to a <a
3338 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3339 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3340 the number or order of execution of this <tt>load</tt> with other
3341 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3344 The optional constant "align" argument specifies the alignment of the operation
3345 (that is, the alignment of the memory address). A value of 0 or an
3346 omitted "align" argument means that the operation has the preferential
3347 alignment for the target. It is the responsibility of the code emitter
3348 to ensure that the alignment information is correct. Overestimating
3349 the alignment results in an undefined behavior. Underestimating the
3350 alignment may produce less efficient code. An alignment of 1 is always
3354 <p>The location of memory pointed to is loaded.</p>
3356 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3358 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3359 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3362 <!-- _______________________________________________________________________ -->
3363 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3364 Instruction</a> </div>
3365 <div class="doc_text">
3367 <pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3368 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3371 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3373 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
3374 to store and an address at which to store it. The type of the '<tt><pointer></tt>'
3375 operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3376 of the '<tt><value></tt>'
3377 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3378 optimizer is not allowed to modify the number or order of execution of
3379 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3380 href="#i_store">store</a></tt> instructions.</p>
3382 The optional constant "align" argument specifies the alignment of the operation
3383 (that is, the alignment of the memory address). A value of 0 or an
3384 omitted "align" argument means that the operation has the preferential
3385 alignment for the target. It is the responsibility of the code emitter
3386 to ensure that the alignment information is correct. Overestimating
3387 the alignment results in an undefined behavior. Underestimating the
3388 alignment may produce less efficient code. An alignment of 1 is always
3392 <p>The contents of memory are updated to contain '<tt><value></tt>'
3393 at the location specified by the '<tt><pointer></tt>' operand.</p>
3395 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3396 store i32 3, i32* %ptr <i>; yields {void}</i>
3397 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
3401 <!-- _______________________________________________________________________ -->
3402 <div class="doc_subsubsection">
3403 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3406 <div class="doc_text">
3409 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
3415 The '<tt>getelementptr</tt>' instruction is used to get the address of a
3416 subelement of an aggregate data structure. It performs address calculation only
3417 and does not access memory.</p>
3421 <p>The first argument is always a pointer, and forms the basis of the
3422 calculation. The remaining arguments are indices, that indicate which of the
3423 elements of the aggregate object are indexed. The interpretation of each index
3424 is dependent on the type being indexed into. The first index always indexes the
3425 pointer value given as the first argument, the second index indexes a value of
3426 the type pointed to (not necessarily the value directly pointed to, since the
3427 first index can be non-zero), etc. The first type indexed into must be a pointer
3428 value, subsequent types can be arrays, vectors and structs. Note that subsequent
3429 types being indexed into can never be pointers, since that would require loading
3430 the pointer before continuing calculation.</p>
3432 <p>The type of each index argument depends on the type it is indexing into.
3433 When indexing into a (packed) structure, only <tt>i32</tt> integer
3434 <b>constants</b> are allowed. When indexing into an array, pointer or vector,
3435 only integers of 32 or 64 bits are allowed (also non-constants). 32-bit values
3436 will be sign extended to 64-bits if required.</p>
3438 <p>For example, let's consider a C code fragment and how it gets
3439 compiled to LLVM:</p>
3441 <div class="doc_code">
3454 int *foo(struct ST *s) {
3455 return &s[1].Z.B[5][13];
3460 <p>The LLVM code generated by the GCC frontend is:</p>
3462 <div class="doc_code">
3464 %RT = <a href="#namedtypes">type</a> { i8 , [10 x [20 x i32]], i8 }
3465 %ST = <a href="#namedtypes">type</a> { i32, double, %RT }
3467 define i32* %foo(%ST* %s) {
3469 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3477 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3478 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3479 }</tt>' type, a structure. The second index indexes into the third element of
3480 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3481 i8 }</tt>' type, another structure. The third index indexes into the second
3482 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3483 array. The two dimensions of the array are subscripted into, yielding an
3484 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3485 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3487 <p>Note that it is perfectly legal to index partially through a
3488 structure, returning a pointer to an inner element. Because of this,
3489 the LLVM code for the given testcase is equivalent to:</p>
3492 define i32* %foo(%ST* %s) {
3493 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3494 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3495 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3496 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3497 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3502 <p>Note that it is undefined to access an array out of bounds: array and
3503 pointer indexes must always be within the defined bounds of the array type.
3504 The one exception for this rule is zero length arrays. These arrays are
3505 defined to be accessible as variable length arrays, which requires access
3506 beyond the zero'th element.</p>
3508 <p>The getelementptr instruction is often confusing. For some more insight
3509 into how it works, see <a href="GetElementPtr.html">the getelementptr
3515 <i>; yields [12 x i8]*:aptr</i>
3516 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
3517 <i>; yields i8*:vptr</i>
3518 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
3519 <i>; yields i8*:eptr</i>
3520 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
3524 <!-- ======================================================================= -->
3525 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3527 <div class="doc_text">
3528 <p>The instructions in this category are the conversion instructions (casting)
3529 which all take a single operand and a type. They perform various bit conversions
3533 <!-- _______________________________________________________________________ -->
3534 <div class="doc_subsubsection">
3535 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3537 <div class="doc_text">
3541 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
3546 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3551 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3552 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3553 and type of the result, which must be an <a href="#t_integer">integer</a>
3554 type. The bit size of <tt>value</tt> must be larger than the bit size of
3555 <tt>ty2</tt>. Equal sized types are not allowed.</p>
3559 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3560 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3561 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3562 It will always truncate bits.</p>
3566 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3567 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3568 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3572 <!-- _______________________________________________________________________ -->
3573 <div class="doc_subsubsection">
3574 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3576 <div class="doc_text">
3580 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
3584 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
3589 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3590 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3591 also be of <a href="#t_integer">integer</a> type. The bit size of the
3592 <tt>value</tt> must be smaller than the bit size of the destination type,
3596 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3597 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3599 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
3603 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3604 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3608 <!-- _______________________________________________________________________ -->
3609 <div class="doc_subsubsection">
3610 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3612 <div class="doc_text">
3616 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
3620 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3624 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3625 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3626 also be of <a href="#t_integer">integer</a> type. The bit size of the
3627 <tt>value</tt> must be smaller than the bit size of the destination type,
3632 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3633 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3634 the type <tt>ty2</tt>.</p>
3636 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
3640 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3641 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3645 <!-- _______________________________________________________________________ -->
3646 <div class="doc_subsubsection">
3647 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3650 <div class="doc_text">
3655 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
3659 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3664 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3665 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3666 cast it to. The size of <tt>value</tt> must be larger than the size of
3667 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3668 <i>no-op cast</i>.</p>
3671 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3672 <a href="#t_floating">floating point</a> type to a smaller
3673 <a href="#t_floating">floating point</a> type. If the value cannot fit within
3674 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3678 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3679 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3683 <!-- _______________________________________________________________________ -->
3684 <div class="doc_subsubsection">
3685 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3687 <div class="doc_text">
3691 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
3695 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3696 floating point value.</p>
3699 <p>The '<tt>fpext</tt>' instruction takes a
3700 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3701 and a <a href="#t_floating">floating point</a> type to cast it to. The source
3702 type must be smaller than the destination type.</p>
3705 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3706 <a href="#t_floating">floating point</a> type to a larger
3707 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3708 used to make a <i>no-op cast</i> because it always changes bits. Use
3709 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3713 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3714 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3718 <!-- _______________________________________________________________________ -->
3719 <div class="doc_subsubsection">
3720 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3722 <div class="doc_text">
3726 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
3730 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
3731 unsigned integer equivalent of type <tt>ty2</tt>.
3735 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
3736 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3737 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3738 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3739 vector integer type with the same number of elements as <tt>ty</tt></p>
3742 <p> The '<tt>fptoui</tt>' instruction converts its
3743 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3744 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3745 the results are undefined.</p>
3749 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
3750 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
3751 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
3755 <!-- _______________________________________________________________________ -->
3756 <div class="doc_subsubsection">
3757 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3759 <div class="doc_text">
3763 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
3767 <p>The '<tt>fptosi</tt>' instruction converts
3768 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3772 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
3773 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3774 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3775 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3776 vector integer type with the same number of elements as <tt>ty</tt></p>
3779 <p>The '<tt>fptosi</tt>' instruction converts its
3780 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3781 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3782 the results are undefined.</p>
3786 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
3787 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
3788 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3792 <!-- _______________________________________________________________________ -->
3793 <div class="doc_subsubsection">
3794 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3796 <div class="doc_text">
3800 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3804 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3805 integer and converts that value to the <tt>ty2</tt> type.</p>
3808 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3809 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3810 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3811 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3812 floating point type with the same number of elements as <tt>ty</tt></p>
3815 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3816 integer quantity and converts it to the corresponding floating point value. If
3817 the value cannot fit in the floating point value, the results are undefined.</p>
3821 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3822 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3826 <!-- _______________________________________________________________________ -->
3827 <div class="doc_subsubsection">
3828 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3830 <div class="doc_text">
3834 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3838 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3839 integer and converts that value to the <tt>ty2</tt> type.</p>
3842 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3843 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3844 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3845 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3846 floating point type with the same number of elements as <tt>ty</tt></p>
3849 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3850 integer quantity and converts it to the corresponding floating point value. If
3851 the value cannot fit in the floating point value, the results are undefined.</p>
3855 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3856 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3860 <!-- _______________________________________________________________________ -->
3861 <div class="doc_subsubsection">
3862 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3864 <div class="doc_text">
3868 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
3872 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3873 the integer type <tt>ty2</tt>.</p>
3876 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3877 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3878 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
3881 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3882 <tt>ty2</tt> by interpreting the pointer value as an integer and either
3883 truncating or zero extending that value to the size of the integer type. If
3884 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3885 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3886 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3891 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3892 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3896 <!-- _______________________________________________________________________ -->
3897 <div class="doc_subsubsection">
3898 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3900 <div class="doc_text">
3904 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
3908 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3909 a pointer type, <tt>ty2</tt>.</p>
3912 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3913 value to cast, and a type to cast it to, which must be a
3914 <a href="#t_pointer">pointer</a> type.</p>
3917 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3918 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
3919 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3920 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3921 the size of a pointer then a zero extension is done. If they are the same size,
3922 nothing is done (<i>no-op cast</i>).</p>
3926 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3927 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3928 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3932 <!-- _______________________________________________________________________ -->
3933 <div class="doc_subsubsection">
3934 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3936 <div class="doc_text">
3940 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3945 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3946 <tt>ty2</tt> without changing any bits.</p>
3950 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3951 a non-aggregate first class value, and a type to cast it to, which must also be
3952 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
3954 and the destination type, <tt>ty2</tt>, must be identical. If the source
3955 type is a pointer, the destination type must also be a pointer. This
3956 instruction supports bitwise conversion of vectors to integers and to vectors
3957 of other types (as long as they have the same size).</p>
3960 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3961 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3962 this conversion. The conversion is done as if the <tt>value</tt> had been
3963 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3964 converted to other pointer types with this instruction. To convert pointers to
3965 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3966 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3970 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3971 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3972 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
3976 <!-- ======================================================================= -->
3977 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3978 <div class="doc_text">
3979 <p>The instructions in this category are the "miscellaneous"
3980 instructions, which defy better classification.</p>
3983 <!-- _______________________________________________________________________ -->
3984 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3986 <div class="doc_text">
3988 <pre> <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
3991 <p>The '<tt>icmp</tt>' instruction returns a boolean value or
3992 a vector of boolean values based on comparison
3993 of its two integer, integer vector, or pointer operands.</p>
3995 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3996 the condition code indicating the kind of comparison to perform. It is not
3997 a value, just a keyword. The possible condition code are:
4000 <li><tt>eq</tt>: equal</li>
4001 <li><tt>ne</tt>: not equal </li>
4002 <li><tt>ugt</tt>: unsigned greater than</li>
4003 <li><tt>uge</tt>: unsigned greater or equal</li>
4004 <li><tt>ult</tt>: unsigned less than</li>
4005 <li><tt>ule</tt>: unsigned less or equal</li>
4006 <li><tt>sgt</tt>: signed greater than</li>
4007 <li><tt>sge</tt>: signed greater or equal</li>
4008 <li><tt>slt</tt>: signed less than</li>
4009 <li><tt>sle</tt>: signed less or equal</li>
4011 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
4012 <a href="#t_pointer">pointer</a>
4013 or integer <a href="#t_vector">vector</a> typed.
4014 They must also be identical types.</p>
4016 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to
4017 the condition code given as <tt>cond</tt>. The comparison performed always
4018 yields either an <a href="#t_primitive"><tt>i1</tt></a> or vector of <tt>i1</tt> result, as follows:
4021 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
4022 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
4024 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
4025 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.</li>
4026 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
4027 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4028 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
4029 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4030 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
4031 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4032 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
4033 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4034 <li><tt>sgt</tt>: interprets the operands as signed values and yields
4035 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4036 <li><tt>sge</tt>: interprets the operands as signed values and yields
4037 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4038 <li><tt>slt</tt>: interprets the operands as signed values and yields
4039 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4040 <li><tt>sle</tt>: interprets the operands as signed values and yields
4041 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4043 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
4044 values are compared as if they were integers.</p>
4045 <p>If the operands are integer vectors, then they are compared
4046 element by element. The result is an <tt>i1</tt> vector with
4047 the same number of elements as the values being compared.
4048 Otherwise, the result is an <tt>i1</tt>.
4052 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
4053 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
4054 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
4055 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4056 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4057 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
4061 <!-- _______________________________________________________________________ -->
4062 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4064 <div class="doc_text">
4066 <pre> <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4069 <p>The '<tt>fcmp</tt>' instruction returns a boolean value
4070 or vector of boolean values based on comparison
4071 of its operands.</p>
4073 If the operands are floating point scalars, then the result
4074 type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
4076 <p>If the operands are floating point vectors, then the result type
4077 is a vector of boolean with the same number of elements as the
4078 operands being compared.</p>
4080 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4081 the condition code indicating the kind of comparison to perform. It is not
4082 a value, just a keyword. The possible condition code are:</p>
4084 <li><tt>false</tt>: no comparison, always returns false</li>
4085 <li><tt>oeq</tt>: ordered and equal</li>
4086 <li><tt>ogt</tt>: ordered and greater than </li>
4087 <li><tt>oge</tt>: ordered and greater than or equal</li>
4088 <li><tt>olt</tt>: ordered and less than </li>
4089 <li><tt>ole</tt>: ordered and less than or equal</li>
4090 <li><tt>one</tt>: ordered and not equal</li>
4091 <li><tt>ord</tt>: ordered (no nans)</li>
4092 <li><tt>ueq</tt>: unordered or equal</li>
4093 <li><tt>ugt</tt>: unordered or greater than </li>
4094 <li><tt>uge</tt>: unordered or greater than or equal</li>
4095 <li><tt>ult</tt>: unordered or less than </li>
4096 <li><tt>ule</tt>: unordered or less than or equal</li>
4097 <li><tt>une</tt>: unordered or not equal</li>
4098 <li><tt>uno</tt>: unordered (either nans)</li>
4099 <li><tt>true</tt>: no comparison, always returns true</li>
4101 <p><i>Ordered</i> means that neither operand is a QNAN while
4102 <i>unordered</i> means that either operand may be a QNAN.</p>
4103 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
4104 either a <a href="#t_floating">floating point</a> type
4105 or a <a href="#t_vector">vector</a> of floating point type.
4106 They must have identical types.</p>
4108 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4109 according to the condition code given as <tt>cond</tt>.
4110 If the operands are vectors, then the vectors are compared
4112 Each comparison performed
4113 always yields an <a href="#t_primitive">i1</a> result, as follows:</p>
4115 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4116 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4117 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4118 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4119 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
4120 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4121 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4122 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4123 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4124 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4125 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4126 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4127 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4128 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4129 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
4130 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4131 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
4132 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4133 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
4134 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4135 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
4136 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4137 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
4138 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4139 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
4140 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4141 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4142 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4146 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
4147 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4148 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4149 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
4153 <!-- _______________________________________________________________________ -->
4154 <div class="doc_subsubsection">
4155 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
4157 <div class="doc_text">
4159 <pre> <result> = vicmp <cond> <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4162 <p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
4163 element-wise comparison of its two integer vector operands.</p>
4165 <p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
4166 the condition code indicating the kind of comparison to perform. It is not
4167 a value, just a keyword. The possible condition code are:</p>
4169 <li><tt>eq</tt>: equal</li>
4170 <li><tt>ne</tt>: not equal </li>
4171 <li><tt>ugt</tt>: unsigned greater than</li>
4172 <li><tt>uge</tt>: unsigned greater or equal</li>
4173 <li><tt>ult</tt>: unsigned less than</li>
4174 <li><tt>ule</tt>: unsigned less or equal</li>
4175 <li><tt>sgt</tt>: signed greater than</li>
4176 <li><tt>sge</tt>: signed greater or equal</li>
4177 <li><tt>slt</tt>: signed less than</li>
4178 <li><tt>sle</tt>: signed less or equal</li>
4180 <p>The remaining two arguments must be <a href="#t_vector">vector</a> or
4181 <a href="#t_integer">integer</a> typed. They must also be identical types.</p>
4183 <p>The '<tt>vicmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4184 according to the condition code given as <tt>cond</tt>. The comparison yields a
4185 <a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
4186 identical type as the values being compared. The most significant bit in each
4187 element is 1 if the element-wise comparison evaluates to true, and is 0
4188 otherwise. All other bits of the result are undefined. The condition codes
4189 are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
4190 instruction</a>.</p>
4194 <result> = vicmp eq <2 x i32> < i32 4, i32 0>, < i32 5, i32 0> <i>; yields: result=<2 x i32> < i32 0, i32 -1 ></i>
4195 <result> = vicmp ult <2 x i8 > < i8 1, i8 2>, < i8 2, i8 2 > <i>; yields: result=<2 x i8> < i8 -1, i8 0 ></i>
4199 <!-- _______________________________________________________________________ -->
4200 <div class="doc_subsubsection">
4201 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4203 <div class="doc_text">
4205 <pre> <result> = vfcmp <cond> <ty> <op1>, <op2></pre>
4207 <p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4208 element-wise comparison of its two floating point vector operands. The output
4209 elements have the same width as the input elements.</p>
4211 <p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4212 the condition code indicating the kind of comparison to perform. It is not
4213 a value, just a keyword. The possible condition code are:</p>
4215 <li><tt>false</tt>: no comparison, always returns false</li>
4216 <li><tt>oeq</tt>: ordered and equal</li>
4217 <li><tt>ogt</tt>: ordered and greater than </li>
4218 <li><tt>oge</tt>: ordered and greater than or equal</li>
4219 <li><tt>olt</tt>: ordered and less than </li>
4220 <li><tt>ole</tt>: ordered and less than or equal</li>
4221 <li><tt>one</tt>: ordered and not equal</li>
4222 <li><tt>ord</tt>: ordered (no nans)</li>
4223 <li><tt>ueq</tt>: unordered or equal</li>
4224 <li><tt>ugt</tt>: unordered or greater than </li>
4225 <li><tt>uge</tt>: unordered or greater than or equal</li>
4226 <li><tt>ult</tt>: unordered or less than </li>
4227 <li><tt>ule</tt>: unordered or less than or equal</li>
4228 <li><tt>une</tt>: unordered or not equal</li>
4229 <li><tt>uno</tt>: unordered (either nans)</li>
4230 <li><tt>true</tt>: no comparison, always returns true</li>
4232 <p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4233 <a href="#t_floating">floating point</a> typed. They must also be identical
4236 <p>The '<tt>vfcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4237 according to the condition code given as <tt>cond</tt>. The comparison yields a
4238 <a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4239 an identical number of elements as the values being compared, and each element
4240 having identical with to the width of the floating point elements. The most
4241 significant bit in each element is 1 if the element-wise comparison evaluates to
4242 true, and is 0 otherwise. All other bits of the result are undefined. The
4243 condition codes are evaluated identically to the
4244 <a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.</p>
4248 <i>; yields: result=<2 x i32> < i32 0, i32 -1 ></i>
4249 <result> = vfcmp oeq <2 x float> < float 4, float 0 >, < float 5, float 0 >
4251 <i>; yields: result=<2 x i64> < i64 -1, i64 0 ></i>
4252 <result> = vfcmp ult <2 x double> < double 1, double 2 >, < double 2, double 2>
4256 <!-- _______________________________________________________________________ -->
4257 <div class="doc_subsubsection">
4258 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4261 <div class="doc_text">
4265 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
4267 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
4268 the SSA graph representing the function.</p>
4271 <p>The type of the incoming values is specified with the first type
4272 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4273 as arguments, with one pair for each predecessor basic block of the
4274 current block. Only values of <a href="#t_firstclass">first class</a>
4275 type may be used as the value arguments to the PHI node. Only labels
4276 may be used as the label arguments.</p>
4278 <p>There must be no non-phi instructions between the start of a basic
4279 block and the PHI instructions: i.e. PHI instructions must be first in
4284 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4285 specified by the pair corresponding to the predecessor basic block that executed
4286 just prior to the current block.</p>
4290 Loop: ; Infinite loop that counts from 0 on up...
4291 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4292 %nextindvar = add i32 %indvar, 1
4297 <!-- _______________________________________________________________________ -->
4298 <div class="doc_subsubsection">
4299 <a name="i_select">'<tt>select</tt>' Instruction</a>
4302 <div class="doc_text">
4307 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
4309 <i>selty</i> is either i1 or {<N x i1>}
4315 The '<tt>select</tt>' instruction is used to choose one value based on a
4316 condition, without branching.
4323 The '<tt>select</tt>' instruction requires an 'i1' value or
4324 a vector of 'i1' values indicating the
4325 condition, and two values of the same <a href="#t_firstclass">first class</a>
4326 type. If the val1/val2 are vectors and
4327 the condition is a scalar, then entire vectors are selected, not
4328 individual elements.
4334 If the condition is an i1 and it evaluates to 1, the instruction returns the first
4335 value argument; otherwise, it returns the second value argument.
4338 If the condition is a vector of i1, then the value arguments must
4339 be vectors of the same size, and the selection is done element
4346 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4351 <!-- _______________________________________________________________________ -->
4352 <div class="doc_subsubsection">
4353 <a name="i_call">'<tt>call</tt>' Instruction</a>
4356 <div class="doc_text">
4360 <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>]
4365 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4369 <p>This instruction requires several arguments:</p>
4373 <p>The optional "tail" marker indicates whether the callee function accesses
4374 any allocas or varargs in the caller. If the "tail" marker is present, the
4375 function call is eligible for tail call optimization. Note that calls may
4376 be marked "tail" even if they do not occur before a <a
4377 href="#i_ret"><tt>ret</tt></a> instruction.</p>
4380 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4381 convention</a> the call should use. If none is specified, the call defaults
4382 to using C calling conventions.</p>
4386 <p>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4387 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
4388 and '<tt>inreg</tt>' attributes are valid here.</p>
4392 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4393 the type of the return value. Functions that return no value are marked
4394 <tt><a href="#t_void">void</a></tt>.</p>
4397 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4398 value being invoked. The argument types must match the types implied by
4399 this signature. This type can be omitted if the function is not varargs
4400 and if the function type does not return a pointer to a function.</p>
4403 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4404 be invoked. In most cases, this is a direct function invocation, but
4405 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4406 to function value.</p>
4409 <p>'<tt>function args</tt>': argument list whose types match the
4410 function signature argument types. All arguments must be of
4411 <a href="#t_firstclass">first class</a> type. If the function signature
4412 indicates the function accepts a variable number of arguments, the extra
4413 arguments can be specified.</p>
4416 <p>The optional <a href="#fnattrs">function attributes</a> list. Only
4417 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
4418 '<tt>readnone</tt>' attributes are valid here.</p>
4424 <p>The '<tt>call</tt>' instruction is used to cause control flow to
4425 transfer to a specified function, with its incoming arguments bound to
4426 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4427 instruction in the called function, control flow continues with the
4428 instruction after the function call, and the return value of the
4429 function is bound to the result argument.</p>
4434 %retval = call i32 @test(i32 %argc)
4435 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4436 %X = tail call i32 @foo() <i>; yields i32</i>
4437 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4438 call void %foo(i8 97 signext)
4440 %struct.A = type { i32, i8 }
4441 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
4442 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4443 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
4444 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
4445 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
4450 <!-- _______________________________________________________________________ -->
4451 <div class="doc_subsubsection">
4452 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4455 <div class="doc_text">
4460 <resultval> = va_arg <va_list*> <arglist>, <argty>
4465 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4466 the "variable argument" area of a function call. It is used to implement the
4467 <tt>va_arg</tt> macro in C.</p>
4471 <p>This instruction takes a <tt>va_list*</tt> value and the type of
4472 the argument. It returns a value of the specified argument type and
4473 increments the <tt>va_list</tt> to point to the next argument. The
4474 actual type of <tt>va_list</tt> is target specific.</p>
4478 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4479 type from the specified <tt>va_list</tt> and causes the
4480 <tt>va_list</tt> to point to the next argument. For more information,
4481 see the variable argument handling <a href="#int_varargs">Intrinsic
4484 <p>It is legal for this instruction to be called in a function which does not
4485 take a variable number of arguments, for example, the <tt>vfprintf</tt>
4488 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4489 href="#intrinsics">intrinsic function</a> because it takes a type as an
4494 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4498 <!-- *********************************************************************** -->
4499 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4500 <!-- *********************************************************************** -->
4502 <div class="doc_text">
4504 <p>LLVM supports the notion of an "intrinsic function". These functions have
4505 well known names and semantics and are required to follow certain restrictions.
4506 Overall, these intrinsics represent an extension mechanism for the LLVM
4507 language that does not require changing all of the transformations in LLVM when
4508 adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4510 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4511 prefix is reserved in LLVM for intrinsic names; thus, function names may not
4512 begin with this prefix. Intrinsic functions must always be external functions:
4513 you cannot define the body of intrinsic functions. Intrinsic functions may
4514 only be used in call or invoke instructions: it is illegal to take the address
4515 of an intrinsic function. Additionally, because intrinsic functions are part
4516 of the LLVM language, it is required if any are added that they be documented
4519 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4520 a family of functions that perform the same operation but on different data
4521 types. Because LLVM can represent over 8 million different integer types,
4522 overloading is used commonly to allow an intrinsic function to operate on any
4523 integer type. One or more of the argument types or the result type can be
4524 overloaded to accept any integer type. Argument types may also be defined as
4525 exactly matching a previous argument's type or the result type. This allows an
4526 intrinsic function which accepts multiple arguments, but needs all of them to
4527 be of the same type, to only be overloaded with respect to a single argument or
4530 <p>Overloaded intrinsics will have the names of its overloaded argument types
4531 encoded into its function name, each preceded by a period. Only those types
4532 which are overloaded result in a name suffix. Arguments whose type is matched
4533 against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4534 take an integer of any width and returns an integer of exactly the same integer
4535 width. This leads to a family of functions such as
4536 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4537 Only one type, the return type, is overloaded, and only one type suffix is
4538 required. Because the argument's type is matched against the return type, it
4539 does not require its own name suffix.</p>
4541 <p>To learn how to add an intrinsic function, please see the
4542 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4547 <!-- ======================================================================= -->
4548 <div class="doc_subsection">
4549 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4552 <div class="doc_text">
4554 <p>Variable argument support is defined in LLVM with the <a
4555 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4556 intrinsic functions. These functions are related to the similarly
4557 named macros defined in the <tt><stdarg.h></tt> header file.</p>
4559 <p>All of these functions operate on arguments that use a
4560 target-specific value type "<tt>va_list</tt>". The LLVM assembly
4561 language reference manual does not define what this type is, so all
4562 transformations should be prepared to handle these functions regardless of
4565 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4566 instruction and the variable argument handling intrinsic functions are
4569 <div class="doc_code">
4571 define i32 @test(i32 %X, ...) {
4572 ; Initialize variable argument processing
4574 %ap2 = bitcast i8** %ap to i8*
4575 call void @llvm.va_start(i8* %ap2)
4577 ; Read a single integer argument
4578 %tmp = va_arg i8** %ap, i32
4580 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4582 %aq2 = bitcast i8** %aq to i8*
4583 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4584 call void @llvm.va_end(i8* %aq2)
4586 ; Stop processing of arguments.
4587 call void @llvm.va_end(i8* %ap2)
4591 declare void @llvm.va_start(i8*)
4592 declare void @llvm.va_copy(i8*, i8*)
4593 declare void @llvm.va_end(i8*)
4599 <!-- _______________________________________________________________________ -->
4600 <div class="doc_subsubsection">
4601 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4605 <div class="doc_text">
4607 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
4609 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes
4610 <tt>*<arglist></tt> for subsequent use by <tt><a
4611 href="#i_va_arg">va_arg</a></tt>.</p>
4615 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
4619 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
4620 macro available in C. In a target-dependent way, it initializes the
4621 <tt>va_list</tt> element to which the argument points, so that the next call to
4622 <tt>va_arg</tt> will produce the first variable argument passed to the function.
4623 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4624 last argument of the function as the compiler can figure that out.</p>
4628 <!-- _______________________________________________________________________ -->
4629 <div class="doc_subsubsection">
4630 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4633 <div class="doc_text">
4635 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
4638 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
4639 which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4640 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4644 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4648 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4649 macro available in C. In a target-dependent way, it destroys the
4650 <tt>va_list</tt> element to which the argument points. Calls to <a
4651 href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4652 <tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4653 <tt>llvm.va_end</tt>.</p>
4657 <!-- _______________________________________________________________________ -->
4658 <div class="doc_subsubsection">
4659 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4662 <div class="doc_text">
4667 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
4672 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4673 from the source argument list to the destination argument list.</p>
4677 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4678 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4683 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4684 macro available in C. In a target-dependent way, it copies the source
4685 <tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4686 intrinsic is necessary because the <tt><a href="#int_va_start">
4687 llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4688 example, memory allocation.</p>
4692 <!-- ======================================================================= -->
4693 <div class="doc_subsection">
4694 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4697 <div class="doc_text">
4700 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
4701 Collection</a> (GC) requires the implementation and generation of these
4703 These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4704 stack</a>, as well as garbage collector implementations that require <a
4705 href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4706 Front-ends for type-safe garbage collected languages should generate these
4707 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4708 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4711 <p>The garbage collection intrinsics only operate on objects in the generic
4712 address space (address space zero).</p>
4716 <!-- _______________________________________________________________________ -->
4717 <div class="doc_subsubsection">
4718 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4721 <div class="doc_text">
4726 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
4731 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4732 the code generator, and allows some metadata to be associated with it.</p>
4736 <p>The first argument specifies the address of a stack object that contains the
4737 root pointer. The second pointer (which must be either a constant or a global
4738 value address) contains the meta-data to be associated with the root.</p>
4742 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
4743 location. At compile-time, the code generator generates information to allow
4744 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4745 intrinsic may only be used in a function which <a href="#gc">specifies a GC
4751 <!-- _______________________________________________________________________ -->
4752 <div class="doc_subsubsection">
4753 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4756 <div class="doc_text">
4761 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
4766 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4767 locations, allowing garbage collector implementations that require read
4772 <p>The second argument is the address to read from, which should be an address
4773 allocated from the garbage collector. The first object is a pointer to the
4774 start of the referenced object, if needed by the language runtime (otherwise
4779 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4780 instruction, but may be replaced with substantially more complex code by the
4781 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4782 may only be used in a function which <a href="#gc">specifies a GC
4788 <!-- _______________________________________________________________________ -->
4789 <div class="doc_subsubsection">
4790 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4793 <div class="doc_text">
4798 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
4803 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4804 locations, allowing garbage collector implementations that require write
4805 barriers (such as generational or reference counting collectors).</p>
4809 <p>The first argument is the reference to store, the second is the start of the
4810 object to store it to, and the third is the address of the field of Obj to
4811 store to. If the runtime does not require a pointer to the object, Obj may be
4816 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4817 instruction, but may be replaced with substantially more complex code by the
4818 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4819 may only be used in a function which <a href="#gc">specifies a GC
4826 <!-- ======================================================================= -->
4827 <div class="doc_subsection">
4828 <a name="int_codegen">Code Generator Intrinsics</a>
4831 <div class="doc_text">
4833 These intrinsics are provided by LLVM to expose special features that may only
4834 be implemented with code generator support.
4839 <!-- _______________________________________________________________________ -->
4840 <div class="doc_subsubsection">
4841 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4844 <div class="doc_text">
4848 declare i8 *@llvm.returnaddress(i32 <level>)
4854 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4855 target-specific value indicating the return address of the current function
4856 or one of its callers.
4862 The argument to this intrinsic indicates which function to return the address
4863 for. Zero indicates the calling function, one indicates its caller, etc. The
4864 argument is <b>required</b> to be a constant integer value.
4870 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4871 the return address of the specified call frame, or zero if it cannot be
4872 identified. The value returned by this intrinsic is likely to be incorrect or 0
4873 for arguments other than zero, so it should only be used for debugging purposes.
4877 Note that calling this intrinsic does not prevent function inlining or other
4878 aggressive transformations, so the value returned may not be that of the obvious
4879 source-language caller.
4884 <!-- _______________________________________________________________________ -->
4885 <div class="doc_subsubsection">
4886 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4889 <div class="doc_text">
4893 declare i8 *@llvm.frameaddress(i32 <level>)
4899 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4900 target-specific frame pointer value for the specified stack frame.
4906 The argument to this intrinsic indicates which function to return the frame
4907 pointer for. Zero indicates the calling function, one indicates its caller,
4908 etc. The argument is <b>required</b> to be a constant integer value.
4914 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4915 the frame address of the specified call frame, or zero if it cannot be
4916 identified. The value returned by this intrinsic is likely to be incorrect or 0
4917 for arguments other than zero, so it should only be used for debugging purposes.
4921 Note that calling this intrinsic does not prevent function inlining or other
4922 aggressive transformations, so the value returned may not be that of the obvious
4923 source-language caller.
4927 <!-- _______________________________________________________________________ -->
4928 <div class="doc_subsubsection">
4929 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4932 <div class="doc_text">
4936 declare i8 *@llvm.stacksave()
4942 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4943 the function stack, for use with <a href="#int_stackrestore">
4944 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4945 features like scoped automatic variable sized arrays in C99.
4951 This intrinsic returns a opaque pointer value that can be passed to <a
4952 href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4953 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4954 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4955 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4956 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4957 that were allocated after the <tt>llvm.stacksave</tt> was executed.
4962 <!-- _______________________________________________________________________ -->
4963 <div class="doc_subsubsection">
4964 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4967 <div class="doc_text">
4971 declare void @llvm.stackrestore(i8 * %ptr)
4977 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4978 the function stack to the state it was in when the corresponding <a
4979 href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4980 useful for implementing language features like scoped automatic variable sized
4987 See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4993 <!-- _______________________________________________________________________ -->
4994 <div class="doc_subsubsection">
4995 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4998 <div class="doc_text">
5002 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
5009 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
5010 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
5012 effect on the behavior of the program but can change its performance
5019 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
5020 determining if the fetch should be for a read (0) or write (1), and
5021 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
5022 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
5023 <tt>locality</tt> arguments must be constant integers.
5029 This intrinsic does not modify the behavior of the program. In particular,
5030 prefetches cannot trap and do not produce a value. On targets that support this
5031 intrinsic, the prefetch can provide hints to the processor cache for better
5037 <!-- _______________________________________________________________________ -->
5038 <div class="doc_subsubsection">
5039 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5042 <div class="doc_text">
5046 declare void @llvm.pcmarker(i32 <id>)
5053 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
5055 code to simulators and other tools. The method is target specific, but it is
5056 expected that the marker will use exported symbols to transmit the PC of the
5058 The marker makes no guarantees that it will remain with any specific instruction
5059 after optimizations. It is possible that the presence of a marker will inhibit
5060 optimizations. The intended use is to be inserted after optimizations to allow
5061 correlations of simulation runs.
5067 <tt>id</tt> is a numerical id identifying the marker.
5073 This intrinsic does not modify the behavior of the program. Backends that do not
5074 support this intrinisic may ignore it.
5079 <!-- _______________________________________________________________________ -->
5080 <div class="doc_subsubsection">
5081 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5084 <div class="doc_text">
5088 declare i64 @llvm.readcyclecounter( )
5095 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5096 counter register (or similar low latency, high accuracy clocks) on those targets
5097 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
5098 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
5099 should only be used for small timings.
5105 When directly supported, reading the cycle counter should not modify any memory.
5106 Implementations are allowed to either return a application specific value or a
5107 system wide value. On backends without support, this is lowered to a constant 0.
5112 <!-- ======================================================================= -->
5113 <div class="doc_subsection">
5114 <a name="int_libc">Standard C Library Intrinsics</a>
5117 <div class="doc_text">
5119 LLVM provides intrinsics for a few important standard C library functions.
5120 These intrinsics allow source-language front-ends to pass information about the
5121 alignment of the pointer arguments to the code generator, providing opportunity
5122 for more efficient code generation.
5127 <!-- _______________________________________________________________________ -->
5128 <div class="doc_subsubsection">
5129 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5132 <div class="doc_text">
5135 <p>This is an overloaded intrinsic. You can use llvm.memcpy on any integer bit
5136 width. Not all targets support all bit widths however.</p>
5138 declare void @llvm.memcpy.i8(i8 * <dest>, i8 * <src>,
5139 i8 <len>, i32 <align>)
5140 declare void @llvm.memcpy.i16(i8 * <dest>, i8 * <src>,
5141 i16 <len>, i32 <align>)
5142 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
5143 i32 <len>, i32 <align>)
5144 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
5145 i64 <len>, i32 <align>)
5151 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5152 location to the destination location.
5156 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5157 intrinsics do not return a value, and takes an extra alignment argument.
5163 The first argument is a pointer to the destination, the second is a pointer to
5164 the source. The third argument is an integer argument
5165 specifying the number of bytes to copy, and the fourth argument is the alignment
5166 of the source and destination locations.
5170 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5171 the caller guarantees that both the source and destination pointers are aligned
5178 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5179 location to the destination location, which are not allowed to overlap. It
5180 copies "len" bytes of memory over. If the argument is known to be aligned to
5181 some boundary, this can be specified as the fourth argument, otherwise it should
5187 <!-- _______________________________________________________________________ -->
5188 <div class="doc_subsubsection">
5189 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5192 <div class="doc_text">
5195 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5196 width. Not all targets support all bit widths however.</p>
5198 declare void @llvm.memmove.i8(i8 * <dest>, i8 * <src>,
5199 i8 <len>, i32 <align>)
5200 declare void @llvm.memmove.i16(i8 * <dest>, i8 * <src>,
5201 i16 <len>, i32 <align>)
5202 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
5203 i32 <len>, i32 <align>)
5204 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
5205 i64 <len>, i32 <align>)
5211 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5212 location to the destination location. It is similar to the
5213 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
5217 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5218 intrinsics do not return a value, and takes an extra alignment argument.
5224 The first argument is a pointer to the destination, the second is a pointer to
5225 the source. The third argument is an integer argument
5226 specifying the number of bytes to copy, and the fourth argument is the alignment
5227 of the source and destination locations.
5231 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5232 the caller guarantees that the source and destination pointers are aligned to
5239 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5240 location to the destination location, which may overlap. It
5241 copies "len" bytes of memory over. If the argument is known to be aligned to
5242 some boundary, this can be specified as the fourth argument, otherwise it should
5248 <!-- _______________________________________________________________________ -->
5249 <div class="doc_subsubsection">
5250 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5253 <div class="doc_text">
5256 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5257 width. Not all targets support all bit widths however.</p>
5259 declare void @llvm.memset.i8(i8 * <dest>, i8 <val>,
5260 i8 <len>, i32 <align>)
5261 declare void @llvm.memset.i16(i8 * <dest>, i8 <val>,
5262 i16 <len>, i32 <align>)
5263 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
5264 i32 <len>, i32 <align>)
5265 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
5266 i64 <len>, i32 <align>)
5272 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5277 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5278 does not return a value, and takes an extra alignment argument.
5284 The first argument is a pointer to the destination to fill, the second is the
5285 byte value to fill it with, the third argument is an integer
5286 argument specifying the number of bytes to fill, and the fourth argument is the
5287 known alignment of destination location.
5291 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5292 the caller guarantees that the destination pointer is aligned to that boundary.
5298 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5300 destination location. If the argument is known to be aligned to some boundary,
5301 this can be specified as the fourth argument, otherwise it should be set to 0 or
5307 <!-- _______________________________________________________________________ -->
5308 <div class="doc_subsubsection">
5309 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5312 <div class="doc_text">
5315 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
5316 floating point or vector of floating point type. Not all targets support all
5319 declare float @llvm.sqrt.f32(float %Val)
5320 declare double @llvm.sqrt.f64(double %Val)
5321 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5322 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5323 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
5329 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
5330 returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
5331 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
5332 negative numbers other than -0.0 (which allows for better optimization, because
5333 there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5334 defined to return -0.0 like IEEE sqrt.
5340 The argument and return value are floating point numbers of the same type.
5346 This function returns the sqrt of the specified operand if it is a nonnegative
5347 floating point number.
5351 <!-- _______________________________________________________________________ -->
5352 <div class="doc_subsubsection">
5353 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5356 <div class="doc_text">
5359 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
5360 floating point or vector of floating point type. Not all targets support all
5363 declare float @llvm.powi.f32(float %Val, i32 %power)
5364 declare double @llvm.powi.f64(double %Val, i32 %power)
5365 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5366 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5367 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
5373 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5374 specified (positive or negative) power. The order of evaluation of
5375 multiplications is not defined. When a vector of floating point type is
5376 used, the second argument remains a scalar integer value.
5382 The second argument is an integer power, and the first is a value to raise to
5389 This function returns the first value raised to the second power with an
5390 unspecified sequence of rounding operations.</p>
5393 <!-- _______________________________________________________________________ -->
5394 <div class="doc_subsubsection">
5395 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5398 <div class="doc_text">
5401 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5402 floating point or vector of floating point type. Not all targets support all
5405 declare float @llvm.sin.f32(float %Val)
5406 declare double @llvm.sin.f64(double %Val)
5407 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5408 declare fp128 @llvm.sin.f128(fp128 %Val)
5409 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5415 The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5421 The argument and return value are floating point numbers of the same type.
5427 This function returns the sine of the specified operand, returning the
5428 same values as the libm <tt>sin</tt> functions would, and handles error
5429 conditions in the same way.</p>
5432 <!-- _______________________________________________________________________ -->
5433 <div class="doc_subsubsection">
5434 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5437 <div class="doc_text">
5440 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5441 floating point or vector of floating point type. Not all targets support all
5444 declare float @llvm.cos.f32(float %Val)
5445 declare double @llvm.cos.f64(double %Val)
5446 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5447 declare fp128 @llvm.cos.f128(fp128 %Val)
5448 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5454 The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5460 The argument and return value are floating point numbers of the same type.
5466 This function returns the cosine of the specified operand, returning the
5467 same values as the libm <tt>cos</tt> functions would, and handles error
5468 conditions in the same way.</p>
5471 <!-- _______________________________________________________________________ -->
5472 <div class="doc_subsubsection">
5473 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5476 <div class="doc_text">
5479 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5480 floating point or vector of floating point type. Not all targets support all
5483 declare float @llvm.pow.f32(float %Val, float %Power)
5484 declare double @llvm.pow.f64(double %Val, double %Power)
5485 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5486 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5487 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5493 The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5494 specified (positive or negative) power.
5500 The second argument is a floating point power, and the first is a value to
5501 raise to that power.
5507 This function returns the first value raised to the second power,
5509 same values as the libm <tt>pow</tt> functions would, and handles error
5510 conditions in the same way.</p>
5514 <!-- ======================================================================= -->
5515 <div class="doc_subsection">
5516 <a name="int_manip">Bit Manipulation Intrinsics</a>
5519 <div class="doc_text">
5521 LLVM provides intrinsics for a few important bit manipulation operations.
5522 These allow efficient code generation for some algorithms.
5527 <!-- _______________________________________________________________________ -->
5528 <div class="doc_subsubsection">
5529 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5532 <div class="doc_text">
5535 <p>This is an overloaded intrinsic function. You can use bswap on any integer
5536 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
5538 declare i16 @llvm.bswap.i16(i16 <id>)
5539 declare i32 @llvm.bswap.i32(i32 <id>)
5540 declare i64 @llvm.bswap.i64(i64 <id>)
5546 The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5547 values with an even number of bytes (positive multiple of 16 bits). These are
5548 useful for performing operations on data that is not in the target's native
5555 The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
5556 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5557 intrinsic returns an i32 value that has the four bytes of the input i32
5558 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
5559 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5560 <tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
5561 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5566 <!-- _______________________________________________________________________ -->
5567 <div class="doc_subsubsection">
5568 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5571 <div class="doc_text">
5574 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
5575 width. Not all targets support all bit widths however.</p>
5577 declare i8 @llvm.ctpop.i8 (i8 <src>)
5578 declare i16 @llvm.ctpop.i16(i16 <src>)
5579 declare i32 @llvm.ctpop.i32(i32 <src>)
5580 declare i64 @llvm.ctpop.i64(i64 <src>)
5581 declare i256 @llvm.ctpop.i256(i256 <src>)
5587 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5594 The only argument is the value to be counted. The argument may be of any
5595 integer type. The return type must match the argument type.
5601 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5605 <!-- _______________________________________________________________________ -->
5606 <div class="doc_subsubsection">
5607 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5610 <div class="doc_text">
5613 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5614 integer bit width. Not all targets support all bit widths however.</p>
5616 declare i8 @llvm.ctlz.i8 (i8 <src>)
5617 declare i16 @llvm.ctlz.i16(i16 <src>)
5618 declare i32 @llvm.ctlz.i32(i32 <src>)
5619 declare i64 @llvm.ctlz.i64(i64 <src>)
5620 declare i256 @llvm.ctlz.i256(i256 <src>)
5626 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5627 leading zeros in a variable.
5633 The only argument is the value to be counted. The argument may be of any
5634 integer type. The return type must match the argument type.
5640 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5641 in a variable. If the src == 0 then the result is the size in bits of the type
5642 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5648 <!-- _______________________________________________________________________ -->
5649 <div class="doc_subsubsection">
5650 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5653 <div class="doc_text">
5656 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5657 integer bit width. Not all targets support all bit widths however.</p>
5659 declare i8 @llvm.cttz.i8 (i8 <src>)
5660 declare i16 @llvm.cttz.i16(i16 <src>)
5661 declare i32 @llvm.cttz.i32(i32 <src>)
5662 declare i64 @llvm.cttz.i64(i64 <src>)
5663 declare i256 @llvm.cttz.i256(i256 <src>)
5669 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5676 The only argument is the value to be counted. The argument may be of any
5677 integer type. The return type must match the argument type.
5683 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5684 in a variable. If the src == 0 then the result is the size in bits of the type
5685 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5689 <!-- _______________________________________________________________________ -->
5690 <div class="doc_subsubsection">
5691 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5694 <div class="doc_text">
5697 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
5698 on any integer bit width.</p>
5700 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5701 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
5705 <p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5706 range of bits from an integer value and returns them in the same bit width as
5707 the original value.</p>
5710 <p>The first argument, <tt>%val</tt> and the result may be integer types of
5711 any bit width but they must have the same bit width. The second and third
5712 arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5715 <p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5716 of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5717 <tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5718 operates in forward mode.</p>
5719 <p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5720 right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5721 only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5723 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5724 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5725 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5726 to determine the number of bits to retain.</li>
5727 <li>A mask of the retained bits is created by shifting a -1 value.</li>
5728 <li>The mask is ANDed with <tt>%val</tt> to produce the result.</li>
5730 <p>In reverse mode, a similar computation is made except that the bits are
5731 returned in the reverse order. So, for example, if <tt>X</tt> has the value
5732 <tt>i16 0x0ACF (101011001111)</tt> and we apply
5733 <tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5734 <tt>i16 0x0026 (000000100110)</tt>.</p>
5737 <div class="doc_subsubsection">
5738 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5741 <div class="doc_text">
5744 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
5745 on any integer bit width.</p>
5747 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5748 declare i29 @llvm.part.set.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
5752 <p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5753 of bits in an integer value with another integer value. It returns the integer
5754 with the replaced bits.</p>
5757 <p>The first argument, <tt>%val</tt> and the result may be integer types of
5758 any bit width but they must have the same bit width. <tt>%val</tt> is the value
5759 whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5760 integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5761 type since they specify only a bit index.</p>
5764 <p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5765 of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5766 <tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5767 operates in forward mode.</p>
5768 <p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5769 truncating it down to the size of the replacement area or zero extending it
5770 up to that size.</p>
5771 <p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5772 are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5773 in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5774 to the <tt>%hi</tt>th bit.</p>
5775 <p>In reverse mode, a similar computation is made except that the bits are
5776 reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5777 <tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.</p>
5780 llvm.part.set(0xFFFF, 0, 4, 7) -> 0xFF0F
5781 llvm.part.set(0xFFFF, 0, 7, 4) -> 0xFF0F
5782 llvm.part.set(0xFFFF, 1, 7, 4) -> 0xFF8F
5783 llvm.part.set(0xFFFF, F, 8, 3) -> 0xFFE7
5784 llvm.part.set(0xFFFF, 0, 3, 8) -> 0xFE07
5788 <!-- ======================================================================= -->
5789 <div class="doc_subsection">
5790 <a name="int_debugger">Debugger Intrinsics</a>
5793 <div class="doc_text">
5795 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5796 are described in the <a
5797 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5798 Debugging</a> document.
5803 <!-- ======================================================================= -->
5804 <div class="doc_subsection">
5805 <a name="int_eh">Exception Handling Intrinsics</a>
5808 <div class="doc_text">
5809 <p> The LLVM exception handling intrinsics (which all start with
5810 <tt>llvm.eh.</tt> prefix), are described in the <a
5811 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5812 Handling</a> document. </p>
5815 <!-- ======================================================================= -->
5816 <div class="doc_subsection">
5817 <a name="int_trampoline">Trampoline Intrinsic</a>
5820 <div class="doc_text">
5822 This intrinsic makes it possible to excise one parameter, marked with
5823 the <tt>nest</tt> attribute, from a function. The result is a callable
5824 function pointer lacking the nest parameter - the caller does not need
5825 to provide a value for it. Instead, the value to use is stored in
5826 advance in a "trampoline", a block of memory usually allocated
5827 on the stack, which also contains code to splice the nest value into the
5828 argument list. This is used to implement the GCC nested function address
5832 For example, if the function is
5833 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
5834 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:</p>
5836 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5837 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5838 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5839 %fp = bitcast i8* %p to i32 (i32, i32)*
5841 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5842 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
5845 <!-- _______________________________________________________________________ -->
5846 <div class="doc_subsubsection">
5847 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5849 <div class="doc_text">
5852 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
5856 This fills the memory pointed to by <tt>tramp</tt> with code
5857 and returns a function pointer suitable for executing it.
5861 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5862 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5863 and sufficiently aligned block of memory; this memory is written to by the
5864 intrinsic. Note that the size and the alignment are target-specific - LLVM
5865 currently provides no portable way of determining them, so a front-end that
5866 generates this intrinsic needs to have some target-specific knowledge.
5867 The <tt>func</tt> argument must hold a function bitcast to an <tt>i8*</tt>.
5871 The block of memory pointed to by <tt>tramp</tt> is filled with target
5872 dependent code, turning it into a function. A pointer to this function is
5873 returned, but needs to be bitcast to an
5874 <a href="#int_trampoline">appropriate function pointer type</a>
5875 before being called. The new function's signature is the same as that of
5876 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5877 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5878 of pointer type. Calling the new function is equivalent to calling
5879 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5880 missing <tt>nest</tt> argument. If, after calling
5881 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5882 modified, then the effect of any later call to the returned function pointer is
5887 <!-- ======================================================================= -->
5888 <div class="doc_subsection">
5889 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
5892 <div class="doc_text">
5894 These intrinsic functions expand the "universal IR" of LLVM to represent
5895 hardware constructs for atomic operations and memory synchronization. This
5896 provides an interface to the hardware, not an interface to the programmer. It
5897 is aimed at a low enough level to allow any programming models or APIs
5898 (Application Programming Interfaces) which
5899 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
5900 hardware behavior. Just as hardware provides a "universal IR" for source
5901 languages, it also provides a starting point for developing a "universal"
5902 atomic operation and synchronization IR.
5905 These do <em>not</em> form an API such as high-level threading libraries,
5906 software transaction memory systems, atomic primitives, and intrinsic
5907 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
5908 application libraries. The hardware interface provided by LLVM should allow
5909 a clean implementation of all of these APIs and parallel programming models.
5910 No one model or paradigm should be selected above others unless the hardware
5911 itself ubiquitously does so.
5916 <!-- _______________________________________________________________________ -->
5917 <div class="doc_subsubsection">
5918 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
5920 <div class="doc_text">
5923 declare void @llvm.memory.barrier( i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>,
5929 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
5930 specific pairs of memory access types.
5934 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
5935 The first four arguments enables a specific barrier as listed below. The fith
5936 argument specifies that the barrier applies to io or device or uncached memory.
5940 <li><tt>ll</tt>: load-load barrier</li>
5941 <li><tt>ls</tt>: load-store barrier</li>
5942 <li><tt>sl</tt>: store-load barrier</li>
5943 <li><tt>ss</tt>: store-store barrier</li>
5944 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
5948 This intrinsic causes the system to enforce some ordering constraints upon
5949 the loads and stores of the program. This barrier does not indicate
5950 <em>when</em> any events will occur, it only enforces an <em>order</em> in
5951 which they occur. For any of the specified pairs of load and store operations
5952 (f.ex. load-load, or store-load), all of the first operations preceding the
5953 barrier will complete before any of the second operations succeeding the
5954 barrier begin. Specifically the semantics for each pairing is as follows:
5957 <li><tt>ll</tt>: All loads before the barrier must complete before any load
5958 after the barrier begins.</li>
5960 <li><tt>ls</tt>: All loads before the barrier must complete before any
5961 store after the barrier begins.</li>
5962 <li><tt>ss</tt>: All stores before the barrier must complete before any
5963 store after the barrier begins.</li>
5964 <li><tt>sl</tt>: All stores before the barrier must complete before any
5965 load after the barrier begins.</li>
5968 These semantics are applied with a logical "and" behavior when more than one
5969 is enabled in a single memory barrier intrinsic.
5972 Backends may implement stronger barriers than those requested when they do not
5973 support as fine grained a barrier as requested. Some architectures do not
5974 need all types of barriers and on such architectures, these become noops.
5981 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
5982 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
5983 <i>; guarantee the above finishes</i>
5984 store i32 8, %ptr <i>; before this begins</i>
5988 <!-- _______________________________________________________________________ -->
5989 <div class="doc_subsubsection">
5990 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
5992 <div class="doc_text">
5995 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
5996 any integer bit width and for different address spaces. Not all targets
5997 support all bit widths however.</p>
6000 declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* <ptr>, i8 <cmp>, i8 <val> )
6001 declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* <ptr>, i16 <cmp>, i16 <val> )
6002 declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* <ptr>, i32 <cmp>, i32 <val> )
6003 declare i64 @llvm.atomic.cmp.swap.i64.p0i64( i64* <ptr>, i64 <cmp>, i64 <val> )
6008 This loads a value in memory and compares it to a given value. If they are
6009 equal, it stores a new value into the memory.
6013 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
6014 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
6015 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
6016 this integer type. While any bit width integer may be used, targets may only
6017 lower representations they support in hardware.
6022 This entire intrinsic must be executed atomically. It first loads the value
6023 in memory pointed to by <tt>ptr</tt> and compares it with the value
6024 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
6025 loaded value is yielded in all cases. This provides the equivalent of an
6026 atomic compare-and-swap operation within the SSA framework.
6034 %val1 = add i32 4, 4
6035 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
6036 <i>; yields {i32}:result1 = 4</i>
6037 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6038 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6040 %val2 = add i32 1, 1
6041 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
6042 <i>; yields {i32}:result2 = 8</i>
6043 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6045 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6049 <!-- _______________________________________________________________________ -->
6050 <div class="doc_subsubsection">
6051 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6053 <div class="doc_text">
6057 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6058 integer bit width. Not all targets support all bit widths however.</p>
6060 declare i8 @llvm.atomic.swap.i8.p0i8( i8* <ptr>, i8 <val> )
6061 declare i16 @llvm.atomic.swap.i16.p0i16( i16* <ptr>, i16 <val> )
6062 declare i32 @llvm.atomic.swap.i32.p0i32( i32* <ptr>, i32 <val> )
6063 declare i64 @llvm.atomic.swap.i64.p0i64( i64* <ptr>, i64 <val> )
6068 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6069 the value from memory. It then stores the value in <tt>val</tt> in the memory
6075 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
6076 <tt>val</tt> argument and the result must be integers of the same bit width.
6077 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6078 integer type. The targets may only lower integer representations they
6083 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6084 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6085 equivalent of an atomic swap operation within the SSA framework.
6093 %val1 = add i32 4, 4
6094 %result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
6095 <i>; yields {i32}:result1 = 4</i>
6096 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6097 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6099 %val2 = add i32 1, 1
6100 %result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
6101 <i>; yields {i32}:result2 = 8</i>
6103 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6104 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6108 <!-- _______________________________________________________________________ -->
6109 <div class="doc_subsubsection">
6110 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
6113 <div class="doc_text">
6116 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on any
6117 integer bit width. Not all targets support all bit widths however.</p>
6119 declare i8 @llvm.atomic.load.add.i8..p0i8( i8* <ptr>, i8 <delta> )
6120 declare i16 @llvm.atomic.load.add.i16..p0i16( i16* <ptr>, i16 <delta> )
6121 declare i32 @llvm.atomic.load.add.i32..p0i32( i32* <ptr>, i32 <delta> )
6122 declare i64 @llvm.atomic.load.add.i64..p0i64( i64* <ptr>, i64 <delta> )
6127 This intrinsic adds <tt>delta</tt> to the value stored in memory at
6128 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6133 The intrinsic takes two arguments, the first a pointer to an integer value
6134 and the second an integer value. The result is also an integer value. These
6135 integer types can have any bit width, but they must all have the same bit
6136 width. The targets may only lower integer representations they support.
6140 This intrinsic does a series of operations atomically. It first loads the
6141 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6142 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6149 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
6150 <i>; yields {i32}:result1 = 4</i>
6151 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
6152 <i>; yields {i32}:result2 = 8</i>
6153 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
6154 <i>; yields {i32}:result3 = 10</i>
6155 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
6159 <!-- _______________________________________________________________________ -->
6160 <div class="doc_subsubsection">
6161 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6164 <div class="doc_text">
6167 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
6168 any integer bit width and for different address spaces. Not all targets
6169 support all bit widths however.</p>
6171 declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* <ptr>, i8 <delta> )
6172 declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* <ptr>, i16 <delta> )
6173 declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* <ptr>, i32 <delta> )
6174 declare i64 @llvm.atomic.load.sub.i64.p0i32( i64* <ptr>, i64 <delta> )
6179 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6180 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6185 The intrinsic takes two arguments, the first a pointer to an integer value
6186 and the second an integer value. The result is also an integer value. These
6187 integer types can have any bit width, but they must all have the same bit
6188 width. The targets may only lower integer representations they support.
6192 This intrinsic does a series of operations atomically. It first loads the
6193 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6194 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6201 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
6202 <i>; yields {i32}:result1 = 8</i>
6203 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
6204 <i>; yields {i32}:result2 = 4</i>
6205 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
6206 <i>; yields {i32}:result3 = 2</i>
6207 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6211 <!-- _______________________________________________________________________ -->
6212 <div class="doc_subsubsection">
6213 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6214 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6215 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6216 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6219 <div class="doc_text">
6222 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6223 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
6224 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6225 address spaces. Not all targets support all bit widths however.</p>
6227 declare i8 @llvm.atomic.load.and.i8.p0i8( i8* <ptr>, i8 <delta> )
6228 declare i16 @llvm.atomic.load.and.i16.p0i16( i16* <ptr>, i16 <delta> )
6229 declare i32 @llvm.atomic.load.and.i32.p0i32( i32* <ptr>, i32 <delta> )
6230 declare i64 @llvm.atomic.load.and.i64.p0i64( i64* <ptr>, i64 <delta> )
6235 declare i8 @llvm.atomic.load.or.i8.p0i8( i8* <ptr>, i8 <delta> )
6236 declare i16 @llvm.atomic.load.or.i16.p0i16( i16* <ptr>, i16 <delta> )
6237 declare i32 @llvm.atomic.load.or.i32.p0i32( i32* <ptr>, i32 <delta> )
6238 declare i64 @llvm.atomic.load.or.i64.p0i64( i64* <ptr>, i64 <delta> )
6243 declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* <ptr>, i8 <delta> )
6244 declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* <ptr>, i16 <delta> )
6245 declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* <ptr>, i32 <delta> )
6246 declare i64 @llvm.atomic.load.nand.i64.p0i32( i64* <ptr>, i64 <delta> )
6251 declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* <ptr>, i8 <delta> )
6252 declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* <ptr>, i16 <delta> )
6253 declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* <ptr>, i32 <delta> )
6254 declare i64 @llvm.atomic.load.xor.i64.p0i32( i64* <ptr>, i64 <delta> )
6259 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6260 the value stored in memory at <tt>ptr</tt>. It yields the original value
6266 These intrinsics take two arguments, the first a pointer to an integer value
6267 and the second an integer value. The result is also an integer value. These
6268 integer types can have any bit width, but they must all have the same bit
6269 width. The targets may only lower integer representations they support.
6273 These intrinsics does a series of operations atomically. They first load the
6274 value stored at <tt>ptr</tt>. They then do the bitwise operation
6275 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6276 value stored at <tt>ptr</tt>.
6282 store i32 0x0F0F, %ptr
6283 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
6284 <i>; yields {i32}:result0 = 0x0F0F</i>
6285 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
6286 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
6287 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
6288 <i>; yields {i32}:result2 = 0xF0</i>
6289 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
6290 <i>; yields {i32}:result3 = FF</i>
6291 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6296 <!-- _______________________________________________________________________ -->
6297 <div class="doc_subsubsection">
6298 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6299 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6300 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6301 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6304 <div class="doc_text">
6307 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6308 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
6309 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6310 address spaces. Not all targets
6311 support all bit widths however.</p>
6313 declare i8 @llvm.atomic.load.max.i8.p0i8( i8* <ptr>, i8 <delta> )
6314 declare i16 @llvm.atomic.load.max.i16.p0i16( i16* <ptr>, i16 <delta> )
6315 declare i32 @llvm.atomic.load.max.i32.p0i32( i32* <ptr>, i32 <delta> )
6316 declare i64 @llvm.atomic.load.max.i64.p0i64( i64* <ptr>, i64 <delta> )
6321 declare i8 @llvm.atomic.load.min.i8.p0i8( i8* <ptr>, i8 <delta> )
6322 declare i16 @llvm.atomic.load.min.i16.p0i16( i16* <ptr>, i16 <delta> )
6323 declare i32 @llvm.atomic.load.min.i32..p0i32( i32* <ptr>, i32 <delta> )
6324 declare i64 @llvm.atomic.load.min.i64..p0i64( i64* <ptr>, i64 <delta> )
6329 declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* <ptr>, i8 <delta> )
6330 declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* <ptr>, i16 <delta> )
6331 declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* <ptr>, i32 <delta> )
6332 declare i64 @llvm.atomic.load.umax.i64.p0i64( i64* <ptr>, i64 <delta> )
6337 declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* <ptr>, i8 <delta> )
6338 declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* <ptr>, i16 <delta> )
6339 declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* <ptr>, i32 <delta> )
6340 declare i64 @llvm.atomic.load.umin.i64..p0i64( i64* <ptr>, i64 <delta> )
6345 These intrinsics takes the signed or unsigned minimum or maximum of
6346 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6347 original value at <tt>ptr</tt>.
6352 These intrinsics take two arguments, the first a pointer to an integer value
6353 and the second an integer value. The result is also an integer value. These
6354 integer types can have any bit width, but they must all have the same bit
6355 width. The targets may only lower integer representations they support.
6359 These intrinsics does a series of operations atomically. They first load the
6360 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6361 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6362 the original value stored at <tt>ptr</tt>.
6369 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
6370 <i>; yields {i32}:result0 = 7</i>
6371 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
6372 <i>; yields {i32}:result1 = -2</i>
6373 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
6374 <i>; yields {i32}:result2 = 8</i>
6375 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
6376 <i>; yields {i32}:result3 = 8</i>
6377 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6381 <!-- ======================================================================= -->
6382 <div class="doc_subsection">
6383 <a name="int_general">General Intrinsics</a>
6386 <div class="doc_text">
6387 <p> This class of intrinsics is designed to be generic and has
6388 no specific purpose. </p>
6391 <!-- _______________________________________________________________________ -->
6392 <div class="doc_subsubsection">
6393 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6396 <div class="doc_text">
6400 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
6406 The '<tt>llvm.var.annotation</tt>' intrinsic
6412 The first argument is a pointer to a value, the second is a pointer to a
6413 global string, the third is a pointer to a global string which is the source
6414 file name, and the last argument is the line number.
6420 This intrinsic allows annotation of local variables with arbitrary strings.
6421 This can be useful for special purpose optimizations that want to look for these
6422 annotations. These have no other defined use, they are ignored by code
6423 generation and optimization.
6427 <!-- _______________________________________________________________________ -->
6428 <div class="doc_subsubsection">
6429 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
6432 <div class="doc_text">
6435 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6436 any integer bit width.
6439 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
6440 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
6441 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
6442 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
6443 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
6449 The '<tt>llvm.annotation</tt>' intrinsic.
6455 The first argument is an integer value (result of some expression),
6456 the second is a pointer to a global string, the third is a pointer to a global
6457 string which is the source file name, and the last argument is the line number.
6458 It returns the value of the first argument.
6464 This intrinsic allows annotations to be put on arbitrary expressions
6465 with arbitrary strings. This can be useful for special purpose optimizations
6466 that want to look for these annotations. These have no other defined use, they
6467 are ignored by code generation and optimization.
6471 <!-- _______________________________________________________________________ -->
6472 <div class="doc_subsubsection">
6473 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6476 <div class="doc_text">
6480 declare void @llvm.trap()
6486 The '<tt>llvm.trap</tt>' intrinsic
6498 This intrinsics is lowered to the target dependent trap instruction. If the
6499 target does not have a trap instruction, this intrinsic will be lowered to the
6500 call of the abort() function.
6504 <!-- _______________________________________________________________________ -->
6505 <div class="doc_subsubsection">
6506 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
6508 <div class="doc_text">
6511 declare void @llvm.stackprotector( i8* <guard>, i8** <slot> )
6516 The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and stores
6517 it onto the stack at <tt>slot</tt>. The stack slot is adjusted to ensure that
6518 it is placed on the stack before local variables.
6522 The <tt>llvm.stackprotector</tt> intrinsic requires two pointer arguments. The
6523 first argument is the value loaded from the stack guard
6524 <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt> that
6525 has enough space to hold the value of the guard.
6529 This intrinsic causes the prologue/epilogue inserter to force the position of
6530 the <tt>AllocaInst</tt> stack slot to be before local variables on the
6531 stack. This is to ensure that if a local variable on the stack is overwritten,
6532 it will destroy the value of the guard. When the function exits, the guard on
6533 the stack is checked against the original guard. If they're different, then
6534 the program aborts by calling the <tt>__stack_chk_fail()</tt> function.
6538 <!-- *********************************************************************** -->
6541 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
6542 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
6543 <a href="http://validator.w3.org/check/referer"><img
6544 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
6546 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
6547 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
6548 Last modified: $Date$