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="#paramattrs">Parameter Attributes</a></li>
28 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
29 <li><a href="#datalayout">Data Layout</a></li>
32 <li><a href="#typesystem">Type System</a>
34 <li><a href="#t_primitive">Primitive Types</a>
36 <li><a href="#t_classifications">Type Classifications</a></li>
39 <li><a href="#t_derived">Derived Types</a>
41 <li><a href="#t_array">Array Type</a></li>
42 <li><a href="#t_function">Function Type</a></li>
43 <li><a href="#t_pointer">Pointer Type</a></li>
44 <li><a href="#t_struct">Structure Type</a></li>
45 <li><a href="#t_pstruct">Packed Structure Type</a></li>
46 <li><a href="#t_vector">Vector Type</a></li>
47 <li><a href="#t_opaque">Opaque Type</a></li>
52 <li><a href="#constants">Constants</a>
54 <li><a href="#simpleconstants">Simple Constants</a>
55 <li><a href="#aggregateconstants">Aggregate Constants</a>
56 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
57 <li><a href="#undefvalues">Undefined Values</a>
58 <li><a href="#constantexprs">Constant Expressions</a>
61 <li><a href="#othervalues">Other Values</a>
63 <li><a href="#inlineasm">Inline Assembler Expressions</a>
66 <li><a href="#instref">Instruction Reference</a>
68 <li><a href="#terminators">Terminator Instructions</a>
70 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
71 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
72 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
73 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
74 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
75 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
78 <li><a href="#binaryops">Binary Operations</a>
80 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
81 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
82 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
83 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
84 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
85 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
86 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
87 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
88 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
91 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
93 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
94 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
95 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
96 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
97 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
98 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
101 <li><a href="#vectorops">Vector Operations</a>
103 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
104 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
105 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
108 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
110 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
111 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
112 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
113 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
114 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
115 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
118 <li><a href="#convertops">Conversion Operations</a>
120 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
121 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
122 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
123 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
124 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
125 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
126 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
127 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
128 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
129 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
130 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
131 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
133 <li><a href="#otherops">Other Operations</a>
135 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
136 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
137 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
138 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
139 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
140 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
145 <li><a href="#intrinsics">Intrinsic Functions</a>
147 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
149 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
150 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
151 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
154 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
156 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
157 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
158 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
161 <li><a href="#int_codegen">Code Generator Intrinsics</a>
163 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
164 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
165 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
166 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
167 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
168 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
169 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
172 <li><a href="#int_libc">Standard C Library Intrinsics</a>
174 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
175 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
176 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
177 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
178 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
181 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
183 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
184 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
185 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
186 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
187 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
188 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
191 <li><a href="#int_debugger">Debugger intrinsics</a></li>
192 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
197 <div class="doc_author">
198 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
199 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
202 <!-- *********************************************************************** -->
203 <div class="doc_section"> <a name="abstract">Abstract </a></div>
204 <!-- *********************************************************************** -->
206 <div class="doc_text">
207 <p>This document is a reference manual for the LLVM assembly language.
208 LLVM is an SSA based representation that provides type safety,
209 low-level operations, flexibility, and the capability of representing
210 'all' high-level languages cleanly. It is the common code
211 representation used throughout all phases of the LLVM compilation
215 <!-- *********************************************************************** -->
216 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
217 <!-- *********************************************************************** -->
219 <div class="doc_text">
221 <p>The LLVM code representation is designed to be used in three
222 different forms: as an in-memory compiler IR, as an on-disk bytecode
223 representation (suitable for fast loading by a Just-In-Time compiler),
224 and as a human readable assembly language representation. This allows
225 LLVM to provide a powerful intermediate representation for efficient
226 compiler transformations and analysis, while providing a natural means
227 to debug and visualize the transformations. The three different forms
228 of LLVM are all equivalent. This document describes the human readable
229 representation and notation.</p>
231 <p>The LLVM representation aims to be light-weight and low-level
232 while being expressive, typed, and extensible at the same time. It
233 aims to be a "universal IR" of sorts, by being at a low enough level
234 that high-level ideas may be cleanly mapped to it (similar to how
235 microprocessors are "universal IR's", allowing many source languages to
236 be mapped to them). By providing type information, LLVM can be used as
237 the target of optimizations: for example, through pointer analysis, it
238 can be proven that a C automatic variable is never accessed outside of
239 the current function... allowing it to be promoted to a simple SSA
240 value instead of a memory location.</p>
244 <!-- _______________________________________________________________________ -->
245 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
247 <div class="doc_text">
249 <p>It is important to note that this document describes 'well formed'
250 LLVM assembly language. There is a difference between what the parser
251 accepts and what is considered 'well formed'. For example, the
252 following instruction is syntactically okay, but not well formed:</p>
255 %x = <a href="#i_add">add</a> i32 1, %x
258 <p>...because the definition of <tt>%x</tt> does not dominate all of
259 its uses. The LLVM infrastructure provides a verification pass that may
260 be used to verify that an LLVM module is well formed. This pass is
261 automatically run by the parser after parsing input assembly and by
262 the optimizer before it outputs bytecode. The violations pointed out
263 by the verifier pass indicate bugs in transformation passes or input to
266 <!-- Describe the typesetting conventions here. --> </div>
268 <!-- *********************************************************************** -->
269 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
270 <!-- *********************************************************************** -->
272 <div class="doc_text">
274 <p>LLVM uses three different forms of identifiers, for different
278 <li>Named values are represented as a string of characters with a '%' prefix.
279 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
280 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
281 Identifiers which require other characters in their names can be surrounded
282 with quotes. In this way, anything except a <tt>"</tt> character can be used
285 <li>Unnamed values are represented as an unsigned numeric value with a '%'
286 prefix. For example, %12, %2, %44.</li>
288 <li>Constants, which are described in a <a href="#constants">section about
289 constants</a>, below.</li>
292 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
293 don't need to worry about name clashes with reserved words, and the set of
294 reserved words may be expanded in the future without penalty. Additionally,
295 unnamed identifiers allow a compiler to quickly come up with a temporary
296 variable without having to avoid symbol table conflicts.</p>
298 <p>Reserved words in LLVM are very similar to reserved words in other
299 languages. There are keywords for different opcodes
300 ('<tt><a href="#i_add">add</a></tt>',
301 '<tt><a href="#i_bitcast">bitcast</a></tt>',
302 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
303 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
304 and others. These reserved words cannot conflict with variable names, because
305 none of them start with a '%' character.</p>
307 <p>Here is an example of LLVM code to multiply the integer variable
308 '<tt>%X</tt>' by 8:</p>
313 %result = <a href="#i_mul">mul</a> i32 %X, 8
316 <p>After strength reduction:</p>
319 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
322 <p>And the hard way:</p>
325 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
326 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
327 %result = <a href="#i_add">add</a> i32 %1, %1
330 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
331 important lexical features of LLVM:</p>
335 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
338 <li>Unnamed temporaries are created when the result of a computation is not
339 assigned to a named value.</li>
341 <li>Unnamed temporaries are numbered sequentially</li>
345 <p>...and it also shows a convention that we follow in this document. When
346 demonstrating instructions, we will follow an instruction with a comment that
347 defines the type and name of value produced. Comments are shown in italic
352 <!-- *********************************************************************** -->
353 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
354 <!-- *********************************************************************** -->
356 <!-- ======================================================================= -->
357 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
360 <div class="doc_text">
362 <p>LLVM programs are composed of "Module"s, each of which is a
363 translation unit of the input programs. Each module consists of
364 functions, global variables, and symbol table entries. Modules may be
365 combined together with the LLVM linker, which merges function (and
366 global variable) definitions, resolves forward declarations, and merges
367 symbol table entries. Here is an example of the "hello world" module:</p>
369 <pre><i>; Declare the string constant as a global constant...</i>
370 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
371 href="#globalvars">constant</a> <a href="#t_array">[13 x i8 ]</a> c"hello world\0A\00" <i>; [13 x i8 ]*</i>
373 <i>; External declaration of the puts function</i>
374 <a href="#functionstructure">declare</a> i32 %puts(i8 *) <i>; i32(i8 *)* </i>
376 <i>; Definition of main function</i>
377 define i32 %main() { <i>; i32()* </i>
378 <i>; Convert [13x i8 ]* to i8 *...</i>
380 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* %.LC0, i64 0, i64 0 <i>; i8 *</i>
382 <i>; Call puts function to write out the string to stdout...</i>
384 href="#i_call">call</a> i32 %puts(i8 * %cast210) <i>; i32</i>
386 href="#i_ret">ret</a> i32 0<br>}<br></pre>
388 <p>This example is made up of a <a href="#globalvars">global variable</a>
389 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
390 function, and a <a href="#functionstructure">function definition</a>
391 for "<tt>main</tt>".</p>
393 <p>In general, a module is made up of a list of global values,
394 where both functions and global variables are global values. Global values are
395 represented by a pointer to a memory location (in this case, a pointer to an
396 array of char, and a pointer to a function), and have one of the following <a
397 href="#linkage">linkage types</a>.</p>
401 <!-- ======================================================================= -->
402 <div class="doc_subsection">
403 <a name="linkage">Linkage Types</a>
406 <div class="doc_text">
409 All Global Variables and Functions have one of the following types of linkage:
414 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
416 <dd>Global values with internal linkage are only directly accessible by
417 objects in the current module. In particular, linking code into a module with
418 an internal global value may cause the internal to be renamed as necessary to
419 avoid collisions. Because the symbol is internal to the module, all
420 references can be updated. This corresponds to the notion of the
421 '<tt>static</tt>' keyword in C.
424 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
426 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
427 the same name when linkage occurs. This is typically used to implement
428 inline functions, templates, or other code which must be generated in each
429 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
430 allowed to be discarded.
433 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
435 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
436 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
437 used for globals that may be emitted in multiple translation units, but that
438 are not guaranteed to be emitted into every translation unit that uses them.
439 One example of this are common globals in C, such as "<tt>int X;</tt>" at
443 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
445 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
446 pointer to array type. When two global variables with appending linkage are
447 linked together, the two global arrays are appended together. This is the
448 LLVM, typesafe, equivalent of having the system linker append together
449 "sections" with identical names when .o files are linked.
452 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
453 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
454 until linked, if not linked, the symbol becomes null instead of being an
458 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
460 <dd>If none of the above identifiers are used, the global is externally
461 visible, meaning that it participates in linkage and can be used to resolve
462 external symbol references.
467 The next two types of linkage are targeted for Microsoft Windows platform
468 only. They are designed to support importing (exporting) symbols from (to)
473 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
475 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
476 or variable via a global pointer to a pointer that is set up by the DLL
477 exporting the symbol. On Microsoft Windows targets, the pointer name is
478 formed by combining <code>_imp__</code> and the function or variable name.
481 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
483 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
484 pointer to a pointer in a DLL, so that it can be referenced with the
485 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
486 name is formed by combining <code>_imp__</code> and the function or variable
492 <p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
493 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
494 variable and was linked with this one, one of the two would be renamed,
495 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
496 external (i.e., lacking any linkage declarations), they are accessible
497 outside of the current module.</p>
498 <p>It is illegal for a function <i>declaration</i>
499 to have any linkage type other than "externally visible", <tt>dllimport</tt>,
500 or <tt>extern_weak</tt>.</p>
501 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
505 <!-- ======================================================================= -->
506 <div class="doc_subsection">
507 <a name="callingconv">Calling Conventions</a>
510 <div class="doc_text">
512 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
513 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
514 specified for the call. The calling convention of any pair of dynamic
515 caller/callee must match, or the behavior of the program is undefined. The
516 following calling conventions are supported by LLVM, and more may be added in
520 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
522 <dd>This calling convention (the default if no other calling convention is
523 specified) matches the target C calling conventions. This calling convention
524 supports varargs function calls and tolerates some mismatch in the declared
525 prototype and implemented declaration of the function (as does normal C).
528 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
530 <dd>This calling convention attempts to make calls as fast as possible
531 (e.g. by passing things in registers). This calling convention allows the
532 target to use whatever tricks it wants to produce fast code for the target,
533 without having to conform to an externally specified ABI. Implementations of
534 this convention should allow arbitrary tail call optimization to be supported.
535 This calling convention does not support varargs and requires the prototype of
536 all callees to exactly match the prototype of the function definition.
539 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
541 <dd>This calling convention attempts to make code in the caller as efficient
542 as possible under the assumption that the call is not commonly executed. As
543 such, these calls often preserve all registers so that the call does not break
544 any live ranges in the caller side. This calling convention does not support
545 varargs and requires the prototype of all callees to exactly match the
546 prototype of the function definition.
549 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
551 <dd>Any calling convention may be specified by number, allowing
552 target-specific calling conventions to be used. Target specific calling
553 conventions start at 64.
557 <p>More calling conventions can be added/defined on an as-needed basis, to
558 support pascal conventions or any other well-known target-independent
563 <!-- ======================================================================= -->
564 <div class="doc_subsection">
565 <a name="visibility">Visibility Styles</a>
568 <div class="doc_text">
571 All Global Variables and Functions have one of the following visibility styles:
575 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
577 <dd>On ELF, default visibility means that the declaration is visible to other
578 modules and, in shared libraries, means that the declared entity may be
579 overridden. On Darwin, default visibility means that the declaration is
580 visible to other modules. Default visibility corresponds to "external
581 linkage" in the language.
584 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
586 <dd>Two declarations of an object with hidden visibility refer to the same
587 object if they are in the same shared object. Usually, hidden visibility
588 indicates that the symbol will not be placed into the dynamic symbol table,
589 so no other module (executable or shared library) can reference it
597 <!-- ======================================================================= -->
598 <div class="doc_subsection">
599 <a name="globalvars">Global Variables</a>
602 <div class="doc_text">
604 <p>Global variables define regions of memory allocated at compilation time
605 instead of run-time. Global variables may optionally be initialized, may have
606 an explicit section to be placed in, and may have an optional explicit alignment
607 specified. A variable may be defined as "thread_local", which means that it
608 will not be shared by threads (each thread will have a separated copy of the
609 variable). A variable may be defined as a global "constant," which indicates
610 that the contents of the variable will <b>never</b> be modified (enabling better
611 optimization, allowing the global data to be placed in the read-only section of
612 an executable, etc). Note that variables that need runtime initialization
613 cannot be marked "constant" as there is a store to the variable.</p>
616 LLVM explicitly allows <em>declarations</em> of global variables to be marked
617 constant, even if the final definition of the global is not. This capability
618 can be used to enable slightly better optimization of the program, but requires
619 the language definition to guarantee that optimizations based on the
620 'constantness' are valid for the translation units that do not include the
624 <p>As SSA values, global variables define pointer values that are in
625 scope (i.e. they dominate) all basic blocks in the program. Global
626 variables always define a pointer to their "content" type because they
627 describe a region of memory, and all memory objects in LLVM are
628 accessed through pointers.</p>
630 <p>LLVM allows an explicit section to be specified for globals. If the target
631 supports it, it will emit globals to the section specified.</p>
633 <p>An explicit alignment may be specified for a global. If not present, or if
634 the alignment is set to zero, the alignment of the global is set by the target
635 to whatever it feels convenient. If an explicit alignment is specified, the
636 global is forced to have at least that much alignment. All alignments must be
639 <p>For example, the following defines a global with an initializer, section,
643 %G = constant float 1.0, section "foo", align 4
649 <!-- ======================================================================= -->
650 <div class="doc_subsection">
651 <a name="functionstructure">Functions</a>
654 <div class="doc_text">
656 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
657 an optional <a href="#linkage">linkage type</a>, an optional
658 <a href="#visibility">visibility style</a>, an optional
659 <a href="#callingconv">calling convention</a>, a return type, an optional
660 <a href="#paramattrs">parameter attribute</a> for the return type, a function
661 name, a (possibly empty) argument list (each with optional
662 <a href="#paramattrs">parameter attributes</a>), an optional section, an
663 optional alignment, an opening curly brace, a list of basic blocks, and a
666 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
667 optional <a href="#linkage">linkage type</a>, an optional
668 <a href="#visibility">visibility style</a>, an optional
669 <a href="#callingconv">calling convention</a>, a return type, an optional
670 <a href="#paramattrs">parameter attribute</a> for the return type, a function
671 name, a possibly empty list of arguments, and an optional alignment.</p>
673 <p>A function definition contains a list of basic blocks, forming the CFG for
674 the function. Each basic block may optionally start with a label (giving the
675 basic block a symbol table entry), contains a list of instructions, and ends
676 with a <a href="#terminators">terminator</a> instruction (such as a branch or
677 function return).</p>
679 <p>The first basic block in a program is special in two ways: it is immediately
680 executed on entrance to the function, and it is not allowed to have predecessor
681 basic blocks (i.e. there can not be any branches to the entry block of a
682 function). Because the block can have no predecessors, it also cannot have any
683 <a href="#i_phi">PHI nodes</a>.</p>
685 <p>LLVM functions are identified by their name and type signature. Hence, two
686 functions with the same name but different parameter lists or return values are
687 considered different functions, and LLVM will resolve references to each
690 <p>LLVM allows an explicit section to be specified for functions. If the target
691 supports it, it will emit functions to the section specified.</p>
693 <p>An explicit alignment may be specified for a function. If not present, or if
694 the alignment is set to zero, the alignment of the function is set by the target
695 to whatever it feels convenient. If an explicit alignment is specified, the
696 function is forced to have at least that much alignment. All alignments must be
702 <!-- ======================================================================= -->
703 <div class="doc_subsection">
704 <a name="aliasstructure">Aliases</a>
706 <div class="doc_text">
707 <p>Aliases act as "second name" for the aliasee value (which can be either
708 function or global variable). Aliases may have an
709 optional <a href="#linkage">linkage type</a>, and an
710 optional <a href="#visibility">visibility style</a>.</p>
715 @<Name> = [Linkage] [Visibility] alias <AliaseeTy> @<Aliasee>
722 <!-- ======================================================================= -->
723 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
724 <div class="doc_text">
725 <p>The return type and each parameter of a function type may have a set of
726 <i>parameter attributes</i> associated with them. Parameter attributes are
727 used to communicate additional information about the result or parameters of
728 a function. Parameter attributes are considered to be part of the function
729 type so two functions types that differ only by the parameter attributes
730 are different function types.</p>
732 <p>Parameter attributes are simple keywords that follow the type specified. If
733 multiple parameter attributes are needed, they are space separated. For
735 %someFunc = i16 (i8 sext %someParam) zext
736 %someFunc = i16 (i8 zext %someParam) zext</pre>
737 <p>Note that the two function types above are unique because the parameter has
738 a different attribute (sext in the first one, zext in the second). Also note
739 that the attribute for the function result (zext) comes immediately after the
742 <p>Currently, only the following parameter attributes are defined:</p>
744 <dt><tt>zext</tt></dt>
745 <dd>This indicates that the parameter should be zero extended just before
746 a call to this function.</dd>
747 <dt><tt>sext</tt></dt>
748 <dd>This indicates that the parameter should be sign extended just before
749 a call to this function.</dd>
750 <dt><tt>inreg</tt></dt>
751 <dd>This indicates that the parameter should be placed in register (if
752 possible) during assembling function call. Support for this attribute is
754 <dt><tt>sret</tt></dt>
755 <dd>This indicates that the parameter specifies the address of a structure
756 that is the return value of the function in the source program.</dd>
757 <dt><tt>noreturn</tt></dt>
758 <dd>This function attribute indicates that the function never returns. This
759 indicates to LLVM that every call to this function should be treated as if
760 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
761 <dt><tt>nounwind</tt></dt>
762 <dd>This function attribute indicates that the function type does not use
763 the unwind instruction and does not allow stack unwinding to propagate
769 <!-- ======================================================================= -->
770 <div class="doc_subsection">
771 <a name="moduleasm">Module-Level Inline Assembly</a>
774 <div class="doc_text">
776 Modules may contain "module-level inline asm" blocks, which corresponds to the
777 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
778 LLVM and treated as a single unit, but may be separated in the .ll file if
779 desired. The syntax is very simple:
782 <div class="doc_code"><pre>
783 module asm "inline asm code goes here"
784 module asm "more can go here"
787 <p>The strings can contain any character by escaping non-printable characters.
788 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
793 The inline asm code is simply printed to the machine code .s file when
794 assembly code is generated.
798 <!-- ======================================================================= -->
799 <div class="doc_subsection">
800 <a name="datalayout">Data Layout</a>
803 <div class="doc_text">
804 <p>A module may specify a target specific data layout string that specifies how
805 data is to be laid out in memory. The syntax for the data layout is simply:</p>
806 <pre> target datalayout = "<i>layout specification</i>"</pre>
807 <p>The <i>layout specification</i> consists of a list of specifications
808 separated by the minus sign character ('-'). Each specification starts with a
809 letter and may include other information after the letter to define some
810 aspect of the data layout. The specifications accepted are as follows: </p>
813 <dd>Specifies that the target lays out data in big-endian form. That is, the
814 bits with the most significance have the lowest address location.</dd>
816 <dd>Specifies that hte target lays out data in little-endian form. That is,
817 the bits with the least significance have the lowest address location.</dd>
818 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
819 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
820 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
821 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
823 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
824 <dd>This specifies the alignment for an integer type of a given bit
825 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
826 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
827 <dd>This specifies the alignment for a vector type of a given bit
829 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
830 <dd>This specifies the alignment for a floating point type of a given bit
831 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
833 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
834 <dd>This specifies the alignment for an aggregate type of a given bit
837 <p>When constructing the data layout for a given target, LLVM starts with a
838 default set of specifications which are then (possibly) overriden by the
839 specifications in the <tt>datalayout</tt> keyword. The default specifications
840 are given in this list:</p>
842 <li><tt>E</tt> - big endian</li>
843 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
844 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
845 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
846 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
847 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
848 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
849 alignment of 64-bits</li>
850 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
851 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
852 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
853 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
854 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
856 <p>When llvm is determining the alignment for a given type, it uses the
859 <li>If the type sought is an exact match for one of the specifications, that
860 specification is used.</li>
861 <li>If no match is found, and the type sought is an integer type, then the
862 smallest integer type that is larger than the bitwidth of the sought type is
863 used. If none of the specifications are larger than the bitwidth then the the
864 largest integer type is used. For example, given the default specifications
865 above, the i7 type will use the alignment of i8 (next largest) while both
866 i65 and i256 will use the alignment of i64 (largest specified).</li>
867 <li>If no match is found, and the type sought is a vector type, then the
868 largest vector type that is smaller than the sought vector type will be used
869 as a fall back. This happens because <128 x double> can be implemented in
870 terms of 64 <2 x double>, for example.</li>
874 <!-- *********************************************************************** -->
875 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
876 <!-- *********************************************************************** -->
878 <div class="doc_text">
880 <p>The LLVM type system is one of the most important features of the
881 intermediate representation. Being typed enables a number of
882 optimizations to be performed on the IR directly, without having to do
883 extra analyses on the side before the transformation. A strong type
884 system makes it easier to read the generated code and enables novel
885 analyses and transformations that are not feasible to perform on normal
886 three address code representations.</p>
890 <!-- ======================================================================= -->
891 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
892 <div class="doc_text">
893 <p>The primitive types are the fundamental building blocks of the LLVM
894 system. The current set of primitive types is as follows:</p>
896 <table class="layout">
901 <tr><th>Type</th><th>Description</th></tr>
902 <tr><td><tt><a name="t_void">void</a></tt></td><td>No value</td></tr>
903 <tr><td><tt>i8</tt></td><td>8-bit value</td></tr>
904 <tr><td><tt>i32</tt></td><td>32-bit value</td></tr>
905 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
906 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
913 <tr><th>Type</th><th>Description</th></tr>
914 <tr><td><tt>i1</tt></td><td>True or False value</td></tr>
915 <tr><td><tt>i16</tt></td><td>16-bit value</td></tr>
916 <tr><td><tt>i64</tt></td><td>64-bit value</td></tr>
917 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
925 <!-- _______________________________________________________________________ -->
926 <div class="doc_subsubsection"> <a name="t_classifications">Type
927 Classifications</a> </div>
928 <div class="doc_text">
929 <p>These different primitive types fall into a few useful
932 <table border="1" cellspacing="0" cellpadding="4">
934 <tr><th>Classification</th><th>Types</th></tr>
936 <td><a name="t_integer">integer</a></td>
937 <td><tt>i1, i8, i16, i32, i64</tt></td>
940 <td><a name="t_floating">floating point</a></td>
941 <td><tt>float, double</tt></td>
944 <td><a name="t_firstclass">first class</a></td>
945 <td><tt>i1, i8, i16, i32, i64, float, double, <br/>
946 <a href="#t_pointer">pointer</a>,<a href="#t_vector">vector</a></tt>
952 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
953 most important. Values of these types are the only ones which can be
954 produced by instructions, passed as arguments, or used as operands to
955 instructions. This means that all structures and arrays must be
956 manipulated either by pointer or by component.</p>
959 <!-- ======================================================================= -->
960 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
962 <div class="doc_text">
964 <p>The real power in LLVM comes from the derived types in the system.
965 This is what allows a programmer to represent arrays, functions,
966 pointers, and other useful types. Note that these derived types may be
967 recursive: For example, it is possible to have a two dimensional array.</p>
971 <!-- _______________________________________________________________________ -->
972 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
974 <div class="doc_text">
978 <p>The array type is a very simple derived type that arranges elements
979 sequentially in memory. The array type requires a size (number of
980 elements) and an underlying data type.</p>
985 [<# elements> x <elementtype>]
988 <p>The number of elements is a constant integer value; elementtype may
989 be any type with a size.</p>
992 <table class="layout">
995 <tt>[40 x i32 ]</tt><br/>
996 <tt>[41 x i32 ]</tt><br/>
997 <tt>[40 x i8]</tt><br/>
1000 Array of 40 32-bit integer values.<br/>
1001 Array of 41 32-bit integer values.<br/>
1002 Array of 40 8-bit integer values.<br/>
1006 <p>Here are some examples of multidimensional arrays:</p>
1007 <table class="layout">
1010 <tt>[3 x [4 x i32]]</tt><br/>
1011 <tt>[12 x [10 x float]]</tt><br/>
1012 <tt>[2 x [3 x [4 x i16]]]</tt><br/>
1015 3x4 array of 32-bit integer values.<br/>
1016 12x10 array of single precision floating point values.<br/>
1017 2x3x4 array of 16-bit integer values.<br/>
1022 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1023 length array. Normally, accesses past the end of an array are undefined in
1024 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1025 As a special case, however, zero length arrays are recognized to be variable
1026 length. This allows implementation of 'pascal style arrays' with the LLVM
1027 type "{ i32, [0 x float]}", for example.</p>
1031 <!-- _______________________________________________________________________ -->
1032 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1033 <div class="doc_text">
1035 <p>The function type can be thought of as a function signature. It
1036 consists of a return type and a list of formal parameter types.
1037 Function types are usually used to build virtual function tables
1038 (which are structures of pointers to functions), for indirect function
1039 calls, and when defining a function.</p>
1041 The return type of a function type cannot be an aggregate type.
1044 <pre> <returntype> (<parameter list>)<br></pre>
1045 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1046 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1047 which indicates that the function takes a variable number of arguments.
1048 Variable argument functions can access their arguments with the <a
1049 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1051 <table class="layout">
1053 <td class="left"><tt>i32 (i32)</tt></td>
1054 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1056 </tr><tr class="layout">
1057 <td class="left"><tt>float (i16 sext, i32 *) *
1059 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1060 an <tt>i16</tt> that should be sign extended and a
1061 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1064 </tr><tr class="layout">
1065 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1066 <td class="left">A vararg function that takes at least one
1067 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1068 which returns an integer. This is the signature for <tt>printf</tt> in
1075 <!-- _______________________________________________________________________ -->
1076 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1077 <div class="doc_text">
1079 <p>The structure type is used to represent a collection of data members
1080 together in memory. The packing of the field types is defined to match
1081 the ABI of the underlying processor. The elements of a structure may
1082 be any type that has a size.</p>
1083 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1084 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1085 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1088 <pre> { <type list> }<br></pre>
1090 <table class="layout">
1092 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1093 <td class="left">A triple of three <tt>i32</tt> values</td>
1094 </tr><tr class="layout">
1095 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1096 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1097 second element is a <a href="#t_pointer">pointer</a> to a
1098 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1099 an <tt>i32</tt>.</td>
1104 <!-- _______________________________________________________________________ -->
1105 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1107 <div class="doc_text">
1109 <p>The packed structure type is used to represent a collection of data members
1110 together in memory. There is no padding between fields. Further, the alignment
1111 of a packed structure is 1 byte. The elements of a packed structure may
1112 be any type that has a size.</p>
1113 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1114 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1115 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1118 <pre> < { <type list> } > <br></pre>
1120 <table class="layout">
1122 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1123 <td class="left">A triple of three <tt>i32</tt> values</td>
1124 </tr><tr class="layout">
1125 <td class="left"><tt>< { float, i32 (i32) * } ></tt></td>
1126 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1127 second element is a <a href="#t_pointer">pointer</a> to a
1128 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1129 an <tt>i32</tt>.</td>
1134 <!-- _______________________________________________________________________ -->
1135 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1136 <div class="doc_text">
1138 <p>As in many languages, the pointer type represents a pointer or
1139 reference to another object, which must live in memory.</p>
1141 <pre> <type> *<br></pre>
1143 <table class="layout">
1146 <tt>[4x i32]*</tt><br/>
1147 <tt>i32 (i32 *) *</tt><br/>
1150 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
1151 four <tt>i32</tt> values<br/>
1152 A <a href="#t_pointer">pointer</a> to a <a
1153 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1160 <!-- _______________________________________________________________________ -->
1161 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1162 <div class="doc_text">
1166 <p>A vector type is a simple derived type that represents a vector
1167 of elements. Vector types are used when multiple primitive data
1168 are operated in parallel using a single instruction (SIMD).
1169 A vector type requires a size (number of
1170 elements) and an underlying primitive data type. Vectors must have a power
1171 of two length (1, 2, 4, 8, 16 ...). Vector types are
1172 considered <a href="#t_firstclass">first class</a>.</p>
1177 < <# elements> x <elementtype> >
1180 <p>The number of elements is a constant integer value; elementtype may
1181 be any integer or floating point type.</p>
1185 <table class="layout">
1188 <tt><4 x i32></tt><br/>
1189 <tt><8 x float></tt><br/>
1190 <tt><2 x i64></tt><br/>
1193 Vector of 4 32-bit integer values.<br/>
1194 Vector of 8 floating-point values.<br/>
1195 Vector of 2 64-bit integer values.<br/>
1201 <!-- _______________________________________________________________________ -->
1202 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1203 <div class="doc_text">
1207 <p>Opaque types are used to represent unknown types in the system. This
1208 corresponds (for example) to the C notion of a foward declared structure type.
1209 In LLVM, opaque types can eventually be resolved to any type (not just a
1210 structure type).</p>
1220 <table class="layout">
1226 An opaque type.<br/>
1233 <!-- *********************************************************************** -->
1234 <div class="doc_section"> <a name="constants">Constants</a> </div>
1235 <!-- *********************************************************************** -->
1237 <div class="doc_text">
1239 <p>LLVM has several different basic types of constants. This section describes
1240 them all and their syntax.</p>
1244 <!-- ======================================================================= -->
1245 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1247 <div class="doc_text">
1250 <dt><b>Boolean constants</b></dt>
1252 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1253 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1256 <dt><b>Integer constants</b></dt>
1258 <dd>Standard integers (such as '4') are constants of the <a
1259 href="#t_integer">integer</a> type. Negative numbers may be used with
1263 <dt><b>Floating point constants</b></dt>
1265 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1266 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1267 notation (see below). Floating point constants must have a <a
1268 href="#t_floating">floating point</a> type. </dd>
1270 <dt><b>Null pointer constants</b></dt>
1272 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1273 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1277 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1278 of floating point constants. For example, the form '<tt>double
1279 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1280 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1281 (and the only time that they are generated by the disassembler) is when a
1282 floating point constant must be emitted but it cannot be represented as a
1283 decimal floating point number. For example, NaN's, infinities, and other
1284 special values are represented in their IEEE hexadecimal format so that
1285 assembly and disassembly do not cause any bits to change in the constants.</p>
1289 <!-- ======================================================================= -->
1290 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1293 <div class="doc_text">
1294 <p>Aggregate constants arise from aggregation of simple constants
1295 and smaller aggregate constants.</p>
1298 <dt><b>Structure constants</b></dt>
1300 <dd>Structure constants are represented with notation similar to structure
1301 type definitions (a comma separated list of elements, surrounded by braces
1302 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* %G }</tt>",
1303 where "<tt>%G</tt>" is declared as "<tt>%G = external global i32</tt>". Structure constants
1304 must have <a href="#t_struct">structure type</a>, and the number and
1305 types of elements must match those specified by the type.
1308 <dt><b>Array constants</b></dt>
1310 <dd>Array constants are represented with notation similar to array type
1311 definitions (a comma separated list of elements, surrounded by square brackets
1312 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1313 constants must have <a href="#t_array">array type</a>, and the number and
1314 types of elements must match those specified by the type.
1317 <dt><b>Vector constants</b></dt>
1319 <dd>Vector constants are represented with notation similar to vector type
1320 definitions (a comma separated list of elements, surrounded by
1321 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1322 i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
1323 href="#t_vector">vector type</a>, and the number and types of elements must
1324 match those specified by the type.
1327 <dt><b>Zero initialization</b></dt>
1329 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1330 value to zero of <em>any</em> type, including scalar and aggregate types.
1331 This is often used to avoid having to print large zero initializers (e.g. for
1332 large arrays) and is always exactly equivalent to using explicit zero
1339 <!-- ======================================================================= -->
1340 <div class="doc_subsection">
1341 <a name="globalconstants">Global Variable and Function Addresses</a>
1344 <div class="doc_text">
1346 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1347 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1348 constants. These constants are explicitly referenced when the <a
1349 href="#identifiers">identifier for the global</a> is used and always have <a
1350 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1356 %Z = global [2 x i32*] [ i32* %X, i32* %Y ]
1361 <!-- ======================================================================= -->
1362 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1363 <div class="doc_text">
1364 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1365 no specific value. Undefined values may be of any type and be used anywhere
1366 a constant is permitted.</p>
1368 <p>Undefined values indicate to the compiler that the program is well defined
1369 no matter what value is used, giving the compiler more freedom to optimize.
1373 <!-- ======================================================================= -->
1374 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1377 <div class="doc_text">
1379 <p>Constant expressions are used to allow expressions involving other constants
1380 to be used as constants. Constant expressions may be of any <a
1381 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1382 that does not have side effects (e.g. load and call are not supported). The
1383 following is the syntax for constant expressions:</p>
1386 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1387 <dd>Truncate a constant to another type. The bit size of CST must be larger
1388 than the bit size of TYPE. Both types must be integers.</dd>
1390 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1391 <dd>Zero extend a constant to another type. The bit size of CST must be
1392 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1394 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1395 <dd>Sign extend a constant to another type. The bit size of CST must be
1396 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1398 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1399 <dd>Truncate a floating point constant to another floating point type. The
1400 size of CST must be larger than the size of TYPE. Both types must be
1401 floating point.</dd>
1403 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1404 <dd>Floating point extend a constant to another type. The size of CST must be
1405 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1407 <dt><b><tt>fp2uint ( CST to TYPE )</tt></b></dt>
1408 <dd>Convert a floating point constant to the corresponding unsigned integer
1409 constant. TYPE must be an integer type. CST must be floating point. If the
1410 value won't fit in the integer type, the results are undefined.</dd>
1412 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1413 <dd>Convert a floating point constant to the corresponding signed integer
1414 constant. TYPE must be an integer type. CST must be floating point. If the
1415 value won't fit in the integer type, the results are undefined.</dd>
1417 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1418 <dd>Convert an unsigned integer constant to the corresponding floating point
1419 constant. TYPE must be floating point. CST must be of integer type. If the
1420 value won't fit in the floating point type, the results are undefined.</dd>
1422 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1423 <dd>Convert a signed integer constant to the corresponding floating point
1424 constant. TYPE must be floating point. CST must be of integer type. If the
1425 value won't fit in the floating point type, the results are undefined.</dd>
1427 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1428 <dd>Convert a pointer typed constant to the corresponding integer constant
1429 TYPE must be an integer type. CST must be of pointer type. The CST value is
1430 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1432 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1433 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1434 pointer type. CST must be of integer type. The CST value is zero extended,
1435 truncated, or unchanged to make it fit in a pointer size. This one is
1436 <i>really</i> dangerous!</dd>
1438 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1439 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1440 identical (same number of bits). The conversion is done as if the CST value
1441 was stored to memory and read back as TYPE. In other words, no bits change
1442 with this operator, just the type. This can be used for conversion of
1443 vector types to any other type, as long as they have the same bit width. For
1444 pointers it is only valid to cast to another pointer type.
1447 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1449 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1450 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1451 instruction, the index list may have zero or more indexes, which are required
1452 to make sense for the type of "CSTPTR".</dd>
1454 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1456 <dd>Perform the <a href="#i_select">select operation</a> on
1459 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1460 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1462 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1463 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1465 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1467 <dd>Perform the <a href="#i_extractelement">extractelement
1468 operation</a> on constants.
1470 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1472 <dd>Perform the <a href="#i_insertelement">insertelement
1473 operation</a> on constants.</dd>
1476 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1478 <dd>Perform the <a href="#i_shufflevector">shufflevector
1479 operation</a> on constants.</dd>
1481 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1483 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1484 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1485 binary</a> operations. The constraints on operands are the same as those for
1486 the corresponding instruction (e.g. no bitwise operations on floating point
1487 values are allowed).</dd>
1491 <!-- *********************************************************************** -->
1492 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1493 <!-- *********************************************************************** -->
1495 <!-- ======================================================================= -->
1496 <div class="doc_subsection">
1497 <a name="inlineasm">Inline Assembler Expressions</a>
1500 <div class="doc_text">
1503 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1504 Module-Level Inline Assembly</a>) through the use of a special value. This
1505 value represents the inline assembler as a string (containing the instructions
1506 to emit), a list of operand constraints (stored as a string), and a flag that
1507 indicates whether or not the inline asm expression has side effects. An example
1508 inline assembler expression is:
1512 i32 (i32) asm "bswap $0", "=r,r"
1516 Inline assembler expressions may <b>only</b> be used as the callee operand of
1517 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1521 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1525 Inline asms with side effects not visible in the constraint list must be marked
1526 as having side effects. This is done through the use of the
1527 '<tt>sideeffect</tt>' keyword, like so:
1531 call void asm sideeffect "eieio", ""()
1534 <p>TODO: The format of the asm and constraints string still need to be
1535 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1536 need to be documented).
1541 <!-- *********************************************************************** -->
1542 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1543 <!-- *********************************************************************** -->
1545 <div class="doc_text">
1547 <p>The LLVM instruction set consists of several different
1548 classifications of instructions: <a href="#terminators">terminator
1549 instructions</a>, <a href="#binaryops">binary instructions</a>,
1550 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1551 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1552 instructions</a>.</p>
1556 <!-- ======================================================================= -->
1557 <div class="doc_subsection"> <a name="terminators">Terminator
1558 Instructions</a> </div>
1560 <div class="doc_text">
1562 <p>As mentioned <a href="#functionstructure">previously</a>, every
1563 basic block in a program ends with a "Terminator" instruction, which
1564 indicates which block should be executed after the current block is
1565 finished. These terminator instructions typically yield a '<tt>void</tt>'
1566 value: they produce control flow, not values (the one exception being
1567 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1568 <p>There are six different terminator instructions: the '<a
1569 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1570 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1571 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1572 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1573 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1577 <!-- _______________________________________________________________________ -->
1578 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1579 Instruction</a> </div>
1580 <div class="doc_text">
1582 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1583 ret void <i>; Return from void function</i>
1586 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1587 value) from a function back to the caller.</p>
1588 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1589 returns a value and then causes control flow, and one that just causes
1590 control flow to occur.</p>
1592 <p>The '<tt>ret</tt>' instruction may return any '<a
1593 href="#t_firstclass">first class</a>' type. Notice that a function is
1594 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1595 instruction inside of the function that returns a value that does not
1596 match the return type of the function.</p>
1598 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1599 returns back to the calling function's context. If the caller is a "<a
1600 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1601 the instruction after the call. If the caller was an "<a
1602 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1603 at the beginning of the "normal" destination block. If the instruction
1604 returns a value, that value shall set the call or invoke instruction's
1607 <pre> ret i32 5 <i>; Return an integer value of 5</i>
1608 ret void <i>; Return from a void function</i>
1611 <!-- _______________________________________________________________________ -->
1612 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1613 <div class="doc_text">
1615 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1618 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1619 transfer to a different basic block in the current function. There are
1620 two forms of this instruction, corresponding to a conditional branch
1621 and an unconditional branch.</p>
1623 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1624 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1625 unconditional form of the '<tt>br</tt>' instruction takes a single
1626 '<tt>label</tt>' value as a target.</p>
1628 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1629 argument is evaluated. If the value is <tt>true</tt>, control flows
1630 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1631 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1633 <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
1634 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1636 <!-- _______________________________________________________________________ -->
1637 <div class="doc_subsubsection">
1638 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1641 <div class="doc_text">
1645 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1650 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1651 several different places. It is a generalization of the '<tt>br</tt>'
1652 instruction, allowing a branch to occur to one of many possible
1658 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1659 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1660 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1661 table is not allowed to contain duplicate constant entries.</p>
1665 <p>The <tt>switch</tt> instruction specifies a table of values and
1666 destinations. When the '<tt>switch</tt>' instruction is executed, this
1667 table is searched for the given value. If the value is found, control flow is
1668 transfered to the corresponding destination; otherwise, control flow is
1669 transfered to the default destination.</p>
1671 <h5>Implementation:</h5>
1673 <p>Depending on properties of the target machine and the particular
1674 <tt>switch</tt> instruction, this instruction may be code generated in different
1675 ways. For example, it could be generated as a series of chained conditional
1676 branches or with a lookup table.</p>
1681 <i>; Emulate a conditional br instruction</i>
1682 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1683 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1685 <i>; Emulate an unconditional br instruction</i>
1686 switch i32 0, label %dest [ ]
1688 <i>; Implement a jump table:</i>
1689 switch i32 %val, label %otherwise [ i32 0, label %onzero
1691 i32 2, label %ontwo ]
1695 <!-- _______________________________________________________________________ -->
1696 <div class="doc_subsubsection">
1697 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1700 <div class="doc_text">
1705 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1706 to label <normal label> unwind label <exception label>
1711 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1712 function, with the possibility of control flow transfer to either the
1713 '<tt>normal</tt>' label or the
1714 '<tt>exception</tt>' label. If the callee function returns with the
1715 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1716 "normal" label. If the callee (or any indirect callees) returns with the "<a
1717 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1718 continued at the dynamically nearest "exception" label.</p>
1722 <p>This instruction requires several arguments:</p>
1726 The optional "cconv" marker indicates which <a href="#callingconv">calling
1727 convention</a> the call should use. If none is specified, the call defaults
1728 to using C calling conventions.
1730 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1731 function value being invoked. In most cases, this is a direct function
1732 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1733 an arbitrary pointer to function value.
1736 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1737 function to be invoked. </li>
1739 <li>'<tt>function args</tt>': argument list whose types match the function
1740 signature argument types. If the function signature indicates the function
1741 accepts a variable number of arguments, the extra arguments can be
1744 <li>'<tt>normal label</tt>': the label reached when the called function
1745 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1747 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1748 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1754 <p>This instruction is designed to operate as a standard '<tt><a
1755 href="#i_call">call</a></tt>' instruction in most regards. The primary
1756 difference is that it establishes an association with a label, which is used by
1757 the runtime library to unwind the stack.</p>
1759 <p>This instruction is used in languages with destructors to ensure that proper
1760 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1761 exception. Additionally, this is important for implementation of
1762 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1766 %retval = invoke i32 %Test(i32 15) to label %Continue
1767 unwind label %TestCleanup <i>; {i32}:retval set</i>
1768 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1769 unwind label %TestCleanup <i>; {i32}:retval set</i>
1774 <!-- _______________________________________________________________________ -->
1776 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1777 Instruction</a> </div>
1779 <div class="doc_text">
1788 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1789 at the first callee in the dynamic call stack which used an <a
1790 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1791 primarily used to implement exception handling.</p>
1795 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1796 immediately halt. The dynamic call stack is then searched for the first <a
1797 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1798 execution continues at the "exceptional" destination block specified by the
1799 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1800 dynamic call chain, undefined behavior results.</p>
1803 <!-- _______________________________________________________________________ -->
1805 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1806 Instruction</a> </div>
1808 <div class="doc_text">
1817 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1818 instruction is used to inform the optimizer that a particular portion of the
1819 code is not reachable. This can be used to indicate that the code after a
1820 no-return function cannot be reached, and other facts.</p>
1824 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1829 <!-- ======================================================================= -->
1830 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1831 <div class="doc_text">
1832 <p>Binary operators are used to do most of the computation in a
1833 program. They require two operands, execute an operation on them, and
1834 produce a single value. The operands might represent
1835 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
1836 The result value of a binary operator is not
1837 necessarily the same type as its operands.</p>
1838 <p>There are several different binary operators:</p>
1840 <!-- _______________________________________________________________________ -->
1841 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1842 Instruction</a> </div>
1843 <div class="doc_text">
1845 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1848 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1850 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1851 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1852 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1853 Both arguments must have identical types.</p>
1855 <p>The value produced is the integer or floating point sum of the two
1858 <pre> <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
1861 <!-- _______________________________________________________________________ -->
1862 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1863 Instruction</a> </div>
1864 <div class="doc_text">
1866 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1869 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1871 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1872 instruction present in most other intermediate representations.</p>
1874 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1875 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1877 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1878 Both arguments must have identical types.</p>
1880 <p>The value produced is the integer or floating point difference of
1881 the two operands.</p>
1883 <pre> <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
1884 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
1887 <!-- _______________________________________________________________________ -->
1888 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1889 Instruction</a> </div>
1890 <div class="doc_text">
1892 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1895 <p>The '<tt>mul</tt>' instruction returns the product of its two
1898 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1899 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1901 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1902 Both arguments must have identical types.</p>
1904 <p>The value produced is the integer or floating point product of the
1906 <p>Because the operands are the same width, the result of an integer
1907 multiplication is the same whether the operands should be deemed unsigned or
1910 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
1913 <!-- _______________________________________________________________________ -->
1914 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
1916 <div class="doc_text">
1918 <pre> <result> = udiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1921 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
1924 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
1925 <a href="#t_integer">integer</a> values. Both arguments must have identical
1926 types. This instruction can also take <a href="#t_vector">vector</a> versions
1927 of the values in which case the elements must be integers.</p>
1929 <p>The value produced is the unsigned integer quotient of the two operands. This
1930 instruction always performs an unsigned division operation, regardless of
1931 whether the arguments are unsigned or not.</p>
1933 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
1936 <!-- _______________________________________________________________________ -->
1937 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
1939 <div class="doc_text">
1941 <pre> <result> = sdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1944 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
1947 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
1948 <a href="#t_integer">integer</a> values. Both arguments must have identical
1949 types. This instruction can also take <a href="#t_vector">vector</a> versions
1950 of the values in which case the elements must be integers.</p>
1952 <p>The value produced is the signed integer quotient of the two operands. This
1953 instruction always performs a signed division operation, regardless of whether
1954 the arguments are signed or not.</p>
1956 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
1959 <!-- _______________________________________________________________________ -->
1960 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
1961 Instruction</a> </div>
1962 <div class="doc_text">
1964 <pre> <result> = fdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1967 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
1970 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
1971 <a href="#t_floating">floating point</a> values. Both arguments must have
1972 identical types. This instruction can also take <a href="#t_vector">vector</a>
1973 versions of floating point values.</p>
1975 <p>The value produced is the floating point quotient of the two operands.</p>
1977 <pre> <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
1980 <!-- _______________________________________________________________________ -->
1981 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
1983 <div class="doc_text">
1985 <pre> <result> = urem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1988 <p>The '<tt>urem</tt>' instruction returns the remainder from the
1989 unsigned division of its two arguments.</p>
1991 <p>The two arguments to the '<tt>urem</tt>' instruction must be
1992 <a href="#t_integer">integer</a> values. Both arguments must have identical
1995 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
1996 This instruction always performs an unsigned division to get the remainder,
1997 regardless of whether the arguments are unsigned or not.</p>
1999 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2003 <!-- _______________________________________________________________________ -->
2004 <div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2005 Instruction</a> </div>
2006 <div class="doc_text">
2008 <pre> <result> = srem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2011 <p>The '<tt>srem</tt>' instruction returns the remainder from the
2012 signed division of its two operands.</p>
2014 <p>The two arguments to the '<tt>srem</tt>' instruction must be
2015 <a href="#t_integer">integer</a> values. Both arguments must have identical
2018 <p>This instruction returns the <i>remainder</i> of a division (where the result
2019 has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2020 operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2021 a value. For more information about the difference, see <a
2022 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2023 Math Forum</a>. For a table of how this is implemented in various languages,
2024 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2025 Wikipedia: modulo operation</a>.</p>
2027 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2031 <!-- _______________________________________________________________________ -->
2032 <div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2033 Instruction</a> </div>
2034 <div class="doc_text">
2036 <pre> <result> = frem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2039 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2040 division of its two operands.</p>
2042 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2043 <a href="#t_floating">floating point</a> values. Both arguments must have
2044 identical types.</p>
2046 <p>This instruction returns the <i>remainder</i> of a division.</p>
2048 <pre> <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2052 <!-- ======================================================================= -->
2053 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2054 Operations</a> </div>
2055 <div class="doc_text">
2056 <p>Bitwise binary operators are used to do various forms of
2057 bit-twiddling in a program. They are generally very efficient
2058 instructions and can commonly be strength reduced from other
2059 instructions. They require two operands, execute an operation on them,
2060 and produce a single value. The resulting value of the bitwise binary
2061 operators is always the same type as its first operand.</p>
2064 <!-- _______________________________________________________________________ -->
2065 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2066 Instruction</a> </div>
2067 <div class="doc_text">
2069 <pre> <result> = shl <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2072 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2073 the left a specified number of bits.</p>
2075 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2076 href="#t_integer">integer</a> type.</p>
2078 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
2079 <h5>Example:</h5><pre>
2080 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2081 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2082 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2085 <!-- _______________________________________________________________________ -->
2086 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2087 Instruction</a> </div>
2088 <div class="doc_text">
2090 <pre> <result> = lshr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2094 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2095 operand shifted to the right a specified number of bits with zero fill.</p>
2098 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2099 <a href="#t_integer">integer</a> type.</p>
2102 <p>This instruction always performs a logical shift right operation. The most
2103 significant bits of the result will be filled with zero bits after the
2108 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2109 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2110 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2111 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2115 <!-- _______________________________________________________________________ -->
2116 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2117 Instruction</a> </div>
2118 <div class="doc_text">
2121 <pre> <result> = ashr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2125 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2126 operand shifted to the right a specified number of bits with sign extension.</p>
2129 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2130 <a href="#t_integer">integer</a> type.</p>
2133 <p>This instruction always performs an arithmetic shift right operation,
2134 The most significant bits of the result will be filled with the sign bit
2135 of <tt>var1</tt>.</p>
2139 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2140 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2141 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2142 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
2146 <!-- _______________________________________________________________________ -->
2147 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2148 Instruction</a> </div>
2149 <div class="doc_text">
2151 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2154 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2155 its two operands.</p>
2157 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
2158 href="#t_integer">integer</a> values. Both arguments must have
2159 identical types.</p>
2161 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2163 <div style="align: center">
2164 <table border="1" cellspacing="0" cellpadding="4">
2195 <pre> <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
2196 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2197 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2200 <!-- _______________________________________________________________________ -->
2201 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2202 <div class="doc_text">
2204 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2207 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2208 or of its two operands.</p>
2210 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
2211 href="#t_integer">integer</a> values. Both arguments must have
2212 identical types.</p>
2214 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2216 <div style="align: center">
2217 <table border="1" cellspacing="0" cellpadding="4">
2248 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2249 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2250 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2253 <!-- _______________________________________________________________________ -->
2254 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2255 Instruction</a> </div>
2256 <div class="doc_text">
2258 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2261 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2262 or of its two operands. The <tt>xor</tt> is used to implement the
2263 "one's complement" operation, which is the "~" operator in C.</p>
2265 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2266 href="#t_integer">integer</a> values. Both arguments must have
2267 identical types.</p>
2269 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2271 <div style="align: center">
2272 <table border="1" cellspacing="0" cellpadding="4">
2304 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2305 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2306 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2307 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2311 <!-- ======================================================================= -->
2312 <div class="doc_subsection">
2313 <a name="vectorops">Vector Operations</a>
2316 <div class="doc_text">
2318 <p>LLVM supports several instructions to represent vector operations in a
2319 target-independent manner. These instructions cover the element-access and
2320 vector-specific operations needed to process vectors effectively. While LLVM
2321 does directly support these vector operations, many sophisticated algorithms
2322 will want to use target-specific intrinsics to take full advantage of a specific
2327 <!-- _______________________________________________________________________ -->
2328 <div class="doc_subsubsection">
2329 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2332 <div class="doc_text">
2337 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
2343 The '<tt>extractelement</tt>' instruction extracts a single scalar
2344 element from a vector at a specified index.
2351 The first operand of an '<tt>extractelement</tt>' instruction is a
2352 value of <a href="#t_vector">vector</a> type. The second operand is
2353 an index indicating the position from which to extract the element.
2354 The index may be a variable.</p>
2359 The result is a scalar of the same type as the element type of
2360 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2361 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2362 results are undefined.
2368 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
2373 <!-- _______________________________________________________________________ -->
2374 <div class="doc_subsubsection">
2375 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2378 <div class="doc_text">
2383 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
2389 The '<tt>insertelement</tt>' instruction inserts a scalar
2390 element into a vector at a specified index.
2397 The first operand of an '<tt>insertelement</tt>' instruction is a
2398 value of <a href="#t_vector">vector</a> type. The second operand is a
2399 scalar value whose type must equal the element type of the first
2400 operand. The third operand is an index indicating the position at
2401 which to insert the value. The index may be a variable.</p>
2406 The result is a vector of the same type as <tt>val</tt>. Its
2407 element values are those of <tt>val</tt> except at position
2408 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2409 exceeds the length of <tt>val</tt>, the results are undefined.
2415 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
2419 <!-- _______________________________________________________________________ -->
2420 <div class="doc_subsubsection">
2421 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2424 <div class="doc_text">
2429 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x i32> <mask> <i>; yields <n x <ty>></i>
2435 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2436 from two input vectors, returning a vector of the same type.
2442 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2443 with types that match each other and types that match the result of the
2444 instruction. The third argument is a shuffle mask, which has the same number
2445 of elements as the other vector type, but whose element type is always 'i32'.
2449 The shuffle mask operand is required to be a constant vector with either
2450 constant integer or undef values.
2456 The elements of the two input vectors are numbered from left to right across
2457 both of the vectors. The shuffle mask operand specifies, for each element of
2458 the result vector, which element of the two input registers the result element
2459 gets. The element selector may be undef (meaning "don't care") and the second
2460 operand may be undef if performing a shuffle from only one vector.
2466 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
2467 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
2468 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
2469 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
2474 <!-- ======================================================================= -->
2475 <div class="doc_subsection">
2476 <a name="memoryops">Memory Access and Addressing Operations</a>
2479 <div class="doc_text">
2481 <p>A key design point of an SSA-based representation is how it
2482 represents memory. In LLVM, no memory locations are in SSA form, which
2483 makes things very simple. This section describes how to read, write,
2484 allocate, and free memory in LLVM.</p>
2488 <!-- _______________________________________________________________________ -->
2489 <div class="doc_subsubsection">
2490 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2493 <div class="doc_text">
2498 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2503 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2504 heap and returns a pointer to it.</p>
2508 <p>The '<tt>malloc</tt>' instruction allocates
2509 <tt>sizeof(<type>)*NumElements</tt>
2510 bytes of memory from the operating system and returns a pointer of the
2511 appropriate type to the program. If "NumElements" is specified, it is the
2512 number of elements allocated. If an alignment is specified, the value result
2513 of the allocation is guaranteed to be aligned to at least that boundary. If
2514 not specified, or if zero, the target can choose to align the allocation on any
2515 convenient boundary.</p>
2517 <p>'<tt>type</tt>' must be a sized type.</p>
2521 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2522 a pointer is returned.</p>
2527 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2529 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2530 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2531 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2532 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2533 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2537 <!-- _______________________________________________________________________ -->
2538 <div class="doc_subsubsection">
2539 <a name="i_free">'<tt>free</tt>' Instruction</a>
2542 <div class="doc_text">
2547 free <type> <value> <i>; yields {void}</i>
2552 <p>The '<tt>free</tt>' instruction returns memory back to the unused
2553 memory heap to be reallocated in the future.</p>
2557 <p>'<tt>value</tt>' shall be a pointer value that points to a value
2558 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2563 <p>Access to the memory pointed to by the pointer is no longer defined
2564 after this instruction executes.</p>
2569 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2570 free [4 x i8]* %array
2574 <!-- _______________________________________________________________________ -->
2575 <div class="doc_subsubsection">
2576 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2579 <div class="doc_text">
2584 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2589 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
2590 currently executing function, to be automatically released when this function
2591 returns to its caller.</p>
2595 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2596 bytes of memory on the runtime stack, returning a pointer of the
2597 appropriate type to the program. If "NumElements" is specified, it is the
2598 number of elements allocated. If an alignment is specified, the value result
2599 of the allocation is guaranteed to be aligned to at least that boundary. If
2600 not specified, or if zero, the target can choose to align the allocation on any
2601 convenient boundary.</p>
2603 <p>'<tt>type</tt>' may be any sized type.</p>
2607 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2608 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2609 instruction is commonly used to represent automatic variables that must
2610 have an address available. When the function returns (either with the <tt><a
2611 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2612 instructions), the memory is reclaimed.</p>
2617 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2618 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2619 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2620 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2624 <!-- _______________________________________________________________________ -->
2625 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2626 Instruction</a> </div>
2627 <div class="doc_text">
2629 <pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
2631 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2633 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2634 address from which to load. The pointer must point to a <a
2635 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2636 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2637 the number or order of execution of this <tt>load</tt> with other
2638 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2641 <p>The location of memory pointed to is loaded.</p>
2643 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2645 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2646 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2649 <!-- _______________________________________________________________________ -->
2650 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2651 Instruction</a> </div>
2652 <div class="doc_text">
2654 <pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2655 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2658 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2660 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2661 to store and an address at which to store it. The type of the '<tt><pointer></tt>'
2662 operand must be a pointer to the type of the '<tt><value></tt>'
2663 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2664 optimizer is not allowed to modify the number or order of execution of
2665 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2666 href="#i_store">store</a></tt> instructions.</p>
2668 <p>The contents of memory are updated to contain '<tt><value></tt>'
2669 at the location specified by the '<tt><pointer></tt>' operand.</p>
2671 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2673 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2674 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2678 <!-- _______________________________________________________________________ -->
2679 <div class="doc_subsubsection">
2680 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2683 <div class="doc_text">
2686 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2692 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2693 subelement of an aggregate data structure.</p>
2697 <p>This instruction takes a list of integer operands that indicate what
2698 elements of the aggregate object to index to. The actual types of the arguments
2699 provided depend on the type of the first pointer argument. The
2700 '<tt>getelementptr</tt>' instruction is used to index down through the type
2701 levels of a structure or to a specific index in an array. When indexing into a
2702 structure, only <tt>i32</tt> integer constants are allowed. When indexing
2703 into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2704 be sign extended to 64-bit values.</p>
2706 <p>For example, let's consider a C code fragment and how it gets
2707 compiled to LLVM:</p>
2721 define i32 *foo(struct ST *s) {
2722 return &s[1].Z.B[5][13];
2726 <p>The LLVM code generated by the GCC frontend is:</p>
2729 %RT = type { i8 , [10 x [20 x i32]], i8 }
2730 %ST = type { i32, double, %RT }
2732 define i32* %foo(%ST* %s) {
2734 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
2741 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2742 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2743 and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
2744 <a href="#t_integer">integer</a> type but the value will always be sign extended
2745 to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
2746 <b>constants</b>.</p>
2748 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2749 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
2750 }</tt>' type, a structure. The second index indexes into the third element of
2751 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
2752 i8 }</tt>' type, another structure. The third index indexes into the second
2753 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
2754 array. The two dimensions of the array are subscripted into, yielding an
2755 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2756 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
2758 <p>Note that it is perfectly legal to index partially through a
2759 structure, returning a pointer to an inner element. Because of this,
2760 the LLVM code for the given testcase is equivalent to:</p>
2763 define i32* %foo(%ST* %s) {
2764 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
2765 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
2766 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
2767 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
2768 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
2773 <p>Note that it is undefined to access an array out of bounds: array and
2774 pointer indexes must always be within the defined bounds of the array type.
2775 The one exception for this rules is zero length arrays. These arrays are
2776 defined to be accessible as variable length arrays, which requires access
2777 beyond the zero'th element.</p>
2779 <p>The getelementptr instruction is often confusing. For some more insight
2780 into how it works, see <a href="GetElementPtr.html">the getelementptr
2786 <i>; yields [12 x i8]*:aptr</i>
2787 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
2791 <!-- ======================================================================= -->
2792 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2794 <div class="doc_text">
2795 <p>The instructions in this category are the conversion instructions (casting)
2796 which all take a single operand and a type. They perform various bit conversions
2800 <!-- _______________________________________________________________________ -->
2801 <div class="doc_subsubsection">
2802 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2804 <div class="doc_text">
2808 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
2813 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2818 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2819 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2820 and type of the result, which must be an <a href="#t_integer">integer</a>
2821 type. The bit size of <tt>value</tt> must be larger than the bit size of
2822 <tt>ty2</tt>. Equal sized types are not allowed.</p>
2826 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
2827 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
2828 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
2829 It will always truncate bits.</p>
2833 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
2834 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
2835 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
2839 <!-- _______________________________________________________________________ -->
2840 <div class="doc_subsubsection">
2841 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
2843 <div class="doc_text">
2847 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
2851 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
2856 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
2857 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
2858 also be of <a href="#t_integer">integer</a> type. The bit size of the
2859 <tt>value</tt> must be smaller than the bit size of the destination type,
2863 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
2864 bits until it reaches the size of the destination type, <tt>ty2</tt>. When the
2865 the operand and the type are the same size, no bit filling is done and the
2866 cast is considered a <i>no-op cast</i> because no bits change (only the type
2869 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
2873 %X = zext i32 257 to i64 <i>; yields i64:257</i>
2874 %Y = zext i1 true to i32 <i>; yields i32:1</i>
2878 <!-- _______________________________________________________________________ -->
2879 <div class="doc_subsubsection">
2880 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
2882 <div class="doc_text">
2886 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
2890 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
2894 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
2895 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
2896 also be of <a href="#t_integer">integer</a> type. The bit size of the
2897 <tt>value</tt> must be smaller than the bit size of the destination type,
2902 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
2903 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
2904 the type <tt>ty2</tt>. When the the operand and the type are the same size,
2905 no bit filling is done and the cast is considered a <i>no-op cast</i> because
2906 no bits change (only the type changes).</p>
2908 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
2912 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
2913 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
2917 <!-- _______________________________________________________________________ -->
2918 <div class="doc_subsubsection">
2919 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
2922 <div class="doc_text">
2927 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
2931 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
2936 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
2937 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
2938 cast it to. The size of <tt>value</tt> must be larger than the size of
2939 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
2940 <i>no-op cast</i>.</p>
2943 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
2944 <a href="#t_floating">floating point</a> type to a smaller
2945 <a href="#t_floating">floating point</a> type. If the value cannot fit within
2946 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
2950 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
2951 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
2955 <!-- _______________________________________________________________________ -->
2956 <div class="doc_subsubsection">
2957 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
2959 <div class="doc_text">
2963 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
2967 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
2968 floating point value.</p>
2971 <p>The '<tt>fpext</tt>' instruction takes a
2972 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
2973 and a <a href="#t_floating">floating point</a> type to cast it to. The source
2974 type must be smaller than the destination type.</p>
2977 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
2978 <a href="#t_floating">floating point</a> type to a larger
2979 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
2980 used to make a <i>no-op cast</i> because it always changes bits. Use
2981 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
2985 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
2986 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
2990 <!-- _______________________________________________________________________ -->
2991 <div class="doc_subsubsection">
2992 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
2994 <div class="doc_text">
2998 <result> = fp2uint <ty> <value> to <ty2> <i>; yields ty2</i>
3002 <p>The '<tt>fp2uint</tt>' converts a floating point <tt>value</tt> to its
3003 unsigned integer equivalent of type <tt>ty2</tt>.
3007 <p>The '<tt>fp2uint</tt>' instruction takes a value to cast, which must be a
3008 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
3009 must be an <a href="#t_integer">integer</a> type.</p>
3012 <p> The '<tt>fp2uint</tt>' instruction converts its
3013 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3014 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3015 the results are undefined.</p>
3017 <p>When converting to i1, the conversion is done as a comparison against
3018 zero. If the <tt>value</tt> was zero, the i1 result will be <tt>false</tt>.
3019 If the <tt>value</tt> was non-zero, the i1 result will be <tt>true</tt>.</p>
3023 %X = fp2uint double 123.0 to i32 <i>; yields i32:123</i>
3024 %Y = fp2uint float 1.0E+300 to i1 <i>; yields i1:true</i>
3025 %X = fp2uint float 1.04E+17 to i8 <i>; yields undefined:1</i>
3029 <!-- _______________________________________________________________________ -->
3030 <div class="doc_subsubsection">
3031 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3033 <div class="doc_text">
3037 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
3041 <p>The '<tt>fptosi</tt>' instruction converts
3042 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3047 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
3048 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
3049 must also be an <a href="#t_integer">integer</a> type.</p>
3052 <p>The '<tt>fptosi</tt>' instruction converts its
3053 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3054 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3055 the results are undefined.</p>
3057 <p>When converting to i1, the conversion is done as a comparison against
3058 zero. If the <tt>value</tt> was zero, the i1 result will be <tt>false</tt>.
3059 If the <tt>value</tt> was non-zero, the i1 result will be <tt>true</tt>.</p>
3063 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
3064 %Y = fptosi float 1.0E-247 to i1 <i>; yields i1:true</i>
3065 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3069 <!-- _______________________________________________________________________ -->
3070 <div class="doc_subsubsection">
3071 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3073 <div class="doc_text">
3077 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3081 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3082 integer and converts that value to the <tt>ty2</tt> type.</p>
3086 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be an
3087 <a href="#t_integer">integer</a> value, and a type to cast it to, which must
3088 be a <a href="#t_floating">floating point</a> type.</p>
3091 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3092 integer quantity and converts it to the corresponding floating point value. If
3093 the value cannot fit in the floating point value, the results are undefined.</p>
3098 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3099 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3103 <!-- _______________________________________________________________________ -->
3104 <div class="doc_subsubsection">
3105 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3107 <div class="doc_text">
3111 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3115 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3116 integer and converts that value to the <tt>ty2</tt> type.</p>
3119 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be an
3120 <a href="#t_integer">integer</a> value, and a type to cast it to, which must be
3121 a <a href="#t_floating">floating point</a> type.</p>
3124 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3125 integer quantity and converts it to the corresponding floating point value. If
3126 the value cannot fit in the floating point value, the results are undefined.</p>
3130 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3131 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3135 <!-- _______________________________________________________________________ -->
3136 <div class="doc_subsubsection">
3137 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3139 <div class="doc_text">
3143 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
3147 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3148 the integer type <tt>ty2</tt>.</p>
3151 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3152 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3153 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3156 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3157 <tt>ty2</tt> by interpreting the pointer value as an integer and either
3158 truncating or zero extending that value to the size of the integer type. If
3159 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3160 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3161 are the same size, then nothing is done (<i>no-op cast</i>).</p>
3165 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit</i>
3166 %Y = ptrtoint i32* %x to i64 <i>; yields zero extend on 32-bit</i>
3170 <!-- _______________________________________________________________________ -->
3171 <div class="doc_subsubsection">
3172 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3174 <div class="doc_text">
3178 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
3182 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3183 a pointer type, <tt>ty2</tt>.</p>
3186 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3187 value to cast, and a type to cast it to, which must be a
3188 <a href="#t_pointer">pointer</a> type.
3191 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3192 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
3193 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3194 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3195 the size of a pointer then a zero extension is done. If they are the same size,
3196 nothing is done (<i>no-op cast</i>).</p>
3200 %X = inttoptr i32 255 to i32* <i>; yields zero extend on 64-bit</i>
3201 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit </i>
3202 %Y = inttoptr i16 0 to i32* <i>; yields zero extend on 32-bit</i>
3206 <!-- _______________________________________________________________________ -->
3207 <div class="doc_subsubsection">
3208 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3210 <div class="doc_text">
3214 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3218 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3219 <tt>ty2</tt> without changing any bits.</p>
3222 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3223 a first class value, and a type to cast it to, which must also be a <a
3224 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3225 and the destination type, <tt>ty2</tt>, must be identical. If the source
3226 type is a pointer, the destination type must also be a pointer.</p>
3229 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3230 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3231 this conversion. The conversion is done as if the <tt>value</tt> had been
3232 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3233 converted to other pointer types with this instruction. To convert pointers to
3234 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3235 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3239 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3240 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3241 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3245 <!-- ======================================================================= -->
3246 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3247 <div class="doc_text">
3248 <p>The instructions in this category are the "miscellaneous"
3249 instructions, which defy better classification.</p>
3252 <!-- _______________________________________________________________________ -->
3253 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3255 <div class="doc_text">
3257 <pre> <result> = icmp <cond> <ty> <var1>, <var2>
3258 <i>; yields {i1}:result</i>
3261 <p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3262 of its two integer operands.</p>
3264 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3265 the condition code which indicates the kind of comparison to perform. It is not
3266 a value, just a keyword. The possibilities for the condition code are:
3268 <li><tt>eq</tt>: equal</li>
3269 <li><tt>ne</tt>: not equal </li>
3270 <li><tt>ugt</tt>: unsigned greater than</li>
3271 <li><tt>uge</tt>: unsigned greater or equal</li>
3272 <li><tt>ult</tt>: unsigned less than</li>
3273 <li><tt>ule</tt>: unsigned less or equal</li>
3274 <li><tt>sgt</tt>: signed greater than</li>
3275 <li><tt>sge</tt>: signed greater or equal</li>
3276 <li><tt>slt</tt>: signed less than</li>
3277 <li><tt>sle</tt>: signed less or equal</li>
3279 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3280 <a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3282 <p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3283 the condition code given as <tt>cond</tt>. The comparison performed always
3284 yields a <a href="#t_primitive">i1</a> result, as follows:
3286 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3287 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3289 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3290 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3291 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3292 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3293 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3294 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3295 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3296 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3297 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3298 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3299 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3300 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3301 <li><tt>sge</tt>: interprets the operands as signed values and yields
3302 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3303 <li><tt>slt</tt>: interprets the operands as signed values and yields
3304 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3305 <li><tt>sle</tt>: interprets the operands as signed values and yields
3306 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3308 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3309 values are treated as integers and then compared.</p>
3312 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
3313 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
3314 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
3315 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3316 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3317 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
3321 <!-- _______________________________________________________________________ -->
3322 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3324 <div class="doc_text">
3326 <pre> <result> = fcmp <cond> <ty> <var1>, <var2>
3327 <i>; yields {i1}:result</i>
3330 <p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3331 of its floating point operands.</p>
3333 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3334 the condition code which indicates the kind of comparison to perform. It is not
3335 a value, just a keyword. The possibilities for the condition code are:
3337 <li><tt>false</tt>: no comparison, always returns false</li>
3338 <li><tt>oeq</tt>: ordered and equal</li>
3339 <li><tt>ogt</tt>: ordered and greater than </li>
3340 <li><tt>oge</tt>: ordered and greater than or equal</li>
3341 <li><tt>olt</tt>: ordered and less than </li>
3342 <li><tt>ole</tt>: ordered and less than or equal</li>
3343 <li><tt>one</tt>: ordered and not equal</li>
3344 <li><tt>ord</tt>: ordered (no nans)</li>
3345 <li><tt>ueq</tt>: unordered or equal</li>
3346 <li><tt>ugt</tt>: unordered or greater than </li>
3347 <li><tt>uge</tt>: unordered or greater than or equal</li>
3348 <li><tt>ult</tt>: unordered or less than </li>
3349 <li><tt>ule</tt>: unordered or less than or equal</li>
3350 <li><tt>une</tt>: unordered or not equal</li>
3351 <li><tt>uno</tt>: unordered (either nans)</li>
3352 <li><tt>true</tt>: no comparison, always returns true</li>
3354 <p>In the preceding, <i>ordered</i> means that neither operand is a QNAN while
3355 <i>unordered</i> means that either operand may be a QNAN.</p>
3356 <p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3357 <a href="#t_floating">floating point</a> typed. They must have identical
3359 <p>In the foregoing, <i>ordered</i> means that neither operand is a QNAN and
3360 <i>unordered</i> means that either operand is a QNAN.</p>
3362 <p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3363 the condition code given as <tt>cond</tt>. The comparison performed always
3364 yields a <a href="#t_primitive">i1</a> result, as follows:
3366 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3367 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3368 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3369 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3370 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3371 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3372 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3373 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3374 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3375 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3376 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3377 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3378 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3379 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3380 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3381 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3382 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3383 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3384 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3385 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3386 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3387 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3388 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3389 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3390 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3391 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3392 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3393 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3397 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3398 <result> = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3399 <result> = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3400 <result> = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3404 <!-- _______________________________________________________________________ -->
3405 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3406 Instruction</a> </div>
3407 <div class="doc_text">
3409 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
3411 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
3412 the SSA graph representing the function.</p>
3414 <p>The type of the incoming values are specified with the first type
3415 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3416 as arguments, with one pair for each predecessor basic block of the
3417 current block. Only values of <a href="#t_firstclass">first class</a>
3418 type may be used as the value arguments to the PHI node. Only labels
3419 may be used as the label arguments.</p>
3420 <p>There must be no non-phi instructions between the start of a basic
3421 block and the PHI instructions: i.e. PHI instructions must be first in
3424 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
3425 value specified by the parameter, depending on which basic block we
3426 came from in the last <a href="#terminators">terminator</a> instruction.</p>
3428 <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>
3431 <!-- _______________________________________________________________________ -->
3432 <div class="doc_subsubsection">
3433 <a name="i_select">'<tt>select</tt>' Instruction</a>
3436 <div class="doc_text">
3441 <result> = select i1 <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
3447 The '<tt>select</tt>' instruction is used to choose one value based on a
3448 condition, without branching.
3455 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.
3461 If the boolean condition evaluates to true, the instruction returns the first
3462 value argument; otherwise, it returns the second value argument.
3468 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3473 <!-- _______________________________________________________________________ -->
3474 <div class="doc_subsubsection">
3475 <a name="i_call">'<tt>call</tt>' Instruction</a>
3478 <div class="doc_text">
3482 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
3487 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3491 <p>This instruction requires several arguments:</p>
3495 <p>The optional "tail" marker indicates whether the callee function accesses
3496 any allocas or varargs in the caller. If the "tail" marker is present, the
3497 function call is eligible for tail call optimization. Note that calls may
3498 be marked "tail" even if they do not occur before a <a
3499 href="#i_ret"><tt>ret</tt></a> instruction.
3502 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3503 convention</a> the call should use. If none is specified, the call defaults
3504 to using C calling conventions.
3507 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
3508 being invoked. The argument types must match the types implied by this
3509 signature. This type can be omitted if the function is not varargs and
3510 if the function type does not return a pointer to a function.</p>
3513 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3514 be invoked. In most cases, this is a direct function invocation, but
3515 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3516 to function value.</p>
3519 <p>'<tt>function args</tt>': argument list whose types match the
3520 function signature argument types. All arguments must be of
3521 <a href="#t_firstclass">first class</a> type. If the function signature
3522 indicates the function accepts a variable number of arguments, the extra
3523 arguments can be specified.</p>
3529 <p>The '<tt>call</tt>' instruction is used to cause control flow to
3530 transfer to a specified function, with its incoming arguments bound to
3531 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3532 instruction in the called function, control flow continues with the
3533 instruction after the function call, and the return value of the
3534 function is bound to the result argument. This is a simpler case of
3535 the <a href="#i_invoke">invoke</a> instruction.</p>
3540 %retval = call i32 %test(i32 %argc)
3541 call i32(i8 *, ...) *%printf(i8 * %msg, i32 12, i8 42);
3542 %X = tail call i32 %foo()
3543 %Y = tail call <a href="#callingconv">fastcc</a> i32 %foo()
3548 <!-- _______________________________________________________________________ -->
3549 <div class="doc_subsubsection">
3550 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3553 <div class="doc_text">
3558 <resultval> = va_arg <va_list*> <arglist>, <argty>
3563 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3564 the "variable argument" area of a function call. It is used to implement the
3565 <tt>va_arg</tt> macro in C.</p>
3569 <p>This instruction takes a <tt>va_list*</tt> value and the type of
3570 the argument. It returns a value of the specified argument type and
3571 increments the <tt>va_list</tt> to point to the next argument. Again, the
3572 actual type of <tt>va_list</tt> is target specific.</p>
3576 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3577 type from the specified <tt>va_list</tt> and causes the
3578 <tt>va_list</tt> to point to the next argument. For more information,
3579 see the variable argument handling <a href="#int_varargs">Intrinsic
3582 <p>It is legal for this instruction to be called in a function which does not
3583 take a variable number of arguments, for example, the <tt>vfprintf</tt>
3586 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3587 href="#intrinsics">intrinsic function</a> because it takes a type as an
3592 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3596 <!-- *********************************************************************** -->
3597 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3598 <!-- *********************************************************************** -->
3600 <div class="doc_text">
3602 <p>LLVM supports the notion of an "intrinsic function". These functions have
3603 well known names and semantics and are required to follow certain restrictions.
3604 Overall, these intrinsics represent an extension mechanism for the LLVM
3605 language that does not require changing all of the transformations in LLVM to
3606 add to the language (or the bytecode reader/writer, the parser,
3609 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3610 prefix is reserved in LLVM for intrinsic names; thus, functions may not be named
3611 this. Intrinsic functions must always be external functions: you cannot define
3612 the body of intrinsic functions. Intrinsic functions may only be used in call
3613 or invoke instructions: it is illegal to take the address of an intrinsic
3614 function. Additionally, because intrinsic functions are part of the LLVM
3615 language, it is required that they all be documented here if any are added.</p>
3617 <p>Some intrinsic functions can be overloaded. That is, the intrinsic represents
3618 a family of functions that perform the same operation but on different data
3619 types. This is most frequent with the integer types. Since LLVM can represent
3620 over 8 million different integer types, there is a way to declare an intrinsic
3621 that can be overloaded based on its arguments. Such intrinsics will have the
3622 names of the arbitrary types encoded into the intrinsic function name, each
3623 preceded by a period. For example, the <tt>llvm.ctpop</tt> function can take an
3624 integer of any width. This leads to a family of functions such as
3625 <tt>i32 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i32 @llvm.ctpop.i29(i29 %val)</tt>.
3629 <p>To learn how to add an intrinsic function, please see the
3630 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3635 <!-- ======================================================================= -->
3636 <div class="doc_subsection">
3637 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3640 <div class="doc_text">
3642 <p>Variable argument support is defined in LLVM with the <a
3643 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3644 intrinsic functions. These functions are related to the similarly
3645 named macros defined in the <tt><stdarg.h></tt> header file.</p>
3647 <p>All of these functions operate on arguments that use a
3648 target-specific value type "<tt>va_list</tt>". The LLVM assembly
3649 language reference manual does not define what this type is, so all
3650 transformations should be prepared to handle intrinsics with any type
3653 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3654 instruction and the variable argument handling intrinsic functions are
3658 define i32 @test(i32 %X, ...) {
3659 ; Initialize variable argument processing
3661 %ap2 = bitcast i8** %ap to i8*
3662 call void @llvm.va_start(i8* %ap2)
3664 ; Read a single integer argument
3665 %tmp = va_arg i8 ** %ap, i32
3667 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3669 %aq2 = bitcast i8** %aq to i8*
3670 call void @llvm.va_copy(i8 *%aq2, i8* %ap2)
3671 call void @llvm.va_end(i8* %aq2)
3673 ; Stop processing of arguments.
3674 call void @llvm.va_end(i8* %ap2)
3678 declare void @llvm.va_start(i8*)
3679 declare void @llvm.va_copy(i8*, i8*)
3680 declare void @llvm.va_end(i8*)
3684 <!-- _______________________________________________________________________ -->
3685 <div class="doc_subsubsection">
3686 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3690 <div class="doc_text">
3692 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
3694 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3695 <tt>*<arglist></tt> for subsequent use by <tt><a
3696 href="#i_va_arg">va_arg</a></tt>.</p>
3700 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3704 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3705 macro available in C. In a target-dependent way, it initializes the
3706 <tt>va_list</tt> element the argument points to, so that the next call to
3707 <tt>va_arg</tt> will produce the first variable argument passed to the function.
3708 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3709 last argument of the function, the compiler can figure that out.</p>
3713 <!-- _______________________________________________________________________ -->
3714 <div class="doc_subsubsection">
3715 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3718 <div class="doc_text">
3720 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
3723 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
3724 which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
3725 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3729 <p>The argument is a <tt>va_list</tt> to destroy.</p>
3733 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3734 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
3735 Calls to <a href="#int_va_start"><tt>llvm.va_start</tt></a> and <a
3736 href="#int_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
3737 with calls to <tt>llvm.va_end</tt>.</p>
3741 <!-- _______________________________________________________________________ -->
3742 <div class="doc_subsubsection">
3743 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3746 <div class="doc_text">
3751 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
3756 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position from
3757 the source argument list to the destination argument list.</p>
3761 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3762 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3767 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt> macro
3768 available in C. In a target-dependent way, it copies the source
3769 <tt>va_list</tt> element into the destination list. This intrinsic is necessary
3770 because the <tt><a href="#int_va_start">llvm.va_start</a></tt> intrinsic may be
3771 arbitrarily complex and require memory allocation, for example.</p>
3775 <!-- ======================================================================= -->
3776 <div class="doc_subsection">
3777 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3780 <div class="doc_text">
3783 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3784 Collection</a> requires the implementation and generation of these intrinsics.
3785 These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
3786 stack</a>, as well as garbage collector implementations that require <a
3787 href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
3788 Front-ends for type-safe garbage collected languages should generate these
3789 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3790 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3794 <!-- _______________________________________________________________________ -->
3795 <div class="doc_subsubsection">
3796 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3799 <div class="doc_text">
3804 declare void @llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
3809 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3810 the code generator, and allows some metadata to be associated with it.</p>
3814 <p>The first argument specifies the address of a stack object that contains the
3815 root pointer. The second pointer (which must be either a constant or a global
3816 value address) contains the meta-data to be associated with the root.</p>
3820 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3821 location. At compile-time, the code generator generates information to allow
3822 the runtime to find the pointer at GC safe points.
3828 <!-- _______________________________________________________________________ -->
3829 <div class="doc_subsubsection">
3830 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
3833 <div class="doc_text">
3838 declare i8 * @llvm.gcread(i8 * %ObjPtr, i8 ** %Ptr)
3843 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
3844 locations, allowing garbage collector implementations that require read
3849 <p>The second argument is the address to read from, which should be an address
3850 allocated from the garbage collector. The first object is a pointer to the
3851 start of the referenced object, if needed by the language runtime (otherwise
3856 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
3857 instruction, but may be replaced with substantially more complex code by the
3858 garbage collector runtime, as needed.</p>
3863 <!-- _______________________________________________________________________ -->
3864 <div class="doc_subsubsection">
3865 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
3868 <div class="doc_text">
3873 declare void @llvm.gcwrite(i8 * %P1, i8 * %Obj, i8 ** %P2)
3878 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
3879 locations, allowing garbage collector implementations that require write
3880 barriers (such as generational or reference counting collectors).</p>
3884 <p>The first argument is the reference to store, the second is the start of the
3885 object to store it to, and the third is the address of the field of Obj to
3886 store to. If the runtime does not require a pointer to the object, Obj may be
3891 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
3892 instruction, but may be replaced with substantially more complex code by the
3893 garbage collector runtime, as needed.</p>
3899 <!-- ======================================================================= -->
3900 <div class="doc_subsection">
3901 <a name="int_codegen">Code Generator Intrinsics</a>
3904 <div class="doc_text">
3906 These intrinsics are provided by LLVM to expose special features that may only
3907 be implemented with code generator support.
3912 <!-- _______________________________________________________________________ -->
3913 <div class="doc_subsubsection">
3914 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
3917 <div class="doc_text">
3921 declare i8 *@llvm.returnaddress(i32 <level>)
3927 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
3928 target-specific value indicating the return address of the current function
3929 or one of its callers.
3935 The argument to this intrinsic indicates which function to return the address
3936 for. Zero indicates the calling function, one indicates its caller, etc. The
3937 argument is <b>required</b> to be a constant integer value.
3943 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
3944 the return address of the specified call frame, or zero if it cannot be
3945 identified. The value returned by this intrinsic is likely to be incorrect or 0
3946 for arguments other than zero, so it should only be used for debugging purposes.
3950 Note that calling this intrinsic does not prevent function inlining or other
3951 aggressive transformations, so the value returned may not be that of the obvious
3952 source-language caller.
3957 <!-- _______________________________________________________________________ -->
3958 <div class="doc_subsubsection">
3959 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
3962 <div class="doc_text">
3966 declare i8 *@llvm.frameaddress(i32 <level>)
3972 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
3973 target-specific frame pointer value for the specified stack frame.
3979 The argument to this intrinsic indicates which function to return the frame
3980 pointer for. Zero indicates the calling function, one indicates its caller,
3981 etc. The argument is <b>required</b> to be a constant integer value.
3987 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
3988 the frame address of the specified call frame, or zero if it cannot be
3989 identified. The value returned by this intrinsic is likely to be incorrect or 0
3990 for arguments other than zero, so it should only be used for debugging purposes.
3994 Note that calling this intrinsic does not prevent function inlining or other
3995 aggressive transformations, so the value returned may not be that of the obvious
3996 source-language caller.
4000 <!-- _______________________________________________________________________ -->
4001 <div class="doc_subsubsection">
4002 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4005 <div class="doc_text">
4009 declare i8 *@llvm.stacksave()
4015 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4016 the function stack, for use with <a href="#int_stackrestore">
4017 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4018 features like scoped automatic variable sized arrays in C99.
4024 This intrinsic returns a opaque pointer value that can be passed to <a
4025 href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4026 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4027 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4028 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4029 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4030 that were allocated after the <tt>llvm.stacksave</tt> was executed.
4035 <!-- _______________________________________________________________________ -->
4036 <div class="doc_subsubsection">
4037 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4040 <div class="doc_text">
4044 declare void @llvm.stackrestore(i8 * %ptr)
4050 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4051 the function stack to the state it was in when the corresponding <a
4052 href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4053 useful for implementing language features like scoped automatic variable sized
4060 See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4066 <!-- _______________________________________________________________________ -->
4067 <div class="doc_subsubsection">
4068 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4071 <div class="doc_text">
4075 declare void @llvm.prefetch(i8 * <address>,
4076 i32 <rw>, i32 <locality>)
4083 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4084 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4086 effect on the behavior of the program but can change its performance
4093 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4094 determining if the fetch should be for a read (0) or write (1), and
4095 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4096 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4097 <tt>locality</tt> arguments must be constant integers.
4103 This intrinsic does not modify the behavior of the program. In particular,
4104 prefetches cannot trap and do not produce a value. On targets that support this
4105 intrinsic, the prefetch can provide hints to the processor cache for better
4111 <!-- _______________________________________________________________________ -->
4112 <div class="doc_subsubsection">
4113 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4116 <div class="doc_text">
4120 declare void @llvm.pcmarker( i32 <id> )
4127 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4129 code to simulators and other tools. The method is target specific, but it is
4130 expected that the marker will use exported symbols to transmit the PC of the marker.
4131 The marker makes no guarantees that it will remain with any specific instruction
4132 after optimizations. It is possible that the presence of a marker will inhibit
4133 optimizations. The intended use is to be inserted after optimizations to allow
4134 correlations of simulation runs.
4140 <tt>id</tt> is a numerical id identifying the marker.
4146 This intrinsic does not modify the behavior of the program. Backends that do not
4147 support this intrinisic may ignore it.
4152 <!-- _______________________________________________________________________ -->
4153 <div class="doc_subsubsection">
4154 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4157 <div class="doc_text">
4161 declare i64 @llvm.readcyclecounter( )
4168 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4169 counter register (or similar low latency, high accuracy clocks) on those targets
4170 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4171 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4172 should only be used for small timings.
4178 When directly supported, reading the cycle counter should not modify any memory.
4179 Implementations are allowed to either return a application specific value or a
4180 system wide value. On backends without support, this is lowered to a constant 0.
4185 <!-- ======================================================================= -->
4186 <div class="doc_subsection">
4187 <a name="int_libc">Standard C Library Intrinsics</a>
4190 <div class="doc_text">
4192 LLVM provides intrinsics for a few important standard C library functions.
4193 These intrinsics allow source-language front-ends to pass information about the
4194 alignment of the pointer arguments to the code generator, providing opportunity
4195 for more efficient code generation.
4200 <!-- _______________________________________________________________________ -->
4201 <div class="doc_subsubsection">
4202 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4205 <div class="doc_text">
4209 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
4210 i32 <len>, i32 <align>)
4211 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
4212 i64 <len>, i32 <align>)
4218 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4219 location to the destination location.
4223 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4224 intrinsics do not return a value, and takes an extra alignment argument.
4230 The first argument is a pointer to the destination, the second is a pointer to
4231 the source. The third argument is an integer argument
4232 specifying the number of bytes to copy, and the fourth argument is the alignment
4233 of the source and destination locations.
4237 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4238 the caller guarantees that both the source and destination pointers are aligned
4245 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4246 location to the destination location, which are not allowed to overlap. It
4247 copies "len" bytes of memory over. If the argument is known to be aligned to
4248 some boundary, this can be specified as the fourth argument, otherwise it should
4254 <!-- _______________________________________________________________________ -->
4255 <div class="doc_subsubsection">
4256 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4259 <div class="doc_text">
4263 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
4264 i32 <len>, i32 <align>)
4265 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
4266 i64 <len>, i32 <align>)
4272 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4273 location to the destination location. It is similar to the
4274 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4278 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4279 intrinsics do not return a value, and takes an extra alignment argument.
4285 The first argument is a pointer to the destination, the second is a pointer to
4286 the source. The third argument is an integer argument
4287 specifying the number of bytes to copy, and the fourth argument is the alignment
4288 of the source and destination locations.
4292 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4293 the caller guarantees that the source and destination pointers are aligned to
4300 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4301 location to the destination location, which may overlap. It
4302 copies "len" bytes of memory over. If the argument is known to be aligned to
4303 some boundary, this can be specified as the fourth argument, otherwise it should
4309 <!-- _______________________________________________________________________ -->
4310 <div class="doc_subsubsection">
4311 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4314 <div class="doc_text">
4318 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
4319 i32 <len>, i32 <align>)
4320 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
4321 i64 <len>, i32 <align>)
4327 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4332 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4333 does not return a value, and takes an extra alignment argument.
4339 The first argument is a pointer to the destination to fill, the second is the
4340 byte value to fill it with, the third argument is an integer
4341 argument specifying the number of bytes to fill, and the fourth argument is the
4342 known alignment of destination location.
4346 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4347 the caller guarantees that the destination pointer is aligned to that boundary.
4353 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4355 destination location. If the argument is known to be aligned to some boundary,
4356 this can be specified as the fourth argument, otherwise it should be set to 0 or
4362 <!-- _______________________________________________________________________ -->
4363 <div class="doc_subsubsection">
4364 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4367 <div class="doc_text">
4371 declare float @llvm.sqrt.f32(float %Val)
4372 declare double @llvm.sqrt.f64(double %Val)
4378 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
4379 returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
4380 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4381 negative numbers (which allows for better optimization).
4387 The argument and return value are floating point numbers of the same type.
4393 This function returns the sqrt of the specified operand if it is a positive
4394 floating point number.
4398 <!-- _______________________________________________________________________ -->
4399 <div class="doc_subsubsection">
4400 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4403 <div class="doc_text">
4407 declare float @llvm.powi.f32(float %Val, i32 %power)
4408 declare double @llvm.powi.f64(double %Val, i32 %power)
4414 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4415 specified (positive or negative) power. The order of evaluation of
4416 multiplications is not defined.
4422 The second argument is an integer power, and the first is a value to raise to
4429 This function returns the first value raised to the second power with an
4430 unspecified sequence of rounding operations.</p>
4434 <!-- ======================================================================= -->
4435 <div class="doc_subsection">
4436 <a name="int_manip">Bit Manipulation Intrinsics</a>
4439 <div class="doc_text">
4441 LLVM provides intrinsics for a few important bit manipulation operations.
4442 These allow efficient code generation for some algorithms.
4447 <!-- _______________________________________________________________________ -->
4448 <div class="doc_subsubsection">
4449 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4452 <div class="doc_text">
4455 <p>This is an overloaded intrinsic function. You can use bswap on any integer
4456 type that is an even number of bytes (i.e. BitWidth % 16 == 0). Note the suffix
4457 that includes the type for the result and the operand.
4459 declare i16 @llvm.bswap.i16.i16(i16 <id>)
4460 declare i32 @llvm.bswap.i32.i32(i32 <id>)
4461 declare i64 @llvm.bswap.i64.i64(i64 <id>)
4467 The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
4468 values with an even number of bytes (positive multiple of 16 bits). These are
4469 useful for performing operations on data that is not in the target's native
4476 The <tt>llvm.bswap.16.i16</tt> intrinsic returns an i16 value that has the high
4477 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4478 intrinsic returns an i32 value that has the four bytes of the input i32
4479 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
4480 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48.i48</tt>,
4481 <tt>llvm.bswap.i64.i64</tt> and other intrinsics extend this concept to
4482 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4487 <!-- _______________________________________________________________________ -->
4488 <div class="doc_subsubsection">
4489 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4492 <div class="doc_text">
4495 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4496 width. Not all targets support all bit widths however.
4498 declare i32 @llvm.ctpop.i8 (i8 <src>)
4499 declare i32 @llvm.ctpop.i16(i16 <src>)
4500 declare i32 @llvm.ctpop.i32(i32 <src>)
4501 declare i32 @llvm.ctpop.i64(i64 <src>)
4502 declare i32 @llvm.ctpop.i256(i256 <src>)
4508 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4515 The only argument is the value to be counted. The argument may be of any
4516 integer type. The return type must match the argument type.
4522 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4526 <!-- _______________________________________________________________________ -->
4527 <div class="doc_subsubsection">
4528 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4531 <div class="doc_text">
4534 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4535 integer bit width. Not all targets support all bit widths however.
4537 declare i32 @llvm.ctlz.i8 (i8 <src>)
4538 declare i32 @llvm.ctlz.i16(i16 <src>)
4539 declare i32 @llvm.ctlz.i32(i32 <src>)
4540 declare i32 @llvm.ctlz.i64(i64 <src>)
4541 declare i32 @llvm.ctlz.i256(i256 <src>)
4547 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4548 leading zeros in a variable.
4554 The only argument is the value to be counted. The argument may be of any
4555 integer type. The return type must match the argument type.
4561 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4562 in a variable. If the src == 0 then the result is the size in bits of the type
4563 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4569 <!-- _______________________________________________________________________ -->
4570 <div class="doc_subsubsection">
4571 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4574 <div class="doc_text">
4577 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
4578 integer bit width. Not all targets support all bit widths however.
4580 declare i32 @llvm.cttz.i8 (i8 <src>)
4581 declare i32 @llvm.cttz.i16(i16 <src>)
4582 declare i32 @llvm.cttz.i32(i32 <src>)
4583 declare i32 @llvm.cttz.i64(i64 <src>)
4584 declare i32 @llvm.cttz.i256(i256 <src>)
4590 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4597 The only argument is the value to be counted. The argument may be of any
4598 integer type. The return type must match the argument type.
4604 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4605 in a variable. If the src == 0 then the result is the size in bits of the type
4606 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4610 <!-- _______________________________________________________________________ -->
4611 <div class="doc_subsubsection">
4612 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
4615 <div class="doc_text">
4618 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
4619 on any integer bit width.
4621 declare i17 @llvm.part.select.i17.i17 (i17 %val, i32 %loBit, i32 %hiBit)
4622 declare i29 @llvm.part.select.i29.i29 (i29 %val, i32 %loBit, i32 %hiBit)
4626 <p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
4627 range of bits from an integer value and returns them in the same bit width as
4628 the original value.</p>
4631 <p>The first argument, <tt>%val</tt> and the result may be integer types of
4632 any bit width but they must have the same bit width. The second and third
4633 arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
4636 <p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
4637 of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
4638 <tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
4639 operates in forward mode.</p>
4640 <p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
4641 right by <tt>%loBit</tt> bits and then ANDing it with a mask with
4642 only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
4644 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
4645 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
4646 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
4647 to determine the number of bits to retain.</li>
4648 <li>A mask of the retained bits is created by shifting a -1 value.</li>
4649 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
4651 <p>In reverse mode, a similar computation is made except that:</p>
4653 <li>The bits selected wrap around to include both the highest and lowest bits.
4654 For example, part.select(i16 X, 4, 7) selects bits from X with a mask of
4655 0x00F0 (forwards case) while part.select(i16 X, 8, 3) selects bits from X
4656 with a mask of 0xFF0F.</li>
4657 <li>The bits returned in the reverse case are reversed. So, if X has the value
4658 0x6ACF and we apply part.select(i16 X, 8, 3) to it, we get back the value
4663 <div class="doc_subsubsection">
4664 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
4667 <div class="doc_text">
4670 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
4671 on any integer bit width.
4673 declare i17 @llvm.part.set.i17.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
4674 declare i29 @llvm.part.set.i29.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
4678 <p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
4679 of bits in an integer value with another integer value. It returns the integer
4680 with the replaced bits.</p>
4683 <p>The first argument, <tt>%val</tt> and the result may be integer types of
4684 any bit width but they must have the same bit width. <tt>%val</tt> is the value
4685 whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
4686 integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
4687 type since they specify only a bit index.</p>
4690 <p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
4691 of operation: forwards and reverse. If <tt>%lo</tt> is greater than
4692 <tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
4693 operates in forward mode.</p>
4694 <p>For both modes, the <tt>%repl</tt> value is prepared for use by either
4695 truncating it down to the size of the replacement area or zero extending it
4696 up to that size.</p>
4697 <p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
4698 are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
4699 in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
4700 to the <tt>%hi</tt>th bit.
4701 <p>In reverse mode, a similar computation is made except that the bits replaced
4702 wrap around to include both the highest and lowest bits. For example, if a
4703 16 bit value is being replaced then <tt>%lo=8</tt> and <tt>%hi=4</tt> would
4704 cause these bits to be set: <tt>0xFF1F</tt>.</p>
4707 llvm.part.set(0xFFFF, 0, 4, 7) -> 0xFF0F
4708 llvm.part.set(0xFFFF, 0, 7, 4) -> 0x0060
4709 llvm.part.set(0xFFFF, 0, 8, 3) -> 0x00F0
4710 llvm.part.set(0xFFFF, 0, 3, 8) -> 0xFE07
4714 <!-- ======================================================================= -->
4715 <div class="doc_subsection">
4716 <a name="int_debugger">Debugger Intrinsics</a>
4719 <div class="doc_text">
4721 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
4722 are described in the <a
4723 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
4724 Debugging</a> document.
4729 <!-- ======================================================================= -->
4730 <div class="doc_subsection">
4731 <a name="int_eh">Exception Handling Intrinsics</a>
4734 <div class="doc_text">
4735 <p> The LLVM exception handling intrinsics (which all start with
4736 <tt>llvm.eh.</tt> prefix), are described in the <a
4737 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
4738 Handling</a> document. </p>
4742 <!-- *********************************************************************** -->
4745 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
4746 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
4747 <a href="http://validator.w3.org/check/referer"><img
4748 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
4750 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
4751 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
4752 Last modified: $Date$