1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
2 <html><head><title>LLVM Assembly Language Reference Manual</title></head>
5 <table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
6 <tr><td> <font size=+5 color="#EEEEFF" face="Georgia,Palatino,Times,Roman"><b>LLVM Language Reference Manual</b></font></td>
10 <li><a href="#abstract">Abstract</a>
11 <li><a href="#introduction">Introduction</a>
12 <li><a href="#identifiers">Identifiers</a>
13 <li><a href="#typesystem">Type System</a>
15 <li><a href="#t_primitive">Primitive Types</a>
17 <li><a href="#t_classifications">Type Classifications</a>
19 <li><a href="#t_derived">Derived Types</a>
21 <li><a href="#t_array" >Array Type</a>
22 <li><a href="#t_function">Function Type</a>
23 <li><a href="#t_pointer">Pointer Type</a>
24 <li><a href="#t_struct" >Structure Type</a>
25 <!-- <li><a href="#t_packed" >Packed Type</a> -->
28 <li><a href="#highlevel">High Level Structure</a>
30 <li><a href="#modulestructure">Module Structure</a>
31 <li><a href="#globalvars">Global Variables</a>
32 <li><a href="#functionstructure">Function Structure</a>
34 <li><a href="#instref">Instruction Reference</a>
36 <li><a href="#terminators">Terminator Instructions</a>
38 <li><a href="#i_ret" >'<tt>ret</tt>' Instruction</a>
39 <li><a href="#i_br" >'<tt>br</tt>' Instruction</a>
40 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a>
41 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a>
42 <li><a href="#i_unwind" >'<tt>unwind</tt>' Instruction</a>
44 <li><a href="#binaryops">Binary Operations</a>
46 <li><a href="#i_add" >'<tt>add</tt>' Instruction</a>
47 <li><a href="#i_sub" >'<tt>sub</tt>' Instruction</a>
48 <li><a href="#i_mul" >'<tt>mul</tt>' Instruction</a>
49 <li><a href="#i_div" >'<tt>div</tt>' Instruction</a>
50 <li><a href="#i_rem" >'<tt>rem</tt>' Instruction</a>
51 <li><a href="#i_setcc">'<tt>set<i>cc</i></tt>' Instructions</a>
53 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
55 <li><a href="#i_and">'<tt>and</tt>' Instruction</a>
56 <li><a href="#i_or" >'<tt>or</tt>' Instruction</a>
57 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a>
58 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a>
59 <li><a href="#i_shr">'<tt>shr</tt>' Instruction</a>
61 <li><a href="#memoryops">Memory Access Operations</a>
63 <li><a href="#i_malloc" >'<tt>malloc</tt>' Instruction</a>
64 <li><a href="#i_free" >'<tt>free</tt>' Instruction</a>
65 <li><a href="#i_alloca" >'<tt>alloca</tt>' Instruction</a>
66 <li><a href="#i_load" >'<tt>load</tt>' Instruction</a>
67 <li><a href="#i_store" >'<tt>store</tt>' Instruction</a>
68 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
70 <li><a href="#otherops">Other Operations</a>
72 <li><a href="#i_phi" >'<tt>phi</tt>' Instruction</a>
73 <li><a href="#i_cast">'<tt>cast .. to</tt>' Instruction</a>
74 <li><a href="#i_call" >'<tt>call</tt>' Instruction</a>
75 <li><a href="#i_vanext">'<tt>vanext</tt>' Instruction</a>
76 <li><a href="#i_vaarg" >'<tt>vaarg</tt>' Instruction</a>
79 <li><a href="#intrinsics">Intrinsic Functions</a>
81 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
83 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
84 <li><a href="#i_va_end" >'<tt>llvm.va_end</tt>' Intrinsic</a>
85 <li><a href="#i_va_copy" >'<tt>llvm.va_copy</tt>' Intrinsic</a>
89 <p><b>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a> and <A href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></b><p>
95 <!-- *********************************************************************** -->
96 <p><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
97 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
98 <a name="abstract">Abstract
99 </b></font></td></tr></table><ul>
100 <!-- *********************************************************************** -->
103 This document is a reference manual for the LLVM assembly language. LLVM is
104 an SSA based representation that provides type safety, low-level operations,
105 flexibility, and the capability of representing 'all' high-level languages
106 cleanly. It is the common code representation used throughout all phases of
107 the LLVM compilation strategy.
113 <!-- *********************************************************************** -->
114 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
115 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
116 <a name="introduction">Introduction
117 </b></font></td></tr></table><ul>
118 <!-- *********************************************************************** -->
120 The LLVM code representation is designed to be used in three different forms: as
121 an in-memory compiler IR, as an on-disk bytecode representation (suitable for
122 fast loading by a Just-In-Time compiler), and as a human readable assembly
123 language representation. This allows LLVM to provide a powerful intermediate
124 representation for efficient compiler transformations and analysis, while
125 providing a natural means to debug and visualize the transformations. The three
126 different forms of LLVM are all equivalent. This document describes the human
127 readable representation and notation.<p>
129 The LLVM representation aims to be a light-weight and low-level while being
130 expressive, typed, and extensible at the same time. It aims to be a "universal
131 IR" of sorts, by being at a low enough level that high-level ideas may be
132 cleanly mapped to it (similar to how microprocessors are "universal IR's",
133 allowing many source languages to be mapped to them). By providing type
134 information, LLVM can be used as the target of optimizations: for example,
135 through pointer analysis, it can be proven that a C automatic variable is never
136 accessed outside of the current function... allowing it to be promoted to a
137 simple SSA value instead of a memory location.<p>
139 <!-- _______________________________________________________________________ -->
140 </ul><a name="wellformed"><h4><hr size=0>Well Formedness</h4><ul>
142 It is important to note that this document describes 'well formed' LLVM assembly
143 language. There is a difference between what the parser accepts and what is
144 considered 'well formed'. For example, the following instruction is
145 syntactically okay, but not well formed:<p>
148 %x = <a href="#i_add">add</a> int 1, %x
151 ...because the definition of <tt>%x</tt> does not dominate all of its uses. The
152 LLVM infrastructure provides a verification pass that may be used to verify that
153 an LLVM module is well formed. This pass is automatically run by the parser
154 after parsing input assembly, and by the optimizer before it outputs bytecode.
155 The violations pointed out by the verifier pass indicate bugs in transformation
156 passes or input to the parser.<p>
158 <!-- Describe the typesetting conventions here. -->
161 <!-- *********************************************************************** -->
162 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
163 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
164 <a name="identifiers">Identifiers
165 </b></font></td></tr></table><ul>
166 <!-- *********************************************************************** -->
168 LLVM uses three different forms of identifiers, for different purposes:<p>
171 <li>Numeric constants are represented as you would expect: 12, -3 123.421, etc.
172 Floating point constants have an optional hexidecimal notation.
174 <li>Named values are represented as a string of characters with a '%' prefix.
175 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
176 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'. Identifiers
177 which require other characters in their names can be surrounded with quotes. In
178 this way, anything except a <tt>"</tt> character can be used in a name.
180 <li>Unnamed values are represented as an unsigned numeric value with a '%'
181 prefix. For example, %12, %2, %44.
184 LLVM requires the values start with a '%' sign for two reasons: Compilers don't
185 need to worry about name clashes with reserved words, and the set of reserved
186 words may be expanded in the future without penalty. Additionally, unnamed
187 identifiers allow a compiler to quickly come up with a temporary variable
188 without having to avoid symbol table conflicts.<p>
190 Reserved words in LLVM are very similar to reserved words in other languages.
191 There are keywords for different opcodes ('<tt><a href="#i_add">add</a></tt>',
192 '<tt><a href="#i_cast">cast</a></tt>', '<tt><a href="#i_ret">ret</a></tt>',
193 etc...), for primitive type names ('<tt><a href="#t_void">void</a></tt>',
194 '<tt><a href="#t_uint">uint</a></tt>', etc...), and others. These reserved
195 words cannot conflict with variable names, because none of them start with a '%'
198 Here is an example of LLVM code to multiply the integer variable '<tt>%X</tt>'
203 %result = <a href="#i_mul">mul</a> uint %X, 8
206 After strength reduction:
208 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
213 <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
214 <a href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
215 %result = <a href="#i_add">add</a> uint %1, %1
218 This last way of multiplying <tt>%X</tt> by 8 illustrates several important lexical features of LLVM:<p>
221 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of line.
222 <li>Unnamed temporaries are created when the result of a computation is not
223 assigned to a named value.
224 <li>Unnamed temporaries are numbered sequentially
227 ...and it also show a convention that we follow in this document. When
228 demonstrating instructions, we will follow an instruction with a comment that
229 defines the type and name of value produced. Comments are shown in italic
232 The one non-intuitive notation for constants is the optional hexidecimal form of
233 floating point constants. For example, the form '<tt>double
234 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
235 4.5e+15</tt>' which is also supported by the parser. The only time hexadecimal
236 floating point constants are useful (and the only time that they are generated
237 by the disassembler) is when an FP constant has to be emitted that is not
238 representable as a decimal floating point number exactly. For example, NaN's,
239 infinities, and other special cases are represented in their IEEE hexadecimal
240 format so that assembly and disassembly do not cause any bits to change in the
244 <!-- *********************************************************************** -->
245 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
246 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
247 <a name="typesystem">Type System
248 </b></font></td></tr></table><ul>
249 <!-- *********************************************************************** -->
251 The LLVM type system is one of the most important features of the intermediate
252 representation. Being typed enables a number of optimizations to be performed
253 on the IR directly, without having to do extra analyses on the side before the
254 transformation. A strong type system makes it easier to read the generated code
255 and enables novel analyses and transformations that are not feasible to perform
256 on normal three address code representations.<p>
258 <!-- The written form for the type system was heavily influenced by the
259 syntactic problems with types in the C language<sup><a
260 href="#rw_stroustrup">1</a></sup>.<p> -->
264 <!-- ======================================================================= -->
265 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
266 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
267 <a name="t_primitive">Primitive Types
268 </b></font></td></tr></table><ul>
270 The primitive types are the fundemental building blocks of the LLVM system. The
271 current set of primitive types are as follows:<p>
273 <table border=0 align=center><tr><td>
275 <table border=1 cellspacing=0 cellpadding=4 align=center>
276 <tr><td><tt>void</tt></td> <td>No value</td></tr>
277 <tr><td><tt>ubyte</tt></td> <td>Unsigned 8 bit value</td></tr>
278 <tr><td><tt>ushort</tt></td><td>Unsigned 16 bit value</td></tr>
279 <tr><td><tt>uint</tt></td> <td>Unsigned 32 bit value</td></tr>
280 <tr><td><tt>ulong</tt></td> <td>Unsigned 64 bit value</td></tr>
281 <tr><td><tt>float</tt></td> <td>32 bit floating point value</td></tr>
282 <tr><td><tt>label</tt></td> <td>Branch destination</td></tr>
287 <table border=1 cellspacing=0 cellpadding=4 align=center>
288 <tr><td><tt>bool</tt></td> <td>True or False value</td></tr>
289 <tr><td><tt>sbyte</tt></td> <td>Signed 8 bit value</td></tr>
290 <tr><td><tt>short</tt></td> <td>Signed 16 bit value</td></tr>
291 <tr><td><tt>int</tt></td> <td>Signed 32 bit value</td></tr>
292 <tr><td><tt>long</tt></td> <td>Signed 64 bit value</td></tr>
293 <tr><td><tt>double</tt></td><td>64 bit floating point value</td></tr>
296 </td></tr></table><p>
300 <!-- _______________________________________________________________________ -->
301 </ul><a name="t_classifications"><h4><hr size=0>Type Classifications</h4><ul>
303 These different primitive types fall into a few useful classifications:<p>
305 <table border=1 cellspacing=0 cellpadding=4 align=center>
306 <tr><td><a name="t_signed">signed</td> <td><tt>sbyte, short, int, long, float, double</tt></td></tr>
307 <tr><td><a name="t_unsigned">unsigned</td><td><tt>ubyte, ushort, uint, ulong</tt></td></tr>
308 <tr><td><a name="t_integer">integer</td><td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td></tr>
309 <tr><td><a name="t_integral">integral</td><td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td></tr>
310 <tr><td><a name="t_floating">floating point</td><td><tt>float, double</tt></td></tr>
311 <tr><td><a name="t_firstclass">first class</td><td><tt>bool, ubyte, sbyte, ushort, short,<br> uint, int, ulong, long, float, double, <a href="#t_pointer">pointer</a></tt></td></tr>
318 <!-- ======================================================================= -->
319 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0><tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
320 <a name="t_derived">Derived Types
321 </b></font></td></tr></table><ul>
323 The real power in LLVM comes from the derived types in the system. This is what
324 allows a programmer to represent arrays, functions, pointers, and other useful
325 types. Note that these derived types may be recursive: For example, it is
326 possible to have a two dimensional array.<p>
330 <!-- _______________________________________________________________________ -->
331 </ul><a name="t_array"><h4><hr size=0>Array Type</h4><ul>
335 The array type is a very simple derived type that arranges elements sequentially
336 in memory. The array type requires a size (number of elements) and an
337 underlying data type.<p>
341 [<# elements> x <elementtype>]
344 The number of elements is a constant integer value, elementtype may be any type
349 <tt>[40 x int ]</tt>: Array of 40 integer values.<br>
350 <tt>[41 x int ]</tt>: Array of 41 integer values.<br>
351 <tt>[40 x uint]</tt>: Array of 40 unsigned integer values.<p>
354 Here are some examples of multidimensional arrays:<p>
356 <table border=0 cellpadding=0 cellspacing=0>
357 <tr><td><tt>[3 x [4 x int]]</tt></td><td>: 3x4 array integer values.</td></tr>
358 <tr><td><tt>[12 x [10 x float]]</tt></td><td>: 12x10 array of single precision floating point values.</td></tr>
359 <tr><td><tt>[2 x [3 x [4 x uint]]]</tt></td><td>: 2x3x4 array of unsigned integer values.</td></tr>
364 <!-- _______________________________________________________________________ -->
365 </ul><a name="t_function"><h4><hr size=0>Function Type</h4><ul>
369 The function type can be thought of as a function signature. It consists of a
370 return type and a list of formal parameter types. Function types are usually
371 used when to build virtual function tables (which are structures of pointers to
372 functions), for indirect function calls, and when defining a function.<p>
376 <returntype> (<parameter list>)
379 Where '<tt><parameter list></tt>' is a comma-separated list of type
380 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
381 which indicates that the function takes a variable number of arguments.
382 Variable argument functions can access their arguments with the <a
383 href="#int_varargs">variable argument handling intrinsic</a> functions.
388 <table border=0 cellpadding=0 cellspacing=0>
390 <tr><td><tt>int (int)</tt></td><td>: function taking an <tt>int</tt>, returning
391 an <tt>int</tt></td></tr>
393 <tr><td><tt>float (int, int *) *</tt></td><td>: <a href="#t_pointer">Pointer</a>
394 to a function that takes an <tt>int</tt> and a <a href="#t_pointer">pointer</a>
395 to <tt>int</tt>, returning <tt>float</tt>.</td></tr>
397 <tr><td><tt>int (sbyte *, ...)</tt></td><td>: A vararg function that takes at
398 least one <a href="#t_pointer">pointer</a> to <tt>sbyte</tt> (signed char in C),
399 which returns an integer. This is the signature for <tt>printf</tt> in
407 <!-- _______________________________________________________________________ -->
408 </ul><a name="t_struct"><h4><hr size=0>Structure Type</h4><ul>
412 The structure type is used to represent a collection of data members together in
413 memory. The packing of the field types is defined to match the ABI of the
414 underlying processor. The elements of a structure may be any type that has a
417 Structures are accessed using '<tt><a href="#i_load">load</a></tt> and '<tt><a
418 href="#i_store">store</a></tt>' by getting a pointer to a field with the '<tt><a
419 href="#i_getelementptr">getelementptr</a></tt>' instruction.<p>
423 { <type list> }
428 <table border=0 cellpadding=0 cellspacing=0>
430 <tr><td><tt>{ int, int, int }</tt></td><td>: a triple of three <tt>int</tt>
433 <tr><td><tt>{ float, int (int) * }</tt></td><td>: A pair, where the first
434 element is a <tt>float</tt> and the second element is a <a
435 href="#t_pointer">pointer</a> to a <a href="t_function">function</a> that takes
436 an <tt>int</tt>, returning an <tt>int</tt>.</td></tr>
441 <!-- _______________________________________________________________________ -->
442 </ul><a name="t_pointer"><h4><hr size=0>Pointer Type</h4><ul>
446 As in many languages, the pointer type represents a pointer or reference to
447 another object, which must live in memory.<p>
456 <table border=0 cellpadding=0 cellspacing=0>
458 <tr><td><tt>[4x int]*</tt></td><td>: <a href="#t_pointer">pointer</a> to <a
459 href="#t_array">array</a> of four <tt>int</tt> values</td></tr>
461 <tr><td><tt>int (int *) *</tt></td><td>: A <a href="#t_pointer">pointer</a> to a
462 <a href="t_function">function</a> that takes an <tt>int</tt>, returning an
463 <tt>int</tt>.</td></tr>
469 <!-- _______________________________________________________________________ -->
471 </ul><a name="t_packed"><h4><hr size=0>Packed Type</h4><ul>
473 Mention/decide that packed types work with saturation or not. Maybe have a packed+saturated type in addition to just a packed type.<p>
475 Packed types should be 'nonsaturated' because standard data types are not saturated. Maybe have a saturated packed type?<p>
480 <!-- *********************************************************************** -->
481 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
482 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
483 <a name="highlevel">High Level Structure
484 </b></font></td></tr></table><ul>
485 <!-- *********************************************************************** -->
488 <!-- ======================================================================= -->
489 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
490 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
491 <a name="modulestructure">Module Structure
492 </b></font></td></tr></table><ul>
494 LLVM programs are composed of "Module"s, each of which is a translation unit of
495 the input programs. Each module consists of functions, global variables, and
496 symbol table entries. Modules may be combined together with the LLVM linker,
497 which merges function (and global variable) definitions, resolves forward
498 declarations, and merges symbol table entries. Here is an example of the "hello world" module:<p>
501 <i>; Declare the string constant as a global constant...</i>
502 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
504 <i>; External declaration of the puts function</i>
505 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
507 <i>; Definition of main function</i>
508 int %main() { <i>; int()* </i>
509 <i>; Convert [13x sbyte]* to sbyte *...</i>
510 %cast210 = <a href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
512 <i>; Call puts function to write out the string to stdout...</i>
513 <a href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
514 <a href="#i_ret">ret</a> int 0
518 This example is made up of a <a href="#globalvars">global variable</a> named
519 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function, and a
520 <a href="#functionstructure">function definition</a> for "<tt>main</tt>".<p>
523 In general, a module is made up of a list of global values, where both functions
524 and global variables are global values. Global values are represented by a
525 pointer to a memory location (in this case, a pointer to an array of char, and a
526 pointer to a function), and have one of the following linkage types:<p>
529 <a name="linkage_internal">
530 <dt><tt><b>internal</b></tt>
532 <dd>Global values with internal linkage are only directly accessible by objects
533 in the current module. In particular, linking code into a module with an
534 internal global value may cause the internal to be renamed as necessary to avoid
535 collisions. Because the symbol is internal to the module, all references can be
536 updated. This corresponds to the notion of the '<tt>static</tt>' keyword in C,
537 or the idea of "anonymous namespaces" in C++.<p>
539 <a name="linkage_linkonce">
540 <dt><tt><b>linkonce</b></tt>:
542 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
543 the twist that linking together two modules defining the same <tt>linkonce</tt>
544 globals will cause one of the globals to be discarded. This is typically used
545 to implement inline functions. Unreferenced <tt>linkonce</tt> globals are
546 allowed to be discarded.<p>
548 <a name="linkage_weak">
549 <dt><tt><b>weak</b></tt>:
551 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
552 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
553 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.<p>
555 <a name="linkage_appending">
556 <dt><tt><b>appending</b></tt>:
558 <dd>"<tt>appending</tt>" linkage may only applied to global variables of pointer
559 to array type. When two global variables with appending linkage are linked
560 together, the two global arrays are appended together. This is the LLVM,
561 typesafe, equivalent of having the system linker append together "sections" with
562 identical names when .o files are linked.<p>
564 <a name="linkage_external">
565 <dt><tt><b>externally visible</b></tt>:
567 <dd>If none of the above identifiers are used, the global is externally visible,
568 meaning that it participates in linkage and can be used to resolve external
569 symbol references.<p>
574 For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
575 another module defined a "<tt>.LC0</tt>" variable and was linked with this one,
576 one of the two would be renamed, preventing a collision. Since "<tt>main</tt>"
577 and "<tt>puts</tt>" are external (i.e., lacking any linkage declarations), they
578 are accessible outside of the current module. It is illegal for a function
579 <i>declaration</i> to have any linkage type other than "externally visible".<p>
582 <!-- ======================================================================= -->
583 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
584 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
585 <a name="globalvars">Global Variables
586 </b></font></td></tr></table><ul>
588 Global variables define regions of memory allocated at compilation time instead
589 of run-time. Global variables may optionally be initialized. A variable may
590 be defined as a global "constant", which indicates that the contents of the
591 variable will never be modified (opening options for optimization). Constants
592 must always have an initial value.<p>
594 As SSA values, global variables define pointer values that are in scope
595 (i.e. they dominate) for all basic blocks in the program. Global variables
596 always define a pointer to their "content" type because they describe a region
597 of memory, and all memory objects in LLVM are accessed through pointers.<p>
601 <!-- ======================================================================= -->
602 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
603 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
604 <a name="functionstructure">Functions
605 </b></font></td></tr></table><ul>
607 LLVM functions definitions are composed of a (possibly empty) argument list, an
608 opening curly brace, a list of basic blocks, and a closing curly brace. LLVM
609 function declarations are defined with the "<tt>declare</tt>" keyword, a
610 function name and a function signature.<p>
612 A function definition contains a list of basic blocks, forming the CFG for the
613 function. Each basic block may optionally start with a label (giving the basic
614 block a symbol table entry), contains a list of instructions, and ends with a <a
615 href="#terminators">terminator</a> instruction (such as a branch or function
618 The first basic block in program is special in two ways: it is immediately
619 executed on entrance to the function, and it is not allowed to have predecessor
620 basic blocks (i.e. there can not be any branches to the entry block of a
621 function). Because the block can have no predecessors, it also cannot have any
622 <a href="#i_phi">PHI nodes</a>.<p>
625 <!-- *********************************************************************** -->
626 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
627 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
628 <a name="instref">Instruction Reference
629 </b></font></td></tr></table><ul>
630 <!-- *********************************************************************** -->
632 The LLVM instruction set consists of several different classifications of
633 instructions: <a href="#terminators">terminator instructions</a>, <a
634 href="#binaryops">binary instructions</a>, <a href="#memoryops">memory
635 instructions</a>, and <a href="#otherops">other instructions</a>.<p>
638 <!-- ======================================================================= -->
639 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
640 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
641 <a name="terminators">Terminator Instructions
642 </b></font></td></tr></table><ul>
644 As mentioned <a href="#functionstructure">previously</a>, every basic block in a
645 program ends with a "Terminator" instruction, which indicates which block should
646 be executed after the current block is finished. These terminator instructions
647 typically yield a '<tt>void</tt>' value: they produce control flow, not values
648 (the one exception being the '<a href="#i_invoke"><tt>invoke</tt></a>'
651 There are five different terminator instructions: the '<a
652 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a
653 href="#i_br"><tt>br</tt></a>' instruction, the '<a
654 href="#i_switch"><tt>switch</tt></a>' instruction, the '<a
655 href="#i_invoke"><tt>invoke</tt></a>' instruction, and the '<a
656 href="#i_unwind"><tt>unwind</tt></a>' instruction.<p>
659 <!-- _______________________________________________________________________ -->
660 </ul><a name="i_ret"><h4><hr size=0>'<tt>ret</tt>' Instruction</h4><ul>
664 ret <type> <value> <i>; Return a value from a non-void function</i>
665 ret void <i>; Return from void function</i>
670 The '<tt>ret</tt>' instruction is used to return control flow (and a value) from
671 a function, back to the caller.<p>
673 There are two forms of the '<tt>ret</tt>' instructruction: one that returns a
674 value and then causes control flow, and one that just causes control flow to
679 The '<tt>ret</tt>' instruction may return any '<a href="#t_firstclass">first
680 class</a>' type. Notice that a function is not <a href="#wellformed">well
681 formed</a> if there exists a '<tt>ret</tt>' instruction inside of the function
682 that returns a value that does not match the return type of the function.<p>
686 When the '<tt>ret</tt>' instruction is executed, control flow returns back to
687 the calling function's context. If the caller is a "<a
688 href="#i_call"><tt>call</tt></a> instruction, execution continues at the
689 instruction after the call. If the caller was an "<a
690 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues at the
691 beginning "normal" of the destination block. If the instruction returns a
692 value, that value shall set the call or invoke instruction's return value.<p>
697 ret int 5 <i>; Return an integer value of 5</i>
698 ret void <i>; Return from a void function</i>
702 <!-- _______________________________________________________________________ -->
703 </ul><a name="i_br"><h4><hr size=0>'<tt>br</tt>' Instruction</h4><ul>
707 br bool <cond>, label <iftrue>, label <iffalse>
708 br label <dest> <i>; Unconditional branch</i>
713 The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
714 different basic block in the current function. There are two forms of this
715 instruction, corresponding to a conditional branch and an unconditional
720 The conditional branch form of the '<tt>br</tt>' instruction takes a single
721 '<tt>bool</tt>' value and two '<tt>label</tt>' values. The unconditional form
722 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
727 Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
728 argument is evaluated. If the value is <tt>true</tt>, control flows to the
729 '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
730 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.<p>
735 %cond = <a href="#i_setcc">seteq</a> int %a, %b
736 br bool %cond, label %IfEqual, label %IfUnequal
738 <a href="#i_ret">ret</a> int 1
740 <a href="#i_ret">ret</a> int 0
744 <!-- _______________________________________________________________________ -->
745 </ul><a name="i_switch"><h4><hr size=0>'<tt>switch</tt>' Instruction</h4><ul>
749 switch uint <value>, label <defaultdest> [ int <val>, label &dest>, ... ]
755 The '<tt>switch</tt>' instruction is used to transfer control flow to one of
756 several different places. It is a generalization of the '<tt>br</tt>'
757 instruction, allowing a branch to occur to one of many possible destinations.<p>
761 The '<tt>switch</tt>' instruction uses three parameters: a '<tt>uint</tt>'
762 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
763 an array of pairs of comparison value constants and '<tt>label</tt>'s.<p>
767 The <tt>switch</tt> instruction specifies a table of values and destinations.
768 When the '<tt>switch</tt>' instruction is executed, this table is searched for
769 the given value. If the value is found, the corresponding destination is
770 branched to, otherwise the default value it transfered to.<p>
772 <h5>Implementation:</h5>
774 Depending on properties of the target machine and the particular <tt>switch</tt>
775 instruction, this instruction may be code generated as a series of chained
776 conditional branches, or with a lookup table.<p>
780 <i>; Emulate a conditional br instruction</i>
781 %Val = <a href="#i_cast">cast</a> bool %value to uint
782 switch uint %Val, label %truedest [int 0, label %falsedest ]
784 <i>; Emulate an unconditional br instruction</i>
785 switch uint 0, label %dest [ ]
787 <i>; Implement a jump table:</i>
788 switch uint %val, label %otherwise [ int 0, label %onzero,
790 int 2, label %ontwo ]
795 <!-- _______________________________________________________________________ -->
796 </ul><a name="i_invoke"><h4><hr size=0>'<tt>invoke</tt>' Instruction</h4><ul>
800 <result> = invoke <ptr to function ty> %<function ptr val>(<function args>)
801 to label <normal label> except label <exception label>
806 The '<tt>invoke</tt>' instruction causes control to transfer to a specified
807 function, with the possibility of control flow transfer to either the
808 '<tt>normal</tt>' <tt>label</tt> label or the '<tt>exception</tt>'
809 <tt>label</tt>. If the callee function returns with the "<tt><a
810 href="#i_ret">ret</a></tt>" instruction, control flow will return to the
811 "normal" label. If the callee (or any indirect callees) returns with the "<a
812 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted, and
813 continued at the dynamically nearest "except" label.<p>
818 This instruction requires several arguments:<p>
821 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
822 function value being invoked. In most cases, this is a direct function
823 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
824 an arbitrary pointer to function value.
826 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
827 function to be invoked.
829 <li>'<tt>function args</tt>': argument list whose types match the function
830 signature argument types. If the function signature indicates the function
831 accepts a variable number of arguments, the extra arguments can be specified.
833 <li>'<tt>normal label</tt>': the label reached when the called function executes
834 a '<tt><a href="#i_ret">ret</a></tt>' instruction.
836 <li>'<tt>exception label</tt>': the label reached when a callee returns with the
837 <a href="#i_unwind"><tt>unwind</tt></a> instruction.
842 This instruction is designed to operate as a standard '<tt><a
843 href="#i_call">call</a></tt>' instruction in most regards. The primary
844 difference is that it establishes an association with a label, which is used by the runtime library to unwind the stack.<p>
846 This instruction is used in languages with destructors to ensure that proper
847 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
848 exception. Additionally, this is important for implementation of
849 '<tt>catch</tt>' clauses in high-level languages that support them.<p>
853 %retval = invoke int %Test(int 15)
855 except label %TestCleanup <i>; {int}:retval set</i>
858 <!-- _______________________________________________________________________ -->
859 </ul><a name="i_unwind"><h4><hr size=0>'<tt>unwind</tt>' Instruction</h4><ul>
868 The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow at
869 the first callee in the dynamic call stack which used an <a
870 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
871 primarily used to implement exception handling.
875 The '<tt>unwind</tt>' intrinsic causes execution of the current function to
876 immediately halt. The dynamic call stack is then searched for the first <a
877 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
878 execution continues at the "exceptional" destination block specified by the
879 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
880 dynamic call chain, undefined behavior results.
884 <!-- ======================================================================= -->
885 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0><tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
886 <a name="binaryops">Binary Operations
887 </b></font></td></tr></table><ul>
889 Binary operators are used to do most of the computation in a program. They
890 require two operands, execute an operation on them, and produce a single value.
891 The result value of a binary operator is not necessarily the same type as its
894 There are several different binary operators:<p>
897 <!-- _______________________________________________________________________ -->
898 </ul><a name="i_add"><h4><hr size=0>'<tt>add</tt>' Instruction</h4><ul>
902 <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
906 The '<tt>add</tt>' instruction returns the sum of its two operands.<p>
909 The two arguments to the '<tt>add</tt>' instruction must be either <a href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values. Both arguments must have identical types.<p>
913 The value produced is the integer or floating point sum of the two operands.<p>
917 <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
921 <!-- _______________________________________________________________________ -->
922 </ul><a name="i_sub"><h4><hr size=0>'<tt>sub</tt>' Instruction</h4><ul>
926 <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
931 The '<tt>sub</tt>' instruction returns the difference of its two operands.<p>
933 Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
934 instruction present in most other intermediate representations.<p>
938 The two arguments to the '<tt>sub</tt>' instruction must be either <a
939 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
940 values. Both arguments must have identical types.<p>
944 The value produced is the integer or floating point difference of the two
949 <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
950 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
953 <!-- _______________________________________________________________________ -->
954 </ul><a name="i_mul"><h4><hr size=0>'<tt>mul</tt>' Instruction</h4><ul>
958 <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
962 The '<tt>mul</tt>' instruction returns the product of its two operands.<p>
965 The two arguments to the '<tt>mul</tt>' instruction must be either <a href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values. Both arguments must have identical types.<p>
969 The value produced is the integer or floating point product of the two
972 There is no signed vs unsigned multiplication. The appropriate action is taken
973 based on the type of the operand. <p>
978 <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
982 <!-- _______________________________________________________________________ -->
983 </ul><a name="i_div"><h4><hr size=0>'<tt>div</tt>' Instruction</h4><ul>
987 <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
992 The '<tt>div</tt>' instruction returns the quotient of its two operands.<p>
996 The two arguments to the '<tt>div</tt>' instruction must be either <a
997 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
998 values. Both arguments must have identical types.<p>
1002 The value produced is the integer or floating point quotient of the two
1007 <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
1011 <!-- _______________________________________________________________________ -->
1012 </ul><a name="i_rem"><h4><hr size=0>'<tt>rem</tt>' Instruction</h4><ul>
1016 <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1020 The '<tt>rem</tt>' instruction returns the remainder from the division of its two operands.<p>
1023 The two arguments to the '<tt>rem</tt>' instruction must be either <a href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values. Both arguments must have identical types.<p>
1027 This returns the <i>remainder</i> of a division (where the result has the same
1028 sign as the divisor), not the <i>modulus</i> (where the result has the same sign
1029 as the dividend) of a value. For more information about the difference, see: <a
1030 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The Math
1035 <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1039 <!-- _______________________________________________________________________ -->
1040 </ul><a name="i_setcc"><h4><hr size=0>'<tt>set<i>cc</i></tt>' Instructions</h4><ul>
1044 <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1045 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1046 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1047 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1048 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1049 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1052 <h5>Overview:</h5> The '<tt>set<i>cc</i></tt>' family of instructions returns a
1053 boolean value based on a comparison of their two operands.<p>
1055 <h5>Arguments:</h5> The two arguments to the '<tt>set<i>cc</i></tt>'
1056 instructions must be of <a href="#t_firstclass">first class</a> or <a
1057 href="#t_pointer">pointer</a> type (it is not possible to compare
1058 '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>' or '<tt>void</tt>'
1059 values, etc...). Both arguments must have identical types.<p>
1063 The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
1064 both operands are equal.<br>
1066 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
1067 both operands are unequal.<br>
1069 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
1070 the first operand is less than the second operand.<br>
1072 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
1073 the first operand is greater than the second operand.<br>
1075 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
1076 the first operand is less than or equal to the second operand.<br>
1078 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
1079 the first operand is greater than or equal to the second operand.<p>
1083 <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1084 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1085 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1086 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1087 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1088 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1093 <!-- ======================================================================= -->
1094 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1095 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
1096 <a name="bitwiseops">Bitwise Binary Operations
1097 </b></font></td></tr></table><ul>
1099 Bitwise binary operators are used to do various forms of bit-twiddling in a
1100 program. They are generally very efficient instructions, and can commonly be
1101 strength reduced from other instructions. They require two operands, execute an
1102 operation on them, and produce a single value. The resulting value of the
1103 bitwise binary operators is always the same type as its first operand.<p>
1105 <!-- _______________________________________________________________________ -->
1106 </ul><a name="i_and"><h4><hr size=0>'<tt>and</tt>' Instruction</h4><ul>
1110 <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1114 The '<tt>and</tt>' instruction returns the bitwise logical and of its two operands.<p>
1118 The two arguments to the '<tt>and</tt>' instruction must be <a
1119 href="#t_integral">integral</a> values. Both arguments must have identical
1125 The truth table used for the '<tt>and</tt>' instruction is:<p>
1127 <center><table border=1 cellspacing=0 cellpadding=4>
1128 <tr><td>In0</td> <td>In1</td> <td>Out</td></tr>
1129 <tr><td>0</td> <td>0</td> <td>0</td></tr>
1130 <tr><td>0</td> <td>1</td> <td>0</td></tr>
1131 <tr><td>1</td> <td>0</td> <td>0</td></tr>
1132 <tr><td>1</td> <td>1</td> <td>1</td></tr>
1133 </table></center><p>
1138 <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1139 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1140 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1145 <!-- _______________________________________________________________________ -->
1146 </ul><a name="i_or"><h4><hr size=0>'<tt>or</tt>' Instruction</h4><ul>
1150 <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1153 <h5>Overview:</h5> The '<tt>or</tt>' instruction returns the bitwise logical
1154 inclusive or of its two operands.<p>
1158 The two arguments to the '<tt>or</tt>' instruction must be <a
1159 href="#t_integral">integral</a> values. Both arguments must have identical
1165 The truth table used for the '<tt>or</tt>' instruction is:<p>
1167 <center><table border=1 cellspacing=0 cellpadding=4>
1168 <tr><td>In0</td> <td>In1</td> <td>Out</td></tr>
1169 <tr><td>0</td> <td>0</td> <td>0</td></tr>
1170 <tr><td>0</td> <td>1</td> <td>1</td></tr>
1171 <tr><td>1</td> <td>0</td> <td>1</td></tr>
1172 <tr><td>1</td> <td>1</td> <td>1</td></tr>
1173 </table></center><p>
1178 <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1179 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1180 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1184 <!-- _______________________________________________________________________ -->
1185 </ul><a name="i_xor"><h4><hr size=0>'<tt>xor</tt>' Instruction</h4><ul>
1189 <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1194 The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of its
1195 two operands. The <tt>xor</tt> is used to implement the "one's complement"
1196 operation, which is the "~" operator in C.<p>
1200 The two arguments to the '<tt>xor</tt>' instruction must be <a
1201 href="#t_integral">integral</a> values. Both arguments must have identical
1207 The truth table used for the '<tt>xor</tt>' instruction is:<p>
1209 <center><table border=1 cellspacing=0 cellpadding=4>
1210 <tr><td>In0</td> <td>In1</td> <td>Out</td></tr>
1211 <tr><td>0</td> <td>0</td> <td>0</td></tr>
1212 <tr><td>0</td> <td>1</td> <td>1</td></tr>
1213 <tr><td>1</td> <td>0</td> <td>1</td></tr>
1214 <tr><td>1</td> <td>1</td> <td>0</td></tr>
1215 </table></center><p>
1220 <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1221 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1222 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1223 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1227 <!-- _______________________________________________________________________ -->
1228 </ul><a name="i_shl"><h4><hr size=0>'<tt>shl</tt>' Instruction</h4><ul>
1232 <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1237 The '<tt>shl</tt>' instruction returns the first operand shifted to the left a
1238 specified number of bits.
1242 The first argument to the '<tt>shl</tt>' instruction must be an <a
1243 href="#t_integer">integer</a> type. The second argument must be an
1244 '<tt>ubyte</tt>' type.<p>
1248 The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.<p>
1253 <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1254 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1255 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1259 <!-- _______________________________________________________________________ -->
1260 </ul><a name="i_shr"><h4><hr size=0>'<tt>shr</tt>' Instruction</h4><ul>
1265 <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1269 The '<tt>shr</tt>' instruction returns the first operand shifted to the right a specified number of bits.
1272 The first argument to the '<tt>shr</tt>' instruction must be an <a href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>' type.<p>
1276 If the first argument is a <a href="#t_signed">signed</a> type, the most
1277 significant bit is duplicated in the newly free'd bit positions. If the first
1278 argument is unsigned, zero bits shall fill the empty positions.<p>
1282 <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1283 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1284 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1285 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1286 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1293 <!-- ======================================================================= -->
1294 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1295 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
1296 <a name="memoryops">Memory Access Operations
1297 </b></font></td></tr></table><ul>
1299 A key design point of an SSA-based representation is how it represents memory.
1300 In LLVM, no memory locations are in SSA form, which makes things very simple.
1301 This section describes how to read, write, allocate and free memory in LLVM.<p>
1304 <!-- _______________________________________________________________________ -->
1305 </ul><a name="i_malloc"><h4><hr size=0>'<tt>malloc</tt>' Instruction</h4><ul>
1309 <result> = malloc <type>, uint <NumElements> <i>; yields {type*}:result</i>
1310 <result> = malloc <type> <i>; yields {type*}:result</i>
1314 The '<tt>malloc</tt>' instruction allocates memory from the system heap and returns a pointer to it.<p>
1318 The the '<tt>malloc</tt>' instruction allocates
1319 <tt>sizeof(<type>)*NumElements</tt> bytes of memory from the operating
1320 system, and returns a pointer of the appropriate type to the program. The
1321 second form of the instruction is a shorter version of the first instruction
1322 that defaults to allocating one element.<p>
1324 '<tt>type</tt>' must be a sized type.<p>
1328 Memory is allocated using the system "<tt>malloc</tt>" function, and a pointer
1333 %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1335 %size = <a href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1336 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1337 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1341 <!-- _______________________________________________________________________ -->
1342 </ul><a name="i_free"><h4><hr size=0>'<tt>free</tt>' Instruction</h4><ul>
1346 free <type> <value> <i>; yields {void}</i>
1351 The '<tt>free</tt>' instruction returns memory back to the unused memory heap, to be reallocated in the future.<p>
1356 '<tt>value</tt>' shall be a pointer value that points to a value that was
1357 allocated with the '<tt><a href="#i_malloc">malloc</a></tt>' instruction.<p>
1362 Access to the memory pointed to by the pointer is not longer defined after this instruction executes.<p>
1366 %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1367 free [4 x ubyte]* %array
1371 <!-- _______________________________________________________________________ -->
1372 </ul><a name="i_alloca"><h4><hr size=0>'<tt>alloca</tt>' Instruction</h4><ul>
1376 <result> = alloca <type>, uint <NumElements> <i>; yields {type*}:result</i>
1377 <result> = alloca <type> <i>; yields {type*}:result</i>
1382 The '<tt>alloca</tt>' instruction allocates memory on the current stack frame of
1383 the procedure that is live until the current function returns to its caller.<p>
1387 The the '<tt>alloca</tt>' instruction allocates
1388 <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the runtime stack,
1389 returning a pointer of the appropriate type to the program. The second form of
1390 the instruction is a shorter version of the first that defaults to allocating
1393 '<tt>type</tt>' may be any sized type.<p>
1397 Memory is allocated, a pointer is returned. '<tt>alloca</tt>'d memory is
1398 automatically released when the function returns. The '<tt>alloca</tt>'
1399 instruction is commonly used to represent automatic variables that must have an
1400 address available. When the function returns (either with the <tt><a
1401 href="#i_ret">ret</a></tt> or <tt><a href="#i_invoke">invoke</a></tt>
1402 instructions), the memory is reclaimed.<p>
1406 %ptr = alloca int <i>; yields {int*}:ptr</i>
1407 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
1411 <!-- _______________________________________________________________________ -->
1412 </ul><a name="i_load"><h4><hr size=0>'<tt>load</tt>' Instruction</h4><ul>
1416 <result> = load <ty>* <pointer>
1417 <result> = volatile load <ty>* <pointer>
1421 The '<tt>load</tt>' instruction is used to read from memory.<p>
1425 The argument to the '<tt>load</tt>' instruction specifies the memory address to
1426 load from. The pointer must point to a <a href="t_firstclass">first class</a>
1427 type. If the <tt>load</tt> is marked as <tt>volatile</tt> then the optimizer is
1428 not allowed to modify the number or order of execution of this <tt>load</tt>
1429 with other volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
1434 The location of memory pointed to is loaded.
1438 %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1439 <a href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1440 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1446 <!-- _______________________________________________________________________ -->
1447 </ul><a name="i_store"><h4><hr size=0>'<tt>store</tt>' Instruction</h4><ul>
1451 store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1452 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1456 The '<tt>store</tt>' instruction is used to write to memory.<p>
1460 There are two arguments to the '<tt>store</tt>' instruction: a value to store
1461 and an address to store it into. The type of the '<tt><pointer></tt>'
1462 operand must be a pointer to the type of the '<tt><value></tt>' operand.
1463 If the <tt>store</tt> is marked as <tt>volatile</tt> then the optimizer is not
1464 allowed to modify the number or order of execution of this <tt>store</tt> with
1465 other volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
1468 <h5>Semantics:</h5> The contents of memory are updated to contain
1469 '<tt><value></tt>' at the location specified by the
1470 '<tt><pointer></tt>' operand.<p>
1474 %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1475 <a href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1476 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1482 <!-- _______________________________________________________________________ -->
1483 </ul><a name="i_getelementptr"><h4><hr size=0>'<tt>getelementptr</tt>' Instruction</h4><ul>
1487 <result> = getelementptr <ty>* <ptrval>{, long <aidx>|, ubyte <sidx>}*
1492 The '<tt>getelementptr</tt>' instruction is used to get the address of a
1493 subelement of an aggregate data structure.<p>
1497 This instruction takes a list of <tt>long</tt> values and <tt>ubyte</tt>
1498 constants that indicate what form of addressing to perform. The actual types of
1499 the arguments provided depend on the type of the first pointer argument. The
1500 '<tt>getelementptr</tt>' instruction is used to index down through the type
1501 levels of a structure.<p>
1503 For example, lets consider a C code fragment and how it gets compiled to
1518 int *foo(struct ST *s) {
1519 return &s[1].Z.B[5][13];
1523 The LLVM code generated by the GCC frontend is:
1526 %RT = type { sbyte, [10 x [20 x int]], sbyte }
1527 %ST = type { int, double, %RT }
1529 int* "foo"(%ST* %s) {
1530 %reg = getelementptr %ST* %s, long 1, ubyte 2, ubyte 1, long 5, long 13
1537 The index types specified for the '<tt>getelementptr</tt>' instruction depend on
1538 the pointer type that is being index into. <a href="t_pointer">Pointer</a> and
1539 <a href="t_array">array</a> types require '<tt>long</tt>' values, and <a
1540 href="t_struct">structure</a> types require '<tt>ubyte</tt>'
1541 <b>constants</b>.<p>
1543 In the example above, the first index is indexing into the '<tt>%ST*</tt>' type,
1544 which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT }</tt>'
1545 type, a structure. The second index indexes into the third element of the
1546 structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]], sbyte
1547 }</tt>' type, another structure. The third index indexes into the second
1548 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
1549 array. The two dimensions of the array are subscripted into, yielding an
1550 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction return a pointer
1551 to this element, thus yielding a '<tt>int*</tt>' type.<p>
1553 Note that it is perfectly legal to index partially through a structure,
1554 returning a pointer to an inner element. Because of this, the LLVM code for the
1555 given testcase is equivalent to:<p>
1558 int* "foo"(%ST* %s) {
1559 %t1 = getelementptr %ST* %s , long 1 <i>; yields %ST*:%t1</i>
1560 %t2 = getelementptr %ST* %t1, long 0, ubyte 2 <i>; yields %RT*:%t2</i>
1561 %t3 = getelementptr %RT* %t2, long 0, ubyte 1 <i>; yields [10 x [20 x int]]*:%t3</i>
1562 %t4 = getelementptr [10 x [20 x int]]* %t3, long 0, long 5 <i>; yields [20 x int]*:%t4</i>
1563 %t5 = getelementptr [20 x int]* %t4, long 0, long 13 <i>; yields int*:%t5</i>
1572 <i>; yields [12 x ubyte]*:aptr</i>
1573 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, ubyte 1
1578 <!-- ======================================================================= -->
1579 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1580 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
1581 <a name="otherops">Other Operations
1582 </b></font></td></tr></table><ul>
1584 The instructions in this catagory are the "miscellaneous" instructions, which defy better classification.<p>
1587 <!-- _______________________________________________________________________ -->
1588 </ul><a name="i_phi"><h4><hr size=0>'<tt>phi</tt>' Instruction</h4><ul>
1592 <result> = phi <ty> [ <val0>, <label0>], ...
1597 The '<tt>phi</tt>' instruction is used to implement the φ node in the SSA
1598 graph representing the function.<p>
1602 The type of the incoming values are specified with the first type field. After
1603 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
1604 one pair for each predecessor basic block of the current block.<p>
1606 There must be no non-phi instructions between the start of a basic block and the
1607 PHI instructions: i.e. PHI instructions must be first in a basic block.<p>
1611 At runtime, the '<tt>phi</tt>' instruction logically takes on the value
1612 specified by the parameter, depending on which basic block we came from in the
1613 last <a href="#terminators">terminator</a> instruction.<p>
1618 Loop: ; Infinite loop that counts from 0 on up...
1619 %indvar = phi uint [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
1620 %nextindvar = add uint %indvar, 1
1625 <!-- _______________________________________________________________________ -->
1626 </ul><a name="i_cast"><h4><hr size=0>'<tt>cast .. to</tt>' Instruction</h4><ul>
1630 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
1635 The '<tt>cast</tt>' instruction is used as the primitive means to convert
1636 integers to floating point, change data type sizes, and break type safety (by
1637 casting pointers).<p>
1641 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
1642 class value, and a type to cast it to, which must also be a first class type.<p>
1646 This instruction follows the C rules for explicit casts when determining how the
1647 data being cast must change to fit in its new container.<p>
1649 When casting to bool, any value that would be considered true in the context of
1650 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
1651 all else are '<tt>false</tt>'.<p>
1653 When extending an integral value from a type of one signness to another (for
1654 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
1655 <b>source</b> value is signed, and zero-extended if the source value is
1656 unsigned. <tt>bool</tt> values are always zero extended into either zero or
1661 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
1662 %Y = cast int 123 to bool <i>; yields bool:true</i>
1667 <!-- _______________________________________________________________________ -->
1668 </ul><a name="i_call"><h4><hr size=0>'<tt>call</tt>' Instruction</h4><ul>
1672 <result> = call <ty>* <fnptrval>(<param list>)
1677 The '<tt>call</tt>' instruction represents a simple function call.<p>
1681 This instruction requires several arguments:<p>
1684 <li>'<tt>ty</tt>': shall be the signature of the pointer to function value being
1685 invoked. The argument types must match the types implied by this signature.<p>
1687 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to be
1688 invoked. In most cases, this is a direct function invocation, but indirect
1689 <tt>call</tt>s are just as possible, calling an arbitrary pointer to function
1692 <li>'<tt>function args</tt>': argument list whose types match the function
1693 signature argument types. If the function signature indicates the function
1694 accepts a variable number of arguments, the extra arguments can be specified.
1699 The '<tt>call</tt>' instruction is used to cause control flow to transfer to a
1700 specified function, with its incoming arguments bound to the specified values.
1701 Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called function,
1702 control flow continues with the instruction after the function call, and the
1703 return value of the function is bound to the result argument. This is a simpler
1704 case of the <a href="#i_invoke">invoke</a> instruction.<p>
1708 %retval = call int %test(int %argc)
1709 call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);
1713 <!-- _______________________________________________________________________ -->
1714 </ul><a name="i_vanext"><h4><hr size=0>'<tt>vanext</tt>' Instruction</h4><ul>
1718 <resultarglist> = vanext <va_list> <arglist>, <argty>
1723 The '<tt>vanext</tt>' instruction is used to access arguments passed through
1724 the "variable argument" area of a function call. It is used to implement the
1725 <tt>va_arg</tt> macro in C.<p>
1729 This instruction takes a <tt>valist</tt> value and the type of the argument. It
1730 returns another <tt>valist</tt>.
1734 The '<tt>vanext</tt>' instruction advances the specified <tt>valist</tt> past
1735 an argument of the specified type. In conjunction with the <a
1736 href="#i_vaarg"><tt>vaarg</tt></a> instruction, it is used to implement the
1737 <tt>va_arg</tt> macro available in C. For more information, see the variable
1738 argument handling <a href="#int_varargs">Intrinsic Functions</a>.<p>
1740 It is legal for this instruction to be called in a function which does not take
1741 a variable number of arguments, for example, the <tt>vfprintf</tt> function.<p>
1743 <tt>vanext</tt> is an LLVM instruction instead of an <a
1744 href="#intrinsics">intrinsic function</a> because it takes an type as an
1749 See the <a href="#int_varargs">variable argument processing</a> section.<p>
1753 <!-- _______________________________________________________________________ -->
1754 </ul><a name="i_vaarg"><h4><hr size=0>'<tt>vaarg</tt>' Instruction</h4><ul>
1758 <resultval> = vaarg <va_list> <arglist>, <argty>
1763 The '<tt>vaarg</tt>' instruction is used to access arguments passed through
1764 the "variable argument" area of a function call. It is used to implement the
1765 <tt>va_arg</tt> macro in C.<p>
1769 This instruction takes a <tt>valist</tt> value and the type of the argument. It
1770 returns a value of the specified argument type.
1774 The '<tt>vaarg</tt>' instruction loads an argument of the specified type from
1775 the specified <tt>va_list</tt>. In conjunction with the <a
1776 href="#i_vanext"><tt>vanext</tt></a> instruction, it is used to implement the
1777 <tt>va_arg</tt> macro available in C. For more information, see the variable
1778 argument handling <a href="#int_varargs">Intrinsic Functions</a>.<p>
1780 It is legal for this instruction to be called in a function which does not take
1781 a variable number of arguments, for example, the <tt>vfprintf</tt> function.<p>
1783 <tt>vaarg</tt> is an LLVM instruction instead of an <a
1784 href="#intrinsics">intrinsic function</a> because it takes an type as an
1789 See the <a href="#int_varargs">variable argument processing</a> section.<p>
1795 <!-- *********************************************************************** -->
1796 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
1797 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
1798 <a name="intrinsics">Intrinsic Functions
1799 </b></font></td></tr></table><ul>
1800 <!-- *********************************************************************** -->
1802 LLVM supports the notion of an "intrinsic function". These functions have well
1803 known names and semantics, and are required to follow certain restrictions.
1804 Overall, these instructions represent an extension mechanism for the LLVM
1805 language that does not require changing all of the transformations in LLVM to
1806 add to the language (or the bytecode reader/writer, the parser, etc...).<p>
1808 Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix, this
1809 prefix is reserved in LLVM for intrinsic names, thus functions may not be named
1810 this. Intrinsic functions must always be external functions: you cannot define
1811 the body of intrinsic functions. Intrinsic functions may only be used in call
1812 or invoke instructions: it is illegal to take the address of an intrinsic
1813 function. Additionally, because intrinsic functions are part of the LLVM
1814 language, it is required that they all be documented here if any are added.<p>
1816 Unless an intrinsic function is target-specific, there must be a lowering pass
1817 to eliminate the intrinsic or all backends must support the intrinsic
1821 <!-- ======================================================================= -->
1822 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1823 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
1824 <a name="int_varargs">Variable Argument Handling Intrinsics
1825 </b></font></td></tr></table><ul>
1827 Variable argument support is defined in LLVM with the <a
1828 href="#i_vanext"><tt>vanext</tt></a> instruction and these three intrinsic
1829 functions. These functions are related to the similarly named macros defined in
1830 the <tt><stdarg.h></tt> header file.<p>
1832 All of these functions operate on arguments that use a target-specific value
1833 type "<tt>va_list</tt>". The LLVM assembly language reference manual does not
1834 define what this type is, so all transformations should be prepared to handle
1835 intrinsics with any type used.<p>
1837 This example shows how the <a href="#i_vanext"><tt>vanext</tt></a> instruction
1838 and the variable argument handling intrinsic functions are used.<p>
1841 int %test(int %X, ...) {
1842 ; Initialize variable argument processing
1843 %ap = call sbyte*()* %<a href="#i_va_start">llvm.va_start</a>()
1845 ; Read a single integer argument
1846 %tmp = vaarg sbyte* %ap, int
1848 ; Advance to the next argument
1849 %ap2 = vanext sbyte* %ap, int
1851 ; Demonstrate usage of llvm.va_copy and llvm.va_end
1852 %aq = call sbyte* (sbyte*)* %<a href="#i_va_copy">llvm.va_copy</a>(sbyte* %ap2)
1853 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %aq)
1855 ; Stop processing of arguments.
1856 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %ap2)
1861 <!-- _______________________________________________________________________ -->
1862 </ul><a name="i_va_start"><h4><hr size=0>'<tt>llvm.va_start</tt>' Intrinsic</h4><ul>
1866 call va_list ()* %llvm.va_start()
1871 The '<tt>llvm.va_start</tt>' intrinsic returns a new <tt><arglist></tt>
1872 for subsequent use by the variable argument intrinsics.<p>
1876 The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
1877 macro available in C. In a target-dependent way, it initializes and returns a
1878 <tt>va_list</tt> element, so that the next <tt>vaarg</tt> will produce the first
1879 variable argument passed to the function. Unlike the C <tt>va_start</tt> macro,
1880 this intrinsic does not need to know the last argument of the function, the
1881 compiler can figure that out.<p>
1883 Note that this intrinsic function is only legal to be called from within the
1884 body of a variable argument function.<p>
1887 <!-- _______________________________________________________________________ -->
1888 </ul><a name="i_va_end"><h4><hr size=0>'<tt>llvm.va_end</tt>' Intrinsic</h4><ul>
1892 call void (va_list)* %llvm.va_end(va_list <arglist>)
1897 The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt> which has
1898 been initialized previously with <tt><a
1899 href="#i_va_start">llvm.va_start</a></tt> or <tt><a
1900 href="#i_va_copy">llvm.va_copy</a></tt>.<p>
1904 The argument is a <tt>va_list</tt> to destroy.<p>
1908 The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt> macro
1909 available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
1910 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
1911 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly with calls
1912 to <tt>llvm.va_end</tt>.<p>
1916 <!-- _______________________________________________________________________ -->
1917 </ul><a name="i_va_copy"><h4><hr size=0>'<tt>llvm.va_copy</tt>' Intrinsic</h4><ul>
1921 call va_list (va_list)* %llvm.va_copy(va_list <destarglist>)
1926 The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position from
1927 the source argument list to the destination argument list.<p>
1931 The argument is the <tt>va_list</tt> to copy.
1935 The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt> macro
1936 available in C. In a target-dependent way, it copies the source
1937 <tt>va_list</tt> element into the returned list. This intrinsic is necessary
1938 because the <tt><a href="i_va_start">llvm.va_start</a></tt> intrinsic may be
1939 arbitrarily complex and require memory allocation, for example.<p>
1942 <!-- *********************************************************************** -->
1944 <!-- *********************************************************************** -->
1949 <address><a href="mailto:sabre@nondot.org">Chris Lattner</a></address>
1950 <!-- Created: Tue Jan 23 15:19:28 CST 2001 -->
1951 <!-- hhmts start -->
1952 Last modified: Tue Oct 21 10:43:36 CDT 2003