1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>LLVM Assembly Language Reference Manual</title>
6 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
8 <meta name="description"
9 content="LLVM Assembly Language Reference Manual.">
10 <link rel="stylesheet" href="llvm.css" type="text/css">
15 <div class="doc_title"> LLVM Language Reference Manual </div>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
20 <li><a href="#highlevel">High Level Structure</a>
22 <li><a href="#modulestructure">Module Structure</a></li>
23 <li><a href="#linkage">Linkage Types</a></li>
24 <li><a href="#callingconv">Calling Conventions</a></li>
25 <li><a href="#globalvars">Global Variables</a></li>
26 <li><a href="#functionstructure">Functions</a></li>
27 <li><a href="#aliasstructure">Aliases</a>
28 <li><a href="#paramattrs">Parameter Attributes</a></li>
29 <li><a href="#moduleasm">Module-Level Inline Assembly</a></li>
30 <li><a href="#datalayout">Data Layout</a></li>
33 <li><a href="#typesystem">Type System</a>
35 <li><a href="#t_primitive">Primitive Types</a>
37 <li><a href="#t_classifications">Type Classifications</a></li>
40 <li><a href="#t_derived">Derived Types</a>
42 <li><a href="#t_array">Array Type</a></li>
43 <li><a href="#t_function">Function Type</a></li>
44 <li><a href="#t_pointer">Pointer Type</a></li>
45 <li><a href="#t_struct">Structure Type</a></li>
46 <li><a href="#t_pstruct">Packed Structure Type</a></li>
47 <li><a href="#t_vector">Vector Type</a></li>
48 <li><a href="#t_opaque">Opaque Type</a></li>
53 <li><a href="#constants">Constants</a>
55 <li><a href="#simpleconstants">Simple Constants</a>
56 <li><a href="#aggregateconstants">Aggregate Constants</a>
57 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
58 <li><a href="#undefvalues">Undefined Values</a>
59 <li><a href="#constantexprs">Constant Expressions</a>
62 <li><a href="#othervalues">Other Values</a>
64 <li><a href="#inlineasm">Inline Assembler Expressions</a>
67 <li><a href="#instref">Instruction Reference</a>
69 <li><a href="#terminators">Terminator Instructions</a>
71 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
72 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
73 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
74 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
75 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
76 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
79 <li><a href="#binaryops">Binary Operations</a>
81 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
82 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
83 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
84 <li><a href="#i_udiv">'<tt>udiv</tt>' Instruction</a></li>
85 <li><a href="#i_sdiv">'<tt>sdiv</tt>' Instruction</a></li>
86 <li><a href="#i_fdiv">'<tt>fdiv</tt>' Instruction</a></li>
87 <li><a href="#i_urem">'<tt>urem</tt>' Instruction</a></li>
88 <li><a href="#i_srem">'<tt>srem</tt>' Instruction</a></li>
89 <li><a href="#i_frem">'<tt>frem</tt>' Instruction</a></li>
92 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
94 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
95 <li><a href="#i_lshr">'<tt>lshr</tt>' Instruction</a></li>
96 <li><a href="#i_ashr">'<tt>ashr</tt>' Instruction</a></li>
97 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
98 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
99 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
102 <li><a href="#vectorops">Vector Operations</a>
104 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
105 <li><a href="#i_insertelement">'<tt>insertelement</tt>' Instruction</a></li>
106 <li><a href="#i_shufflevector">'<tt>shufflevector</tt>' Instruction</a></li>
109 <li><a href="#memoryops">Memory Access and Addressing Operations</a>
111 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
112 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
113 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
114 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
115 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
116 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
119 <li><a href="#convertops">Conversion Operations</a>
121 <li><a href="#i_trunc">'<tt>trunc .. to</tt>' Instruction</a></li>
122 <li><a href="#i_zext">'<tt>zext .. to</tt>' Instruction</a></li>
123 <li><a href="#i_sext">'<tt>sext .. to</tt>' Instruction</a></li>
124 <li><a href="#i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a></li>
125 <li><a href="#i_fpext">'<tt>fpext .. to</tt>' Instruction</a></li>
126 <li><a href="#i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a></li>
127 <li><a href="#i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a></li>
128 <li><a href="#i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a></li>
129 <li><a href="#i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a></li>
130 <li><a href="#i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a></li>
131 <li><a href="#i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a></li>
132 <li><a href="#i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a></li>
134 <li><a href="#otherops">Other Operations</a>
136 <li><a href="#i_icmp">'<tt>icmp</tt>' Instruction</a></li>
137 <li><a href="#i_fcmp">'<tt>fcmp</tt>' Instruction</a></li>
138 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
139 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
140 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
141 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a></li>
146 <li><a href="#intrinsics">Intrinsic Functions</a>
148 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
150 <li><a href="#int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
151 <li><a href="#int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
152 <li><a href="#int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
155 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
157 <li><a href="#int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
158 <li><a href="#int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
159 <li><a href="#int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
162 <li><a href="#int_codegen">Code Generator Intrinsics</a>
164 <li><a href="#int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
165 <li><a href="#int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
166 <li><a href="#int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a></li>
167 <li><a href="#int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a></li>
168 <li><a href="#int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
169 <li><a href="#int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
170 <li><a href="#int_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
173 <li><a href="#int_libc">Standard C Library Intrinsics</a>
175 <li><a href="#int_memcpy">'<tt>llvm.memcpy.*</tt>' Intrinsic</a></li>
176 <li><a href="#int_memmove">'<tt>llvm.memmove.*</tt>' Intrinsic</a></li>
177 <li><a href="#int_memset">'<tt>llvm.memset.*</tt>' Intrinsic</a></li>
178 <li><a href="#int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a></li>
179 <li><a href="#int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a></li>
182 <li><a href="#int_manip">Bit Manipulation Intrinsics</a>
184 <li><a href="#int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a></li>
185 <li><a href="#int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic </a></li>
186 <li><a href="#int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic </a></li>
187 <li><a href="#int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic </a></li>
188 <li><a href="#int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic </a></li>
189 <li><a href="#int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic </a></li>
192 <li><a href="#int_debugger">Debugger intrinsics</a></li>
193 <li><a href="#int_eh">Exception Handling intrinsics</a></li>
198 <div class="doc_author">
199 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
200 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
203 <!-- *********************************************************************** -->
204 <div class="doc_section"> <a name="abstract">Abstract </a></div>
205 <!-- *********************************************************************** -->
207 <div class="doc_text">
208 <p>This document is a reference manual for the LLVM assembly language.
209 LLVM is an SSA based representation that provides type safety,
210 low-level operations, flexibility, and the capability of representing
211 'all' high-level languages cleanly. It is the common code
212 representation used throughout all phases of the LLVM compilation
216 <!-- *********************************************************************** -->
217 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
218 <!-- *********************************************************************** -->
220 <div class="doc_text">
222 <p>The LLVM code representation is designed to be used in three
223 different forms: as an in-memory compiler IR, as an on-disk bytecode
224 representation (suitable for fast loading by a Just-In-Time compiler),
225 and as a human readable assembly language representation. This allows
226 LLVM to provide a powerful intermediate representation for efficient
227 compiler transformations and analysis, while providing a natural means
228 to debug and visualize the transformations. The three different forms
229 of LLVM are all equivalent. This document describes the human readable
230 representation and notation.</p>
232 <p>The LLVM representation aims to be light-weight and low-level
233 while being expressive, typed, and extensible at the same time. It
234 aims to be a "universal IR" of sorts, by being at a low enough level
235 that high-level ideas may be cleanly mapped to it (similar to how
236 microprocessors are "universal IR's", allowing many source languages to
237 be mapped to them). By providing type information, LLVM can be used as
238 the target of optimizations: for example, through pointer analysis, it
239 can be proven that a C automatic variable is never accessed outside of
240 the current function... allowing it to be promoted to a simple SSA
241 value instead of a memory location.</p>
245 <!-- _______________________________________________________________________ -->
246 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
248 <div class="doc_text">
250 <p>It is important to note that this document describes 'well formed'
251 LLVM assembly language. There is a difference between what the parser
252 accepts and what is considered 'well formed'. For example, the
253 following instruction is syntactically okay, but not well formed:</p>
255 <div class="doc_code">
257 %x = <a href="#i_add">add</a> i32 1, %x
261 <p>...because the definition of <tt>%x</tt> does not dominate all of
262 its uses. The LLVM infrastructure provides a verification pass that may
263 be used to verify that an LLVM module is well formed. This pass is
264 automatically run by the parser after parsing input assembly and by
265 the optimizer before it outputs bytecode. The violations pointed out
266 by the verifier pass indicate bugs in transformation passes or input to
270 <!-- Describe the typesetting conventions here. --> </div>
272 <!-- *********************************************************************** -->
273 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
274 <!-- *********************************************************************** -->
276 <div class="doc_text">
278 <p>LLVM uses three different forms of identifiers, for different
282 <li>Named values are represented as a string of characters with a '%' prefix.
283 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
284 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
285 Identifiers which require other characters in their names can be surrounded
286 with quotes. In this way, anything except a <tt>"</tt> character can be used
289 <li>Unnamed values are represented as an unsigned numeric value with a '%'
290 prefix. For example, %12, %2, %44.</li>
292 <li>Constants, which are described in a <a href="#constants">section about
293 constants</a>, below.</li>
296 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
297 don't need to worry about name clashes with reserved words, and the set of
298 reserved words may be expanded in the future without penalty. Additionally,
299 unnamed identifiers allow a compiler to quickly come up with a temporary
300 variable without having to avoid symbol table conflicts.</p>
302 <p>Reserved words in LLVM are very similar to reserved words in other
303 languages. There are keywords for different opcodes
304 ('<tt><a href="#i_add">add</a></tt>',
305 '<tt><a href="#i_bitcast">bitcast</a></tt>',
306 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
307 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
308 and others. These reserved words cannot conflict with variable names, because
309 none of them start with a '%' character.</p>
311 <p>Here is an example of LLVM code to multiply the integer variable
312 '<tt>%X</tt>' by 8:</p>
316 <div class="doc_code">
318 %result = <a href="#i_mul">mul</a> i32 %X, 8
322 <p>After strength reduction:</p>
324 <div class="doc_code">
326 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
330 <p>And the hard way:</p>
332 <div class="doc_code">
334 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
335 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
336 %result = <a href="#i_add">add</a> i32 %1, %1
340 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
341 important lexical features of LLVM:</p>
345 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
348 <li>Unnamed temporaries are created when the result of a computation is not
349 assigned to a named value.</li>
351 <li>Unnamed temporaries are numbered sequentially</li>
355 <p>...and it also shows a convention that we follow in this document. When
356 demonstrating instructions, we will follow an instruction with a comment that
357 defines the type and name of value produced. Comments are shown in italic
362 <!-- *********************************************************************** -->
363 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
364 <!-- *********************************************************************** -->
366 <!-- ======================================================================= -->
367 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
370 <div class="doc_text">
372 <p>LLVM programs are composed of "Module"s, each of which is a
373 translation unit of the input programs. Each module consists of
374 functions, global variables, and symbol table entries. Modules may be
375 combined together with the LLVM linker, which merges function (and
376 global variable) definitions, resolves forward declarations, and merges
377 symbol table entries. Here is an example of the "hello world" module:</p>
379 <div class="doc_code">
380 <pre><i>; Declare the string constant as a global constant...</i>
381 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
382 href="#globalvars">constant</a> <a href="#t_array">[13 x i8 ]</a> c"hello world\0A\00" <i>; [13 x i8 ]*</i>
384 <i>; External declaration of the puts function</i>
385 <a href="#functionstructure">declare</a> i32 %puts(i8 *) <i>; i32(i8 *)* </i>
387 <i>; Definition of main function</i>
388 define i32 %main() { <i>; i32()* </i>
389 <i>; Convert [13x i8 ]* to i8 *...</i>
391 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* %.LC0, i64 0, i64 0 <i>; i8 *</i>
393 <i>; Call puts function to write out the string to stdout...</i>
395 href="#i_call">call</a> i32 %puts(i8 * %cast210) <i>; i32</i>
397 href="#i_ret">ret</a> i32 0<br>}<br>
401 <p>This example is made up of a <a href="#globalvars">global variable</a>
402 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
403 function, and a <a href="#functionstructure">function definition</a>
404 for "<tt>main</tt>".</p>
406 <p>In general, a module is made up of a list of global values,
407 where both functions and global variables are global values. Global values are
408 represented by a pointer to a memory location (in this case, a pointer to an
409 array of char, and a pointer to a function), and have one of the following <a
410 href="#linkage">linkage types</a>.</p>
414 <!-- ======================================================================= -->
415 <div class="doc_subsection">
416 <a name="linkage">Linkage Types</a>
419 <div class="doc_text">
422 All Global Variables and Functions have one of the following types of linkage:
427 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
429 <dd>Global values with internal linkage are only directly accessible by
430 objects in the current module. In particular, linking code into a module with
431 an internal global value may cause the internal to be renamed as necessary to
432 avoid collisions. Because the symbol is internal to the module, all
433 references can be updated. This corresponds to the notion of the
434 '<tt>static</tt>' keyword in C.
437 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
439 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
440 the same name when linkage occurs. This is typically used to implement
441 inline functions, templates, or other code which must be generated in each
442 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
443 allowed to be discarded.
446 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
448 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
449 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
450 used for globals that may be emitted in multiple translation units, but that
451 are not guaranteed to be emitted into every translation unit that uses them.
452 One example of this are common globals in C, such as "<tt>int X;</tt>" at
456 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
458 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
459 pointer to array type. When two global variables with appending linkage are
460 linked together, the two global arrays are appended together. This is the
461 LLVM, typesafe, equivalent of having the system linker append together
462 "sections" with identical names when .o files are linked.
465 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
466 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
467 until linked, if not linked, the symbol becomes null instead of being an
471 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
473 <dd>If none of the above identifiers are used, the global is externally
474 visible, meaning that it participates in linkage and can be used to resolve
475 external symbol references.
480 The next two types of linkage are targeted for Microsoft Windows platform
481 only. They are designed to support importing (exporting) symbols from (to)
486 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
488 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
489 or variable via a global pointer to a pointer that is set up by the DLL
490 exporting the symbol. On Microsoft Windows targets, the pointer name is
491 formed by combining <code>_imp__</code> and the function or variable name.
494 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
496 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
497 pointer to a pointer in a DLL, so that it can be referenced with the
498 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
499 name is formed by combining <code>_imp__</code> and the function or variable
505 <p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
506 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
507 variable and was linked with this one, one of the two would be renamed,
508 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
509 external (i.e., lacking any linkage declarations), they are accessible
510 outside of the current module.</p>
511 <p>It is illegal for a function <i>declaration</i>
512 to have any linkage type other than "externally visible", <tt>dllimport</tt>,
513 or <tt>extern_weak</tt>.</p>
514 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
518 <!-- ======================================================================= -->
519 <div class="doc_subsection">
520 <a name="callingconv">Calling Conventions</a>
523 <div class="doc_text">
525 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
526 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
527 specified for the call. The calling convention of any pair of dynamic
528 caller/callee must match, or the behavior of the program is undefined. The
529 following calling conventions are supported by LLVM, and more may be added in
533 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
535 <dd>This calling convention (the default if no other calling convention is
536 specified) matches the target C calling conventions. This calling convention
537 supports varargs function calls and tolerates some mismatch in the declared
538 prototype and implemented declaration of the function (as does normal C).
541 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
543 <dd>This calling convention attempts to make calls as fast as possible
544 (e.g. by passing things in registers). This calling convention allows the
545 target to use whatever tricks it wants to produce fast code for the target,
546 without having to conform to an externally specified ABI. Implementations of
547 this convention should allow arbitrary tail call optimization to be supported.
548 This calling convention does not support varargs and requires the prototype of
549 all callees to exactly match the prototype of the function definition.
552 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
554 <dd>This calling convention attempts to make code in the caller as efficient
555 as possible under the assumption that the call is not commonly executed. As
556 such, these calls often preserve all registers so that the call does not break
557 any live ranges in the caller side. This calling convention does not support
558 varargs and requires the prototype of all callees to exactly match the
559 prototype of the function definition.
562 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
564 <dd>Any calling convention may be specified by number, allowing
565 target-specific calling conventions to be used. Target specific calling
566 conventions start at 64.
570 <p>More calling conventions can be added/defined on an as-needed basis, to
571 support pascal conventions or any other well-known target-independent
576 <!-- ======================================================================= -->
577 <div class="doc_subsection">
578 <a name="visibility">Visibility Styles</a>
581 <div class="doc_text">
584 All Global Variables and Functions have one of the following visibility styles:
588 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
590 <dd>On ELF, default visibility means that the declaration is visible to other
591 modules and, in shared libraries, means that the declared entity may be
592 overridden. On Darwin, default visibility means that the declaration is
593 visible to other modules. Default visibility corresponds to "external
594 linkage" in the language.
597 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
599 <dd>Two declarations of an object with hidden visibility refer to the same
600 object if they are in the same shared object. Usually, hidden visibility
601 indicates that the symbol will not be placed into the dynamic symbol table,
602 so no other module (executable or shared library) can reference it
606 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
608 <dd>On ELF, protected visibility indicates that the symbol will be placed in
609 the dynamic symbol table, but that references within the defining module will
610 bind to the local symbol. That is, the symbol cannot be overridden by another
617 <!-- ======================================================================= -->
618 <div class="doc_subsection">
619 <a name="globalvars">Global Variables</a>
622 <div class="doc_text">
624 <p>Global variables define regions of memory allocated at compilation time
625 instead of run-time. Global variables may optionally be initialized, may have
626 an explicit section to be placed in, and may have an optional explicit alignment
627 specified. A variable may be defined as "thread_local", which means that it
628 will not be shared by threads (each thread will have a separated copy of the
629 variable). A variable may be defined as a global "constant," which indicates
630 that the contents of the variable will <b>never</b> be modified (enabling better
631 optimization, allowing the global data to be placed in the read-only section of
632 an executable, etc). Note that variables that need runtime initialization
633 cannot be marked "constant" as there is a store to the variable.</p>
636 LLVM explicitly allows <em>declarations</em> of global variables to be marked
637 constant, even if the final definition of the global is not. This capability
638 can be used to enable slightly better optimization of the program, but requires
639 the language definition to guarantee that optimizations based on the
640 'constantness' are valid for the translation units that do not include the
644 <p>As SSA values, global variables define pointer values that are in
645 scope (i.e. they dominate) all basic blocks in the program. Global
646 variables always define a pointer to their "content" type because they
647 describe a region of memory, and all memory objects in LLVM are
648 accessed through pointers.</p>
650 <p>LLVM allows an explicit section to be specified for globals. If the target
651 supports it, it will emit globals to the section specified.</p>
653 <p>An explicit alignment may be specified for a global. If not present, or if
654 the alignment is set to zero, the alignment of the global is set by the target
655 to whatever it feels convenient. If an explicit alignment is specified, the
656 global is forced to have at least that much alignment. All alignments must be
659 <p>For example, the following defines a global with an initializer, section,
662 <div class="doc_code">
664 %G = constant float 1.0, section "foo", align 4
671 <!-- ======================================================================= -->
672 <div class="doc_subsection">
673 <a name="functionstructure">Functions</a>
676 <div class="doc_text">
678 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
679 an optional <a href="#linkage">linkage type</a>, an optional
680 <a href="#visibility">visibility style</a>, an optional
681 <a href="#callingconv">calling convention</a>, a return type, an optional
682 <a href="#paramattrs">parameter attribute</a> for the return type, a function
683 name, a (possibly empty) argument list (each with optional
684 <a href="#paramattrs">parameter attributes</a>), an optional section, an
685 optional alignment, an opening curly brace, a list of basic blocks, and a
688 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
689 optional <a href="#linkage">linkage type</a>, an optional
690 <a href="#visibility">visibility style</a>, an optional
691 <a href="#callingconv">calling convention</a>, a return type, an optional
692 <a href="#paramattrs">parameter attribute</a> for the return type, a function
693 name, a possibly empty list of arguments, and an optional alignment.</p>
695 <p>A function definition contains a list of basic blocks, forming the CFG for
696 the function. Each basic block may optionally start with a label (giving the
697 basic block a symbol table entry), contains a list of instructions, and ends
698 with a <a href="#terminators">terminator</a> instruction (such as a branch or
699 function return).</p>
701 <p>The first basic block in a program is special in two ways: it is immediately
702 executed on entrance to the function, and it is not allowed to have predecessor
703 basic blocks (i.e. there can not be any branches to the entry block of a
704 function). Because the block can have no predecessors, it also cannot have any
705 <a href="#i_phi">PHI nodes</a>.</p>
707 <p>LLVM functions are identified by their name and type signature. Hence, two
708 functions with the same name but different parameter lists or return values are
709 considered different functions, and LLVM will resolve references to each
712 <p>LLVM allows an explicit section to be specified for functions. If the target
713 supports it, it will emit functions to the section specified.</p>
715 <p>An explicit alignment may be specified for a function. If not present, or if
716 the alignment is set to zero, the alignment of the function is set by the target
717 to whatever it feels convenient. If an explicit alignment is specified, the
718 function is forced to have at least that much alignment. All alignments must be
724 <!-- ======================================================================= -->
725 <div class="doc_subsection">
726 <a name="aliasstructure">Aliases</a>
728 <div class="doc_text">
729 <p>Aliases act as "second name" for the aliasee value (which can be either
730 function or global variable or bitcast of global value). Aliases may have an
731 optional <a href="#linkage">linkage type</a>, and an
732 optional <a href="#visibility">visibility style</a>.</p>
736 <div class="doc_code">
738 @<Name> = [Linkage] [Visibility] alias <AliaseeTy> @<Aliasee>
746 <!-- ======================================================================= -->
747 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
748 <div class="doc_text">
749 <p>The return type and each parameter of a function type may have a set of
750 <i>parameter attributes</i> associated with them. Parameter attributes are
751 used to communicate additional information about the result or parameters of
752 a function. Parameter attributes are considered to be part of the function
753 type so two functions types that differ only by the parameter attributes
754 are different function types.</p>
756 <p>Parameter attributes are simple keywords that follow the type specified. If
757 multiple parameter attributes are needed, they are space separated. For
760 <div class="doc_code">
762 %someFunc = i16 (i8 sext %someParam) zext
763 %someFunc = i16 (i8 zext %someParam) zext
767 <p>Note that the two function types above are unique because the parameter has
768 a different attribute (sext in the first one, zext in the second). Also note
769 that the attribute for the function result (zext) comes immediately after the
772 <p>Currently, only the following parameter attributes are defined:</p>
774 <dt><tt>zext</tt></dt>
775 <dd>This indicates that the parameter should be zero extended just before
776 a call to this function.</dd>
777 <dt><tt>sext</tt></dt>
778 <dd>This indicates that the parameter should be sign extended just before
779 a call to this function.</dd>
780 <dt><tt>inreg</tt></dt>
781 <dd>This indicates that the parameter should be placed in register (if
782 possible) during assembling function call. Support for this attribute is
784 <dt><tt>sret</tt></dt>
785 <dd>This indicates that the parameter specifies the address of a structure
786 that is the return value of the function in the source program.</dd>
787 <dt><tt>noreturn</tt></dt>
788 <dd>This function attribute indicates that the function never returns. This
789 indicates to LLVM that every call to this function should be treated as if
790 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
791 <dt><tt>nounwind</tt></dt>
792 <dd>This function attribute indicates that the function type does not use
793 the unwind instruction and does not allow stack unwinding to propagate
799 <!-- ======================================================================= -->
800 <div class="doc_subsection">
801 <a name="moduleasm">Module-Level Inline Assembly</a>
804 <div class="doc_text">
806 Modules may contain "module-level inline asm" blocks, which corresponds to the
807 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
808 LLVM and treated as a single unit, but may be separated in the .ll file if
809 desired. The syntax is very simple:
812 <div class="doc_code">
814 module asm "inline asm code goes here"
815 module asm "more can go here"
819 <p>The strings can contain any character by escaping non-printable characters.
820 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
825 The inline asm code is simply printed to the machine code .s file when
826 assembly code is generated.
830 <!-- ======================================================================= -->
831 <div class="doc_subsection">
832 <a name="datalayout">Data Layout</a>
835 <div class="doc_text">
836 <p>A module may specify a target specific data layout string that specifies how
837 data is to be laid out in memory. The syntax for the data layout is simply:</p>
838 <pre> target datalayout = "<i>layout specification</i>"</pre>
839 <p>The <i>layout specification</i> consists of a list of specifications
840 separated by the minus sign character ('-'). Each specification starts with a
841 letter and may include other information after the letter to define some
842 aspect of the data layout. The specifications accepted are as follows: </p>
845 <dd>Specifies that the target lays out data in big-endian form. That is, the
846 bits with the most significance have the lowest address location.</dd>
848 <dd>Specifies that hte target lays out data in little-endian form. That is,
849 the bits with the least significance have the lowest address location.</dd>
850 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
851 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
852 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
853 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
855 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
856 <dd>This specifies the alignment for an integer type of a given bit
857 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
858 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
859 <dd>This specifies the alignment for a vector type of a given bit
861 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
862 <dd>This specifies the alignment for a floating point type of a given bit
863 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
865 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
866 <dd>This specifies the alignment for an aggregate type of a given bit
869 <p>When constructing the data layout for a given target, LLVM starts with a
870 default set of specifications which are then (possibly) overriden by the
871 specifications in the <tt>datalayout</tt> keyword. The default specifications
872 are given in this list:</p>
874 <li><tt>E</tt> - big endian</li>
875 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
876 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
877 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
878 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
879 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
880 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
881 alignment of 64-bits</li>
882 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
883 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
884 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
885 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
886 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
888 <p>When llvm is determining the alignment for a given type, it uses the
891 <li>If the type sought is an exact match for one of the specifications, that
892 specification is used.</li>
893 <li>If no match is found, and the type sought is an integer type, then the
894 smallest integer type that is larger than the bitwidth of the sought type is
895 used. If none of the specifications are larger than the bitwidth then the the
896 largest integer type is used. For example, given the default specifications
897 above, the i7 type will use the alignment of i8 (next largest) while both
898 i65 and i256 will use the alignment of i64 (largest specified).</li>
899 <li>If no match is found, and the type sought is a vector type, then the
900 largest vector type that is smaller than the sought vector type will be used
901 as a fall back. This happens because <128 x double> can be implemented in
902 terms of 64 <2 x double>, for example.</li>
906 <!-- *********************************************************************** -->
907 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
908 <!-- *********************************************************************** -->
910 <div class="doc_text">
912 <p>The LLVM type system is one of the most important features of the
913 intermediate representation. Being typed enables a number of
914 optimizations to be performed on the IR directly, without having to do
915 extra analyses on the side before the transformation. A strong type
916 system makes it easier to read the generated code and enables novel
917 analyses and transformations that are not feasible to perform on normal
918 three address code representations.</p>
922 <!-- ======================================================================= -->
923 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
924 <div class="doc_text">
925 <p>The primitive types are the fundamental building blocks of the LLVM
926 system. The current set of primitive types is as follows:</p>
928 <table class="layout">
933 <tr><th>Type</th><th>Description</th></tr>
934 <tr><td><tt><a name="t_void">void</a></tt></td><td>No value</td></tr>
935 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
942 <tr><th>Type</th><th>Description</th></tr>
943 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
944 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
952 <!-- _______________________________________________________________________ -->
953 <div class="doc_subsubsection"> <a name="t_classifications">Type
954 Classifications</a> </div>
955 <div class="doc_text">
956 <p>These different primitive types fall into a few useful
959 <table border="1" cellspacing="0" cellpadding="4">
961 <tr><th>Classification</th><th>Types</th></tr>
963 <td><a name="t_integer">integer</a></td>
964 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
967 <td><a name="t_floating">floating point</a></td>
968 <td><tt>float, double</tt></td>
971 <td><a name="t_firstclass">first class</a></td>
972 <td><tt>i1, ..., float, double, <br/>
973 <a href="#t_pointer">pointer</a>,<a href="#t_vector">vector</a></tt>
979 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
980 most important. Values of these types are the only ones which can be
981 produced by instructions, passed as arguments, or used as operands to
982 instructions. This means that all structures and arrays must be
983 manipulated either by pointer or by component.</p>
986 <!-- ======================================================================= -->
987 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
989 <div class="doc_text">
991 <p>The real power in LLVM comes from the derived types in the system.
992 This is what allows a programmer to represent arrays, functions,
993 pointers, and other useful types. Note that these derived types may be
994 recursive: For example, it is possible to have a two dimensional array.</p>
998 <!-- _______________________________________________________________________ -->
999 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1001 <div class="doc_text">
1004 <p>The integer type is a very simple derived type that simply specifies an
1005 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1006 2^23-1 (about 8 million) can be specified.</p>
1014 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1018 <table class="layout">
1028 <tt>i1942652</tt><br/>
1031 A boolean integer of 1 bit<br/>
1032 A nibble sized integer of 4 bits.<br/>
1033 A byte sized integer of 8 bits.<br/>
1034 A half word sized integer of 16 bits.<br/>
1035 A word sized integer of 32 bits.<br/>
1036 An integer whose bit width is the answer. <br/>
1037 A double word sized integer of 64 bits.<br/>
1038 A really big integer of over 1 million bits.<br/>
1044 <!-- _______________________________________________________________________ -->
1045 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1047 <div class="doc_text">
1051 <p>The array type is a very simple derived type that arranges elements
1052 sequentially in memory. The array type requires a size (number of
1053 elements) and an underlying data type.</p>
1058 [<# elements> x <elementtype>]
1061 <p>The number of elements is a constant integer value; elementtype may
1062 be any type with a size.</p>
1065 <table class="layout">
1068 <tt>[40 x i32 ]</tt><br/>
1069 <tt>[41 x i32 ]</tt><br/>
1070 <tt>[40 x i8]</tt><br/>
1073 Array of 40 32-bit integer values.<br/>
1074 Array of 41 32-bit integer values.<br/>
1075 Array of 40 8-bit integer values.<br/>
1079 <p>Here are some examples of multidimensional arrays:</p>
1080 <table class="layout">
1083 <tt>[3 x [4 x i32]]</tt><br/>
1084 <tt>[12 x [10 x float]]</tt><br/>
1085 <tt>[2 x [3 x [4 x i16]]]</tt><br/>
1088 3x4 array of 32-bit integer values.<br/>
1089 12x10 array of single precision floating point values.<br/>
1090 2x3x4 array of 16-bit integer values.<br/>
1095 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1096 length array. Normally, accesses past the end of an array are undefined in
1097 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1098 As a special case, however, zero length arrays are recognized to be variable
1099 length. This allows implementation of 'pascal style arrays' with the LLVM
1100 type "{ i32, [0 x float]}", for example.</p>
1104 <!-- _______________________________________________________________________ -->
1105 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1106 <div class="doc_text">
1108 <p>The function type can be thought of as a function signature. It
1109 consists of a return type and a list of formal parameter types.
1110 Function types are usually used to build virtual function tables
1111 (which are structures of pointers to functions), for indirect function
1112 calls, and when defining a function.</p>
1114 The return type of a function type cannot be an aggregate type.
1117 <pre> <returntype> (<parameter list>)<br></pre>
1118 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1119 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1120 which indicates that the function takes a variable number of arguments.
1121 Variable argument functions can access their arguments with the <a
1122 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1124 <table class="layout">
1126 <td class="left"><tt>i32 (i32)</tt></td>
1127 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1129 </tr><tr class="layout">
1130 <td class="left"><tt>float (i16 sext, i32 *) *
1132 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1133 an <tt>i16</tt> that should be sign extended and a
1134 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1137 </tr><tr class="layout">
1138 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1139 <td class="left">A vararg function that takes at least one
1140 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1141 which returns an integer. This is the signature for <tt>printf</tt> in
1148 <!-- _______________________________________________________________________ -->
1149 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1150 <div class="doc_text">
1152 <p>The structure type is used to represent a collection of data members
1153 together in memory. The packing of the field types is defined to match
1154 the ABI of the underlying processor. The elements of a structure may
1155 be any type that has a size.</p>
1156 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1157 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1158 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1161 <pre> { <type list> }<br></pre>
1163 <table class="layout">
1165 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1166 <td class="left">A triple of three <tt>i32</tt> values</td>
1167 </tr><tr class="layout">
1168 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1169 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1170 second element is a <a href="#t_pointer">pointer</a> to a
1171 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1172 an <tt>i32</tt>.</td>
1177 <!-- _______________________________________________________________________ -->
1178 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1180 <div class="doc_text">
1182 <p>The packed structure type is used to represent a collection of data members
1183 together in memory. There is no padding between fields. Further, the alignment
1184 of a packed structure is 1 byte. The elements of a packed structure may
1185 be any type that has a size.</p>
1186 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1187 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1188 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1191 <pre> < { <type list> } > <br></pre>
1193 <table class="layout">
1195 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1196 <td class="left">A triple of three <tt>i32</tt> values</td>
1197 </tr><tr class="layout">
1198 <td class="left"><tt>< { float, i32 (i32) * } ></tt></td>
1199 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1200 second element is a <a href="#t_pointer">pointer</a> to a
1201 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1202 an <tt>i32</tt>.</td>
1207 <!-- _______________________________________________________________________ -->
1208 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1209 <div class="doc_text">
1211 <p>As in many languages, the pointer type represents a pointer or
1212 reference to another object, which must live in memory.</p>
1214 <pre> <type> *<br></pre>
1216 <table class="layout">
1219 <tt>[4x i32]*</tt><br/>
1220 <tt>i32 (i32 *) *</tt><br/>
1223 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
1224 four <tt>i32</tt> values<br/>
1225 A <a href="#t_pointer">pointer</a> to a <a
1226 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1233 <!-- _______________________________________________________________________ -->
1234 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1235 <div class="doc_text">
1239 <p>A vector type is a simple derived type that represents a vector
1240 of elements. Vector types are used when multiple primitive data
1241 are operated in parallel using a single instruction (SIMD).
1242 A vector type requires a size (number of
1243 elements) and an underlying primitive data type. Vectors must have a power
1244 of two length (1, 2, 4, 8, 16 ...). Vector types are
1245 considered <a href="#t_firstclass">first class</a>.</p>
1250 < <# elements> x <elementtype> >
1253 <p>The number of elements is a constant integer value; elementtype may
1254 be any integer or floating point type.</p>
1258 <table class="layout">
1261 <tt><4 x i32></tt><br/>
1262 <tt><8 x float></tt><br/>
1263 <tt><2 x i64></tt><br/>
1266 Vector of 4 32-bit integer values.<br/>
1267 Vector of 8 floating-point values.<br/>
1268 Vector of 2 64-bit integer values.<br/>
1274 <!-- _______________________________________________________________________ -->
1275 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1276 <div class="doc_text">
1280 <p>Opaque types are used to represent unknown types in the system. This
1281 corresponds (for example) to the C notion of a foward declared structure type.
1282 In LLVM, opaque types can eventually be resolved to any type (not just a
1283 structure type).</p>
1293 <table class="layout">
1299 An opaque type.<br/>
1306 <!-- *********************************************************************** -->
1307 <div class="doc_section"> <a name="constants">Constants</a> </div>
1308 <!-- *********************************************************************** -->
1310 <div class="doc_text">
1312 <p>LLVM has several different basic types of constants. This section describes
1313 them all and their syntax.</p>
1317 <!-- ======================================================================= -->
1318 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1320 <div class="doc_text">
1323 <dt><b>Boolean constants</b></dt>
1325 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1326 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1329 <dt><b>Integer constants</b></dt>
1331 <dd>Standard integers (such as '4') are constants of the <a
1332 href="#t_integer">integer</a> type. Negative numbers may be used with
1336 <dt><b>Floating point constants</b></dt>
1338 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1339 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1340 notation (see below). Floating point constants must have a <a
1341 href="#t_floating">floating point</a> type. </dd>
1343 <dt><b>Null pointer constants</b></dt>
1345 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1346 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1350 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1351 of floating point constants. For example, the form '<tt>double
1352 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1353 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1354 (and the only time that they are generated by the disassembler) is when a
1355 floating point constant must be emitted but it cannot be represented as a
1356 decimal floating point number. For example, NaN's, infinities, and other
1357 special values are represented in their IEEE hexadecimal format so that
1358 assembly and disassembly do not cause any bits to change in the constants.</p>
1362 <!-- ======================================================================= -->
1363 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1366 <div class="doc_text">
1367 <p>Aggregate constants arise from aggregation of simple constants
1368 and smaller aggregate constants.</p>
1371 <dt><b>Structure constants</b></dt>
1373 <dd>Structure constants are represented with notation similar to structure
1374 type definitions (a comma separated list of elements, surrounded by braces
1375 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* %G }</tt>",
1376 where "<tt>%G</tt>" is declared as "<tt>%G = external global i32</tt>". Structure constants
1377 must have <a href="#t_struct">structure type</a>, and the number and
1378 types of elements must match those specified by the type.
1381 <dt><b>Array constants</b></dt>
1383 <dd>Array constants are represented with notation similar to array type
1384 definitions (a comma separated list of elements, surrounded by square brackets
1385 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1386 constants must have <a href="#t_array">array type</a>, and the number and
1387 types of elements must match those specified by the type.
1390 <dt><b>Vector constants</b></dt>
1392 <dd>Vector constants are represented with notation similar to vector type
1393 definitions (a comma separated list of elements, surrounded by
1394 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1395 i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
1396 href="#t_vector">vector type</a>, and the number and types of elements must
1397 match those specified by the type.
1400 <dt><b>Zero initialization</b></dt>
1402 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1403 value to zero of <em>any</em> type, including scalar and aggregate types.
1404 This is often used to avoid having to print large zero initializers (e.g. for
1405 large arrays) and is always exactly equivalent to using explicit zero
1412 <!-- ======================================================================= -->
1413 <div class="doc_subsection">
1414 <a name="globalconstants">Global Variable and Function Addresses</a>
1417 <div class="doc_text">
1419 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1420 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1421 constants. These constants are explicitly referenced when the <a
1422 href="#identifiers">identifier for the global</a> is used and always have <a
1423 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1426 <div class="doc_code">
1430 %Z = global [2 x i32*] [ i32* %X, i32* %Y ]
1436 <!-- ======================================================================= -->
1437 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1438 <div class="doc_text">
1439 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1440 no specific value. Undefined values may be of any type and be used anywhere
1441 a constant is permitted.</p>
1443 <p>Undefined values indicate to the compiler that the program is well defined
1444 no matter what value is used, giving the compiler more freedom to optimize.
1448 <!-- ======================================================================= -->
1449 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1452 <div class="doc_text">
1454 <p>Constant expressions are used to allow expressions involving other constants
1455 to be used as constants. Constant expressions may be of any <a
1456 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1457 that does not have side effects (e.g. load and call are not supported). The
1458 following is the syntax for constant expressions:</p>
1461 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1462 <dd>Truncate a constant to another type. The bit size of CST must be larger
1463 than the bit size of TYPE. Both types must be integers.</dd>
1465 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1466 <dd>Zero extend a constant to another type. The bit size of CST must be
1467 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1469 <dt><b><tt>sext ( CST to TYPE )</tt></b></dt>
1470 <dd>Sign extend a constant to another type. The bit size of CST must be
1471 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1473 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1474 <dd>Truncate a floating point constant to another floating point type. The
1475 size of CST must be larger than the size of TYPE. Both types must be
1476 floating point.</dd>
1478 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1479 <dd>Floating point extend a constant to another type. The size of CST must be
1480 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1482 <dt><b><tt>fp2uint ( CST to TYPE )</tt></b></dt>
1483 <dd>Convert a floating point constant to the corresponding unsigned integer
1484 constant. TYPE must be an integer type. CST must be floating point. If the
1485 value won't fit in the integer type, the results are undefined.</dd>
1487 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1488 <dd>Convert a floating point constant to the corresponding signed integer
1489 constant. TYPE must be an integer type. CST must be floating point. If the
1490 value won't fit in the integer type, the results are undefined.</dd>
1492 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1493 <dd>Convert an unsigned integer constant to the corresponding floating point
1494 constant. TYPE must be floating point. CST must be of integer type. If the
1495 value won't fit in the floating point type, the results are undefined.</dd>
1497 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1498 <dd>Convert a signed integer constant to the corresponding floating point
1499 constant. TYPE must be floating point. CST must be of integer type. If the
1500 value won't fit in the floating point type, the results are undefined.</dd>
1502 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1503 <dd>Convert a pointer typed constant to the corresponding integer constant
1504 TYPE must be an integer type. CST must be of pointer type. The CST value is
1505 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1507 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1508 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1509 pointer type. CST must be of integer type. The CST value is zero extended,
1510 truncated, or unchanged to make it fit in a pointer size. This one is
1511 <i>really</i> dangerous!</dd>
1513 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1514 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1515 identical (same number of bits). The conversion is done as if the CST value
1516 was stored to memory and read back as TYPE. In other words, no bits change
1517 with this operator, just the type. This can be used for conversion of
1518 vector types to any other type, as long as they have the same bit width. For
1519 pointers it is only valid to cast to another pointer type.
1522 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1524 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1525 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1526 instruction, the index list may have zero or more indexes, which are required
1527 to make sense for the type of "CSTPTR".</dd>
1529 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1531 <dd>Perform the <a href="#i_select">select operation</a> on
1534 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1535 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1537 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1538 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1540 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1542 <dd>Perform the <a href="#i_extractelement">extractelement
1543 operation</a> on constants.
1545 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1547 <dd>Perform the <a href="#i_insertelement">insertelement
1548 operation</a> on constants.</dd>
1551 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1553 <dd>Perform the <a href="#i_shufflevector">shufflevector
1554 operation</a> on constants.</dd>
1556 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1558 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1559 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1560 binary</a> operations. The constraints on operands are the same as those for
1561 the corresponding instruction (e.g. no bitwise operations on floating point
1562 values are allowed).</dd>
1566 <!-- *********************************************************************** -->
1567 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1568 <!-- *********************************************************************** -->
1570 <!-- ======================================================================= -->
1571 <div class="doc_subsection">
1572 <a name="inlineasm">Inline Assembler Expressions</a>
1575 <div class="doc_text">
1578 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1579 Module-Level Inline Assembly</a>) through the use of a special value. This
1580 value represents the inline assembler as a string (containing the instructions
1581 to emit), a list of operand constraints (stored as a string), and a flag that
1582 indicates whether or not the inline asm expression has side effects. An example
1583 inline assembler expression is:
1586 <div class="doc_code">
1588 i32 (i32) asm "bswap $0", "=r,r"
1593 Inline assembler expressions may <b>only</b> be used as the callee operand of
1594 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1597 <div class="doc_code">
1599 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1604 Inline asms with side effects not visible in the constraint list must be marked
1605 as having side effects. This is done through the use of the
1606 '<tt>sideeffect</tt>' keyword, like so:
1609 <div class="doc_code">
1611 call void asm sideeffect "eieio", ""()
1615 <p>TODO: The format of the asm and constraints string still need to be
1616 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1617 need to be documented).
1622 <!-- *********************************************************************** -->
1623 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1624 <!-- *********************************************************************** -->
1626 <div class="doc_text">
1628 <p>The LLVM instruction set consists of several different
1629 classifications of instructions: <a href="#terminators">terminator
1630 instructions</a>, <a href="#binaryops">binary instructions</a>,
1631 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1632 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1633 instructions</a>.</p>
1637 <!-- ======================================================================= -->
1638 <div class="doc_subsection"> <a name="terminators">Terminator
1639 Instructions</a> </div>
1641 <div class="doc_text">
1643 <p>As mentioned <a href="#functionstructure">previously</a>, every
1644 basic block in a program ends with a "Terminator" instruction, which
1645 indicates which block should be executed after the current block is
1646 finished. These terminator instructions typically yield a '<tt>void</tt>'
1647 value: they produce control flow, not values (the one exception being
1648 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1649 <p>There are six different terminator instructions: the '<a
1650 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1651 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1652 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1653 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1654 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1658 <!-- _______________________________________________________________________ -->
1659 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1660 Instruction</a> </div>
1661 <div class="doc_text">
1663 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1664 ret void <i>; Return from void function</i>
1667 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1668 value) from a function back to the caller.</p>
1669 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1670 returns a value and then causes control flow, and one that just causes
1671 control flow to occur.</p>
1673 <p>The '<tt>ret</tt>' instruction may return any '<a
1674 href="#t_firstclass">first class</a>' type. Notice that a function is
1675 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1676 instruction inside of the function that returns a value that does not
1677 match the return type of the function.</p>
1679 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1680 returns back to the calling function's context. If the caller is a "<a
1681 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1682 the instruction after the call. If the caller was an "<a
1683 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1684 at the beginning of the "normal" destination block. If the instruction
1685 returns a value, that value shall set the call or invoke instruction's
1688 <pre> ret i32 5 <i>; Return an integer value of 5</i>
1689 ret void <i>; Return from a void function</i>
1692 <!-- _______________________________________________________________________ -->
1693 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1694 <div class="doc_text">
1696 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1699 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1700 transfer to a different basic block in the current function. There are
1701 two forms of this instruction, corresponding to a conditional branch
1702 and an unconditional branch.</p>
1704 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1705 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1706 unconditional form of the '<tt>br</tt>' instruction takes a single
1707 '<tt>label</tt>' value as a target.</p>
1709 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1710 argument is evaluated. If the value is <tt>true</tt>, control flows
1711 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1712 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1714 <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
1715 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1717 <!-- _______________________________________________________________________ -->
1718 <div class="doc_subsubsection">
1719 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1722 <div class="doc_text">
1726 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1731 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1732 several different places. It is a generalization of the '<tt>br</tt>'
1733 instruction, allowing a branch to occur to one of many possible
1739 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1740 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1741 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1742 table is not allowed to contain duplicate constant entries.</p>
1746 <p>The <tt>switch</tt> instruction specifies a table of values and
1747 destinations. When the '<tt>switch</tt>' instruction is executed, this
1748 table is searched for the given value. If the value is found, control flow is
1749 transfered to the corresponding destination; otherwise, control flow is
1750 transfered to the default destination.</p>
1752 <h5>Implementation:</h5>
1754 <p>Depending on properties of the target machine and the particular
1755 <tt>switch</tt> instruction, this instruction may be code generated in different
1756 ways. For example, it could be generated as a series of chained conditional
1757 branches or with a lookup table.</p>
1762 <i>; Emulate a conditional br instruction</i>
1763 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1764 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1766 <i>; Emulate an unconditional br instruction</i>
1767 switch i32 0, label %dest [ ]
1769 <i>; Implement a jump table:</i>
1770 switch i32 %val, label %otherwise [ i32 0, label %onzero
1772 i32 2, label %ontwo ]
1776 <!-- _______________________________________________________________________ -->
1777 <div class="doc_subsubsection">
1778 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1781 <div class="doc_text">
1786 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1787 to label <normal label> unwind label <exception label>
1792 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1793 function, with the possibility of control flow transfer to either the
1794 '<tt>normal</tt>' label or the
1795 '<tt>exception</tt>' label. If the callee function returns with the
1796 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1797 "normal" label. If the callee (or any indirect callees) returns with the "<a
1798 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1799 continued at the dynamically nearest "exception" label.</p>
1803 <p>This instruction requires several arguments:</p>
1807 The optional "cconv" marker indicates which <a href="#callingconv">calling
1808 convention</a> the call should use. If none is specified, the call defaults
1809 to using C calling conventions.
1811 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1812 function value being invoked. In most cases, this is a direct function
1813 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1814 an arbitrary pointer to function value.
1817 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1818 function to be invoked. </li>
1820 <li>'<tt>function args</tt>': argument list whose types match the function
1821 signature argument types. If the function signature indicates the function
1822 accepts a variable number of arguments, the extra arguments can be
1825 <li>'<tt>normal label</tt>': the label reached when the called function
1826 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1828 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1829 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1835 <p>This instruction is designed to operate as a standard '<tt><a
1836 href="#i_call">call</a></tt>' instruction in most regards. The primary
1837 difference is that it establishes an association with a label, which is used by
1838 the runtime library to unwind the stack.</p>
1840 <p>This instruction is used in languages with destructors to ensure that proper
1841 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1842 exception. Additionally, this is important for implementation of
1843 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1847 %retval = invoke i32 %Test(i32 15) to label %Continue
1848 unwind label %TestCleanup <i>; {i32}:retval set</i>
1849 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1850 unwind label %TestCleanup <i>; {i32}:retval set</i>
1855 <!-- _______________________________________________________________________ -->
1857 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1858 Instruction</a> </div>
1860 <div class="doc_text">
1869 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1870 at the first callee in the dynamic call stack which used an <a
1871 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1872 primarily used to implement exception handling.</p>
1876 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1877 immediately halt. The dynamic call stack is then searched for the first <a
1878 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1879 execution continues at the "exceptional" destination block specified by the
1880 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1881 dynamic call chain, undefined behavior results.</p>
1884 <!-- _______________________________________________________________________ -->
1886 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1887 Instruction</a> </div>
1889 <div class="doc_text">
1898 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1899 instruction is used to inform the optimizer that a particular portion of the
1900 code is not reachable. This can be used to indicate that the code after a
1901 no-return function cannot be reached, and other facts.</p>
1905 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1910 <!-- ======================================================================= -->
1911 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1912 <div class="doc_text">
1913 <p>Binary operators are used to do most of the computation in a
1914 program. They require two operands, execute an operation on them, and
1915 produce a single value. The operands might represent
1916 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
1917 The result value of a binary operator is not
1918 necessarily the same type as its operands.</p>
1919 <p>There are several different binary operators:</p>
1921 <!-- _______________________________________________________________________ -->
1922 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1923 Instruction</a> </div>
1924 <div class="doc_text">
1926 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1929 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1931 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1932 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1933 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1934 Both arguments must have identical types.</p>
1936 <p>The value produced is the integer or floating point sum of the two
1939 <pre> <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
1942 <!-- _______________________________________________________________________ -->
1943 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1944 Instruction</a> </div>
1945 <div class="doc_text">
1947 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1950 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1952 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1953 instruction present in most other intermediate representations.</p>
1955 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1956 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1958 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1959 Both arguments must have identical types.</p>
1961 <p>The value produced is the integer or floating point difference of
1962 the two operands.</p>
1965 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
1966 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
1969 <!-- _______________________________________________________________________ -->
1970 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1971 Instruction</a> </div>
1972 <div class="doc_text">
1974 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1977 <p>The '<tt>mul</tt>' instruction returns the product of its two
1980 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1981 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1983 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1984 Both arguments must have identical types.</p>
1986 <p>The value produced is the integer or floating point product of the
1988 <p>Because the operands are the same width, the result of an integer
1989 multiplication is the same whether the operands should be deemed unsigned or
1992 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
1995 <!-- _______________________________________________________________________ -->
1996 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
1998 <div class="doc_text">
2000 <pre> <result> = udiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2003 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2006 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2007 <a href="#t_integer">integer</a> values. Both arguments must have identical
2008 types. This instruction can also take <a href="#t_vector">vector</a> versions
2009 of the values in which case the elements must be integers.</p>
2011 <p>The value produced is the unsigned integer quotient of the two operands. This
2012 instruction always performs an unsigned division operation, regardless of
2013 whether the arguments are unsigned or not.</p>
2015 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2018 <!-- _______________________________________________________________________ -->
2019 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2021 <div class="doc_text">
2023 <pre> <result> = sdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2026 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2029 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2030 <a href="#t_integer">integer</a> values. Both arguments must have identical
2031 types. This instruction can also take <a href="#t_vector">vector</a> versions
2032 of the values in which case the elements must be integers.</p>
2034 <p>The value produced is the signed integer quotient of the two operands. This
2035 instruction always performs a signed division operation, regardless of whether
2036 the arguments are signed or not.</p>
2038 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2041 <!-- _______________________________________________________________________ -->
2042 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2043 Instruction</a> </div>
2044 <div class="doc_text">
2046 <pre> <result> = fdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2049 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2052 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2053 <a href="#t_floating">floating point</a> values. Both arguments must have
2054 identical types. This instruction can also take <a href="#t_vector">vector</a>
2055 versions of floating point values.</p>
2057 <p>The value produced is the floating point quotient of the two operands.</p>
2059 <pre> <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2062 <!-- _______________________________________________________________________ -->
2063 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2065 <div class="doc_text">
2067 <pre> <result> = urem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2070 <p>The '<tt>urem</tt>' instruction returns the remainder from the
2071 unsigned division of its two arguments.</p>
2073 <p>The two arguments to the '<tt>urem</tt>' instruction must be
2074 <a href="#t_integer">integer</a> values. Both arguments must have identical
2077 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2078 This instruction always performs an unsigned division to get the remainder,
2079 regardless of whether the arguments are unsigned or not.</p>
2081 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2085 <!-- _______________________________________________________________________ -->
2086 <div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2087 Instruction</a> </div>
2088 <div class="doc_text">
2090 <pre> <result> = srem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2093 <p>The '<tt>srem</tt>' instruction returns the remainder from the
2094 signed division of its two operands.</p>
2096 <p>The two arguments to the '<tt>srem</tt>' instruction must be
2097 <a href="#t_integer">integer</a> values. Both arguments must have identical
2100 <p>This instruction returns the <i>remainder</i> of a division (where the result
2101 has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2102 operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2103 a value. For more information about the difference, see <a
2104 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2105 Math Forum</a>. For a table of how this is implemented in various languages,
2106 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2107 Wikipedia: modulo operation</a>.</p>
2109 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2113 <!-- _______________________________________________________________________ -->
2114 <div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2115 Instruction</a> </div>
2116 <div class="doc_text">
2118 <pre> <result> = frem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2121 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2122 division of its two operands.</p>
2124 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2125 <a href="#t_floating">floating point</a> values. Both arguments must have
2126 identical types.</p>
2128 <p>This instruction returns the <i>remainder</i> of a division.</p>
2130 <pre> <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2134 <!-- ======================================================================= -->
2135 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2136 Operations</a> </div>
2137 <div class="doc_text">
2138 <p>Bitwise binary operators are used to do various forms of
2139 bit-twiddling in a program. They are generally very efficient
2140 instructions and can commonly be strength reduced from other
2141 instructions. They require two operands, execute an operation on them,
2142 and produce a single value. The resulting value of the bitwise binary
2143 operators is always the same type as its first operand.</p>
2146 <!-- _______________________________________________________________________ -->
2147 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2148 Instruction</a> </div>
2149 <div class="doc_text">
2151 <pre> <result> = shl <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2154 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2155 the left a specified number of bits.</p>
2157 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2158 href="#t_integer">integer</a> type.</p>
2160 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
2161 <h5>Example:</h5><pre>
2162 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2163 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2164 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2167 <!-- _______________________________________________________________________ -->
2168 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2169 Instruction</a> </div>
2170 <div class="doc_text">
2172 <pre> <result> = lshr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2176 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2177 operand shifted to the right a specified number of bits with zero fill.</p>
2180 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2181 <a href="#t_integer">integer</a> type.</p>
2184 <p>This instruction always performs a logical shift right operation. The most
2185 significant bits of the result will be filled with zero bits after the
2190 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2191 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2192 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2193 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2197 <!-- _______________________________________________________________________ -->
2198 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2199 Instruction</a> </div>
2200 <div class="doc_text">
2203 <pre> <result> = ashr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2207 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2208 operand shifted to the right a specified number of bits with sign extension.</p>
2211 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2212 <a href="#t_integer">integer</a> type.</p>
2215 <p>This instruction always performs an arithmetic shift right operation,
2216 The most significant bits of the result will be filled with the sign bit
2217 of <tt>var1</tt>.</p>
2221 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2222 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2223 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2224 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
2228 <!-- _______________________________________________________________________ -->
2229 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2230 Instruction</a> </div>
2231 <div class="doc_text">
2233 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2236 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2237 its two operands.</p>
2239 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
2240 href="#t_integer">integer</a> values. Both arguments must have
2241 identical types.</p>
2243 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2245 <div style="align: center">
2246 <table border="1" cellspacing="0" cellpadding="4">
2277 <pre> <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
2278 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2279 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2282 <!-- _______________________________________________________________________ -->
2283 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2284 <div class="doc_text">
2286 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2289 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2290 or of its two operands.</p>
2292 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
2293 href="#t_integer">integer</a> values. Both arguments must have
2294 identical types.</p>
2296 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2298 <div style="align: center">
2299 <table border="1" cellspacing="0" cellpadding="4">
2330 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2331 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2332 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2335 <!-- _______________________________________________________________________ -->
2336 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2337 Instruction</a> </div>
2338 <div class="doc_text">
2340 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2343 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2344 or of its two operands. The <tt>xor</tt> is used to implement the
2345 "one's complement" operation, which is the "~" operator in C.</p>
2347 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2348 href="#t_integer">integer</a> values. Both arguments must have
2349 identical types.</p>
2351 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2353 <div style="align: center">
2354 <table border="1" cellspacing="0" cellpadding="4">
2386 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2387 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2388 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2389 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2393 <!-- ======================================================================= -->
2394 <div class="doc_subsection">
2395 <a name="vectorops">Vector Operations</a>
2398 <div class="doc_text">
2400 <p>LLVM supports several instructions to represent vector operations in a
2401 target-independent manner. These instructions cover the element-access and
2402 vector-specific operations needed to process vectors effectively. While LLVM
2403 does directly support these vector operations, many sophisticated algorithms
2404 will want to use target-specific intrinsics to take full advantage of a specific
2409 <!-- _______________________________________________________________________ -->
2410 <div class="doc_subsubsection">
2411 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2414 <div class="doc_text">
2419 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
2425 The '<tt>extractelement</tt>' instruction extracts a single scalar
2426 element from a vector at a specified index.
2433 The first operand of an '<tt>extractelement</tt>' instruction is a
2434 value of <a href="#t_vector">vector</a> type. The second operand is
2435 an index indicating the position from which to extract the element.
2436 The index may be a variable.</p>
2441 The result is a scalar of the same type as the element type of
2442 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2443 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2444 results are undefined.
2450 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
2455 <!-- _______________________________________________________________________ -->
2456 <div class="doc_subsubsection">
2457 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2460 <div class="doc_text">
2465 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
2471 The '<tt>insertelement</tt>' instruction inserts a scalar
2472 element into a vector at a specified index.
2479 The first operand of an '<tt>insertelement</tt>' instruction is a
2480 value of <a href="#t_vector">vector</a> type. The second operand is a
2481 scalar value whose type must equal the element type of the first
2482 operand. The third operand is an index indicating the position at
2483 which to insert the value. The index may be a variable.</p>
2488 The result is a vector of the same type as <tt>val</tt>. Its
2489 element values are those of <tt>val</tt> except at position
2490 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2491 exceeds the length of <tt>val</tt>, the results are undefined.
2497 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
2501 <!-- _______________________________________________________________________ -->
2502 <div class="doc_subsubsection">
2503 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2506 <div class="doc_text">
2511 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x i32> <mask> <i>; yields <n x <ty>></i>
2517 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2518 from two input vectors, returning a vector of the same type.
2524 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2525 with types that match each other and types that match the result of the
2526 instruction. The third argument is a shuffle mask, which has the same number
2527 of elements as the other vector type, but whose element type is always 'i32'.
2531 The shuffle mask operand is required to be a constant vector with either
2532 constant integer or undef values.
2538 The elements of the two input vectors are numbered from left to right across
2539 both of the vectors. The shuffle mask operand specifies, for each element of
2540 the result vector, which element of the two input registers the result element
2541 gets. The element selector may be undef (meaning "don't care") and the second
2542 operand may be undef if performing a shuffle from only one vector.
2548 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
2549 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
2550 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
2551 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
2556 <!-- ======================================================================= -->
2557 <div class="doc_subsection">
2558 <a name="memoryops">Memory Access and Addressing Operations</a>
2561 <div class="doc_text">
2563 <p>A key design point of an SSA-based representation is how it
2564 represents memory. In LLVM, no memory locations are in SSA form, which
2565 makes things very simple. This section describes how to read, write,
2566 allocate, and free memory in LLVM.</p>
2570 <!-- _______________________________________________________________________ -->
2571 <div class="doc_subsubsection">
2572 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2575 <div class="doc_text">
2580 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2585 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2586 heap and returns a pointer to it.</p>
2590 <p>The '<tt>malloc</tt>' instruction allocates
2591 <tt>sizeof(<type>)*NumElements</tt>
2592 bytes of memory from the operating system and returns a pointer of the
2593 appropriate type to the program. If "NumElements" is specified, it is the
2594 number of elements allocated. If an alignment is specified, the value result
2595 of the allocation is guaranteed to be aligned to at least that boundary. If
2596 not specified, or if zero, the target can choose to align the allocation on any
2597 convenient boundary.</p>
2599 <p>'<tt>type</tt>' must be a sized type.</p>
2603 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2604 a pointer is returned.</p>
2609 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2611 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2612 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2613 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2614 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2615 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2619 <!-- _______________________________________________________________________ -->
2620 <div class="doc_subsubsection">
2621 <a name="i_free">'<tt>free</tt>' Instruction</a>
2624 <div class="doc_text">
2629 free <type> <value> <i>; yields {void}</i>
2634 <p>The '<tt>free</tt>' instruction returns memory back to the unused
2635 memory heap to be reallocated in the future.</p>
2639 <p>'<tt>value</tt>' shall be a pointer value that points to a value
2640 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2645 <p>Access to the memory pointed to by the pointer is no longer defined
2646 after this instruction executes.</p>
2651 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2652 free [4 x i8]* %array
2656 <!-- _______________________________________________________________________ -->
2657 <div class="doc_subsubsection">
2658 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2661 <div class="doc_text">
2666 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2671 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
2672 currently executing function, to be automatically released when this function
2673 returns to its caller.</p>
2677 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2678 bytes of memory on the runtime stack, returning a pointer of the
2679 appropriate type to the program. If "NumElements" is specified, it is the
2680 number of elements allocated. If an alignment is specified, the value result
2681 of the allocation is guaranteed to be aligned to at least that boundary. If
2682 not specified, or if zero, the target can choose to align the allocation on any
2683 convenient boundary.</p>
2685 <p>'<tt>type</tt>' may be any sized type.</p>
2689 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2690 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2691 instruction is commonly used to represent automatic variables that must
2692 have an address available. When the function returns (either with the <tt><a
2693 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2694 instructions), the memory is reclaimed.</p>
2699 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2700 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2701 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2702 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2706 <!-- _______________________________________________________________________ -->
2707 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2708 Instruction</a> </div>
2709 <div class="doc_text">
2711 <pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
2713 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2715 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2716 address from which to load. The pointer must point to a <a
2717 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2718 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2719 the number or order of execution of this <tt>load</tt> with other
2720 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2723 <p>The location of memory pointed to is loaded.</p>
2725 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2727 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2728 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2731 <!-- _______________________________________________________________________ -->
2732 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2733 Instruction</a> </div>
2734 <div class="doc_text">
2736 <pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2737 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2740 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2742 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2743 to store and an address at which to store it. The type of the '<tt><pointer></tt>'
2744 operand must be a pointer to the type of the '<tt><value></tt>'
2745 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2746 optimizer is not allowed to modify the number or order of execution of
2747 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2748 href="#i_store">store</a></tt> instructions.</p>
2750 <p>The contents of memory are updated to contain '<tt><value></tt>'
2751 at the location specified by the '<tt><pointer></tt>' operand.</p>
2753 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2755 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2756 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2760 <!-- _______________________________________________________________________ -->
2761 <div class="doc_subsubsection">
2762 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2765 <div class="doc_text">
2768 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2774 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2775 subelement of an aggregate data structure.</p>
2779 <p>This instruction takes a list of integer operands that indicate what
2780 elements of the aggregate object to index to. The actual types of the arguments
2781 provided depend on the type of the first pointer argument. The
2782 '<tt>getelementptr</tt>' instruction is used to index down through the type
2783 levels of a structure or to a specific index in an array. When indexing into a
2784 structure, only <tt>i32</tt> integer constants are allowed. When indexing
2785 into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2786 be sign extended to 64-bit values.</p>
2788 <p>For example, let's consider a C code fragment and how it gets
2789 compiled to LLVM:</p>
2791 <div class="doc_code">
2804 int *foo(struct ST *s) {
2805 return &s[1].Z.B[5][13];
2810 <p>The LLVM code generated by the GCC frontend is:</p>
2812 <div class="doc_code">
2814 %RT = type { i8 , [10 x [20 x i32]], i8 }
2815 %ST = type { i32, double, %RT }
2817 define i32* %foo(%ST* %s) {
2819 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
2827 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2828 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2829 and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
2830 <a href="#t_integer">integer</a> type but the value will always be sign extended
2831 to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
2832 <b>constants</b>.</p>
2834 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2835 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
2836 }</tt>' type, a structure. The second index indexes into the third element of
2837 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
2838 i8 }</tt>' type, another structure. The third index indexes into the second
2839 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
2840 array. The two dimensions of the array are subscripted into, yielding an
2841 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2842 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
2844 <p>Note that it is perfectly legal to index partially through a
2845 structure, returning a pointer to an inner element. Because of this,
2846 the LLVM code for the given testcase is equivalent to:</p>
2849 define i32* %foo(%ST* %s) {
2850 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
2851 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
2852 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
2853 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
2854 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
2859 <p>Note that it is undefined to access an array out of bounds: array and
2860 pointer indexes must always be within the defined bounds of the array type.
2861 The one exception for this rules is zero length arrays. These arrays are
2862 defined to be accessible as variable length arrays, which requires access
2863 beyond the zero'th element.</p>
2865 <p>The getelementptr instruction is often confusing. For some more insight
2866 into how it works, see <a href="GetElementPtr.html">the getelementptr
2872 <i>; yields [12 x i8]*:aptr</i>
2873 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
2877 <!-- ======================================================================= -->
2878 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2880 <div class="doc_text">
2881 <p>The instructions in this category are the conversion instructions (casting)
2882 which all take a single operand and a type. They perform various bit conversions
2886 <!-- _______________________________________________________________________ -->
2887 <div class="doc_subsubsection">
2888 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2890 <div class="doc_text">
2894 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
2899 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2904 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2905 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2906 and type of the result, which must be an <a href="#t_integer">integer</a>
2907 type. The bit size of <tt>value</tt> must be larger than the bit size of
2908 <tt>ty2</tt>. Equal sized types are not allowed.</p>
2912 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
2913 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
2914 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
2915 It will always truncate bits.</p>
2919 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
2920 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
2921 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
2925 <!-- _______________________________________________________________________ -->
2926 <div class="doc_subsubsection">
2927 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
2929 <div class="doc_text">
2933 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
2937 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
2942 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
2943 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
2944 also be of <a href="#t_integer">integer</a> type. The bit size of the
2945 <tt>value</tt> must be smaller than the bit size of the destination type,
2949 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
2950 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
2952 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
2956 %X = zext i32 257 to i64 <i>; yields i64:257</i>
2957 %Y = zext i1 true to i32 <i>; yields i32:1</i>
2961 <!-- _______________________________________________________________________ -->
2962 <div class="doc_subsubsection">
2963 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
2965 <div class="doc_text">
2969 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
2973 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
2977 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
2978 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
2979 also be of <a href="#t_integer">integer</a> type. The bit size of the
2980 <tt>value</tt> must be smaller than the bit size of the destination type,
2985 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
2986 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
2987 the type <tt>ty2</tt>.</p>
2989 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
2993 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
2994 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
2998 <!-- _______________________________________________________________________ -->
2999 <div class="doc_subsubsection">
3000 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3003 <div class="doc_text">
3008 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
3012 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3017 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3018 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3019 cast it to. The size of <tt>value</tt> must be larger than the size of
3020 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3021 <i>no-op cast</i>.</p>
3024 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3025 <a href="#t_floating">floating point</a> type to a smaller
3026 <a href="#t_floating">floating point</a> type. If the value cannot fit within
3027 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3031 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3032 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3036 <!-- _______________________________________________________________________ -->
3037 <div class="doc_subsubsection">
3038 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3040 <div class="doc_text">
3044 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
3048 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3049 floating point value.</p>
3052 <p>The '<tt>fpext</tt>' instruction takes a
3053 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3054 and a <a href="#t_floating">floating point</a> type to cast it to. The source
3055 type must be smaller than the destination type.</p>
3058 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3059 <a href="#t_floating">floating point</a> type to a larger
3060 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3061 used to make a <i>no-op cast</i> because it always changes bits. Use
3062 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3066 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3067 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3071 <!-- _______________________________________________________________________ -->
3072 <div class="doc_subsubsection">
3073 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3075 <div class="doc_text">
3079 <result> = fp2uint <ty> <value> to <ty2> <i>; yields ty2</i>
3083 <p>The '<tt>fp2uint</tt>' converts a floating point <tt>value</tt> to its
3084 unsigned integer equivalent of type <tt>ty2</tt>.
3088 <p>The '<tt>fp2uint</tt>' instruction takes a value to cast, which must be a
3089 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
3090 must be an <a href="#t_integer">integer</a> type.</p>
3093 <p> The '<tt>fp2uint</tt>' instruction converts its
3094 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3095 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3096 the results are undefined.</p>
3098 <p>When converting to i1, the conversion is done as a comparison against
3099 zero. If the <tt>value</tt> was zero, the i1 result will be <tt>false</tt>.
3100 If the <tt>value</tt> was non-zero, the i1 result will be <tt>true</tt>.</p>
3104 %X = fp2uint double 123.0 to i32 <i>; yields i32:123</i>
3105 %Y = fp2uint float 1.0E+300 to i1 <i>; yields i1:true</i>
3106 %X = fp2uint float 1.04E+17 to i8 <i>; yields undefined:1</i>
3110 <!-- _______________________________________________________________________ -->
3111 <div class="doc_subsubsection">
3112 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3114 <div class="doc_text">
3118 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
3122 <p>The '<tt>fptosi</tt>' instruction converts
3123 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3128 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
3129 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
3130 must also be an <a href="#t_integer">integer</a> type.</p>
3133 <p>The '<tt>fptosi</tt>' instruction converts its
3134 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3135 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3136 the results are undefined.</p>
3138 <p>When converting to i1, the conversion is done as a comparison against
3139 zero. If the <tt>value</tt> was zero, the i1 result will be <tt>false</tt>.
3140 If the <tt>value</tt> was non-zero, the i1 result will be <tt>true</tt>.</p>
3144 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
3145 %Y = fptosi float 1.0E-247 to i1 <i>; yields i1:true</i>
3146 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3150 <!-- _______________________________________________________________________ -->
3151 <div class="doc_subsubsection">
3152 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3154 <div class="doc_text">
3158 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3162 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3163 integer and converts that value to the <tt>ty2</tt> type.</p>
3167 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be an
3168 <a href="#t_integer">integer</a> value, and a type to cast it to, which must
3169 be a <a href="#t_floating">floating point</a> type.</p>
3172 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3173 integer quantity and converts it to the corresponding floating point value. If
3174 the value cannot fit in the floating point value, the results are undefined.</p>
3179 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3180 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3184 <!-- _______________________________________________________________________ -->
3185 <div class="doc_subsubsection">
3186 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3188 <div class="doc_text">
3192 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3196 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3197 integer and converts that value to the <tt>ty2</tt> type.</p>
3200 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be an
3201 <a href="#t_integer">integer</a> value, and a type to cast it to, which must be
3202 a <a href="#t_floating">floating point</a> type.</p>
3205 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3206 integer quantity and converts it to the corresponding floating point value. If
3207 the value cannot fit in the floating point value, the results are undefined.</p>
3211 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3212 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3216 <!-- _______________________________________________________________________ -->
3217 <div class="doc_subsubsection">
3218 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3220 <div class="doc_text">
3224 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
3228 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3229 the integer type <tt>ty2</tt>.</p>
3232 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3233 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3234 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3237 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3238 <tt>ty2</tt> by interpreting the pointer value as an integer and either
3239 truncating or zero extending that value to the size of the integer type. If
3240 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3241 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3242 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3247 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3248 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3252 <!-- _______________________________________________________________________ -->
3253 <div class="doc_subsubsection">
3254 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3256 <div class="doc_text">
3260 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
3264 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3265 a pointer type, <tt>ty2</tt>.</p>
3268 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3269 value to cast, and a type to cast it to, which must be a
3270 <a href="#t_pointer">pointer</a> type.
3273 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3274 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
3275 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3276 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3277 the size of a pointer then a zero extension is done. If they are the same size,
3278 nothing is done (<i>no-op cast</i>).</p>
3282 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3283 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3284 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3288 <!-- _______________________________________________________________________ -->
3289 <div class="doc_subsubsection">
3290 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3292 <div class="doc_text">
3296 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3300 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3301 <tt>ty2</tt> without changing any bits.</p>
3304 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3305 a first class value, and a type to cast it to, which must also be a <a
3306 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3307 and the destination type, <tt>ty2</tt>, must be identical. If the source
3308 type is a pointer, the destination type must also be a pointer.</p>
3311 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3312 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3313 this conversion. The conversion is done as if the <tt>value</tt> had been
3314 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3315 converted to other pointer types with this instruction. To convert pointers to
3316 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3317 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3321 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3322 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3323 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3327 <!-- ======================================================================= -->
3328 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3329 <div class="doc_text">
3330 <p>The instructions in this category are the "miscellaneous"
3331 instructions, which defy better classification.</p>
3334 <!-- _______________________________________________________________________ -->
3335 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3337 <div class="doc_text">
3339 <pre> <result> = icmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
3342 <p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3343 of its two integer operands.</p>
3345 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3346 the condition code indicating the kind of comparison to perform. It is not
3347 a value, just a keyword. The possible condition code are:
3349 <li><tt>eq</tt>: equal</li>
3350 <li><tt>ne</tt>: not equal </li>
3351 <li><tt>ugt</tt>: unsigned greater than</li>
3352 <li><tt>uge</tt>: unsigned greater or equal</li>
3353 <li><tt>ult</tt>: unsigned less than</li>
3354 <li><tt>ule</tt>: unsigned less or equal</li>
3355 <li><tt>sgt</tt>: signed greater than</li>
3356 <li><tt>sge</tt>: signed greater or equal</li>
3357 <li><tt>slt</tt>: signed less than</li>
3358 <li><tt>sle</tt>: signed less or equal</li>
3360 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3361 <a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3363 <p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3364 the condition code given as <tt>cond</tt>. The comparison performed always
3365 yields a <a href="#t_primitive">i1</a> result, as follows:
3367 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3368 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3370 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3371 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3372 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3373 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3374 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3375 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3376 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3377 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3378 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3379 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3380 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3381 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3382 <li><tt>sge</tt>: interprets the operands as signed values and yields
3383 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3384 <li><tt>slt</tt>: interprets the operands as signed values and yields
3385 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3386 <li><tt>sle</tt>: interprets the operands as signed values and yields
3387 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3389 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3390 values are compared as if they were integers.</p>
3393 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
3394 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
3395 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
3396 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3397 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3398 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
3402 <!-- _______________________________________________________________________ -->
3403 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3405 <div class="doc_text">
3407 <pre> <result> = fcmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
3410 <p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3411 of its floating point operands.</p>
3413 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3414 the condition code indicating the kind of comparison to perform. It is not
3415 a value, just a keyword. The possible condition code are:
3417 <li><tt>false</tt>: no comparison, always returns false</li>
3418 <li><tt>oeq</tt>: ordered and equal</li>
3419 <li><tt>ogt</tt>: ordered and greater than </li>
3420 <li><tt>oge</tt>: ordered and greater than or equal</li>
3421 <li><tt>olt</tt>: ordered and less than </li>
3422 <li><tt>ole</tt>: ordered and less than or equal</li>
3423 <li><tt>one</tt>: ordered and not equal</li>
3424 <li><tt>ord</tt>: ordered (no nans)</li>
3425 <li><tt>ueq</tt>: unordered or equal</li>
3426 <li><tt>ugt</tt>: unordered or greater than </li>
3427 <li><tt>uge</tt>: unordered or greater than or equal</li>
3428 <li><tt>ult</tt>: unordered or less than </li>
3429 <li><tt>ule</tt>: unordered or less than or equal</li>
3430 <li><tt>une</tt>: unordered or not equal</li>
3431 <li><tt>uno</tt>: unordered (either nans)</li>
3432 <li><tt>true</tt>: no comparison, always returns true</li>
3434 <p><i>Ordered</i> means that neither operand is a QNAN while
3435 <i>unordered</i> means that either operand may be a QNAN.</p>
3436 <p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3437 <a href="#t_floating">floating point</a> typed. They must have identical
3440 <p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3441 the condition code given as <tt>cond</tt>. The comparison performed always
3442 yields a <a href="#t_primitive">i1</a> result, as follows:
3444 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3445 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3446 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3447 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3448 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3449 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3450 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3451 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3452 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3453 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3454 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3455 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3456 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3457 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3458 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3459 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3460 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3461 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3462 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3463 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3464 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3465 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3466 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3467 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3468 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3469 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3470 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3471 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3475 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3476 <result> = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3477 <result> = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3478 <result> = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3482 <!-- _______________________________________________________________________ -->
3483 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3484 Instruction</a> </div>
3485 <div class="doc_text">
3487 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
3489 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
3490 the SSA graph representing the function.</p>
3492 <p>The type of the incoming values is specified with the first type
3493 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3494 as arguments, with one pair for each predecessor basic block of the
3495 current block. Only values of <a href="#t_firstclass">first class</a>
3496 type may be used as the value arguments to the PHI node. Only labels
3497 may be used as the label arguments.</p>
3498 <p>There must be no non-phi instructions between the start of a basic
3499 block and the PHI instructions: i.e. PHI instructions must be first in
3502 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3503 specified by the pair corresponding to the predecessor basic block that executed
3504 just prior to the current block.</p>
3506 <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>
3509 <!-- _______________________________________________________________________ -->
3510 <div class="doc_subsubsection">
3511 <a name="i_select">'<tt>select</tt>' Instruction</a>
3514 <div class="doc_text">
3519 <result> = select i1 <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
3525 The '<tt>select</tt>' instruction is used to choose one value based on a
3526 condition, without branching.
3533 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.
3539 If the boolean condition evaluates to true, the instruction returns the first
3540 value argument; otherwise, it returns the second value argument.
3546 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3551 <!-- _______________________________________________________________________ -->
3552 <div class="doc_subsubsection">
3553 <a name="i_call">'<tt>call</tt>' Instruction</a>
3556 <div class="doc_text">
3560 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
3565 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3569 <p>This instruction requires several arguments:</p>
3573 <p>The optional "tail" marker indicates whether the callee function accesses
3574 any allocas or varargs in the caller. If the "tail" marker is present, the
3575 function call is eligible for tail call optimization. Note that calls may
3576 be marked "tail" even if they do not occur before a <a
3577 href="#i_ret"><tt>ret</tt></a> instruction.
3580 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3581 convention</a> the call should use. If none is specified, the call defaults
3582 to using C calling conventions.
3585 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
3586 being invoked. The argument types must match the types implied by this
3587 signature. This type can be omitted if the function is not varargs and
3588 if the function type does not return a pointer to a function.</p>
3591 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3592 be invoked. In most cases, this is a direct function invocation, but
3593 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3594 to function value.</p>
3597 <p>'<tt>function args</tt>': argument list whose types match the
3598 function signature argument types. All arguments must be of
3599 <a href="#t_firstclass">first class</a> type. If the function signature
3600 indicates the function accepts a variable number of arguments, the extra
3601 arguments can be specified.</p>
3607 <p>The '<tt>call</tt>' instruction is used to cause control flow to
3608 transfer to a specified function, with its incoming arguments bound to
3609 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3610 instruction in the called function, control flow continues with the
3611 instruction after the function call, and the return value of the
3612 function is bound to the result argument. This is a simpler case of
3613 the <a href="#i_invoke">invoke</a> instruction.</p>
3618 %retval = call i32 %test(i32 %argc)
3619 call i32(i8 *, ...) *%printf(i8 * %msg, i32 12, i8 42);
3620 %X = tail call i32 %foo()
3621 %Y = tail call <a href="#callingconv">fastcc</a> i32 %foo()
3626 <!-- _______________________________________________________________________ -->
3627 <div class="doc_subsubsection">
3628 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3631 <div class="doc_text">
3636 <resultval> = va_arg <va_list*> <arglist>, <argty>
3641 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3642 the "variable argument" area of a function call. It is used to implement the
3643 <tt>va_arg</tt> macro in C.</p>
3647 <p>This instruction takes a <tt>va_list*</tt> value and the type of
3648 the argument. It returns a value of the specified argument type and
3649 increments the <tt>va_list</tt> to point to the next argument. The
3650 actual type of <tt>va_list</tt> is target specific.</p>
3654 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3655 type from the specified <tt>va_list</tt> and causes the
3656 <tt>va_list</tt> to point to the next argument. For more information,
3657 see the variable argument handling <a href="#int_varargs">Intrinsic
3660 <p>It is legal for this instruction to be called in a function which does not
3661 take a variable number of arguments, for example, the <tt>vfprintf</tt>
3664 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3665 href="#intrinsics">intrinsic function</a> because it takes a type as an
3670 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3674 <!-- *********************************************************************** -->
3675 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3676 <!-- *********************************************************************** -->
3678 <div class="doc_text">
3680 <p>LLVM supports the notion of an "intrinsic function". These functions have
3681 well known names and semantics and are required to follow certain restrictions.
3682 Overall, these intrinsics represent an extension mechanism for the LLVM
3683 language that does not require changing all of the transformations in LLVM when
3684 adding to the language (or the bytecode reader/writer, the parser, etc...).</p>
3686 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3687 prefix is reserved in LLVM for intrinsic names; thus, function names may not
3688 begin with this prefix. Intrinsic functions must always be external functions:
3689 you cannot define the body of intrinsic functions. Intrinsic functions may
3690 only be used in call or invoke instructions: it is illegal to take the address
3691 of an intrinsic function. Additionally, because intrinsic functions are part
3692 of the LLVM language, it is required if any are added that they be documented
3695 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
3696 a family of functions that perform the same operation but on different data
3697 types. This is most frequent with the integer types. Since LLVM can represent
3698 over 8 million different integer types, there is a way to declare an intrinsic
3699 that can be overloaded based on its arguments. Such an intrinsic will have the
3700 names of its argument types encoded into its function name, each
3701 preceded by a period. For example, the <tt>llvm.ctpop</tt> function can take an
3702 integer of any width. This leads to a family of functions such as
3703 <tt>i32 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i32 @llvm.ctpop.i29(i29 %val)</tt>.
3707 <p>To learn how to add an intrinsic function, please see the
3708 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3713 <!-- ======================================================================= -->
3714 <div class="doc_subsection">
3715 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3718 <div class="doc_text">
3720 <p>Variable argument support is defined in LLVM with the <a
3721 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3722 intrinsic functions. These functions are related to the similarly
3723 named macros defined in the <tt><stdarg.h></tt> header file.</p>
3725 <p>All of these functions operate on arguments that use a
3726 target-specific value type "<tt>va_list</tt>". The LLVM assembly
3727 language reference manual does not define what this type is, so all
3728 transformations should be prepared to handle these functions regardless of
3731 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3732 instruction and the variable argument handling intrinsic functions are
3735 <div class="doc_code">
3737 define i32 @test(i32 %X, ...) {
3738 ; Initialize variable argument processing
3740 %ap2 = bitcast i8** %ap to i8*
3741 call void @llvm.va_start(i8* %ap2)
3743 ; Read a single integer argument
3744 %tmp = va_arg i8** %ap, i32
3746 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3748 %aq2 = bitcast i8** %aq to i8*
3749 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
3750 call void @llvm.va_end(i8* %aq2)
3752 ; Stop processing of arguments.
3753 call void @llvm.va_end(i8* %ap2)
3757 declare void @llvm.va_start(i8*)
3758 declare void @llvm.va_copy(i8*, i8*)
3759 declare void @llvm.va_end(i8*)
3765 <!-- _______________________________________________________________________ -->
3766 <div class="doc_subsubsection">
3767 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3771 <div class="doc_text">
3773 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
3775 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3776 <tt>*<arglist></tt> for subsequent use by <tt><a
3777 href="#i_va_arg">va_arg</a></tt>.</p>
3781 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3785 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3786 macro available in C. In a target-dependent way, it initializes the
3787 <tt>va_list</tt> element to which the argument points, so that the next call to
3788 <tt>va_arg</tt> will produce the first variable argument passed to the function.
3789 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3790 last argument of the function as the compiler can figure that out.</p>
3794 <!-- _______________________________________________________________________ -->
3795 <div class="doc_subsubsection">
3796 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3799 <div class="doc_text">
3801 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
3804 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
3805 which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
3806 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3810 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
3814 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3815 macro available in C. In a target-dependent way, it destroys the
3816 <tt>va_list</tt> element to which the argument points. Calls to <a
3817 href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
3818 <tt>llvm.va_copy</tt></a> must be matched exactly with calls to
3819 <tt>llvm.va_end</tt>.</p>
3823 <!-- _______________________________________________________________________ -->
3824 <div class="doc_subsubsection">
3825 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3828 <div class="doc_text">
3833 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
3838 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
3839 from the source argument list to the destination argument list.</p>
3843 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3844 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3849 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
3850 macro available in C. In a target-dependent way, it copies the source
3851 <tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
3852 intrinsic is necessary because the <tt><a href="#int_va_start">
3853 llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
3854 example, memory allocation.</p>
3858 <!-- ======================================================================= -->
3859 <div class="doc_subsection">
3860 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3863 <div class="doc_text">
3866 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3867 Collection</a> requires the implementation and generation of these intrinsics.
3868 These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
3869 stack</a>, as well as garbage collector implementations that require <a
3870 href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
3871 Front-ends for type-safe garbage collected languages should generate these
3872 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3873 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3877 <!-- _______________________________________________________________________ -->
3878 <div class="doc_subsubsection">
3879 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3882 <div class="doc_text">
3887 declare void @llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
3892 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3893 the code generator, and allows some metadata to be associated with it.</p>
3897 <p>The first argument specifies the address of a stack object that contains the
3898 root pointer. The second pointer (which must be either a constant or a global
3899 value address) contains the meta-data to be associated with the root.</p>
3903 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3904 location. At compile-time, the code generator generates information to allow
3905 the runtime to find the pointer at GC safe points.
3911 <!-- _______________________________________________________________________ -->
3912 <div class="doc_subsubsection">
3913 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
3916 <div class="doc_text">
3921 declare i8 * @llvm.gcread(i8 * %ObjPtr, i8 ** %Ptr)
3926 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
3927 locations, allowing garbage collector implementations that require read
3932 <p>The second argument is the address to read from, which should be an address
3933 allocated from the garbage collector. The first object is a pointer to the
3934 start of the referenced object, if needed by the language runtime (otherwise
3939 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
3940 instruction, but may be replaced with substantially more complex code by the
3941 garbage collector runtime, as needed.</p>
3946 <!-- _______________________________________________________________________ -->
3947 <div class="doc_subsubsection">
3948 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
3951 <div class="doc_text">
3956 declare void @llvm.gcwrite(i8 * %P1, i8 * %Obj, i8 ** %P2)
3961 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
3962 locations, allowing garbage collector implementations that require write
3963 barriers (such as generational or reference counting collectors).</p>
3967 <p>The first argument is the reference to store, the second is the start of the
3968 object to store it to, and the third is the address of the field of Obj to
3969 store to. If the runtime does not require a pointer to the object, Obj may be
3974 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
3975 instruction, but may be replaced with substantially more complex code by the
3976 garbage collector runtime, as needed.</p>
3982 <!-- ======================================================================= -->
3983 <div class="doc_subsection">
3984 <a name="int_codegen">Code Generator Intrinsics</a>
3987 <div class="doc_text">
3989 These intrinsics are provided by LLVM to expose special features that may only
3990 be implemented with code generator support.
3995 <!-- _______________________________________________________________________ -->
3996 <div class="doc_subsubsection">
3997 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4000 <div class="doc_text">
4004 declare i8 *@llvm.returnaddress(i32 <level>)
4010 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4011 target-specific value indicating the return address of the current function
4012 or one of its callers.
4018 The argument to this intrinsic indicates which function to return the address
4019 for. Zero indicates the calling function, one indicates its caller, etc. The
4020 argument is <b>required</b> to be a constant integer value.
4026 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4027 the return address of the specified call frame, or zero if it cannot be
4028 identified. The value returned by this intrinsic is likely to be incorrect or 0
4029 for arguments other than zero, so it should only be used for debugging purposes.
4033 Note that calling this intrinsic does not prevent function inlining or other
4034 aggressive transformations, so the value returned may not be that of the obvious
4035 source-language caller.
4040 <!-- _______________________________________________________________________ -->
4041 <div class="doc_subsubsection">
4042 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4045 <div class="doc_text">
4049 declare i8 *@llvm.frameaddress(i32 <level>)
4055 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4056 target-specific frame pointer value for the specified stack frame.
4062 The argument to this intrinsic indicates which function to return the frame
4063 pointer for. Zero indicates the calling function, one indicates its caller,
4064 etc. The argument is <b>required</b> to be a constant integer value.
4070 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4071 the frame address of the specified call frame, or zero if it cannot be
4072 identified. The value returned by this intrinsic is likely to be incorrect or 0
4073 for arguments other than zero, so it should only be used for debugging purposes.
4077 Note that calling this intrinsic does not prevent function inlining or other
4078 aggressive transformations, so the value returned may not be that of the obvious
4079 source-language caller.
4083 <!-- _______________________________________________________________________ -->
4084 <div class="doc_subsubsection">
4085 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4088 <div class="doc_text">
4092 declare i8 *@llvm.stacksave()
4098 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4099 the function stack, for use with <a href="#int_stackrestore">
4100 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4101 features like scoped automatic variable sized arrays in C99.
4107 This intrinsic returns a opaque pointer value that can be passed to <a
4108 href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4109 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4110 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4111 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4112 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4113 that were allocated after the <tt>llvm.stacksave</tt> was executed.
4118 <!-- _______________________________________________________________________ -->
4119 <div class="doc_subsubsection">
4120 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4123 <div class="doc_text">
4127 declare void @llvm.stackrestore(i8 * %ptr)
4133 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4134 the function stack to the state it was in when the corresponding <a
4135 href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4136 useful for implementing language features like scoped automatic variable sized
4143 See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4149 <!-- _______________________________________________________________________ -->
4150 <div class="doc_subsubsection">
4151 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4154 <div class="doc_text">
4158 declare void @llvm.prefetch(i8 * <address>,
4159 i32 <rw>, i32 <locality>)
4166 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4167 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4169 effect on the behavior of the program but can change its performance
4176 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4177 determining if the fetch should be for a read (0) or write (1), and
4178 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4179 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4180 <tt>locality</tt> arguments must be constant integers.
4186 This intrinsic does not modify the behavior of the program. In particular,
4187 prefetches cannot trap and do not produce a value. On targets that support this
4188 intrinsic, the prefetch can provide hints to the processor cache for better
4194 <!-- _______________________________________________________________________ -->
4195 <div class="doc_subsubsection">
4196 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4199 <div class="doc_text">
4203 declare void @llvm.pcmarker( i32 <id> )
4210 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4212 code to simulators and other tools. The method is target specific, but it is
4213 expected that the marker will use exported symbols to transmit the PC of the marker.
4214 The marker makes no guarantees that it will remain with any specific instruction
4215 after optimizations. It is possible that the presence of a marker will inhibit
4216 optimizations. The intended use is to be inserted after optimizations to allow
4217 correlations of simulation runs.
4223 <tt>id</tt> is a numerical id identifying the marker.
4229 This intrinsic does not modify the behavior of the program. Backends that do not
4230 support this intrinisic may ignore it.
4235 <!-- _______________________________________________________________________ -->
4236 <div class="doc_subsubsection">
4237 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4240 <div class="doc_text">
4244 declare i64 @llvm.readcyclecounter( )
4251 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4252 counter register (or similar low latency, high accuracy clocks) on those targets
4253 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4254 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4255 should only be used for small timings.
4261 When directly supported, reading the cycle counter should not modify any memory.
4262 Implementations are allowed to either return a application specific value or a
4263 system wide value. On backends without support, this is lowered to a constant 0.
4268 <!-- ======================================================================= -->
4269 <div class="doc_subsection">
4270 <a name="int_libc">Standard C Library Intrinsics</a>
4273 <div class="doc_text">
4275 LLVM provides intrinsics for a few important standard C library functions.
4276 These intrinsics allow source-language front-ends to pass information about the
4277 alignment of the pointer arguments to the code generator, providing opportunity
4278 for more efficient code generation.
4283 <!-- _______________________________________________________________________ -->
4284 <div class="doc_subsubsection">
4285 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4288 <div class="doc_text">
4292 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
4293 i32 <len>, i32 <align>)
4294 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
4295 i64 <len>, i32 <align>)
4301 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4302 location to the destination location.
4306 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4307 intrinsics do not return a value, and takes an extra alignment argument.
4313 The first argument is a pointer to the destination, the second is a pointer to
4314 the source. The third argument is an integer argument
4315 specifying the number of bytes to copy, and the fourth argument is the alignment
4316 of the source and destination locations.
4320 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4321 the caller guarantees that both the source and destination pointers are aligned
4328 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4329 location to the destination location, which are not allowed to overlap. It
4330 copies "len" bytes of memory over. If the argument is known to be aligned to
4331 some boundary, this can be specified as the fourth argument, otherwise it should
4337 <!-- _______________________________________________________________________ -->
4338 <div class="doc_subsubsection">
4339 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4342 <div class="doc_text">
4346 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
4347 i32 <len>, i32 <align>)
4348 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
4349 i64 <len>, i32 <align>)
4355 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4356 location to the destination location. It is similar to the
4357 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4361 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4362 intrinsics do not return a value, and takes an extra alignment argument.
4368 The first argument is a pointer to the destination, the second is a pointer to
4369 the source. The third argument is an integer argument
4370 specifying the number of bytes to copy, and the fourth argument is the alignment
4371 of the source and destination locations.
4375 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4376 the caller guarantees that the source and destination pointers are aligned to
4383 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4384 location to the destination location, which may overlap. It
4385 copies "len" bytes of memory over. If the argument is known to be aligned to
4386 some boundary, this can be specified as the fourth argument, otherwise it should
4392 <!-- _______________________________________________________________________ -->
4393 <div class="doc_subsubsection">
4394 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4397 <div class="doc_text">
4401 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
4402 i32 <len>, i32 <align>)
4403 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
4404 i64 <len>, i32 <align>)
4410 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4415 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4416 does not return a value, and takes an extra alignment argument.
4422 The first argument is a pointer to the destination to fill, the second is the
4423 byte value to fill it with, the third argument is an integer
4424 argument specifying the number of bytes to fill, and the fourth argument is the
4425 known alignment of destination location.
4429 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4430 the caller guarantees that the destination pointer is aligned to that boundary.
4436 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4438 destination location. If the argument is known to be aligned to some boundary,
4439 this can be specified as the fourth argument, otherwise it should be set to 0 or
4445 <!-- _______________________________________________________________________ -->
4446 <div class="doc_subsubsection">
4447 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4450 <div class="doc_text">
4454 declare float @llvm.sqrt.f32(float %Val)
4455 declare double @llvm.sqrt.f64(double %Val)
4461 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
4462 returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
4463 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4464 negative numbers (which allows for better optimization).
4470 The argument and return value are floating point numbers of the same type.
4476 This function returns the sqrt of the specified operand if it is a positive
4477 floating point number.
4481 <!-- _______________________________________________________________________ -->
4482 <div class="doc_subsubsection">
4483 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4486 <div class="doc_text">
4490 declare float @llvm.powi.f32(float %Val, i32 %power)
4491 declare double @llvm.powi.f64(double %Val, i32 %power)
4497 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4498 specified (positive or negative) power. The order of evaluation of
4499 multiplications is not defined.
4505 The second argument is an integer power, and the first is a value to raise to
4512 This function returns the first value raised to the second power with an
4513 unspecified sequence of rounding operations.</p>
4517 <!-- ======================================================================= -->
4518 <div class="doc_subsection">
4519 <a name="int_manip">Bit Manipulation Intrinsics</a>
4522 <div class="doc_text">
4524 LLVM provides intrinsics for a few important bit manipulation operations.
4525 These allow efficient code generation for some algorithms.
4530 <!-- _______________________________________________________________________ -->
4531 <div class="doc_subsubsection">
4532 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4535 <div class="doc_text">
4538 <p>This is an overloaded intrinsic function. You can use bswap on any integer
4539 type that is an even number of bytes (i.e. BitWidth % 16 == 0). Note the suffix
4540 that includes the type for the result and the operand.
4542 declare i16 @llvm.bswap.i16.i16(i16 <id>)
4543 declare i32 @llvm.bswap.i32.i32(i32 <id>)
4544 declare i64 @llvm.bswap.i64.i64(i64 <id>)
4550 The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
4551 values with an even number of bytes (positive multiple of 16 bits). These are
4552 useful for performing operations on data that is not in the target's native
4559 The <tt>llvm.bswap.16.i16</tt> intrinsic returns an i16 value that has the high
4560 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4561 intrinsic returns an i32 value that has the four bytes of the input i32
4562 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
4563 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48.i48</tt>,
4564 <tt>llvm.bswap.i64.i64</tt> and other intrinsics extend this concept to
4565 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4570 <!-- _______________________________________________________________________ -->
4571 <div class="doc_subsubsection">
4572 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4575 <div class="doc_text">
4578 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4579 width. Not all targets support all bit widths however.
4581 declare i32 @llvm.ctpop.i8 (i8 <src>)
4582 declare i32 @llvm.ctpop.i16(i16 <src>)
4583 declare i32 @llvm.ctpop.i32(i32 <src>)
4584 declare i32 @llvm.ctpop.i64(i64 <src>)
4585 declare i32 @llvm.ctpop.i256(i256 <src>)
4591 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4598 The only argument is the value to be counted. The argument may be of any
4599 integer type. The return type must match the argument type.
4605 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4609 <!-- _______________________________________________________________________ -->
4610 <div class="doc_subsubsection">
4611 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4614 <div class="doc_text">
4617 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4618 integer bit width. Not all targets support all bit widths however.
4620 declare i32 @llvm.ctlz.i8 (i8 <src>)
4621 declare i32 @llvm.ctlz.i16(i16 <src>)
4622 declare i32 @llvm.ctlz.i32(i32 <src>)
4623 declare i32 @llvm.ctlz.i64(i64 <src>)
4624 declare i32 @llvm.ctlz.i256(i256 <src>)
4630 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4631 leading zeros in a variable.
4637 The only argument is the value to be counted. The argument may be of any
4638 integer type. The return type must match the argument type.
4644 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4645 in a variable. If the src == 0 then the result is the size in bits of the type
4646 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4652 <!-- _______________________________________________________________________ -->
4653 <div class="doc_subsubsection">
4654 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4657 <div class="doc_text">
4660 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
4661 integer bit width. Not all targets support all bit widths however.
4663 declare i32 @llvm.cttz.i8 (i8 <src>)
4664 declare i32 @llvm.cttz.i16(i16 <src>)
4665 declare i32 @llvm.cttz.i32(i32 <src>)
4666 declare i32 @llvm.cttz.i64(i64 <src>)
4667 declare i32 @llvm.cttz.i256(i256 <src>)
4673 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4680 The only argument is the value to be counted. The argument may be of any
4681 integer type. The return type must match the argument type.
4687 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4688 in a variable. If the src == 0 then the result is the size in bits of the type
4689 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4693 <!-- _______________________________________________________________________ -->
4694 <div class="doc_subsubsection">
4695 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
4698 <div class="doc_text">
4701 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
4702 on any integer bit width.
4704 declare i17 @llvm.part.select.i17.i17 (i17 %val, i32 %loBit, i32 %hiBit)
4705 declare i29 @llvm.part.select.i29.i29 (i29 %val, i32 %loBit, i32 %hiBit)
4709 <p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
4710 range of bits from an integer value and returns them in the same bit width as
4711 the original value.</p>
4714 <p>The first argument, <tt>%val</tt> and the result may be integer types of
4715 any bit width but they must have the same bit width. The second and third
4716 arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
4719 <p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
4720 of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
4721 <tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
4722 operates in forward mode.</p>
4723 <p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
4724 right by <tt>%loBit</tt> bits and then ANDing it with a mask with
4725 only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
4727 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
4728 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
4729 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
4730 to determine the number of bits to retain.</li>
4731 <li>A mask of the retained bits is created by shifting a -1 value.</li>
4732 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
4734 <p>In reverse mode, a similar computation is made except that the bits are
4735 returned in the reverse order. So, for example, if <tt>X</tt> has the value
4736 <tt>i16 0x0ACF (101011001111)</tt> and we apply
4737 <tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
4738 <tt>i16 0x0026 (000000100110)</tt>.</p>
4741 <div class="doc_subsubsection">
4742 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
4745 <div class="doc_text">
4748 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
4749 on any integer bit width.
4751 declare i17 @llvm.part.set.i17.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
4752 declare i29 @llvm.part.set.i29.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
4756 <p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
4757 of bits in an integer value with another integer value. It returns the integer
4758 with the replaced bits.</p>
4761 <p>The first argument, <tt>%val</tt> and the result may be integer types of
4762 any bit width but they must have the same bit width. <tt>%val</tt> is the value
4763 whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
4764 integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
4765 type since they specify only a bit index.</p>
4768 <p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
4769 of operation: forwards and reverse. If <tt>%lo</tt> is greater than
4770 <tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
4771 operates in forward mode.</p>
4772 <p>For both modes, the <tt>%repl</tt> value is prepared for use by either
4773 truncating it down to the size of the replacement area or zero extending it
4774 up to that size.</p>
4775 <p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
4776 are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
4777 in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
4778 to the <tt>%hi</tt>th bit.
4779 <p>In reverse mode, a similar computation is made except that the bits are
4780 reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
4781 <tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
4784 llvm.part.set(0xFFFF, 0, 4, 7) -> 0xFF0F
4785 llvm.part.set(0xFFFF, 0, 7, 4) -> 0xFF0F
4786 llvm.part.set(0xFFFF, 1, 7, 4) -> 0xFF8F
4787 llvm.part.set(0xFFFF, F, 8, 3) -> 0xFFE7
4788 llvm.part.set(0xFFFF, 0, 3, 8) -> 0xFE07
4792 <!-- ======================================================================= -->
4793 <div class="doc_subsection">
4794 <a name="int_debugger">Debugger Intrinsics</a>
4797 <div class="doc_text">
4799 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
4800 are described in the <a
4801 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
4802 Debugging</a> document.
4807 <!-- ======================================================================= -->
4808 <div class="doc_subsection">
4809 <a name="int_eh">Exception Handling Intrinsics</a>
4812 <div class="doc_text">
4813 <p> The LLVM exception handling intrinsics (which all start with
4814 <tt>llvm.eh.</tt> prefix), are described in the <a
4815 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
4816 Handling</a> document. </p>
4820 <!-- *********************************************************************** -->
4823 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
4824 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
4825 <a href="http://validator.w3.org/check/referer"><img
4826 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
4828 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
4829 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
4830 Last modified: $Date$