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="#globalvars">Global Variables</a></li>
26 <li><a href="#functionstructure">Functions</a></li>
27 <li><a href="#aliasstructure">Aliases</a></li>
28 <li><a href="#paramattrs">Parameter Attributes</a></li>
29 <li><a href="#fnattrs">Function Attributes</a></li>
30 <li><a href="#gc">Garbage Collector Names</a></li>
31 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
32 <li><a href="#datalayout">Data Layout</a></li>
35 <li><a href="#typesystem">Type System</a>
37 <li><a href="#t_classifications">Type Classifications</a></li>
38 <li><a href="#t_primitive">Primitive Types</a>
40 <li><a href="#t_floating">Floating Point Types</a></li>
41 <li><a href="#t_void">Void Type</a></li>
42 <li><a href="#t_label">Label Type</a></li>
45 <li><a href="#t_derived">Derived Types</a>
47 <li><a href="#t_integer">Integer Type</a></li>
48 <li><a href="#t_array">Array Type</a></li>
49 <li><a href="#t_function">Function Type</a></li>
50 <li><a href="#t_pointer">Pointer Type</a></li>
51 <li><a href="#t_struct">Structure Type</a></li>
52 <li><a href="#t_pstruct">Packed Structure Type</a></li>
53 <li><a href="#t_vector">Vector Type</a></li>
54 <li><a href="#t_opaque">Opaque Type</a></li>
59 <li><a href="#constants">Constants</a>
61 <li><a href="#simpleconstants">Simple Constants</a></li>
62 <li><a href="#aggregateconstants">Aggregate Constants</a></li>
63 <li><a href="#globalconstants">Global Variable and Function Addresses</a></li>
64 <li><a href="#undefvalues">Undefined Values</a></li>
65 <li><a href="#constantexprs">Constant Expressions</a></li>
68 <li><a href="#othervalues">Other Values</a>
70 <li><a href="#inlineasm">Inline Assembler Expressions</a></li>
73 <li><a href="#instref">Instruction Reference</a>
75 <li><a href="#terminators">Terminator Instructions</a>
77 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
78 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
79 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
80 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
81 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
82 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
85 <li><a href="#binaryops">Binary Operations</a>
87 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
88 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
89 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
90 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
91 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
92 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
93 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
94 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
95 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
98 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
100 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
101 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
102 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
103 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
104 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
105 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
108 <li><a href="#vectorops">Vector Operations</a>
110 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
111 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
112 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
115 <li><a href="#aggregateops">Aggregate Operations</a>
117 <li><a href="#i_extractvalue">'<tt>extractvalue</tt>' Instruction</a></li>
118 <li><a href="#i_insertvalue">'<tt>insertvalue</tt>' Instruction</a></li>
121 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
123 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
124 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
125 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
126 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
127 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
128 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
131 <li><a href="#convertops">Conversion Operations</a>
133 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
134 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
135 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
136 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
137 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
138 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
139 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
140 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
141 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
142 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
143 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
144 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
147 <li><a href="#otherops">Other Operations</a>
149 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
150 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
151 <li><a href="#i_vicmp">'<tt>vicmp</tt>' Instruction</a></li>
152 <li><a href="#i_vfcmp">'<tt>vfcmp</tt>' Instruction</a></li>
153 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
154 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
155 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
156 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
161 <li><a href="#intrinsics">Intrinsic Functions</a>
163 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
165 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
166 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
167 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
170 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
172 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
173 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
174 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
177 <li><a href="#int_codegen">Code Generator Intrinsics</a>
179 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
180 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
181 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
182 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
183 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
184 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
185 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
188 <li><a href="#int_libc">Standard C Library Intrinsics</a>
190 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
191 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
192 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
193 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
194 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
195 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
196 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
197 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
200 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
202 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
203 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
204 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
205 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
206 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
207 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
210 <li><a href="#int_debugger">Debugger intrinsics</a></li>
211 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
212 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
214 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
217 <li><a href="#int_atomics">Atomic intrinsics</a>
219 <li><a href="#int_memory_barrier"><tt>llvm.memory_barrier</tt></a></li>
220 <li><a href="#int_atomic_cmp_swap"><tt>llvm.atomic.cmp.swap</tt></a></li>
221 <li><a href="#int_atomic_swap"><tt>llvm.atomic.swap</tt></a></li>
222 <li><a href="#int_atomic_load_add"><tt>llvm.atomic.load.add</tt></a></li>
223 <li><a href="#int_atomic_load_sub"><tt>llvm.atomic.load.sub</tt></a></li>
224 <li><a href="#int_atomic_load_and"><tt>llvm.atomic.load.and</tt></a></li>
225 <li><a href="#int_atomic_load_nand"><tt>llvm.atomic.load.nand</tt></a></li>
226 <li><a href="#int_atomic_load_or"><tt>llvm.atomic.load.or</tt></a></li>
227 <li><a href="#int_atomic_load_xor"><tt>llvm.atomic.load.xor</tt></a></li>
228 <li><a href="#int_atomic_load_max"><tt>llvm.atomic.load.max</tt></a></li>
229 <li><a href="#int_atomic_load_min"><tt>llvm.atomic.load.min</tt></a></li>
230 <li><a href="#int_atomic_load_umax"><tt>llvm.atomic.load.umax</tt></a></li>
231 <li><a href="#int_atomic_load_umin"><tt>llvm.atomic.load.umin</tt></a></li>
234 <li><a href="#int_general">General intrinsics</a>
236 <li><a href="#int_var_annotation">
237 '<tt>llvm.var.annotation</tt>' Intrinsic</a></li>
238 <li><a href="#int_annotation">
239 '<tt>llvm.annotation.*</tt>' Intrinsic</a></li>
240 <li><a href="#int_trap">
241 '<tt>llvm.trap</tt>' Intrinsic</a></li>
242 <li><a href="#int_stackprotector">
243 '<tt>llvm.stackprotector</tt>' Intrinsic</a></li>
250 <div class="doc_author">
251 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
252 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
255 <!-- *********************************************************************** -->
256 <div class="doc_section"> <a name="abstract">Abstract </a></div>
257 <!-- *********************************************************************** -->
259 <div class="doc_text">
260 <p>This document is a reference manual for the LLVM assembly language.
261 LLVM is a Static Single Assignment (SSA) based representation that provides
262 type safety, low-level operations, flexibility, and the capability of
263 representing 'all' high-level languages cleanly. It is the common code
264 representation used throughout all phases of the LLVM compilation
268 <!-- *********************************************************************** -->
269 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
270 <!-- *********************************************************************** -->
272 <div class="doc_text">
274 <p>The LLVM code representation is designed to be used in three
275 different forms: as an in-memory compiler IR, as an on-disk bitcode
276 representation (suitable for fast loading by a Just-In-Time compiler),
277 and as a human readable assembly language representation. This allows
278 LLVM to provide a powerful intermediate representation for efficient
279 compiler transformations and analysis, while providing a natural means
280 to debug and visualize the transformations. The three different forms
281 of LLVM are all equivalent. This document describes the human readable
282 representation and notation.</p>
284 <p>The LLVM representation aims to be light-weight and low-level
285 while being expressive, typed, and extensible at the same time. It
286 aims to be a "universal IR" of sorts, by being at a low enough level
287 that high-level ideas may be cleanly mapped to it (similar to how
288 microprocessors are "universal IR's", allowing many source languages to
289 be mapped to them). By providing type information, LLVM can be used as
290 the target of optimizations: for example, through pointer analysis, it
291 can be proven that a C automatic variable is never accessed outside of
292 the current function... allowing it to be promoted to a simple SSA
293 value instead of a memory location.</p>
297 <!-- _______________________________________________________________________ -->
298 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
300 <div class="doc_text">
302 <p>It is important to note that this document describes 'well formed'
303 LLVM assembly language. There is a difference between what the parser
304 accepts and what is considered 'well formed'. For example, the
305 following instruction is syntactically okay, but not well formed:</p>
307 <div class="doc_code">
309 %x = <a href="#i_add">add</a> i32 1, %x
313 <p>...because the definition of <tt>%x</tt> does not dominate all of
314 its uses. The LLVM infrastructure provides a verification pass that may
315 be used to verify that an LLVM module is well formed. This pass is
316 automatically run by the parser after parsing input assembly and by
317 the optimizer before it outputs bitcode. The violations pointed out
318 by the verifier pass indicate bugs in transformation passes or input to
322 <!-- Describe the typesetting conventions here. -->
324 <!-- *********************************************************************** -->
325 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
326 <!-- *********************************************************************** -->
328 <div class="doc_text">
330 <p>LLVM identifiers come in two basic types: global and local. Global
331 identifiers (functions, global variables) begin with the @ character. Local
332 identifiers (register names, types) begin with the % character. Additionally,
333 there are three different formats for identifiers, for different purposes:</p>
336 <li>Named values are represented as a string of characters with their prefix.
337 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
338 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
339 Identifiers which require other characters in their names can be surrounded
340 with quotes. Special characters may be escaped using "\xx" where xx is the
341 ASCII code for the character in hexadecimal. In this way, any character can
342 be used in a name value, even quotes themselves.
344 <li>Unnamed values are represented as an unsigned numeric value with their
345 prefix. For example, %12, @2, %44.</li>
347 <li>Constants, which are described in a <a href="#constants">section about
348 constants</a>, below.</li>
351 <p>LLVM requires that values start with a prefix for two reasons: Compilers
352 don't need to worry about name clashes with reserved words, and the set of
353 reserved words may be expanded in the future without penalty. Additionally,
354 unnamed identifiers allow a compiler to quickly come up with a temporary
355 variable without having to avoid symbol table conflicts.</p>
357 <p>Reserved words in LLVM are very similar to reserved words in other
358 languages. There are keywords for different opcodes
359 ('<tt><a href="#i_add">add</a></tt>',
360 '<tt><a href="#i_bitcast">bitcast</a></tt>',
361 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
362 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
363 and others. These reserved words cannot conflict with variable names, because
364 none of them start with a prefix character ('%' or '@').</p>
366 <p>Here is an example of LLVM code to multiply the integer variable
367 '<tt>%X</tt>' by 8:</p>
371 <div class="doc_code">
373 %result = <a href="#i_mul">mul</a> i32 %X, 8
377 <p>After strength reduction:</p>
379 <div class="doc_code">
381 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
385 <p>And the hard way:</p>
387 <div class="doc_code">
389 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
390 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
391 %result = <a href="#i_add">add</a> i32 %1, %1
395 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
396 important lexical features of LLVM:</p>
400 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
403 <li>Unnamed temporaries are created when the result of a computation is not
404 assigned to a named value.</li>
406 <li>Unnamed temporaries are numbered sequentially</li>
410 <p>...and it also shows a convention that we follow in this document. When
411 demonstrating instructions, we will follow an instruction with a comment that
412 defines the type and name of value produced. Comments are shown in italic
417 <!-- *********************************************************************** -->
418 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
419 <!-- *********************************************************************** -->
421 <!-- ======================================================================= -->
422 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
425 <div class="doc_text">
427 <p>LLVM programs are composed of "Module"s, each of which is a
428 translation unit of the input programs. Each module consists of
429 functions, global variables, and symbol table entries. Modules may be
430 combined together with the LLVM linker, which merges function (and
431 global variable) definitions, resolves forward declarations, and merges
432 symbol table entries. Here is an example of the "hello world" module:</p>
434 <div class="doc_code">
435 <pre><i>; Declare the string constant as a global constant...</i>
436 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
437 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
439 <i>; External declaration of the puts function</i>
440 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
442 <i>; Definition of main function</i>
443 define i32 @main() { <i>; i32()* </i>
444 <i>; Convert [13x i8 ]* to i8 *...</i>
446 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
448 <i>; Call puts function to write out the string to stdout...</i>
450 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
452 href="#i_ret">ret</a> i32 0<br>}<br>
456 <p>This example is made up of a <a href="#globalvars">global variable</a>
457 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
458 function, and a <a href="#functionstructure">function definition</a>
459 for "<tt>main</tt>".</p>
461 <p>In general, a module is made up of a list of global values,
462 where both functions and global variables are global values. Global values are
463 represented by a pointer to a memory location (in this case, a pointer to an
464 array of char, and a pointer to a function), and have one of the following <a
465 href="#linkage">linkage types</a>.</p>
469 <!-- ======================================================================= -->
470 <div class="doc_subsection">
471 <a name="linkage">Linkage Types</a>
474 <div class="doc_text">
477 All Global Variables and Functions have one of the following types of linkage:
482 <dt><tt><b><a name="linkage_internal">internal</a></b></tt>: </dt>
484 <dd>Global values with internal linkage are only directly accessible by
485 objects in the current module. In particular, linking code into a module with
486 an internal global value may cause the internal to be renamed as necessary to
487 avoid collisions. Because the symbol is internal to the module, all
488 references can be updated. This corresponds to the notion of the
489 '<tt>static</tt>' keyword in C.
492 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
494 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
495 the same name when linkage occurs. This is typically used to implement
496 inline functions, templates, or other code which must be generated in each
497 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
498 allowed to be discarded.
501 <dt><tt><b><a name="linkage_common">common</a></b></tt>: </dt>
503 <dd>"<tt>common</tt>" linkage is exactly the same as <tt>linkonce</tt>
504 linkage, except that unreferenced <tt>common</tt> globals may not be
505 discarded. This is used for globals that may be emitted in multiple
506 translation units, but that are not guaranteed to be emitted into every
507 translation unit that uses them. One example of this is tentative
508 definitions in C, such as "<tt>int X;</tt>" at global scope.
511 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
513 <dd>"<tt>weak</tt>" linkage is the same as <tt>common</tt> linkage, except
514 that some targets may choose to emit different assembly sequences for them
515 for target-dependent reasons. This is used for globals that are declared
516 "weak" in C source code.
519 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
521 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
522 pointer to array type. When two global variables with appending linkage are
523 linked together, the two global arrays are appended together. This is the
524 LLVM, typesafe, equivalent of having the system linker append together
525 "sections" with identical names when .o files are linked.
528 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
529 <dd>The semantics of this linkage follow the ELF object file model: the
530 symbol is weak until linked, if not linked, the symbol becomes null instead
531 of being an undefined reference.
534 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
536 <dd>If none of the above identifiers are used, the global is externally
537 visible, meaning that it participates in linkage and can be used to resolve
538 external symbol references.
543 The next two types of linkage are targeted for Microsoft Windows platform
544 only. They are designed to support importing (exporting) symbols from (to)
545 DLLs (Dynamic Link Libraries).
549 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
551 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
552 or variable via a global pointer to a pointer that is set up by the DLL
553 exporting the symbol. On Microsoft Windows targets, the pointer name is
554 formed by combining <code>_imp__</code> and the function or variable name.
557 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
559 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
560 pointer to a pointer in a DLL, so that it can be referenced with the
561 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
562 name is formed by combining <code>_imp__</code> and the function or variable
568 <p>For example, since the "<tt>.LC0</tt>"
569 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
570 variable and was linked with this one, one of the two would be renamed,
571 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
572 external (i.e., lacking any linkage declarations), they are accessible
573 outside of the current module.</p>
574 <p>It is illegal for a function <i>declaration</i>
575 to have any linkage type other than "externally visible", <tt>dllimport</tt>,
576 or <tt>extern_weak</tt>.</p>
577 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
581 <!-- ======================================================================= -->
582 <div class="doc_subsection">
583 <a name="callingconv">Calling Conventions</a>
586 <div class="doc_text">
588 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
589 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
590 specified for the call. The calling convention of any pair of dynamic
591 caller/callee must match, or the behavior of the program is undefined. The
592 following calling conventions are supported by LLVM, and more may be added in
596 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
598 <dd>This calling convention (the default if no other calling convention is
599 specified) matches the target C calling conventions. This calling convention
600 supports varargs function calls and tolerates some mismatch in the declared
601 prototype and implemented declaration of the function (as does normal C).
604 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
606 <dd>This calling convention attempts to make calls as fast as possible
607 (e.g. by passing things in registers). This calling convention allows the
608 target to use whatever tricks it wants to produce fast code for the target,
609 without having to conform to an externally specified ABI (Application Binary
610 Interface). Implementations of this convention should allow arbitrary
611 <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> to be
612 supported. This calling convention does not support varargs and requires the
613 prototype of all callees to exactly match the prototype of the function
617 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
619 <dd>This calling convention attempts to make code in the caller as efficient
620 as possible under the assumption that the call is not commonly executed. As
621 such, these calls often preserve all registers so that the call does not break
622 any live ranges in the caller side. This calling convention does not support
623 varargs and requires the prototype of all callees to exactly match the
624 prototype of the function definition.
627 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
629 <dd>Any calling convention may be specified by number, allowing
630 target-specific calling conventions to be used. Target specific calling
631 conventions start at 64.
635 <p>More calling conventions can be added/defined on an as-needed basis, to
636 support pascal conventions or any other well-known target-independent
641 <!-- ======================================================================= -->
642 <div class="doc_subsection">
643 <a name="visibility">Visibility Styles</a>
646 <div class="doc_text">
649 All Global Variables and Functions have one of the following visibility styles:
653 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
655 <dd>On targets that use the ELF object file format, default visibility means
656 that the declaration is visible to other
657 modules and, in shared libraries, means that the declared entity may be
658 overridden. On Darwin, default visibility means that the declaration is
659 visible to other modules. Default visibility corresponds to "external
660 linkage" in the language.
663 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
665 <dd>Two declarations of an object with hidden visibility refer to the same
666 object if they are in the same shared object. Usually, hidden visibility
667 indicates that the symbol will not be placed into the dynamic symbol table,
668 so no other module (executable or shared library) can reference it
672 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
674 <dd>On ELF, protected visibility indicates that the symbol will be placed in
675 the dynamic symbol table, but that references within the defining module will
676 bind to the local symbol. That is, the symbol cannot be overridden by another
683 <!-- ======================================================================= -->
684 <div class="doc_subsection">
685 <a name="globalvars">Global Variables</a>
688 <div class="doc_text">
690 <p>Global variables define regions of memory allocated at compilation time
691 instead of run-time. Global variables may optionally be initialized, may have
692 an explicit section to be placed in, and may have an optional explicit alignment
693 specified. A variable may be defined as "thread_local", which means that it
694 will not be shared by threads (each thread will have a separated copy of the
695 variable). A variable may be defined as a global "constant," which indicates
696 that the contents of the variable will <b>never</b> be modified (enabling better
697 optimization, allowing the global data to be placed in the read-only section of
698 an executable, etc). Note that variables that need runtime initialization
699 cannot be marked "constant" as there is a store to the variable.</p>
702 LLVM explicitly allows <em>declarations</em> of global variables to be marked
703 constant, even if the final definition of the global is not. This capability
704 can be used to enable slightly better optimization of the program, but requires
705 the language definition to guarantee that optimizations based on the
706 'constantness' are valid for the translation units that do not include the
710 <p>As SSA values, global variables define pointer values that are in
711 scope (i.e. they dominate) all basic blocks in the program. Global
712 variables always define a pointer to their "content" type because they
713 describe a region of memory, and all memory objects in LLVM are
714 accessed through pointers.</p>
716 <p>A global variable may be declared to reside in a target-specifc numbered
717 address space. For targets that support them, address spaces may affect how
718 optimizations are performed and/or what target instructions are used to access
719 the variable. The default address space is zero. The address space qualifier
720 must precede any other attributes.</p>
722 <p>LLVM allows an explicit section to be specified for globals. If the target
723 supports it, it will emit globals to the section specified.</p>
725 <p>An explicit alignment may be specified for a global. If not present, or if
726 the alignment is set to zero, the alignment of the global is set by the target
727 to whatever it feels convenient. If an explicit alignment is specified, the
728 global is forced to have at least that much alignment. All alignments must be
731 <p>For example, the following defines a global in a numbered address space with
732 an initializer, section, and alignment:</p>
734 <div class="doc_code">
736 @G = constant float 1.0 addrspace(5), section "foo", align 4
743 <!-- ======================================================================= -->
744 <div class="doc_subsection">
745 <a name="functionstructure">Functions</a>
748 <div class="doc_text">
750 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
751 an optional <a href="#linkage">linkage type</a>, an optional
752 <a href="#visibility">visibility style</a>, an optional
753 <a href="#callingconv">calling convention</a>, a return type, an optional
754 <a href="#paramattrs">parameter attribute</a> for the return type, a function
755 name, a (possibly empty) argument list (each with optional
756 <a href="#paramattrs">parameter attributes</a>), optional
757 <a href="#fnattrs">function attributes</a>, an optional section,
758 an optional alignment, an optional <a href="#gc">garbage collector name</a>,
759 an opening curly brace, a list of basic blocks, and a closing curly brace.
761 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
762 optional <a href="#linkage">linkage type</a>, an optional
763 <a href="#visibility">visibility style</a>, an optional
764 <a href="#callingconv">calling convention</a>, a return type, an optional
765 <a href="#paramattrs">parameter attribute</a> for the return type, a function
766 name, a possibly empty list of arguments, an optional alignment, and an optional
767 <a href="#gc">garbage collector name</a>.</p>
769 <p>A function definition contains a list of basic blocks, forming the CFG
770 (Control Flow Graph) for
771 the function. Each basic block may optionally start with a label (giving the
772 basic block a symbol table entry), contains a list of instructions, and ends
773 with a <a href="#terminators">terminator</a> instruction (such as a branch or
774 function return).</p>
776 <p>The first basic block in a function is special in two ways: it is immediately
777 executed on entrance to the function, and it is not allowed to have predecessor
778 basic blocks (i.e. there can not be any branches to the entry block of a
779 function). Because the block can have no predecessors, it also cannot have any
780 <a href="#i_phi">PHI nodes</a>.</p>
782 <p>LLVM allows an explicit section to be specified for functions. If the target
783 supports it, it will emit functions to the section specified.</p>
785 <p>An explicit alignment may be specified for a function. If not present, or if
786 the alignment is set to zero, the alignment of the function is set by the target
787 to whatever it feels convenient. If an explicit alignment is specified, the
788 function is forced to have at least that much alignment. All alignments must be
793 <div class="doc_code">
795 define [<a href="#linkage">linkage</a>] [<a href="#visibility">visibility</a>]
796 [<a href="#callingconv">cconv</a>] [<a href="#paramattrs">ret attrs</a>]
797 <ResultType> @<FunctionName> ([argument list])
798 [<a href="#fnattrs">fn Attrs</a>] [section "name"] [align N]
799 [<a href="#gc">gc</a>] { ... }
806 <!-- ======================================================================= -->
807 <div class="doc_subsection">
808 <a name="aliasstructure">Aliases</a>
810 <div class="doc_text">
811 <p>Aliases act as "second name" for the aliasee value (which can be either
812 function, global variable, another alias or bitcast of global value). Aliases
813 may have an optional <a href="#linkage">linkage type</a>, and an
814 optional <a href="#visibility">visibility style</a>.</p>
818 <div class="doc_code">
820 @<Name> = alias [Linkage] [Visibility] <AliaseeTy> @<Aliasee>
828 <!-- ======================================================================= -->
829 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
830 <div class="doc_text">
831 <p>The return type and each parameter of a function type may have a set of
832 <i>parameter attributes</i> associated with them. Parameter attributes are
833 used to communicate additional information about the result or parameters of
834 a function. Parameter attributes are considered to be part of the function,
835 not of the function type, so functions with different parameter attributes
836 can have the same function type.</p>
838 <p>Parameter attributes are simple keywords that follow the type specified. If
839 multiple parameter attributes are needed, they are space separated. For
842 <div class="doc_code">
844 declare i32 @printf(i8* noalias , ...)
845 declare i32 @atoi(i8 zeroext)
846 declare signext i8 @returns_signed_char()
850 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
851 <tt>readonly</tt>) come immediately after the argument list.</p>
853 <p>Currently, only the following parameter attributes are defined:</p>
855 <dt><tt>zeroext</tt></dt>
856 <dd>This indicates to the code generator that the parameter or return value
857 should be zero-extended to a 32-bit value by the caller (for a parameter)
858 or the callee (for a return value).</dd>
860 <dt><tt>signext</tt></dt>
861 <dd>This indicates to the code generator that the parameter or return value
862 should be sign-extended to a 32-bit value by the caller (for a parameter)
863 or the callee (for a return value).</dd>
865 <dt><tt>inreg</tt></dt>
866 <dd>This indicates that this parameter or return value should be treated
867 in a special target-dependent fashion during while emitting code for a
868 function call or return (usually, by putting it in a register as opposed
869 to memory, though some targets use it to distinguish between two different
870 kinds of registers). Use of this attribute is target-specific.</dd>
872 <dt><tt><a name="byval">byval</a></tt></dt>
873 <dd>This indicates that the pointer parameter should really be passed by
874 value to the function. The attribute implies that a hidden copy of the
875 pointee is made between the caller and the callee, so the callee is unable
876 to modify the value in the callee. This attribute is only valid on LLVM
877 pointer arguments. It is generally used to pass structs and arrays by
878 value, but is also valid on pointers to scalars. The copy is considered to
879 belong to the caller not the callee (for example,
880 <tt><a href="#readonly">readonly</a></tt> functions should not write to
881 <tt>byval</tt> parameters). This is not a valid attribute for return
884 <dt><tt>sret</tt></dt>
885 <dd>This indicates that the pointer parameter specifies the address of a
886 structure that is the return value of the function in the source program.
887 This pointer must be guaranteed by the caller to be valid: loads and stores
888 to the structure may be assumed by the callee to not to trap. This may only
889 be applied to the first parameter. This is not a valid attribute for
892 <dt><tt>noalias</tt></dt>
893 <dd>This indicates that the pointer does not alias any global or any other
894 parameter. The caller is responsible for ensuring that this is the
895 case. On a function return value, <tt>noalias</tt> additionally indicates
896 that the pointer does not alias any other pointers visible to the
897 caller. For further details, please see the discussion of the NoAlias
899 <a href="http://llvm.org/docs/AliasAnalysis.html#MustMayNo">alias
902 <dt><tt>nocapture</tt></dt>
903 <dd>This indicates that the callee does not make any copies of the pointer
904 that outlive the callee itself. This is not a valid attribute for return
907 <dt><tt>nest</tt></dt>
908 <dd>This indicates that the pointer parameter can be excised using the
909 <a href="#int_trampoline">trampoline intrinsics</a>. This is not a valid
910 attribute for return values.</dd>
915 <!-- ======================================================================= -->
916 <div class="doc_subsection">
917 <a name="gc">Garbage Collector Names</a>
920 <div class="doc_text">
921 <p>Each function may specify a garbage collector name, which is simply a
924 <div class="doc_code"><pre
925 >define void @f() gc "name" { ...</pre></div>
927 <p>The compiler declares the supported values of <i>name</i>. Specifying a
928 collector which will cause the compiler to alter its output in order to support
929 the named garbage collection algorithm.</p>
932 <!-- ======================================================================= -->
933 <div class="doc_subsection">
934 <a name="fnattrs">Function Attributes</a>
937 <div class="doc_text">
939 <p>Function attributes are set to communicate additional information about
940 a function. Function attributes are considered to be part of the function,
941 not of the function type, so functions with different parameter attributes
942 can have the same function type.</p>
944 <p>Function attributes are simple keywords that follow the type specified. If
945 multiple attributes are needed, they are space separated. For
948 <div class="doc_code">
950 define void @f() noinline { ... }
951 define void @f() alwaysinline { ... }
952 define void @f() alwaysinline optsize { ... }
953 define void @f() optsize
958 <dt><tt>alwaysinline</tt></dt>
959 <dd>This attribute indicates that the inliner should attempt to inline this
960 function into callers whenever possible, ignoring any active inlining size
961 threshold for this caller.</dd>
963 <dt><tt>noinline</tt></dt>
964 <dd>This attribute indicates that the inliner should never inline this function
965 in any situation. This attribute may not be used together with the
966 <tt>alwaysinline</tt> attribute.</dd>
968 <dt><tt>optsize</tt></dt>
969 <dd>This attribute suggests that optimization passes and code generator passes
970 make choices that keep the code size of this function low, and otherwise do
971 optimizations specifically to reduce code size.</dd>
973 <dt><tt>noreturn</tt></dt>
974 <dd>This function attribute indicates that the function never returns normally.
975 This produces undefined behavior at runtime if the function ever does
976 dynamically return.</dd>
978 <dt><tt>nounwind</tt></dt>
979 <dd>This function attribute indicates that the function never returns with an
980 unwind or exceptional control flow. If the function does unwind, its runtime
981 behavior is undefined.</dd>
983 <dt><tt>readnone</tt></dt>
984 <dd>This attribute indicates that the function computes its result (or the
985 exception it throws) based strictly on its arguments, without dereferencing any
986 pointer arguments or otherwise accessing any mutable state (e.g. memory, control
987 registers, etc) visible to caller functions. It does not write through any
988 pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments) and
989 never changes any state visible to callers.</dd>
991 <dt><tt><a name="readonly">readonly</a></tt></dt>
992 <dd>This attribute indicates that the function does not write through any
993 pointer arguments (including <tt><a href="#byval">byval</a></tt> arguments)
994 or otherwise modify any state (e.g. memory, control registers, etc) visible to
995 caller functions. It may dereference pointer arguments and read state that may
996 be set in the caller. A readonly function always returns the same value (or
997 throws the same exception) when called with the same set of arguments and global
1000 <dt><tt><a name="ssp">ssp</a></tt></dt>
1001 <dd>This attribute indicates that the function should emit a stack smashing
1002 protector. It is in the form of a "canary"—a random value placed on the
1003 stack before the local variables that's checked upon return from the function to
1004 see if it has been overwritten. A heuristic is used to determine if a function
1005 needs stack protectors or not.
1007 <p>If a function that has an <tt>ssp</tt> attribute is inlined into a function
1008 that doesn't have an <tt>ssp</tt> attribute, then the resulting function will
1009 have an <tt>ssp</tt> attribute.</p></dd>
1011 <dt><tt>sspreq</tt></dt>
1012 <dd>This attribute indicates that the function should <em>always</em> emit a
1013 stack smashing protector. This overrides the <tt><a href="#ssp">ssp</a></tt>
1016 <p>If a function that has an <tt>sspreq</tt> attribute is inlined into a
1017 function that doesn't have an <tt>sspreq</tt> attribute or which has
1018 an <tt>ssp</tt> attribute, then the resulting function will have
1019 an <tt>sspreq</tt> attribute.</p></dd>
1024 <!-- ======================================================================= -->
1025 <div class="doc_subsection">
1026 <a name="moduleasm">Module-Level Inline Assembly</a>
1029 <div class="doc_text">
1031 Modules may contain "module-level inline asm" blocks, which corresponds to the
1032 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
1033 LLVM and treated as a single unit, but may be separated in the .ll file if
1034 desired. The syntax is very simple:
1037 <div class="doc_code">
1039 module asm "inline asm code goes here"
1040 module asm "more can go here"
1044 <p>The strings can contain any character by escaping non-printable characters.
1045 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
1050 The inline asm code is simply printed to the machine code .s file when
1051 assembly code is generated.
1055 <!-- ======================================================================= -->
1056 <div class="doc_subsection">
1057 <a name="datalayout">Data Layout</a>
1060 <div class="doc_text">
1061 <p>A module may specify a target specific data layout string that specifies how
1062 data is to be laid out in memory. The syntax for the data layout is simply:</p>
1063 <pre> target datalayout = "<i>layout specification</i>"</pre>
1064 <p>The <i>layout specification</i> consists of a list of specifications
1065 separated by the minus sign character ('-'). Each specification starts with a
1066 letter and may include other information after the letter to define some
1067 aspect of the data layout. The specifications accepted are as follows: </p>
1070 <dd>Specifies that the target lays out data in big-endian form. That is, the
1071 bits with the most significance have the lowest address location.</dd>
1073 <dd>Specifies that the target lays out data in little-endian form. That is,
1074 the bits with the least significance have the lowest address location.</dd>
1075 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1076 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
1077 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
1078 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
1080 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1081 <dd>This specifies the alignment for an integer type of a given bit
1082 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
1083 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1084 <dd>This specifies the alignment for a vector type of a given bit
1086 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1087 <dd>This specifies the alignment for a floating point type of a given bit
1088 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
1090 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
1091 <dd>This specifies the alignment for an aggregate type of a given bit
1094 <p>When constructing the data layout for a given target, LLVM starts with a
1095 default set of specifications which are then (possibly) overriden by the
1096 specifications in the <tt>datalayout</tt> keyword. The default specifications
1097 are given in this list:</p>
1099 <li><tt>E</tt> - big endian</li>
1100 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
1101 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
1102 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
1103 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
1104 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
1105 <li><tt>i64:32:64</tt> - i64 has ABI alignment of 32-bits but preferred
1106 alignment of 64-bits</li>
1107 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
1108 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
1109 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
1110 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
1111 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
1113 <p>When LLVM is determining the alignment for a given type, it uses the
1114 following rules:</p>
1116 <li>If the type sought is an exact match for one of the specifications, that
1117 specification is used.</li>
1118 <li>If no match is found, and the type sought is an integer type, then the
1119 smallest integer type that is larger than the bitwidth of the sought type is
1120 used. If none of the specifications are larger than the bitwidth then the the
1121 largest integer type is used. For example, given the default specifications
1122 above, the i7 type will use the alignment of i8 (next largest) while both
1123 i65 and i256 will use the alignment of i64 (largest specified).</li>
1124 <li>If no match is found, and the type sought is a vector type, then the
1125 largest vector type that is smaller than the sought vector type will be used
1126 as a fall back. This happens because <128 x double> can be implemented
1127 in terms of 64 <2 x double>, for example.</li>
1131 <!-- *********************************************************************** -->
1132 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
1133 <!-- *********************************************************************** -->
1135 <div class="doc_text">
1137 <p>The LLVM type system is one of the most important features of the
1138 intermediate representation. Being typed enables a number of
1139 optimizations to be performed on the intermediate representation directly,
1140 without having to do
1141 extra analyses on the side before the transformation. A strong type
1142 system makes it easier to read the generated code and enables novel
1143 analyses and transformations that are not feasible to perform on normal
1144 three address code representations.</p>
1148 <!-- ======================================================================= -->
1149 <div class="doc_subsection"> <a name="t_classifications">Type
1150 Classifications</a> </div>
1151 <div class="doc_text">
1152 <p>The types fall into a few useful
1153 classifications:</p>
1155 <table border="1" cellspacing="0" cellpadding="4">
1157 <tr><th>Classification</th><th>Types</th></tr>
1159 <td><a href="#t_integer">integer</a></td>
1160 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1163 <td><a href="#t_floating">floating point</a></td>
1164 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1167 <td><a name="t_firstclass">first class</a></td>
1168 <td><a href="#t_integer">integer</a>,
1169 <a href="#t_floating">floating point</a>,
1170 <a href="#t_pointer">pointer</a>,
1171 <a href="#t_vector">vector</a>,
1172 <a href="#t_struct">structure</a>,
1173 <a href="#t_array">array</a>,
1174 <a href="#t_label">label</a>.
1178 <td><a href="#t_primitive">primitive</a></td>
1179 <td><a href="#t_label">label</a>,
1180 <a href="#t_void">void</a>,
1181 <a href="#t_floating">floating point</a>.</td>
1184 <td><a href="#t_derived">derived</a></td>
1185 <td><a href="#t_integer">integer</a>,
1186 <a href="#t_array">array</a>,
1187 <a href="#t_function">function</a>,
1188 <a href="#t_pointer">pointer</a>,
1189 <a href="#t_struct">structure</a>,
1190 <a href="#t_pstruct">packed structure</a>,
1191 <a href="#t_vector">vector</a>,
1192 <a href="#t_opaque">opaque</a>.
1198 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
1199 most important. Values of these types are the only ones which can be
1200 produced by instructions, passed as arguments, or used as operands to
1204 <!-- ======================================================================= -->
1205 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1207 <div class="doc_text">
1208 <p>The primitive types are the fundamental building blocks of the LLVM
1213 <!-- _______________________________________________________________________ -->
1214 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1216 <div class="doc_text">
1219 <tr><th>Type</th><th>Description</th></tr>
1220 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1221 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1222 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1223 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1224 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1229 <!-- _______________________________________________________________________ -->
1230 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1232 <div class="doc_text">
1234 <p>The void type does not represent any value and has no size.</p>
1243 <!-- _______________________________________________________________________ -->
1244 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1246 <div class="doc_text">
1248 <p>The label type represents code labels.</p>
1258 <!-- ======================================================================= -->
1259 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1261 <div class="doc_text">
1263 <p>The real power in LLVM comes from the derived types in the system.
1264 This is what allows a programmer to represent arrays, functions,
1265 pointers, and other useful types. Note that these derived types may be
1266 recursive: For example, it is possible to have a two dimensional array.</p>
1270 <!-- _______________________________________________________________________ -->
1271 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1273 <div class="doc_text">
1276 <p>The integer type is a very simple derived type that simply specifies an
1277 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1278 2^23-1 (about 8 million) can be specified.</p>
1286 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1290 <table class="layout">
1293 <td><tt>i1</tt></td>
1294 <td>a single-bit integer.</td>
1296 <td><tt>i32</tt></td>
1297 <td>a 32-bit integer.</td>
1299 <td><tt>i1942652</tt></td>
1300 <td>a really big integer of over 1 million bits.</td>
1306 <!-- _______________________________________________________________________ -->
1307 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1309 <div class="doc_text">
1313 <p>The array type is a very simple derived type that arranges elements
1314 sequentially in memory. The array type requires a size (number of
1315 elements) and an underlying data type.</p>
1320 [<# elements> x <elementtype>]
1323 <p>The number of elements is a constant integer value; elementtype may
1324 be any type with a size.</p>
1327 <table class="layout">
1329 <td class="left"><tt>[40 x i32]</tt></td>
1330 <td class="left">Array of 40 32-bit integer values.</td>
1333 <td class="left"><tt>[41 x i32]</tt></td>
1334 <td class="left">Array of 41 32-bit integer values.</td>
1337 <td class="left"><tt>[4 x i8]</tt></td>
1338 <td class="left">Array of 4 8-bit integer values.</td>
1341 <p>Here are some examples of multidimensional arrays:</p>
1342 <table class="layout">
1344 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1345 <td class="left">3x4 array of 32-bit integer values.</td>
1348 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1349 <td class="left">12x10 array of single precision floating point values.</td>
1352 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1353 <td class="left">2x3x4 array of 16-bit integer values.</td>
1357 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1358 length array. Normally, accesses past the end of an array are undefined in
1359 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1360 As a special case, however, zero length arrays are recognized to be variable
1361 length. This allows implementation of 'pascal style arrays' with the LLVM
1362 type "{ i32, [0 x float]}", for example.</p>
1366 <!-- _______________________________________________________________________ -->
1367 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1368 <div class="doc_text">
1372 <p>The function type can be thought of as a function signature. It
1373 consists of a return type and a list of formal parameter types. The
1374 return type of a function type is a scalar type, a void type, or a struct type.
1375 If the return type is a struct type then all struct elements must be of first
1376 class types, and the struct must have at least one element.</p>
1381 <returntype list> (<parameter list>)
1384 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1385 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1386 which indicates that the function takes a variable number of arguments.
1387 Variable argument functions can access their arguments with the <a
1388 href="#int_varargs">variable argument handling intrinsic</a> functions.
1389 '<tt><returntype list></tt>' is a comma-separated list of
1390 <a href="#t_firstclass">first class</a> type specifiers.</p>
1393 <table class="layout">
1395 <td class="left"><tt>i32 (i32)</tt></td>
1396 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1398 </tr><tr class="layout">
1399 <td class="left"><tt>float (i16 signext, i32 *) *
1401 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1402 an <tt>i16</tt> that should be sign extended and a
1403 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1406 </tr><tr class="layout">
1407 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1408 <td class="left">A vararg function that takes at least one
1409 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1410 which returns an integer. This is the signature for <tt>printf</tt> in
1413 </tr><tr class="layout">
1414 <td class="left"><tt>{i32, i32} (i32)</tt></td>
1415 <td class="left">A function taking an <tt>i32</tt>, returning two
1416 <tt>i32</tt> values as an aggregate of type <tt>{ i32, i32 }</tt>
1422 <!-- _______________________________________________________________________ -->
1423 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1424 <div class="doc_text">
1426 <p>The structure type is used to represent a collection of data members
1427 together in memory. The packing of the field types is defined to match
1428 the ABI of the underlying processor. The elements of a structure may
1429 be any type that has a size.</p>
1430 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1431 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1432 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1435 <pre> { <type list> }<br></pre>
1437 <table class="layout">
1439 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1440 <td class="left">A triple of three <tt>i32</tt> values</td>
1441 </tr><tr class="layout">
1442 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1443 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1444 second element is a <a href="#t_pointer">pointer</a> to a
1445 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1446 an <tt>i32</tt>.</td>
1451 <!-- _______________________________________________________________________ -->
1452 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1454 <div class="doc_text">
1456 <p>The packed structure type is used to represent a collection of data members
1457 together in memory. There is no padding between fields. Further, the alignment
1458 of a packed structure is 1 byte. The elements of a packed structure may
1459 be any type that has a size.</p>
1460 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1461 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1462 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1465 <pre> < { <type list> } > <br></pre>
1467 <table class="layout">
1469 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1470 <td class="left">A triple of three <tt>i32</tt> values</td>
1471 </tr><tr class="layout">
1473 <tt>< { float, i32 (i32)* } ></tt></td>
1474 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1475 second element is a <a href="#t_pointer">pointer</a> to a
1476 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1477 an <tt>i32</tt>.</td>
1482 <!-- _______________________________________________________________________ -->
1483 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1484 <div class="doc_text">
1486 <p>As in many languages, the pointer type represents a pointer or
1487 reference to another object, which must live in memory. Pointer types may have
1488 an optional address space attribute defining the target-specific numbered
1489 address space where the pointed-to object resides. The default address space is
1492 <pre> <type> *<br></pre>
1494 <table class="layout">
1496 <td class="left"><tt>[4x i32]*</tt></td>
1497 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1498 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1501 <td class="left"><tt>i32 (i32 *) *</tt></td>
1502 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1503 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1507 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1508 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1509 that resides in address space #5.</td>
1514 <!-- _______________________________________________________________________ -->
1515 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1516 <div class="doc_text">
1520 <p>A vector type is a simple derived type that represents a vector
1521 of elements. Vector types are used when multiple primitive data
1522 are operated in parallel using a single instruction (SIMD).
1523 A vector type requires a size (number of
1524 elements) and an underlying primitive data type. Vectors must have a power
1525 of two length (1, 2, 4, 8, 16 ...). Vector types are
1526 considered <a href="#t_firstclass">first class</a>.</p>
1531 < <# elements> x <elementtype> >
1534 <p>The number of elements is a constant integer value; elementtype may
1535 be any integer or floating point type.</p>
1539 <table class="layout">
1541 <td class="left"><tt><4 x i32></tt></td>
1542 <td class="left">Vector of 4 32-bit integer values.</td>
1545 <td class="left"><tt><8 x float></tt></td>
1546 <td class="left">Vector of 8 32-bit floating-point values.</td>
1549 <td class="left"><tt><2 x i64></tt></td>
1550 <td class="left">Vector of 2 64-bit integer values.</td>
1555 <!-- _______________________________________________________________________ -->
1556 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1557 <div class="doc_text">
1561 <p>Opaque types are used to represent unknown types in the system. This
1562 corresponds (for example) to the C notion of a forward declared structure type.
1563 In LLVM, opaque types can eventually be resolved to any type (not just a
1564 structure type).</p>
1574 <table class="layout">
1576 <td class="left"><tt>opaque</tt></td>
1577 <td class="left">An opaque type.</td>
1583 <!-- *********************************************************************** -->
1584 <div class="doc_section"> <a name="constants">Constants</a> </div>
1585 <!-- *********************************************************************** -->
1587 <div class="doc_text">
1589 <p>LLVM has several different basic types of constants. This section describes
1590 them all and their syntax.</p>
1594 <!-- ======================================================================= -->
1595 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1597 <div class="doc_text">
1600 <dt><b>Boolean constants</b></dt>
1602 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1603 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1606 <dt><b>Integer constants</b></dt>
1608 <dd>Standard integers (such as '4') are constants of the <a
1609 href="#t_integer">integer</a> type. Negative numbers may be used with
1613 <dt><b>Floating point constants</b></dt>
1615 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1616 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1617 notation (see below). The assembler requires the exact decimal value of
1618 a floating-point constant. For example, the assembler accepts 1.25 but
1619 rejects 1.3 because 1.3 is a repeating decimal in binary. Floating point
1620 constants must have a <a href="#t_floating">floating point</a> type. </dd>
1622 <dt><b>Null pointer constants</b></dt>
1624 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1625 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1629 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1630 of floating point constants. For example, the form '<tt>double
1631 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1632 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1633 (and the only time that they are generated by the disassembler) is when a
1634 floating point constant must be emitted but it cannot be represented as a
1635 decimal floating point number. For example, NaN's, infinities, and other
1636 special values are represented in their IEEE hexadecimal format so that
1637 assembly and disassembly do not cause any bits to change in the constants.</p>
1641 <!-- ======================================================================= -->
1642 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1645 <div class="doc_text">
1646 <p>Aggregate constants arise from aggregation of simple constants
1647 and smaller aggregate constants.</p>
1650 <dt><b>Structure constants</b></dt>
1652 <dd>Structure constants are represented with notation similar to structure
1653 type definitions (a comma separated list of elements, surrounded by braces
1654 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1655 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1656 must have <a href="#t_struct">structure type</a>, and the number and
1657 types of elements must match those specified by the type.
1660 <dt><b>Array constants</b></dt>
1662 <dd>Array constants are represented with notation similar to array type
1663 definitions (a comma separated list of elements, surrounded by square brackets
1664 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1665 constants must have <a href="#t_array">array type</a>, and the number and
1666 types of elements must match those specified by the type.
1669 <dt><b>Vector constants</b></dt>
1671 <dd>Vector constants are represented with notation similar to vector type
1672 definitions (a comma separated list of elements, surrounded by
1673 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1674 i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
1675 href="#t_vector">vector type</a>, and the number and types of elements must
1676 match those specified by the type.
1679 <dt><b>Zero initialization</b></dt>
1681 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1682 value to zero of <em>any</em> type, including scalar and aggregate types.
1683 This is often used to avoid having to print large zero initializers (e.g. for
1684 large arrays) and is always exactly equivalent to using explicit zero
1691 <!-- ======================================================================= -->
1692 <div class="doc_subsection">
1693 <a name="globalconstants">Global Variable and Function Addresses</a>
1696 <div class="doc_text">
1698 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1699 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1700 constants. These constants are explicitly referenced when the <a
1701 href="#identifiers">identifier for the global</a> is used and always have <a
1702 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1705 <div class="doc_code">
1709 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1715 <!-- ======================================================================= -->
1716 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1717 <div class="doc_text">
1718 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1719 no specific value. Undefined values may be of any type and be used anywhere
1720 a constant is permitted.</p>
1722 <p>Undefined values indicate to the compiler that the program is well defined
1723 no matter what value is used, giving the compiler more freedom to optimize.
1727 <!-- ======================================================================= -->
1728 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1731 <div class="doc_text">
1733 <p>Constant expressions are used to allow expressions involving other constants
1734 to be used as constants. Constant expressions may be of any <a
1735 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1736 that does not have side effects (e.g. load and call are not supported). The
1737 following is the syntax for constant expressions:</p>
1740 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1741 <dd>Truncate a constant to another type. The bit size of CST must be larger
1742 than the bit size of TYPE. Both types must be integers.</dd>
1744 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1745 <dd>Zero extend a constant to another type. The bit size of CST must be
1746 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1748 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1749 <dd>Sign extend a constant to another type. The bit size of CST must be
1750 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1752 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1753 <dd>Truncate a floating point constant to another floating point type. The
1754 size of CST must be larger than the size of TYPE. Both types must be
1755 floating point.</dd>
1757 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1758 <dd>Floating point extend a constant to another type. The size of CST must be
1759 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1761 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
1762 <dd>Convert a floating point constant to the corresponding unsigned integer
1763 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1764 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1765 of the same number of elements. If the value won't fit in the integer type,
1766 the results are undefined.</dd>
1768 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1769 <dd>Convert a floating point constant to the corresponding signed integer
1770 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1771 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1772 of the same number of elements. If the value won't fit in the integer type,
1773 the results are undefined.</dd>
1775 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1776 <dd>Convert an unsigned integer constant to the corresponding floating point
1777 constant. TYPE must be a scalar or vector floating point type. CST must be of
1778 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1779 of the same number of elements. If the value won't fit in the floating point
1780 type, the results are undefined.</dd>
1782 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1783 <dd>Convert a signed integer constant to the corresponding floating point
1784 constant. TYPE must be a scalar or vector floating point type. CST must be of
1785 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1786 of the same number of elements. If the value won't fit in the floating point
1787 type, the results are undefined.</dd>
1789 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1790 <dd>Convert a pointer typed constant to the corresponding integer constant
1791 TYPE must be an integer type. CST must be of pointer type. The CST value is
1792 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1794 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1795 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1796 pointer type. CST must be of integer type. The CST value is zero extended,
1797 truncated, or unchanged to make it fit in a pointer size. This one is
1798 <i>really</i> dangerous!</dd>
1800 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1801 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1802 identical (same number of bits). The conversion is done as if the CST value
1803 was stored to memory and read back as TYPE. In other words, no bits change
1804 with this operator, just the type. This can be used for conversion of
1805 vector types to any other type, as long as they have the same bit width. For
1806 pointers it is only valid to cast to another pointer type. It is not valid
1807 to bitcast to or from an aggregate type.
1810 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1812 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1813 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1814 instruction, the index list may have zero or more indexes, which are required
1815 to make sense for the type of "CSTPTR".</dd>
1817 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1819 <dd>Perform the <a href="#i_select">select operation</a> on
1822 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1823 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1825 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1826 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1828 <dt><b><tt>vicmp COND ( VAL1, VAL2 )</tt></b></dt>
1829 <dd>Performs the <a href="#i_vicmp">vicmp operation</a> on constants.</dd>
1831 <dt><b><tt>vfcmp COND ( VAL1, VAL2 )</tt></b></dt>
1832 <dd>Performs the <a href="#i_vfcmp">vfcmp operation</a> on constants.</dd>
1834 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1836 <dd>Perform the <a href="#i_extractelement">extractelement
1837 operation</a> on constants.</dd>
1839 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1841 <dd>Perform the <a href="#i_insertelement">insertelement
1842 operation</a> on constants.</dd>
1845 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1847 <dd>Perform the <a href="#i_shufflevector">shufflevector
1848 operation</a> on constants.</dd>
1850 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1852 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1853 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1854 binary</a> operations. The constraints on operands are the same as those for
1855 the corresponding instruction (e.g. no bitwise operations on floating point
1856 values are allowed).</dd>
1860 <!-- *********************************************************************** -->
1861 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1862 <!-- *********************************************************************** -->
1864 <!-- ======================================================================= -->
1865 <div class="doc_subsection">
1866 <a name="inlineasm">Inline Assembler Expressions</a>
1869 <div class="doc_text">
1872 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1873 Module-Level Inline Assembly</a>) through the use of a special value. This
1874 value represents the inline assembler as a string (containing the instructions
1875 to emit), a list of operand constraints (stored as a string), and a flag that
1876 indicates whether or not the inline asm expression has side effects. An example
1877 inline assembler expression is:
1880 <div class="doc_code">
1882 i32 (i32) asm "bswap $0", "=r,r"
1887 Inline assembler expressions may <b>only</b> be used as the callee operand of
1888 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1891 <div class="doc_code">
1893 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1898 Inline asms with side effects not visible in the constraint list must be marked
1899 as having side effects. This is done through the use of the
1900 '<tt>sideeffect</tt>' keyword, like so:
1903 <div class="doc_code">
1905 call void asm sideeffect "eieio", ""()
1909 <p>TODO: The format of the asm and constraints string still need to be
1910 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1911 need to be documented). This is probably best done by reference to another
1912 document that covers inline asm from a holistic perspective.
1917 <!-- *********************************************************************** -->
1918 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1919 <!-- *********************************************************************** -->
1921 <div class="doc_text">
1923 <p>The LLVM instruction set consists of several different
1924 classifications of instructions: <a href="#terminators">terminator
1925 instructions</a>, <a href="#binaryops">binary instructions</a>,
1926 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1927 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1928 instructions</a>.</p>
1932 <!-- ======================================================================= -->
1933 <div class="doc_subsection"> <a name="terminators">Terminator
1934 Instructions</a> </div>
1936 <div class="doc_text">
1938 <p>As mentioned <a href="#functionstructure">previously</a>, every
1939 basic block in a program ends with a "Terminator" instruction, which
1940 indicates which block should be executed after the current block is
1941 finished. These terminator instructions typically yield a '<tt>void</tt>'
1942 value: they produce control flow, not values (the one exception being
1943 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1944 <p>There are six different terminator instructions: the '<a
1945 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1946 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1947 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1948 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1949 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1953 <!-- _______________________________________________________________________ -->
1954 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1955 Instruction</a> </div>
1956 <div class="doc_text">
1959 ret <type> <value> <i>; Return a value from a non-void function</i>
1960 ret void <i>; Return from void function</i>
1965 <p>The '<tt>ret</tt>' instruction is used to return control flow (and
1966 optionally a value) from a function back to the caller.</p>
1967 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1968 returns a value and then causes control flow, and one that just causes
1969 control flow to occur.</p>
1973 <p>The '<tt>ret</tt>' instruction optionally accepts a single argument,
1974 the return value. The type of the return value must be a
1975 '<a href="#t_firstclass">first class</a>' type.</p>
1977 <p>A function is not <a href="#wellformed">well formed</a> if
1978 it it has a non-void return type and contains a '<tt>ret</tt>'
1979 instruction with no return value or a return value with a type that
1980 does not match its type, or if it has a void return type and contains
1981 a '<tt>ret</tt>' instruction with a return value.</p>
1985 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1986 returns back to the calling function's context. If the caller is a "<a
1987 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1988 the instruction after the call. If the caller was an "<a
1989 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1990 at the beginning of the "normal" destination block. If the instruction
1991 returns a value, that value shall set the call or invoke instruction's
1997 ret i32 5 <i>; Return an integer value of 5</i>
1998 ret void <i>; Return from a void function</i>
1999 ret { i32, i8 } { i32 4, i8 2 } <i>; Return an aggregate of values 4 and 2</i>
2002 <!-- _______________________________________________________________________ -->
2003 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
2004 <div class="doc_text">
2006 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
2009 <p>The '<tt>br</tt>' instruction is used to cause control flow to
2010 transfer to a different basic block in the current function. There are
2011 two forms of this instruction, corresponding to a conditional branch
2012 and an unconditional branch.</p>
2014 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
2015 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
2016 unconditional form of the '<tt>br</tt>' instruction takes a single
2017 '<tt>label</tt>' value as a target.</p>
2019 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
2020 argument is evaluated. If the value is <tt>true</tt>, control flows
2021 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
2022 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
2024 <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
2025 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
2027 <!-- _______________________________________________________________________ -->
2028 <div class="doc_subsubsection">
2029 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
2032 <div class="doc_text">
2036 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
2041 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
2042 several different places. It is a generalization of the '<tt>br</tt>'
2043 instruction, allowing a branch to occur to one of many possible
2049 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
2050 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
2051 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
2052 table is not allowed to contain duplicate constant entries.</p>
2056 <p>The <tt>switch</tt> instruction specifies a table of values and
2057 destinations. When the '<tt>switch</tt>' instruction is executed, this
2058 table is searched for the given value. If the value is found, control flow is
2059 transfered to the corresponding destination; otherwise, control flow is
2060 transfered to the default destination.</p>
2062 <h5>Implementation:</h5>
2064 <p>Depending on properties of the target machine and the particular
2065 <tt>switch</tt> instruction, this instruction may be code generated in different
2066 ways. For example, it could be generated as a series of chained conditional
2067 branches or with a lookup table.</p>
2072 <i>; Emulate a conditional br instruction</i>
2073 %Val = <a href="#i_zext">zext</a> i1 %value to i32
2074 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
2076 <i>; Emulate an unconditional br instruction</i>
2077 switch i32 0, label %dest [ ]
2079 <i>; Implement a jump table:</i>
2080 switch i32 %val, label %otherwise [ i32 0, label %onzero
2082 i32 2, label %ontwo ]
2086 <!-- _______________________________________________________________________ -->
2087 <div class="doc_subsubsection">
2088 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
2091 <div class="doc_text">
2096 <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>]
2097 to label <normal label> unwind label <exception label>
2102 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
2103 function, with the possibility of control flow transfer to either the
2104 '<tt>normal</tt>' label or the
2105 '<tt>exception</tt>' label. If the callee function returns with the
2106 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
2107 "normal" label. If the callee (or any indirect callees) returns with the "<a
2108 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
2109 continued at the dynamically nearest "exception" label.</p>
2113 <p>This instruction requires several arguments:</p>
2117 The optional "cconv" marker indicates which <a href="#callingconv">calling
2118 convention</a> the call should use. If none is specified, the call defaults
2119 to using C calling conventions.
2122 <li>The optional <a href="#paramattrs">Parameter Attributes</a> list for
2123 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
2124 and '<tt>inreg</tt>' attributes are valid here.</li>
2126 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
2127 function value being invoked. In most cases, this is a direct function
2128 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
2129 an arbitrary pointer to function value.
2132 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
2133 function to be invoked. </li>
2135 <li>'<tt>function args</tt>': argument list whose types match the function
2136 signature argument types. If the function signature indicates the function
2137 accepts a variable number of arguments, the extra arguments can be
2140 <li>'<tt>normal label</tt>': the label reached when the called function
2141 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
2143 <li>'<tt>exception label</tt>': the label reached when a callee returns with
2144 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
2146 <li>The optional <a href="#fnattrs">function attributes</a> list. Only
2147 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
2148 '<tt>readnone</tt>' attributes are valid here.</li>
2153 <p>This instruction is designed to operate as a standard '<tt><a
2154 href="#i_call">call</a></tt>' instruction in most regards. The primary
2155 difference is that it establishes an association with a label, which is used by
2156 the runtime library to unwind the stack.</p>
2158 <p>This instruction is used in languages with destructors to ensure that proper
2159 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
2160 exception. Additionally, this is important for implementation of
2161 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
2165 %retval = invoke i32 @Test(i32 15) to label %Continue
2166 unwind label %TestCleanup <i>; {i32}:retval set</i>
2167 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Testfnptr(i32 15) to label %Continue
2168 unwind label %TestCleanup <i>; {i32}:retval set</i>
2173 <!-- _______________________________________________________________________ -->
2175 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
2176 Instruction</a> </div>
2178 <div class="doc_text">
2187 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
2188 at the first callee in the dynamic call stack which used an <a
2189 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
2190 primarily used to implement exception handling.</p>
2194 <p>The '<tt>unwind</tt>' instruction causes execution of the current function to
2195 immediately halt. The dynamic call stack is then searched for the first <a
2196 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
2197 execution continues at the "exceptional" destination block specified by the
2198 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
2199 dynamic call chain, undefined behavior results.</p>
2202 <!-- _______________________________________________________________________ -->
2204 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2205 Instruction</a> </div>
2207 <div class="doc_text">
2216 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2217 instruction is used to inform the optimizer that a particular portion of the
2218 code is not reachable. This can be used to indicate that the code after a
2219 no-return function cannot be reached, and other facts.</p>
2223 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2228 <!-- ======================================================================= -->
2229 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2230 <div class="doc_text">
2231 <p>Binary operators are used to do most of the computation in a
2232 program. They require two operands of the same type, execute an operation on them, and
2233 produce a single value. The operands might represent
2234 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
2235 The result value has the same type as its operands.</p>
2236 <p>There are several different binary operators:</p>
2238 <!-- _______________________________________________________________________ -->
2239 <div class="doc_subsubsection">
2240 <a name="i_add">'<tt>add</tt>' Instruction</a>
2243 <div class="doc_text">
2248 <result> = add <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2253 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2257 <p>The two arguments to the '<tt>add</tt>' instruction must be <a
2258 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>, or
2259 <a href="#t_vector">vector</a> values. Both arguments must have identical
2264 <p>The value produced is the integer or floating point sum of the two
2267 <p>If an integer sum has unsigned overflow, the result returned is the
2268 mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2271 <p>Because LLVM integers use a two's complement representation, this
2272 instruction is appropriate for both signed and unsigned integers.</p>
2277 <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2280 <!-- _______________________________________________________________________ -->
2281 <div class="doc_subsubsection">
2282 <a name="i_sub">'<tt>sub</tt>' Instruction</a>
2285 <div class="doc_text">
2290 <result> = sub <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2295 <p>The '<tt>sub</tt>' instruction returns the difference of its two
2298 <p>Note that the '<tt>sub</tt>' instruction is used to represent the
2299 '<tt>neg</tt>' instruction present in most other intermediate
2300 representations.</p>
2304 <p>The two arguments to the '<tt>sub</tt>' instruction must be <a
2305 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2306 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2311 <p>The value produced is the integer or floating point difference of
2312 the two operands.</p>
2314 <p>If an integer difference has unsigned overflow, the result returned is the
2315 mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2318 <p>Because LLVM integers use a two's complement representation, this
2319 instruction is appropriate for both signed and unsigned integers.</p>
2323 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2324 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2328 <!-- _______________________________________________________________________ -->
2329 <div class="doc_subsubsection">
2330 <a name="i_mul">'<tt>mul</tt>' Instruction</a>
2333 <div class="doc_text">
2336 <pre> <result> = mul <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2339 <p>The '<tt>mul</tt>' instruction returns the product of its two
2344 <p>The two arguments to the '<tt>mul</tt>' instruction must be <a
2345 href="#t_integer">integer</a>, <a href="#t_floating">floating point</a>,
2346 or <a href="#t_vector">vector</a> values. Both arguments must have identical
2351 <p>The value produced is the integer or floating point product of the
2354 <p>If the result of an integer multiplication has unsigned overflow,
2355 the result returned is the mathematical result modulo
2356 2<sup>n</sup>, where n is the bit width of the result.</p>
2357 <p>Because LLVM integers use a two's complement representation, and the
2358 result is the same width as the operands, this instruction returns the
2359 correct result for both signed and unsigned integers. If a full product
2360 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2361 should be sign-extended or zero-extended as appropriate to the
2362 width of the full product.</p>
2364 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2368 <!-- _______________________________________________________________________ -->
2369 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2371 <div class="doc_text">
2373 <pre> <result> = udiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2376 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2381 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2382 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2383 values. Both arguments must have identical types.</p>
2387 <p>The value produced is the unsigned integer quotient of the two operands.</p>
2388 <p>Note that unsigned integer division and signed integer division are distinct
2389 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2390 <p>Division by zero leads to undefined behavior.</p>
2392 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2395 <!-- _______________________________________________________________________ -->
2396 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2398 <div class="doc_text">
2401 <result> = sdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2406 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2411 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2412 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2413 values. Both arguments must have identical types.</p>
2416 <p>The value produced is the signed integer quotient of the two operands rounded towards zero.</p>
2417 <p>Note that signed integer division and unsigned integer division are distinct
2418 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2419 <p>Division by zero leads to undefined behavior. Overflow also leads to
2420 undefined behavior; this is a rare case, but can occur, for example,
2421 by doing a 32-bit division of -2147483648 by -1.</p>
2423 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2426 <!-- _______________________________________________________________________ -->
2427 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2428 Instruction</a> </div>
2429 <div class="doc_text">
2432 <result> = fdiv <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2436 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2441 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2442 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2443 of floating point values. Both arguments must have identical types.</p>
2447 <p>The value produced is the floating point quotient of the two operands.</p>
2452 <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2456 <!-- _______________________________________________________________________ -->
2457 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2459 <div class="doc_text">
2461 <pre> <result> = urem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2464 <p>The '<tt>urem</tt>' instruction returns the remainder from the
2465 unsigned division of its two arguments.</p>
2467 <p>The two arguments to the '<tt>urem</tt>' instruction must be
2468 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2469 values. Both arguments must have identical types.</p>
2471 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2472 This instruction always performs an unsigned division to get the remainder.</p>
2473 <p>Note that unsigned integer remainder and signed integer remainder are
2474 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2475 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
2477 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2481 <!-- _______________________________________________________________________ -->
2482 <div class="doc_subsubsection">
2483 <a name="i_srem">'<tt>srem</tt>' Instruction</a>
2486 <div class="doc_text">
2491 <result> = srem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2496 <p>The '<tt>srem</tt>' instruction returns the remainder from the
2497 signed division of its two operands. This instruction can also take
2498 <a href="#t_vector">vector</a> versions of the values in which case
2499 the elements must be integers.</p>
2503 <p>The two arguments to the '<tt>srem</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>
2509 <p>This instruction returns the <i>remainder</i> of a division (where the result
2510 has the same sign as the dividend, <tt>op1</tt>), not the <i>modulo</i>
2511 operator (where the result has the same sign as the divisor, <tt>op2</tt>) of
2512 a value. For more information about the difference, see <a
2513 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2514 Math Forum</a>. For a table of how this is implemented in various languages,
2515 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2516 Wikipedia: modulo operation</a>.</p>
2517 <p>Note that signed integer remainder and unsigned integer remainder are
2518 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2519 <p>Taking the remainder of a division by zero leads to undefined behavior.
2520 Overflow also leads to undefined behavior; this is a rare case, but can occur,
2521 for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2522 (The remainder doesn't actually overflow, but this rule lets srem be
2523 implemented using instructions that return both the result of the division
2524 and the remainder.)</p>
2526 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2530 <!-- _______________________________________________________________________ -->
2531 <div class="doc_subsubsection">
2532 <a name="i_frem">'<tt>frem</tt>' Instruction</a> </div>
2534 <div class="doc_text">
2537 <pre> <result> = frem <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2540 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2541 division of its two operands.</p>
2543 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2544 <a href="#t_floating">floating point</a> or <a href="#t_vector">vector</a>
2545 of floating point values. Both arguments must have identical types.</p>
2549 <p>This instruction returns the <i>remainder</i> of a division.
2550 The remainder has the same sign as the dividend.</p>
2555 <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2559 <!-- ======================================================================= -->
2560 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2561 Operations</a> </div>
2562 <div class="doc_text">
2563 <p>Bitwise binary operators are used to do various forms of
2564 bit-twiddling in a program. They are generally very efficient
2565 instructions and can commonly be strength reduced from other
2566 instructions. They require two operands of the same type, execute an operation on them,
2567 and produce a single value. The resulting value is the same type as its operands.</p>
2570 <!-- _______________________________________________________________________ -->
2571 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2572 Instruction</a> </div>
2573 <div class="doc_text">
2575 <pre> <result> = shl <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2580 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2581 the left a specified number of bits.</p>
2585 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2586 href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2587 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2591 <p>The value produced is <tt>op1</tt> * 2<sup><tt>op2</tt></sup> mod 2<sup>n</sup>,
2592 where n is the width of the result. If <tt>op2</tt> is (statically or dynamically) negative or
2593 equal to or larger than the number of bits in <tt>op1</tt>, the result is undefined.
2594 If the arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2595 corresponding shift amount in <tt>op2</tt>.</p>
2597 <h5>Example:</h5><pre>
2598 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2599 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2600 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2601 <result> = shl i32 1, 32 <i>; undefined</i>
2602 <result> = shl <2 x i32> < i32 1, i32 1>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 2, i32 4></i>
2605 <!-- _______________________________________________________________________ -->
2606 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2607 Instruction</a> </div>
2608 <div class="doc_text">
2610 <pre> <result> = lshr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2614 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2615 operand shifted to the right a specified number of bits with zero fill.</p>
2618 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2619 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2620 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2624 <p>This instruction always performs a logical shift right operation. The most
2625 significant bits of the result will be filled with zero bits after the
2626 shift. If <tt>op2</tt> is (statically or dynamically) equal to or larger than
2627 the number of bits in <tt>op1</tt>, the result is undefined. If the arguments are
2628 vectors, each vector element of <tt>op1</tt> is shifted by the corresponding shift
2629 amount in <tt>op2</tt>.</p>
2633 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2634 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2635 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2636 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2637 <result> = lshr i32 1, 32 <i>; undefined</i>
2638 <result> = lshr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 2> <i>; yields: result=<2 x i32> < i32 0x7FFFFFFF, i32 1></i>
2642 <!-- _______________________________________________________________________ -->
2643 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2644 Instruction</a> </div>
2645 <div class="doc_text">
2648 <pre> <result> = ashr <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2652 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2653 operand shifted to the right a specified number of bits with sign extension.</p>
2656 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2657 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2658 type. '<tt>op2</tt>' is treated as an unsigned value.</p>
2661 <p>This instruction always performs an arithmetic shift right operation,
2662 The most significant bits of the result will be filled with the sign bit
2663 of <tt>op1</tt>. If <tt>op2</tt> is (statically or dynamically) equal to or
2664 larger than the number of bits in <tt>op1</tt>, the result is undefined. If the
2665 arguments are vectors, each vector element of <tt>op1</tt> is shifted by the
2666 corresponding shift amount in <tt>op2</tt>.</p>
2670 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2671 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2672 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2673 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
2674 <result> = ashr i32 1, 32 <i>; undefined</i>
2675 <result> = ashr <2 x i32> < i32 -2, i32 4>, < i32 1, i32 3> <i>; yields: result=<2 x i32> < i32 -1, i32 0></i>
2679 <!-- _______________________________________________________________________ -->
2680 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2681 Instruction</a> </div>
2683 <div class="doc_text">
2688 <result> = and <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2693 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2694 its two operands.</p>
2698 <p>The two arguments to the '<tt>and</tt>' instruction must be
2699 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2700 values. Both arguments must have identical types.</p>
2703 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2706 <table border="1" cellspacing="0" cellpadding="4">
2738 <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
2739 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2740 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2743 <!-- _______________________________________________________________________ -->
2744 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2745 <div class="doc_text">
2747 <pre> <result> = or <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2750 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2751 or of its two operands.</p>
2754 <p>The two arguments to the '<tt>or</tt>' instruction must be
2755 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2756 values. Both arguments must have identical types.</p>
2758 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2761 <table border="1" cellspacing="0" cellpadding="4">
2792 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2793 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2794 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2797 <!-- _______________________________________________________________________ -->
2798 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2799 Instruction</a> </div>
2800 <div class="doc_text">
2802 <pre> <result> = xor <ty> <op1>, <op2> <i>; yields {ty}:result</i>
2805 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2806 or of its two operands. The <tt>xor</tt> is used to implement the
2807 "one's complement" operation, which is the "~" operator in C.</p>
2809 <p>The two arguments to the '<tt>xor</tt>' instruction must be
2810 <a href="#t_integer">integer</a> or <a href="#t_vector">vector</a> of integer
2811 values. Both arguments must have identical types.</p>
2815 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2818 <table border="1" cellspacing="0" cellpadding="4">
2850 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2851 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2852 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2853 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2857 <!-- ======================================================================= -->
2858 <div class="doc_subsection">
2859 <a name="vectorops">Vector Operations</a>
2862 <div class="doc_text">
2864 <p>LLVM supports several instructions to represent vector operations in a
2865 target-independent manner. These instructions cover the element-access and
2866 vector-specific operations needed to process vectors effectively. While LLVM
2867 does directly support these vector operations, many sophisticated algorithms
2868 will want to use target-specific intrinsics to take full advantage of a specific
2873 <!-- _______________________________________________________________________ -->
2874 <div class="doc_subsubsection">
2875 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2878 <div class="doc_text">
2883 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
2889 The '<tt>extractelement</tt>' instruction extracts a single scalar
2890 element from a vector at a specified index.
2897 The first operand of an '<tt>extractelement</tt>' instruction is a
2898 value of <a href="#t_vector">vector</a> type. The second operand is
2899 an index indicating the position from which to extract the element.
2900 The index may be a variable.</p>
2905 The result is a scalar of the same type as the element type of
2906 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2907 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2908 results are undefined.
2914 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
2919 <!-- _______________________________________________________________________ -->
2920 <div class="doc_subsubsection">
2921 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2924 <div class="doc_text">
2929 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
2935 The '<tt>insertelement</tt>' instruction inserts a scalar
2936 element into a vector at a specified index.
2943 The first operand of an '<tt>insertelement</tt>' instruction is a
2944 value of <a href="#t_vector">vector</a> type. The second operand is a
2945 scalar value whose type must equal the element type of the first
2946 operand. The third operand is an index indicating the position at
2947 which to insert the value. The index may be a variable.</p>
2952 The result is a vector of the same type as <tt>val</tt>. Its
2953 element values are those of <tt>val</tt> except at position
2954 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2955 exceeds the length of <tt>val</tt>, the results are undefined.
2961 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
2965 <!-- _______________________________________________________________________ -->
2966 <div class="doc_subsubsection">
2967 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2970 <div class="doc_text">
2975 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <m x i32> <mask> <i>; yields <m x <ty>></i>
2981 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2982 from two input vectors, returning a vector with the same element type as
2983 the input and length that is the same as the shuffle mask.
2989 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2990 with types that match each other. The third argument is a shuffle mask whose
2991 element type is always 'i32'. The result of the instruction is a vector whose
2992 length is the same as the shuffle mask and whose element type is the same as
2993 the element type of the first two operands.
2997 The shuffle mask operand is required to be a constant vector with either
2998 constant integer or undef values.
3004 The elements of the two input vectors are numbered from left to right across
3005 both of the vectors. The shuffle mask operand specifies, for each element of
3006 the result vector, which element of the two input vectors the result element
3007 gets. The element selector may be undef (meaning "don't care") and the second
3008 operand may be undef if performing a shuffle from only one vector.
3014 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3015 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
3016 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
3017 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
3018 %result = shufflevector <8 x i32> %v1, <8 x i32> undef,
3019 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i>
3020 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
3021 <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>
3026 <!-- ======================================================================= -->
3027 <div class="doc_subsection">
3028 <a name="aggregateops">Aggregate Operations</a>
3031 <div class="doc_text">
3033 <p>LLVM supports several instructions for working with aggregate values.
3038 <!-- _______________________________________________________________________ -->
3039 <div class="doc_subsubsection">
3040 <a name="i_extractvalue">'<tt>extractvalue</tt>' Instruction</a>
3043 <div class="doc_text">
3048 <result> = extractvalue <aggregate type> <val>, <idx>{, <idx>}*
3054 The '<tt>extractvalue</tt>' instruction extracts the value of a struct field
3055 or array element from an aggregate value.
3062 The first operand of an '<tt>extractvalue</tt>' instruction is a
3063 value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a>
3064 type. The operands are constant indices to specify which value to extract
3065 in a similar manner as indices in a
3066 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3072 The result is the value at the position in the aggregate specified by
3079 %result = extractvalue {i32, float} %agg, 0 <i>; yields i32</i>
3084 <!-- _______________________________________________________________________ -->
3085 <div class="doc_subsubsection">
3086 <a name="i_insertvalue">'<tt>insertvalue</tt>' Instruction</a>
3089 <div class="doc_text">
3094 <result> = insertvalue <aggregate type> <val>, <ty> <val>, <idx> <i>; yields <n x <ty>></i>
3100 The '<tt>insertvalue</tt>' instruction inserts a value
3101 into a struct field or array element in an aggregate.
3108 The first operand of an '<tt>insertvalue</tt>' instruction is a
3109 value of <a href="#t_struct">struct</a> or <a href="#t_array">array</a> type.
3110 The second operand is a first-class value to insert.
3111 The following operands are constant indices
3112 indicating the position at which to insert the value in a similar manner as
3114 '<tt><a href="#i_getelementptr">getelementptr</a></tt>' instruction.
3115 The value to insert must have the same type as the value identified
3122 The result is an aggregate of the same type as <tt>val</tt>. Its
3123 value is that of <tt>val</tt> except that the value at the position
3124 specified by the indices is that of <tt>elt</tt>.
3130 %result = insertvalue {i32, float} %agg, i32 1, 0 <i>; yields {i32, float}</i>
3135 <!-- ======================================================================= -->
3136 <div class="doc_subsection">
3137 <a name="memoryops">Memory Access and Addressing Operations</a>
3140 <div class="doc_text">
3142 <p>A key design point of an SSA-based representation is how it
3143 represents memory. In LLVM, no memory locations are in SSA form, which
3144 makes things very simple. This section describes how to read, write,
3145 allocate, and free memory in LLVM.</p>
3149 <!-- _______________________________________________________________________ -->
3150 <div class="doc_subsubsection">
3151 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
3154 <div class="doc_text">
3159 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3164 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
3165 heap and returns a pointer to it. The object is always allocated in the generic
3166 address space (address space zero).</p>
3170 <p>The '<tt>malloc</tt>' instruction allocates
3171 <tt>sizeof(<type>)*NumElements</tt>
3172 bytes of memory from the operating system and returns a pointer of the
3173 appropriate type to the program. If "NumElements" is specified, it is the
3174 number of elements allocated, otherwise "NumElements" is defaulted to be one.
3175 If a constant alignment is specified, the value result of the allocation is guaranteed to
3176 be aligned to at least that boundary. If not specified, or if zero, the target can
3177 choose to align the allocation on any convenient boundary.</p>
3179 <p>'<tt>type</tt>' must be a sized type.</p>
3183 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
3184 a pointer is returned. The result of a zero byte allocation is undefined. The
3185 result is null if there is insufficient memory available.</p>
3190 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
3192 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
3193 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
3194 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
3195 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
3196 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
3200 <!-- _______________________________________________________________________ -->
3201 <div class="doc_subsubsection">
3202 <a name="i_free">'<tt>free</tt>' Instruction</a>
3205 <div class="doc_text">
3210 free <type> <value> <i>; yields {void}</i>
3215 <p>The '<tt>free</tt>' instruction returns memory back to the unused
3216 memory heap to be reallocated in the future.</p>
3220 <p>'<tt>value</tt>' shall be a pointer value that points to a value
3221 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
3226 <p>Access to the memory pointed to by the pointer is no longer defined
3227 after this instruction executes. If the pointer is null, the operation
3233 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
3234 free [4 x i8]* %array
3238 <!-- _______________________________________________________________________ -->
3239 <div class="doc_subsubsection">
3240 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
3243 <div class="doc_text">
3248 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
3253 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
3254 currently executing function, to be automatically released when this function
3255 returns to its caller. The object is always allocated in the generic address
3256 space (address space zero).</p>
3260 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
3261 bytes of memory on the runtime stack, returning a pointer of the
3262 appropriate type to the program. If "NumElements" is specified, it is the
3263 number of elements allocated, otherwise "NumElements" is defaulted to be one.
3264 If a constant alignment is specified, the value result of the allocation is guaranteed
3265 to be aligned to at least that boundary. If not specified, or if zero, the target
3266 can choose to align the allocation on any convenient boundary.</p>
3268 <p>'<tt>type</tt>' may be any sized type.</p>
3272 <p>Memory is allocated; a pointer is returned. The operation is undefiend if
3273 there is insufficient stack space for the allocation. '<tt>alloca</tt>'d
3274 memory is automatically released when the function returns. The '<tt>alloca</tt>'
3275 instruction is commonly used to represent automatic variables that must
3276 have an address available. When the function returns (either with the <tt><a
3277 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
3278 instructions), the memory is reclaimed. Allocating zero bytes
3279 is legal, but the result is undefined.</p>
3284 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
3285 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
3286 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
3287 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
3291 <!-- _______________________________________________________________________ -->
3292 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
3293 Instruction</a> </div>
3294 <div class="doc_text">
3296 <pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
3298 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
3300 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
3301 address from which to load. The pointer must point to a <a
3302 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
3303 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
3304 the number or order of execution of this <tt>load</tt> with other
3305 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
3308 The optional constant "align" argument specifies the alignment of the operation
3309 (that is, the alignment of the memory address). A value of 0 or an
3310 omitted "align" argument means that the operation has the preferential
3311 alignment for the target. It is the responsibility of the code emitter
3312 to ensure that the alignment information is correct. Overestimating
3313 the alignment results in an undefined behavior. Underestimating the
3314 alignment may produce less efficient code. An alignment of 1 is always
3318 <p>The location of memory pointed to is loaded.</p>
3320 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3322 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
3323 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
3326 <!-- _______________________________________________________________________ -->
3327 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
3328 Instruction</a> </div>
3329 <div class="doc_text">
3331 <pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3332 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
3335 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
3337 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
3338 to store and an address at which to store it. The type of the '<tt><pointer></tt>'
3339 operand must be a pointer to the <a href="#t_firstclass">first class</a> type
3340 of the '<tt><value></tt>'
3341 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
3342 optimizer is not allowed to modify the number or order of execution of
3343 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
3344 href="#i_store">store</a></tt> instructions.</p>
3346 The optional constant "align" argument specifies the alignment of the operation
3347 (that is, the alignment of the memory address). A value of 0 or an
3348 omitted "align" argument means that the operation has the preferential
3349 alignment for the target. It is the responsibility of the code emitter
3350 to ensure that the alignment information is correct. Overestimating
3351 the alignment results in an undefined behavior. Underestimating the
3352 alignment may produce less efficient code. An alignment of 1 is always
3356 <p>The contents of memory are updated to contain '<tt><value></tt>'
3357 at the location specified by the '<tt><pointer></tt>' operand.</p>
3359 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
3360 store i32 3, i32* %ptr <i>; yields {void}</i>
3361 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
3365 <!-- _______________________________________________________________________ -->
3366 <div class="doc_subsubsection">
3367 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
3370 <div class="doc_text">
3373 <result> = getelementptr <pty>* <ptrval>{, <ty> <idx>}*
3379 The '<tt>getelementptr</tt>' instruction is used to get the address of a
3380 subelement of an aggregate data structure. It performs address calculation only
3381 and does not access memory.</p>
3385 <p>The first argument is always a pointer, and forms the basis of the
3386 calculation. The remaining arguments are indices, that indicate which of the
3387 elements of the aggregate object are indexed. The interpretation of each index
3388 is dependent on the type being indexed into. The first index always indexes the
3389 pointer value given as the first argument, the second index indexes a value of
3390 the type pointed to (not necessarily the value directly pointed to, since the
3391 first index can be non-zero), etc. The first type indexed into must be a pointer
3392 value, subsequent types can be arrays, vectors and structs. Note that subsequent
3393 types being indexed into can never be pointers, since that would require loading
3394 the pointer before continuing calculation.</p>
3396 <p>The type of each index argument depends on the type it is indexing into.
3397 When indexing into a (packed) structure, only <tt>i32</tt> integer
3398 <b>constants</b> are allowed. When indexing into an array, pointer or vector,
3399 only integers of 32 or 64 bits are allowed (also non-constants). 32-bit values
3400 will be sign extended to 64-bits if required.</p>
3402 <p>For example, let's consider a C code fragment and how it gets
3403 compiled to LLVM:</p>
3405 <div class="doc_code">
3418 int *foo(struct ST *s) {
3419 return &s[1].Z.B[5][13];
3424 <p>The LLVM code generated by the GCC frontend is:</p>
3426 <div class="doc_code">
3428 %RT = type { i8 , [10 x [20 x i32]], i8 }
3429 %ST = type { i32, double, %RT }
3431 define i32* %foo(%ST* %s) {
3433 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3441 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3442 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3443 }</tt>' type, a structure. The second index indexes into the third element of
3444 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3445 i8 }</tt>' type, another structure. The third index indexes into the second
3446 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3447 array. The two dimensions of the array are subscripted into, yielding an
3448 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3449 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3451 <p>Note that it is perfectly legal to index partially through a
3452 structure, returning a pointer to an inner element. Because of this,
3453 the LLVM code for the given testcase is equivalent to:</p>
3456 define i32* %foo(%ST* %s) {
3457 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3458 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3459 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3460 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3461 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3466 <p>Note that it is undefined to access an array out of bounds: array and
3467 pointer indexes must always be within the defined bounds of the array type.
3468 The one exception for this rule is zero length arrays. These arrays are
3469 defined to be accessible as variable length arrays, which requires access
3470 beyond the zero'th element.</p>
3472 <p>The getelementptr instruction is often confusing. For some more insight
3473 into how it works, see <a href="GetElementPtr.html">the getelementptr
3479 <i>; yields [12 x i8]*:aptr</i>
3480 %aptr = getelementptr {i32, [12 x i8]}* %saptr, i64 0, i32 1
3481 <i>; yields i8*:vptr</i>
3482 %vptr = getelementptr {i32, <2 x i8>}* %svptr, i64 0, i32 1, i32 1
3483 <i>; yields i8*:eptr</i>
3484 %eptr = getelementptr [12 x i8]* %aptr, i64 0, i32 1
3488 <!-- ======================================================================= -->
3489 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3491 <div class="doc_text">
3492 <p>The instructions in this category are the conversion instructions (casting)
3493 which all take a single operand and a type. They perform various bit conversions
3497 <!-- _______________________________________________________________________ -->
3498 <div class="doc_subsubsection">
3499 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3501 <div class="doc_text">
3505 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
3510 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3515 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3516 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3517 and type of the result, which must be an <a href="#t_integer">integer</a>
3518 type. The bit size of <tt>value</tt> must be larger than the bit size of
3519 <tt>ty2</tt>. Equal sized types are not allowed.</p>
3523 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3524 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3525 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3526 It will always truncate bits.</p>
3530 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3531 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3532 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3536 <!-- _______________________________________________________________________ -->
3537 <div class="doc_subsubsection">
3538 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3540 <div class="doc_text">
3544 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
3548 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
3553 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3554 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3555 also be of <a href="#t_integer">integer</a> type. The bit size of the
3556 <tt>value</tt> must be smaller than the bit size of the destination type,
3560 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3561 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3563 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
3567 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3568 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3572 <!-- _______________________________________________________________________ -->
3573 <div class="doc_subsubsection">
3574 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3576 <div class="doc_text">
3580 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
3584 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3588 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3589 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3590 also be of <a href="#t_integer">integer</a> type. The bit size of the
3591 <tt>value</tt> must be smaller than the bit size of the destination type,
3596 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3597 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3598 the type <tt>ty2</tt>.</p>
3600 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
3604 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3605 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3609 <!-- _______________________________________________________________________ -->
3610 <div class="doc_subsubsection">
3611 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3614 <div class="doc_text">
3619 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
3623 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3628 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3629 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3630 cast it to. The size of <tt>value</tt> must be larger than the size of
3631 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3632 <i>no-op cast</i>.</p>
3635 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3636 <a href="#t_floating">floating point</a> type to a smaller
3637 <a href="#t_floating">floating point</a> type. If the value cannot fit within
3638 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3642 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3643 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3647 <!-- _______________________________________________________________________ -->
3648 <div class="doc_subsubsection">
3649 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3651 <div class="doc_text">
3655 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
3659 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3660 floating point value.</p>
3663 <p>The '<tt>fpext</tt>' instruction takes a
3664 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3665 and a <a href="#t_floating">floating point</a> type to cast it to. The source
3666 type must be smaller than the destination type.</p>
3669 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3670 <a href="#t_floating">floating point</a> type to a larger
3671 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3672 used to make a <i>no-op cast</i> because it always changes bits. Use
3673 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3677 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3678 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3682 <!-- _______________________________________________________________________ -->
3683 <div class="doc_subsubsection">
3684 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3686 <div class="doc_text">
3690 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
3694 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
3695 unsigned integer equivalent of type <tt>ty2</tt>.
3699 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
3700 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3701 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3702 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3703 vector integer type with the same number of elements as <tt>ty</tt></p>
3706 <p> The '<tt>fptoui</tt>' instruction converts its
3707 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3708 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3709 the results are undefined.</p>
3713 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
3714 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
3715 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
3719 <!-- _______________________________________________________________________ -->
3720 <div class="doc_subsubsection">
3721 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3723 <div class="doc_text">
3727 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
3731 <p>The '<tt>fptosi</tt>' instruction converts
3732 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3736 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
3737 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3738 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3739 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3740 vector integer type with the same number of elements as <tt>ty</tt></p>
3743 <p>The '<tt>fptosi</tt>' instruction converts its
3744 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3745 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3746 the results are undefined.</p>
3750 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
3751 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
3752 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3756 <!-- _______________________________________________________________________ -->
3757 <div class="doc_subsubsection">
3758 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3760 <div class="doc_text">
3764 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3768 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3769 integer and converts that value to the <tt>ty2</tt> type.</p>
3772 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3773 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3774 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3775 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3776 floating point type with the same number of elements as <tt>ty</tt></p>
3779 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3780 integer quantity and converts it to the corresponding floating point value. If
3781 the value cannot fit in the floating point value, the results are undefined.</p>
3785 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3786 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3790 <!-- _______________________________________________________________________ -->
3791 <div class="doc_subsubsection">
3792 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3794 <div class="doc_text">
3798 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3802 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3803 integer and converts that value to the <tt>ty2</tt> type.</p>
3806 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3807 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3808 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3809 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3810 floating point type with the same number of elements as <tt>ty</tt></p>
3813 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3814 integer quantity and converts it to the corresponding floating point value. If
3815 the value cannot fit in the floating point value, the results are undefined.</p>
3819 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3820 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3824 <!-- _______________________________________________________________________ -->
3825 <div class="doc_subsubsection">
3826 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3828 <div class="doc_text">
3832 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
3836 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3837 the integer type <tt>ty2</tt>.</p>
3840 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3841 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3842 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.</p>
3845 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3846 <tt>ty2</tt> by interpreting the pointer value as an integer and either
3847 truncating or zero extending that value to the size of the integer type. If
3848 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3849 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3850 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3855 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3856 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3860 <!-- _______________________________________________________________________ -->
3861 <div class="doc_subsubsection">
3862 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3864 <div class="doc_text">
3868 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
3872 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3873 a pointer type, <tt>ty2</tt>.</p>
3876 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3877 value to cast, and a type to cast it to, which must be a
3878 <a href="#t_pointer">pointer</a> type.</p>
3881 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3882 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
3883 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3884 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3885 the size of a pointer then a zero extension is done. If they are the same size,
3886 nothing is done (<i>no-op cast</i>).</p>
3890 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3891 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3892 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3896 <!-- _______________________________________________________________________ -->
3897 <div class="doc_subsubsection">
3898 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3900 <div class="doc_text">
3904 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3909 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3910 <tt>ty2</tt> without changing any bits.</p>
3914 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3915 a non-aggregate first class value, and a type to cast it to, which must also be
3916 a non-aggregate <a href="#t_firstclass">first class</a> type. The bit sizes of
3918 and the destination type, <tt>ty2</tt>, must be identical. If the source
3919 type is a pointer, the destination type must also be a pointer. This
3920 instruction supports bitwise conversion of vectors to integers and to vectors
3921 of other types (as long as they have the same size).</p>
3924 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3925 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3926 this conversion. The conversion is done as if the <tt>value</tt> had been
3927 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3928 converted to other pointer types with this instruction. To convert pointers to
3929 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3930 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3934 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3935 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3936 %Z = bitcast <2 x int> %V to i64; <i>; yields i64: %V</i>
3940 <!-- ======================================================================= -->
3941 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3942 <div class="doc_text">
3943 <p>The instructions in this category are the "miscellaneous"
3944 instructions, which defy better classification.</p>
3947 <!-- _______________________________________________________________________ -->
3948 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3950 <div class="doc_text">
3952 <pre> <result> = icmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
3955 <p>The '<tt>icmp</tt>' instruction returns a boolean value or
3956 a vector of boolean values based on comparison
3957 of its two integer, integer vector, or pointer operands.</p>
3959 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3960 the condition code indicating the kind of comparison to perform. It is not
3961 a value, just a keyword. The possible condition code are:
3964 <li><tt>eq</tt>: equal</li>
3965 <li><tt>ne</tt>: not equal </li>
3966 <li><tt>ugt</tt>: unsigned greater than</li>
3967 <li><tt>uge</tt>: unsigned greater or equal</li>
3968 <li><tt>ult</tt>: unsigned less than</li>
3969 <li><tt>ule</tt>: unsigned less or equal</li>
3970 <li><tt>sgt</tt>: signed greater than</li>
3971 <li><tt>sge</tt>: signed greater or equal</li>
3972 <li><tt>slt</tt>: signed less than</li>
3973 <li><tt>sle</tt>: signed less or equal</li>
3975 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3976 <a href="#t_pointer">pointer</a>
3977 or integer <a href="#t_vector">vector</a> typed.
3978 They must also be identical types.</p>
3980 <p>The '<tt>icmp</tt>' compares <tt>op1</tt> and <tt>op2</tt> according to
3981 the condition code given as <tt>cond</tt>. The comparison performed always
3982 yields either an <a href="#t_primitive"><tt>i1</tt></a> or vector of <tt>i1</tt> result, as follows:
3985 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3986 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3988 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3989 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.</li>
3990 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3991 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
3992 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3993 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
3994 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3995 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
3996 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3997 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
3998 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3999 <tt>true</tt> if <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4000 <li><tt>sge</tt>: interprets the operands as signed values and yields
4001 <tt>true</tt> if <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4002 <li><tt>slt</tt>: interprets the operands as signed values and yields
4003 <tt>true</tt> if <tt>op1</tt> is less than <tt>op2</tt>.</li>
4004 <li><tt>sle</tt>: interprets the operands as signed values and yields
4005 <tt>true</tt> if <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4007 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
4008 values are compared as if they were integers.</p>
4009 <p>If the operands are integer vectors, then they are compared
4010 element by element. The result is an <tt>i1</tt> vector with
4011 the same number of elements as the values being compared.
4012 Otherwise, the result is an <tt>i1</tt>.
4016 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
4017 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
4018 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
4019 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
4020 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
4021 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
4025 <!-- _______________________________________________________________________ -->
4026 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
4028 <div class="doc_text">
4030 <pre> <result> = fcmp <cond> <ty> <op1>, <op2> <i>; yields {i1} or {<N x i1>}:result</i>
4033 <p>The '<tt>fcmp</tt>' instruction returns a boolean value
4034 or vector of boolean values based on comparison
4035 of its operands.</p>
4037 If the operands are floating point scalars, then the result
4038 type is a boolean (<a href="#t_primitive"><tt>i1</tt></a>).
4040 <p>If the operands are floating point vectors, then the result type
4041 is a vector of boolean with the same number of elements as the
4042 operands being compared.</p>
4044 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
4045 the condition code indicating the kind of comparison to perform. It is not
4046 a value, just a keyword. The possible condition code are:</p>
4048 <li><tt>false</tt>: no comparison, always returns false</li>
4049 <li><tt>oeq</tt>: ordered and equal</li>
4050 <li><tt>ogt</tt>: ordered and greater than </li>
4051 <li><tt>oge</tt>: ordered and greater than or equal</li>
4052 <li><tt>olt</tt>: ordered and less than </li>
4053 <li><tt>ole</tt>: ordered and less than or equal</li>
4054 <li><tt>one</tt>: ordered and not equal</li>
4055 <li><tt>ord</tt>: ordered (no nans)</li>
4056 <li><tt>ueq</tt>: unordered or equal</li>
4057 <li><tt>ugt</tt>: unordered or greater than </li>
4058 <li><tt>uge</tt>: unordered or greater than or equal</li>
4059 <li><tt>ult</tt>: unordered or less than </li>
4060 <li><tt>ule</tt>: unordered or less than or equal</li>
4061 <li><tt>une</tt>: unordered or not equal</li>
4062 <li><tt>uno</tt>: unordered (either nans)</li>
4063 <li><tt>true</tt>: no comparison, always returns true</li>
4065 <p><i>Ordered</i> means that neither operand is a QNAN while
4066 <i>unordered</i> means that either operand may be a QNAN.</p>
4067 <p>Each of <tt>val1</tt> and <tt>val2</tt> arguments must be
4068 either a <a href="#t_floating">floating point</a> type
4069 or a <a href="#t_vector">vector</a> of floating point type.
4070 They must have identical types.</p>
4072 <p>The '<tt>fcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4073 according to the condition code given as <tt>cond</tt>.
4074 If the operands are vectors, then the vectors are compared
4076 Each comparison performed
4077 always yields an <a href="#t_primitive">i1</a> result, as follows:</p>
4079 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
4080 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4081 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4082 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4083 <tt>op1</tt> is greather than <tt>op2</tt>.</li>
4084 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4085 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4086 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4087 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4088 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4089 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4090 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
4091 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4092 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
4093 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
4094 <tt>op1</tt> is equal to <tt>op2</tt>.</li>
4095 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
4096 <tt>op1</tt> is greater than <tt>op2</tt>.</li>
4097 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
4098 <tt>op1</tt> is greater than or equal to <tt>op2</tt>.</li>
4099 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
4100 <tt>op1</tt> is less than <tt>op2</tt>.</li>
4101 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
4102 <tt>op1</tt> is less than or equal to <tt>op2</tt>.</li>
4103 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
4104 <tt>op1</tt> is not equal to <tt>op2</tt>.</li>
4105 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
4106 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
4110 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
4111 <result> = fcmp one float 4.0, 5.0 <i>; yields: result=true</i>
4112 <result> = fcmp olt float 4.0, 5.0 <i>; yields: result=true</i>
4113 <result> = fcmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
4117 <!-- _______________________________________________________________________ -->
4118 <div class="doc_subsubsection">
4119 <a name="i_vicmp">'<tt>vicmp</tt>' Instruction</a>
4121 <div class="doc_text">
4123 <pre> <result> = vicmp <cond> <ty> <op1>, <op2> <i>; yields {ty}:result</i>
4126 <p>The '<tt>vicmp</tt>' instruction returns an integer vector value based on
4127 element-wise comparison of its two integer vector operands.</p>
4129 <p>The '<tt>vicmp</tt>' instruction takes three operands. The first operand is
4130 the condition code indicating the kind of comparison to perform. It is not
4131 a value, just a keyword. The possible condition code are:</p>
4133 <li><tt>eq</tt>: equal</li>
4134 <li><tt>ne</tt>: not equal </li>
4135 <li><tt>ugt</tt>: unsigned greater than</li>
4136 <li><tt>uge</tt>: unsigned greater or equal</li>
4137 <li><tt>ult</tt>: unsigned less than</li>
4138 <li><tt>ule</tt>: unsigned less or equal</li>
4139 <li><tt>sgt</tt>: signed greater than</li>
4140 <li><tt>sge</tt>: signed greater or equal</li>
4141 <li><tt>slt</tt>: signed less than</li>
4142 <li><tt>sle</tt>: signed less or equal</li>
4144 <p>The remaining two arguments must be <a href="#t_vector">vector</a> or
4145 <a href="#t_integer">integer</a> typed. They must also be identical types.</p>
4147 <p>The '<tt>vicmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4148 according to the condition code given as <tt>cond</tt>. The comparison yields a
4149 <a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, of
4150 identical type as the values being compared. The most significant bit in each
4151 element is 1 if the element-wise comparison evaluates to true, and is 0
4152 otherwise. All other bits of the result are undefined. The condition codes
4153 are evaluated identically to the <a href="#i_icmp">'<tt>icmp</tt>'
4154 instruction</a>.</p>
4158 <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>
4159 <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>
4163 <!-- _______________________________________________________________________ -->
4164 <div class="doc_subsubsection">
4165 <a name="i_vfcmp">'<tt>vfcmp</tt>' Instruction</a>
4167 <div class="doc_text">
4169 <pre> <result> = vfcmp <cond> <ty> <op1>, <op2></pre>
4171 <p>The '<tt>vfcmp</tt>' instruction returns an integer vector value based on
4172 element-wise comparison of its two floating point vector operands. The output
4173 elements have the same width as the input elements.</p>
4175 <p>The '<tt>vfcmp</tt>' instruction takes three operands. The first operand is
4176 the condition code indicating the kind of comparison to perform. It is not
4177 a value, just a keyword. The possible condition code are:</p>
4179 <li><tt>false</tt>: no comparison, always returns false</li>
4180 <li><tt>oeq</tt>: ordered and equal</li>
4181 <li><tt>ogt</tt>: ordered and greater than </li>
4182 <li><tt>oge</tt>: ordered and greater than or equal</li>
4183 <li><tt>olt</tt>: ordered and less than </li>
4184 <li><tt>ole</tt>: ordered and less than or equal</li>
4185 <li><tt>one</tt>: ordered and not equal</li>
4186 <li><tt>ord</tt>: ordered (no nans)</li>
4187 <li><tt>ueq</tt>: unordered or equal</li>
4188 <li><tt>ugt</tt>: unordered or greater than </li>
4189 <li><tt>uge</tt>: unordered or greater than or equal</li>
4190 <li><tt>ult</tt>: unordered or less than </li>
4191 <li><tt>ule</tt>: unordered or less than or equal</li>
4192 <li><tt>une</tt>: unordered or not equal</li>
4193 <li><tt>uno</tt>: unordered (either nans)</li>
4194 <li><tt>true</tt>: no comparison, always returns true</li>
4196 <p>The remaining two arguments must be <a href="#t_vector">vector</a> of
4197 <a href="#t_floating">floating point</a> typed. They must also be identical
4200 <p>The '<tt>vfcmp</tt>' instruction compares <tt>op1</tt> and <tt>op2</tt>
4201 according to the condition code given as <tt>cond</tt>. The comparison yields a
4202 <a href="#t_vector">vector</a> of <a href="#t_integer">integer</a> result, with
4203 an identical number of elements as the values being compared, and each element
4204 having identical with to the width of the floating point elements. The most
4205 significant bit in each element is 1 if the element-wise comparison evaluates to
4206 true, and is 0 otherwise. All other bits of the result are undefined. The
4207 condition codes are evaluated identically to the
4208 <a href="#i_fcmp">'<tt>fcmp</tt>' instruction</a>.</p>
4212 <i>; yields: result=<2 x i32> < i32 0, i32 -1 ></i>
4213 <result> = vfcmp oeq <2 x float> < float 4, float 0 >, < float 5, float 0 >
4215 <i>; yields: result=<2 x i64> < i64 -1, i64 0 ></i>
4216 <result> = vfcmp ult <2 x double> < double 1, double 2 >, < double 2, double 2>
4220 <!-- _______________________________________________________________________ -->
4221 <div class="doc_subsubsection">
4222 <a name="i_phi">'<tt>phi</tt>' Instruction</a>
4225 <div class="doc_text">
4229 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
4231 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
4232 the SSA graph representing the function.</p>
4235 <p>The type of the incoming values is specified with the first type
4236 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
4237 as arguments, with one pair for each predecessor basic block of the
4238 current block. Only values of <a href="#t_firstclass">first class</a>
4239 type may be used as the value arguments to the PHI node. Only labels
4240 may be used as the label arguments.</p>
4242 <p>There must be no non-phi instructions between the start of a basic
4243 block and the PHI instructions: i.e. PHI instructions must be first in
4248 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
4249 specified by the pair corresponding to the predecessor basic block that executed
4250 just prior to the current block.</p>
4254 Loop: ; Infinite loop that counts from 0 on up...
4255 %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
4256 %nextindvar = add i32 %indvar, 1
4261 <!-- _______________________________________________________________________ -->
4262 <div class="doc_subsubsection">
4263 <a name="i_select">'<tt>select</tt>' Instruction</a>
4266 <div class="doc_text">
4271 <result> = select <i>selty</i> <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
4273 <i>selty</i> is either i1 or {<N x i1>}
4279 The '<tt>select</tt>' instruction is used to choose one value based on a
4280 condition, without branching.
4287 The '<tt>select</tt>' instruction requires an 'i1' value or
4288 a vector of 'i1' values indicating the
4289 condition, and two values of the same <a href="#t_firstclass">first class</a>
4290 type. If the val1/val2 are vectors and
4291 the condition is a scalar, then entire vectors are selected, not
4292 individual elements.
4298 If the condition is an i1 and it evaluates to 1, the instruction returns the first
4299 value argument; otherwise, it returns the second value argument.
4302 If the condition is a vector of i1, then the value arguments must
4303 be vectors of the same size, and the selection is done element
4310 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
4315 <!-- _______________________________________________________________________ -->
4316 <div class="doc_subsubsection">
4317 <a name="i_call">'<tt>call</tt>' Instruction</a>
4320 <div class="doc_text">
4324 <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>]
4329 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
4333 <p>This instruction requires several arguments:</p>
4337 <p>The optional "tail" marker indicates whether the callee function accesses
4338 any allocas or varargs in the caller. If the "tail" marker is present, the
4339 function call is eligible for tail call optimization. Note that calls may
4340 be marked "tail" even if they do not occur before a <a
4341 href="#i_ret"><tt>ret</tt></a> instruction.</p>
4344 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
4345 convention</a> the call should use. If none is specified, the call defaults
4346 to using C calling conventions.</p>
4350 <p>The optional <a href="#paramattrs">Parameter Attributes</a> list for
4351 return values. Only '<tt>zeroext</tt>', '<tt>signext</tt>',
4352 and '<tt>inreg</tt>' attributes are valid here.</p>
4356 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
4357 the type of the return value. Functions that return no value are marked
4358 <tt><a href="#t_void">void</a></tt>.</p>
4361 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
4362 value being invoked. The argument types must match the types implied by
4363 this signature. This type can be omitted if the function is not varargs
4364 and if the function type does not return a pointer to a function.</p>
4367 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
4368 be invoked. In most cases, this is a direct function invocation, but
4369 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
4370 to function value.</p>
4373 <p>'<tt>function args</tt>': argument list whose types match the
4374 function signature argument types. All arguments must be of
4375 <a href="#t_firstclass">first class</a> type. If the function signature
4376 indicates the function accepts a variable number of arguments, the extra
4377 arguments can be specified.</p>
4380 <p>The optional <a href="#fnattrs">function attributes</a> list. Only
4381 '<tt>noreturn</tt>', '<tt>nounwind</tt>', '<tt>readonly</tt>' and
4382 '<tt>readnone</tt>' attributes are valid here.</p>
4388 <p>The '<tt>call</tt>' instruction is used to cause control flow to
4389 transfer to a specified function, with its incoming arguments bound to
4390 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
4391 instruction in the called function, control flow continues with the
4392 instruction after the function call, and the return value of the
4393 function is bound to the result argument.</p>
4398 %retval = call i32 @test(i32 %argc)
4399 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42) <i>; yields i32</i>
4400 %X = tail call i32 @foo() <i>; yields i32</i>
4401 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo() <i>; yields i32</i>
4402 call void %foo(i8 97 signext)
4404 %struct.A = type { i32, i8 }
4405 %r = call %struct.A @foo() <i>; yields { 32, i8 }</i>
4406 %gr = extractvalue %struct.A %r, 0 <i>; yields i32</i>
4407 %gr1 = extractvalue %struct.A %r, 1 <i>; yields i8</i>
4408 %Z = call void @foo() noreturn <i>; indicates that %foo never returns normally</i>
4409 %ZZ = call zeroext i32 @bar() <i>; Return value is %zero extended</i>
4414 <!-- _______________________________________________________________________ -->
4415 <div class="doc_subsubsection">
4416 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
4419 <div class="doc_text">
4424 <resultval> = va_arg <va_list*> <arglist>, <argty>
4429 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
4430 the "variable argument" area of a function call. It is used to implement the
4431 <tt>va_arg</tt> macro in C.</p>
4435 <p>This instruction takes a <tt>va_list*</tt> value and the type of
4436 the argument. It returns a value of the specified argument type and
4437 increments the <tt>va_list</tt> to point to the next argument. The
4438 actual type of <tt>va_list</tt> is target specific.</p>
4442 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
4443 type from the specified <tt>va_list</tt> and causes the
4444 <tt>va_list</tt> to point to the next argument. For more information,
4445 see the variable argument handling <a href="#int_varargs">Intrinsic
4448 <p>It is legal for this instruction to be called in a function which does not
4449 take a variable number of arguments, for example, the <tt>vfprintf</tt>
4452 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
4453 href="#intrinsics">intrinsic function</a> because it takes a type as an
4458 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
4462 <!-- *********************************************************************** -->
4463 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
4464 <!-- *********************************************************************** -->
4466 <div class="doc_text">
4468 <p>LLVM supports the notion of an "intrinsic function". These functions have
4469 well known names and semantics and are required to follow certain restrictions.
4470 Overall, these intrinsics represent an extension mechanism for the LLVM
4471 language that does not require changing all of the transformations in LLVM when
4472 adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
4474 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
4475 prefix is reserved in LLVM for intrinsic names; thus, function names may not
4476 begin with this prefix. Intrinsic functions must always be external functions:
4477 you cannot define the body of intrinsic functions. Intrinsic functions may
4478 only be used in call or invoke instructions: it is illegal to take the address
4479 of an intrinsic function. Additionally, because intrinsic functions are part
4480 of the LLVM language, it is required if any are added that they be documented
4483 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
4484 a family of functions that perform the same operation but on different data
4485 types. Because LLVM can represent over 8 million different integer types,
4486 overloading is used commonly to allow an intrinsic function to operate on any
4487 integer type. One or more of the argument types or the result type can be
4488 overloaded to accept any integer type. Argument types may also be defined as
4489 exactly matching a previous argument's type or the result type. This allows an
4490 intrinsic function which accepts multiple arguments, but needs all of them to
4491 be of the same type, to only be overloaded with respect to a single argument or
4494 <p>Overloaded intrinsics will have the names of its overloaded argument types
4495 encoded into its function name, each preceded by a period. Only those types
4496 which are overloaded result in a name suffix. Arguments whose type is matched
4497 against another type do not. For example, the <tt>llvm.ctpop</tt> function can
4498 take an integer of any width and returns an integer of exactly the same integer
4499 width. This leads to a family of functions such as
4500 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
4501 Only one type, the return type, is overloaded, and only one type suffix is
4502 required. Because the argument's type is matched against the return type, it
4503 does not require its own name suffix.</p>
4505 <p>To learn how to add an intrinsic function, please see the
4506 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
4511 <!-- ======================================================================= -->
4512 <div class="doc_subsection">
4513 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
4516 <div class="doc_text">
4518 <p>Variable argument support is defined in LLVM with the <a
4519 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
4520 intrinsic functions. These functions are related to the similarly
4521 named macros defined in the <tt><stdarg.h></tt> header file.</p>
4523 <p>All of these functions operate on arguments that use a
4524 target-specific value type "<tt>va_list</tt>". The LLVM assembly
4525 language reference manual does not define what this type is, so all
4526 transformations should be prepared to handle these functions regardless of
4529 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
4530 instruction and the variable argument handling intrinsic functions are
4533 <div class="doc_code">
4535 define i32 @test(i32 %X, ...) {
4536 ; Initialize variable argument processing
4538 %ap2 = bitcast i8** %ap to i8*
4539 call void @llvm.va_start(i8* %ap2)
4541 ; Read a single integer argument
4542 %tmp = va_arg i8** %ap, i32
4544 ; Demonstrate usage of llvm.va_copy and llvm.va_end
4546 %aq2 = bitcast i8** %aq to i8*
4547 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
4548 call void @llvm.va_end(i8* %aq2)
4550 ; Stop processing of arguments.
4551 call void @llvm.va_end(i8* %ap2)
4555 declare void @llvm.va_start(i8*)
4556 declare void @llvm.va_copy(i8*, i8*)
4557 declare void @llvm.va_end(i8*)
4563 <!-- _______________________________________________________________________ -->
4564 <div class="doc_subsubsection">
4565 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
4569 <div class="doc_text">
4571 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
4573 <p>The '<tt>llvm.va_start</tt>' intrinsic initializes
4574 <tt>*<arglist></tt> for subsequent use by <tt><a
4575 href="#i_va_arg">va_arg</a></tt>.</p>
4579 <p>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
4583 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
4584 macro available in C. In a target-dependent way, it initializes the
4585 <tt>va_list</tt> element to which the argument points, so that the next call to
4586 <tt>va_arg</tt> will produce the first variable argument passed to the function.
4587 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
4588 last argument of the function as the compiler can figure that out.</p>
4592 <!-- _______________________________________________________________________ -->
4593 <div class="doc_subsubsection">
4594 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4597 <div class="doc_text">
4599 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
4602 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
4603 which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4604 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4608 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4612 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4613 macro available in C. In a target-dependent way, it destroys the
4614 <tt>va_list</tt> element to which the argument points. Calls to <a
4615 href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4616 <tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4617 <tt>llvm.va_end</tt>.</p>
4621 <!-- _______________________________________________________________________ -->
4622 <div class="doc_subsubsection">
4623 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4626 <div class="doc_text">
4631 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
4636 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4637 from the source argument list to the destination argument list.</p>
4641 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4642 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4647 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4648 macro available in C. In a target-dependent way, it copies the source
4649 <tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4650 intrinsic is necessary because the <tt><a href="#int_va_start">
4651 llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4652 example, memory allocation.</p>
4656 <!-- ======================================================================= -->
4657 <div class="doc_subsection">
4658 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4661 <div class="doc_text">
4664 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
4665 Collection</a> (GC) requires the implementation and generation of these
4667 These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4668 stack</a>, as well as garbage collector implementations that require <a
4669 href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4670 Front-ends for type-safe garbage collected languages should generate these
4671 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4672 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4675 <p>The garbage collection intrinsics only operate on objects in the generic
4676 address space (address space zero).</p>
4680 <!-- _______________________________________________________________________ -->
4681 <div class="doc_subsubsection">
4682 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4685 <div class="doc_text">
4690 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
4695 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4696 the code generator, and allows some metadata to be associated with it.</p>
4700 <p>The first argument specifies the address of a stack object that contains the
4701 root pointer. The second pointer (which must be either a constant or a global
4702 value address) contains the meta-data to be associated with the root.</p>
4706 <p>At runtime, a call to this intrinsic stores a null pointer into the "ptrloc"
4707 location. At compile-time, the code generator generates information to allow
4708 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4709 intrinsic may only be used in a function which <a href="#gc">specifies a GC
4715 <!-- _______________________________________________________________________ -->
4716 <div class="doc_subsubsection">
4717 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4720 <div class="doc_text">
4725 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
4730 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4731 locations, allowing garbage collector implementations that require read
4736 <p>The second argument is the address to read from, which should be an address
4737 allocated from the garbage collector. The first object is a pointer to the
4738 start of the referenced object, if needed by the language runtime (otherwise
4743 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4744 instruction, but may be replaced with substantially more complex code by the
4745 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4746 may only be used in a function which <a href="#gc">specifies a GC
4752 <!-- _______________________________________________________________________ -->
4753 <div class="doc_subsubsection">
4754 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4757 <div class="doc_text">
4762 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
4767 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4768 locations, allowing garbage collector implementations that require write
4769 barriers (such as generational or reference counting collectors).</p>
4773 <p>The first argument is the reference to store, the second is the start of the
4774 object to store it to, and the third is the address of the field of Obj to
4775 store to. If the runtime does not require a pointer to the object, Obj may be
4780 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4781 instruction, but may be replaced with substantially more complex code by the
4782 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4783 may only be used in a function which <a href="#gc">specifies a GC
4790 <!-- ======================================================================= -->
4791 <div class="doc_subsection">
4792 <a name="int_codegen">Code Generator Intrinsics</a>
4795 <div class="doc_text">
4797 These intrinsics are provided by LLVM to expose special features that may only
4798 be implemented with code generator support.
4803 <!-- _______________________________________________________________________ -->
4804 <div class="doc_subsubsection">
4805 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4808 <div class="doc_text">
4812 declare i8 *@llvm.returnaddress(i32 <level>)
4818 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4819 target-specific value indicating the return address of the current function
4820 or one of its callers.
4826 The argument to this intrinsic indicates which function to return the address
4827 for. Zero indicates the calling function, one indicates its caller, etc. The
4828 argument is <b>required</b> to be a constant integer value.
4834 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4835 the return address of the specified call frame, or zero if it cannot be
4836 identified. The value returned by this intrinsic is likely to be incorrect or 0
4837 for arguments other than zero, so it should only be used for debugging purposes.
4841 Note that calling this intrinsic does not prevent function inlining or other
4842 aggressive transformations, so the value returned may not be that of the obvious
4843 source-language caller.
4848 <!-- _______________________________________________________________________ -->
4849 <div class="doc_subsubsection">
4850 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4853 <div class="doc_text">
4857 declare i8 *@llvm.frameaddress(i32 <level>)
4863 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4864 target-specific frame pointer value for the specified stack frame.
4870 The argument to this intrinsic indicates which function to return the frame
4871 pointer for. Zero indicates the calling function, one indicates its caller,
4872 etc. The argument is <b>required</b> to be a constant integer value.
4878 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4879 the frame address of the specified call frame, or zero if it cannot be
4880 identified. The value returned by this intrinsic is likely to be incorrect or 0
4881 for arguments other than zero, so it should only be used for debugging purposes.
4885 Note that calling this intrinsic does not prevent function inlining or other
4886 aggressive transformations, so the value returned may not be that of the obvious
4887 source-language caller.
4891 <!-- _______________________________________________________________________ -->
4892 <div class="doc_subsubsection">
4893 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4896 <div class="doc_text">
4900 declare i8 *@llvm.stacksave()
4906 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4907 the function stack, for use with <a href="#int_stackrestore">
4908 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4909 features like scoped automatic variable sized arrays in C99.
4915 This intrinsic returns a opaque pointer value that can be passed to <a
4916 href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4917 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4918 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4919 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4920 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4921 that were allocated after the <tt>llvm.stacksave</tt> was executed.
4926 <!-- _______________________________________________________________________ -->
4927 <div class="doc_subsubsection">
4928 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4931 <div class="doc_text">
4935 declare void @llvm.stackrestore(i8 * %ptr)
4941 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4942 the function stack to the state it was in when the corresponding <a
4943 href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4944 useful for implementing language features like scoped automatic variable sized
4951 See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4957 <!-- _______________________________________________________________________ -->
4958 <div class="doc_subsubsection">
4959 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4962 <div class="doc_text">
4966 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
4973 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4974 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4976 effect on the behavior of the program but can change its performance
4983 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4984 determining if the fetch should be for a read (0) or write (1), and
4985 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4986 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4987 <tt>locality</tt> arguments must be constant integers.
4993 This intrinsic does not modify the behavior of the program. In particular,
4994 prefetches cannot trap and do not produce a value. On targets that support this
4995 intrinsic, the prefetch can provide hints to the processor cache for better
5001 <!-- _______________________________________________________________________ -->
5002 <div class="doc_subsubsection">
5003 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
5006 <div class="doc_text">
5010 declare void @llvm.pcmarker(i32 <id>)
5017 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
5019 code to simulators and other tools. The method is target specific, but it is
5020 expected that the marker will use exported symbols to transmit the PC of the
5022 The marker makes no guarantees that it will remain with any specific instruction
5023 after optimizations. It is possible that the presence of a marker will inhibit
5024 optimizations. The intended use is to be inserted after optimizations to allow
5025 correlations of simulation runs.
5031 <tt>id</tt> is a numerical id identifying the marker.
5037 This intrinsic does not modify the behavior of the program. Backends that do not
5038 support this intrinisic may ignore it.
5043 <!-- _______________________________________________________________________ -->
5044 <div class="doc_subsubsection">
5045 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
5048 <div class="doc_text">
5052 declare i64 @llvm.readcyclecounter( )
5059 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
5060 counter register (or similar low latency, high accuracy clocks) on those targets
5061 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
5062 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
5063 should only be used for small timings.
5069 When directly supported, reading the cycle counter should not modify any memory.
5070 Implementations are allowed to either return a application specific value or a
5071 system wide value. On backends without support, this is lowered to a constant 0.
5076 <!-- ======================================================================= -->
5077 <div class="doc_subsection">
5078 <a name="int_libc">Standard C Library Intrinsics</a>
5081 <div class="doc_text">
5083 LLVM provides intrinsics for a few important standard C library functions.
5084 These intrinsics allow source-language front-ends to pass information about the
5085 alignment of the pointer arguments to the code generator, providing opportunity
5086 for more efficient code generation.
5091 <!-- _______________________________________________________________________ -->
5092 <div class="doc_subsubsection">
5093 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
5096 <div class="doc_text">
5099 <p>This is an overloaded intrinsic. You can use llvm.memcpy on any integer bit
5100 width. Not all targets support all bit widths however.</p>
5102 declare void @llvm.memcpy.i8(i8 * <dest>, i8 * <src>,
5103 i8 <len>, i32 <align>)
5104 declare void @llvm.memcpy.i16(i8 * <dest>, i8 * <src>,
5105 i16 <len>, i32 <align>)
5106 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
5107 i32 <len>, i32 <align>)
5108 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
5109 i64 <len>, i32 <align>)
5115 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5116 location to the destination location.
5120 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
5121 intrinsics do not return a value, and takes an extra alignment argument.
5127 The first argument is a pointer to the destination, the second is a pointer to
5128 the source. The third argument is an integer argument
5129 specifying the number of bytes to copy, and the fourth argument is the alignment
5130 of the source and destination locations.
5134 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5135 the caller guarantees that both the source and destination pointers are aligned
5142 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
5143 location to the destination location, which are not allowed to overlap. It
5144 copies "len" bytes of memory over. If the argument is known to be aligned to
5145 some boundary, this can be specified as the fourth argument, otherwise it should
5151 <!-- _______________________________________________________________________ -->
5152 <div class="doc_subsubsection">
5153 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
5156 <div class="doc_text">
5159 <p>This is an overloaded intrinsic. You can use llvm.memmove on any integer bit
5160 width. Not all targets support all bit widths however.</p>
5162 declare void @llvm.memmove.i8(i8 * <dest>, i8 * <src>,
5163 i8 <len>, i32 <align>)
5164 declare void @llvm.memmove.i16(i8 * <dest>, i8 * <src>,
5165 i16 <len>, i32 <align>)
5166 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
5167 i32 <len>, i32 <align>)
5168 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
5169 i64 <len>, i32 <align>)
5175 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
5176 location to the destination location. It is similar to the
5177 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
5181 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
5182 intrinsics do not return a value, and takes an extra alignment argument.
5188 The first argument is a pointer to the destination, the second is a pointer to
5189 the source. The third argument is an integer argument
5190 specifying the number of bytes to copy, and the fourth argument is the alignment
5191 of the source and destination locations.
5195 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5196 the caller guarantees that the source and destination pointers are aligned to
5203 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
5204 location to the destination location, which may overlap. It
5205 copies "len" bytes of memory over. If the argument is known to be aligned to
5206 some boundary, this can be specified as the fourth argument, otherwise it should
5212 <!-- _______________________________________________________________________ -->
5213 <div class="doc_subsubsection">
5214 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
5217 <div class="doc_text">
5220 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
5221 width. Not all targets support all bit widths however.</p>
5223 declare void @llvm.memset.i8(i8 * <dest>, i8 <val>,
5224 i8 <len>, i32 <align>)
5225 declare void @llvm.memset.i16(i8 * <dest>, i8 <val>,
5226 i16 <len>, i32 <align>)
5227 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
5228 i32 <len>, i32 <align>)
5229 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
5230 i64 <len>, i32 <align>)
5236 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
5241 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
5242 does not return a value, and takes an extra alignment argument.
5248 The first argument is a pointer to the destination to fill, the second is the
5249 byte value to fill it with, the third argument is an integer
5250 argument specifying the number of bytes to fill, and the fourth argument is the
5251 known alignment of destination location.
5255 If the call to this intrinisic has an alignment value that is not 0 or 1, then
5256 the caller guarantees that the destination pointer is aligned to that boundary.
5262 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
5264 destination location. If the argument is known to be aligned to some boundary,
5265 this can be specified as the fourth argument, otherwise it should be set to 0 or
5271 <!-- _______________________________________________________________________ -->
5272 <div class="doc_subsubsection">
5273 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
5276 <div class="doc_text">
5279 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
5280 floating point or vector of floating point type. Not all targets support all
5283 declare float @llvm.sqrt.f32(float %Val)
5284 declare double @llvm.sqrt.f64(double %Val)
5285 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
5286 declare fp128 @llvm.sqrt.f128(fp128 %Val)
5287 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
5293 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
5294 returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
5295 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
5296 negative numbers other than -0.0 (which allows for better optimization, because
5297 there is no need to worry about errno being set). <tt>llvm.sqrt(-0.0)</tt> is
5298 defined to return -0.0 like IEEE sqrt.
5304 The argument and return value are floating point numbers of the same type.
5310 This function returns the sqrt of the specified operand if it is a nonnegative
5311 floating point number.
5315 <!-- _______________________________________________________________________ -->
5316 <div class="doc_subsubsection">
5317 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
5320 <div class="doc_text">
5323 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
5324 floating point or vector of floating point type. Not all targets support all
5327 declare float @llvm.powi.f32(float %Val, i32 %power)
5328 declare double @llvm.powi.f64(double %Val, i32 %power)
5329 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
5330 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
5331 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
5337 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
5338 specified (positive or negative) power. The order of evaluation of
5339 multiplications is not defined. When a vector of floating point type is
5340 used, the second argument remains a scalar integer value.
5346 The second argument is an integer power, and the first is a value to raise to
5353 This function returns the first value raised to the second power with an
5354 unspecified sequence of rounding operations.</p>
5357 <!-- _______________________________________________________________________ -->
5358 <div class="doc_subsubsection">
5359 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
5362 <div class="doc_text">
5365 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
5366 floating point or vector of floating point type. Not all targets support all
5369 declare float @llvm.sin.f32(float %Val)
5370 declare double @llvm.sin.f64(double %Val)
5371 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
5372 declare fp128 @llvm.sin.f128(fp128 %Val)
5373 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
5379 The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
5385 The argument and return value are floating point numbers of the same type.
5391 This function returns the sine of the specified operand, returning the
5392 same values as the libm <tt>sin</tt> functions would, and handles error
5393 conditions in the same way.</p>
5396 <!-- _______________________________________________________________________ -->
5397 <div class="doc_subsubsection">
5398 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
5401 <div class="doc_text">
5404 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
5405 floating point or vector of floating point type. Not all targets support all
5408 declare float @llvm.cos.f32(float %Val)
5409 declare double @llvm.cos.f64(double %Val)
5410 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
5411 declare fp128 @llvm.cos.f128(fp128 %Val)
5412 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
5418 The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
5424 The argument and return value are floating point numbers of the same type.
5430 This function returns the cosine of the specified operand, returning the
5431 same values as the libm <tt>cos</tt> functions would, and handles error
5432 conditions in the same way.</p>
5435 <!-- _______________________________________________________________________ -->
5436 <div class="doc_subsubsection">
5437 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
5440 <div class="doc_text">
5443 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
5444 floating point or vector of floating point type. Not all targets support all
5447 declare float @llvm.pow.f32(float %Val, float %Power)
5448 declare double @llvm.pow.f64(double %Val, double %Power)
5449 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
5450 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
5451 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
5457 The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
5458 specified (positive or negative) power.
5464 The second argument is a floating point power, and the first is a value to
5465 raise to that power.
5471 This function returns the first value raised to the second power,
5473 same values as the libm <tt>pow</tt> functions would, and handles error
5474 conditions in the same way.</p>
5478 <!-- ======================================================================= -->
5479 <div class="doc_subsection">
5480 <a name="int_manip">Bit Manipulation Intrinsics</a>
5483 <div class="doc_text">
5485 LLVM provides intrinsics for a few important bit manipulation operations.
5486 These allow efficient code generation for some algorithms.
5491 <!-- _______________________________________________________________________ -->
5492 <div class="doc_subsubsection">
5493 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
5496 <div class="doc_text">
5499 <p>This is an overloaded intrinsic function. You can use bswap on any integer
5500 type that is an even number of bytes (i.e. BitWidth % 16 == 0).</p>
5502 declare i16 @llvm.bswap.i16(i16 <id>)
5503 declare i32 @llvm.bswap.i32(i32 <id>)
5504 declare i64 @llvm.bswap.i64(i64 <id>)
5510 The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
5511 values with an even number of bytes (positive multiple of 16 bits). These are
5512 useful for performing operations on data that is not in the target's native
5519 The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
5520 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
5521 intrinsic returns an i32 value that has the four bytes of the input i32
5522 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
5523 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
5524 <tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
5525 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
5530 <!-- _______________________________________________________________________ -->
5531 <div class="doc_subsubsection">
5532 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
5535 <div class="doc_text">
5538 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
5539 width. Not all targets support all bit widths however.</p>
5541 declare i8 @llvm.ctpop.i8 (i8 <src>)
5542 declare i16 @llvm.ctpop.i16(i16 <src>)
5543 declare i32 @llvm.ctpop.i32(i32 <src>)
5544 declare i64 @llvm.ctpop.i64(i64 <src>)
5545 declare i256 @llvm.ctpop.i256(i256 <src>)
5551 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
5558 The only argument is the value to be counted. The argument may be of any
5559 integer type. The return type must match the argument type.
5565 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
5569 <!-- _______________________________________________________________________ -->
5570 <div class="doc_subsubsection">
5571 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
5574 <div class="doc_text">
5577 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
5578 integer bit width. Not all targets support all bit widths however.</p>
5580 declare i8 @llvm.ctlz.i8 (i8 <src>)
5581 declare i16 @llvm.ctlz.i16(i16 <src>)
5582 declare i32 @llvm.ctlz.i32(i32 <src>)
5583 declare i64 @llvm.ctlz.i64(i64 <src>)
5584 declare i256 @llvm.ctlz.i256(i256 <src>)
5590 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
5591 leading zeros in a variable.
5597 The only argument is the value to be counted. The argument may be of any
5598 integer type. The return type must match the argument type.
5604 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
5605 in a variable. If the src == 0 then the result is the size in bits of the type
5606 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
5612 <!-- _______________________________________________________________________ -->
5613 <div class="doc_subsubsection">
5614 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
5617 <div class="doc_text">
5620 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5621 integer bit width. Not all targets support all bit widths however.</p>
5623 declare i8 @llvm.cttz.i8 (i8 <src>)
5624 declare i16 @llvm.cttz.i16(i16 <src>)
5625 declare i32 @llvm.cttz.i32(i32 <src>)
5626 declare i64 @llvm.cttz.i64(i64 <src>)
5627 declare i256 @llvm.cttz.i256(i256 <src>)
5633 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5640 The only argument is the value to be counted. The argument may be of any
5641 integer type. The return type must match the argument type.
5647 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5648 in a variable. If the src == 0 then the result is the size in bits of the type
5649 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5653 <!-- _______________________________________________________________________ -->
5654 <div class="doc_subsubsection">
5655 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5658 <div class="doc_text">
5661 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
5662 on any integer bit width.</p>
5664 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5665 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
5669 <p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5670 range of bits from an integer value and returns them in the same bit width as
5671 the original value.</p>
5674 <p>The first argument, <tt>%val</tt> and the result may be integer types of
5675 any bit width but they must have the same bit width. The second and third
5676 arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5679 <p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5680 of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5681 <tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5682 operates in forward mode.</p>
5683 <p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5684 right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5685 only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5687 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5688 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5689 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5690 to determine the number of bits to retain.</li>
5691 <li>A mask of the retained bits is created by shifting a -1 value.</li>
5692 <li>The mask is ANDed with <tt>%val</tt> to produce the result.</li>
5694 <p>In reverse mode, a similar computation is made except that the bits are
5695 returned in the reverse order. So, for example, if <tt>X</tt> has the value
5696 <tt>i16 0x0ACF (101011001111)</tt> and we apply
5697 <tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5698 <tt>i16 0x0026 (000000100110)</tt>.</p>
5701 <div class="doc_subsubsection">
5702 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5705 <div class="doc_text">
5708 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
5709 on any integer bit width.</p>
5711 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5712 declare i29 @llvm.part.set.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
5716 <p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5717 of bits in an integer value with another integer value. It returns the integer
5718 with the replaced bits.</p>
5721 <p>The first argument, <tt>%val</tt> and the result may be integer types of
5722 any bit width but they must have the same bit width. <tt>%val</tt> is the value
5723 whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5724 integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5725 type since they specify only a bit index.</p>
5728 <p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5729 of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5730 <tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5731 operates in forward mode.</p>
5732 <p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5733 truncating it down to the size of the replacement area or zero extending it
5734 up to that size.</p>
5735 <p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5736 are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5737 in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5738 to the <tt>%hi</tt>th bit.</p>
5739 <p>In reverse mode, a similar computation is made except that the bits are
5740 reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5741 <tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.</p>
5744 llvm.part.set(0xFFFF, 0, 4, 7) -> 0xFF0F
5745 llvm.part.set(0xFFFF, 0, 7, 4) -> 0xFF0F
5746 llvm.part.set(0xFFFF, 1, 7, 4) -> 0xFF8F
5747 llvm.part.set(0xFFFF, F, 8, 3) -> 0xFFE7
5748 llvm.part.set(0xFFFF, 0, 3, 8) -> 0xFE07
5752 <!-- ======================================================================= -->
5753 <div class="doc_subsection">
5754 <a name="int_debugger">Debugger Intrinsics</a>
5757 <div class="doc_text">
5759 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5760 are described in the <a
5761 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5762 Debugging</a> document.
5767 <!-- ======================================================================= -->
5768 <div class="doc_subsection">
5769 <a name="int_eh">Exception Handling Intrinsics</a>
5772 <div class="doc_text">
5773 <p> The LLVM exception handling intrinsics (which all start with
5774 <tt>llvm.eh.</tt> prefix), are described in the <a
5775 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5776 Handling</a> document. </p>
5779 <!-- ======================================================================= -->
5780 <div class="doc_subsection">
5781 <a name="int_trampoline">Trampoline Intrinsic</a>
5784 <div class="doc_text">
5786 This intrinsic makes it possible to excise one parameter, marked with
5787 the <tt>nest</tt> attribute, from a function. The result is a callable
5788 function pointer lacking the nest parameter - the caller does not need
5789 to provide a value for it. Instead, the value to use is stored in
5790 advance in a "trampoline", a block of memory usually allocated
5791 on the stack, which also contains code to splice the nest value into the
5792 argument list. This is used to implement the GCC nested function address
5796 For example, if the function is
5797 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
5798 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:</p>
5800 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5801 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5802 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5803 %fp = bitcast i8* %p to i32 (i32, i32)*
5805 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5806 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
5809 <!-- _______________________________________________________________________ -->
5810 <div class="doc_subsubsection">
5811 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5813 <div class="doc_text">
5816 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
5820 This fills the memory pointed to by <tt>tramp</tt> with code
5821 and returns a function pointer suitable for executing it.
5825 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5826 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5827 and sufficiently aligned block of memory; this memory is written to by the
5828 intrinsic. Note that the size and the alignment are target-specific - LLVM
5829 currently provides no portable way of determining them, so a front-end that
5830 generates this intrinsic needs to have some target-specific knowledge.
5831 The <tt>func</tt> argument must hold a function bitcast to an <tt>i8*</tt>.
5835 The block of memory pointed to by <tt>tramp</tt> is filled with target
5836 dependent code, turning it into a function. A pointer to this function is
5837 returned, but needs to be bitcast to an
5838 <a href="#int_trampoline">appropriate function pointer type</a>
5839 before being called. The new function's signature is the same as that of
5840 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5841 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5842 of pointer type. Calling the new function is equivalent to calling
5843 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5844 missing <tt>nest</tt> argument. If, after calling
5845 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5846 modified, then the effect of any later call to the returned function pointer is
5851 <!-- ======================================================================= -->
5852 <div class="doc_subsection">
5853 <a name="int_atomics">Atomic Operations and Synchronization Intrinsics</a>
5856 <div class="doc_text">
5858 These intrinsic functions expand the "universal IR" of LLVM to represent
5859 hardware constructs for atomic operations and memory synchronization. This
5860 provides an interface to the hardware, not an interface to the programmer. It
5861 is aimed at a low enough level to allow any programming models or APIs
5862 (Application Programming Interfaces) which
5863 need atomic behaviors to map cleanly onto it. It is also modeled primarily on
5864 hardware behavior. Just as hardware provides a "universal IR" for source
5865 languages, it also provides a starting point for developing a "universal"
5866 atomic operation and synchronization IR.
5869 These do <em>not</em> form an API such as high-level threading libraries,
5870 software transaction memory systems, atomic primitives, and intrinsic
5871 functions as found in BSD, GNU libc, atomic_ops, APR, and other system and
5872 application libraries. The hardware interface provided by LLVM should allow
5873 a clean implementation of all of these APIs and parallel programming models.
5874 No one model or paradigm should be selected above others unless the hardware
5875 itself ubiquitously does so.
5880 <!-- _______________________________________________________________________ -->
5881 <div class="doc_subsubsection">
5882 <a name="int_memory_barrier">'<tt>llvm.memory.barrier</tt>' Intrinsic</a>
5884 <div class="doc_text">
5887 declare void @llvm.memory.barrier( i1 <ll>, i1 <ls>, i1 <sl>, i1 <ss>,
5893 The <tt>llvm.memory.barrier</tt> intrinsic guarantees ordering between
5894 specific pairs of memory access types.
5898 The <tt>llvm.memory.barrier</tt> intrinsic requires five boolean arguments.
5899 The first four arguments enables a specific barrier as listed below. The fith
5900 argument specifies that the barrier applies to io or device or uncached memory.
5904 <li><tt>ll</tt>: load-load barrier</li>
5905 <li><tt>ls</tt>: load-store barrier</li>
5906 <li><tt>sl</tt>: store-load barrier</li>
5907 <li><tt>ss</tt>: store-store barrier</li>
5908 <li><tt>device</tt>: barrier applies to device and uncached memory also.</li>
5912 This intrinsic causes the system to enforce some ordering constraints upon
5913 the loads and stores of the program. This barrier does not indicate
5914 <em>when</em> any events will occur, it only enforces an <em>order</em> in
5915 which they occur. For any of the specified pairs of load and store operations
5916 (f.ex. load-load, or store-load), all of the first operations preceding the
5917 barrier will complete before any of the second operations succeeding the
5918 barrier begin. Specifically the semantics for each pairing is as follows:
5921 <li><tt>ll</tt>: All loads before the barrier must complete before any load
5922 after the barrier begins.</li>
5924 <li><tt>ls</tt>: All loads before the barrier must complete before any
5925 store after the barrier begins.</li>
5926 <li><tt>ss</tt>: All stores before the barrier must complete before any
5927 store after the barrier begins.</li>
5928 <li><tt>sl</tt>: All stores before the barrier must complete before any
5929 load after the barrier begins.</li>
5932 These semantics are applied with a logical "and" behavior when more than one
5933 is enabled in a single memory barrier intrinsic.
5936 Backends may implement stronger barriers than those requested when they do not
5937 support as fine grained a barrier as requested. Some architectures do not
5938 need all types of barriers and on such architectures, these become noops.
5945 %result1 = load i32* %ptr <i>; yields {i32}:result1 = 4</i>
5946 call void @llvm.memory.barrier( i1 false, i1 true, i1 false, i1 false )
5947 <i>; guarantee the above finishes</i>
5948 store i32 8, %ptr <i>; before this begins</i>
5952 <!-- _______________________________________________________________________ -->
5953 <div class="doc_subsubsection">
5954 <a name="int_atomic_cmp_swap">'<tt>llvm.atomic.cmp.swap.*</tt>' Intrinsic</a>
5956 <div class="doc_text">
5959 This is an overloaded intrinsic. You can use <tt>llvm.atomic.cmp.swap</tt> on
5960 any integer bit width and for different address spaces. Not all targets
5961 support all bit widths however.</p>
5964 declare i8 @llvm.atomic.cmp.swap.i8.p0i8( i8* <ptr>, i8 <cmp>, i8 <val> )
5965 declare i16 @llvm.atomic.cmp.swap.i16.p0i16( i16* <ptr>, i16 <cmp>, i16 <val> )
5966 declare i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* <ptr>, i32 <cmp>, i32 <val> )
5967 declare i64 @llvm.atomic.cmp.swap.i64.p0i64( i64* <ptr>, i64 <cmp>, i64 <val> )
5972 This loads a value in memory and compares it to a given value. If they are
5973 equal, it stores a new value into the memory.
5977 The <tt>llvm.atomic.cmp.swap</tt> intrinsic takes three arguments. The result as
5978 well as both <tt>cmp</tt> and <tt>val</tt> must be integer values with the
5979 same bit width. The <tt>ptr</tt> argument must be a pointer to a value of
5980 this integer type. While any bit width integer may be used, targets may only
5981 lower representations they support in hardware.
5986 This entire intrinsic must be executed atomically. It first loads the value
5987 in memory pointed to by <tt>ptr</tt> and compares it with the value
5988 <tt>cmp</tt>. If they are equal, <tt>val</tt> is stored into the memory. The
5989 loaded value is yielded in all cases. This provides the equivalent of an
5990 atomic compare-and-swap operation within the SSA framework.
5998 %val1 = add i32 4, 4
5999 %result1 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 4, %val1 )
6000 <i>; yields {i32}:result1 = 4</i>
6001 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6002 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6004 %val2 = add i32 1, 1
6005 %result2 = call i32 @llvm.atomic.cmp.swap.i32.p0i32( i32* %ptr, i32 5, %val2 )
6006 <i>; yields {i32}:result2 = 8</i>
6007 %stored2 = icmp eq i32 %result2, 5 <i>; yields {i1}:stored2 = false</i>
6009 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 8</i>
6013 <!-- _______________________________________________________________________ -->
6014 <div class="doc_subsubsection">
6015 <a name="int_atomic_swap">'<tt>llvm.atomic.swap.*</tt>' Intrinsic</a>
6017 <div class="doc_text">
6021 This is an overloaded intrinsic. You can use <tt>llvm.atomic.swap</tt> on any
6022 integer bit width. Not all targets support all bit widths however.</p>
6024 declare i8 @llvm.atomic.swap.i8.p0i8( i8* <ptr>, i8 <val> )
6025 declare i16 @llvm.atomic.swap.i16.p0i16( i16* <ptr>, i16 <val> )
6026 declare i32 @llvm.atomic.swap.i32.p0i32( i32* <ptr>, i32 <val> )
6027 declare i64 @llvm.atomic.swap.i64.p0i64( i64* <ptr>, i64 <val> )
6032 This intrinsic loads the value stored in memory at <tt>ptr</tt> and yields
6033 the value from memory. It then stores the value in <tt>val</tt> in the memory
6039 The <tt>llvm.atomic.swap</tt> intrinsic takes two arguments. Both the
6040 <tt>val</tt> argument and the result must be integers of the same bit width.
6041 The first argument, <tt>ptr</tt>, must be a pointer to a value of this
6042 integer type. The targets may only lower integer representations they
6047 This intrinsic loads the value pointed to by <tt>ptr</tt>, yields it, and
6048 stores <tt>val</tt> back into <tt>ptr</tt> atomically. This provides the
6049 equivalent of an atomic swap operation within the SSA framework.
6057 %val1 = add i32 4, 4
6058 %result1 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val1 )
6059 <i>; yields {i32}:result1 = 4</i>
6060 %stored1 = icmp eq i32 %result1, 4 <i>; yields {i1}:stored1 = true</i>
6061 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 8</i>
6063 %val2 = add i32 1, 1
6064 %result2 = call i32 @llvm.atomic.swap.i32.p0i32( i32* %ptr, i32 %val2 )
6065 <i>; yields {i32}:result2 = 8</i>
6067 %stored2 = icmp eq i32 %result2, 8 <i>; yields {i1}:stored2 = true</i>
6068 %memval2 = load i32* %ptr <i>; yields {i32}:memval2 = 2</i>
6072 <!-- _______________________________________________________________________ -->
6073 <div class="doc_subsubsection">
6074 <a name="int_atomic_load_add">'<tt>llvm.atomic.load.add.*</tt>' Intrinsic</a>
6077 <div class="doc_text">
6080 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.add</tt> on any
6081 integer bit width. Not all targets support all bit widths however.</p>
6083 declare i8 @llvm.atomic.load.add.i8..p0i8( i8* <ptr>, i8 <delta> )
6084 declare i16 @llvm.atomic.load.add.i16..p0i16( i16* <ptr>, i16 <delta> )
6085 declare i32 @llvm.atomic.load.add.i32..p0i32( i32* <ptr>, i32 <delta> )
6086 declare i64 @llvm.atomic.load.add.i64..p0i64( i64* <ptr>, i64 <delta> )
6091 This intrinsic adds <tt>delta</tt> to the value stored in memory at
6092 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6097 The intrinsic takes two arguments, the first a pointer to an integer value
6098 and the second an integer value. The result is also an integer value. These
6099 integer types can have any bit width, but they must all have the same bit
6100 width. The targets may only lower integer representations they support.
6104 This intrinsic does a series of operations atomically. It first loads the
6105 value stored at <tt>ptr</tt>. It then adds <tt>delta</tt>, stores the result
6106 to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6113 %result1 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 4 )
6114 <i>; yields {i32}:result1 = 4</i>
6115 %result2 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 2 )
6116 <i>; yields {i32}:result2 = 8</i>
6117 %result3 = call i32 @llvm.atomic.load.add.i32.p0i32( i32* %ptr, i32 5 )
6118 <i>; yields {i32}:result3 = 10</i>
6119 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 15</i>
6123 <!-- _______________________________________________________________________ -->
6124 <div class="doc_subsubsection">
6125 <a name="int_atomic_load_sub">'<tt>llvm.atomic.load.sub.*</tt>' Intrinsic</a>
6128 <div class="doc_text">
6131 This is an overloaded intrinsic. You can use <tt>llvm.atomic.load.sub</tt> on
6132 any integer bit width and for different address spaces. Not all targets
6133 support all bit widths however.</p>
6135 declare i8 @llvm.atomic.load.sub.i8.p0i32( i8* <ptr>, i8 <delta> )
6136 declare i16 @llvm.atomic.load.sub.i16.p0i32( i16* <ptr>, i16 <delta> )
6137 declare i32 @llvm.atomic.load.sub.i32.p0i32( i32* <ptr>, i32 <delta> )
6138 declare i64 @llvm.atomic.load.sub.i64.p0i32( i64* <ptr>, i64 <delta> )
6143 This intrinsic subtracts <tt>delta</tt> to the value stored in memory at
6144 <tt>ptr</tt>. It yields the original value at <tt>ptr</tt>.
6149 The intrinsic takes two arguments, the first a pointer to an integer value
6150 and the second an integer value. The result is also an integer value. These
6151 integer types can have any bit width, but they must all have the same bit
6152 width. The targets may only lower integer representations they support.
6156 This intrinsic does a series of operations atomically. It first loads the
6157 value stored at <tt>ptr</tt>. It then subtracts <tt>delta</tt>, stores the
6158 result to <tt>ptr</tt>. It yields the original value stored at <tt>ptr</tt>.
6165 %result1 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 4 )
6166 <i>; yields {i32}:result1 = 8</i>
6167 %result2 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 2 )
6168 <i>; yields {i32}:result2 = 4</i>
6169 %result3 = call i32 @llvm.atomic.load.sub.i32.p0i32( i32* %ptr, i32 5 )
6170 <i>; yields {i32}:result3 = 2</i>
6171 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = -3</i>
6175 <!-- _______________________________________________________________________ -->
6176 <div class="doc_subsubsection">
6177 <a name="int_atomic_load_and">'<tt>llvm.atomic.load.and.*</tt>' Intrinsic</a><br>
6178 <a name="int_atomic_load_nand">'<tt>llvm.atomic.load.nand.*</tt>' Intrinsic</a><br>
6179 <a name="int_atomic_load_or">'<tt>llvm.atomic.load.or.*</tt>' Intrinsic</a><br>
6180 <a name="int_atomic_load_xor">'<tt>llvm.atomic.load.xor.*</tt>' Intrinsic</a><br>
6183 <div class="doc_text">
6186 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_and</tt>,
6187 <tt>llvm.atomic.load_nand</tt>, <tt>llvm.atomic.load_or</tt>, and
6188 <tt>llvm.atomic.load_xor</tt> on any integer bit width and for different
6189 address spaces. Not all targets support all bit widths however.</p>
6191 declare i8 @llvm.atomic.load.and.i8.p0i8( i8* <ptr>, i8 <delta> )
6192 declare i16 @llvm.atomic.load.and.i16.p0i16( i16* <ptr>, i16 <delta> )
6193 declare i32 @llvm.atomic.load.and.i32.p0i32( i32* <ptr>, i32 <delta> )
6194 declare i64 @llvm.atomic.load.and.i64.p0i64( i64* <ptr>, i64 <delta> )
6199 declare i8 @llvm.atomic.load.or.i8.p0i8( i8* <ptr>, i8 <delta> )
6200 declare i16 @llvm.atomic.load.or.i16.p0i16( i16* <ptr>, i16 <delta> )
6201 declare i32 @llvm.atomic.load.or.i32.p0i32( i32* <ptr>, i32 <delta> )
6202 declare i64 @llvm.atomic.load.or.i64.p0i64( i64* <ptr>, i64 <delta> )
6207 declare i8 @llvm.atomic.load.nand.i8.p0i32( i8* <ptr>, i8 <delta> )
6208 declare i16 @llvm.atomic.load.nand.i16.p0i32( i16* <ptr>, i16 <delta> )
6209 declare i32 @llvm.atomic.load.nand.i32.p0i32( i32* <ptr>, i32 <delta> )
6210 declare i64 @llvm.atomic.load.nand.i64.p0i32( i64* <ptr>, i64 <delta> )
6215 declare i8 @llvm.atomic.load.xor.i8.p0i32( i8* <ptr>, i8 <delta> )
6216 declare i16 @llvm.atomic.load.xor.i16.p0i32( i16* <ptr>, i16 <delta> )
6217 declare i32 @llvm.atomic.load.xor.i32.p0i32( i32* <ptr>, i32 <delta> )
6218 declare i64 @llvm.atomic.load.xor.i64.p0i32( i64* <ptr>, i64 <delta> )
6223 These intrinsics bitwise the operation (and, nand, or, xor) <tt>delta</tt> to
6224 the value stored in memory at <tt>ptr</tt>. It yields the original value
6230 These intrinsics take two arguments, the first a pointer to an integer value
6231 and the second an integer value. The result is also an integer value. These
6232 integer types can have any bit width, but they must all have the same bit
6233 width. The targets may only lower integer representations they support.
6237 These intrinsics does a series of operations atomically. They first load the
6238 value stored at <tt>ptr</tt>. They then do the bitwise operation
6239 <tt>delta</tt>, store the result to <tt>ptr</tt>. They yield the original
6240 value stored at <tt>ptr</tt>.
6246 store i32 0x0F0F, %ptr
6247 %result0 = call i32 @llvm.atomic.load.nand.i32.p0i32( i32* %ptr, i32 0xFF )
6248 <i>; yields {i32}:result0 = 0x0F0F</i>
6249 %result1 = call i32 @llvm.atomic.load.and.i32.p0i32( i32* %ptr, i32 0xFF )
6250 <i>; yields {i32}:result1 = 0xFFFFFFF0</i>
6251 %result2 = call i32 @llvm.atomic.load.or.i32.p0i32( i32* %ptr, i32 0F )
6252 <i>; yields {i32}:result2 = 0xF0</i>
6253 %result3 = call i32 @llvm.atomic.load.xor.i32.p0i32( i32* %ptr, i32 0F )
6254 <i>; yields {i32}:result3 = FF</i>
6255 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = F0</i>
6260 <!-- _______________________________________________________________________ -->
6261 <div class="doc_subsubsection">
6262 <a name="int_atomic_load_max">'<tt>llvm.atomic.load.max.*</tt>' Intrinsic</a><br>
6263 <a name="int_atomic_load_min">'<tt>llvm.atomic.load.min.*</tt>' Intrinsic</a><br>
6264 <a name="int_atomic_load_umax">'<tt>llvm.atomic.load.umax.*</tt>' Intrinsic</a><br>
6265 <a name="int_atomic_load_umin">'<tt>llvm.atomic.load.umin.*</tt>' Intrinsic</a><br>
6268 <div class="doc_text">
6271 These are overloaded intrinsics. You can use <tt>llvm.atomic.load_max</tt>,
6272 <tt>llvm.atomic.load_min</tt>, <tt>llvm.atomic.load_umax</tt>, and
6273 <tt>llvm.atomic.load_umin</tt> on any integer bit width and for different
6274 address spaces. Not all targets
6275 support all bit widths however.</p>
6277 declare i8 @llvm.atomic.load.max.i8.p0i8( i8* <ptr>, i8 <delta> )
6278 declare i16 @llvm.atomic.load.max.i16.p0i16( i16* <ptr>, i16 <delta> )
6279 declare i32 @llvm.atomic.load.max.i32.p0i32( i32* <ptr>, i32 <delta> )
6280 declare i64 @llvm.atomic.load.max.i64.p0i64( i64* <ptr>, i64 <delta> )
6285 declare i8 @llvm.atomic.load.min.i8.p0i8( i8* <ptr>, i8 <delta> )
6286 declare i16 @llvm.atomic.load.min.i16.p0i16( i16* <ptr>, i16 <delta> )
6287 declare i32 @llvm.atomic.load.min.i32..p0i32( i32* <ptr>, i32 <delta> )
6288 declare i64 @llvm.atomic.load.min.i64..p0i64( i64* <ptr>, i64 <delta> )
6293 declare i8 @llvm.atomic.load.umax.i8.p0i8( i8* <ptr>, i8 <delta> )
6294 declare i16 @llvm.atomic.load.umax.i16.p0i16( i16* <ptr>, i16 <delta> )
6295 declare i32 @llvm.atomic.load.umax.i32.p0i32( i32* <ptr>, i32 <delta> )
6296 declare i64 @llvm.atomic.load.umax.i64.p0i64( i64* <ptr>, i64 <delta> )
6301 declare i8 @llvm.atomic.load.umin.i8..p0i8( i8* <ptr>, i8 <delta> )
6302 declare i16 @llvm.atomic.load.umin.i16.p0i16( i16* <ptr>, i16 <delta> )
6303 declare i32 @llvm.atomic.load.umin.i32..p0i32( i32* <ptr>, i32 <delta> )
6304 declare i64 @llvm.atomic.load.umin.i64..p0i64( i64* <ptr>, i64 <delta> )
6309 These intrinsics takes the signed or unsigned minimum or maximum of
6310 <tt>delta</tt> and the value stored in memory at <tt>ptr</tt>. It yields the
6311 original value at <tt>ptr</tt>.
6316 These intrinsics take two arguments, the first a pointer to an integer value
6317 and the second an integer value. The result is also an integer value. These
6318 integer types can have any bit width, but they must all have the same bit
6319 width. The targets may only lower integer representations they support.
6323 These intrinsics does a series of operations atomically. They first load the
6324 value stored at <tt>ptr</tt>. They then do the signed or unsigned min or max
6325 <tt>delta</tt> and the value, store the result to <tt>ptr</tt>. They yield
6326 the original value stored at <tt>ptr</tt>.
6333 %result0 = call i32 @llvm.atomic.load.min.i32.p0i32( i32* %ptr, i32 -2 )
6334 <i>; yields {i32}:result0 = 7</i>
6335 %result1 = call i32 @llvm.atomic.load.max.i32.p0i32( i32* %ptr, i32 8 )
6336 <i>; yields {i32}:result1 = -2</i>
6337 %result2 = call i32 @llvm.atomic.load.umin.i32.p0i32( i32* %ptr, i32 10 )
6338 <i>; yields {i32}:result2 = 8</i>
6339 %result3 = call i32 @llvm.atomic.load.umax.i32.p0i32( i32* %ptr, i32 30 )
6340 <i>; yields {i32}:result3 = 8</i>
6341 %memval1 = load i32* %ptr <i>; yields {i32}:memval1 = 30</i>
6345 <!-- ======================================================================= -->
6346 <div class="doc_subsection">
6347 <a name="int_general">General Intrinsics</a>
6350 <div class="doc_text">
6351 <p> This class of intrinsics is designed to be generic and has
6352 no specific purpose. </p>
6355 <!-- _______________________________________________________________________ -->
6356 <div class="doc_subsubsection">
6357 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
6360 <div class="doc_text">
6364 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
6370 The '<tt>llvm.var.annotation</tt>' intrinsic
6376 The first argument is a pointer to a value, the second is a pointer to a
6377 global string, the third is a pointer to a global string which is the source
6378 file name, and the last argument is the line number.
6384 This intrinsic allows annotation of local variables with arbitrary strings.
6385 This can be useful for special purpose optimizations that want to look for these
6386 annotations. These have no other defined use, they are ignored by code
6387 generation and optimization.
6391 <!-- _______________________________________________________________________ -->
6392 <div class="doc_subsubsection">
6393 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
6396 <div class="doc_text">
6399 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
6400 any integer bit width.
6403 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
6404 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
6405 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
6406 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
6407 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
6413 The '<tt>llvm.annotation</tt>' intrinsic.
6419 The first argument is an integer value (result of some expression),
6420 the second is a pointer to a global string, the third is a pointer to a global
6421 string which is the source file name, and the last argument is the line number.
6422 It returns the value of the first argument.
6428 This intrinsic allows annotations to be put on arbitrary expressions
6429 with arbitrary strings. This can be useful for special purpose optimizations
6430 that want to look for these annotations. These have no other defined use, they
6431 are ignored by code generation and optimization.
6435 <!-- _______________________________________________________________________ -->
6436 <div class="doc_subsubsection">
6437 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
6440 <div class="doc_text">
6444 declare void @llvm.trap()
6450 The '<tt>llvm.trap</tt>' intrinsic
6462 This intrinsics is lowered to the target dependent trap instruction. If the
6463 target does not have a trap instruction, this intrinsic will be lowered to the
6464 call of the abort() function.
6468 <!-- _______________________________________________________________________ -->
6469 <div class="doc_subsubsection">
6470 <a name="int_stackprotector">'<tt>llvm.stackprotector</tt>' Intrinsic</a>
6472 <div class="doc_text">
6475 declare void @llvm.stackprotector( i8* <guard>, i8** <slot> )
6480 The <tt>llvm.stackprotector</tt> intrinsic takes the <tt>guard</tt> and stores
6481 it onto the stack at <tt>slot</tt>. The stack slot is adjusted to ensure that
6482 it is placed on the stack before local variables.
6486 The <tt>llvm.stackprotector</tt> intrinsic requires two pointer arguments. The
6487 first argument is the value loaded from the stack guard
6488 <tt>@__stack_chk_guard</tt>. The second variable is an <tt>alloca</tt> that
6489 has enough space to hold the value of the guard.
6493 This intrinsic causes the prologue/epilogue inserter to force the position of
6494 the <tt>AllocaInst</tt> stack slot to be before local variables on the
6495 stack. This is to ensure that if a local variable on the stack is overwritten,
6496 it will destroy the value of the guard. When the function exits, the guard on
6497 the stack is checked against the original guard. If they're different, then
6498 the program aborts by calling the <tt>__stack_chk_fail()</tt> function.
6502 <!-- *********************************************************************** -->
6505 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
6506 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
6507 <a href="http://validator.w3.org/check/referer"><img
6508 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
6510 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
6511 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
6512 Last modified: $Date$