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>
28 <li><a href="#paramattrs">Parameter Attributes</a></li>
29 <li><a href="#gc">Garbage Collector Names</a></li>
30 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
31 <li><a href="#datalayout">Data Layout</a></li>
34 <li><a href="#typesystem">Type System</a>
36 <li><a href="#t_classifications">Type Classifications</a></li>
37 <li><a href="#t_primitive">Primitive Types</a>
39 <li><a href="#t_floating">Floating Point Types</a></li>
40 <li><a href="#t_void">Void Type</a></li>
41 <li><a href="#t_label">Label Type</a></li>
44 <li><a href="#t_derived">Derived Types</a>
46 <li><a href="#t_integer">Integer Type</a></li>
47 <li><a href="#t_array">Array Type</a></li>
48 <li><a href="#t_function">Function Type</a></li>
49 <li><a href="#t_pointer">Pointer Type</a></li>
50 <li><a href="#t_struct">Structure Type</a></li>
51 <li><a href="#t_pstruct">Packed Structure Type</a></li>
52 <li><a href="#t_vector">Vector Type</a></li>
53 <li><a href="#t_opaque">Opaque Type</a></li>
58 <li><a href="#constants">Constants</a>
60 <li><a href="#simpleconstants">Simple Constants</a>
61 <li><a href="#aggregateconstants">Aggregate Constants</a>
62 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
63 <li><a href="#undefvalues">Undefined Values</a>
64 <li><a href="#constantexprs">Constant Expressions</a>
67 <li><a href="#othervalues">Other Values</a>
69 <li><a href="#inlineasm">Inline Assembler Expressions</a>
72 <li><a href="#instref">Instruction Reference</a>
74 <li><a href="#terminators">Terminator Instructions</a>
76 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
77 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
78 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
79 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
80 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
81 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
84 <li><a href="#binaryops">Binary Operations</a>
86 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
87 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
88 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
89 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
90 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
91 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
92 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
93 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
94 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
97 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
99 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
100 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
101 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
102 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
103 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
104 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
107 <li><a href="#vectorops">Vector Operations</a>
109 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
110 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
111 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
114 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
116 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
117 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
118 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
119 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
120 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
121 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
124 <li><a href="#convertops">Conversion Operations</a>
126 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
127 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
128 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
129 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
130 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
131 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
132 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
133 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
134 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
135 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
136 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
137 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
139 <li><a href="#otherops">Other Operations</a>
141 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
142 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
143 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
144 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
145 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
146 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
151 <li><a href="#intrinsics">Intrinsic Functions</a>
153 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
155 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
156 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
157 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
160 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
162 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
163 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
164 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
167 <li><a href="#int_codegen">Code Generator Intrinsics</a>
169 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
170 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
171 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
172 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
173 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
174 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
175 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
178 <li><a href="#int_libc">Standard C Library Intrinsics</a>
180 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
181 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
182 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
183 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
184 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
185 <li><a href="#int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a></li>
186 <li><a href="#int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a></li>
187 <li><a href="#int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a></li>
190 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
192 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
193 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
194 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
195 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
196 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
197 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
200 <li><a href="#int_debugger">Debugger intrinsics</a></li>
201 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
202 <li><a href="#int_trampoline">Trampoline Intrinsic</a>
204 <li><a href="#int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a></li>
207 <li><a href="#int_general">General intrinsics</a>
209 <li><a href="#int_var_annotation">
210 <tt>llvm.var.annotation</tt>' Intrinsic</a></li>
211 <li><a href="#int_annotation">
212 <tt>llvm.annotation.*</tt>' Intrinsic</a></li>
213 <li><a href="#int_trap">
214 <tt>llvm.trap</tt>' Intrinsic</a></li>
221 <div class="doc_author">
222 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
223 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
226 <!-- *********************************************************************** -->
227 <div class="doc_section"> <a name="abstract">Abstract </a></div>
228 <!-- *********************************************************************** -->
230 <div class="doc_text">
231 <p>This document is a reference manual for the LLVM assembly language.
232 LLVM is an SSA based representation that provides type safety,
233 low-level operations, flexibility, and the capability of representing
234 'all' high-level languages cleanly. It is the common code
235 representation used throughout all phases of the LLVM compilation
239 <!-- *********************************************************************** -->
240 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
241 <!-- *********************************************************************** -->
243 <div class="doc_text">
245 <p>The LLVM code representation is designed to be used in three
246 different forms: as an in-memory compiler IR, as an on-disk bitcode
247 representation (suitable for fast loading by a Just-In-Time compiler),
248 and as a human readable assembly language representation. This allows
249 LLVM to provide a powerful intermediate representation for efficient
250 compiler transformations and analysis, while providing a natural means
251 to debug and visualize the transformations. The three different forms
252 of LLVM are all equivalent. This document describes the human readable
253 representation and notation.</p>
255 <p>The LLVM representation aims to be light-weight and low-level
256 while being expressive, typed, and extensible at the same time. It
257 aims to be a "universal IR" of sorts, by being at a low enough level
258 that high-level ideas may be cleanly mapped to it (similar to how
259 microprocessors are "universal IR's", allowing many source languages to
260 be mapped to them). By providing type information, LLVM can be used as
261 the target of optimizations: for example, through pointer analysis, it
262 can be proven that a C automatic variable is never accessed outside of
263 the current function... allowing it to be promoted to a simple SSA
264 value instead of a memory location.</p>
268 <!-- _______________________________________________________________________ -->
269 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
271 <div class="doc_text">
273 <p>It is important to note that this document describes 'well formed'
274 LLVM assembly language. There is a difference between what the parser
275 accepts and what is considered 'well formed'. For example, the
276 following instruction is syntactically okay, but not well formed:</p>
278 <div class="doc_code">
280 %x = <a href="#i_add">add</a> i32 1, %x
284 <p>...because the definition of <tt>%x</tt> does not dominate all of
285 its uses. The LLVM infrastructure provides a verification pass that may
286 be used to verify that an LLVM module is well formed. This pass is
287 automatically run by the parser after parsing input assembly and by
288 the optimizer before it outputs bitcode. The violations pointed out
289 by the verifier pass indicate bugs in transformation passes or input to
293 <!-- Describe the typesetting conventions here. -->
295 <!-- *********************************************************************** -->
296 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
297 <!-- *********************************************************************** -->
299 <div class="doc_text">
301 <p>LLVM identifiers come in two basic types: global and local. Global
302 identifiers (functions, global variables) begin with the @ character. Local
303 identifiers (register names, types) begin with the % character. Additionally,
304 there are three different formats for identifiers, for different purposes:
307 <li>Named values are represented as a string of characters with their prefix.
308 For example, %foo, @DivisionByZero, %a.really.long.identifier. The actual
309 regular expression used is '<tt>[%@][a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
310 Identifiers which require other characters in their names can be surrounded
311 with quotes. In this way, anything except a <tt>"</tt> character can
312 be used in a named value.</li>
314 <li>Unnamed values are represented as an unsigned numeric value with their
315 prefix. For example, %12, @2, %44.</li>
317 <li>Constants, which are described in a <a href="#constants">section about
318 constants</a>, below.</li>
321 <p>LLVM requires that values start with a prefix for two reasons: Compilers
322 don't need to worry about name clashes with reserved words, and the set of
323 reserved words may be expanded in the future without penalty. Additionally,
324 unnamed identifiers allow a compiler to quickly come up with a temporary
325 variable without having to avoid symbol table conflicts.</p>
327 <p>Reserved words in LLVM are very similar to reserved words in other
328 languages. There are keywords for different opcodes
329 ('<tt><a href="#i_add">add</a></tt>',
330 '<tt><a href="#i_bitcast">bitcast</a></tt>',
331 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
332 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
333 and others. These reserved words cannot conflict with variable names, because
334 none of them start with a prefix character ('%' or '@').</p>
336 <p>Here is an example of LLVM code to multiply the integer variable
337 '<tt>%X</tt>' by 8:</p>
341 <div class="doc_code">
343 %result = <a href="#i_mul">mul</a> i32 %X, 8
347 <p>After strength reduction:</p>
349 <div class="doc_code">
351 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
355 <p>And the hard way:</p>
357 <div class="doc_code">
359 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
360 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
361 %result = <a href="#i_add">add</a> i32 %1, %1
365 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
366 important lexical features of LLVM:</p>
370 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
373 <li>Unnamed temporaries are created when the result of a computation is not
374 assigned to a named value.</li>
376 <li>Unnamed temporaries are numbered sequentially</li>
380 <p>...and it also shows a convention that we follow in this document. When
381 demonstrating instructions, we will follow an instruction with a comment that
382 defines the type and name of value produced. Comments are shown in italic
387 <!-- *********************************************************************** -->
388 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
389 <!-- *********************************************************************** -->
391 <!-- ======================================================================= -->
392 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
395 <div class="doc_text">
397 <p>LLVM programs are composed of "Module"s, each of which is a
398 translation unit of the input programs. Each module consists of
399 functions, global variables, and symbol table entries. Modules may be
400 combined together with the LLVM linker, which merges function (and
401 global variable) definitions, resolves forward declarations, and merges
402 symbol table entries. Here is an example of the "hello world" module:</p>
404 <div class="doc_code">
405 <pre><i>; Declare the string constant as a global constant...</i>
406 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
407 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
409 <i>; External declaration of the puts function</i>
410 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
412 <i>; Definition of main function</i>
413 define i32 @main() { <i>; i32()* </i>
414 <i>; Convert [13x i8 ]* to i8 *...</i>
416 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
418 <i>; Call puts function to write out the string to stdout...</i>
420 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
422 href="#i_ret">ret</a> i32 0<br>}<br>
426 <p>This example is made up of a <a href="#globalvars">global variable</a>
427 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
428 function, and a <a href="#functionstructure">function definition</a>
429 for "<tt>main</tt>".</p>
431 <p>In general, a module is made up of a list of global values,
432 where both functions and global variables are global values. Global values are
433 represented by a pointer to a memory location (in this case, a pointer to an
434 array of char, and a pointer to a function), and have one of the following <a
435 href="#linkage">linkage types</a>.</p>
439 <!-- ======================================================================= -->
440 <div class="doc_subsection">
441 <a name="linkage">Linkage Types</a>
444 <div class="doc_text">
447 All Global Variables and Functions have one of the following types of linkage:
452 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
454 <dd>Global values with internal linkage are only directly accessible by
455 objects in the current module. In particular, linking code into a module with
456 an internal global value may cause the internal to be renamed as necessary to
457 avoid collisions. Because the symbol is internal to the module, all
458 references can be updated. This corresponds to the notion of the
459 '<tt>static</tt>' keyword in C.
462 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
464 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
465 the same name when linkage occurs. This is typically used to implement
466 inline functions, templates, or other code which must be generated in each
467 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
468 allowed to be discarded.
471 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
473 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
474 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
475 used for globals that may be emitted in multiple translation units, but that
476 are not guaranteed to be emitted into every translation unit that uses them.
477 One example of this are common globals in C, such as "<tt>int X;</tt>" at
481 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
483 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
484 pointer to array type. When two global variables with appending linkage are
485 linked together, the two global arrays are appended together. This is the
486 LLVM, typesafe, equivalent of having the system linker append together
487 "sections" with identical names when .o files are linked.
490 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
491 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
492 until linked, if not linked, the symbol becomes null instead of being an
496 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
498 <dd>If none of the above identifiers are used, the global is externally
499 visible, meaning that it participates in linkage and can be used to resolve
500 external symbol references.
505 The next two types of linkage are targeted for Microsoft Windows platform
506 only. They are designed to support importing (exporting) symbols from (to)
511 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
513 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
514 or variable via a global pointer to a pointer that is set up by the DLL
515 exporting the symbol. On Microsoft Windows targets, the pointer name is
516 formed by combining <code>_imp__</code> and the function or variable name.
519 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
521 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
522 pointer to a pointer in a DLL, so that it can be referenced with the
523 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
524 name is formed by combining <code>_imp__</code> and the function or variable
530 <p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
531 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
532 variable and was linked with this one, one of the two would be renamed,
533 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
534 external (i.e., lacking any linkage declarations), they are accessible
535 outside of the current module.</p>
536 <p>It is illegal for a function <i>declaration</i>
537 to have any linkage type other than "externally visible", <tt>dllimport</tt>,
538 or <tt>extern_weak</tt>.</p>
539 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
543 <!-- ======================================================================= -->
544 <div class="doc_subsection">
545 <a name="callingconv">Calling Conventions</a>
548 <div class="doc_text">
550 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
551 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
552 specified for the call. The calling convention of any pair of dynamic
553 caller/callee must match, or the behavior of the program is undefined. The
554 following calling conventions are supported by LLVM, and more may be added in
558 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
560 <dd>This calling convention (the default if no other calling convention is
561 specified) matches the target C calling conventions. This calling convention
562 supports varargs function calls and tolerates some mismatch in the declared
563 prototype and implemented declaration of the function (as does normal C).
566 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
568 <dd>This calling convention attempts to make calls as fast as possible
569 (e.g. by passing things in registers). This calling convention allows the
570 target to use whatever tricks it wants to produce fast code for the target,
571 without having to conform to an externally specified ABI. Implementations of
572 this convention should allow arbitrary tail call optimization to be supported.
573 This calling convention does not support varargs and requires the prototype of
574 all callees to exactly match the prototype of the function definition.
577 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
579 <dd>This calling convention attempts to make code in the caller as efficient
580 as possible under the assumption that the call is not commonly executed. As
581 such, these calls often preserve all registers so that the call does not break
582 any live ranges in the caller side. This calling convention does not support
583 varargs and requires the prototype of all callees to exactly match the
584 prototype of the function definition.
587 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
589 <dd>Any calling convention may be specified by number, allowing
590 target-specific calling conventions to be used. Target specific calling
591 conventions start at 64.
595 <p>More calling conventions can be added/defined on an as-needed basis, to
596 support pascal conventions or any other well-known target-independent
601 <!-- ======================================================================= -->
602 <div class="doc_subsection">
603 <a name="visibility">Visibility Styles</a>
606 <div class="doc_text">
609 All Global Variables and Functions have one of the following visibility styles:
613 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
615 <dd>On ELF, default visibility means that the declaration is visible to other
616 modules and, in shared libraries, means that the declared entity may be
617 overridden. On Darwin, default visibility means that the declaration is
618 visible to other modules. Default visibility corresponds to "external
619 linkage" in the language.
622 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
624 <dd>Two declarations of an object with hidden visibility refer to the same
625 object if they are in the same shared object. Usually, hidden visibility
626 indicates that the symbol will not be placed into the dynamic symbol table,
627 so no other module (executable or shared library) can reference it
631 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
633 <dd>On ELF, protected visibility indicates that the symbol will be placed in
634 the dynamic symbol table, but that references within the defining module will
635 bind to the local symbol. That is, the symbol cannot be overridden by another
642 <!-- ======================================================================= -->
643 <div class="doc_subsection">
644 <a name="globalvars">Global Variables</a>
647 <div class="doc_text">
649 <p>Global variables define regions of memory allocated at compilation time
650 instead of run-time. Global variables may optionally be initialized, may have
651 an explicit section to be placed in, and may have an optional explicit alignment
652 specified. A variable may be defined as "thread_local", which means that it
653 will not be shared by threads (each thread will have a separated copy of the
654 variable). A variable may be defined as a global "constant," which indicates
655 that the contents of the variable will <b>never</b> be modified (enabling better
656 optimization, allowing the global data to be placed in the read-only section of
657 an executable, etc). Note that variables that need runtime initialization
658 cannot be marked "constant" as there is a store to the variable.</p>
661 LLVM explicitly allows <em>declarations</em> of global variables to be marked
662 constant, even if the final definition of the global is not. This capability
663 can be used to enable slightly better optimization of the program, but requires
664 the language definition to guarantee that optimizations based on the
665 'constantness' are valid for the translation units that do not include the
669 <p>As SSA values, global variables define pointer values that are in
670 scope (i.e. they dominate) all basic blocks in the program. Global
671 variables always define a pointer to their "content" type because they
672 describe a region of memory, and all memory objects in LLVM are
673 accessed through pointers.</p>
675 <p>A global variable may be declared to reside in a target-specifc numbered
676 address space. For targets that support them, address spaces may affect how
677 optimizations are performed and/or what target instructions are used to access
678 the variable. The default address space is zero. The address space qualifier
679 must precede any other attributes.</p>
681 <p>LLVM allows an explicit section to be specified for globals. If the target
682 supports it, it will emit globals to the section specified.</p>
684 <p>An explicit alignment may be specified for a global. If not present, or if
685 the alignment is set to zero, the alignment of the global is set by the target
686 to whatever it feels convenient. If an explicit alignment is specified, the
687 global is forced to have at least that much alignment. All alignments must be
690 <p>For example, the following defines a global in a numbered address space with
691 an initializer, section, and alignment:</p>
693 <div class="doc_code">
695 @G = constant float 1.0 addrspace(5), section "foo", align 4
702 <!-- ======================================================================= -->
703 <div class="doc_subsection">
704 <a name="functionstructure">Functions</a>
707 <div class="doc_text">
709 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
710 an optional <a href="#linkage">linkage type</a>, an optional
711 <a href="#visibility">visibility style</a>, an optional
712 <a href="#callingconv">calling convention</a>, a return type, an optional
713 <a href="#paramattrs">parameter attribute</a> for the return type, a function
714 name, a (possibly empty) argument list (each with optional
715 <a href="#paramattrs">parameter attributes</a>), an optional section, an
716 optional alignment, an optional <a href="#gc">garbage collector name</a>, an
717 opening curly brace, a list of basic blocks, and a closing curly brace.
719 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
720 optional <a href="#linkage">linkage type</a>, an optional
721 <a href="#visibility">visibility style</a>, an optional
722 <a href="#callingconv">calling convention</a>, a return type, an optional
723 <a href="#paramattrs">parameter attribute</a> for the return type, a function
724 name, a possibly empty list of arguments, an optional alignment, and an optional
725 <a href="#gc">garbage collector name</a>.</p>
727 <p>A function definition contains a list of basic blocks, forming the CFG for
728 the function. Each basic block may optionally start with a label (giving the
729 basic block a symbol table entry), contains a list of instructions, and ends
730 with a <a href="#terminators">terminator</a> instruction (such as a branch or
731 function return).</p>
733 <p>The first basic block in a function is special in two ways: it is immediately
734 executed on entrance to the function, and it is not allowed to have predecessor
735 basic blocks (i.e. there can not be any branches to the entry block of a
736 function). Because the block can have no predecessors, it also cannot have any
737 <a href="#i_phi">PHI nodes</a>.</p>
739 <p>LLVM allows an explicit section to be specified for functions. If the target
740 supports it, it will emit functions to the section specified.</p>
742 <p>An explicit alignment may be specified for a function. If not present, or if
743 the alignment is set to zero, the alignment of the function is set by the target
744 to whatever it feels convenient. If an explicit alignment is specified, the
745 function is forced to have at least that much alignment. All alignments must be
751 <!-- ======================================================================= -->
752 <div class="doc_subsection">
753 <a name="aliasstructure">Aliases</a>
755 <div class="doc_text">
756 <p>Aliases act as "second name" for the aliasee value (which can be either
757 function or global variable or bitcast of global value). Aliases may have an
758 optional <a href="#linkage">linkage type</a>, and an
759 optional <a href="#visibility">visibility style</a>.</p>
763 <div class="doc_code">
765 @<Name> = [Linkage] [Visibility] alias <AliaseeTy> @<Aliasee>
773 <!-- ======================================================================= -->
774 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
775 <div class="doc_text">
776 <p>The return type and each parameter of a function type may have a set of
777 <i>parameter attributes</i> associated with them. Parameter attributes are
778 used to communicate additional information about the result or parameters of
779 a function. Parameter attributes are considered to be part of the function,
780 not of the function type, so functions with different parameter attributes
781 can have the same function type.</p>
783 <p>Parameter attributes are simple keywords that follow the type specified. If
784 multiple parameter attributes are needed, they are space separated. For
787 <div class="doc_code">
789 declare i32 @printf(i8* noalias , ...) nounwind
790 declare i32 @atoi(i8*) nounwind readonly
794 <p>Note that any attributes for the function result (<tt>nounwind</tt>,
795 <tt>readonly</tt>) come immediately after the argument list.</p>
797 <p>Currently, only the following parameter attributes are defined:</p>
799 <dt><tt>zeroext</tt></dt>
800 <dd>This indicates that the parameter should be zero extended just before
801 a call to this function.</dd>
803 <dt><tt>signext</tt></dt>
804 <dd>This indicates that the parameter should be sign extended just before
805 a call to this function.</dd>
807 <dt><tt>inreg</tt></dt>
808 <dd>This indicates that the parameter should be placed in register (if
809 possible) during assembling function call. Support for this attribute is
812 <dt><tt>byval</tt></dt>
813 <dd>This indicates that the pointer parameter should really be passed by
814 value to the function. The attribute implies that a hidden copy of the
815 pointee is made between the caller and the callee, so the callee is unable
816 to modify the value in the callee. This attribute is only valid on llvm
817 pointer arguments. It is generally used to pass structs and arrays by
818 value, but is also valid on scalars (even though this is silly).</dd>
820 <dt><tt>sret</tt></dt>
821 <dd>This indicates that the parameter specifies the address of a structure
822 that is the return value of the function in the source program.</dd>
824 <dt><tt>noalias</tt></dt>
825 <dd>This indicates that the parameter not alias any other object or any
826 other "noalias" objects during the function call.
828 <dt><tt>noreturn</tt></dt>
829 <dd>This function attribute indicates that the function never returns. This
830 indicates to LLVM that every call to this function should be treated as if
831 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
833 <dt><tt>nounwind</tt></dt>
834 <dd>This function attribute indicates that the function type does not use
835 the unwind instruction and does not allow stack unwinding to propagate
838 <dt><tt>nest</tt></dt>
839 <dd>This indicates that the parameter can be excised using the
840 <a href="#int_trampoline">trampoline intrinsics</a>.</dd>
841 <dt><tt>readonly</tt></dt>
842 <dd>This function attribute indicates that the function has no side-effects
843 except for producing a return value or throwing an exception. The value
844 returned must only depend on the function arguments and/or global variables.
845 It may use values obtained by dereferencing pointers.</dd>
846 <dt><tt>readnone</tt></dt>
847 <dd>A <tt>readnone</tt> function has the same restrictions as a <tt>readonly</tt>
848 function, but in addition it is not allowed to dereference any pointer arguments
854 <!-- ======================================================================= -->
855 <div class="doc_subsection">
856 <a name="gc">Garbage Collector Names</a>
859 <div class="doc_text">
860 <p>Each function may specify a garbage collector name, which is simply a
863 <div class="doc_code"><pre
864 >define void @f() gc "name" { ...</pre></div>
866 <p>The compiler declares the supported values of <i>name</i>. Specifying a
867 collector which will cause the compiler to alter its output in order to support
868 the named garbage collection algorithm.</p>
871 <!-- ======================================================================= -->
872 <div class="doc_subsection">
873 <a name="moduleasm">Module-Level Inline Assembly</a>
876 <div class="doc_text">
878 Modules may contain "module-level inline asm" blocks, which corresponds to the
879 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
880 LLVM and treated as a single unit, but may be separated in the .ll file if
881 desired. The syntax is very simple:
884 <div class="doc_code">
886 module asm "inline asm code goes here"
887 module asm "more can go here"
891 <p>The strings can contain any character by escaping non-printable characters.
892 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
897 The inline asm code is simply printed to the machine code .s file when
898 assembly code is generated.
902 <!-- ======================================================================= -->
903 <div class="doc_subsection">
904 <a name="datalayout">Data Layout</a>
907 <div class="doc_text">
908 <p>A module may specify a target specific data layout string that specifies how
909 data is to be laid out in memory. The syntax for the data layout is simply:</p>
910 <pre> target datalayout = "<i>layout specification</i>"</pre>
911 <p>The <i>layout specification</i> consists of a list of specifications
912 separated by the minus sign character ('-'). Each specification starts with a
913 letter and may include other information after the letter to define some
914 aspect of the data layout. The specifications accepted are as follows: </p>
917 <dd>Specifies that the target lays out data in big-endian form. That is, the
918 bits with the most significance have the lowest address location.</dd>
920 <dd>Specifies that hte target lays out data in little-endian form. That is,
921 the bits with the least significance have the lowest address location.</dd>
922 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
923 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
924 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
925 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
927 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
928 <dd>This specifies the alignment for an integer type of a given bit
929 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
930 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
931 <dd>This specifies the alignment for a vector type of a given bit
933 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
934 <dd>This specifies the alignment for a floating point type of a given bit
935 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
937 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
938 <dd>This specifies the alignment for an aggregate type of a given bit
941 <p>When constructing the data layout for a given target, LLVM starts with a
942 default set of specifications which are then (possibly) overriden by the
943 specifications in the <tt>datalayout</tt> keyword. The default specifications
944 are given in this list:</p>
946 <li><tt>E</tt> - big endian</li>
947 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
948 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
949 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
950 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
951 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
952 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
953 alignment of 64-bits</li>
954 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
955 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
956 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
957 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
958 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
960 <p>When llvm is determining the alignment for a given type, it uses the
963 <li>If the type sought is an exact match for one of the specifications, that
964 specification is used.</li>
965 <li>If no match is found, and the type sought is an integer type, then the
966 smallest integer type that is larger than the bitwidth of the sought type is
967 used. If none of the specifications are larger than the bitwidth then the the
968 largest integer type is used. For example, given the default specifications
969 above, the i7 type will use the alignment of i8 (next largest) while both
970 i65 and i256 will use the alignment of i64 (largest specified).</li>
971 <li>If no match is found, and the type sought is a vector type, then the
972 largest vector type that is smaller than the sought vector type will be used
973 as a fall back. This happens because <128 x double> can be implemented in
974 terms of 64 <2 x double>, for example.</li>
978 <!-- *********************************************************************** -->
979 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
980 <!-- *********************************************************************** -->
982 <div class="doc_text">
984 <p>The LLVM type system is one of the most important features of the
985 intermediate representation. Being typed enables a number of
986 optimizations to be performed on the IR directly, without having to do
987 extra analyses on the side before the transformation. A strong type
988 system makes it easier to read the generated code and enables novel
989 analyses and transformations that are not feasible to perform on normal
990 three address code representations.</p>
994 <!-- ======================================================================= -->
995 <div class="doc_subsection"> <a name="t_classifications">Type
996 Classifications</a> </div>
997 <div class="doc_text">
998 <p>The types fall into a few useful
1001 <table border="1" cellspacing="0" cellpadding="4">
1003 <tr><th>Classification</th><th>Types</th></tr>
1005 <td><a href="#t_integer">integer</a></td>
1006 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
1009 <td><a href="#t_floating">floating point</a></td>
1010 <td><tt>float, double, x86_fp80, fp128, ppc_fp128</tt></td>
1013 <td><a name="t_firstclass">first class</a></td>
1014 <td><a href="#t_integer">integer</a>,
1015 <a href="#t_floating">floating point</a>,
1016 <a href="#t_pointer">pointer</a>,
1017 <a href="#t_vector">vector</a>
1021 <td><a href="#t_primitive">primitive</a></td>
1022 <td><a href="#t_label">label</a>,
1023 <a href="#t_void">void</a>,
1024 <a href="#t_integer">integer</a>,
1025 <a href="#t_floating">floating point</a>.</td>
1028 <td><a href="#t_derived">derived</a></td>
1029 <td><a href="#t_integer">integer</a>,
1030 <a href="#t_array">array</a>,
1031 <a href="#t_function">function</a>,
1032 <a href="#t_pointer">pointer</a>,
1033 <a href="#t_struct">structure</a>,
1034 <a href="#t_pstruct">packed structure</a>,
1035 <a href="#t_vector">vector</a>,
1036 <a href="#t_opaque">opaque</a>.
1041 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
1042 most important. Values of these types are the only ones which can be
1043 produced by instructions, passed as arguments, or used as operands to
1044 instructions. This means that all structures and arrays must be
1045 manipulated either by pointer or by component.</p>
1048 <!-- ======================================================================= -->
1049 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
1051 <div class="doc_text">
1052 <p>The primitive types are the fundamental building blocks of the LLVM
1057 <!-- _______________________________________________________________________ -->
1058 <div class="doc_subsubsection"> <a name="t_floating">Floating Point Types</a> </div>
1060 <div class="doc_text">
1063 <tr><th>Type</th><th>Description</th></tr>
1064 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
1065 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
1066 <tr><td><tt>fp128</tt></td><td>128-bit floating point value (112-bit mantissa)</td></tr>
1067 <tr><td><tt>x86_fp80</tt></td><td>80-bit floating point value (X87)</td></tr>
1068 <tr><td><tt>ppc_fp128</tt></td><td>128-bit floating point value (two 64-bits)</td></tr>
1073 <!-- _______________________________________________________________________ -->
1074 <div class="doc_subsubsection"> <a name="t_void">Void Type</a> </div>
1076 <div class="doc_text">
1078 <p>The void type does not represent any value and has no size.</p>
1087 <!-- _______________________________________________________________________ -->
1088 <div class="doc_subsubsection"> <a name="t_label">Label Type</a> </div>
1090 <div class="doc_text">
1092 <p>The label type represents code labels.</p>
1102 <!-- ======================================================================= -->
1103 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
1105 <div class="doc_text">
1107 <p>The real power in LLVM comes from the derived types in the system.
1108 This is what allows a programmer to represent arrays, functions,
1109 pointers, and other useful types. Note that these derived types may be
1110 recursive: For example, it is possible to have a two dimensional array.</p>
1114 <!-- _______________________________________________________________________ -->
1115 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1117 <div class="doc_text">
1120 <p>The integer type is a very simple derived type that simply specifies an
1121 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1122 2^23-1 (about 8 million) can be specified.</p>
1130 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1134 <table class="layout">
1137 <td><tt>i1</tt></td>
1138 <td>a single-bit integer.</td>
1140 <td><tt>i32</tt></td>
1141 <td>a 32-bit integer.</td>
1143 <td><tt>i1942652</tt></td>
1144 <td>a really big integer of over 1 million bits.</td>
1150 <!-- _______________________________________________________________________ -->
1151 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1153 <div class="doc_text">
1157 <p>The array type is a very simple derived type that arranges elements
1158 sequentially in memory. The array type requires a size (number of
1159 elements) and an underlying data type.</p>
1164 [<# elements> x <elementtype>]
1167 <p>The number of elements is a constant integer value; elementtype may
1168 be any type with a size.</p>
1171 <table class="layout">
1173 <td class="left"><tt>[40 x i32]</tt></td>
1174 <td class="left">Array of 40 32-bit integer values.</td>
1177 <td class="left"><tt>[41 x i32]</tt></td>
1178 <td class="left">Array of 41 32-bit integer values.</td>
1181 <td class="left"><tt>[4 x i8]</tt></td>
1182 <td class="left">Array of 4 8-bit integer values.</td>
1185 <p>Here are some examples of multidimensional arrays:</p>
1186 <table class="layout">
1188 <td class="left"><tt>[3 x [4 x i32]]</tt></td>
1189 <td class="left">3x4 array of 32-bit integer values.</td>
1192 <td class="left"><tt>[12 x [10 x float]]</tt></td>
1193 <td class="left">12x10 array of single precision floating point values.</td>
1196 <td class="left"><tt>[2 x [3 x [4 x i16]]]</tt></td>
1197 <td class="left">2x3x4 array of 16-bit integer values.</td>
1201 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1202 length array. Normally, accesses past the end of an array are undefined in
1203 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1204 As a special case, however, zero length arrays are recognized to be variable
1205 length. This allows implementation of 'pascal style arrays' with the LLVM
1206 type "{ i32, [0 x float]}", for example.</p>
1210 <!-- _______________________________________________________________________ -->
1211 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1212 <div class="doc_text">
1214 <p>The function type can be thought of as a function signature. It
1215 consists of a return type and a list of formal parameter types.
1216 Function types are usually used to build virtual function tables
1217 (which are structures of pointers to functions), for indirect function
1218 calls, and when defining a function.</p>
1220 The return type of a function type cannot be an aggregate type.
1223 <pre> <returntype> (<parameter list>)<br></pre>
1224 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1225 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1226 which indicates that the function takes a variable number of arguments.
1227 Variable argument functions can access their arguments with the <a
1228 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1230 <table class="layout">
1232 <td class="left"><tt>i32 (i32)</tt></td>
1233 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1235 </tr><tr class="layout">
1236 <td class="left"><tt>float (i16 signext, i32 *) *
1238 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1239 an <tt>i16</tt> that should be sign extended and a
1240 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1243 </tr><tr class="layout">
1244 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1245 <td class="left">A vararg function that takes at least one
1246 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1247 which returns an integer. This is the signature for <tt>printf</tt> in
1254 <!-- _______________________________________________________________________ -->
1255 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1256 <div class="doc_text">
1258 <p>The structure type is used to represent a collection of data members
1259 together in memory. The packing of the field types is defined to match
1260 the ABI of the underlying processor. The elements of a structure may
1261 be any type that has a size.</p>
1262 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1263 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1264 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1267 <pre> { <type list> }<br></pre>
1269 <table class="layout">
1271 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1272 <td class="left">A triple of three <tt>i32</tt> values</td>
1273 </tr><tr class="layout">
1274 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1275 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1276 second element is a <a href="#t_pointer">pointer</a> to a
1277 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1278 an <tt>i32</tt>.</td>
1283 <!-- _______________________________________________________________________ -->
1284 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1286 <div class="doc_text">
1288 <p>The packed structure type is used to represent a collection of data members
1289 together in memory. There is no padding between fields. Further, the alignment
1290 of a packed structure is 1 byte. The elements of a packed structure may
1291 be any type that has a size.</p>
1292 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1293 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1294 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1297 <pre> < { <type list> } > <br></pre>
1299 <table class="layout">
1301 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1302 <td class="left">A triple of three <tt>i32</tt> values</td>
1303 </tr><tr class="layout">
1304 <td class="left"><tt>< { float, i32 (i32)* } ></tt></td>
1305 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1306 second element is a <a href="#t_pointer">pointer</a> to a
1307 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1308 an <tt>i32</tt>.</td>
1313 <!-- _______________________________________________________________________ -->
1314 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1315 <div class="doc_text">
1317 <p>As in many languages, the pointer type represents a pointer or
1318 reference to another object, which must live in memory. Pointer types may have
1319 an optional address space attribute defining the target-specific numbered
1320 address space where the pointed-to object resides. The default address space is
1323 <pre> <type> *<br></pre>
1325 <table class="layout">
1327 <td class="left"><tt>[4x i32]*</tt></td>
1328 <td class="left">A <a href="#t_pointer">pointer</a> to <a
1329 href="#t_array">array</a> of four <tt>i32</tt> values.</td>
1332 <td class="left"><tt>i32 (i32 *) *</tt></td>
1333 <td class="left"> A <a href="#t_pointer">pointer</a> to a <a
1334 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1338 <td class="left"><tt>i32 addrspace(5)*</tt></td>
1339 <td class="left">A <a href="#t_pointer">pointer</a> to an <tt>i32</tt> value
1340 that resides in address space #5.</td>
1345 <!-- _______________________________________________________________________ -->
1346 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1347 <div class="doc_text">
1351 <p>A vector type is a simple derived type that represents a vector
1352 of elements. Vector types are used when multiple primitive data
1353 are operated in parallel using a single instruction (SIMD).
1354 A vector type requires a size (number of
1355 elements) and an underlying primitive data type. Vectors must have a power
1356 of two length (1, 2, 4, 8, 16 ...). Vector types are
1357 considered <a href="#t_firstclass">first class</a>.</p>
1362 < <# elements> x <elementtype> >
1365 <p>The number of elements is a constant integer value; elementtype may
1366 be any integer or floating point type.</p>
1370 <table class="layout">
1372 <td class="left"><tt><4 x i32></tt></td>
1373 <td class="left">Vector of 4 32-bit integer values.</td>
1376 <td class="left"><tt><8 x float></tt></td>
1377 <td class="left">Vector of 8 32-bit floating-point values.</td>
1380 <td class="left"><tt><2 x i64></tt></td>
1381 <td class="left">Vector of 2 64-bit integer values.</td>
1386 <!-- _______________________________________________________________________ -->
1387 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1388 <div class="doc_text">
1392 <p>Opaque types are used to represent unknown types in the system. This
1393 corresponds (for example) to the C notion of a forward declared structure type.
1394 In LLVM, opaque types can eventually be resolved to any type (not just a
1395 structure type).</p>
1405 <table class="layout">
1407 <td class="left"><tt>opaque</tt></td>
1408 <td class="left">An opaque type.</td>
1414 <!-- *********************************************************************** -->
1415 <div class="doc_section"> <a name="constants">Constants</a> </div>
1416 <!-- *********************************************************************** -->
1418 <div class="doc_text">
1420 <p>LLVM has several different basic types of constants. This section describes
1421 them all and their syntax.</p>
1425 <!-- ======================================================================= -->
1426 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1428 <div class="doc_text">
1431 <dt><b>Boolean constants</b></dt>
1433 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1434 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1437 <dt><b>Integer constants</b></dt>
1439 <dd>Standard integers (such as '4') are constants of the <a
1440 href="#t_integer">integer</a> type. Negative numbers may be used with
1444 <dt><b>Floating point constants</b></dt>
1446 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1447 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1448 notation (see below). Floating point constants must have a <a
1449 href="#t_floating">floating point</a> type. </dd>
1451 <dt><b>Null pointer constants</b></dt>
1453 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1454 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1458 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1459 of floating point constants. For example, the form '<tt>double
1460 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1461 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1462 (and the only time that they are generated by the disassembler) is when a
1463 floating point constant must be emitted but it cannot be represented as a
1464 decimal floating point number. For example, NaN's, infinities, and other
1465 special values are represented in their IEEE hexadecimal format so that
1466 assembly and disassembly do not cause any bits to change in the constants.</p>
1470 <!-- ======================================================================= -->
1471 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1474 <div class="doc_text">
1475 <p>Aggregate constants arise from aggregation of simple constants
1476 and smaller aggregate constants.</p>
1479 <dt><b>Structure constants</b></dt>
1481 <dd>Structure constants are represented with notation similar to structure
1482 type definitions (a comma separated list of elements, surrounded by braces
1483 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* @G }</tt>",
1484 where "<tt>@G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1485 must have <a href="#t_struct">structure type</a>, and the number and
1486 types of elements must match those specified by the type.
1489 <dt><b>Array constants</b></dt>
1491 <dd>Array constants are represented with notation similar to array type
1492 definitions (a comma separated list of elements, surrounded by square brackets
1493 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1494 constants must have <a href="#t_array">array type</a>, and the number and
1495 types of elements must match those specified by the type.
1498 <dt><b>Vector constants</b></dt>
1500 <dd>Vector constants are represented with notation similar to vector type
1501 definitions (a comma separated list of elements, surrounded by
1502 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1503 i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
1504 href="#t_vector">vector type</a>, and the number and types of elements must
1505 match those specified by the type.
1508 <dt><b>Zero initialization</b></dt>
1510 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1511 value to zero of <em>any</em> type, including scalar and aggregate types.
1512 This is often used to avoid having to print large zero initializers (e.g. for
1513 large arrays) and is always exactly equivalent to using explicit zero
1520 <!-- ======================================================================= -->
1521 <div class="doc_subsection">
1522 <a name="globalconstants">Global Variable and Function Addresses</a>
1525 <div class="doc_text">
1527 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1528 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1529 constants. These constants are explicitly referenced when the <a
1530 href="#identifiers">identifier for the global</a> is used and always have <a
1531 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1534 <div class="doc_code">
1538 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1544 <!-- ======================================================================= -->
1545 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1546 <div class="doc_text">
1547 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1548 no specific value. Undefined values may be of any type and be used anywhere
1549 a constant is permitted.</p>
1551 <p>Undefined values indicate to the compiler that the program is well defined
1552 no matter what value is used, giving the compiler more freedom to optimize.
1556 <!-- ======================================================================= -->
1557 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1560 <div class="doc_text">
1562 <p>Constant expressions are used to allow expressions involving other constants
1563 to be used as constants. Constant expressions may be of any <a
1564 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1565 that does not have side effects (e.g. load and call are not supported). The
1566 following is the syntax for constant expressions:</p>
1569 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1570 <dd>Truncate a constant to another type. The bit size of CST must be larger
1571 than the bit size of TYPE. Both types must be integers.</dd>
1573 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1574 <dd>Zero extend a constant to another type. The bit size of CST must be
1575 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1577 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1578 <dd>Sign extend a constant to another type. The bit size of CST must be
1579 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1581 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1582 <dd>Truncate a floating point constant to another floating point type. The
1583 size of CST must be larger than the size of TYPE. Both types must be
1584 floating point.</dd>
1586 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1587 <dd>Floating point extend a constant to another type. The size of CST must be
1588 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1590 <dt><b><tt>fptoui ( CST to TYPE )</tt></b></dt>
1591 <dd>Convert a floating point constant to the corresponding unsigned integer
1592 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1593 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1594 of the same number of elements. If the value won't fit in the integer type,
1595 the results are undefined.</dd>
1597 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1598 <dd>Convert a floating point constant to the corresponding signed integer
1599 constant. TYPE must be a scalar or vector integer type. CST must be of scalar
1600 or vector floating point type. Both CST and TYPE must be scalars, or vectors
1601 of the same number of elements. If the value won't fit in the integer type,
1602 the results are undefined.</dd>
1604 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1605 <dd>Convert an unsigned integer constant to the corresponding floating point
1606 constant. TYPE must be a scalar or vector floating point type. CST must be of
1607 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1608 of the same number of elements. If the value won't fit in the floating point
1609 type, the results are undefined.</dd>
1611 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1612 <dd>Convert a signed integer constant to the corresponding floating point
1613 constant. TYPE must be a scalar or vector floating point type. CST must be of
1614 scalar or vector integer type. Both CST and TYPE must be scalars, or vectors
1615 of the same number of elements. If the value won't fit in the floating point
1616 type, the results are undefined.</dd>
1618 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1619 <dd>Convert a pointer typed constant to the corresponding integer constant
1620 TYPE must be an integer type. CST must be of pointer type. The CST value is
1621 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1623 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1624 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1625 pointer type. CST must be of integer type. The CST value is zero extended,
1626 truncated, or unchanged to make it fit in a pointer size. This one is
1627 <i>really</i> dangerous!</dd>
1629 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1630 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1631 identical (same number of bits). The conversion is done as if the CST value
1632 was stored to memory and read back as TYPE. In other words, no bits change
1633 with this operator, just the type. This can be used for conversion of
1634 vector types to any other type, as long as they have the same bit width. For
1635 pointers it is only valid to cast to another pointer type.
1638 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1640 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1641 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1642 instruction, the index list may have zero or more indexes, which are required
1643 to make sense for the type of "CSTPTR".</dd>
1645 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1647 <dd>Perform the <a href="#i_select">select operation</a> on
1650 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1651 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1653 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1654 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1656 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1658 <dd>Perform the <a href="#i_extractelement">extractelement
1659 operation</a> on constants.
1661 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1663 <dd>Perform the <a href="#i_insertelement">insertelement
1664 operation</a> on constants.</dd>
1667 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1669 <dd>Perform the <a href="#i_shufflevector">shufflevector
1670 operation</a> on constants.</dd>
1672 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1674 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1675 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1676 binary</a> operations. The constraints on operands are the same as those for
1677 the corresponding instruction (e.g. no bitwise operations on floating point
1678 values are allowed).</dd>
1682 <!-- *********************************************************************** -->
1683 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1684 <!-- *********************************************************************** -->
1686 <!-- ======================================================================= -->
1687 <div class="doc_subsection">
1688 <a name="inlineasm">Inline Assembler Expressions</a>
1691 <div class="doc_text">
1694 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1695 Module-Level Inline Assembly</a>) through the use of a special value. This
1696 value represents the inline assembler as a string (containing the instructions
1697 to emit), a list of operand constraints (stored as a string), and a flag that
1698 indicates whether or not the inline asm expression has side effects. An example
1699 inline assembler expression is:
1702 <div class="doc_code">
1704 i32 (i32) asm "bswap $0", "=r,r"
1709 Inline assembler expressions may <b>only</b> be used as the callee operand of
1710 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1713 <div class="doc_code">
1715 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1720 Inline asms with side effects not visible in the constraint list must be marked
1721 as having side effects. This is done through the use of the
1722 '<tt>sideeffect</tt>' keyword, like so:
1725 <div class="doc_code">
1727 call void asm sideeffect "eieio", ""()
1731 <p>TODO: The format of the asm and constraints string still need to be
1732 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1733 need to be documented).
1738 <!-- *********************************************************************** -->
1739 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1740 <!-- *********************************************************************** -->
1742 <div class="doc_text">
1744 <p>The LLVM instruction set consists of several different
1745 classifications of instructions: <a href="#terminators">terminator
1746 instructions</a>, <a href="#binaryops">binary instructions</a>,
1747 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1748 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1749 instructions</a>.</p>
1753 <!-- ======================================================================= -->
1754 <div class="doc_subsection"> <a name="terminators">Terminator
1755 Instructions</a> </div>
1757 <div class="doc_text">
1759 <p>As mentioned <a href="#functionstructure">previously</a>, every
1760 basic block in a program ends with a "Terminator" instruction, which
1761 indicates which block should be executed after the current block is
1762 finished. These terminator instructions typically yield a '<tt>void</tt>'
1763 value: they produce control flow, not values (the one exception being
1764 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1765 <p>There are six different terminator instructions: the '<a
1766 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1767 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1768 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1769 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1770 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1774 <!-- _______________________________________________________________________ -->
1775 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1776 Instruction</a> </div>
1777 <div class="doc_text">
1779 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1780 ret void <i>; Return from void function</i>
1783 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1784 value) from a function back to the caller.</p>
1785 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1786 returns a value and then causes control flow, and one that just causes
1787 control flow to occur.</p>
1789 <p>The '<tt>ret</tt>' instruction may return any '<a
1790 href="#t_firstclass">first class</a>' type. Notice that a function is
1791 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1792 instruction inside of the function that returns a value that does not
1793 match the return type of the function.</p>
1795 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1796 returns back to the calling function's context. If the caller is a "<a
1797 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1798 the instruction after the call. If the caller was an "<a
1799 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1800 at the beginning of the "normal" destination block. If the instruction
1801 returns a value, that value shall set the call or invoke instruction's
1804 <pre> ret i32 5 <i>; Return an integer value of 5</i>
1805 ret void <i>; Return from a void function</i>
1808 <!-- _______________________________________________________________________ -->
1809 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1810 <div class="doc_text">
1812 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1815 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1816 transfer to a different basic block in the current function. There are
1817 two forms of this instruction, corresponding to a conditional branch
1818 and an unconditional branch.</p>
1820 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1821 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1822 unconditional form of the '<tt>br</tt>' instruction takes a single
1823 '<tt>label</tt>' value as a target.</p>
1825 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1826 argument is evaluated. If the value is <tt>true</tt>, control flows
1827 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1828 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1830 <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
1831 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1833 <!-- _______________________________________________________________________ -->
1834 <div class="doc_subsubsection">
1835 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1838 <div class="doc_text">
1842 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1847 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1848 several different places. It is a generalization of the '<tt>br</tt>'
1849 instruction, allowing a branch to occur to one of many possible
1855 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1856 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1857 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1858 table is not allowed to contain duplicate constant entries.</p>
1862 <p>The <tt>switch</tt> instruction specifies a table of values and
1863 destinations. When the '<tt>switch</tt>' instruction is executed, this
1864 table is searched for the given value. If the value is found, control flow is
1865 transfered to the corresponding destination; otherwise, control flow is
1866 transfered to the default destination.</p>
1868 <h5>Implementation:</h5>
1870 <p>Depending on properties of the target machine and the particular
1871 <tt>switch</tt> instruction, this instruction may be code generated in different
1872 ways. For example, it could be generated as a series of chained conditional
1873 branches or with a lookup table.</p>
1878 <i>; Emulate a conditional br instruction</i>
1879 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1880 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1882 <i>; Emulate an unconditional br instruction</i>
1883 switch i32 0, label %dest [ ]
1885 <i>; Implement a jump table:</i>
1886 switch i32 %val, label %otherwise [ i32 0, label %onzero
1888 i32 2, label %ontwo ]
1892 <!-- _______________________________________________________________________ -->
1893 <div class="doc_subsubsection">
1894 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1897 <div class="doc_text">
1902 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1903 to label <normal label> unwind label <exception label>
1908 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1909 function, with the possibility of control flow transfer to either the
1910 '<tt>normal</tt>' label or the
1911 '<tt>exception</tt>' label. If the callee function returns with the
1912 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1913 "normal" label. If the callee (or any indirect callees) returns with the "<a
1914 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1915 continued at the dynamically nearest "exception" label.</p>
1919 <p>This instruction requires several arguments:</p>
1923 The optional "cconv" marker indicates which <a href="#callingconv">calling
1924 convention</a> the call should use. If none is specified, the call defaults
1925 to using C calling conventions.
1927 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1928 function value being invoked. In most cases, this is a direct function
1929 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1930 an arbitrary pointer to function value.
1933 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1934 function to be invoked. </li>
1936 <li>'<tt>function args</tt>': argument list whose types match the function
1937 signature argument types. If the function signature indicates the function
1938 accepts a variable number of arguments, the extra arguments can be
1941 <li>'<tt>normal label</tt>': the label reached when the called function
1942 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1944 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1945 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1951 <p>This instruction is designed to operate as a standard '<tt><a
1952 href="#i_call">call</a></tt>' instruction in most regards. The primary
1953 difference is that it establishes an association with a label, which is used by
1954 the runtime library to unwind the stack.</p>
1956 <p>This instruction is used in languages with destructors to ensure that proper
1957 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1958 exception. Additionally, this is important for implementation of
1959 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1963 %retval = invoke i32 %Test(i32 15) to label %Continue
1964 unwind label %TestCleanup <i>; {i32}:retval set</i>
1965 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1966 unwind label %TestCleanup <i>; {i32}:retval set</i>
1971 <!-- _______________________________________________________________________ -->
1973 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1974 Instruction</a> </div>
1976 <div class="doc_text">
1985 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1986 at the first callee in the dynamic call stack which used an <a
1987 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1988 primarily used to implement exception handling.</p>
1992 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1993 immediately halt. The dynamic call stack is then searched for the first <a
1994 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1995 execution continues at the "exceptional" destination block specified by the
1996 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1997 dynamic call chain, undefined behavior results.</p>
2000 <!-- _______________________________________________________________________ -->
2002 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
2003 Instruction</a> </div>
2005 <div class="doc_text">
2014 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
2015 instruction is used to inform the optimizer that a particular portion of the
2016 code is not reachable. This can be used to indicate that the code after a
2017 no-return function cannot be reached, and other facts.</p>
2021 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
2026 <!-- ======================================================================= -->
2027 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
2028 <div class="doc_text">
2029 <p>Binary operators are used to do most of the computation in a
2030 program. They require two operands, execute an operation on them, and
2031 produce a single value. The operands might represent
2032 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
2033 The result value of a binary operator is not
2034 necessarily the same type as its operands.</p>
2035 <p>There are several different binary operators:</p>
2037 <!-- _______________________________________________________________________ -->
2038 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
2039 Instruction</a> </div>
2040 <div class="doc_text">
2042 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2045 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
2047 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
2048 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
2049 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2050 Both arguments must have identical types.</p>
2052 <p>The value produced is the integer or floating point sum of the two
2054 <p>If an integer sum has unsigned overflow, the result returned is the
2055 mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2057 <p>Because LLVM integers use a two's complement representation, this
2058 instruction is appropriate for both signed and unsigned integers.</p>
2060 <pre> <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
2063 <!-- _______________________________________________________________________ -->
2064 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
2065 Instruction</a> </div>
2066 <div class="doc_text">
2068 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2071 <p>The '<tt>sub</tt>' instruction returns the difference of its two
2073 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
2074 instruction present in most other intermediate representations.</p>
2076 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
2077 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2079 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2080 Both arguments must have identical types.</p>
2082 <p>The value produced is the integer or floating point difference of
2083 the two operands.</p>
2084 <p>If an integer difference has unsigned overflow, the result returned is the
2085 mathematical result modulo 2<sup>n</sup>, where n is the bit width of
2087 <p>Because LLVM integers use a two's complement representation, this
2088 instruction is appropriate for both signed and unsigned integers.</p>
2091 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
2092 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
2095 <!-- _______________________________________________________________________ -->
2096 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
2097 Instruction</a> </div>
2098 <div class="doc_text">
2100 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2103 <p>The '<tt>mul</tt>' instruction returns the product of its two
2106 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
2107 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
2109 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
2110 Both arguments must have identical types.</p>
2112 <p>The value produced is the integer or floating point product of the
2114 <p>If the result of an integer multiplication has unsigned overflow,
2115 the result returned is the mathematical result modulo
2116 2<sup>n</sup>, where n is the bit width of the result.</p>
2117 <p>Because LLVM integers use a two's complement representation, and the
2118 result is the same width as the operands, this instruction returns the
2119 correct result for both signed and unsigned integers. If a full product
2120 (e.g. <tt>i32</tt>x<tt>i32</tt>-><tt>i64</tt>) is needed, the operands
2121 should be sign-extended or zero-extended as appropriate to the
2122 width of the full product.</p>
2124 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
2127 <!-- _______________________________________________________________________ -->
2128 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2130 <div class="doc_text">
2132 <pre> <result> = udiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2135 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2138 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2139 <a href="#t_integer">integer</a> values. Both arguments must have identical
2140 types. This instruction can also take <a href="#t_vector">vector</a> versions
2141 of the values in which case the elements must be integers.</p>
2143 <p>The value produced is the unsigned integer quotient of the two operands.</p>
2144 <p>Note that unsigned integer division and signed integer division are distinct
2145 operations; for signed integer division, use '<tt>sdiv</tt>'.</p>
2146 <p>Division by zero leads to undefined behavior.</p>
2148 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2151 <!-- _______________________________________________________________________ -->
2152 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2154 <div class="doc_text">
2156 <pre> <result> = sdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2159 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2162 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2163 <a href="#t_integer">integer</a> values. Both arguments must have identical
2164 types. This instruction can also take <a href="#t_vector">vector</a> versions
2165 of the values in which case the elements must be integers.</p>
2167 <p>The value produced is the signed integer quotient of the two operands.</p>
2168 <p>Note that signed integer division and unsigned integer division are distinct
2169 operations; for unsigned integer division, use '<tt>udiv</tt>'.</p>
2170 <p>Division by zero leads to undefined behavior. Overflow also leads to
2171 undefined behavior; this is a rare case, but can occur, for example,
2172 by doing a 32-bit division of -2147483648 by -1.</p>
2174 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2177 <!-- _______________________________________________________________________ -->
2178 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2179 Instruction</a> </div>
2180 <div class="doc_text">
2182 <pre> <result> = fdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2185 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2188 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2189 <a href="#t_floating">floating point</a> values. Both arguments must have
2190 identical types. This instruction can also take <a href="#t_vector">vector</a>
2191 versions of floating point values.</p>
2193 <p>The value produced is the floating point quotient of the two operands.</p>
2195 <pre> <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2198 <!-- _______________________________________________________________________ -->
2199 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2201 <div class="doc_text">
2203 <pre> <result> = urem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2206 <p>The '<tt>urem</tt>' instruction returns the remainder from the
2207 unsigned division of its two arguments.</p>
2209 <p>The two arguments to the '<tt>urem</tt>' instruction must be
2210 <a href="#t_integer">integer</a> values. Both arguments must have identical
2211 types. This instruction can also take <a href="#t_vector">vector</a> versions
2212 of the values in which case the elements must be integers.</p>
2214 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2215 This instruction always performs an unsigned division to get the remainder,
2216 regardless of whether the arguments are unsigned or not.</p>
2217 <p>Note that unsigned integer remainder and signed integer remainder are
2218 distinct operations; for signed integer remainder, use '<tt>srem</tt>'.</p>
2219 <p>Taking the remainder of a division by zero leads to undefined behavior.</p>
2221 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2225 <!-- _______________________________________________________________________ -->
2226 <div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2227 Instruction</a> </div>
2228 <div class="doc_text">
2230 <pre> <result> = srem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2233 <p>The '<tt>srem</tt>' instruction returns the remainder from the
2234 signed division of its two operands. This instruction can also take
2235 <a href="#t_vector">vector</a> versions of the values in which case
2236 the elements must be integers.</p>
2239 <p>The two arguments to the '<tt>srem</tt>' instruction must be
2240 <a href="#t_integer">integer</a> values. Both arguments must have identical
2243 <p>This instruction returns the <i>remainder</i> of a division (where the result
2244 has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2245 operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2246 a value. For more information about the difference, see <a
2247 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2248 Math Forum</a>. For a table of how this is implemented in various languages,
2249 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2250 Wikipedia: modulo operation</a>.</p>
2251 <p>Note that signed integer remainder and unsigned integer remainder are
2252 distinct operations; for unsigned integer remainder, use '<tt>urem</tt>'.</p>
2253 <p>Taking the remainder of a division by zero leads to undefined behavior.
2254 Overflow also leads to undefined behavior; this is a rare case, but can occur,
2255 for example, by taking the remainder of a 32-bit division of -2147483648 by -1.
2256 (The remainder doesn't actually overflow, but this rule lets srem be
2257 implemented using instructions that return both the result of the division
2258 and the remainder.)</p>
2260 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2264 <!-- _______________________________________________________________________ -->
2265 <div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2266 Instruction</a> </div>
2267 <div class="doc_text">
2269 <pre> <result> = frem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2272 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2273 division of its two operands.</p>
2275 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2276 <a href="#t_floating">floating point</a> values. Both arguments must have
2277 identical types. This instruction can also take <a href="#t_vector">vector</a>
2278 versions of floating point values.</p>
2280 <p>This instruction returns the <i>remainder</i> of a division.</p>
2282 <pre> <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2286 <!-- ======================================================================= -->
2287 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2288 Operations</a> </div>
2289 <div class="doc_text">
2290 <p>Bitwise binary operators are used to do various forms of
2291 bit-twiddling in a program. They are generally very efficient
2292 instructions and can commonly be strength reduced from other
2293 instructions. They require two operands, execute an operation on them,
2294 and produce a single value. The resulting value of the bitwise binary
2295 operators is always the same type as its first operand.</p>
2298 <!-- _______________________________________________________________________ -->
2299 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2300 Instruction</a> </div>
2301 <div class="doc_text">
2303 <pre> <result> = shl <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2308 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2309 the left a specified number of bits.</p>
2313 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2314 href="#t_integer">integer</a> type.</p>
2318 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>. If
2319 <tt>var2</tt> is (statically or dynamically) equal to or larger than the number
2320 of bits in <tt>var1</tt>, the result is undefined.</p>
2322 <h5>Example:</h5><pre>
2323 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2324 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2325 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2326 <result> = shl i32 1, 32 <i>; undefined</i>
2329 <!-- _______________________________________________________________________ -->
2330 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2331 Instruction</a> </div>
2332 <div class="doc_text">
2334 <pre> <result> = lshr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2338 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2339 operand shifted to the right a specified number of bits with zero fill.</p>
2342 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2343 <a href="#t_integer">integer</a> type.</p>
2347 <p>This instruction always performs a logical shift right operation. The most
2348 significant bits of the result will be filled with zero bits after the
2349 shift. If <tt>var2</tt> is (statically or dynamically) equal to or larger than
2350 the number of bits in <tt>var1</tt>, the result is undefined.</p>
2354 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2355 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2356 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2357 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2358 <result> = lshr i32 1, 32 <i>; undefined</i>
2362 <!-- _______________________________________________________________________ -->
2363 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2364 Instruction</a> </div>
2365 <div class="doc_text">
2368 <pre> <result> = ashr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2372 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2373 operand shifted to the right a specified number of bits with sign extension.</p>
2376 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2377 <a href="#t_integer">integer</a> type.</p>
2380 <p>This instruction always performs an arithmetic shift right operation,
2381 The most significant bits of the result will be filled with the sign bit
2382 of <tt>var1</tt>. If <tt>var2</tt> is (statically or dynamically) equal to or
2383 larger than the number of bits in <tt>var1</tt>, the result is undefined.
2388 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2389 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2390 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2391 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
2392 <result> = ashr i32 1, 32 <i>; undefined</i>
2396 <!-- _______________________________________________________________________ -->
2397 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2398 Instruction</a> </div>
2399 <div class="doc_text">
2401 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2404 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2405 its two operands.</p>
2407 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
2408 href="#t_integer">integer</a> values. Both arguments must have
2409 identical types.</p>
2411 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2413 <div style="align: center">
2414 <table border="1" cellspacing="0" cellpadding="4">
2445 <pre> <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
2446 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2447 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2450 <!-- _______________________________________________________________________ -->
2451 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2452 <div class="doc_text">
2454 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2457 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2458 or of its two operands.</p>
2460 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
2461 href="#t_integer">integer</a> values. Both arguments must have
2462 identical types.</p>
2464 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2466 <div style="align: center">
2467 <table border="1" cellspacing="0" cellpadding="4">
2498 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2499 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2500 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2503 <!-- _______________________________________________________________________ -->
2504 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2505 Instruction</a> </div>
2506 <div class="doc_text">
2508 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2511 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2512 or of its two operands. The <tt>xor</tt> is used to implement the
2513 "one's complement" operation, which is the "~" operator in C.</p>
2515 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2516 href="#t_integer">integer</a> values. Both arguments must have
2517 identical types.</p>
2519 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2521 <div style="align: center">
2522 <table border="1" cellspacing="0" cellpadding="4">
2554 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2555 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2556 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2557 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2561 <!-- ======================================================================= -->
2562 <div class="doc_subsection">
2563 <a name="vectorops">Vector Operations</a>
2566 <div class="doc_text">
2568 <p>LLVM supports several instructions to represent vector operations in a
2569 target-independent manner. These instructions cover the element-access and
2570 vector-specific operations needed to process vectors effectively. While LLVM
2571 does directly support these vector operations, many sophisticated algorithms
2572 will want to use target-specific intrinsics to take full advantage of a specific
2577 <!-- _______________________________________________________________________ -->
2578 <div class="doc_subsubsection">
2579 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2582 <div class="doc_text">
2587 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
2593 The '<tt>extractelement</tt>' instruction extracts a single scalar
2594 element from a vector at a specified index.
2601 The first operand of an '<tt>extractelement</tt>' instruction is a
2602 value of <a href="#t_vector">vector</a> type. The second operand is
2603 an index indicating the position from which to extract the element.
2604 The index may be a variable.</p>
2609 The result is a scalar of the same type as the element type of
2610 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2611 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2612 results are undefined.
2618 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
2623 <!-- _______________________________________________________________________ -->
2624 <div class="doc_subsubsection">
2625 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2628 <div class="doc_text">
2633 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
2639 The '<tt>insertelement</tt>' instruction inserts a scalar
2640 element into a vector at a specified index.
2647 The first operand of an '<tt>insertelement</tt>' instruction is a
2648 value of <a href="#t_vector">vector</a> type. The second operand is a
2649 scalar value whose type must equal the element type of the first
2650 operand. The third operand is an index indicating the position at
2651 which to insert the value. The index may be a variable.</p>
2656 The result is a vector of the same type as <tt>val</tt>. Its
2657 element values are those of <tt>val</tt> except at position
2658 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2659 exceeds the length of <tt>val</tt>, the results are undefined.
2665 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
2669 <!-- _______________________________________________________________________ -->
2670 <div class="doc_subsubsection">
2671 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2674 <div class="doc_text">
2679 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x i32> <mask> <i>; yields <n x <ty>></i>
2685 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2686 from two input vectors, returning a vector of the same type.
2692 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2693 with types that match each other and types that match the result of the
2694 instruction. The third argument is a shuffle mask, which has the same number
2695 of elements as the other vector type, but whose element type is always 'i32'.
2699 The shuffle mask operand is required to be a constant vector with either
2700 constant integer or undef values.
2706 The elements of the two input vectors are numbered from left to right across
2707 both of the vectors. The shuffle mask operand specifies, for each element of
2708 the result vector, which element of the two input registers the result element
2709 gets. The element selector may be undef (meaning "don't care") and the second
2710 operand may be undef if performing a shuffle from only one vector.
2716 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
2717 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
2718 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
2719 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
2724 <!-- ======================================================================= -->
2725 <div class="doc_subsection">
2726 <a name="memoryops">Memory Access and Addressing Operations</a>
2729 <div class="doc_text">
2731 <p>A key design point of an SSA-based representation is how it
2732 represents memory. In LLVM, no memory locations are in SSA form, which
2733 makes things very simple. This section describes how to read, write,
2734 allocate, and free memory in LLVM.</p>
2738 <!-- _______________________________________________________________________ -->
2739 <div class="doc_subsubsection">
2740 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2743 <div class="doc_text">
2748 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2753 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2754 heap and returns a pointer to it. The object is always allocated in the generic
2755 address space (address space zero).</p>
2759 <p>The '<tt>malloc</tt>' instruction allocates
2760 <tt>sizeof(<type>)*NumElements</tt>
2761 bytes of memory from the operating system and returns a pointer of the
2762 appropriate type to the program. If "NumElements" is specified, it is the
2763 number of elements allocated. If an alignment is specified, the value result
2764 of the allocation is guaranteed to be aligned to at least that boundary. If
2765 not specified, or if zero, the target can choose to align the allocation on any
2766 convenient boundary.</p>
2768 <p>'<tt>type</tt>' must be a sized type.</p>
2772 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2773 a pointer is returned.</p>
2778 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2780 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2781 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2782 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2783 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2784 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2788 <!-- _______________________________________________________________________ -->
2789 <div class="doc_subsubsection">
2790 <a name="i_free">'<tt>free</tt>' Instruction</a>
2793 <div class="doc_text">
2798 free <type> <value> <i>; yields {void}</i>
2803 <p>The '<tt>free</tt>' instruction returns memory back to the unused
2804 memory heap to be reallocated in the future.</p>
2808 <p>'<tt>value</tt>' shall be a pointer value that points to a value
2809 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2814 <p>Access to the memory pointed to by the pointer is no longer defined
2815 after this instruction executes.</p>
2820 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2821 free [4 x i8]* %array
2825 <!-- _______________________________________________________________________ -->
2826 <div class="doc_subsubsection">
2827 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2830 <div class="doc_text">
2835 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2840 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
2841 currently executing function, to be automatically released when this function
2842 returns to its caller. The object is always allocated in the generic address
2843 space (address space zero).</p>
2847 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2848 bytes of memory on the runtime stack, returning a pointer of the
2849 appropriate type to the program. If "NumElements" is specified, it is the
2850 number of elements allocated. If an alignment is specified, the value result
2851 of the allocation is guaranteed to be aligned to at least that boundary. If
2852 not specified, or if zero, the target can choose to align the allocation on any
2853 convenient boundary.</p>
2855 <p>'<tt>type</tt>' may be any sized type.</p>
2859 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2860 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2861 instruction is commonly used to represent automatic variables that must
2862 have an address available. When the function returns (either with the <tt><a
2863 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2864 instructions), the memory is reclaimed.</p>
2869 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2870 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2871 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2872 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2876 <!-- _______________________________________________________________________ -->
2877 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2878 Instruction</a> </div>
2879 <div class="doc_text">
2881 <pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
2883 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2885 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2886 address from which to load. The pointer must point to a <a
2887 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2888 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2889 the number or order of execution of this <tt>load</tt> with other
2890 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2893 The optional "align" argument specifies the alignment of the operation
2894 (that is, the alignment of the memory address). A value of 0 or an
2895 omitted "align" argument means that the operation has the preferential
2896 alignment for the target. It is the responsibility of the code emitter
2897 to ensure that the alignment information is correct. Overestimating
2898 the alignment results in an undefined behavior. Underestimating the
2899 alignment may produce less efficient code. An alignment of 1 is always
2903 <p>The location of memory pointed to is loaded.</p>
2905 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2907 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2908 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2911 <!-- _______________________________________________________________________ -->
2912 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2913 Instruction</a> </div>
2914 <div class="doc_text">
2916 <pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2917 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2920 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2922 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2923 to store and an address at which to store it. The type of the '<tt><pointer></tt>'
2924 operand must be a pointer to the type of the '<tt><value></tt>'
2925 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2926 optimizer is not allowed to modify the number or order of execution of
2927 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2928 href="#i_store">store</a></tt> instructions.</p>
2930 The optional "align" argument specifies the alignment of the operation
2931 (that is, the alignment of the memory address). A value of 0 or an
2932 omitted "align" argument means that the operation has the preferential
2933 alignment for the target. It is the responsibility of the code emitter
2934 to ensure that the alignment information is correct. Overestimating
2935 the alignment results in an undefined behavior. Underestimating the
2936 alignment may produce less efficient code. An alignment of 1 is always
2940 <p>The contents of memory are updated to contain '<tt><value></tt>'
2941 at the location specified by the '<tt><pointer></tt>' operand.</p>
2943 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2944 store i32 3, i32* %ptr <i>; yields {void}</i>
2945 %val = <a href="#i_load">load</a> i32* %ptr <i>; yields {i32}:val = i32 3</i>
2949 <!-- _______________________________________________________________________ -->
2950 <div class="doc_subsubsection">
2951 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2954 <div class="doc_text">
2957 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2963 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2964 subelement of an aggregate data structure.</p>
2968 <p>This instruction takes a list of integer operands that indicate what
2969 elements of the aggregate object to index to. The actual types of the arguments
2970 provided depend on the type of the first pointer argument. The
2971 '<tt>getelementptr</tt>' instruction is used to index down through the type
2972 levels of a structure or to a specific index in an array. When indexing into a
2973 structure, only <tt>i32</tt> integer constants are allowed. When indexing
2974 into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2975 be sign extended to 64-bit values.</p>
2977 <p>For example, let's consider a C code fragment and how it gets
2978 compiled to LLVM:</p>
2980 <div class="doc_code">
2993 int *foo(struct ST *s) {
2994 return &s[1].Z.B[5][13];
2999 <p>The LLVM code generated by the GCC frontend is:</p>
3001 <div class="doc_code">
3003 %RT = type { i8 , [10 x [20 x i32]], i8 }
3004 %ST = type { i32, double, %RT }
3006 define i32* %foo(%ST* %s) {
3008 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
3016 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
3017 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
3018 and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
3019 <a href="#t_integer">integer</a> type but the value will always be sign extended
3020 to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
3021 <b>constants</b>.</p>
3023 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
3024 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
3025 }</tt>' type, a structure. The second index indexes into the third element of
3026 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
3027 i8 }</tt>' type, another structure. The third index indexes into the second
3028 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
3029 array. The two dimensions of the array are subscripted into, yielding an
3030 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
3031 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
3033 <p>Note that it is perfectly legal to index partially through a
3034 structure, returning a pointer to an inner element. Because of this,
3035 the LLVM code for the given testcase is equivalent to:</p>
3038 define i32* %foo(%ST* %s) {
3039 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
3040 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
3041 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
3042 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
3043 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
3048 <p>Note that it is undefined to access an array out of bounds: array and
3049 pointer indexes must always be within the defined bounds of the array type.
3050 The one exception for this rules is zero length arrays. These arrays are
3051 defined to be accessible as variable length arrays, which requires access
3052 beyond the zero'th element.</p>
3054 <p>The getelementptr instruction is often confusing. For some more insight
3055 into how it works, see <a href="GetElementPtr.html">the getelementptr
3061 <i>; yields [12 x i8]*:aptr</i>
3062 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
3066 <!-- ======================================================================= -->
3067 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
3069 <div class="doc_text">
3070 <p>The instructions in this category are the conversion instructions (casting)
3071 which all take a single operand and a type. They perform various bit conversions
3075 <!-- _______________________________________________________________________ -->
3076 <div class="doc_subsubsection">
3077 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
3079 <div class="doc_text">
3083 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
3088 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
3093 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
3094 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
3095 and type of the result, which must be an <a href="#t_integer">integer</a>
3096 type. The bit size of <tt>value</tt> must be larger than the bit size of
3097 <tt>ty2</tt>. Equal sized types are not allowed.</p>
3101 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
3102 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
3103 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
3104 It will always truncate bits.</p>
3108 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
3109 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
3110 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
3114 <!-- _______________________________________________________________________ -->
3115 <div class="doc_subsubsection">
3116 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
3118 <div class="doc_text">
3122 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
3126 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
3131 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
3132 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3133 also be of <a href="#t_integer">integer</a> type. The bit size of the
3134 <tt>value</tt> must be smaller than the bit size of the destination type,
3138 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
3139 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
3141 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
3145 %X = zext i32 257 to i64 <i>; yields i64:257</i>
3146 %Y = zext i1 true to i32 <i>; yields i32:1</i>
3150 <!-- _______________________________________________________________________ -->
3151 <div class="doc_subsubsection">
3152 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
3154 <div class="doc_text">
3158 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
3162 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
3166 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
3167 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
3168 also be of <a href="#t_integer">integer</a> type. The bit size of the
3169 <tt>value</tt> must be smaller than the bit size of the destination type,
3174 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
3175 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
3176 the type <tt>ty2</tt>.</p>
3178 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
3182 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
3183 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3187 <!-- _______________________________________________________________________ -->
3188 <div class="doc_subsubsection">
3189 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3192 <div class="doc_text">
3197 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
3201 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3206 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3207 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3208 cast it to. The size of <tt>value</tt> must be larger than the size of
3209 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3210 <i>no-op cast</i>.</p>
3213 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3214 <a href="#t_floating">floating point</a> type to a smaller
3215 <a href="#t_floating">floating point</a> type. If the value cannot fit within
3216 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3220 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3221 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3225 <!-- _______________________________________________________________________ -->
3226 <div class="doc_subsubsection">
3227 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3229 <div class="doc_text">
3233 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
3237 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3238 floating point value.</p>
3241 <p>The '<tt>fpext</tt>' instruction takes a
3242 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3243 and a <a href="#t_floating">floating point</a> type to cast it to. The source
3244 type must be smaller than the destination type.</p>
3247 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3248 <a href="#t_floating">floating point</a> type to a larger
3249 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3250 used to make a <i>no-op cast</i> because it always changes bits. Use
3251 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3255 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3256 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3260 <!-- _______________________________________________________________________ -->
3261 <div class="doc_subsubsection">
3262 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3264 <div class="doc_text">
3268 <result> = fptoui <ty> <value> to <ty2> <i>; yields ty2</i>
3272 <p>The '<tt>fptoui</tt>' converts a floating point <tt>value</tt> to its
3273 unsigned integer equivalent of type <tt>ty2</tt>.
3277 <p>The '<tt>fptoui</tt>' instruction takes a value to cast, which must be a
3278 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3279 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3280 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3281 vector integer type with the same number of elements as <tt>ty</tt></p>
3284 <p> The '<tt>fptoui</tt>' instruction converts its
3285 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3286 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3287 the results are undefined.</p>
3291 %X = fptoui double 123.0 to i32 <i>; yields i32:123</i>
3292 %Y = fptoui float 1.0E+300 to i1 <i>; yields undefined:1</i>
3293 %X = fptoui float 1.04E+17 to i8 <i>; yields undefined:1</i>
3297 <!-- _______________________________________________________________________ -->
3298 <div class="doc_subsubsection">
3299 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3301 <div class="doc_text">
3305 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
3309 <p>The '<tt>fptosi</tt>' instruction converts
3310 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3314 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
3315 scalar or vector <a href="#t_floating">floating point</a> value, and a type
3316 to cast it to <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a>
3317 type. If <tt>ty</tt> is a vector floating point type, <tt>ty2</tt> must be a
3318 vector integer type with the same number of elements as <tt>ty</tt></p>
3321 <p>The '<tt>fptosi</tt>' instruction converts its
3322 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3323 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3324 the results are undefined.</p>
3328 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
3329 %Y = fptosi float 1.0E-247 to i1 <i>; yields undefined:1</i>
3330 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3334 <!-- _______________________________________________________________________ -->
3335 <div class="doc_subsubsection">
3336 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3338 <div class="doc_text">
3342 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3346 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3347 integer and converts that value to the <tt>ty2</tt> type.</p>
3350 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be a
3351 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3352 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3353 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3354 floating point type with the same number of elements as <tt>ty</tt></p>
3357 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3358 integer quantity and converts it to the corresponding floating point value. If
3359 the value cannot fit in the floating point value, the results are undefined.</p>
3363 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3364 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3368 <!-- _______________________________________________________________________ -->
3369 <div class="doc_subsubsection">
3370 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3372 <div class="doc_text">
3376 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3380 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3381 integer and converts that value to the <tt>ty2</tt> type.</p>
3384 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be a
3385 scalar or vector <a href="#t_integer">integer</a> value, and a type to cast it
3386 to <tt>ty2</tt>, which must be an <a href="#t_floating">floating point</a>
3387 type. If <tt>ty</tt> is a vector integer type, <tt>ty2</tt> must be a vector
3388 floating point type with the same number of elements as <tt>ty</tt></p>
3391 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3392 integer quantity and converts it to the corresponding floating point value. If
3393 the value cannot fit in the floating point value, the results are undefined.</p>
3397 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3398 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3402 <!-- _______________________________________________________________________ -->
3403 <div class="doc_subsubsection">
3404 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3406 <div class="doc_text">
3410 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
3414 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3415 the integer type <tt>ty2</tt>.</p>
3418 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3419 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3420 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3423 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3424 <tt>ty2</tt> by interpreting the pointer value as an integer and either
3425 truncating or zero extending that value to the size of the integer type. If
3426 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3427 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3428 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3433 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3434 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3438 <!-- _______________________________________________________________________ -->
3439 <div class="doc_subsubsection">
3440 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3442 <div class="doc_text">
3446 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
3450 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3451 a pointer type, <tt>ty2</tt>.</p>
3454 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3455 value to cast, and a type to cast it to, which must be a
3456 <a href="#t_pointer">pointer</a> type.
3459 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3460 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
3461 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3462 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3463 the size of a pointer then a zero extension is done. If they are the same size,
3464 nothing is done (<i>no-op cast</i>).</p>
3468 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3469 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3470 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3474 <!-- _______________________________________________________________________ -->
3475 <div class="doc_subsubsection">
3476 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3478 <div class="doc_text">
3482 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3486 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3487 <tt>ty2</tt> without changing any bits.</p>
3490 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3491 a first class value, and a type to cast it to, which must also be a <a
3492 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3493 and the destination type, <tt>ty2</tt>, must be identical. If the source
3494 type is a pointer, the destination type must also be a pointer.</p>
3497 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3498 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3499 this conversion. The conversion is done as if the <tt>value</tt> had been
3500 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3501 converted to other pointer types with this instruction. To convert pointers to
3502 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3503 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3507 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3508 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3509 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3513 <!-- ======================================================================= -->
3514 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3515 <div class="doc_text">
3516 <p>The instructions in this category are the "miscellaneous"
3517 instructions, which defy better classification.</p>
3520 <!-- _______________________________________________________________________ -->
3521 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3523 <div class="doc_text">
3525 <pre> <result> = icmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
3528 <p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3529 of its two integer operands.</p>
3531 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3532 the condition code indicating the kind of comparison to perform. It is not
3533 a value, just a keyword. The possible condition code are:
3535 <li><tt>eq</tt>: equal</li>
3536 <li><tt>ne</tt>: not equal </li>
3537 <li><tt>ugt</tt>: unsigned greater than</li>
3538 <li><tt>uge</tt>: unsigned greater or equal</li>
3539 <li><tt>ult</tt>: unsigned less than</li>
3540 <li><tt>ule</tt>: unsigned less or equal</li>
3541 <li><tt>sgt</tt>: signed greater than</li>
3542 <li><tt>sge</tt>: signed greater or equal</li>
3543 <li><tt>slt</tt>: signed less than</li>
3544 <li><tt>sle</tt>: signed less or equal</li>
3546 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3547 <a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3549 <p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3550 the condition code given as <tt>cond</tt>. The comparison performed always
3551 yields a <a href="#t_primitive">i1</a> result, as follows:
3553 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3554 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3556 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3557 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3558 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3559 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3560 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3561 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3562 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3563 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3564 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3565 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3566 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3567 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3568 <li><tt>sge</tt>: interprets the operands as signed values and yields
3569 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3570 <li><tt>slt</tt>: interprets the operands as signed values and yields
3571 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3572 <li><tt>sle</tt>: interprets the operands as signed values and yields
3573 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3575 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3576 values are compared as if they were integers.</p>
3579 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
3580 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
3581 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
3582 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3583 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3584 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
3588 <!-- _______________________________________________________________________ -->
3589 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3591 <div class="doc_text">
3593 <pre> <result> = fcmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
3596 <p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3597 of its floating point operands.</p>
3599 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3600 the condition code indicating the kind of comparison to perform. It is not
3601 a value, just a keyword. The possible condition code are:
3603 <li><tt>false</tt>: no comparison, always returns false</li>
3604 <li><tt>oeq</tt>: ordered and equal</li>
3605 <li><tt>ogt</tt>: ordered and greater than </li>
3606 <li><tt>oge</tt>: ordered and greater than or equal</li>
3607 <li><tt>olt</tt>: ordered and less than </li>
3608 <li><tt>ole</tt>: ordered and less than or equal</li>
3609 <li><tt>one</tt>: ordered and not equal</li>
3610 <li><tt>ord</tt>: ordered (no nans)</li>
3611 <li><tt>ueq</tt>: unordered or equal</li>
3612 <li><tt>ugt</tt>: unordered or greater than </li>
3613 <li><tt>uge</tt>: unordered or greater than or equal</li>
3614 <li><tt>ult</tt>: unordered or less than </li>
3615 <li><tt>ule</tt>: unordered or less than or equal</li>
3616 <li><tt>une</tt>: unordered or not equal</li>
3617 <li><tt>uno</tt>: unordered (either nans)</li>
3618 <li><tt>true</tt>: no comparison, always returns true</li>
3620 <p><i>Ordered</i> means that neither operand is a QNAN while
3621 <i>unordered</i> means that either operand may be a QNAN.</p>
3622 <p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3623 <a href="#t_floating">floating point</a> typed. They must have identical
3626 <p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3627 the condition code given as <tt>cond</tt>. The comparison performed always
3628 yields a <a href="#t_primitive">i1</a> result, as follows:
3630 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3631 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3632 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3633 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3634 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3635 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3636 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3637 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3638 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3639 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3640 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3641 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3642 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3643 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3644 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3645 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3646 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3647 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3648 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3649 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3650 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3651 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3652 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3653 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3654 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3655 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3656 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3657 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3661 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3662 <result> = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3663 <result> = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3664 <result> = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3668 <!-- _______________________________________________________________________ -->
3669 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3670 Instruction</a> </div>
3671 <div class="doc_text">
3673 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
3675 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
3676 the SSA graph representing the function.</p>
3678 <p>The type of the incoming values is specified with the first type
3679 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3680 as arguments, with one pair for each predecessor basic block of the
3681 current block. Only values of <a href="#t_firstclass">first class</a>
3682 type may be used as the value arguments to the PHI node. Only labels
3683 may be used as the label arguments.</p>
3684 <p>There must be no non-phi instructions between the start of a basic
3685 block and the PHI instructions: i.e. PHI instructions must be first in
3688 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3689 specified by the pair corresponding to the predecessor basic block that executed
3690 just prior to the current block.</p>
3692 <pre>Loop: ; Infinite loop that counts from 0 on up...<br> %indvar = phi i32 [ 0, %LoopHeader ], [ %nextindvar, %Loop ]<br> %nextindvar = add i32 %indvar, 1<br> br label %Loop<br></pre>
3695 <!-- _______________________________________________________________________ -->
3696 <div class="doc_subsubsection">
3697 <a name="i_select">'<tt>select</tt>' Instruction</a>
3700 <div class="doc_text">
3705 <result> = select i1 <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
3711 The '<tt>select</tt>' instruction is used to choose one value based on a
3712 condition, without branching.
3719 The '<tt>select</tt>' instruction requires a boolean value indicating the condition, and two values of the same <a href="#t_firstclass">first class</a> type.
3725 If the boolean condition evaluates to true, the instruction returns the first
3726 value argument; otherwise, it returns the second value argument.
3732 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3737 <!-- _______________________________________________________________________ -->
3738 <div class="doc_subsubsection">
3739 <a name="i_call">'<tt>call</tt>' Instruction</a>
3742 <div class="doc_text">
3746 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty> [<fnty>*] <fnptrval>(<param list>)
3751 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3755 <p>This instruction requires several arguments:</p>
3759 <p>The optional "tail" marker indicates whether the callee function accesses
3760 any allocas or varargs in the caller. If the "tail" marker is present, the
3761 function call is eligible for tail call optimization. Note that calls may
3762 be marked "tail" even if they do not occur before a <a
3763 href="#i_ret"><tt>ret</tt></a> instruction.
3766 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3767 convention</a> the call should use. If none is specified, the call defaults
3768 to using C calling conventions.
3771 <p>'<tt>ty</tt>': the type of the call instruction itself which is also
3772 the type of the return value. Functions that return no value are marked
3773 <tt><a href="#t_void">void</a></tt>.</p>
3776 <p>'<tt>fnty</tt>': shall be the signature of the pointer to function
3777 value being invoked. The argument types must match the types implied by
3778 this signature. This type can be omitted if the function is not varargs
3779 and if the function type does not return a pointer to a function.</p>
3782 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3783 be invoked. In most cases, this is a direct function invocation, but
3784 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3785 to function value.</p>
3788 <p>'<tt>function args</tt>': argument list whose types match the
3789 function signature argument types. All arguments must be of
3790 <a href="#t_firstclass">first class</a> type. If the function signature
3791 indicates the function accepts a variable number of arguments, the extra
3792 arguments can be specified.</p>
3798 <p>The '<tt>call</tt>' instruction is used to cause control flow to
3799 transfer to a specified function, with its incoming arguments bound to
3800 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3801 instruction in the called function, control flow continues with the
3802 instruction after the function call, and the return value of the
3803 function is bound to the result argument. This is a simpler case of
3804 the <a href="#i_invoke">invoke</a> instruction.</p>
3809 %retval = call i32 @test(i32 %argc)
3810 call i32 (i8 *, ...)* @printf(i8 * %msg, i32 12, i8 42);
3811 %X = tail call i32 @foo()
3812 %Y = tail call <a href="#callingconv">fastcc</a> i32 @foo()
3813 %Z = call void %foo(i8 97 signext)
3818 <!-- _______________________________________________________________________ -->
3819 <div class="doc_subsubsection">
3820 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3823 <div class="doc_text">
3828 <resultval> = va_arg <va_list*> <arglist>, <argty>
3833 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3834 the "variable argument" area of a function call. It is used to implement the
3835 <tt>va_arg</tt> macro in C.</p>
3839 <p>This instruction takes a <tt>va_list*</tt> value and the type of
3840 the argument. It returns a value of the specified argument type and
3841 increments the <tt>va_list</tt> to point to the next argument. The
3842 actual type of <tt>va_list</tt> is target specific.</p>
3846 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3847 type from the specified <tt>va_list</tt> and causes the
3848 <tt>va_list</tt> to point to the next argument. For more information,
3849 see the variable argument handling <a href="#int_varargs">Intrinsic
3852 <p>It is legal for this instruction to be called in a function which does not
3853 take a variable number of arguments, for example, the <tt>vfprintf</tt>
3856 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3857 href="#intrinsics">intrinsic function</a> because it takes a type as an
3862 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3866 <!-- *********************************************************************** -->
3867 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3868 <!-- *********************************************************************** -->
3870 <div class="doc_text">
3872 <p>LLVM supports the notion of an "intrinsic function". These functions have
3873 well known names and semantics and are required to follow certain restrictions.
3874 Overall, these intrinsics represent an extension mechanism for the LLVM
3875 language that does not require changing all of the transformations in LLVM when
3876 adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
3878 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3879 prefix is reserved in LLVM for intrinsic names; thus, function names may not
3880 begin with this prefix. Intrinsic functions must always be external functions:
3881 you cannot define the body of intrinsic functions. Intrinsic functions may
3882 only be used in call or invoke instructions: it is illegal to take the address
3883 of an intrinsic function. Additionally, because intrinsic functions are part
3884 of the LLVM language, it is required if any are added that they be documented
3887 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
3888 a family of functions that perform the same operation but on different data
3889 types. Because LLVM can represent over 8 million different integer types,
3890 overloading is used commonly to allow an intrinsic function to operate on any
3891 integer type. One or more of the argument types or the result type can be
3892 overloaded to accept any integer type. Argument types may also be defined as
3893 exactly matching a previous argument's type or the result type. This allows an
3894 intrinsic function which accepts multiple arguments, but needs all of them to
3895 be of the same type, to only be overloaded with respect to a single argument or
3898 <p>Overloaded intrinsics will have the names of its overloaded argument types
3899 encoded into its function name, each preceded by a period. Only those types
3900 which are overloaded result in a name suffix. Arguments whose type is matched
3901 against another type do not. For example, the <tt>llvm.ctpop</tt> function can
3902 take an integer of any width and returns an integer of exactly the same integer
3903 width. This leads to a family of functions such as
3904 <tt>i8 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i29 @llvm.ctpop.i29(i29 %val)</tt>.
3905 Only one type, the return type, is overloaded, and only one type suffix is
3906 required. Because the argument's type is matched against the return type, it
3907 does not require its own name suffix.</p>
3909 <p>To learn how to add an intrinsic function, please see the
3910 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3915 <!-- ======================================================================= -->
3916 <div class="doc_subsection">
3917 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3920 <div class="doc_text">
3922 <p>Variable argument support is defined in LLVM with the <a
3923 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3924 intrinsic functions. These functions are related to the similarly
3925 named macros defined in the <tt><stdarg.h></tt> header file.</p>
3927 <p>All of these functions operate on arguments that use a
3928 target-specific value type "<tt>va_list</tt>". The LLVM assembly
3929 language reference manual does not define what this type is, so all
3930 transformations should be prepared to handle these functions regardless of
3933 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3934 instruction and the variable argument handling intrinsic functions are
3937 <div class="doc_code">
3939 define i32 @test(i32 %X, ...) {
3940 ; Initialize variable argument processing
3942 %ap2 = bitcast i8** %ap to i8*
3943 call void @llvm.va_start(i8* %ap2)
3945 ; Read a single integer argument
3946 %tmp = va_arg i8** %ap, i32
3948 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3950 %aq2 = bitcast i8** %aq to i8*
3951 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
3952 call void @llvm.va_end(i8* %aq2)
3954 ; Stop processing of arguments.
3955 call void @llvm.va_end(i8* %ap2)
3959 declare void @llvm.va_start(i8*)
3960 declare void @llvm.va_copy(i8*, i8*)
3961 declare void @llvm.va_end(i8*)
3967 <!-- _______________________________________________________________________ -->
3968 <div class="doc_subsubsection">
3969 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3973 <div class="doc_text">
3975 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
3977 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3978 <tt>*<arglist></tt> for subsequent use by <tt><a
3979 href="#i_va_arg">va_arg</a></tt>.</p>
3983 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3987 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3988 macro available in C. In a target-dependent way, it initializes the
3989 <tt>va_list</tt> element to which the argument points, so that the next call to
3990 <tt>va_arg</tt> will produce the first variable argument passed to the function.
3991 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3992 last argument of the function as the compiler can figure that out.</p>
3996 <!-- _______________________________________________________________________ -->
3997 <div class="doc_subsubsection">
3998 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
4001 <div class="doc_text">
4003 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
4006 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
4007 which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
4008 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
4012 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
4016 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
4017 macro available in C. In a target-dependent way, it destroys the
4018 <tt>va_list</tt> element to which the argument points. Calls to <a
4019 href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
4020 <tt>llvm.va_copy</tt></a> must be matched exactly with calls to
4021 <tt>llvm.va_end</tt>.</p>
4025 <!-- _______________________________________________________________________ -->
4026 <div class="doc_subsubsection">
4027 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
4030 <div class="doc_text">
4035 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
4040 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
4041 from the source argument list to the destination argument list.</p>
4045 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
4046 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
4051 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
4052 macro available in C. In a target-dependent way, it copies the source
4053 <tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
4054 intrinsic is necessary because the <tt><a href="#int_va_start">
4055 llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
4056 example, memory allocation.</p>
4060 <!-- ======================================================================= -->
4061 <div class="doc_subsection">
4062 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
4065 <div class="doc_text">
4068 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
4069 Collection</a> requires the implementation and generation of these intrinsics.
4070 These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
4071 stack</a>, as well as garbage collector implementations that require <a
4072 href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
4073 Front-ends for type-safe garbage collected languages should generate these
4074 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
4075 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
4078 <p>The garbage collection intrinsics only operate on objects in the generic
4079 address space (address space zero).</p>
4083 <!-- _______________________________________________________________________ -->
4084 <div class="doc_subsubsection">
4085 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
4088 <div class="doc_text">
4093 declare void @llvm.gcroot(i8** %ptrloc, i8* %metadata)
4098 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
4099 the code generator, and allows some metadata to be associated with it.</p>
4103 <p>The first argument specifies the address of a stack object that contains the
4104 root pointer. The second pointer (which must be either a constant or a global
4105 value address) contains the meta-data to be associated with the root.</p>
4109 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
4110 location. At compile-time, the code generator generates information to allow
4111 the runtime to find the pointer at GC safe points. The '<tt>llvm.gcroot</tt>'
4112 intrinsic may only be used in a function which <a href="#gc">specifies a GC
4118 <!-- _______________________________________________________________________ -->
4119 <div class="doc_subsubsection">
4120 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
4123 <div class="doc_text">
4128 declare i8* @llvm.gcread(i8* %ObjPtr, i8** %Ptr)
4133 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
4134 locations, allowing garbage collector implementations that require read
4139 <p>The second argument is the address to read from, which should be an address
4140 allocated from the garbage collector. The first object is a pointer to the
4141 start of the referenced object, if needed by the language runtime (otherwise
4146 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
4147 instruction, but may be replaced with substantially more complex code by the
4148 garbage collector runtime, as needed. The '<tt>llvm.gcread</tt>' intrinsic
4149 may only be used in a function which <a href="#gc">specifies a GC
4155 <!-- _______________________________________________________________________ -->
4156 <div class="doc_subsubsection">
4157 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
4160 <div class="doc_text">
4165 declare void @llvm.gcwrite(i8* %P1, i8* %Obj, i8** %P2)
4170 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
4171 locations, allowing garbage collector implementations that require write
4172 barriers (such as generational or reference counting collectors).</p>
4176 <p>The first argument is the reference to store, the second is the start of the
4177 object to store it to, and the third is the address of the field of Obj to
4178 store to. If the runtime does not require a pointer to the object, Obj may be
4183 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
4184 instruction, but may be replaced with substantially more complex code by the
4185 garbage collector runtime, as needed. The '<tt>llvm.gcwrite</tt>' intrinsic
4186 may only be used in a function which <a href="#gc">specifies a GC
4193 <!-- ======================================================================= -->
4194 <div class="doc_subsection">
4195 <a name="int_codegen">Code Generator Intrinsics</a>
4198 <div class="doc_text">
4200 These intrinsics are provided by LLVM to expose special features that may only
4201 be implemented with code generator support.
4206 <!-- _______________________________________________________________________ -->
4207 <div class="doc_subsubsection">
4208 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4211 <div class="doc_text">
4215 declare i8 *@llvm.returnaddress(i32 <level>)
4221 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4222 target-specific value indicating the return address of the current function
4223 or one of its callers.
4229 The argument to this intrinsic indicates which function to return the address
4230 for. Zero indicates the calling function, one indicates its caller, etc. The
4231 argument is <b>required</b> to be a constant integer value.
4237 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4238 the return address of the specified call frame, or zero if it cannot be
4239 identified. The value returned by this intrinsic is likely to be incorrect or 0
4240 for arguments other than zero, so it should only be used for debugging purposes.
4244 Note that calling this intrinsic does not prevent function inlining or other
4245 aggressive transformations, so the value returned may not be that of the obvious
4246 source-language caller.
4251 <!-- _______________________________________________________________________ -->
4252 <div class="doc_subsubsection">
4253 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4256 <div class="doc_text">
4260 declare i8 *@llvm.frameaddress(i32 <level>)
4266 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4267 target-specific frame pointer value for the specified stack frame.
4273 The argument to this intrinsic indicates which function to return the frame
4274 pointer for. Zero indicates the calling function, one indicates its caller,
4275 etc. The argument is <b>required</b> to be a constant integer value.
4281 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4282 the frame address of the specified call frame, or zero if it cannot be
4283 identified. The value returned by this intrinsic is likely to be incorrect or 0
4284 for arguments other than zero, so it should only be used for debugging purposes.
4288 Note that calling this intrinsic does not prevent function inlining or other
4289 aggressive transformations, so the value returned may not be that of the obvious
4290 source-language caller.
4294 <!-- _______________________________________________________________________ -->
4295 <div class="doc_subsubsection">
4296 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4299 <div class="doc_text">
4303 declare i8 *@llvm.stacksave()
4309 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4310 the function stack, for use with <a href="#int_stackrestore">
4311 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4312 features like scoped automatic variable sized arrays in C99.
4318 This intrinsic returns a opaque pointer value that can be passed to <a
4319 href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4320 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4321 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4322 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4323 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4324 that were allocated after the <tt>llvm.stacksave</tt> was executed.
4329 <!-- _______________________________________________________________________ -->
4330 <div class="doc_subsubsection">
4331 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4334 <div class="doc_text">
4338 declare void @llvm.stackrestore(i8 * %ptr)
4344 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4345 the function stack to the state it was in when the corresponding <a
4346 href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4347 useful for implementing language features like scoped automatic variable sized
4354 See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4360 <!-- _______________________________________________________________________ -->
4361 <div class="doc_subsubsection">
4362 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4365 <div class="doc_text">
4369 declare void @llvm.prefetch(i8* <address>, i32 <rw>, i32 <locality>)
4376 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4377 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4379 effect on the behavior of the program but can change its performance
4386 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4387 determining if the fetch should be for a read (0) or write (1), and
4388 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4389 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4390 <tt>locality</tt> arguments must be constant integers.
4396 This intrinsic does not modify the behavior of the program. In particular,
4397 prefetches cannot trap and do not produce a value. On targets that support this
4398 intrinsic, the prefetch can provide hints to the processor cache for better
4404 <!-- _______________________________________________________________________ -->
4405 <div class="doc_subsubsection">
4406 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4409 <div class="doc_text">
4413 declare void @llvm.pcmarker(i32 <id>)
4420 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4422 code to simulators and other tools. The method is target specific, but it is
4423 expected that the marker will use exported symbols to transmit the PC of the marker.
4424 The marker makes no guarantees that it will remain with any specific instruction
4425 after optimizations. It is possible that the presence of a marker will inhibit
4426 optimizations. The intended use is to be inserted after optimizations to allow
4427 correlations of simulation runs.
4433 <tt>id</tt> is a numerical id identifying the marker.
4439 This intrinsic does not modify the behavior of the program. Backends that do not
4440 support this intrinisic may ignore it.
4445 <!-- _______________________________________________________________________ -->
4446 <div class="doc_subsubsection">
4447 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4450 <div class="doc_text">
4454 declare i64 @llvm.readcyclecounter( )
4461 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4462 counter register (or similar low latency, high accuracy clocks) on those targets
4463 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4464 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4465 should only be used for small timings.
4471 When directly supported, reading the cycle counter should not modify any memory.
4472 Implementations are allowed to either return a application specific value or a
4473 system wide value. On backends without support, this is lowered to a constant 0.
4478 <!-- ======================================================================= -->
4479 <div class="doc_subsection">
4480 <a name="int_libc">Standard C Library Intrinsics</a>
4483 <div class="doc_text">
4485 LLVM provides intrinsics for a few important standard C library functions.
4486 These intrinsics allow source-language front-ends to pass information about the
4487 alignment of the pointer arguments to the code generator, providing opportunity
4488 for more efficient code generation.
4493 <!-- _______________________________________________________________________ -->
4494 <div class="doc_subsubsection">
4495 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4498 <div class="doc_text">
4502 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
4503 i32 <len>, i32 <align>)
4504 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
4505 i64 <len>, i32 <align>)
4511 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4512 location to the destination location.
4516 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4517 intrinsics do not return a value, and takes an extra alignment argument.
4523 The first argument is a pointer to the destination, the second is a pointer to
4524 the source. The third argument is an integer argument
4525 specifying the number of bytes to copy, and the fourth argument is the alignment
4526 of the source and destination locations.
4530 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4531 the caller guarantees that both the source and destination pointers are aligned
4538 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4539 location to the destination location, which are not allowed to overlap. It
4540 copies "len" bytes of memory over. If the argument is known to be aligned to
4541 some boundary, this can be specified as the fourth argument, otherwise it should
4547 <!-- _______________________________________________________________________ -->
4548 <div class="doc_subsubsection">
4549 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4552 <div class="doc_text">
4556 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
4557 i32 <len>, i32 <align>)
4558 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
4559 i64 <len>, i32 <align>)
4565 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4566 location to the destination location. It is similar to the
4567 '<tt>llvm.memcpy</tt>' intrinsic but allows the two memory locations to overlap.
4571 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4572 intrinsics do not return a value, and takes an extra alignment argument.
4578 The first argument is a pointer to the destination, the second is a pointer to
4579 the source. The third argument is an integer argument
4580 specifying the number of bytes to copy, and the fourth argument is the alignment
4581 of the source and destination locations.
4585 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4586 the caller guarantees that the source and destination pointers are aligned to
4593 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4594 location to the destination location, which may overlap. It
4595 copies "len" bytes of memory over. If the argument is known to be aligned to
4596 some boundary, this can be specified as the fourth argument, otherwise it should
4602 <!-- _______________________________________________________________________ -->
4603 <div class="doc_subsubsection">
4604 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4607 <div class="doc_text">
4611 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
4612 i32 <len>, i32 <align>)
4613 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
4614 i64 <len>, i32 <align>)
4620 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4625 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4626 does not return a value, and takes an extra alignment argument.
4632 The first argument is a pointer to the destination to fill, the second is the
4633 byte value to fill it with, the third argument is an integer
4634 argument specifying the number of bytes to fill, and the fourth argument is the
4635 known alignment of destination location.
4639 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4640 the caller guarantees that the destination pointer is aligned to that boundary.
4646 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4648 destination location. If the argument is known to be aligned to some boundary,
4649 this can be specified as the fourth argument, otherwise it should be set to 0 or
4655 <!-- _______________________________________________________________________ -->
4656 <div class="doc_subsubsection">
4657 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4660 <div class="doc_text">
4663 <p>This is an overloaded intrinsic. You can use <tt>llvm.sqrt</tt> on any
4664 floating point or vector of floating point type. Not all targets support all
4667 declare float @llvm.sqrt.f32(float %Val)
4668 declare double @llvm.sqrt.f64(double %Val)
4669 declare x86_fp80 @llvm.sqrt.f80(x86_fp80 %Val)
4670 declare fp128 @llvm.sqrt.f128(fp128 %Val)
4671 declare ppc_fp128 @llvm.sqrt.ppcf128(ppc_fp128 %Val)
4677 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
4678 returning the same value as the libm '<tt>sqrt</tt>' functions would. Unlike
4679 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4680 negative numbers (which allows for better optimization).
4686 The argument and return value are floating point numbers of the same type.
4692 This function returns the sqrt of the specified operand if it is a nonnegative
4693 floating point number.
4697 <!-- _______________________________________________________________________ -->
4698 <div class="doc_subsubsection">
4699 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4702 <div class="doc_text">
4705 <p>This is an overloaded intrinsic. You can use <tt>llvm.powi</tt> on any
4706 floating point or vector of floating point type. Not all targets support all
4709 declare float @llvm.powi.f32(float %Val, i32 %power)
4710 declare double @llvm.powi.f64(double %Val, i32 %power)
4711 declare x86_fp80 @llvm.powi.f80(x86_fp80 %Val, i32 %power)
4712 declare fp128 @llvm.powi.f128(fp128 %Val, i32 %power)
4713 declare ppc_fp128 @llvm.powi.ppcf128(ppc_fp128 %Val, i32 %power)
4719 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4720 specified (positive or negative) power. The order of evaluation of
4721 multiplications is not defined. When a vector of floating point type is
4722 used, the second argument remains a scalar integer value.
4728 The second argument is an integer power, and the first is a value to raise to
4735 This function returns the first value raised to the second power with an
4736 unspecified sequence of rounding operations.</p>
4739 <!-- _______________________________________________________________________ -->
4740 <div class="doc_subsubsection">
4741 <a name="int_sin">'<tt>llvm.sin.*</tt>' Intrinsic</a>
4744 <div class="doc_text">
4747 <p>This is an overloaded intrinsic. You can use <tt>llvm.sin</tt> on any
4748 floating point or vector of floating point type. Not all targets support all
4751 declare float @llvm.sin.f32(float %Val)
4752 declare double @llvm.sin.f64(double %Val)
4753 declare x86_fp80 @llvm.sin.f80(x86_fp80 %Val)
4754 declare fp128 @llvm.sin.f128(fp128 %Val)
4755 declare ppc_fp128 @llvm.sin.ppcf128(ppc_fp128 %Val)
4761 The '<tt>llvm.sin.*</tt>' intrinsics return the sine of the operand.
4767 The argument and return value are floating point numbers of the same type.
4773 This function returns the sine of the specified operand, returning the
4774 same values as the libm <tt>sin</tt> functions would, and handles error
4775 conditions in the same way.</p>
4778 <!-- _______________________________________________________________________ -->
4779 <div class="doc_subsubsection">
4780 <a name="int_cos">'<tt>llvm.cos.*</tt>' Intrinsic</a>
4783 <div class="doc_text">
4786 <p>This is an overloaded intrinsic. You can use <tt>llvm.cos</tt> on any
4787 floating point or vector of floating point type. Not all targets support all
4790 declare float @llvm.cos.f32(float %Val)
4791 declare double @llvm.cos.f64(double %Val)
4792 declare x86_fp80 @llvm.cos.f80(x86_fp80 %Val)
4793 declare fp128 @llvm.cos.f128(fp128 %Val)
4794 declare ppc_fp128 @llvm.cos.ppcf128(ppc_fp128 %Val)
4800 The '<tt>llvm.cos.*</tt>' intrinsics return the cosine of the operand.
4806 The argument and return value are floating point numbers of the same type.
4812 This function returns the cosine of the specified operand, returning the
4813 same values as the libm <tt>cos</tt> functions would, and handles error
4814 conditions in the same way.</p>
4817 <!-- _______________________________________________________________________ -->
4818 <div class="doc_subsubsection">
4819 <a name="int_pow">'<tt>llvm.pow.*</tt>' Intrinsic</a>
4822 <div class="doc_text">
4825 <p>This is an overloaded intrinsic. You can use <tt>llvm.pow</tt> on any
4826 floating point or vector of floating point type. Not all targets support all
4829 declare float @llvm.pow.f32(float %Val, float %Power)
4830 declare double @llvm.pow.f64(double %Val, double %Power)
4831 declare x86_fp80 @llvm.pow.f80(x86_fp80 %Val, x86_fp80 %Power)
4832 declare fp128 @llvm.pow.f128(fp128 %Val, fp128 %Power)
4833 declare ppc_fp128 @llvm.pow.ppcf128(ppc_fp128 %Val, ppc_fp128 Power)
4839 The '<tt>llvm.pow.*</tt>' intrinsics return the first operand raised to the
4840 specified (positive or negative) power.
4846 The second argument is a floating point power, and the first is a value to
4847 raise to that power.
4853 This function returns the first value raised to the second power,
4855 same values as the libm <tt>pow</tt> functions would, and handles error
4856 conditions in the same way.</p>
4860 <!-- ======================================================================= -->
4861 <div class="doc_subsection">
4862 <a name="int_manip">Bit Manipulation Intrinsics</a>
4865 <div class="doc_text">
4867 LLVM provides intrinsics for a few important bit manipulation operations.
4868 These allow efficient code generation for some algorithms.
4873 <!-- _______________________________________________________________________ -->
4874 <div class="doc_subsubsection">
4875 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4878 <div class="doc_text">
4881 <p>This is an overloaded intrinsic function. You can use bswap on any integer
4882 type that is an even number of bytes (i.e. BitWidth % 16 == 0).
4884 declare i16 @llvm.bswap.i16(i16 <id>)
4885 declare i32 @llvm.bswap.i32(i32 <id>)
4886 declare i64 @llvm.bswap.i64(i64 <id>)
4892 The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
4893 values with an even number of bytes (positive multiple of 16 bits). These are
4894 useful for performing operations on data that is not in the target's native
4901 The <tt>llvm.bswap.i16</tt> intrinsic returns an i16 value that has the high
4902 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4903 intrinsic returns an i32 value that has the four bytes of the input i32
4904 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
4905 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48</tt>,
4906 <tt>llvm.bswap.i64</tt> and other intrinsics extend this concept to
4907 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4912 <!-- _______________________________________________________________________ -->
4913 <div class="doc_subsubsection">
4914 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4917 <div class="doc_text">
4920 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4921 width. Not all targets support all bit widths however.
4923 declare i8 @llvm.ctpop.i8 (i8 <src>)
4924 declare i16 @llvm.ctpop.i16(i16 <src>)
4925 declare i32 @llvm.ctpop.i32(i32 <src>)
4926 declare i64 @llvm.ctpop.i64(i64 <src>)
4927 declare i256 @llvm.ctpop.i256(i256 <src>)
4933 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4940 The only argument is the value to be counted. The argument may be of any
4941 integer type. The return type must match the argument type.
4947 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4951 <!-- _______________________________________________________________________ -->
4952 <div class="doc_subsubsection">
4953 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4956 <div class="doc_text">
4959 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4960 integer bit width. Not all targets support all bit widths however.
4962 declare i8 @llvm.ctlz.i8 (i8 <src>)
4963 declare i16 @llvm.ctlz.i16(i16 <src>)
4964 declare i32 @llvm.ctlz.i32(i32 <src>)
4965 declare i64 @llvm.ctlz.i64(i64 <src>)
4966 declare i256 @llvm.ctlz.i256(i256 <src>)
4972 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4973 leading zeros in a variable.
4979 The only argument is the value to be counted. The argument may be of any
4980 integer type. The return type must match the argument type.
4986 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4987 in a variable. If the src == 0 then the result is the size in bits of the type
4988 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4994 <!-- _______________________________________________________________________ -->
4995 <div class="doc_subsubsection">
4996 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4999 <div class="doc_text">
5002 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
5003 integer bit width. Not all targets support all bit widths however.
5005 declare i8 @llvm.cttz.i8 (i8 <src>)
5006 declare i16 @llvm.cttz.i16(i16 <src>)
5007 declare i32 @llvm.cttz.i32(i32 <src>)
5008 declare i64 @llvm.cttz.i64(i64 <src>)
5009 declare i256 @llvm.cttz.i256(i256 <src>)
5015 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
5022 The only argument is the value to be counted. The argument may be of any
5023 integer type. The return type must match the argument type.
5029 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
5030 in a variable. If the src == 0 then the result is the size in bits of the type
5031 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
5035 <!-- _______________________________________________________________________ -->
5036 <div class="doc_subsubsection">
5037 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
5040 <div class="doc_text">
5043 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
5044 on any integer bit width.
5046 declare i17 @llvm.part.select.i17 (i17 %val, i32 %loBit, i32 %hiBit)
5047 declare i29 @llvm.part.select.i29 (i29 %val, i32 %loBit, i32 %hiBit)
5051 <p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
5052 range of bits from an integer value and returns them in the same bit width as
5053 the original value.</p>
5056 <p>The first argument, <tt>%val</tt> and the result may be integer types of
5057 any bit width but they must have the same bit width. The second and third
5058 arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
5061 <p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
5062 of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
5063 <tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
5064 operates in forward mode.</p>
5065 <p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
5066 right by <tt>%loBit</tt> bits and then ANDing it with a mask with
5067 only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
5069 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
5070 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
5071 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
5072 to determine the number of bits to retain.</li>
5073 <li>A mask of the retained bits is created by shifting a -1 value.</li>
5074 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
5076 <p>In reverse mode, a similar computation is made except that the bits are
5077 returned in the reverse order. So, for example, if <tt>X</tt> has the value
5078 <tt>i16 0x0ACF (101011001111)</tt> and we apply
5079 <tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
5080 <tt>i16 0x0026 (000000100110)</tt>.</p>
5083 <div class="doc_subsubsection">
5084 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
5087 <div class="doc_text">
5090 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
5091 on any integer bit width.
5093 declare i17 @llvm.part.set.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
5094 declare i29 @llvm.part.set.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
5098 <p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
5099 of bits in an integer value with another integer value. It returns the integer
5100 with the replaced bits.</p>
5103 <p>The first argument, <tt>%val</tt> and the result may be integer types of
5104 any bit width but they must have the same bit width. <tt>%val</tt> is the value
5105 whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
5106 integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
5107 type since they specify only a bit index.</p>
5110 <p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
5111 of operation: forwards and reverse. If <tt>%lo</tt> is greater than
5112 <tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
5113 operates in forward mode.</p>
5114 <p>For both modes, the <tt>%repl</tt> value is prepared for use by either
5115 truncating it down to the size of the replacement area or zero extending it
5116 up to that size.</p>
5117 <p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
5118 are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
5119 in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
5120 to the <tt>%hi</tt>th bit.
5121 <p>In reverse mode, a similar computation is made except that the bits are
5122 reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
5123 <tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
5126 llvm.part.set(0xFFFF, 0, 4, 7) -> 0xFF0F
5127 llvm.part.set(0xFFFF, 0, 7, 4) -> 0xFF0F
5128 llvm.part.set(0xFFFF, 1, 7, 4) -> 0xFF8F
5129 llvm.part.set(0xFFFF, F, 8, 3) -> 0xFFE7
5130 llvm.part.set(0xFFFF, 0, 3, 8) -> 0xFE07
5134 <!-- ======================================================================= -->
5135 <div class="doc_subsection">
5136 <a name="int_debugger">Debugger Intrinsics</a>
5139 <div class="doc_text">
5141 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
5142 are described in the <a
5143 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
5144 Debugging</a> document.
5149 <!-- ======================================================================= -->
5150 <div class="doc_subsection">
5151 <a name="int_eh">Exception Handling Intrinsics</a>
5154 <div class="doc_text">
5155 <p> The LLVM exception handling intrinsics (which all start with
5156 <tt>llvm.eh.</tt> prefix), are described in the <a
5157 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
5158 Handling</a> document. </p>
5161 <!-- ======================================================================= -->
5162 <div class="doc_subsection">
5163 <a name="int_trampoline">Trampoline Intrinsic</a>
5166 <div class="doc_text">
5168 This intrinsic makes it possible to excise one parameter, marked with
5169 the <tt>nest</tt> attribute, from a function. The result is a callable
5170 function pointer lacking the nest parameter - the caller does not need
5171 to provide a value for it. Instead, the value to use is stored in
5172 advance in a "trampoline", a block of memory usually allocated
5173 on the stack, which also contains code to splice the nest value into the
5174 argument list. This is used to implement the GCC nested function address
5178 For example, if the function is
5179 <tt>i32 f(i8* nest %c, i32 %x, i32 %y)</tt> then the resulting function
5180 pointer has signature <tt>i32 (i32, i32)*</tt>. It can be created as follows:</p>
5182 %tramp = alloca [10 x i8], align 4 ; size and alignment only correct for X86
5183 %tramp1 = getelementptr [10 x i8]* %tramp, i32 0, i32 0
5184 %p = call i8* @llvm.init.trampoline( i8* %tramp1, i8* bitcast (i32 (i8* nest , i32, i32)* @f to i8*), i8* %nval )
5185 %fp = bitcast i8* %p to i32 (i32, i32)*
5187 <p>The call <tt>%val = call i32 %fp( i32 %x, i32 %y )</tt> is then equivalent
5188 to <tt>%val = call i32 %f( i8* %nval, i32 %x, i32 %y )</tt>.</p>
5191 <!-- _______________________________________________________________________ -->
5192 <div class="doc_subsubsection">
5193 <a name="int_it">'<tt>llvm.init.trampoline</tt>' Intrinsic</a>
5195 <div class="doc_text">
5198 declare i8* @llvm.init.trampoline(i8* <tramp>, i8* <func>, i8* <nval>)
5202 This fills the memory pointed to by <tt>tramp</tt> with code
5203 and returns a function pointer suitable for executing it.
5207 The <tt>llvm.init.trampoline</tt> intrinsic takes three arguments, all
5208 pointers. The <tt>tramp</tt> argument must point to a sufficiently large
5209 and sufficiently aligned block of memory; this memory is written to by the
5210 intrinsic. Note that the size and the alignment are target-specific - LLVM
5211 currently provides no portable way of determining them, so a front-end that
5212 generates this intrinsic needs to have some target-specific knowledge.
5213 The <tt>func</tt> argument must hold a function bitcast to an <tt>i8*</tt>.
5217 The block of memory pointed to by <tt>tramp</tt> is filled with target
5218 dependent code, turning it into a function. A pointer to this function is
5219 returned, but needs to be bitcast to an
5220 <a href="#int_trampoline">appropriate function pointer type</a>
5221 before being called. The new function's signature is the same as that of
5222 <tt>func</tt> with any arguments marked with the <tt>nest</tt> attribute
5223 removed. At most one such <tt>nest</tt> argument is allowed, and it must be
5224 of pointer type. Calling the new function is equivalent to calling
5225 <tt>func</tt> with the same argument list, but with <tt>nval</tt> used for the
5226 missing <tt>nest</tt> argument. If, after calling
5227 <tt>llvm.init.trampoline</tt>, the memory pointed to by <tt>tramp</tt> is
5228 modified, then the effect of any later call to the returned function pointer is
5233 <!-- ======================================================================= -->
5234 <div class="doc_subsection">
5235 <a name="int_general">General Intrinsics</a>
5238 <div class="doc_text">
5239 <p> This class of intrinsics is designed to be generic and has
5240 no specific purpose. </p>
5243 <!-- _______________________________________________________________________ -->
5244 <div class="doc_subsubsection">
5245 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
5248 <div class="doc_text">
5252 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
5258 The '<tt>llvm.var.annotation</tt>' intrinsic
5264 The first argument is a pointer to a value, the second is a pointer to a
5265 global string, the third is a pointer to a global string which is the source
5266 file name, and the last argument is the line number.
5272 This intrinsic allows annotation of local variables with arbitrary strings.
5273 This can be useful for special purpose optimizations that want to look for these
5274 annotations. These have no other defined use, they are ignored by code
5275 generation and optimization.
5279 <!-- _______________________________________________________________________ -->
5280 <div class="doc_subsubsection">
5281 <a name="int_annotation">'<tt>llvm.annotation.*</tt>' Intrinsic</a>
5284 <div class="doc_text">
5287 <p>This is an overloaded intrinsic. You can use '<tt>llvm.annotation</tt>' on
5288 any integer bit width.
5291 declare i8 @llvm.annotation.i8(i8 <val>, i8* <str>, i8* <str>, i32 <int> )
5292 declare i16 @llvm.annotation.i16(i16 <val>, i8* <str>, i8* <str>, i32 <int> )
5293 declare i32 @llvm.annotation.i32(i32 <val>, i8* <str>, i8* <str>, i32 <int> )
5294 declare i64 @llvm.annotation.i64(i64 <val>, i8* <str>, i8* <str>, i32 <int> )
5295 declare i256 @llvm.annotation.i256(i256 <val>, i8* <str>, i8* <str>, i32 <int> )
5301 The '<tt>llvm.annotation</tt>' intrinsic.
5307 The first argument is an integer value (result of some expression),
5308 the second is a pointer to a global string, the third is a pointer to a global
5309 string which is the source file name, and the last argument is the line number.
5310 It returns the value of the first argument.
5316 This intrinsic allows annotations to be put on arbitrary expressions
5317 with arbitrary strings. This can be useful for special purpose optimizations
5318 that want to look for these annotations. These have no other defined use, they
5319 are ignored by code generation and optimization.
5322 <!-- _______________________________________________________________________ -->
5323 <div class="doc_subsubsection">
5324 <a name="int_trap">'<tt>llvm.trap</tt>' Intrinsic</a>
5327 <div class="doc_text">
5331 declare void @llvm.trap()
5337 The '<tt>llvm.trap</tt>' intrinsic
5349 This intrinsics is lowered to the target dependent trap instruction. If the
5350 target does not have a trap instruction, this intrinsic will be lowered to the
5351 call of the abort() function.
5355 <!-- *********************************************************************** -->
5358 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
5359 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
5360 <a href="http://validator.w3.org/check/referer"><img
5361 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
5363 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
5364 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
5365 Last modified: $Date$