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>
194 <li><a href="#int_general">General intrinsics</a></li>
196 <li><a href="#int_var_annotation">'<tt>llvm.var.annotation</tt>'
204 <div class="doc_author">
205 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
206 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
209 <!-- *********************************************************************** -->
210 <div class="doc_section"> <a name="abstract">Abstract </a></div>
211 <!-- *********************************************************************** -->
213 <div class="doc_text">
214 <p>This document is a reference manual for the LLVM assembly language.
215 LLVM is an SSA based representation that provides type safety,
216 low-level operations, flexibility, and the capability of representing
217 'all' high-level languages cleanly. It is the common code
218 representation used throughout all phases of the LLVM compilation
222 <!-- *********************************************************************** -->
223 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
224 <!-- *********************************************************************** -->
226 <div class="doc_text">
228 <p>The LLVM code representation is designed to be used in three
229 different forms: as an in-memory compiler IR, as an on-disk bitcode
230 representation (suitable for fast loading by a Just-In-Time compiler),
231 and as a human readable assembly language representation. This allows
232 LLVM to provide a powerful intermediate representation for efficient
233 compiler transformations and analysis, while providing a natural means
234 to debug and visualize the transformations. The three different forms
235 of LLVM are all equivalent. This document describes the human readable
236 representation and notation.</p>
238 <p>The LLVM representation aims to be light-weight and low-level
239 while being expressive, typed, and extensible at the same time. It
240 aims to be a "universal IR" of sorts, by being at a low enough level
241 that high-level ideas may be cleanly mapped to it (similar to how
242 microprocessors are "universal IR's", allowing many source languages to
243 be mapped to them). By providing type information, LLVM can be used as
244 the target of optimizations: for example, through pointer analysis, it
245 can be proven that a C automatic variable is never accessed outside of
246 the current function... allowing it to be promoted to a simple SSA
247 value instead of a memory location.</p>
251 <!-- _______________________________________________________________________ -->
252 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
254 <div class="doc_text">
256 <p>It is important to note that this document describes 'well formed'
257 LLVM assembly language. There is a difference between what the parser
258 accepts and what is considered 'well formed'. For example, the
259 following instruction is syntactically okay, but not well formed:</p>
261 <div class="doc_code">
263 %x = <a href="#i_add">add</a> i32 1, %x
267 <p>...because the definition of <tt>%x</tt> does not dominate all of
268 its uses. The LLVM infrastructure provides a verification pass that may
269 be used to verify that an LLVM module is well formed. This pass is
270 automatically run by the parser after parsing input assembly and by
271 the optimizer before it outputs bitcode. The violations pointed out
272 by the verifier pass indicate bugs in transformation passes or input to
276 <!-- Describe the typesetting conventions here. --> </div>
278 <!-- *********************************************************************** -->
279 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
280 <!-- *********************************************************************** -->
282 <div class="doc_text">
284 <p>LLVM uses three different forms of identifiers, for different
288 <li>Named values are represented as a string of characters with a '%' prefix.
289 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
290 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
291 Identifiers which require other characters in their names can be surrounded
292 with quotes. In this way, anything except a <tt>"</tt> character can be used
295 <li>Unnamed values are represented as an unsigned numeric value with a '%'
296 prefix. For example, %12, %2, %44.</li>
298 <li>Constants, which are described in a <a href="#constants">section about
299 constants</a>, below.</li>
302 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
303 don't need to worry about name clashes with reserved words, and the set of
304 reserved words may be expanded in the future without penalty. Additionally,
305 unnamed identifiers allow a compiler to quickly come up with a temporary
306 variable without having to avoid symbol table conflicts.</p>
308 <p>Reserved words in LLVM are very similar to reserved words in other
309 languages. There are keywords for different opcodes
310 ('<tt><a href="#i_add">add</a></tt>',
311 '<tt><a href="#i_bitcast">bitcast</a></tt>',
312 '<tt><a href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
313 href="#t_void">void</a></tt>', '<tt><a href="#t_primitive">i32</a></tt>', etc...),
314 and others. These reserved words cannot conflict with variable names, because
315 none of them start with a '%' character.</p>
317 <p>Here is an example of LLVM code to multiply the integer variable
318 '<tt>%X</tt>' by 8:</p>
322 <div class="doc_code">
324 %result = <a href="#i_mul">mul</a> i32 %X, 8
328 <p>After strength reduction:</p>
330 <div class="doc_code">
332 %result = <a href="#i_shl">shl</a> i32 %X, i8 3
336 <p>And the hard way:</p>
338 <div class="doc_code">
340 <a href="#i_add">add</a> i32 %X, %X <i>; yields {i32}:%0</i>
341 <a href="#i_add">add</a> i32 %0, %0 <i>; yields {i32}:%1</i>
342 %result = <a href="#i_add">add</a> i32 %1, %1
346 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
347 important lexical features of LLVM:</p>
351 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
354 <li>Unnamed temporaries are created when the result of a computation is not
355 assigned to a named value.</li>
357 <li>Unnamed temporaries are numbered sequentially</li>
361 <p>...and it also shows a convention that we follow in this document. When
362 demonstrating instructions, we will follow an instruction with a comment that
363 defines the type and name of value produced. Comments are shown in italic
368 <!-- *********************************************************************** -->
369 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
370 <!-- *********************************************************************** -->
372 <!-- ======================================================================= -->
373 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
376 <div class="doc_text">
378 <p>LLVM programs are composed of "Module"s, each of which is a
379 translation unit of the input programs. Each module consists of
380 functions, global variables, and symbol table entries. Modules may be
381 combined together with the LLVM linker, which merges function (and
382 global variable) definitions, resolves forward declarations, and merges
383 symbol table entries. Here is an example of the "hello world" module:</p>
385 <div class="doc_code">
386 <pre><i>; Declare the string constant as a global constant...</i>
387 <a href="#identifiers">@.LC0</a> = <a href="#linkage_internal">internal</a> <a
388 href="#globalvars">constant</a> <a href="#t_array">[13 x i8]</a> c"hello world\0A\00" <i>; [13 x i8]*</i>
390 <i>; External declaration of the puts function</i>
391 <a href="#functionstructure">declare</a> i32 @puts(i8 *) <i>; i32(i8 *)* </i>
393 <i>; Definition of main function</i>
394 define i32 @main() { <i>; i32()* </i>
395 <i>; Convert [13x i8 ]* to i8 *...</i>
397 href="#i_getelementptr">getelementptr</a> [13 x i8 ]* @.LC0, i64 0, i64 0 <i>; i8 *</i>
399 <i>; Call puts function to write out the string to stdout...</i>
401 href="#i_call">call</a> i32 @puts(i8 * %cast210) <i>; i32</i>
403 href="#i_ret">ret</a> i32 0<br>}<br>
407 <p>This example is made up of a <a href="#globalvars">global variable</a>
408 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
409 function, and a <a href="#functionstructure">function definition</a>
410 for "<tt>main</tt>".</p>
412 <p>In general, a module is made up of a list of global values,
413 where both functions and global variables are global values. Global values are
414 represented by a pointer to a memory location (in this case, a pointer to an
415 array of char, and a pointer to a function), and have one of the following <a
416 href="#linkage">linkage types</a>.</p>
420 <!-- ======================================================================= -->
421 <div class="doc_subsection">
422 <a name="linkage">Linkage Types</a>
425 <div class="doc_text">
428 All Global Variables and Functions have one of the following types of linkage:
433 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
435 <dd>Global values with internal linkage are only directly accessible by
436 objects in the current module. In particular, linking code into a module with
437 an internal global value may cause the internal to be renamed as necessary to
438 avoid collisions. Because the symbol is internal to the module, all
439 references can be updated. This corresponds to the notion of the
440 '<tt>static</tt>' keyword in C.
443 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
445 <dd>Globals with "<tt>linkonce</tt>" linkage are merged with other globals of
446 the same name when linkage occurs. This is typically used to implement
447 inline functions, templates, or other code which must be generated in each
448 translation unit that uses it. Unreferenced <tt>linkonce</tt> globals are
449 allowed to be discarded.
452 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
454 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
455 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
456 used for globals that may be emitted in multiple translation units, but that
457 are not guaranteed to be emitted into every translation unit that uses them.
458 One example of this are common globals in C, such as "<tt>int X;</tt>" at
462 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
464 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
465 pointer to array type. When two global variables with appending linkage are
466 linked together, the two global arrays are appended together. This is the
467 LLVM, typesafe, equivalent of having the system linker append together
468 "sections" with identical names when .o files are linked.
471 <dt><tt><b><a name="linkage_externweak">extern_weak</a></b></tt>: </dt>
472 <dd>The semantics of this linkage follow the ELF model: the symbol is weak
473 until linked, if not linked, the symbol becomes null instead of being an
477 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
479 <dd>If none of the above identifiers are used, the global is externally
480 visible, meaning that it participates in linkage and can be used to resolve
481 external symbol references.
486 The next two types of linkage are targeted for Microsoft Windows platform
487 only. They are designed to support importing (exporting) symbols from (to)
492 <dt><tt><b><a name="linkage_dllimport">dllimport</a></b></tt>: </dt>
494 <dd>"<tt>dllimport</tt>" linkage causes the compiler to reference a function
495 or variable via a global pointer to a pointer that is set up by the DLL
496 exporting the symbol. On Microsoft Windows targets, the pointer name is
497 formed by combining <code>_imp__</code> and the function or variable name.
500 <dt><tt><b><a name="linkage_dllexport">dllexport</a></b></tt>: </dt>
502 <dd>"<tt>dllexport</tt>" linkage causes the compiler to provide a global
503 pointer to a pointer in a DLL, so that it can be referenced with the
504 <tt>dllimport</tt> attribute. On Microsoft Windows targets, the pointer
505 name is formed by combining <code>_imp__</code> and the function or variable
511 <p><a name="linkage_external"></a>For example, since the "<tt>.LC0</tt>"
512 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
513 variable and was linked with this one, one of the two would be renamed,
514 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
515 external (i.e., lacking any linkage declarations), they are accessible
516 outside of the current module.</p>
517 <p>It is illegal for a function <i>declaration</i>
518 to have any linkage type other than "externally visible", <tt>dllimport</tt>,
519 or <tt>extern_weak</tt>.</p>
520 <p>Aliases can have only <tt>external</tt>, <tt>internal</tt> and <tt>weak</tt>
524 <!-- ======================================================================= -->
525 <div class="doc_subsection">
526 <a name="callingconv">Calling Conventions</a>
529 <div class="doc_text">
531 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
532 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
533 specified for the call. The calling convention of any pair of dynamic
534 caller/callee must match, or the behavior of the program is undefined. The
535 following calling conventions are supported by LLVM, and more may be added in
539 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
541 <dd>This calling convention (the default if no other calling convention is
542 specified) matches the target C calling conventions. This calling convention
543 supports varargs function calls and tolerates some mismatch in the declared
544 prototype and implemented declaration of the function (as does normal C).
547 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
549 <dd>This calling convention attempts to make calls as fast as possible
550 (e.g. by passing things in registers). This calling convention allows the
551 target to use whatever tricks it wants to produce fast code for the target,
552 without having to conform to an externally specified ABI. Implementations of
553 this convention should allow arbitrary tail call optimization to be supported.
554 This calling convention does not support varargs and requires the prototype of
555 all callees to exactly match the prototype of the function definition.
558 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
560 <dd>This calling convention attempts to make code in the caller as efficient
561 as possible under the assumption that the call is not commonly executed. As
562 such, these calls often preserve all registers so that the call does not break
563 any live ranges in the caller side. This calling convention does not support
564 varargs and requires the prototype of all callees to exactly match the
565 prototype of the function definition.
568 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
570 <dd>Any calling convention may be specified by number, allowing
571 target-specific calling conventions to be used. Target specific calling
572 conventions start at 64.
576 <p>More calling conventions can be added/defined on an as-needed basis, to
577 support pascal conventions or any other well-known target-independent
582 <!-- ======================================================================= -->
583 <div class="doc_subsection">
584 <a name="visibility">Visibility Styles</a>
587 <div class="doc_text">
590 All Global Variables and Functions have one of the following visibility styles:
594 <dt><b>"<tt>default</tt>" - Default style</b>:</dt>
596 <dd>On ELF, default visibility means that the declaration is visible to other
597 modules and, in shared libraries, means that the declared entity may be
598 overridden. On Darwin, default visibility means that the declaration is
599 visible to other modules. Default visibility corresponds to "external
600 linkage" in the language.
603 <dt><b>"<tt>hidden</tt>" - Hidden style</b>:</dt>
605 <dd>Two declarations of an object with hidden visibility refer to the same
606 object if they are in the same shared object. Usually, hidden visibility
607 indicates that the symbol will not be placed into the dynamic symbol table,
608 so no other module (executable or shared library) can reference it
612 <dt><b>"<tt>protected</tt>" - Protected style</b>:</dt>
614 <dd>On ELF, protected visibility indicates that the symbol will be placed in
615 the dynamic symbol table, but that references within the defining module will
616 bind to the local symbol. That is, the symbol cannot be overridden by another
623 <!-- ======================================================================= -->
624 <div class="doc_subsection">
625 <a name="globalvars">Global Variables</a>
628 <div class="doc_text">
630 <p>Global variables define regions of memory allocated at compilation time
631 instead of run-time. Global variables may optionally be initialized, may have
632 an explicit section to be placed in, and may have an optional explicit alignment
633 specified. A variable may be defined as "thread_local", which means that it
634 will not be shared by threads (each thread will have a separated copy of the
635 variable). A variable may be defined as a global "constant," which indicates
636 that the contents of the variable will <b>never</b> be modified (enabling better
637 optimization, allowing the global data to be placed in the read-only section of
638 an executable, etc). Note that variables that need runtime initialization
639 cannot be marked "constant" as there is a store to the variable.</p>
642 LLVM explicitly allows <em>declarations</em> of global variables to be marked
643 constant, even if the final definition of the global is not. This capability
644 can be used to enable slightly better optimization of the program, but requires
645 the language definition to guarantee that optimizations based on the
646 'constantness' are valid for the translation units that do not include the
650 <p>As SSA values, global variables define pointer values that are in
651 scope (i.e. they dominate) all basic blocks in the program. Global
652 variables always define a pointer to their "content" type because they
653 describe a region of memory, and all memory objects in LLVM are
654 accessed through pointers.</p>
656 <p>LLVM allows an explicit section to be specified for globals. If the target
657 supports it, it will emit globals to the section specified.</p>
659 <p>An explicit alignment may be specified for a global. If not present, or if
660 the alignment is set to zero, the alignment of the global is set by the target
661 to whatever it feels convenient. If an explicit alignment is specified, the
662 global is forced to have at least that much alignment. All alignments must be
665 <p>For example, the following defines a global with an initializer, section,
668 <div class="doc_code">
670 @G = constant float 1.0, section "foo", align 4
677 <!-- ======================================================================= -->
678 <div class="doc_subsection">
679 <a name="functionstructure">Functions</a>
682 <div class="doc_text">
684 <p>LLVM function definitions consist of the "<tt>define</tt>" keyord,
685 an optional <a href="#linkage">linkage type</a>, an optional
686 <a href="#visibility">visibility style</a>, an optional
687 <a href="#callingconv">calling convention</a>, a return type, an optional
688 <a href="#paramattrs">parameter attribute</a> for the return type, a function
689 name, a (possibly empty) argument list (each with optional
690 <a href="#paramattrs">parameter attributes</a>), an optional section, an
691 optional alignment, an opening curly brace, a list of basic blocks, and a
694 LLVM function declarations consist of the "<tt>declare</tt>" keyword, an
695 optional <a href="#linkage">linkage type</a>, an optional
696 <a href="#visibility">visibility style</a>, an optional
697 <a href="#callingconv">calling convention</a>, a return type, an optional
698 <a href="#paramattrs">parameter attribute</a> for the return type, a function
699 name, a possibly empty list of arguments, and an optional alignment.</p>
701 <p>A function definition contains a list of basic blocks, forming the CFG for
702 the function. Each basic block may optionally start with a label (giving the
703 basic block a symbol table entry), contains a list of instructions, and ends
704 with a <a href="#terminators">terminator</a> instruction (such as a branch or
705 function return).</p>
707 <p>The first basic block in a function is special in two ways: it is immediately
708 executed on entrance to the function, and it is not allowed to have predecessor
709 basic blocks (i.e. there can not be any branches to the entry block of a
710 function). Because the block can have no predecessors, it also cannot have any
711 <a href="#i_phi">PHI nodes</a>.</p>
713 <p>LLVM allows an explicit section to be specified for functions. If the target
714 supports it, it will emit functions to the section specified.</p>
716 <p>An explicit alignment may be specified for a function. If not present, or if
717 the alignment is set to zero, the alignment of the function is set by the target
718 to whatever it feels convenient. If an explicit alignment is specified, the
719 function is forced to have at least that much alignment. All alignments must be
725 <!-- ======================================================================= -->
726 <div class="doc_subsection">
727 <a name="aliasstructure">Aliases</a>
729 <div class="doc_text">
730 <p>Aliases act as "second name" for the aliasee value (which can be either
731 function or global variable or bitcast of global value). Aliases may have an
732 optional <a href="#linkage">linkage type</a>, and an
733 optional <a href="#visibility">visibility style</a>.</p>
737 <div class="doc_code">
739 @<Name> = [Linkage] [Visibility] alias <AliaseeTy> @<Aliasee>
747 <!-- ======================================================================= -->
748 <div class="doc_subsection"><a name="paramattrs">Parameter Attributes</a></div>
749 <div class="doc_text">
750 <p>The return type and each parameter of a function type may have a set of
751 <i>parameter attributes</i> associated with them. Parameter attributes are
752 used to communicate additional information about the result or parameters of
753 a function. Parameter attributes are considered to be part of the function
754 type so two functions types that differ only by the parameter attributes
755 are different function types.</p>
757 <p>Parameter attributes are simple keywords that follow the type specified. If
758 multiple parameter attributes are needed, they are space separated. For
761 <div class="doc_code">
763 %someFunc = i16 (i8 sext %someParam) zext
764 %someFunc = i16 (i8 zext %someParam) zext
768 <p>Note that the two function types above are unique because the parameter has
769 a different attribute (sext in the first one, zext in the second). Also note
770 that the attribute for the function result (zext) comes immediately after the
773 <p>Currently, only the following parameter attributes are defined:</p>
775 <dt><tt>zext</tt></dt>
776 <dd>This indicates that the parameter should be zero extended just before
777 a call to this function.</dd>
778 <dt><tt>sext</tt></dt>
779 <dd>This indicates that the parameter should be sign extended just before
780 a call to this function.</dd>
781 <dt><tt>inreg</tt></dt>
782 <dd>This indicates that the parameter should be placed in register (if
783 possible) during assembling function call. Support for this attribute is
785 <dt><tt>sret</tt></dt>
786 <dd>This indicates that the parameter specifies the address of a structure
787 that is the return value of the function in the source program.</dd>
788 <dt><tt>noalias</tt></dt>
789 <dd>This indicates that the parameter not alias any other object or any
790 other "noalias" objects during the function call.
791 <dt><tt>noreturn</tt></dt>
792 <dd>This function attribute indicates that the function never returns. This
793 indicates to LLVM that every call to this function should be treated as if
794 an <tt>unreachable</tt> instruction immediately followed the call.</dd>
795 <dt><tt>nounwind</tt></dt>
796 <dd>This function attribute indicates that the function type does not use
797 the unwind instruction and does not allow stack unwinding to propagate
803 <!-- ======================================================================= -->
804 <div class="doc_subsection">
805 <a name="moduleasm">Module-Level Inline Assembly</a>
808 <div class="doc_text">
810 Modules may contain "module-level inline asm" blocks, which corresponds to the
811 GCC "file scope inline asm" blocks. These blocks are internally concatenated by
812 LLVM and treated as a single unit, but may be separated in the .ll file if
813 desired. The syntax is very simple:
816 <div class="doc_code">
818 module asm "inline asm code goes here"
819 module asm "more can go here"
823 <p>The strings can contain any character by escaping non-printable characters.
824 The escape sequence used is simply "\xx" where "xx" is the two digit hex code
829 The inline asm code is simply printed to the machine code .s file when
830 assembly code is generated.
834 <!-- ======================================================================= -->
835 <div class="doc_subsection">
836 <a name="datalayout">Data Layout</a>
839 <div class="doc_text">
840 <p>A module may specify a target specific data layout string that specifies how
841 data is to be laid out in memory. The syntax for the data layout is simply:</p>
842 <pre> target datalayout = "<i>layout specification</i>"</pre>
843 <p>The <i>layout specification</i> consists of a list of specifications
844 separated by the minus sign character ('-'). Each specification starts with a
845 letter and may include other information after the letter to define some
846 aspect of the data layout. The specifications accepted are as follows: </p>
849 <dd>Specifies that the target lays out data in big-endian form. That is, the
850 bits with the most significance have the lowest address location.</dd>
852 <dd>Specifies that hte target lays out data in little-endian form. That is,
853 the bits with the least significance have the lowest address location.</dd>
854 <dt><tt>p:<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
855 <dd>This specifies the <i>size</i> of a pointer and its <i>abi</i> and
856 <i>preferred</i> alignments. All sizes are in bits. Specifying the <i>pref</i>
857 alignment is optional. If omitted, the preceding <tt>:</tt> should be omitted
859 <dt><tt>i<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
860 <dd>This specifies the alignment for an integer type of a given bit
861 <i>size</i>. The value of <i>size</i> must be in the range [1,2^23).</dd>
862 <dt><tt>v<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
863 <dd>This specifies the alignment for a vector type of a given bit
865 <dt><tt>f<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
866 <dd>This specifies the alignment for a floating point type of a given bit
867 <i>size</i>. The value of <i>size</i> must be either 32 (float) or 64
869 <dt><tt>a<i>size</i>:<i>abi</i>:<i>pref</i></tt></dt>
870 <dd>This specifies the alignment for an aggregate type of a given bit
873 <p>When constructing the data layout for a given target, LLVM starts with a
874 default set of specifications which are then (possibly) overriden by the
875 specifications in the <tt>datalayout</tt> keyword. The default specifications
876 are given in this list:</p>
878 <li><tt>E</tt> - big endian</li>
879 <li><tt>p:32:64:64</tt> - 32-bit pointers with 64-bit alignment</li>
880 <li><tt>i1:8:8</tt> - i1 is 8-bit (byte) aligned</li>
881 <li><tt>i8:8:8</tt> - i8 is 8-bit (byte) aligned</li>
882 <li><tt>i16:16:16</tt> - i16 is 16-bit aligned</li>
883 <li><tt>i32:32:32</tt> - i32 is 32-bit aligned</li>
884 <li><tt>i64:32:64</tt> - i64 has abi alignment of 32-bits but preferred
885 alignment of 64-bits</li>
886 <li><tt>f32:32:32</tt> - float is 32-bit aligned</li>
887 <li><tt>f64:64:64</tt> - double is 64-bit aligned</li>
888 <li><tt>v64:64:64</tt> - 64-bit vector is 64-bit aligned</li>
889 <li><tt>v128:128:128</tt> - 128-bit vector is 128-bit aligned</li>
890 <li><tt>a0:0:1</tt> - aggregates are 8-bit aligned</li>
892 <p>When llvm is determining the alignment for a given type, it uses the
895 <li>If the type sought is an exact match for one of the specifications, that
896 specification is used.</li>
897 <li>If no match is found, and the type sought is an integer type, then the
898 smallest integer type that is larger than the bitwidth of the sought type is
899 used. If none of the specifications are larger than the bitwidth then the the
900 largest integer type is used. For example, given the default specifications
901 above, the i7 type will use the alignment of i8 (next largest) while both
902 i65 and i256 will use the alignment of i64 (largest specified).</li>
903 <li>If no match is found, and the type sought is a vector type, then the
904 largest vector type that is smaller than the sought vector type will be used
905 as a fall back. This happens because <128 x double> can be implemented in
906 terms of 64 <2 x double>, for example.</li>
910 <!-- *********************************************************************** -->
911 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
912 <!-- *********************************************************************** -->
914 <div class="doc_text">
916 <p>The LLVM type system is one of the most important features of the
917 intermediate representation. Being typed enables a number of
918 optimizations to be performed on the IR directly, without having to do
919 extra analyses on the side before the transformation. A strong type
920 system makes it easier to read the generated code and enables novel
921 analyses and transformations that are not feasible to perform on normal
922 three address code representations.</p>
926 <!-- ======================================================================= -->
927 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
928 <div class="doc_text">
929 <p>The primitive types are the fundamental building blocks of the LLVM
930 system. The current set of primitive types is as follows:</p>
932 <table class="layout">
937 <tr><th>Type</th><th>Description</th></tr>
938 <tr><td><tt><a name="t_void">void</a></tt></td><td>No value</td></tr>
939 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
946 <tr><th>Type</th><th>Description</th></tr>
947 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
948 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
956 <!-- _______________________________________________________________________ -->
957 <div class="doc_subsubsection"> <a name="t_classifications">Type
958 Classifications</a> </div>
959 <div class="doc_text">
960 <p>These different primitive types fall into a few useful
963 <table border="1" cellspacing="0" cellpadding="4">
965 <tr><th>Classification</th><th>Types</th></tr>
967 <td><a name="t_integer">integer</a></td>
968 <td><tt>i1, i2, i3, ... i8, ... i16, ... i32, ... i64, ... </tt></td>
971 <td><a name="t_floating">floating point</a></td>
972 <td><tt>float, double</tt></td>
975 <td><a name="t_firstclass">first class</a></td>
976 <td><tt>i1, ..., float, double, <br/>
977 <a href="#t_pointer">pointer</a>,<a href="#t_vector">vector</a></tt>
983 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
984 most important. Values of these types are the only ones which can be
985 produced by instructions, passed as arguments, or used as operands to
986 instructions. This means that all structures and arrays must be
987 manipulated either by pointer or by component.</p>
990 <!-- ======================================================================= -->
991 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
993 <div class="doc_text">
995 <p>The real power in LLVM comes from the derived types in the system.
996 This is what allows a programmer to represent arrays, functions,
997 pointers, and other useful types. Note that these derived types may be
998 recursive: For example, it is possible to have a two dimensional array.</p>
1002 <!-- _______________________________________________________________________ -->
1003 <div class="doc_subsubsection"> <a name="t_integer">Integer Type</a> </div>
1005 <div class="doc_text">
1008 <p>The integer type is a very simple derived type that simply specifies an
1009 arbitrary bit width for the integer type desired. Any bit width from 1 bit to
1010 2^23-1 (about 8 million) can be specified.</p>
1018 <p>The number of bits the integer will occupy is specified by the <tt>N</tt>
1022 <table class="layout">
1032 <tt>i1942652</tt><br/>
1035 A boolean integer of 1 bit<br/>
1036 A nibble sized integer of 4 bits.<br/>
1037 A byte sized integer of 8 bits.<br/>
1038 A half word sized integer of 16 bits.<br/>
1039 A word sized integer of 32 bits.<br/>
1040 An integer whose bit width is the answer. <br/>
1041 A double word sized integer of 64 bits.<br/>
1042 A really big integer of over 1 million bits.<br/>
1048 <!-- _______________________________________________________________________ -->
1049 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
1051 <div class="doc_text">
1055 <p>The array type is a very simple derived type that arranges elements
1056 sequentially in memory. The array type requires a size (number of
1057 elements) and an underlying data type.</p>
1062 [<# elements> x <elementtype>]
1065 <p>The number of elements is a constant integer value; elementtype may
1066 be any type with a size.</p>
1069 <table class="layout">
1072 <tt>[40 x i32 ]</tt><br/>
1073 <tt>[41 x i32 ]</tt><br/>
1074 <tt>[40 x i8]</tt><br/>
1077 Array of 40 32-bit integer values.<br/>
1078 Array of 41 32-bit integer values.<br/>
1079 Array of 40 8-bit integer values.<br/>
1083 <p>Here are some examples of multidimensional arrays:</p>
1084 <table class="layout">
1087 <tt>[3 x [4 x i32]]</tt><br/>
1088 <tt>[12 x [10 x float]]</tt><br/>
1089 <tt>[2 x [3 x [4 x i16]]]</tt><br/>
1092 3x4 array of 32-bit integer values.<br/>
1093 12x10 array of single precision floating point values.<br/>
1094 2x3x4 array of 16-bit integer values.<br/>
1099 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
1100 length array. Normally, accesses past the end of an array are undefined in
1101 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
1102 As a special case, however, zero length arrays are recognized to be variable
1103 length. This allows implementation of 'pascal style arrays' with the LLVM
1104 type "{ i32, [0 x float]}", for example.</p>
1108 <!-- _______________________________________________________________________ -->
1109 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
1110 <div class="doc_text">
1112 <p>The function type can be thought of as a function signature. It
1113 consists of a return type and a list of formal parameter types.
1114 Function types are usually used to build virtual function tables
1115 (which are structures of pointers to functions), for indirect function
1116 calls, and when defining a function.</p>
1118 The return type of a function type cannot be an aggregate type.
1121 <pre> <returntype> (<parameter list>)<br></pre>
1122 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
1123 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
1124 which indicates that the function takes a variable number of arguments.
1125 Variable argument functions can access their arguments with the <a
1126 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
1128 <table class="layout">
1130 <td class="left"><tt>i32 (i32)</tt></td>
1131 <td class="left">function taking an <tt>i32</tt>, returning an <tt>i32</tt>
1133 </tr><tr class="layout">
1134 <td class="left"><tt>float (i16 sext, i32 *) *
1136 <td class="left"><a href="#t_pointer">Pointer</a> to a function that takes
1137 an <tt>i16</tt> that should be sign extended and a
1138 <a href="#t_pointer">pointer</a> to <tt>i32</tt>, returning
1141 </tr><tr class="layout">
1142 <td class="left"><tt>i32 (i8*, ...)</tt></td>
1143 <td class="left">A vararg function that takes at least one
1144 <a href="#t_pointer">pointer</a> to <tt>i8 </tt> (char in C),
1145 which returns an integer. This is the signature for <tt>printf</tt> in
1152 <!-- _______________________________________________________________________ -->
1153 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
1154 <div class="doc_text">
1156 <p>The structure type is used to represent a collection of data members
1157 together in memory. The packing of the field types is defined to match
1158 the ABI of the underlying processor. The elements of a structure may
1159 be any type that has a size.</p>
1160 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1161 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1162 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1165 <pre> { <type list> }<br></pre>
1167 <table class="layout">
1169 <td class="left"><tt>{ i32, i32, i32 }</tt></td>
1170 <td class="left">A triple of three <tt>i32</tt> values</td>
1171 </tr><tr class="layout">
1172 <td class="left"><tt>{ float, i32 (i32) * }</tt></td>
1173 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1174 second element is a <a href="#t_pointer">pointer</a> to a
1175 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1176 an <tt>i32</tt>.</td>
1181 <!-- _______________________________________________________________________ -->
1182 <div class="doc_subsubsection"> <a name="t_pstruct">Packed Structure Type</a>
1184 <div class="doc_text">
1186 <p>The packed structure type is used to represent a collection of data members
1187 together in memory. There is no padding between fields. Further, the alignment
1188 of a packed structure is 1 byte. The elements of a packed structure may
1189 be any type that has a size.</p>
1190 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
1191 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
1192 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
1195 <pre> < { <type list> } > <br></pre>
1197 <table class="layout">
1199 <td class="left"><tt>< { i32, i32, i32 } ></tt></td>
1200 <td class="left">A triple of three <tt>i32</tt> values</td>
1201 </tr><tr class="layout">
1202 <td class="left"><tt>< { float, i32 (i32) * } ></tt></td>
1203 <td class="left">A pair, where the first element is a <tt>float</tt> and the
1204 second element is a <a href="#t_pointer">pointer</a> to a
1205 <a href="#t_function">function</a> that takes an <tt>i32</tt>, returning
1206 an <tt>i32</tt>.</td>
1211 <!-- _______________________________________________________________________ -->
1212 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
1213 <div class="doc_text">
1215 <p>As in many languages, the pointer type represents a pointer or
1216 reference to another object, which must live in memory.</p>
1218 <pre> <type> *<br></pre>
1220 <table class="layout">
1223 <tt>[4x i32]*</tt><br/>
1224 <tt>i32 (i32 *) *</tt><br/>
1227 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
1228 four <tt>i32</tt> values<br/>
1229 A <a href="#t_pointer">pointer</a> to a <a
1230 href="#t_function">function</a> that takes an <tt>i32*</tt>, returning an
1237 <!-- _______________________________________________________________________ -->
1238 <div class="doc_subsubsection"> <a name="t_vector">Vector Type</a> </div>
1239 <div class="doc_text">
1243 <p>A vector type is a simple derived type that represents a vector
1244 of elements. Vector types are used when multiple primitive data
1245 are operated in parallel using a single instruction (SIMD).
1246 A vector type requires a size (number of
1247 elements) and an underlying primitive data type. Vectors must have a power
1248 of two length (1, 2, 4, 8, 16 ...). Vector types are
1249 considered <a href="#t_firstclass">first class</a>.</p>
1254 < <# elements> x <elementtype> >
1257 <p>The number of elements is a constant integer value; elementtype may
1258 be any integer or floating point type.</p>
1262 <table class="layout">
1265 <tt><4 x i32></tt><br/>
1266 <tt><8 x float></tt><br/>
1267 <tt><2 x i64></tt><br/>
1270 Vector of 4 32-bit integer values.<br/>
1271 Vector of 8 floating-point values.<br/>
1272 Vector of 2 64-bit integer values.<br/>
1278 <!-- _______________________________________________________________________ -->
1279 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
1280 <div class="doc_text">
1284 <p>Opaque types are used to represent unknown types in the system. This
1285 corresponds (for example) to the C notion of a foward declared structure type.
1286 In LLVM, opaque types can eventually be resolved to any type (not just a
1287 structure type).</p>
1297 <table class="layout">
1303 An opaque type.<br/>
1310 <!-- *********************************************************************** -->
1311 <div class="doc_section"> <a name="constants">Constants</a> </div>
1312 <!-- *********************************************************************** -->
1314 <div class="doc_text">
1316 <p>LLVM has several different basic types of constants. This section describes
1317 them all and their syntax.</p>
1321 <!-- ======================================================================= -->
1322 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
1324 <div class="doc_text">
1327 <dt><b>Boolean constants</b></dt>
1329 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
1330 constants of the <tt><a href="#t_primitive">i1</a></tt> type.
1333 <dt><b>Integer constants</b></dt>
1335 <dd>Standard integers (such as '4') are constants of the <a
1336 href="#t_integer">integer</a> type. Negative numbers may be used with
1340 <dt><b>Floating point constants</b></dt>
1342 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
1343 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
1344 notation (see below). Floating point constants must have a <a
1345 href="#t_floating">floating point</a> type. </dd>
1347 <dt><b>Null pointer constants</b></dt>
1349 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
1350 and must be of <a href="#t_pointer">pointer type</a>.</dd>
1354 <p>The one non-intuitive notation for constants is the optional hexadecimal form
1355 of floating point constants. For example, the form '<tt>double
1356 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
1357 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
1358 (and the only time that they are generated by the disassembler) is when a
1359 floating point constant must be emitted but it cannot be represented as a
1360 decimal floating point number. For example, NaN's, infinities, and other
1361 special values are represented in their IEEE hexadecimal format so that
1362 assembly and disassembly do not cause any bits to change in the constants.</p>
1366 <!-- ======================================================================= -->
1367 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
1370 <div class="doc_text">
1371 <p>Aggregate constants arise from aggregation of simple constants
1372 and smaller aggregate constants.</p>
1375 <dt><b>Structure constants</b></dt>
1377 <dd>Structure constants are represented with notation similar to structure
1378 type definitions (a comma separated list of elements, surrounded by braces
1379 (<tt>{}</tt>)). For example: "<tt>{ i32 4, float 17.0, i32* %G }</tt>",
1380 where "<tt>%G</tt>" is declared as "<tt>@G = external global i32</tt>". Structure constants
1381 must have <a href="#t_struct">structure type</a>, and the number and
1382 types of elements must match those specified by the type.
1385 <dt><b>Array constants</b></dt>
1387 <dd>Array constants are represented with notation similar to array type
1388 definitions (a comma separated list of elements, surrounded by square brackets
1389 (<tt>[]</tt>)). For example: "<tt>[ i32 42, i32 11, i32 74 ]</tt>". Array
1390 constants must have <a href="#t_array">array type</a>, and the number and
1391 types of elements must match those specified by the type.
1394 <dt><b>Vector constants</b></dt>
1396 <dd>Vector constants are represented with notation similar to vector type
1397 definitions (a comma separated list of elements, surrounded by
1398 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< i32 42,
1399 i32 11, i32 74, i32 100 ></tt>". Vector constants must have <a
1400 href="#t_vector">vector type</a>, and the number and types of elements must
1401 match those specified by the type.
1404 <dt><b>Zero initialization</b></dt>
1406 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1407 value to zero of <em>any</em> type, including scalar and aggregate types.
1408 This is often used to avoid having to print large zero initializers (e.g. for
1409 large arrays) and is always exactly equivalent to using explicit zero
1416 <!-- ======================================================================= -->
1417 <div class="doc_subsection">
1418 <a name="globalconstants">Global Variable and Function Addresses</a>
1421 <div class="doc_text">
1423 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1424 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1425 constants. These constants are explicitly referenced when the <a
1426 href="#identifiers">identifier for the global</a> is used and always have <a
1427 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1430 <div class="doc_code">
1434 @Z = global [2 x i32*] [ i32* @X, i32* @Y ]
1440 <!-- ======================================================================= -->
1441 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1442 <div class="doc_text">
1443 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1444 no specific value. Undefined values may be of any type and be used anywhere
1445 a constant is permitted.</p>
1447 <p>Undefined values indicate to the compiler that the program is well defined
1448 no matter what value is used, giving the compiler more freedom to optimize.
1452 <!-- ======================================================================= -->
1453 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1456 <div class="doc_text">
1458 <p>Constant expressions are used to allow expressions involving other constants
1459 to be used as constants. Constant expressions may be of any <a
1460 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1461 that does not have side effects (e.g. load and call are not supported). The
1462 following is the syntax for constant expressions:</p>
1465 <dt><b><tt>trunc ( CST to TYPE )</tt></b></dt>
1466 <dd>Truncate a constant to another type. The bit size of CST must be larger
1467 than the bit size of TYPE. Both types must be integers.</dd>
1469 <dt><b><tt>zext ( CST to TYPE )</tt></b></dt>
1470 <dd>Zero 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>sext ( CST to TYPE )</tt></b></dt>
1474 <dd>Sign extend a constant to another type. The bit size of CST must be
1475 smaller or equal to the bit size of TYPE. Both types must be integers.</dd>
1477 <dt><b><tt>fptrunc ( CST to TYPE )</tt></b></dt>
1478 <dd>Truncate a floating point constant to another floating point type. The
1479 size of CST must be larger than the size of TYPE. Both types must be
1480 floating point.</dd>
1482 <dt><b><tt>fpext ( CST to TYPE )</tt></b></dt>
1483 <dd>Floating point extend a constant to another type. The size of CST must be
1484 smaller or equal to the size of TYPE. Both types must be floating point.</dd>
1486 <dt><b><tt>fp2uint ( CST to TYPE )</tt></b></dt>
1487 <dd>Convert a floating point constant to the corresponding unsigned integer
1488 constant. TYPE must be an integer type. CST must be floating point. If the
1489 value won't fit in the integer type, the results are undefined.</dd>
1491 <dt><b><tt>fptosi ( CST to TYPE )</tt></b></dt>
1492 <dd>Convert a floating point constant to the corresponding signed integer
1493 constant. TYPE must be an integer type. CST must be floating point. If the
1494 value won't fit in the integer type, the results are undefined.</dd>
1496 <dt><b><tt>uitofp ( CST to TYPE )</tt></b></dt>
1497 <dd>Convert an unsigned integer constant to the corresponding floating point
1498 constant. TYPE must be floating point. CST must be of integer type. If the
1499 value won't fit in the floating point type, the results are undefined.</dd>
1501 <dt><b><tt>sitofp ( CST to TYPE )</tt></b></dt>
1502 <dd>Convert a signed integer constant to the corresponding floating point
1503 constant. TYPE must be floating point. CST must be of integer type. If the
1504 value won't fit in the floating point type, the results are undefined.</dd>
1506 <dt><b><tt>ptrtoint ( CST to TYPE )</tt></b></dt>
1507 <dd>Convert a pointer typed constant to the corresponding integer constant
1508 TYPE must be an integer type. CST must be of pointer type. The CST value is
1509 zero extended, truncated, or unchanged to make it fit in TYPE.</dd>
1511 <dt><b><tt>inttoptr ( CST to TYPE )</tt></b></dt>
1512 <dd>Convert a integer constant to a pointer constant. TYPE must be a
1513 pointer type. CST must be of integer type. The CST value is zero extended,
1514 truncated, or unchanged to make it fit in a pointer size. This one is
1515 <i>really</i> dangerous!</dd>
1517 <dt><b><tt>bitcast ( CST to TYPE )</tt></b></dt>
1518 <dd>Convert a constant, CST, to another TYPE. The size of CST and TYPE must be
1519 identical (same number of bits). The conversion is done as if the CST value
1520 was stored to memory and read back as TYPE. In other words, no bits change
1521 with this operator, just the type. This can be used for conversion of
1522 vector types to any other type, as long as they have the same bit width. For
1523 pointers it is only valid to cast to another pointer type.
1526 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1528 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1529 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1530 instruction, the index list may have zero or more indexes, which are required
1531 to make sense for the type of "CSTPTR".</dd>
1533 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1535 <dd>Perform the <a href="#i_select">select operation</a> on
1538 <dt><b><tt>icmp COND ( VAL1, VAL2 )</tt></b></dt>
1539 <dd>Performs the <a href="#i_icmp">icmp operation</a> on constants.</dd>
1541 <dt><b><tt>fcmp COND ( VAL1, VAL2 )</tt></b></dt>
1542 <dd>Performs the <a href="#i_fcmp">fcmp operation</a> on constants.</dd>
1544 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1546 <dd>Perform the <a href="#i_extractelement">extractelement
1547 operation</a> on constants.
1549 <dt><b><tt>insertelement ( VAL, ELT, IDX )</tt></b></dt>
1551 <dd>Perform the <a href="#i_insertelement">insertelement
1552 operation</a> on constants.</dd>
1555 <dt><b><tt>shufflevector ( VEC1, VEC2, IDXMASK )</tt></b></dt>
1557 <dd>Perform the <a href="#i_shufflevector">shufflevector
1558 operation</a> on constants.</dd>
1560 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1562 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1563 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1564 binary</a> operations. The constraints on operands are the same as those for
1565 the corresponding instruction (e.g. no bitwise operations on floating point
1566 values are allowed).</dd>
1570 <!-- *********************************************************************** -->
1571 <div class="doc_section"> <a name="othervalues">Other Values</a> </div>
1572 <!-- *********************************************************************** -->
1574 <!-- ======================================================================= -->
1575 <div class="doc_subsection">
1576 <a name="inlineasm">Inline Assembler Expressions</a>
1579 <div class="doc_text">
1582 LLVM supports inline assembler expressions (as opposed to <a href="#moduleasm">
1583 Module-Level Inline Assembly</a>) through the use of a special value. This
1584 value represents the inline assembler as a string (containing the instructions
1585 to emit), a list of operand constraints (stored as a string), and a flag that
1586 indicates whether or not the inline asm expression has side effects. An example
1587 inline assembler expression is:
1590 <div class="doc_code">
1592 i32 (i32) asm "bswap $0", "=r,r"
1597 Inline assembler expressions may <b>only</b> be used as the callee operand of
1598 a <a href="#i_call"><tt>call</tt> instruction</a>. Thus, typically we have:
1601 <div class="doc_code">
1603 %X = call i32 asm "<a href="#int_bswap">bswap</a> $0", "=r,r"(i32 %Y)
1608 Inline asms with side effects not visible in the constraint list must be marked
1609 as having side effects. This is done through the use of the
1610 '<tt>sideeffect</tt>' keyword, like so:
1613 <div class="doc_code">
1615 call void asm sideeffect "eieio", ""()
1619 <p>TODO: The format of the asm and constraints string still need to be
1620 documented here. Constraints on what can be done (e.g. duplication, moving, etc
1621 need to be documented).
1626 <!-- *********************************************************************** -->
1627 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1628 <!-- *********************************************************************** -->
1630 <div class="doc_text">
1632 <p>The LLVM instruction set consists of several different
1633 classifications of instructions: <a href="#terminators">terminator
1634 instructions</a>, <a href="#binaryops">binary instructions</a>,
1635 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1636 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1637 instructions</a>.</p>
1641 <!-- ======================================================================= -->
1642 <div class="doc_subsection"> <a name="terminators">Terminator
1643 Instructions</a> </div>
1645 <div class="doc_text">
1647 <p>As mentioned <a href="#functionstructure">previously</a>, every
1648 basic block in a program ends with a "Terminator" instruction, which
1649 indicates which block should be executed after the current block is
1650 finished. These terminator instructions typically yield a '<tt>void</tt>'
1651 value: they produce control flow, not values (the one exception being
1652 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1653 <p>There are six different terminator instructions: the '<a
1654 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1655 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1656 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1657 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1658 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1662 <!-- _______________________________________________________________________ -->
1663 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1664 Instruction</a> </div>
1665 <div class="doc_text">
1667 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1668 ret void <i>; Return from void function</i>
1671 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1672 value) from a function back to the caller.</p>
1673 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1674 returns a value and then causes control flow, and one that just causes
1675 control flow to occur.</p>
1677 <p>The '<tt>ret</tt>' instruction may return any '<a
1678 href="#t_firstclass">first class</a>' type. Notice that a function is
1679 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1680 instruction inside of the function that returns a value that does not
1681 match the return type of the function.</p>
1683 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1684 returns back to the calling function's context. If the caller is a "<a
1685 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1686 the instruction after the call. If the caller was an "<a
1687 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1688 at the beginning of the "normal" destination block. If the instruction
1689 returns a value, that value shall set the call or invoke instruction's
1692 <pre> ret i32 5 <i>; Return an integer value of 5</i>
1693 ret void <i>; Return from a void function</i>
1696 <!-- _______________________________________________________________________ -->
1697 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1698 <div class="doc_text">
1700 <pre> br i1 <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1703 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1704 transfer to a different basic block in the current function. There are
1705 two forms of this instruction, corresponding to a conditional branch
1706 and an unconditional branch.</p>
1708 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1709 single '<tt>i1</tt>' value and two '<tt>label</tt>' values. The
1710 unconditional form of the '<tt>br</tt>' instruction takes a single
1711 '<tt>label</tt>' value as a target.</p>
1713 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>i1</tt>'
1714 argument is evaluated. If the value is <tt>true</tt>, control flows
1715 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1716 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1718 <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
1719 href="#i_ret">ret</a> i32 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> i32 0<br></pre>
1721 <!-- _______________________________________________________________________ -->
1722 <div class="doc_subsubsection">
1723 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1726 <div class="doc_text">
1730 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1735 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1736 several different places. It is a generalization of the '<tt>br</tt>'
1737 instruction, allowing a branch to occur to one of many possible
1743 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1744 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1745 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1746 table is not allowed to contain duplicate constant entries.</p>
1750 <p>The <tt>switch</tt> instruction specifies a table of values and
1751 destinations. When the '<tt>switch</tt>' instruction is executed, this
1752 table is searched for the given value. If the value is found, control flow is
1753 transfered to the corresponding destination; otherwise, control flow is
1754 transfered to the default destination.</p>
1756 <h5>Implementation:</h5>
1758 <p>Depending on properties of the target machine and the particular
1759 <tt>switch</tt> instruction, this instruction may be code generated in different
1760 ways. For example, it could be generated as a series of chained conditional
1761 branches or with a lookup table.</p>
1766 <i>; Emulate a conditional br instruction</i>
1767 %Val = <a href="#i_zext">zext</a> i1 %value to i32
1768 switch i32 %Val, label %truedest [i32 0, label %falsedest ]
1770 <i>; Emulate an unconditional br instruction</i>
1771 switch i32 0, label %dest [ ]
1773 <i>; Implement a jump table:</i>
1774 switch i32 %val, label %otherwise [ i32 0, label %onzero
1776 i32 2, label %ontwo ]
1780 <!-- _______________________________________________________________________ -->
1781 <div class="doc_subsubsection">
1782 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1785 <div class="doc_text">
1790 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1791 to label <normal label> unwind label <exception label>
1796 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1797 function, with the possibility of control flow transfer to either the
1798 '<tt>normal</tt>' label or the
1799 '<tt>exception</tt>' label. If the callee function returns with the
1800 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1801 "normal" label. If the callee (or any indirect callees) returns with the "<a
1802 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1803 continued at the dynamically nearest "exception" label.</p>
1807 <p>This instruction requires several arguments:</p>
1811 The optional "cconv" marker indicates which <a href="#callingconv">calling
1812 convention</a> the call should use. If none is specified, the call defaults
1813 to using C calling conventions.
1815 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1816 function value being invoked. In most cases, this is a direct function
1817 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1818 an arbitrary pointer to function value.
1821 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1822 function to be invoked. </li>
1824 <li>'<tt>function args</tt>': argument list whose types match the function
1825 signature argument types. If the function signature indicates the function
1826 accepts a variable number of arguments, the extra arguments can be
1829 <li>'<tt>normal label</tt>': the label reached when the called function
1830 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1832 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1833 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1839 <p>This instruction is designed to operate as a standard '<tt><a
1840 href="#i_call">call</a></tt>' instruction in most regards. The primary
1841 difference is that it establishes an association with a label, which is used by
1842 the runtime library to unwind the stack.</p>
1844 <p>This instruction is used in languages with destructors to ensure that proper
1845 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1846 exception. Additionally, this is important for implementation of
1847 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1851 %retval = invoke i32 %Test(i32 15) to label %Continue
1852 unwind label %TestCleanup <i>; {i32}:retval set</i>
1853 %retval = invoke <a href="#callingconv">coldcc</a> i32 %Test(i32 15) to label %Continue
1854 unwind label %TestCleanup <i>; {i32}:retval set</i>
1859 <!-- _______________________________________________________________________ -->
1861 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1862 Instruction</a> </div>
1864 <div class="doc_text">
1873 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1874 at the first callee in the dynamic call stack which used an <a
1875 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1876 primarily used to implement exception handling.</p>
1880 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1881 immediately halt. The dynamic call stack is then searched for the first <a
1882 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1883 execution continues at the "exceptional" destination block specified by the
1884 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1885 dynamic call chain, undefined behavior results.</p>
1888 <!-- _______________________________________________________________________ -->
1890 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1891 Instruction</a> </div>
1893 <div class="doc_text">
1902 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1903 instruction is used to inform the optimizer that a particular portion of the
1904 code is not reachable. This can be used to indicate that the code after a
1905 no-return function cannot be reached, and other facts.</p>
1909 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1914 <!-- ======================================================================= -->
1915 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1916 <div class="doc_text">
1917 <p>Binary operators are used to do most of the computation in a
1918 program. They require two operands, execute an operation on them, and
1919 produce a single value. The operands might represent
1920 multiple data, as is the case with the <a href="#t_vector">vector</a> data type.
1921 The result value of a binary operator is not
1922 necessarily the same type as its operands.</p>
1923 <p>There are several different binary operators:</p>
1925 <!-- _______________________________________________________________________ -->
1926 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1927 Instruction</a> </div>
1928 <div class="doc_text">
1930 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1933 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1935 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1936 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1937 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1938 Both arguments must have identical types.</p>
1940 <p>The value produced is the integer or floating point sum of the two
1943 <pre> <result> = add i32 4, %var <i>; yields {i32}:result = 4 + %var</i>
1946 <!-- _______________________________________________________________________ -->
1947 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1948 Instruction</a> </div>
1949 <div class="doc_text">
1951 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1954 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1956 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1957 instruction present in most other intermediate representations.</p>
1959 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1960 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1962 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1963 Both arguments must have identical types.</p>
1965 <p>The value produced is the integer or floating point difference of
1966 the two operands.</p>
1969 <result> = sub i32 4, %var <i>; yields {i32}:result = 4 - %var</i>
1970 <result> = sub i32 0, %val <i>; yields {i32}:result = -%var</i>
1973 <!-- _______________________________________________________________________ -->
1974 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1975 Instruction</a> </div>
1976 <div class="doc_text">
1978 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1981 <p>The '<tt>mul</tt>' instruction returns the product of its two
1984 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1985 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1987 This instruction can also take <a href="#t_vector">vector</a> versions of the values.
1988 Both arguments must have identical types.</p>
1990 <p>The value produced is the integer or floating point product of the
1992 <p>Because the operands are the same width, the result of an integer
1993 multiplication is the same whether the operands should be deemed unsigned or
1996 <pre> <result> = mul i32 4, %var <i>; yields {i32}:result = 4 * %var</i>
1999 <!-- _______________________________________________________________________ -->
2000 <div class="doc_subsubsection"> <a name="i_udiv">'<tt>udiv</tt>' Instruction
2002 <div class="doc_text">
2004 <pre> <result> = udiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2007 <p>The '<tt>udiv</tt>' instruction returns the quotient of its two
2010 <p>The two arguments to the '<tt>udiv</tt>' instruction must be
2011 <a href="#t_integer">integer</a> values. Both arguments must have identical
2012 types. This instruction can also take <a href="#t_vector">vector</a> versions
2013 of the values in which case the elements must be integers.</p>
2015 <p>The value produced is the unsigned integer quotient of the two operands. This
2016 instruction always performs an unsigned division operation, regardless of
2017 whether the arguments are unsigned or not.</p>
2019 <pre> <result> = udiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2022 <!-- _______________________________________________________________________ -->
2023 <div class="doc_subsubsection"> <a name="i_sdiv">'<tt>sdiv</tt>' Instruction
2025 <div class="doc_text">
2027 <pre> <result> = sdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2030 <p>The '<tt>sdiv</tt>' instruction returns the quotient of its two
2033 <p>The two arguments to the '<tt>sdiv</tt>' instruction must be
2034 <a href="#t_integer">integer</a> values. Both arguments must have identical
2035 types. This instruction can also take <a href="#t_vector">vector</a> versions
2036 of the values in which case the elements must be integers.</p>
2038 <p>The value produced is the signed integer quotient of the two operands. This
2039 instruction always performs a signed division operation, regardless of whether
2040 the arguments are signed or not.</p>
2042 <pre> <result> = sdiv i32 4, %var <i>; yields {i32}:result = 4 / %var</i>
2045 <!-- _______________________________________________________________________ -->
2046 <div class="doc_subsubsection"> <a name="i_fdiv">'<tt>fdiv</tt>'
2047 Instruction</a> </div>
2048 <div class="doc_text">
2050 <pre> <result> = fdiv <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2053 <p>The '<tt>fdiv</tt>' instruction returns the quotient of its two
2056 <p>The two arguments to the '<tt>fdiv</tt>' instruction must be
2057 <a href="#t_floating">floating point</a> values. Both arguments must have
2058 identical types. This instruction can also take <a href="#t_vector">vector</a>
2059 versions of floating point values.</p>
2061 <p>The value produced is the floating point quotient of the two operands.</p>
2063 <pre> <result> = fdiv float 4.0, %var <i>; yields {float}:result = 4.0 / %var</i>
2066 <!-- _______________________________________________________________________ -->
2067 <div class="doc_subsubsection"> <a name="i_urem">'<tt>urem</tt>' Instruction</a>
2069 <div class="doc_text">
2071 <pre> <result> = urem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2074 <p>The '<tt>urem</tt>' instruction returns the remainder from the
2075 unsigned division of its two arguments.</p>
2077 <p>The two arguments to the '<tt>urem</tt>' instruction must be
2078 <a href="#t_integer">integer</a> values. Both arguments must have identical
2081 <p>This instruction returns the unsigned integer <i>remainder</i> of a division.
2082 This instruction always performs an unsigned division to get the remainder,
2083 regardless of whether the arguments are unsigned or not.</p>
2085 <pre> <result> = urem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2089 <!-- _______________________________________________________________________ -->
2090 <div class="doc_subsubsection"> <a name="i_srem">'<tt>srem</tt>'
2091 Instruction</a> </div>
2092 <div class="doc_text">
2094 <pre> <result> = srem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2097 <p>The '<tt>srem</tt>' instruction returns the remainder from the
2098 signed division of its two operands.</p>
2100 <p>The two arguments to the '<tt>srem</tt>' instruction must be
2101 <a href="#t_integer">integer</a> values. Both arguments must have identical
2104 <p>This instruction returns the <i>remainder</i> of a division (where the result
2105 has the same sign as the dividend, <tt>var1</tt>), not the <i>modulo</i>
2106 operator (where the result has the same sign as the divisor, <tt>var2</tt>) of
2107 a value. For more information about the difference, see <a
2108 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
2109 Math Forum</a>. For a table of how this is implemented in various languages,
2110 please see <a href="http://en.wikipedia.org/wiki/Modulo_operation">
2111 Wikipedia: modulo operation</a>.</p>
2113 <pre> <result> = srem i32 4, %var <i>; yields {i32}:result = 4 % %var</i>
2117 <!-- _______________________________________________________________________ -->
2118 <div class="doc_subsubsection"> <a name="i_frem">'<tt>frem</tt>'
2119 Instruction</a> </div>
2120 <div class="doc_text">
2122 <pre> <result> = frem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2125 <p>The '<tt>frem</tt>' instruction returns the remainder from the
2126 division of its two operands.</p>
2128 <p>The two arguments to the '<tt>frem</tt>' instruction must be
2129 <a href="#t_floating">floating point</a> values. Both arguments must have
2130 identical types.</p>
2132 <p>This instruction returns the <i>remainder</i> of a division.</p>
2134 <pre> <result> = frem float 4.0, %var <i>; yields {float}:result = 4.0 % %var</i>
2138 <!-- ======================================================================= -->
2139 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
2140 Operations</a> </div>
2141 <div class="doc_text">
2142 <p>Bitwise binary operators are used to do various forms of
2143 bit-twiddling in a program. They are generally very efficient
2144 instructions and can commonly be strength reduced from other
2145 instructions. They require two operands, execute an operation on them,
2146 and produce a single value. The resulting value of the bitwise binary
2147 operators is always the same type as its first operand.</p>
2150 <!-- _______________________________________________________________________ -->
2151 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
2152 Instruction</a> </div>
2153 <div class="doc_text">
2155 <pre> <result> = shl <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2158 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
2159 the left a specified number of bits.</p>
2161 <p>Both arguments to the '<tt>shl</tt>' instruction must be the same <a
2162 href="#t_integer">integer</a> type.</p>
2164 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
2165 <h5>Example:</h5><pre>
2166 <result> = shl i32 4, %var <i>; yields {i32}: 4 << %var</i>
2167 <result> = shl i32 4, 2 <i>; yields {i32}: 16</i>
2168 <result> = shl i32 1, 10 <i>; yields {i32}: 1024</i>
2171 <!-- _______________________________________________________________________ -->
2172 <div class="doc_subsubsection"> <a name="i_lshr">'<tt>lshr</tt>'
2173 Instruction</a> </div>
2174 <div class="doc_text">
2176 <pre> <result> = lshr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2180 <p>The '<tt>lshr</tt>' instruction (logical shift right) returns the first
2181 operand shifted to the right a specified number of bits with zero fill.</p>
2184 <p>Both arguments to the '<tt>lshr</tt>' instruction must be the same
2185 <a href="#t_integer">integer</a> type.</p>
2188 <p>This instruction always performs a logical shift right operation. The most
2189 significant bits of the result will be filled with zero bits after the
2194 <result> = lshr i32 4, 1 <i>; yields {i32}:result = 2</i>
2195 <result> = lshr i32 4, 2 <i>; yields {i32}:result = 1</i>
2196 <result> = lshr i8 4, 3 <i>; yields {i8}:result = 0</i>
2197 <result> = lshr i8 -2, 1 <i>; yields {i8}:result = 0x7FFFFFFF </i>
2201 <!-- _______________________________________________________________________ -->
2202 <div class="doc_subsubsection"> <a name="i_ashr">'<tt>ashr</tt>'
2203 Instruction</a> </div>
2204 <div class="doc_text">
2207 <pre> <result> = ashr <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2211 <p>The '<tt>ashr</tt>' instruction (arithmetic shift right) returns the first
2212 operand shifted to the right a specified number of bits with sign extension.</p>
2215 <p>Both arguments to the '<tt>ashr</tt>' instruction must be the same
2216 <a href="#t_integer">integer</a> type.</p>
2219 <p>This instruction always performs an arithmetic shift right operation,
2220 The most significant bits of the result will be filled with the sign bit
2221 of <tt>var1</tt>.</p>
2225 <result> = ashr i32 4, 1 <i>; yields {i32}:result = 2</i>
2226 <result> = ashr i32 4, 2 <i>; yields {i32}:result = 1</i>
2227 <result> = ashr i8 4, 3 <i>; yields {i8}:result = 0</i>
2228 <result> = ashr i8 -2, 1 <i>; yields {i8}:result = -1</i>
2232 <!-- _______________________________________________________________________ -->
2233 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
2234 Instruction</a> </div>
2235 <div class="doc_text">
2237 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2240 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
2241 its two operands.</p>
2243 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
2244 href="#t_integer">integer</a> values. Both arguments must have
2245 identical types.</p>
2247 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
2249 <div style="align: center">
2250 <table border="1" cellspacing="0" cellpadding="4">
2281 <pre> <result> = and i32 4, %var <i>; yields {i32}:result = 4 & %var</i>
2282 <result> = and i32 15, 40 <i>; yields {i32}:result = 8</i>
2283 <result> = and i32 4, 8 <i>; yields {i32}:result = 0</i>
2286 <!-- _______________________________________________________________________ -->
2287 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
2288 <div class="doc_text">
2290 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2293 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
2294 or of its two operands.</p>
2296 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
2297 href="#t_integer">integer</a> values. Both arguments must have
2298 identical types.</p>
2300 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
2302 <div style="align: center">
2303 <table border="1" cellspacing="0" cellpadding="4">
2334 <pre> <result> = or i32 4, %var <i>; yields {i32}:result = 4 | %var</i>
2335 <result> = or i32 15, 40 <i>; yields {i32}:result = 47</i>
2336 <result> = or i32 4, 8 <i>; yields {i32}:result = 12</i>
2339 <!-- _______________________________________________________________________ -->
2340 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
2341 Instruction</a> </div>
2342 <div class="doc_text">
2344 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
2347 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
2348 or of its two operands. The <tt>xor</tt> is used to implement the
2349 "one's complement" operation, which is the "~" operator in C.</p>
2351 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
2352 href="#t_integer">integer</a> values. Both arguments must have
2353 identical types.</p>
2355 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
2357 <div style="align: center">
2358 <table border="1" cellspacing="0" cellpadding="4">
2390 <pre> <result> = xor i32 4, %var <i>; yields {i32}:result = 4 ^ %var</i>
2391 <result> = xor i32 15, 40 <i>; yields {i32}:result = 39</i>
2392 <result> = xor i32 4, 8 <i>; yields {i32}:result = 12</i>
2393 <result> = xor i32 %V, -1 <i>; yields {i32}:result = ~%V</i>
2397 <!-- ======================================================================= -->
2398 <div class="doc_subsection">
2399 <a name="vectorops">Vector Operations</a>
2402 <div class="doc_text">
2404 <p>LLVM supports several instructions to represent vector operations in a
2405 target-independent manner. These instructions cover the element-access and
2406 vector-specific operations needed to process vectors effectively. While LLVM
2407 does directly support these vector operations, many sophisticated algorithms
2408 will want to use target-specific intrinsics to take full advantage of a specific
2413 <!-- _______________________________________________________________________ -->
2414 <div class="doc_subsubsection">
2415 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2418 <div class="doc_text">
2423 <result> = extractelement <n x <ty>> <val>, i32 <idx> <i>; yields <ty></i>
2429 The '<tt>extractelement</tt>' instruction extracts a single scalar
2430 element from a vector at a specified index.
2437 The first operand of an '<tt>extractelement</tt>' instruction is a
2438 value of <a href="#t_vector">vector</a> type. The second operand is
2439 an index indicating the position from which to extract the element.
2440 The index may be a variable.</p>
2445 The result is a scalar of the same type as the element type of
2446 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2447 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2448 results are undefined.
2454 %result = extractelement <4 x i32> %vec, i32 0 <i>; yields i32</i>
2459 <!-- _______________________________________________________________________ -->
2460 <div class="doc_subsubsection">
2461 <a name="i_insertelement">'<tt>insertelement</tt>' Instruction</a>
2464 <div class="doc_text">
2469 <result> = insertelement <n x <ty>> <val>, <ty> <elt>, i32 <idx> <i>; yields <n x <ty>></i>
2475 The '<tt>insertelement</tt>' instruction inserts a scalar
2476 element into a vector at a specified index.
2483 The first operand of an '<tt>insertelement</tt>' instruction is a
2484 value of <a href="#t_vector">vector</a> type. The second operand is a
2485 scalar value whose type must equal the element type of the first
2486 operand. The third operand is an index indicating the position at
2487 which to insert the value. The index may be a variable.</p>
2492 The result is a vector of the same type as <tt>val</tt>. Its
2493 element values are those of <tt>val</tt> except at position
2494 <tt>idx</tt>, where it gets the value <tt>elt</tt>. If <tt>idx</tt>
2495 exceeds the length of <tt>val</tt>, the results are undefined.
2501 %result = insertelement <4 x i32> %vec, i32 1, i32 0 <i>; yields <4 x i32></i>
2505 <!-- _______________________________________________________________________ -->
2506 <div class="doc_subsubsection">
2507 <a name="i_shufflevector">'<tt>shufflevector</tt>' Instruction</a>
2510 <div class="doc_text">
2515 <result> = shufflevector <n x <ty>> <v1>, <n x <ty>> <v2>, <n x i32> <mask> <i>; yields <n x <ty>></i>
2521 The '<tt>shufflevector</tt>' instruction constructs a permutation of elements
2522 from two input vectors, returning a vector of the same type.
2528 The first two operands of a '<tt>shufflevector</tt>' instruction are vectors
2529 with types that match each other and types that match the result of the
2530 instruction. The third argument is a shuffle mask, which has the same number
2531 of elements as the other vector type, but whose element type is always 'i32'.
2535 The shuffle mask operand is required to be a constant vector with either
2536 constant integer or undef values.
2542 The elements of the two input vectors are numbered from left to right across
2543 both of the vectors. The shuffle mask operand specifies, for each element of
2544 the result vector, which element of the two input registers the result element
2545 gets. The element selector may be undef (meaning "don't care") and the second
2546 operand may be undef if performing a shuffle from only one vector.
2552 %result = shufflevector <4 x i32> %v1, <4 x i32> %v2,
2553 <4 x i32> <i32 0, i32 4, i32 1, i32 5> <i>; yields <4 x i32></i>
2554 %result = shufflevector <4 x i32> %v1, <4 x i32> undef,
2555 <4 x i32> <i32 0, i32 1, i32 2, i32 3> <i>; yields <4 x i32></i> - Identity shuffle.
2560 <!-- ======================================================================= -->
2561 <div class="doc_subsection">
2562 <a name="memoryops">Memory Access and Addressing Operations</a>
2565 <div class="doc_text">
2567 <p>A key design point of an SSA-based representation is how it
2568 represents memory. In LLVM, no memory locations are in SSA form, which
2569 makes things very simple. This section describes how to read, write,
2570 allocate, and free memory in LLVM.</p>
2574 <!-- _______________________________________________________________________ -->
2575 <div class="doc_subsubsection">
2576 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
2579 <div class="doc_text">
2584 <result> = malloc <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2589 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
2590 heap and returns a pointer to it.</p>
2594 <p>The '<tt>malloc</tt>' instruction allocates
2595 <tt>sizeof(<type>)*NumElements</tt>
2596 bytes of memory from the operating system and returns a pointer of the
2597 appropriate type to the program. If "NumElements" is specified, it is the
2598 number of elements allocated. If an alignment is specified, the value result
2599 of the allocation is guaranteed to be aligned to at least that boundary. If
2600 not specified, or if zero, the target can choose to align the allocation on any
2601 convenient boundary.</p>
2603 <p>'<tt>type</tt>' must be a sized type.</p>
2607 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
2608 a pointer is returned.</p>
2613 %array = malloc [4 x i8 ] <i>; yields {[%4 x i8]*}:array</i>
2615 %size = <a href="#i_add">add</a> i32 2, 2 <i>; yields {i32}:size = i32 4</i>
2616 %array1 = malloc i8, i32 4 <i>; yields {i8*}:array1</i>
2617 %array2 = malloc [12 x i8], i32 %size <i>; yields {[12 x i8]*}:array2</i>
2618 %array3 = malloc i32, i32 4, align 1024 <i>; yields {i32*}:array3</i>
2619 %array4 = malloc i32, align 1024 <i>; yields {i32*}:array4</i>
2623 <!-- _______________________________________________________________________ -->
2624 <div class="doc_subsubsection">
2625 <a name="i_free">'<tt>free</tt>' Instruction</a>
2628 <div class="doc_text">
2633 free <type> <value> <i>; yields {void}</i>
2638 <p>The '<tt>free</tt>' instruction returns memory back to the unused
2639 memory heap to be reallocated in the future.</p>
2643 <p>'<tt>value</tt>' shall be a pointer value that points to a value
2644 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
2649 <p>Access to the memory pointed to by the pointer is no longer defined
2650 after this instruction executes.</p>
2655 %array = <a href="#i_malloc">malloc</a> [4 x i8] <i>; yields {[4 x i8]*}:array</i>
2656 free [4 x i8]* %array
2660 <!-- _______________________________________________________________________ -->
2661 <div class="doc_subsubsection">
2662 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
2665 <div class="doc_text">
2670 <result> = alloca <type>[, i32 <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
2675 <p>The '<tt>alloca</tt>' instruction allocates memory on the stack frame of the
2676 currently executing function, to be automatically released when this function
2677 returns to its caller.</p>
2681 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
2682 bytes of memory on the runtime stack, returning a pointer of the
2683 appropriate type to the program. If "NumElements" is specified, it is the
2684 number of elements allocated. If an alignment is specified, the value result
2685 of the allocation is guaranteed to be aligned to at least that boundary. If
2686 not specified, or if zero, the target can choose to align the allocation on any
2687 convenient boundary.</p>
2689 <p>'<tt>type</tt>' may be any sized type.</p>
2693 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
2694 memory is automatically released when the function returns. The '<tt>alloca</tt>'
2695 instruction is commonly used to represent automatic variables that must
2696 have an address available. When the function returns (either with the <tt><a
2697 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
2698 instructions), the memory is reclaimed.</p>
2703 %ptr = alloca i32 <i>; yields {i32*}:ptr</i>
2704 %ptr = alloca i32, i32 4 <i>; yields {i32*}:ptr</i>
2705 %ptr = alloca i32, i32 4, align 1024 <i>; yields {i32*}:ptr</i>
2706 %ptr = alloca i32, align 1024 <i>; yields {i32*}:ptr</i>
2710 <!-- _______________________________________________________________________ -->
2711 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
2712 Instruction</a> </div>
2713 <div class="doc_text">
2715 <pre> <result> = load <ty>* <pointer>[, align <alignment>]<br> <result> = volatile load <ty>* <pointer>[, align <alignment>]<br></pre>
2717 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
2719 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
2720 address from which to load. The pointer must point to a <a
2721 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
2722 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
2723 the number or order of execution of this <tt>load</tt> with other
2724 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
2727 <p>The location of memory pointed to is loaded.</p>
2729 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2731 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2732 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2735 <!-- _______________________________________________________________________ -->
2736 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
2737 Instruction</a> </div>
2738 <div class="doc_text">
2740 <pre> store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2741 volatile store <ty> <value>, <ty>* <pointer>[, align <alignment>] <i>; yields {void}</i>
2744 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
2746 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
2747 to store and an address at which to store it. The type of the '<tt><pointer></tt>'
2748 operand must be a pointer to the type of the '<tt><value></tt>'
2749 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
2750 optimizer is not allowed to modify the number or order of execution of
2751 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
2752 href="#i_store">store</a></tt> instructions.</p>
2754 <p>The contents of memory are updated to contain '<tt><value></tt>'
2755 at the location specified by the '<tt><pointer></tt>' operand.</p>
2757 <pre> %ptr = <a href="#i_alloca">alloca</a> i32 <i>; yields {i32*}:ptr</i>
2759 href="#i_store">store</a> i32 3, i32* %ptr <i>; yields {void}</i>
2760 %val = load i32* %ptr <i>; yields {i32}:val = i32 3</i>
2764 <!-- _______________________________________________________________________ -->
2765 <div class="doc_subsubsection">
2766 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2769 <div class="doc_text">
2772 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2778 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2779 subelement of an aggregate data structure.</p>
2783 <p>This instruction takes a list of integer operands that indicate what
2784 elements of the aggregate object to index to. The actual types of the arguments
2785 provided depend on the type of the first pointer argument. The
2786 '<tt>getelementptr</tt>' instruction is used to index down through the type
2787 levels of a structure or to a specific index in an array. When indexing into a
2788 structure, only <tt>i32</tt> integer constants are allowed. When indexing
2789 into an array or pointer, only integers of 32 or 64 bits are allowed, and will
2790 be sign extended to 64-bit values.</p>
2792 <p>For example, let's consider a C code fragment and how it gets
2793 compiled to LLVM:</p>
2795 <div class="doc_code">
2808 int *foo(struct ST *s) {
2809 return &s[1].Z.B[5][13];
2814 <p>The LLVM code generated by the GCC frontend is:</p>
2816 <div class="doc_code">
2818 %RT = type { i8 , [10 x [20 x i32]], i8 }
2819 %ST = type { i32, double, %RT }
2821 define i32* %foo(%ST* %s) {
2823 %reg = getelementptr %ST* %s, i32 1, i32 2, i32 1, i32 5, i32 13
2831 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2832 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2833 and <a href="#t_array">array</a> types can use a 32-bit or 64-bit
2834 <a href="#t_integer">integer</a> type but the value will always be sign extended
2835 to 64-bits. <a href="#t_struct">Structure</a> types require <tt>i32</tt>
2836 <b>constants</b>.</p>
2838 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2839 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ i32, double, %RT
2840 }</tt>' type, a structure. The second index indexes into the third element of
2841 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ i8 , [10 x [20 x i32]],
2842 i8 }</tt>' type, another structure. The third index indexes into the second
2843 element of the structure, yielding a '<tt>[10 x [20 x i32]]</tt>' type, an
2844 array. The two dimensions of the array are subscripted into, yielding an
2845 '<tt>i32</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2846 to this element, thus computing a value of '<tt>i32*</tt>' type.</p>
2848 <p>Note that it is perfectly legal to index partially through a
2849 structure, returning a pointer to an inner element. Because of this,
2850 the LLVM code for the given testcase is equivalent to:</p>
2853 define i32* %foo(%ST* %s) {
2854 %t1 = getelementptr %ST* %s, i32 1 <i>; yields %ST*:%t1</i>
2855 %t2 = getelementptr %ST* %t1, i32 0, i32 2 <i>; yields %RT*:%t2</i>
2856 %t3 = getelementptr %RT* %t2, i32 0, i32 1 <i>; yields [10 x [20 x i32]]*:%t3</i>
2857 %t4 = getelementptr [10 x [20 x i32]]* %t3, i32 0, i32 5 <i>; yields [20 x i32]*:%t4</i>
2858 %t5 = getelementptr [20 x i32]* %t4, i32 0, i32 13 <i>; yields i32*:%t5</i>
2863 <p>Note that it is undefined to access an array out of bounds: array and
2864 pointer indexes must always be within the defined bounds of the array type.
2865 The one exception for this rules is zero length arrays. These arrays are
2866 defined to be accessible as variable length arrays, which requires access
2867 beyond the zero'th element.</p>
2869 <p>The getelementptr instruction is often confusing. For some more insight
2870 into how it works, see <a href="GetElementPtr.html">the getelementptr
2876 <i>; yields [12 x i8]*:aptr</i>
2877 %aptr = getelementptr {i32, [12 x i8]}* %sptr, i64 0, i32 1
2881 <!-- ======================================================================= -->
2882 <div class="doc_subsection"> <a name="convertops">Conversion Operations</a>
2884 <div class="doc_text">
2885 <p>The instructions in this category are the conversion instructions (casting)
2886 which all take a single operand and a type. They perform various bit conversions
2890 <!-- _______________________________________________________________________ -->
2891 <div class="doc_subsubsection">
2892 <a name="i_trunc">'<tt>trunc .. to</tt>' Instruction</a>
2894 <div class="doc_text">
2898 <result> = trunc <ty> <value> to <ty2> <i>; yields ty2</i>
2903 The '<tt>trunc</tt>' instruction truncates its operand to the type <tt>ty2</tt>.
2908 The '<tt>trunc</tt>' instruction takes a <tt>value</tt> to trunc, which must
2909 be an <a href="#t_integer">integer</a> type, and a type that specifies the size
2910 and type of the result, which must be an <a href="#t_integer">integer</a>
2911 type. The bit size of <tt>value</tt> must be larger than the bit size of
2912 <tt>ty2</tt>. Equal sized types are not allowed.</p>
2916 The '<tt>trunc</tt>' instruction truncates the high order bits in <tt>value</tt>
2917 and converts the remaining bits to <tt>ty2</tt>. Since the source size must be
2918 larger than the destination size, <tt>trunc</tt> cannot be a <i>no-op cast</i>.
2919 It will always truncate bits.</p>
2923 %X = trunc i32 257 to i8 <i>; yields i8:1</i>
2924 %Y = trunc i32 123 to i1 <i>; yields i1:true</i>
2925 %Y = trunc i32 122 to i1 <i>; yields i1:false</i>
2929 <!-- _______________________________________________________________________ -->
2930 <div class="doc_subsubsection">
2931 <a name="i_zext">'<tt>zext .. to</tt>' Instruction</a>
2933 <div class="doc_text">
2937 <result> = zext <ty> <value> to <ty2> <i>; yields ty2</i>
2941 <p>The '<tt>zext</tt>' instruction zero extends its operand to type
2946 <p>The '<tt>zext</tt>' instruction takes a value to cast, which must be of
2947 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
2948 also be of <a href="#t_integer">integer</a> type. The bit size of the
2949 <tt>value</tt> must be smaller than the bit size of the destination type,
2953 <p>The <tt>zext</tt> fills the high order bits of the <tt>value</tt> with zero
2954 bits until it reaches the size of the destination type, <tt>ty2</tt>.</p>
2956 <p>When zero extending from i1, the result will always be either 0 or 1.</p>
2960 %X = zext i32 257 to i64 <i>; yields i64:257</i>
2961 %Y = zext i1 true to i32 <i>; yields i32:1</i>
2965 <!-- _______________________________________________________________________ -->
2966 <div class="doc_subsubsection">
2967 <a name="i_sext">'<tt>sext .. to</tt>' Instruction</a>
2969 <div class="doc_text">
2973 <result> = sext <ty> <value> to <ty2> <i>; yields ty2</i>
2977 <p>The '<tt>sext</tt>' sign extends <tt>value</tt> to the type <tt>ty2</tt>.</p>
2981 The '<tt>sext</tt>' instruction takes a value to cast, which must be of
2982 <a href="#t_integer">integer</a> type, and a type to cast it to, which must
2983 also be of <a href="#t_integer">integer</a> type. The bit size of the
2984 <tt>value</tt> must be smaller than the bit size of the destination type,
2989 The '<tt>sext</tt>' instruction performs a sign extension by copying the sign
2990 bit (highest order bit) of the <tt>value</tt> until it reaches the bit size of
2991 the type <tt>ty2</tt>.</p>
2993 <p>When sign extending from i1, the extension always results in -1 or 0.</p>
2997 %X = sext i8 -1 to i16 <i>; yields i16 :65535</i>
2998 %Y = sext i1 true to i32 <i>; yields i32:-1</i>
3002 <!-- _______________________________________________________________________ -->
3003 <div class="doc_subsubsection">
3004 <a name="i_fptrunc">'<tt>fptrunc .. to</tt>' Instruction</a>
3007 <div class="doc_text">
3012 <result> = fptrunc <ty> <value> to <ty2> <i>; yields ty2</i>
3016 <p>The '<tt>fptrunc</tt>' instruction truncates <tt>value</tt> to type
3021 <p>The '<tt>fptrunc</tt>' instruction takes a <a href="#t_floating">floating
3022 point</a> value to cast and a <a href="#t_floating">floating point</a> type to
3023 cast it to. The size of <tt>value</tt> must be larger than the size of
3024 <tt>ty2</tt>. This implies that <tt>fptrunc</tt> cannot be used to make a
3025 <i>no-op cast</i>.</p>
3028 <p> The '<tt>fptrunc</tt>' instruction truncates a <tt>value</tt> from a larger
3029 <a href="#t_floating">floating point</a> type to a smaller
3030 <a href="#t_floating">floating point</a> type. If the value cannot fit within
3031 the destination type, <tt>ty2</tt>, then the results are undefined.</p>
3035 %X = fptrunc double 123.0 to float <i>; yields float:123.0</i>
3036 %Y = fptrunc double 1.0E+300 to float <i>; yields undefined</i>
3040 <!-- _______________________________________________________________________ -->
3041 <div class="doc_subsubsection">
3042 <a name="i_fpext">'<tt>fpext .. to</tt>' Instruction</a>
3044 <div class="doc_text">
3048 <result> = fpext <ty> <value> to <ty2> <i>; yields ty2</i>
3052 <p>The '<tt>fpext</tt>' extends a floating point <tt>value</tt> to a larger
3053 floating point value.</p>
3056 <p>The '<tt>fpext</tt>' instruction takes a
3057 <a href="#t_floating">floating point</a> <tt>value</tt> to cast,
3058 and a <a href="#t_floating">floating point</a> type to cast it to. The source
3059 type must be smaller than the destination type.</p>
3062 <p>The '<tt>fpext</tt>' instruction extends the <tt>value</tt> from a smaller
3063 <a href="#t_floating">floating point</a> type to a larger
3064 <a href="#t_floating">floating point</a> type. The <tt>fpext</tt> cannot be
3065 used to make a <i>no-op cast</i> because it always changes bits. Use
3066 <tt>bitcast</tt> to make a <i>no-op cast</i> for a floating point cast.</p>
3070 %X = fpext float 3.1415 to double <i>; yields double:3.1415</i>
3071 %Y = fpext float 1.0 to float <i>; yields float:1.0 (no-op)</i>
3075 <!-- _______________________________________________________________________ -->
3076 <div class="doc_subsubsection">
3077 <a name="i_fptoui">'<tt>fptoui .. to</tt>' Instruction</a>
3079 <div class="doc_text">
3083 <result> = fp2uint <ty> <value> to <ty2> <i>; yields ty2</i>
3087 <p>The '<tt>fp2uint</tt>' converts a floating point <tt>value</tt> to its
3088 unsigned integer equivalent of type <tt>ty2</tt>.
3092 <p>The '<tt>fp2uint</tt>' instruction takes a value to cast, which must be a
3093 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
3094 must be an <a href="#t_integer">integer</a> type.</p>
3097 <p> The '<tt>fp2uint</tt>' instruction converts its
3098 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3099 towards zero) unsigned integer value. If the value cannot fit in <tt>ty2</tt>,
3100 the results are undefined.</p>
3102 <p>When converting to i1, the conversion is done as a comparison against
3103 zero. If the <tt>value</tt> was zero, the i1 result will be <tt>false</tt>.
3104 If the <tt>value</tt> was non-zero, the i1 result will be <tt>true</tt>.</p>
3108 %X = fp2uint double 123.0 to i32 <i>; yields i32:123</i>
3109 %Y = fp2uint float 1.0E+300 to i1 <i>; yields i1:true</i>
3110 %X = fp2uint float 1.04E+17 to i8 <i>; yields undefined:1</i>
3114 <!-- _______________________________________________________________________ -->
3115 <div class="doc_subsubsection">
3116 <a name="i_fptosi">'<tt>fptosi .. to</tt>' Instruction</a>
3118 <div class="doc_text">
3122 <result> = fptosi <ty> <value> to <ty2> <i>; yields ty2</i>
3126 <p>The '<tt>fptosi</tt>' instruction converts
3127 <a href="#t_floating">floating point</a> <tt>value</tt> to type <tt>ty2</tt>.
3132 <p> The '<tt>fptosi</tt>' instruction takes a value to cast, which must be a
3133 <a href="#t_floating">floating point</a> value, and a type to cast it to, which
3134 must also be an <a href="#t_integer">integer</a> type.</p>
3137 <p>The '<tt>fptosi</tt>' instruction converts its
3138 <a href="#t_floating">floating point</a> operand into the nearest (rounding
3139 towards zero) signed integer value. If the value cannot fit in <tt>ty2</tt>,
3140 the results are undefined.</p>
3142 <p>When converting to i1, the conversion is done as a comparison against
3143 zero. If the <tt>value</tt> was zero, the i1 result will be <tt>false</tt>.
3144 If the <tt>value</tt> was non-zero, the i1 result will be <tt>true</tt>.</p>
3148 %X = fptosi double -123.0 to i32 <i>; yields i32:-123</i>
3149 %Y = fptosi float 1.0E-247 to i1 <i>; yields i1:true</i>
3150 %X = fptosi float 1.04E+17 to i8 <i>; yields undefined:1</i>
3154 <!-- _______________________________________________________________________ -->
3155 <div class="doc_subsubsection">
3156 <a name="i_uitofp">'<tt>uitofp .. to</tt>' Instruction</a>
3158 <div class="doc_text">
3162 <result> = uitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3166 <p>The '<tt>uitofp</tt>' instruction regards <tt>value</tt> as an unsigned
3167 integer and converts that value to the <tt>ty2</tt> type.</p>
3171 <p>The '<tt>uitofp</tt>' instruction takes a value to cast, which must be an
3172 <a href="#t_integer">integer</a> value, and a type to cast it to, which must
3173 be a <a href="#t_floating">floating point</a> type.</p>
3176 <p>The '<tt>uitofp</tt>' instruction interprets its operand as an unsigned
3177 integer quantity and converts it to the corresponding floating point value. If
3178 the value cannot fit in the floating point value, the results are undefined.</p>
3183 %X = uitofp i32 257 to float <i>; yields float:257.0</i>
3184 %Y = uitofp i8 -1 to double <i>; yields double:255.0</i>
3188 <!-- _______________________________________________________________________ -->
3189 <div class="doc_subsubsection">
3190 <a name="i_sitofp">'<tt>sitofp .. to</tt>' Instruction</a>
3192 <div class="doc_text">
3196 <result> = sitofp <ty> <value> to <ty2> <i>; yields ty2</i>
3200 <p>The '<tt>sitofp</tt>' instruction regards <tt>value</tt> as a signed
3201 integer and converts that value to the <tt>ty2</tt> type.</p>
3204 <p>The '<tt>sitofp</tt>' instruction takes a value to cast, which must be an
3205 <a href="#t_integer">integer</a> value, and a type to cast it to, which must be
3206 a <a href="#t_floating">floating point</a> type.</p>
3209 <p>The '<tt>sitofp</tt>' instruction interprets its operand as a signed
3210 integer quantity and converts it to the corresponding floating point value. If
3211 the value cannot fit in the floating point value, the results are undefined.</p>
3215 %X = sitofp i32 257 to float <i>; yields float:257.0</i>
3216 %Y = sitofp i8 -1 to double <i>; yields double:-1.0</i>
3220 <!-- _______________________________________________________________________ -->
3221 <div class="doc_subsubsection">
3222 <a name="i_ptrtoint">'<tt>ptrtoint .. to</tt>' Instruction</a>
3224 <div class="doc_text">
3228 <result> = ptrtoint <ty> <value> to <ty2> <i>; yields ty2</i>
3232 <p>The '<tt>ptrtoint</tt>' instruction converts the pointer <tt>value</tt> to
3233 the integer type <tt>ty2</tt>.</p>
3236 <p>The '<tt>ptrtoint</tt>' instruction takes a <tt>value</tt> to cast, which
3237 must be a <a href="#t_pointer">pointer</a> value, and a type to cast it to
3238 <tt>ty2</tt>, which must be an <a href="#t_integer">integer</a> type.
3241 <p>The '<tt>ptrtoint</tt>' instruction converts <tt>value</tt> to integer type
3242 <tt>ty2</tt> by interpreting the pointer value as an integer and either
3243 truncating or zero extending that value to the size of the integer type. If
3244 <tt>value</tt> is smaller than <tt>ty2</tt> then a zero extension is done. If
3245 <tt>value</tt> is larger than <tt>ty2</tt> then a truncation is done. If they
3246 are the same size, then nothing is done (<i>no-op cast</i>) other than a type
3251 %X = ptrtoint i32* %X to i8 <i>; yields truncation on 32-bit architecture</i>
3252 %Y = ptrtoint i32* %x to i64 <i>; yields zero extension on 32-bit architecture</i>
3256 <!-- _______________________________________________________________________ -->
3257 <div class="doc_subsubsection">
3258 <a name="i_inttoptr">'<tt>inttoptr .. to</tt>' Instruction</a>
3260 <div class="doc_text">
3264 <result> = inttoptr <ty> <value> to <ty2> <i>; yields ty2</i>
3268 <p>The '<tt>inttoptr</tt>' instruction converts an integer <tt>value</tt> to
3269 a pointer type, <tt>ty2</tt>.</p>
3272 <p>The '<tt>inttoptr</tt>' instruction takes an <a href="#t_integer">integer</a>
3273 value to cast, and a type to cast it to, which must be a
3274 <a href="#t_pointer">pointer</a> type.
3277 <p>The '<tt>inttoptr</tt>' instruction converts <tt>value</tt> to type
3278 <tt>ty2</tt> by applying either a zero extension or a truncation depending on
3279 the size of the integer <tt>value</tt>. If <tt>value</tt> is larger than the
3280 size of a pointer then a truncation is done. If <tt>value</tt> is smaller than
3281 the size of a pointer then a zero extension is done. If they are the same size,
3282 nothing is done (<i>no-op cast</i>).</p>
3286 %X = inttoptr i32 255 to i32* <i>; yields zero extension on 64-bit architecture</i>
3287 %X = inttoptr i32 255 to i32* <i>; yields no-op on 32-bit architecture</i>
3288 %Y = inttoptr i64 0 to i32* <i>; yields truncation on 32-bit architecture</i>
3292 <!-- _______________________________________________________________________ -->
3293 <div class="doc_subsubsection">
3294 <a name="i_bitcast">'<tt>bitcast .. to</tt>' Instruction</a>
3296 <div class="doc_text">
3300 <result> = bitcast <ty> <value> to <ty2> <i>; yields ty2</i>
3304 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3305 <tt>ty2</tt> without changing any bits.</p>
3308 <p>The '<tt>bitcast</tt>' instruction takes a value to cast, which must be
3309 a first class value, and a type to cast it to, which must also be a <a
3310 href="#t_firstclass">first class</a> type. The bit sizes of <tt>value</tt>
3311 and the destination type, <tt>ty2</tt>, must be identical. If the source
3312 type is a pointer, the destination type must also be a pointer.</p>
3315 <p>The '<tt>bitcast</tt>' instruction converts <tt>value</tt> to type
3316 <tt>ty2</tt>. It is always a <i>no-op cast</i> because no bits change with
3317 this conversion. The conversion is done as if the <tt>value</tt> had been
3318 stored to memory and read back as type <tt>ty2</tt>. Pointer types may only be
3319 converted to other pointer types with this instruction. To convert pointers to
3320 other types, use the <a href="#i_inttoptr">inttoptr</a> or
3321 <a href="#i_ptrtoint">ptrtoint</a> instructions first.</p>
3325 %X = bitcast i8 255 to i8 <i>; yields i8 :-1</i>
3326 %Y = bitcast i32* %x to sint* <i>; yields sint*:%x</i>
3327 %Z = bitcast <2xint> %V to i64; <i>; yields i64: %V</i>
3331 <!-- ======================================================================= -->
3332 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
3333 <div class="doc_text">
3334 <p>The instructions in this category are the "miscellaneous"
3335 instructions, which defy better classification.</p>
3338 <!-- _______________________________________________________________________ -->
3339 <div class="doc_subsubsection"><a name="i_icmp">'<tt>icmp</tt>' Instruction</a>
3341 <div class="doc_text">
3343 <pre> <result> = icmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
3346 <p>The '<tt>icmp</tt>' instruction returns a boolean value based on comparison
3347 of its two integer operands.</p>
3349 <p>The '<tt>icmp</tt>' instruction takes three operands. The first operand is
3350 the condition code indicating the kind of comparison to perform. It is not
3351 a value, just a keyword. The possible condition code are:
3353 <li><tt>eq</tt>: equal</li>
3354 <li><tt>ne</tt>: not equal </li>
3355 <li><tt>ugt</tt>: unsigned greater than</li>
3356 <li><tt>uge</tt>: unsigned greater or equal</li>
3357 <li><tt>ult</tt>: unsigned less than</li>
3358 <li><tt>ule</tt>: unsigned less or equal</li>
3359 <li><tt>sgt</tt>: signed greater than</li>
3360 <li><tt>sge</tt>: signed greater or equal</li>
3361 <li><tt>slt</tt>: signed less than</li>
3362 <li><tt>sle</tt>: signed less or equal</li>
3364 <p>The remaining two arguments must be <a href="#t_integer">integer</a> or
3365 <a href="#t_pointer">pointer</a> typed. They must also be identical types.</p>
3367 <p>The '<tt>icmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3368 the condition code given as <tt>cond</tt>. The comparison performed always
3369 yields a <a href="#t_primitive">i1</a> result, as follows:
3371 <li><tt>eq</tt>: yields <tt>true</tt> if the operands are equal,
3372 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3374 <li><tt>ne</tt>: yields <tt>true</tt> if the operands are unequal,
3375 <tt>false</tt> otherwise. No sign interpretation is necessary or performed.
3376 <li><tt>ugt</tt>: interprets the operands as unsigned values and yields
3377 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3378 <li><tt>uge</tt>: interprets the operands as unsigned values and yields
3379 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3380 <li><tt>ult</tt>: interprets the operands as unsigned values and yields
3381 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3382 <li><tt>ule</tt>: interprets the operands as unsigned values and yields
3383 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3384 <li><tt>sgt</tt>: interprets the operands as signed values and yields
3385 <tt>true</tt> if <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3386 <li><tt>sge</tt>: interprets the operands as signed values and yields
3387 <tt>true</tt> if <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3388 <li><tt>slt</tt>: interprets the operands as signed values and yields
3389 <tt>true</tt> if <tt>var1</tt> is less than <tt>var2</tt>.</li>
3390 <li><tt>sle</tt>: interprets the operands as signed values and yields
3391 <tt>true</tt> if <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3393 <p>If the operands are <a href="#t_pointer">pointer</a> typed, the pointer
3394 values are compared as if they were integers.</p>
3397 <pre> <result> = icmp eq i32 4, 5 <i>; yields: result=false</i>
3398 <result> = icmp ne float* %X, %X <i>; yields: result=false</i>
3399 <result> = icmp ult i16 4, 5 <i>; yields: result=true</i>
3400 <result> = icmp sgt i16 4, 5 <i>; yields: result=false</i>
3401 <result> = icmp ule i16 -4, 5 <i>; yields: result=false</i>
3402 <result> = icmp sge i16 4, 5 <i>; yields: result=false</i>
3406 <!-- _______________________________________________________________________ -->
3407 <div class="doc_subsubsection"><a name="i_fcmp">'<tt>fcmp</tt>' Instruction</a>
3409 <div class="doc_text">
3411 <pre> <result> = fcmp <cond> <ty> <var1>, <var2> <i>; yields {i1}:result</i>
3414 <p>The '<tt>fcmp</tt>' instruction returns a boolean value based on comparison
3415 of its floating point operands.</p>
3417 <p>The '<tt>fcmp</tt>' instruction takes three operands. The first operand is
3418 the condition code indicating the kind of comparison to perform. It is not
3419 a value, just a keyword. The possible condition code are:
3421 <li><tt>false</tt>: no comparison, always returns false</li>
3422 <li><tt>oeq</tt>: ordered and equal</li>
3423 <li><tt>ogt</tt>: ordered and greater than </li>
3424 <li><tt>oge</tt>: ordered and greater than or equal</li>
3425 <li><tt>olt</tt>: ordered and less than </li>
3426 <li><tt>ole</tt>: ordered and less than or equal</li>
3427 <li><tt>one</tt>: ordered and not equal</li>
3428 <li><tt>ord</tt>: ordered (no nans)</li>
3429 <li><tt>ueq</tt>: unordered or equal</li>
3430 <li><tt>ugt</tt>: unordered or greater than </li>
3431 <li><tt>uge</tt>: unordered or greater than or equal</li>
3432 <li><tt>ult</tt>: unordered or less than </li>
3433 <li><tt>ule</tt>: unordered or less than or equal</li>
3434 <li><tt>une</tt>: unordered or not equal</li>
3435 <li><tt>uno</tt>: unordered (either nans)</li>
3436 <li><tt>true</tt>: no comparison, always returns true</li>
3438 <p><i>Ordered</i> means that neither operand is a QNAN while
3439 <i>unordered</i> means that either operand may be a QNAN.</p>
3440 <p>The <tt>val1</tt> and <tt>val2</tt> arguments must be
3441 <a href="#t_floating">floating point</a> typed. They must have identical
3444 <p>The '<tt>fcmp</tt>' compares <tt>var1</tt> and <tt>var2</tt> according to
3445 the condition code given as <tt>cond</tt>. The comparison performed always
3446 yields a <a href="#t_primitive">i1</a> result, as follows:
3448 <li><tt>false</tt>: always yields <tt>false</tt>, regardless of operands.</li>
3449 <li><tt>oeq</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3450 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3451 <li><tt>ogt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3452 <tt>var1</tt> is greather than <tt>var2</tt>.</li>
3453 <li><tt>oge</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3454 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3455 <li><tt>olt</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3456 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3457 <li><tt>ole</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3458 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3459 <li><tt>one</tt>: yields <tt>true</tt> if both operands are not a QNAN and
3460 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3461 <li><tt>ord</tt>: yields <tt>true</tt> if both operands are not a QNAN.</li>
3462 <li><tt>ueq</tt>: yields <tt>true</tt> if either operand is a QNAN or
3463 <tt>var1</tt> is equal to <tt>var2</tt>.</li>
3464 <li><tt>ugt</tt>: yields <tt>true</tt> if either operand is a QNAN or
3465 <tt>var1</tt> is greater than <tt>var2</tt>.</li>
3466 <li><tt>uge</tt>: yields <tt>true</tt> if either operand is a QNAN or
3467 <tt>var1</tt> is greater than or equal to <tt>var2</tt>.</li>
3468 <li><tt>ult</tt>: yields <tt>true</tt> if either operand is a QNAN or
3469 <tt>var1</tt> is less than <tt>var2</tt>.</li>
3470 <li><tt>ule</tt>: yields <tt>true</tt> if either operand is a QNAN or
3471 <tt>var1</tt> is less than or equal to <tt>var2</tt>.</li>
3472 <li><tt>une</tt>: yields <tt>true</tt> if either operand is a QNAN or
3473 <tt>var1</tt> is not equal to <tt>var2</tt>.</li>
3474 <li><tt>uno</tt>: yields <tt>true</tt> if either operand is a QNAN.</li>
3475 <li><tt>true</tt>: always yields <tt>true</tt>, regardless of operands.</li>
3479 <pre> <result> = fcmp oeq float 4.0, 5.0 <i>; yields: result=false</i>
3480 <result> = icmp one float 4.0, 5.0 <i>; yields: result=true</i>
3481 <result> = icmp olt float 4.0, 5.0 <i>; yields: result=true</i>
3482 <result> = icmp ueq double 1.0, 2.0 <i>; yields: result=false</i>
3486 <!-- _______________________________________________________________________ -->
3487 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
3488 Instruction</a> </div>
3489 <div class="doc_text">
3491 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
3493 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
3494 the SSA graph representing the function.</p>
3496 <p>The type of the incoming values is specified with the first type
3497 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
3498 as arguments, with one pair for each predecessor basic block of the
3499 current block. Only values of <a href="#t_firstclass">first class</a>
3500 type may be used as the value arguments to the PHI node. Only labels
3501 may be used as the label arguments.</p>
3502 <p>There must be no non-phi instructions between the start of a basic
3503 block and the PHI instructions: i.e. PHI instructions must be first in
3506 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the value
3507 specified by the pair corresponding to the predecessor basic block that executed
3508 just prior to the current block.</p>
3510 <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>
3513 <!-- _______________________________________________________________________ -->
3514 <div class="doc_subsubsection">
3515 <a name="i_select">'<tt>select</tt>' Instruction</a>
3518 <div class="doc_text">
3523 <result> = select i1 <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
3529 The '<tt>select</tt>' instruction is used to choose one value based on a
3530 condition, without branching.
3537 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.
3543 If the boolean condition evaluates to true, the instruction returns the first
3544 value argument; otherwise, it returns the second value argument.
3550 %X = select i1 true, i8 17, i8 42 <i>; yields i8:17</i>
3555 <!-- _______________________________________________________________________ -->
3556 <div class="doc_subsubsection">
3557 <a name="i_call">'<tt>call</tt>' Instruction</a>
3560 <div class="doc_text">
3564 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
3569 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
3573 <p>This instruction requires several arguments:</p>
3577 <p>The optional "tail" marker indicates whether the callee function accesses
3578 any allocas or varargs in the caller. If the "tail" marker is present, the
3579 function call is eligible for tail call optimization. Note that calls may
3580 be marked "tail" even if they do not occur before a <a
3581 href="#i_ret"><tt>ret</tt></a> instruction.
3584 <p>The optional "cconv" marker indicates which <a href="#callingconv">calling
3585 convention</a> the call should use. If none is specified, the call defaults
3586 to using C calling conventions.
3589 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
3590 being invoked. The argument types must match the types implied by this
3591 signature. This type can be omitted if the function is not varargs and
3592 if the function type does not return a pointer to a function.</p>
3595 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
3596 be invoked. In most cases, this is a direct function invocation, but
3597 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
3598 to function value.</p>
3601 <p>'<tt>function args</tt>': argument list whose types match the
3602 function signature argument types. All arguments must be of
3603 <a href="#t_firstclass">first class</a> type. If the function signature
3604 indicates the function accepts a variable number of arguments, the extra
3605 arguments can be specified.</p>
3611 <p>The '<tt>call</tt>' instruction is used to cause control flow to
3612 transfer to a specified function, with its incoming arguments bound to
3613 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
3614 instruction in the called function, control flow continues with the
3615 instruction after the function call, and the return value of the
3616 function is bound to the result argument. This is a simpler case of
3617 the <a href="#i_invoke">invoke</a> instruction.</p>
3622 %retval = call i32 %test(i32 %argc)
3623 call i32(i8 *, ...) *%printf(i8 * %msg, i32 12, i8 42);
3624 %X = tail call i32 %foo()
3625 %Y = tail call <a href="#callingconv">fastcc</a> i32 %foo()
3630 <!-- _______________________________________________________________________ -->
3631 <div class="doc_subsubsection">
3632 <a name="i_va_arg">'<tt>va_arg</tt>' Instruction</a>
3635 <div class="doc_text">
3640 <resultval> = va_arg <va_list*> <arglist>, <argty>
3645 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
3646 the "variable argument" area of a function call. It is used to implement the
3647 <tt>va_arg</tt> macro in C.</p>
3651 <p>This instruction takes a <tt>va_list*</tt> value and the type of
3652 the argument. It returns a value of the specified argument type and
3653 increments the <tt>va_list</tt> to point to the next argument. The
3654 actual type of <tt>va_list</tt> is target specific.</p>
3658 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
3659 type from the specified <tt>va_list</tt> and causes the
3660 <tt>va_list</tt> to point to the next argument. For more information,
3661 see the variable argument handling <a href="#int_varargs">Intrinsic
3664 <p>It is legal for this instruction to be called in a function which does not
3665 take a variable number of arguments, for example, the <tt>vfprintf</tt>
3668 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
3669 href="#intrinsics">intrinsic function</a> because it takes a type as an
3674 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
3678 <!-- *********************************************************************** -->
3679 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
3680 <!-- *********************************************************************** -->
3682 <div class="doc_text">
3684 <p>LLVM supports the notion of an "intrinsic function". These functions have
3685 well known names and semantics and are required to follow certain restrictions.
3686 Overall, these intrinsics represent an extension mechanism for the LLVM
3687 language that does not require changing all of the transformations in LLVM when
3688 adding to the language (or the bitcode reader/writer, the parser, etc...).</p>
3690 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
3691 prefix is reserved in LLVM for intrinsic names; thus, function names may not
3692 begin with this prefix. Intrinsic functions must always be external functions:
3693 you cannot define the body of intrinsic functions. Intrinsic functions may
3694 only be used in call or invoke instructions: it is illegal to take the address
3695 of an intrinsic function. Additionally, because intrinsic functions are part
3696 of the LLVM language, it is required if any are added that they be documented
3699 <p>Some intrinsic functions can be overloaded, i.e., the intrinsic represents
3700 a family of functions that perform the same operation but on different data
3701 types. This is most frequent with the integer types. Since LLVM can represent
3702 over 8 million different integer types, there is a way to declare an intrinsic
3703 that can be overloaded based on its arguments. Such an intrinsic will have the
3704 names of its argument types encoded into its function name, each
3705 preceded by a period. For example, the <tt>llvm.ctpop</tt> function can take an
3706 integer of any width. This leads to a family of functions such as
3707 <tt>i32 @llvm.ctpop.i8(i8 %val)</tt> and <tt>i32 @llvm.ctpop.i29(i29 %val)</tt>.
3711 <p>To learn how to add an intrinsic function, please see the
3712 <a href="ExtendingLLVM.html">Extending LLVM Guide</a>.
3717 <!-- ======================================================================= -->
3718 <div class="doc_subsection">
3719 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
3722 <div class="doc_text">
3724 <p>Variable argument support is defined in LLVM with the <a
3725 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three
3726 intrinsic functions. These functions are related to the similarly
3727 named macros defined in the <tt><stdarg.h></tt> header file.</p>
3729 <p>All of these functions operate on arguments that use a
3730 target-specific value type "<tt>va_list</tt>". The LLVM assembly
3731 language reference manual does not define what this type is, so all
3732 transformations should be prepared to handle these functions regardless of
3735 <p>This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a>
3736 instruction and the variable argument handling intrinsic functions are
3739 <div class="doc_code">
3741 define i32 @test(i32 %X, ...) {
3742 ; Initialize variable argument processing
3744 %ap2 = bitcast i8** %ap to i8*
3745 call void @llvm.va_start(i8* %ap2)
3747 ; Read a single integer argument
3748 %tmp = va_arg i8** %ap, i32
3750 ; Demonstrate usage of llvm.va_copy and llvm.va_end
3752 %aq2 = bitcast i8** %aq to i8*
3753 call void @llvm.va_copy(i8* %aq2, i8* %ap2)
3754 call void @llvm.va_end(i8* %aq2)
3756 ; Stop processing of arguments.
3757 call void @llvm.va_end(i8* %ap2)
3761 declare void @llvm.va_start(i8*)
3762 declare void @llvm.va_copy(i8*, i8*)
3763 declare void @llvm.va_end(i8*)
3769 <!-- _______________________________________________________________________ -->
3770 <div class="doc_subsubsection">
3771 <a name="int_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
3775 <div class="doc_text">
3777 <pre> declare void %llvm.va_start(i8* <arglist>)<br></pre>
3779 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
3780 <tt>*<arglist></tt> for subsequent use by <tt><a
3781 href="#i_va_arg">va_arg</a></tt>.</p>
3785 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
3789 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
3790 macro available in C. In a target-dependent way, it initializes the
3791 <tt>va_list</tt> element to which the argument points, so that the next call to
3792 <tt>va_arg</tt> will produce the first variable argument passed to the function.
3793 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
3794 last argument of the function as the compiler can figure that out.</p>
3798 <!-- _______________________________________________________________________ -->
3799 <div class="doc_subsubsection">
3800 <a name="int_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
3803 <div class="doc_text">
3805 <pre> declare void @llvm.va_end(i8* <arglist>)<br></pre>
3808 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt>,
3809 which has been initialized previously with <tt><a href="#int_va_start">llvm.va_start</a></tt>
3810 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
3814 <p>The argument is a pointer to a <tt>va_list</tt> to destroy.</p>
3818 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
3819 macro available in C. In a target-dependent way, it destroys the
3820 <tt>va_list</tt> element to which the argument points. Calls to <a
3821 href="#int_va_start"><tt>llvm.va_start</tt></a> and <a href="#int_va_copy">
3822 <tt>llvm.va_copy</tt></a> must be matched exactly with calls to
3823 <tt>llvm.va_end</tt>.</p>
3827 <!-- _______________________________________________________________________ -->
3828 <div class="doc_subsubsection">
3829 <a name="int_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
3832 <div class="doc_text">
3837 declare void @llvm.va_copy(i8* <destarglist>, i8* <srcarglist>)
3842 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
3843 from the source argument list to the destination argument list.</p>
3847 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
3848 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
3853 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
3854 macro available in C. In a target-dependent way, it copies the source
3855 <tt>va_list</tt> element into the destination <tt>va_list</tt> element. This
3856 intrinsic is necessary because the <tt><a href="#int_va_start">
3857 llvm.va_start</a></tt> intrinsic may be arbitrarily complex and require, for
3858 example, memory allocation.</p>
3862 <!-- ======================================================================= -->
3863 <div class="doc_subsection">
3864 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
3867 <div class="doc_text">
3870 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
3871 Collection</a> requires the implementation and generation of these intrinsics.
3872 These intrinsics allow identification of <a href="#int_gcroot">GC roots on the
3873 stack</a>, as well as garbage collector implementations that require <a
3874 href="#int_gcread">read</a> and <a href="#int_gcwrite">write</a> barriers.
3875 Front-ends for type-safe garbage collected languages should generate these
3876 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
3877 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
3881 <!-- _______________________________________________________________________ -->
3882 <div class="doc_subsubsection">
3883 <a name="int_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
3886 <div class="doc_text">
3891 declare void @llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
3896 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
3897 the code generator, and allows some metadata to be associated with it.</p>
3901 <p>The first argument specifies the address of a stack object that contains the
3902 root pointer. The second pointer (which must be either a constant or a global
3903 value address) contains the meta-data to be associated with the root.</p>
3907 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
3908 location. At compile-time, the code generator generates information to allow
3909 the runtime to find the pointer at GC safe points.
3915 <!-- _______________________________________________________________________ -->
3916 <div class="doc_subsubsection">
3917 <a name="int_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
3920 <div class="doc_text">
3925 declare i8 * @llvm.gcread(i8 * %ObjPtr, i8 ** %Ptr)
3930 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
3931 locations, allowing garbage collector implementations that require read
3936 <p>The second argument is the address to read from, which should be an address
3937 allocated from the garbage collector. The first object is a pointer to the
3938 start of the referenced object, if needed by the language runtime (otherwise
3943 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
3944 instruction, but may be replaced with substantially more complex code by the
3945 garbage collector runtime, as needed.</p>
3950 <!-- _______________________________________________________________________ -->
3951 <div class="doc_subsubsection">
3952 <a name="int_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
3955 <div class="doc_text">
3960 declare void @llvm.gcwrite(i8 * %P1, i8 * %Obj, i8 ** %P2)
3965 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
3966 locations, allowing garbage collector implementations that require write
3967 barriers (such as generational or reference counting collectors).</p>
3971 <p>The first argument is the reference to store, the second is the start of the
3972 object to store it to, and the third is the address of the field of Obj to
3973 store to. If the runtime does not require a pointer to the object, Obj may be
3978 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
3979 instruction, but may be replaced with substantially more complex code by the
3980 garbage collector runtime, as needed.</p>
3986 <!-- ======================================================================= -->
3987 <div class="doc_subsection">
3988 <a name="int_codegen">Code Generator Intrinsics</a>
3991 <div class="doc_text">
3993 These intrinsics are provided by LLVM to expose special features that may only
3994 be implemented with code generator support.
3999 <!-- _______________________________________________________________________ -->
4000 <div class="doc_subsubsection">
4001 <a name="int_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
4004 <div class="doc_text">
4008 declare i8 *@llvm.returnaddress(i32 <level>)
4014 The '<tt>llvm.returnaddress</tt>' intrinsic attempts to compute a
4015 target-specific value indicating the return address of the current function
4016 or one of its callers.
4022 The argument to this intrinsic indicates which function to return the address
4023 for. Zero indicates the calling function, one indicates its caller, etc. The
4024 argument is <b>required</b> to be a constant integer value.
4030 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
4031 the return address of the specified call frame, or zero if it cannot be
4032 identified. The value returned by this intrinsic is likely to be incorrect or 0
4033 for arguments other than zero, so it should only be used for debugging purposes.
4037 Note that calling this intrinsic does not prevent function inlining or other
4038 aggressive transformations, so the value returned may not be that of the obvious
4039 source-language caller.
4044 <!-- _______________________________________________________________________ -->
4045 <div class="doc_subsubsection">
4046 <a name="int_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
4049 <div class="doc_text">
4053 declare i8 *@llvm.frameaddress(i32 <level>)
4059 The '<tt>llvm.frameaddress</tt>' intrinsic attempts to return the
4060 target-specific frame pointer value for the specified stack frame.
4066 The argument to this intrinsic indicates which function to return the frame
4067 pointer for. Zero indicates the calling function, one indicates its caller,
4068 etc. The argument is <b>required</b> to be a constant integer value.
4074 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
4075 the frame address of the specified call frame, or zero if it cannot be
4076 identified. The value returned by this intrinsic is likely to be incorrect or 0
4077 for arguments other than zero, so it should only be used for debugging purposes.
4081 Note that calling this intrinsic does not prevent function inlining or other
4082 aggressive transformations, so the value returned may not be that of the obvious
4083 source-language caller.
4087 <!-- _______________________________________________________________________ -->
4088 <div class="doc_subsubsection">
4089 <a name="int_stacksave">'<tt>llvm.stacksave</tt>' Intrinsic</a>
4092 <div class="doc_text">
4096 declare i8 *@llvm.stacksave()
4102 The '<tt>llvm.stacksave</tt>' intrinsic is used to remember the current state of
4103 the function stack, for use with <a href="#int_stackrestore">
4104 <tt>llvm.stackrestore</tt></a>. This is useful for implementing language
4105 features like scoped automatic variable sized arrays in C99.
4111 This intrinsic returns a opaque pointer value that can be passed to <a
4112 href="#int_stackrestore"><tt>llvm.stackrestore</tt></a>. When an
4113 <tt>llvm.stackrestore</tt> intrinsic is executed with a value saved from
4114 <tt>llvm.stacksave</tt>, it effectively restores the state of the stack to the
4115 state it was in when the <tt>llvm.stacksave</tt> intrinsic executed. In
4116 practice, this pops any <a href="#i_alloca">alloca</a> blocks from the stack
4117 that were allocated after the <tt>llvm.stacksave</tt> was executed.
4122 <!-- _______________________________________________________________________ -->
4123 <div class="doc_subsubsection">
4124 <a name="int_stackrestore">'<tt>llvm.stackrestore</tt>' Intrinsic</a>
4127 <div class="doc_text">
4131 declare void @llvm.stackrestore(i8 * %ptr)
4137 The '<tt>llvm.stackrestore</tt>' intrinsic is used to restore the state of
4138 the function stack to the state it was in when the corresponding <a
4139 href="#int_stacksave"><tt>llvm.stacksave</tt></a> intrinsic executed. This is
4140 useful for implementing language features like scoped automatic variable sized
4147 See the description for <a href="#int_stacksave"><tt>llvm.stacksave</tt></a>.
4153 <!-- _______________________________________________________________________ -->
4154 <div class="doc_subsubsection">
4155 <a name="int_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
4158 <div class="doc_text">
4162 declare void @llvm.prefetch(i8 * <address>,
4163 i32 <rw>, i32 <locality>)
4170 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
4171 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
4173 effect on the behavior of the program but can change its performance
4180 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
4181 determining if the fetch should be for a read (0) or write (1), and
4182 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
4183 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
4184 <tt>locality</tt> arguments must be constant integers.
4190 This intrinsic does not modify the behavior of the program. In particular,
4191 prefetches cannot trap and do not produce a value. On targets that support this
4192 intrinsic, the prefetch can provide hints to the processor cache for better
4198 <!-- _______________________________________________________________________ -->
4199 <div class="doc_subsubsection">
4200 <a name="int_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
4203 <div class="doc_text">
4207 declare void @llvm.pcmarker( i32 <id> )
4214 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
4216 code to simulators and other tools. The method is target specific, but it is
4217 expected that the marker will use exported symbols to transmit the PC of the marker.
4218 The marker makes no guarantees that it will remain with any specific instruction
4219 after optimizations. It is possible that the presence of a marker will inhibit
4220 optimizations. The intended use is to be inserted after optimizations to allow
4221 correlations of simulation runs.
4227 <tt>id</tt> is a numerical id identifying the marker.
4233 This intrinsic does not modify the behavior of the program. Backends that do not
4234 support this intrinisic may ignore it.
4239 <!-- _______________________________________________________________________ -->
4240 <div class="doc_subsubsection">
4241 <a name="int_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
4244 <div class="doc_text">
4248 declare i64 @llvm.readcyclecounter( )
4255 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
4256 counter register (or similar low latency, high accuracy clocks) on those targets
4257 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
4258 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
4259 should only be used for small timings.
4265 When directly supported, reading the cycle counter should not modify any memory.
4266 Implementations are allowed to either return a application specific value or a
4267 system wide value. On backends without support, this is lowered to a constant 0.
4272 <!-- ======================================================================= -->
4273 <div class="doc_subsection">
4274 <a name="int_libc">Standard C Library Intrinsics</a>
4277 <div class="doc_text">
4279 LLVM provides intrinsics for a few important standard C library functions.
4280 These intrinsics allow source-language front-ends to pass information about the
4281 alignment of the pointer arguments to the code generator, providing opportunity
4282 for more efficient code generation.
4287 <!-- _______________________________________________________________________ -->
4288 <div class="doc_subsubsection">
4289 <a name="int_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
4292 <div class="doc_text">
4296 declare void @llvm.memcpy.i32(i8 * <dest>, i8 * <src>,
4297 i32 <len>, i32 <align>)
4298 declare void @llvm.memcpy.i64(i8 * <dest>, i8 * <src>,
4299 i64 <len>, i32 <align>)
4305 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4306 location to the destination location.
4310 Note that, unlike the standard libc function, the <tt>llvm.memcpy.*</tt>
4311 intrinsics do not return a value, and takes an extra alignment argument.
4317 The first argument is a pointer to the destination, the second is a pointer to
4318 the source. The third argument is an integer argument
4319 specifying the number of bytes to copy, and the fourth argument is the alignment
4320 of the source and destination locations.
4324 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4325 the caller guarantees that both the source and destination pointers are aligned
4332 The '<tt>llvm.memcpy.*</tt>' intrinsics copy a block of memory from the source
4333 location to the destination location, which are not allowed to overlap. It
4334 copies "len" bytes of memory over. If the argument is known to be aligned to
4335 some boundary, this can be specified as the fourth argument, otherwise it should
4341 <!-- _______________________________________________________________________ -->
4342 <div class="doc_subsubsection">
4343 <a name="int_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
4346 <div class="doc_text">
4350 declare void @llvm.memmove.i32(i8 * <dest>, i8 * <src>,
4351 i32 <len>, i32 <align>)
4352 declare void @llvm.memmove.i64(i8 * <dest>, i8 * <src>,
4353 i64 <len>, i32 <align>)
4359 The '<tt>llvm.memmove.*</tt>' intrinsics move a block of memory from the source
4360 location to the destination location. It is similar to the
4361 '<tt>llvm.memcmp</tt>' intrinsic but allows the two memory locations to overlap.
4365 Note that, unlike the standard libc function, the <tt>llvm.memmove.*</tt>
4366 intrinsics do not return a value, and takes an extra alignment argument.
4372 The first argument is a pointer to the destination, the second is a pointer to
4373 the source. The third argument is an integer argument
4374 specifying the number of bytes to copy, and the fourth argument is the alignment
4375 of the source and destination locations.
4379 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4380 the caller guarantees that the source and destination pointers are aligned to
4387 The '<tt>llvm.memmove.*</tt>' intrinsics copy a block of memory from the source
4388 location to the destination location, which may overlap. It
4389 copies "len" bytes of memory over. If the argument is known to be aligned to
4390 some boundary, this can be specified as the fourth argument, otherwise it should
4396 <!-- _______________________________________________________________________ -->
4397 <div class="doc_subsubsection">
4398 <a name="int_memset">'<tt>llvm.memset.*</tt>' Intrinsics</a>
4401 <div class="doc_text">
4405 declare void @llvm.memset.i32(i8 * <dest>, i8 <val>,
4406 i32 <len>, i32 <align>)
4407 declare void @llvm.memset.i64(i8 * <dest>, i8 <val>,
4408 i64 <len>, i32 <align>)
4414 The '<tt>llvm.memset.*</tt>' intrinsics fill a block of memory with a particular
4419 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
4420 does not return a value, and takes an extra alignment argument.
4426 The first argument is a pointer to the destination to fill, the second is the
4427 byte value to fill it with, the third argument is an integer
4428 argument specifying the number of bytes to fill, and the fourth argument is the
4429 known alignment of destination location.
4433 If the call to this intrinisic has an alignment value that is not 0 or 1, then
4434 the caller guarantees that the destination pointer is aligned to that boundary.
4440 The '<tt>llvm.memset.*</tt>' intrinsics fill "len" bytes of memory starting at
4442 destination location. If the argument is known to be aligned to some boundary,
4443 this can be specified as the fourth argument, otherwise it should be set to 0 or
4449 <!-- _______________________________________________________________________ -->
4450 <div class="doc_subsubsection">
4451 <a name="int_sqrt">'<tt>llvm.sqrt.*</tt>' Intrinsic</a>
4454 <div class="doc_text">
4458 declare float @llvm.sqrt.f32(float %Val)
4459 declare double @llvm.sqrt.f64(double %Val)
4465 The '<tt>llvm.sqrt</tt>' intrinsics return the sqrt of the specified operand,
4466 returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
4467 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
4468 negative numbers (which allows for better optimization).
4474 The argument and return value are floating point numbers of the same type.
4480 This function returns the sqrt of the specified operand if it is a nonnegative
4481 floating point number.
4485 <!-- _______________________________________________________________________ -->
4486 <div class="doc_subsubsection">
4487 <a name="int_powi">'<tt>llvm.powi.*</tt>' Intrinsic</a>
4490 <div class="doc_text">
4494 declare float @llvm.powi.f32(float %Val, i32 %power)
4495 declare double @llvm.powi.f64(double %Val, i32 %power)
4501 The '<tt>llvm.powi.*</tt>' intrinsics return the first operand raised to the
4502 specified (positive or negative) power. The order of evaluation of
4503 multiplications is not defined.
4509 The second argument is an integer power, and the first is a value to raise to
4516 This function returns the first value raised to the second power with an
4517 unspecified sequence of rounding operations.</p>
4521 <!-- ======================================================================= -->
4522 <div class="doc_subsection">
4523 <a name="int_manip">Bit Manipulation Intrinsics</a>
4526 <div class="doc_text">
4528 LLVM provides intrinsics for a few important bit manipulation operations.
4529 These allow efficient code generation for some algorithms.
4534 <!-- _______________________________________________________________________ -->
4535 <div class="doc_subsubsection">
4536 <a name="int_bswap">'<tt>llvm.bswap.*</tt>' Intrinsics</a>
4539 <div class="doc_text">
4542 <p>This is an overloaded intrinsic function. You can use bswap on any integer
4543 type that is an even number of bytes (i.e. BitWidth % 16 == 0). Note the suffix
4544 that includes the type for the result and the operand.
4546 declare i16 @llvm.bswap.i16.i16(i16 <id>)
4547 declare i32 @llvm.bswap.i32.i32(i32 <id>)
4548 declare i64 @llvm.bswap.i64.i64(i64 <id>)
4554 The '<tt>llvm.bswap</tt>' family of intrinsics is used to byte swap integer
4555 values with an even number of bytes (positive multiple of 16 bits). These are
4556 useful for performing operations on data that is not in the target's native
4563 The <tt>llvm.bswap.16.i16</tt> intrinsic returns an i16 value that has the high
4564 and low byte of the input i16 swapped. Similarly, the <tt>llvm.bswap.i32</tt>
4565 intrinsic returns an i32 value that has the four bytes of the input i32
4566 swapped, so that if the input bytes are numbered 0, 1, 2, 3 then the returned
4567 i32 will have its bytes in 3, 2, 1, 0 order. The <tt>llvm.bswap.i48.i48</tt>,
4568 <tt>llvm.bswap.i64.i64</tt> and other intrinsics extend this concept to
4569 additional even-byte lengths (6 bytes, 8 bytes and more, respectively).
4574 <!-- _______________________________________________________________________ -->
4575 <div class="doc_subsubsection">
4576 <a name="int_ctpop">'<tt>llvm.ctpop.*</tt>' Intrinsic</a>
4579 <div class="doc_text">
4582 <p>This is an overloaded intrinsic. You can use llvm.ctpop on any integer bit
4583 width. Not all targets support all bit widths however.
4585 declare i32 @llvm.ctpop.i8 (i8 <src>)
4586 declare i32 @llvm.ctpop.i16(i16 <src>)
4587 declare i32 @llvm.ctpop.i32(i32 <src>)
4588 declare i32 @llvm.ctpop.i64(i64 <src>)
4589 declare i32 @llvm.ctpop.i256(i256 <src>)
4595 The '<tt>llvm.ctpop</tt>' family of intrinsics counts the number of bits set in a
4602 The only argument is the value to be counted. The argument may be of any
4603 integer type. The return type must match the argument type.
4609 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
4613 <!-- _______________________________________________________________________ -->
4614 <div class="doc_subsubsection">
4615 <a name="int_ctlz">'<tt>llvm.ctlz.*</tt>' Intrinsic</a>
4618 <div class="doc_text">
4621 <p>This is an overloaded intrinsic. You can use <tt>llvm.ctlz</tt> on any
4622 integer bit width. Not all targets support all bit widths however.
4624 declare i32 @llvm.ctlz.i8 (i8 <src>)
4625 declare i32 @llvm.ctlz.i16(i16 <src>)
4626 declare i32 @llvm.ctlz.i32(i32 <src>)
4627 declare i32 @llvm.ctlz.i64(i64 <src>)
4628 declare i32 @llvm.ctlz.i256(i256 <src>)
4634 The '<tt>llvm.ctlz</tt>' family of intrinsic functions counts the number of
4635 leading zeros in a variable.
4641 The only argument is the value to be counted. The argument may be of any
4642 integer type. The return type must match the argument type.
4648 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
4649 in a variable. If the src == 0 then the result is the size in bits of the type
4650 of src. For example, <tt>llvm.ctlz(i32 2) = 30</tt>.
4656 <!-- _______________________________________________________________________ -->
4657 <div class="doc_subsubsection">
4658 <a name="int_cttz">'<tt>llvm.cttz.*</tt>' Intrinsic</a>
4661 <div class="doc_text">
4664 <p>This is an overloaded intrinsic. You can use <tt>llvm.cttz</tt> on any
4665 integer bit width. Not all targets support all bit widths however.
4667 declare i32 @llvm.cttz.i8 (i8 <src>)
4668 declare i32 @llvm.cttz.i16(i16 <src>)
4669 declare i32 @llvm.cttz.i32(i32 <src>)
4670 declare i32 @llvm.cttz.i64(i64 <src>)
4671 declare i32 @llvm.cttz.i256(i256 <src>)
4677 The '<tt>llvm.cttz</tt>' family of intrinsic functions counts the number of
4684 The only argument is the value to be counted. The argument may be of any
4685 integer type. The return type must match the argument type.
4691 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
4692 in a variable. If the src == 0 then the result is the size in bits of the type
4693 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
4697 <!-- _______________________________________________________________________ -->
4698 <div class="doc_subsubsection">
4699 <a name="int_part_select">'<tt>llvm.part.select.*</tt>' Intrinsic</a>
4702 <div class="doc_text">
4705 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.select</tt>
4706 on any integer bit width.
4708 declare i17 @llvm.part.select.i17.i17 (i17 %val, i32 %loBit, i32 %hiBit)
4709 declare i29 @llvm.part.select.i29.i29 (i29 %val, i32 %loBit, i32 %hiBit)
4713 <p>The '<tt>llvm.part.select</tt>' family of intrinsic functions selects a
4714 range of bits from an integer value and returns them in the same bit width as
4715 the original value.</p>
4718 <p>The first argument, <tt>%val</tt> and the result may be integer types of
4719 any bit width but they must have the same bit width. The second and third
4720 arguments must be <tt>i32</tt> type since they specify only a bit index.</p>
4723 <p>The operation of the '<tt>llvm.part.select</tt>' intrinsic has two modes
4724 of operation: forwards and reverse. If <tt>%loBit</tt> is greater than
4725 <tt>%hiBits</tt> then the intrinsic operates in reverse mode. Otherwise it
4726 operates in forward mode.</p>
4727 <p>In forward mode, this intrinsic is the equivalent of shifting <tt>%val</tt>
4728 right by <tt>%loBit</tt> bits and then ANDing it with a mask with
4729 only the <tt>%hiBit - %loBit</tt> bits set, as follows:</p>
4731 <li>The <tt>%val</tt> is shifted right (LSHR) by the number of bits specified
4732 by <tt>%loBits</tt>. This normalizes the value to the low order bits.</li>
4733 <li>The <tt>%loBits</tt> value is subtracted from the <tt>%hiBits</tt> value
4734 to determine the number of bits to retain.</li>
4735 <li>A mask of the retained bits is created by shifting a -1 value.</li>
4736 <li>The mask is ANDed with <tt>%val</tt> to produce the result.
4738 <p>In reverse mode, a similar computation is made except that the bits are
4739 returned in the reverse order. So, for example, if <tt>X</tt> has the value
4740 <tt>i16 0x0ACF (101011001111)</tt> and we apply
4741 <tt>part.select(i16 X, 8, 3)</tt> to it, we get back the value
4742 <tt>i16 0x0026 (000000100110)</tt>.</p>
4745 <div class="doc_subsubsection">
4746 <a name="int_part_set">'<tt>llvm.part.set.*</tt>' Intrinsic</a>
4749 <div class="doc_text">
4752 <p>This is an overloaded intrinsic. You can use <tt>llvm.part.set</tt>
4753 on any integer bit width.
4755 declare i17 @llvm.part.set.i17.i17.i9 (i17 %val, i9 %repl, i32 %lo, i32 %hi)
4756 declare i29 @llvm.part.set.i29.i29.i9 (i29 %val, i9 %repl, i32 %lo, i32 %hi)
4760 <p>The '<tt>llvm.part.set</tt>' family of intrinsic functions replaces a range
4761 of bits in an integer value with another integer value. It returns the integer
4762 with the replaced bits.</p>
4765 <p>The first argument, <tt>%val</tt> and the result may be integer types of
4766 any bit width but they must have the same bit width. <tt>%val</tt> is the value
4767 whose bits will be replaced. The second argument, <tt>%repl</tt> may be an
4768 integer of any bit width. The third and fourth arguments must be <tt>i32</tt>
4769 type since they specify only a bit index.</p>
4772 <p>The operation of the '<tt>llvm.part.set</tt>' intrinsic has two modes
4773 of operation: forwards and reverse. If <tt>%lo</tt> is greater than
4774 <tt>%hi</tt> then the intrinsic operates in reverse mode. Otherwise it
4775 operates in forward mode.</p>
4776 <p>For both modes, the <tt>%repl</tt> value is prepared for use by either
4777 truncating it down to the size of the replacement area or zero extending it
4778 up to that size.</p>
4779 <p>In forward mode, the bits between <tt>%lo</tt> and <tt>%hi</tt> (inclusive)
4780 are replaced with corresponding bits from <tt>%repl</tt>. That is the 0th bit
4781 in <tt>%repl</tt> replaces the <tt>%lo</tt>th bit in <tt>%val</tt> and etc. up
4782 to the <tt>%hi</tt>th bit.
4783 <p>In reverse mode, a similar computation is made except that the bits are
4784 reversed. That is, the <tt>0</tt>th bit in <tt>%repl</tt> replaces the
4785 <tt>%hi</tt> bit in <tt>%val</tt> and etc. down to the <tt>%lo</tt>th bit.
4788 llvm.part.set(0xFFFF, 0, 4, 7) -> 0xFF0F
4789 llvm.part.set(0xFFFF, 0, 7, 4) -> 0xFF0F
4790 llvm.part.set(0xFFFF, 1, 7, 4) -> 0xFF8F
4791 llvm.part.set(0xFFFF, F, 8, 3) -> 0xFFE7
4792 llvm.part.set(0xFFFF, 0, 3, 8) -> 0xFE07
4796 <!-- ======================================================================= -->
4797 <div class="doc_subsection">
4798 <a name="int_debugger">Debugger Intrinsics</a>
4801 <div class="doc_text">
4803 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
4804 are described in the <a
4805 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
4806 Debugging</a> document.
4811 <!-- ======================================================================= -->
4812 <div class="doc_subsection">
4813 <a name="int_eh">Exception Handling Intrinsics</a>
4816 <div class="doc_text">
4817 <p> The LLVM exception handling intrinsics (which all start with
4818 <tt>llvm.eh.</tt> prefix), are described in the <a
4819 href="ExceptionHandling.html#format_common_intrinsics">LLVM Exception
4820 Handling</a> document. </p>
4823 <!-- ======================================================================= -->
4824 <div class="doc_subsection">
4825 <a name="int_general">General Intrinsics</a>
4828 <div class="doc_text">
4829 <p> This class of intrinsics is designed to be generic and has
4830 no specific purpose. </p>
4833 <!-- _______________________________________________________________________ -->
4834 <div class="doc_subsubsection">
4835 <a name="int_var_annotation">'<tt>llvm.var.annotation</tt>' Intrinsic</a>
4838 <div class="doc_text">
4842 declare void @llvm.var.annotation(i8* <val>, i8* <str>, i8* <str>, i32 <int> )
4848 The '<tt>llvm.var.annotation</tt>' intrinsic
4854 The first argument is a pointer to a value, the second is a pointer to a
4855 global string, the third is a pointer to a global string which is the source
4856 file name, and the last argument is the line number.
4862 This intrinsic allows annotation of local variables with arbitrary strings.
4863 This can be useful for special purpose optimizations that want to look for these
4864 annotations. These have no other defined use, they are ignored by code
4865 generation and optimization.
4869 <!-- *********************************************************************** -->
4872 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
4873 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
4874 <a href="http://validator.w3.org/check/referer"><img
4875 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
4877 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
4878 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
4879 Last modified: $Date$