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>
43 <li><a href="#binaryops">Binary Operations</a>
45 <li><a href="#i_add" >'<tt>add</tt>' Instruction</a>
46 <li><a href="#i_sub" >'<tt>sub</tt>' Instruction</a>
47 <li><a href="#i_mul" >'<tt>mul</tt>' Instruction</a>
48 <li><a href="#i_div" >'<tt>div</tt>' Instruction</a>
49 <li><a href="#i_rem" >'<tt>rem</tt>' Instruction</a>
50 <li><a href="#i_setcc">'<tt>set<i>cc</i></tt>' Instructions</a>
52 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
54 <li><a href="#i_and">'<tt>and</tt>' Instruction</a>
55 <li><a href="#i_or" >'<tt>or</tt>' Instruction</a>
56 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a>
57 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a>
58 <li><a href="#i_shr">'<tt>shr</tt>' Instruction</a>
60 <li><a href="#memoryops">Memory Access Operations</a>
62 <li><a href="#i_malloc" >'<tt>malloc</tt>' Instruction</a>
63 <li><a href="#i_free" >'<tt>free</tt>' Instruction</a>
64 <li><a href="#i_alloca" >'<tt>alloca</tt>' Instruction</a>
65 <li><a href="#i_load" >'<tt>load</tt>' Instruction</a>
66 <li><a href="#i_store" >'<tt>store</tt>' Instruction</a>
67 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
69 <li><a href="#otherops">Other Operations</a>
71 <li><a href="#i_phi" >'<tt>phi</tt>' Instruction</a>
72 <li><a href="#i_cast">'<tt>cast .. to</tt>' Instruction</a>
73 <li><a href="#i_call" >'<tt>call</tt>' Instruction</a>
74 <li><a href="#i_va_arg">'<tt>va_arg</tt>' Instruction</a>
77 <li><a href="#intrinsics">Intrinsic Functions</a>
79 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
81 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
82 <li><a href="#i_va_end" >'<tt>llvm.va_end</tt>' Intrinsic</a>
83 <li><a href="#i_va_copy" >'<tt>llvm.va_copy</tt>' Intrinsic</a>
87 <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>
93 <!-- *********************************************************************** -->
94 <p><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
95 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
96 <a name="abstract">Abstract
97 </b></font></td></tr></table><ul>
98 <!-- *********************************************************************** -->
101 This document is a reference manual for the LLVM assembly language. LLVM is
102 an SSA based representation that provides type safety, low level operations,
103 flexibility, and the capability of representing 'all' high level languages
104 cleanly. It is the common code representation used throughout all phases of
105 the LLVM compilation strategy.
111 <!-- *********************************************************************** -->
112 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
113 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
114 <a name="introduction">Introduction
115 </b></font></td></tr></table><ul>
116 <!-- *********************************************************************** -->
118 The LLVM code representation is designed to be used in three different forms: as
119 an in-memory compiler IR, as an on-disk bytecode representation, suitable for
120 fast loading by a dynamic compiler, and as a human readable assembly language
121 representation. This allows LLVM to provide a powerful intermediate
122 representation for efficient compiler transformations and analysis, while
123 providing a natural means to debug and visualize the transformations. The three
124 different forms of LLVM are all equivalent. This document describes the human
125 readable representation and notation.<p>
127 The LLVM representation aims to be a light weight and low level while being
128 expressive, typed, and extensible at the same time. It aims to be a "universal
129 IR" of sorts, by being at a low enough level that high level ideas may be
130 cleanly mapped to it (similar to how microprocessors are "universal IR's",
131 allowing many source languages to be mapped to them). By providing type
132 information, LLVM can be used as the target of optimizations: for example,
133 through pointer analysis, it can be proven that a C automatic variable is never
134 accessed outside of the current function... allowing it to be promoted to a
135 simple SSA value instead of a memory location.<p>
137 <!-- _______________________________________________________________________ -->
138 </ul><a name="wellformed"><h4><hr size=0>Well Formedness</h4><ul>
140 It is important to note that this document describes 'well formed' LLVM assembly
141 language. There is a difference between what the parser accepts and what is
142 considered 'well formed'. For example, the following instruction is
143 syntactically okay, but not well formed:<p>
146 %x = <a href="#i_add">add</a> int 1, %x
149 ...because the definition of <tt>%x</tt> does not dominate all of its uses. The
150 LLVM infrastructure provides a verification pass that may be used to verify that
151 an LLVM module is well formed. This pass is automatically run by the parser
152 after parsing input assembly, and by the optimizer before it outputs bytecode.
153 The violations pointed out by the verifier pass indicate bugs in transformation
154 passes or input to the parser.<p>
156 <!-- Describe the typesetting conventions here. -->
159 <!-- *********************************************************************** -->
160 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
161 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
162 <a name="identifiers">Identifiers
163 </b></font></td></tr></table><ul>
164 <!-- *********************************************************************** -->
166 LLVM uses three different forms of identifiers, for different purposes:<p>
169 <li>Numeric constants are represented as you would expect: 12, -3 123.421, etc. Floating point constants have an optional hexidecimal notation.
170 <li>Named values are represented as a string of characters with a '%' prefix. For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
171 <li>Unnamed values are represented as an unsigned numeric value with a '%' prefix. For example, %12, %2, %44.
174 LLVM requires the values start with a '%' sign for two reasons: Compilers don't
175 need to worry about name clashes with reserved words, and the set of reserved
176 words may be expanded in the future without penalty. Additionally, unnamed
177 identifiers allow a compiler to quickly come up with a temporary variable
178 without having to avoid symbol table conflicts.<p>
180 Reserved words in LLVM are very similar to reserved words in other languages.
181 There are keywords for different opcodes ('<tt><a href="#i_add">add</a></tt>',
182 '<tt><a href="#i_cast">cast</a></tt>', '<tt><a href="#i_ret">ret</a></tt>',
183 etc...), for primitive type names ('<tt><a href="#t_void">void</a></tt>',
184 '<tt><a href="#t_uint">uint</a></tt>', etc...), and others. These reserved
185 words cannot conflict with variable names, because none of them start with a '%'
188 Here is an example of LLVM code to multiply the integer variable '<tt>%X</tt>'
193 %result = <a href="#i_mul">mul</a> uint %X, 8
196 After strength reduction:
198 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
203 <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
204 <a href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
205 %result = <a href="#i_add">add</a> uint %1, %1
208 This last way of multiplying <tt>%X</tt> by 8 illustrates several important lexical features of LLVM:<p>
211 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of line.
212 <li>Unnamed temporaries are created when the result of a computation is not
213 assigned to a named value.
214 <li>Unnamed temporaries are numbered sequentially
217 ...and it also show a convention that we follow in this document. When
218 demonstrating instructions, we will follow an instruction with a comment that
219 defines the type and name of value produced. Comments are shown in italic
222 The one non-intuitive notation for constants is the optional hexidecimal form of
223 floating point constants. For example, the form '<tt>double
224 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
225 4.5e+15</tt>' which is also supported by the parser. The only time hexadecimal
226 floating point constants are useful (and the only time that they are generated
227 by the disassembler) is when an FP constant has to be emitted that is not
228 representable as a decimal floating point number exactly. For example, NaN's,
229 infinities, and other special cases are represented in their IEEE hexadecimal
230 format so that assembly and disassembly do not cause any bits to change in the
234 <!-- *********************************************************************** -->
235 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
236 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
237 <a name="typesystem">Type System
238 </b></font></td></tr></table><ul>
239 <!-- *********************************************************************** -->
241 The LLVM type system is one of the most important features of the intermediate
242 representation. Being typed enables a number of optimizations to be performed
243 on the IR directly, without having to do extra analyses on the side before the
244 transformation. A strong type system makes it easier to read the generated code
245 and enables novel analyses and transformations that are not feasible to perform
246 on normal three address code representations.<p>
248 <!-- The written form for the type system was heavily influenced by the
249 syntactic problems with types in the C language<sup><a
250 href="#rw_stroustrup">1</a></sup>.<p> -->
254 <!-- ======================================================================= -->
255 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
256 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
257 <a name="t_primitive">Primitive Types
258 </b></font></td></tr></table><ul>
260 The primitive types are the fundemental building blocks of the LLVM system. The
261 current set of primitive types are as follows:<p>
263 <table border=0 align=center><tr><td>
265 <table border=1 cellspacing=0 cellpadding=4 align=center>
266 <tr><td><tt>void</tt></td> <td>No value</td></tr>
267 <tr><td><tt>ubyte</tt></td> <td>Unsigned 8 bit value</td></tr>
268 <tr><td><tt>ushort</tt></td><td>Unsigned 16 bit value</td></tr>
269 <tr><td><tt>uint</tt></td> <td>Unsigned 32 bit value</td></tr>
270 <tr><td><tt>ulong</tt></td> <td>Unsigned 64 bit value</td></tr>
271 <tr><td><tt>float</tt></td> <td>32 bit floating point value</td></tr>
272 <tr><td><tt>label</tt></td> <td>Branch destination</td></tr>
277 <table border=1 cellspacing=0 cellpadding=4 align=center>
278 <tr><td><tt>bool</tt></td> <td>True or False value</td></tr>
279 <tr><td><tt>sbyte</tt></td> <td>Signed 8 bit value</td></tr>
280 <tr><td><tt>short</tt></td> <td>Signed 16 bit value</td></tr>
281 <tr><td><tt>int</tt></td> <td>Signed 32 bit value</td></tr>
282 <tr><td><tt>long</tt></td> <td>Signed 64 bit value</td></tr>
283 <tr><td><tt>double</tt></td><td>64 bit floating point value</td></tr>
286 </td></tr></table><p>
290 <!-- _______________________________________________________________________ -->
291 </ul><a name="t_classifications"><h4><hr size=0>Type Classifications</h4><ul>
293 These different primitive types fall into a few useful classifications:<p>
295 <table border=1 cellspacing=0 cellpadding=4 align=center>
296 <tr><td><a name="t_signed">signed</td> <td><tt>sbyte, short, int, long, float, double</tt></td></tr>
297 <tr><td><a name="t_unsigned">unsigned</td><td><tt>ubyte, ushort, uint, ulong</tt></td></tr>
298 <tr><td><a name="t_integral">integer</td><td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td></tr>
299 <tr><td><a name="t_integral">integral</td><td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td></tr>
300 <tr><td><a name="t_floating">floating point</td><td><tt>float, double</tt></td></tr>
301 <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>
308 <!-- ======================================================================= -->
309 </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>
310 <a name="t_derived">Derived Types
311 </b></font></td></tr></table><ul>
313 The real power in LLVM comes from the derived types in the system. This is what
314 allows a programmer to represent arrays, functions, pointers, and other useful
315 types. Note that these derived types may be recursive: For example, it is
316 possible to have a two dimensional array.<p>
320 <!-- _______________________________________________________________________ -->
321 </ul><a name="t_array"><h4><hr size=0>Array Type</h4><ul>
325 The array type is a very simple derived type that arranges elements sequentially
326 in memory. The array type requires a size (number of elements) and an
327 underlying data type.<p>
331 [<# elements> x <elementtype>]
334 The number of elements is a constant integer value, elementtype may be any type
339 <tt>[40 x int ]</tt>: Array of 40 integer values.<br>
340 <tt>[41 x int ]</tt>: Array of 41 integer values.<br>
341 <tt>[40 x uint]</tt>: Array of 40 unsigned integer values.<p>
344 Here are some examples of multidimensional arrays:<p>
346 <table border=0 cellpadding=0 cellspacing=0>
347 <tr><td><tt>[3 x [4 x int]]</tt></td><td>: 3x4 array integer values.</td></tr>
348 <tr><td><tt>[12 x [10 x float]]</tt></td><td>: 2x10 array of single precision floating point values.</td></tr>
349 <tr><td><tt>[2 x [3 x [4 x uint]]]</tt></td><td>: 2x3x4 array of unsigned integer values.</td></tr>
354 <!-- _______________________________________________________________________ -->
355 </ul><a name="t_function"><h4><hr size=0>Function Type</h4><ul>
359 The function type can be thought of as a function signature. It consists of a
360 return type and a list of formal parameter types. Function types are usually
361 used when to build virtual function tables (which are structures of pointers to
362 functions), for indirect function calls, and when defining a function.<p>
366 <returntype> (<parameter list>)
369 Where '<tt><parameter list></tt>' is a comma seperated list of type
370 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
371 which indicates that the function takes a variable number of arguments. Note
372 that there currently is no way to define a function in LLVM that takes a
373 variable number of arguments, but it is possible to <b>call</b> a function that
378 <table border=0 cellpadding=0 cellspacing=0>
380 <tr><td><tt>int (int)</tt></td><td>: function taking an <tt>int</tt>, returning
381 an <tt>int</tt></td></tr>
383 <tr><td><tt>float (int, int *) *</tt></td><td>: <a href="#t_pointer">Pointer</a>
384 to a function that takes an <tt>int</tt> and a <a href="#t_pointer">pointer</a>
385 to <tt>int</tt>, returning <tt>float</tt>.</td></tr>
387 <tr><td><tt>int (sbyte *, ...)</tt></td><td>: A vararg function that takes at
388 least one <a href="#t_pointer">pointer</a> to <tt>sbyte</tt> (signed char in C),
389 which returns an integer. This is the signature for <tt>printf</tt> in
397 <!-- _______________________________________________________________________ -->
398 </ul><a name="t_struct"><h4><hr size=0>Structure Type</h4><ul>
402 The structure type is used to represent a collection of data members together in
403 memory. The packing of the field types is defined to match the ABI of the
404 underlying processor. The elements of a structure may be any type that has a
407 Structures are accessed using '<tt><a href="#i_load">load</a></tt> and '<tt><a
408 href="#i_store">store</a></tt>' by getting a pointer to a field with the '<tt><a
409 href="#i_getelementptr">getelementptr</a></tt>' instruction.<p>
413 { <type list> }
418 <table border=0 cellpadding=0 cellspacing=0>
420 <tr><td><tt>{ int, int, int }</tt></td><td>: a triple of three <tt>int</tt>
423 <tr><td><tt>{ float, int (int) * }</tt></td><td>: A pair, where the first
424 element is a <tt>float</tt> and the second element is a <a
425 href="#t_pointer">pointer</a> to a <a href="t_function">function</a> that takes
426 an <tt>int</tt>, returning an <tt>int</tt>.</td></tr>
431 <!-- _______________________________________________________________________ -->
432 </ul><a name="t_pointer"><h4><hr size=0>Pointer Type</h4><ul>
436 As in many languages, the pointer type represents a pointer or reference to
437 another object, which must live in memory.<p>
446 <table border=0 cellpadding=0 cellspacing=0>
448 <tr><td><tt>[4x int]*</tt></td><td>: <a href="#t_pointer">pointer</a> to <a
449 href="#t_array">array</a> of four <tt>int</tt> values</td></tr>
451 <tr><td><tt>int (int *) *</tt></td><td>: A <a href="#t_pointer">pointer</a> to a
452 <a href="t_function">function</a> that takes an <tt>int</tt>, returning an
453 <tt>int</tt>.</td></tr>
459 <!-- _______________________________________________________________________ -->
461 </ul><a name="t_packed"><h4><hr size=0>Packed Type</h4><ul>
463 Mention/decide that packed types work with saturation or not. Maybe have a packed+saturated type in addition to just a packed type.<p>
465 Packed types should be 'nonsaturated' because standard data types are not saturated. Maybe have a saturated packed type?<p>
470 <!-- *********************************************************************** -->
471 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
472 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
473 <a name="highlevel">High Level Structure
474 </b></font></td></tr></table><ul>
475 <!-- *********************************************************************** -->
478 <!-- ======================================================================= -->
479 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
480 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
481 <a name="modulestructure">Module Structure
482 </b></font></td></tr></table><ul>
484 LLVM programs are composed of "Module"s, each of which is a translation unit of
485 the input programs. Each module consists of functions, global variables, and
486 symbol table entries. Modules may be combined together with the LLVM linker,
487 which merges function (and global variable) definitions, resolves forward
488 declarations, and merges symbol table entries. Here is an example of the "hello world" module:<p>
491 <i>; Declare the string constant as a global constant...</i>
492 <a href="#identifiers">%.LC0</a> = <a href="#linkage_decl">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>
494 <i>; Forward declaration of puts</i>
495 <a href="#functionstructure">declare</a> int "puts"(sbyte*) <i>; int(sbyte*)* </i>
497 <i>; Definition of main function</i>
498 int "main"() { <i>; int()* </i>
499 <i>; Convert [13x sbyte]* to sbyte *...</i>
500 %cast210 = <a href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
502 <i>; Call puts function to write out the string to stdout...</i>
503 <a href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
504 <a href="#i_ret">ret</a> int 0
508 This example is made up of a <a href="#globalvars">global variable</a> named
509 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function, and a
510 <a href="#functionstructure">function definition</a> for "<tt>main</tt>".<p>
512 <a name="linkage_decl">
513 In general, a module is made up of a list of global values, where both functions
514 and global variables are global values. Global values are represented by a
515 pointer to a memory location (in this case, a pointer to an array of char, and a
516 pointer to a function), and can be either "internal" or externally accessible
517 (which corresponds to the static keyword in C, when used at global scope).<p>
519 For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
520 another module defined a "<tt>.LC0</tt>" variable and was linked with this one,
521 one of the two would be renamed, preventing a collision. Since "<tt>main</tt>"
522 and "<tt>puts</tt>" are external (i.e., lacking "<tt>internal</tt>"
523 declarations), they are accessible outside of the current module. It is illegal
524 for a function declaration to be "<tt>internal</tt>".<p>
527 <!-- ======================================================================= -->
528 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
529 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
530 <a name="globalvars">Global Variables
531 </b></font></td></tr></table><ul>
533 Global variables define regions of memory allocated at compilation time instead
534 of run-time. Global variables may optionally be initialized. A variable may
535 be defined as a global "constant", which indicates that the contents of the
536 variable will never be modified (opening options for optimization). Constants
537 must always have an initial value.<p>
539 As SSA values, global variables define pointer values that are in scope
540 (i.e. they dominate) for all basic blocks in the program. Global variables
541 always define a pointer to their "content" type because they describe a region
542 of memory, and all memory objects in LLVM are accessed through pointers.<p>
546 <!-- ======================================================================= -->
547 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
548 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
549 <a name="functionstructure">Function Structure
550 </b></font></td></tr></table><ul>
552 LLVM functions definitions are composed of a (possibly empty) argument list, an
553 opening curly brace, a list of basic blocks, and a closing curly brace. LLVM
554 function declarations are defined with the "<tt>declare</tt>" keyword, a
555 function name and a function signature.<p>
557 A function definition contains a list of basic blocks, forming the CFG for the
558 function. Each basic block may optionally start with a label (giving the basic
559 block a symbol table entry), contains a list of instructions, and ends with a <a
560 href="#terminators">terminator</a> instruction (such as a branch or function
563 The first basic block in program is special in two ways: it is immediately
564 executed on entrance to the function, and it is not allowed to have predecessor
565 basic blocks (i.e. there can not be any branches to the entry block of a
569 <!-- *********************************************************************** -->
570 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
571 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
572 <a name="instref">Instruction Reference
573 </b></font></td></tr></table><ul>
574 <!-- *********************************************************************** -->
576 The LLVM instruction set consists of several different classifications of
577 instructions: <a href="#terminators">terminator instructions</a>, <a
578 href="#binaryops">binary instructions</a>, <a href="#memoryops">memory
579 instructions</a>, and <a href="#otherops">other instructions</a>.<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="terminators">Terminator Instructions
586 </b></font></td></tr></table><ul>
588 As mentioned <a href="#functionstructure">previously</a>, every basic block in a
589 program ends with a "Terminator" instruction, which indicates which block should
590 be executed after the current block is finished. These terminator instructions
591 typically yield a '<tt>void</tt>' value: they produce control flow, not values
592 (the one exception being the '<a href="#i_invoke"><tt>invoke</tt></a>'
595 There are four different terminator instructions: the '<a
596 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a
597 href="#i_br"><tt>br</tt></a>' instruction, the '<a
598 href="#i_switch"><tt>switch</tt></a>' instruction, and the '<a
599 href="#i_invoke"><tt>invoke</tt></a>' instruction.<p>
602 <!-- _______________________________________________________________________ -->
603 </ul><a name="i_ret"><h4><hr size=0>'<tt>ret</tt>' Instruction</h4><ul>
607 ret <type> <value> <i>; Return a value from a non-void function</i>
608 ret void <i>; Return from void function</i>
613 The '<tt>ret</tt>' instruction is used to return control flow (and a value) from
614 a function, back to the caller.<p>
616 There are two forms of the '<tt>ret</tt>' instructruction: one that returns a
617 value and then causes control flow, and one that just causes control flow to
622 The '<tt>ret</tt>' instruction may return any '<a href="#t_firstclass">first
623 class</a>' type. Notice that a function is not <a href="#wellformed">well
624 formed</a> if there exists a '<tt>ret</tt>' instruction inside of the function
625 that returns a value that does not match the return type of the function.<p>
629 When the '<tt>ret</tt>' instruction is executed, control flow returns back to
630 the calling function's context. If the instruction returns a value, that value
631 shall be propagated into the calling function's data space.<p>
635 ret int 5 <i>; Return an integer value of 5</i>
636 ret void <i>; Return from a void function</i>
640 <!-- _______________________________________________________________________ -->
641 </ul><a name="i_br"><h4><hr size=0>'<tt>br</tt>' Instruction</h4><ul>
645 br bool <cond>, label <iftrue>, label <iffalse>
646 br label <dest> <i>; Unconditional branch</i>
651 The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
652 different basic block in the current function. There are two forms of this
653 instruction, corresponding to a conditional branch and an unconditional
658 The conditional branch form of the '<tt>br</tt>' instruction takes a single
659 '<tt>bool</tt>' value and two '<tt>label</tt>' values. The unconditional form
660 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
665 Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
666 argument is evaluated. If the value is <tt>true</tt>, control flows to the
667 '<tt>iftrue</tt>' '<tt>label</tt>' argument. If "cond" is <tt>false</tt>,
668 control flows to the '<tt>iffalse</tt>' '<tt>label</tt>' argument.<p>
673 %cond = <a href="#i_setcc">seteq</a> int %a, %b
674 br bool %cond, label %IfEqual, label %IfUnequal
676 <a href="#i_ret">ret</a> int 1
678 <a href="#i_ret">ret</a> int 0
682 <!-- _______________________________________________________________________ -->
683 </ul><a name="i_switch"><h4><hr size=0>'<tt>switch</tt>' Instruction</h4><ul>
687 switch int <value>, label <defaultdest> [ int <val>, label &dest>, ... ]
693 The '<tt>switch</tt>' instruction is used to transfer control flow to one of
694 several different places. It is a generalization of the '<tt>br</tt>'
695 instruction, allowing a branch to occur to one of many possible destinations.<p>
699 The '<tt>switch</tt>' instruction uses three parameters: a '<tt>uint</tt>'
700 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
701 an array of pairs of comparison value constants and '<tt>label</tt>'s.<p>
705 The <tt>switch</tt> instruction specifies a table of values and destinations.
706 When the '<tt>switch</tt>' instruction is executed, this table is searched for
707 the given value. If the value is found, the corresponding destination is
708 branched to, otherwise the default value it transfered to.<p>
710 <h5>Implementation:</h5>
712 Depending on properties of the target machine and the particular <tt>switch</tt>
713 instruction, this instruction may be code generated as a series of chained
714 conditional branches, or with a lookup table.<p>
718 <i>; Emulate a conditional br instruction</i>
719 %Val = <a href="#i_cast">cast</a> bool %value to uint
720 switch int %Val, label %truedest [int 0, label %falsedest ]
722 <i>; Emulate an unconditional br instruction</i>
723 switch int 0, label %dest [ ]
725 <i>; Implement a jump table:</i>
726 switch int %val, label %otherwise [ int 0, label %onzero,
728 int 2, label %ontwo ]
733 <!-- _______________________________________________________________________ -->
734 </ul><a name="i_invoke"><h4><hr size=0>'<tt>invoke</tt>' Instruction</h4><ul>
738 <result> = invoke <ptr to function ty> %<function ptr val>(<function args>)
739 to label <normal label> except label <exception label>
744 The '<tt>invoke</tt>' instruction is used to cause control flow to transfer to a
745 specified function, with the possibility of control flow transfer to either the
746 '<tt>normal label</tt>' label or the '<tt>exception label</tt>'. The '<tt><a
747 href="#i_call">call</a></tt>' instruction is closely related, but guarantees
748 that control flow either never returns from the called function, or that it
749 returns to the instruction following the '<tt><a href="#i_call">call</a></tt>'
754 This instruction requires several arguments:<p>
757 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
758 function value being invoked. In most cases, this is a direct function
759 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
760 an arbitrary pointer to function value.<p>
762 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
763 function to be invoked.
765 <li>'<tt>function args</tt>': argument list whose types match the function
766 signature argument types. If the function signature indicates the function
767 accepts a variable number of arguments, the extra arguments can be specified.
769 <li>'<tt>normal label</tt>': the label reached when the called function executes
770 a '<tt><a href="#i_ret">ret</a></tt>' instruction.
772 <li>'<tt>exception label</tt>': the label reached when an exception is thrown.
777 This instruction is designed to operate as a standard '<tt><a
778 href="#i_call">call</a></tt>' instruction in most regards. The primary
779 difference is that it associates a label with the function invocation that may
780 be accessed via the runtime library provided by the execution environment. This
781 instruction is used in languages with destructors to ensure that proper cleanup
782 is performed in the case of either a <tt>longjmp</tt> or a thrown exception.
783 Additionally, this is important for implementation of '<tt>catch</tt>' clauses
784 in high-level languages that support them.<p>
786 <!-- For a more comprehensive explanation of how this instruction is used, look in the llvm/docs/2001-05-18-ExceptionHandling.txt document.<p> -->
790 %retval = invoke int %Test(int 15)
791 to label %Continue except label %TestCleanup <i>; {int}:retval set</i>
796 <!-- ======================================================================= -->
797 </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>
798 <a name="binaryops">Binary Operations
799 </b></font></td></tr></table><ul>
801 Binary operators are used to do most of the computation in a program. They
802 require two operands, execute an operation on them, and produce a single value.
803 The result value of a binary operator is not neccesarily the same type as its
806 There are several different binary operators:<p>
809 <!-- _______________________________________________________________________ -->
810 </ul><a name="i_add"><h4><hr size=0>'<tt>add</tt>' Instruction</h4><ul>
814 <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
818 The '<tt>add</tt>' instruction returns the sum of its two operands.<p>
821 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>
825 The value produced is the integer or floating point sum of the two operands.<p>
829 <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
833 <!-- _______________________________________________________________________ -->
834 </ul><a name="i_sub"><h4><hr size=0>'<tt>sub</tt>' Instruction</h4><ul>
838 <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
843 The '<tt>sub</tt>' instruction returns the difference of its two operands.<p>
845 Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
846 instruction present in most other intermediate representations.<p>
850 The two arguments to the '<tt>sub</tt>' instruction must be either <a
851 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
852 values. Both arguments must have identical types.<p>
856 The value produced is the integer or floating point difference of the two
861 <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
862 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
865 <!-- _______________________________________________________________________ -->
866 </ul><a name="i_mul"><h4><hr size=0>'<tt>mul</tt>' Instruction</h4><ul>
870 <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
874 The '<tt>mul</tt>' instruction returns the product of its two operands.<p>
877 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>
881 The value produced is the integer or floating point product of the two
884 There is no signed vs unsigned multiplication. The appropriate action is taken
885 based on the type of the operand. <p>
890 <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
894 <!-- _______________________________________________________________________ -->
895 </ul><a name="i_div"><h4><hr size=0>'<tt>div</tt>' Instruction</h4><ul>
899 <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
904 The '<tt>div</tt>' instruction returns the quotient of its two operands.<p>
908 The two arguments to the '<tt>div</tt>' instruction must be either <a
909 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
910 values. Both arguments must have identical types.<p>
914 The value produced is the integer or floating point quotient of the two
919 <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
923 <!-- _______________________________________________________________________ -->
924 </ul><a name="i_rem"><h4><hr size=0>'<tt>rem</tt>' Instruction</h4><ul>
928 <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
932 The '<tt>rem</tt>' instruction returns the remainder from the division of its two operands.<p>
935 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>
939 This returns the <i>remainder</i> of a division (where the result has the same
940 sign as the divisor), not the <i>modulus</i> (where the result has the same sign
941 as the dividend) of a value. For more information about the difference, see: <a
942 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The Math
947 <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
951 <!-- _______________________________________________________________________ -->
952 </ul><a name="i_setcc"><h4><hr size=0>'<tt>set<i>cc</i></tt>' Instructions</h4><ul>
956 <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
957 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
958 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
959 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
960 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
961 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
964 <h5>Overview:</h5> The '<tt>set<i>cc</i></tt>' family of instructions returns a
965 boolean value based on a comparison of their two operands.<p>
967 <h5>Arguments:</h5> The two arguments to the '<tt>set<i>cc</i></tt>'
968 instructions must be of <a href="#t_firstclass">first class</a> or <a
969 href="#t_pointer">pointer</a> type (it is not possible to compare
970 '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>' or '<tt>void</tt>'
971 values, etc...). Both arguments must have identical types.<p>
973 The '<tt>setlt</tt>', '<tt>setgt</tt>', '<tt>setle</tt>', and '<tt>setge</tt>'
974 instructions do not operate on '<tt>bool</tt>' typed arguments.<p>
978 The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
979 both operands are equal.<br>
981 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
982 both operands are unequal.<br>
984 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
985 the first operand is less than the second operand.<br>
987 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
988 the first operand is greater than the second operand.<br>
990 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
991 the first operand is less than or equal to the second operand.<br>
993 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
994 the first operand is greater than or equal to the second operand.<p>
998 <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
999 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1000 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1001 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1002 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1003 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1008 <!-- ======================================================================= -->
1009 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1010 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
1011 <a name="bitwiseops">Bitwise Binary Operations
1012 </b></font></td></tr></table><ul>
1014 Bitwise binary operators are used to do various forms of bit-twiddling in a
1015 program. They are generally very efficient instructions, and can commonly be
1016 strength reduced from other instructions. They require two operands, execute an
1017 operation on them, and produce a single value. The resulting value of the
1018 bitwise binary operators is always the same type as its first operand.<p>
1020 <!-- _______________________________________________________________________ -->
1021 </ul><a name="i_and"><h4><hr size=0>'<tt>and</tt>' Instruction</h4><ul>
1025 <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1029 The '<tt>and</tt>' instruction returns the bitwise logical and of its two operands.<p>
1033 The two arguments to the '<tt>and</tt>' instruction must be <a
1034 href="#t_integral">integral</a> values. Both arguments must have identical
1040 The truth table used for the '<tt>and</tt>' instruction is:<p>
1042 <center><table border=1 cellspacing=0 cellpadding=4>
1043 <tr><td>In0</td> <td>In1</td> <td>Out</td></tr>
1044 <tr><td>0</td> <td>0</td> <td>0</td></tr>
1045 <tr><td>0</td> <td>1</td> <td>0</td></tr>
1046 <tr><td>1</td> <td>0</td> <td>0</td></tr>
1047 <tr><td>1</td> <td>1</td> <td>1</td></tr>
1048 </table></center><p>
1053 <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1054 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1055 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1060 <!-- _______________________________________________________________________ -->
1061 </ul><a name="i_or"><h4><hr size=0>'<tt>or</tt>' Instruction</h4><ul>
1065 <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1068 <h5>Overview:</h5> The '<tt>or</tt>' instruction returns the bitwise logical
1069 inclusive or of its two operands.<p>
1073 The two arguments to the '<tt>or</tt>' instruction must be <a
1074 href="#t_integral">integral</a> values. Both arguments must have identical
1080 The truth table used for the '<tt>or</tt>' instruction is:<p>
1082 <center><table border=1 cellspacing=0 cellpadding=4>
1083 <tr><td>In0</td> <td>In1</td> <td>Out</td></tr>
1084 <tr><td>0</td> <td>0</td> <td>0</td></tr>
1085 <tr><td>0</td> <td>1</td> <td>1</td></tr>
1086 <tr><td>1</td> <td>0</td> <td>1</td></tr>
1087 <tr><td>1</td> <td>1</td> <td>1</td></tr>
1088 </table></center><p>
1093 <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1094 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1095 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1099 <!-- _______________________________________________________________________ -->
1100 </ul><a name="i_xor"><h4><hr size=0>'<tt>xor</tt>' Instruction</h4><ul>
1104 <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1109 The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of its
1114 The two arguments to the '<tt>xor</tt>' instruction must be <a
1115 href="#t_integral">integral</a> values. Both arguments must have identical
1121 The truth table used for the '<tt>xor</tt>' instruction is:<p>
1123 <center><table border=1 cellspacing=0 cellpadding=4>
1124 <tr><td>In0</td> <td>In1</td> <td>Out</td></tr>
1125 <tr><td>0</td> <td>0</td> <td>0</td></tr>
1126 <tr><td>0</td> <td>1</td> <td>1</td></tr>
1127 <tr><td>1</td> <td>0</td> <td>1</td></tr>
1128 <tr><td>1</td> <td>1</td> <td>0</td></tr>
1129 </table></center><p>
1134 <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1135 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1136 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1140 <!-- _______________________________________________________________________ -->
1141 </ul><a name="i_shl"><h4><hr size=0>'<tt>shl</tt>' Instruction</h4><ul>
1145 <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1150 The '<tt>shl</tt>' instruction returns the first operand shifted to the left a
1151 specified number of bits.
1155 The first argument to the '<tt>shl</tt>' instruction must be an <a
1156 href="#t_integer">integer</a> type. The second argument must be an
1157 '<tt>ubyte</tt>' type.<p>
1161 The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.<p>
1166 <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1167 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1168 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1172 <!-- _______________________________________________________________________ -->
1173 </ul><a name="i_shr"><h4><hr size=0>'<tt>shr</tt>' Instruction</h4><ul>
1178 <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1182 The '<tt>shr</tt>' instruction returns the first operand shifted to the right a specified number of bits.
1185 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>
1189 If the first argument is a <a href="#t_signed">signed</a> type, the most
1190 significant bit is duplicated in the newly free'd bit positions. If the first
1191 argument is unsigned, zero bits shall fill the empty positions.<p>
1195 <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1196 <result> = shr int 4, ubyte 1 <i>; yields {int}:result = 2</i>
1197 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1198 <result> = shr int 4, ubyte 3 <i>; yields {int}:result = 0</i>
1205 <!-- ======================================================================= -->
1206 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1207 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
1208 <a name="memoryops">Memory Access Operations
1209 </b></font></td></tr></table><ul>
1211 Accessing memory in SSA form is, well, sticky at best. This section describes how to read, write, allocate and free memory in LLVM.<p>
1214 <!-- _______________________________________________________________________ -->
1215 </ul><a name="i_malloc"><h4><hr size=0>'<tt>malloc</tt>' Instruction</h4><ul>
1219 <result> = malloc <type>, uint <NumElements> <i>; yields {type*}:result</i>
1220 <result> = malloc <type> <i>; yields {type*}:result</i>
1224 The '<tt>malloc</tt>' instruction allocates memory from the system heap and returns a pointer to it.<p>
1228 The the '<tt>malloc</tt>' instruction allocates
1229 <tt>sizeof(<type>)*NumElements</tt> bytes of memory from the operating
1230 system, and returns a pointer of the appropriate type to the program. The
1231 second form of the instruction is a shorter version of the first instruction
1232 that defaults to allocating one element.<p>
1234 '<tt>type</tt>' must be a sized type<p>
1237 Memory is allocated, a pointer is returned.<p>
1241 %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1243 %size = <a href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1244 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1245 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1249 <!-- _______________________________________________________________________ -->
1250 </ul><a name="i_free"><h4><hr size=0>'<tt>free</tt>' Instruction</h4><ul>
1254 free <type> <value> <i>; yields {void}</i>
1259 The '<tt>free</tt>' instruction returns memory back to the unused memory heap, to be reallocated in the future.<p>
1264 '<tt>value</tt>' shall be a pointer value that points to a value that was
1265 allocated with the '<tt><a href="#i_malloc">malloc</a></tt>' instruction.<p>
1270 Access to the memory pointed to by the pointer is not longer defined after this instruction executes.<p>
1274 %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1275 free [4 x ubyte]* %array
1279 <!-- _______________________________________________________________________ -->
1280 </ul><a name="i_alloca"><h4><hr size=0>'<tt>alloca</tt>' Instruction</h4><ul>
1284 <result> = alloca <type>, uint <NumElements> <i>; yields {type*}:result</i>
1285 <result> = alloca <type> <i>; yields {type*}:result</i>
1290 The '<tt>alloca</tt>' instruction allocates memory on the current stack frame of
1291 the procedure that is live until the current function returns to its caller.<p>
1295 The the '<tt>alloca</tt>' instruction allocates
1296 <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the runtime stack,
1297 returning a pointer of the appropriate type to the program. The second form of
1298 the instruction is a shorter version of the first that defaults to allocating
1301 '<tt>type</tt>' may be any sized type.<p>
1305 Memory is allocated, a pointer is returned. '<tt>alloca</tt>'d memory is
1306 automatically released when the function returns. The '<tt>alloca</tt>'
1307 instruction is commonly used to represent automatic variables that must have an
1308 address available, as well as spilled variables.<p>
1312 %ptr = alloca int <i>; yields {int*}:ptr</i>
1313 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
1317 <!-- _______________________________________________________________________ -->
1318 </ul><a name="i_load"><h4><hr size=0>'<tt>load</tt>' Instruction</h4><ul>
1322 <result> = load <ty>* <pointer>
1326 The '<tt>load</tt>' instruction is used to read from memory.<p>
1330 The argument to the '<tt>load</tt>' instruction specifies the memory address to load from. The pointer must point to a <a href="t_firstclass">first class</a> type.<p>
1334 The location of memory pointed to is loaded.
1338 %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1339 <a href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1340 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1346 <!-- _______________________________________________________________________ -->
1347 </ul><a name="i_store"><h4><hr size=0>'<tt>store</tt>' Instruction</h4><ul>
1351 store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1355 The '<tt>store</tt>' instruction is used to write to memory.<p>
1359 There are two arguments to the '<tt>store</tt>' instruction: a value to store
1360 and an address to store it into. The type of the '<tt><pointer></tt>'
1361 operand must be a pointer to the type of the '<tt><value></tt>'
1364 <h5>Semantics:</h5> The contents of memory are updated to contain
1365 '<tt><value></tt>' at the location specified by the
1366 '<tt><pointer></tt>' operand.<p>
1370 %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1371 <a href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1372 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1378 <!-- _______________________________________________________________________ -->
1379 </ul><a name="i_getelementptr"><h4><hr size=0>'<tt>getelementptr</tt>' Instruction</h4><ul>
1383 <result> = getelementptr <ty>* <ptrval>{, long <aidx>|, ubyte <sidx>}*
1388 The '<tt>getelementptr</tt>' instruction is used to get the address of a
1389 subelement of an aggregate data structure.<p>
1393 This instruction takes a list of <tt>long</tt> values and <tt>ubyte</tt>
1394 constants that indicate what form of addressing to perform. The actual types of
1395 the arguments provided depend on the type of the first pointer argument. The
1396 '<tt>getelementptr</tt>' instruction is used to index down through the type
1397 levels of a structure.<p>
1399 For example, lets consider a C code fragment and how it gets compiled to
1414 int *foo(struct ST *s) {
1415 return &s[1].Z.B[5][13];
1419 The LLVM code generated by the GCC frontend is:
1422 %RT = type { sbyte, [10 x [20 x int]], sbyte }
1423 %ST = type { int, double, %RT }
1425 int* "foo"(%ST* %s) {
1426 %reg = getelementptr %ST* %s, long 1, ubyte 2, ubyte 1, long 5, long 13
1433 The index types specified for the '<tt>getelementptr</tt>' instruction depend on
1434 the pointer type that is being index into. <a href="t_pointer">Pointer</a> and
1435 <a href="t_array">array</a> types require '<tt>long</tt>' values, and <a
1436 href="t_struct">structure</a> types require '<tt>ubyte</tt>'
1437 <b>constants</b>.<p>
1439 In the example above, the first index is indexing into the '<tt>%ST*</tt>' type,
1440 which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT }</tt>'
1441 type, a structure. The second index indexes into the third element of the
1442 structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]], sbyte
1443 }</tt>' type, another structure. The third index indexes into the second
1444 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
1445 array. The two dimensions of the array are subscripted into, yielding an
1446 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction return a pointer
1447 to this element, thus yielding a '<tt>int*</tt>' type.<p>
1449 Note that it is perfectly legal to index partially through a structure,
1450 returning a pointer to an inner element. Because of this, the LLVM code for the
1451 given testcase is equivalent to:<p>
1454 int* "foo"(%ST* %s) {
1455 %t1 = getelementptr %ST* %s , long 1 <i>; yields %ST*:%t1</i>
1456 %t2 = getelementptr %ST* %t1, long 0, ubyte 2 <i>; yields %RT*:%t2</i>
1457 %t3 = getelementptr %RT* %t2, long 0, ubyte 1 <i>; yields [10 x [20 x int]]*:%t3</i>
1458 %t4 = getelementptr [10 x [20 x int]]* %t3, long 0, long 5 <i>; yields [20 x int]*:%t4</i>
1459 %t5 = getelementptr [20 x int]* %t4, long 0, long 13 <i>; yields int*:%t5</i>
1468 <i>; yields [12 x ubyte]*:aptr</i>
1469 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, ubyte 1
1474 <!-- ======================================================================= -->
1475 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1476 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
1477 <a name="otherops">Other Operations
1478 </b></font></td></tr></table><ul>
1480 The instructions in this catagory are the "miscellaneous" functions, that defy better classification.<p>
1483 <!-- _______________________________________________________________________ -->
1484 </ul><a name="i_phi"><h4><hr size=0>'<tt>phi</tt>' Instruction</h4><ul>
1488 <result> = phi <ty> [ <val0>, <label0>], ...
1493 The '<tt>phi</tt>' instruction is used to implement the φ node in the SSA
1494 graph representing the function.<p>
1498 The type of the incoming values are specified with the first type field. After
1499 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
1500 one pair for each predecessor basic block of the current block.<p>
1502 There must be no non-phi instructions between the start of a basic block and the
1503 PHI instructions: i.e. PHI instructions must be first in a basic block.<p>
1507 At runtime, the '<tt>phi</tt>' instruction logically takes on the value
1508 specified by the parameter, depending on which basic block we came from in the
1509 last <a href="#terminators">terminator</a> instruction.<p>
1514 Loop: ; Infinite loop that counts from 0 on up...
1515 %indvar = phi uint [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
1516 %nextindvar = add uint %indvar, 1
1521 <!-- _______________________________________________________________________ -->
1522 </ul><a name="i_cast"><h4><hr size=0>'<tt>cast .. to</tt>' Instruction</h4><ul>
1526 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
1531 The '<tt>cast</tt>' instruction is used as the primitive means to convert
1532 integers to floating point, change data type sizes, and break type safety (by
1533 casting pointers).<p>
1537 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
1538 class value, and a type to cast it to, which must also be a first class type.<p>
1542 This instruction follows the C rules for explicit casts when determining how the
1543 data being cast must change to fit in its new container.<p>
1545 When casting to bool, any value that would be considered true in the context of
1546 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
1547 all else are '<tt>false</tt>'.<p>
1549 When extending an integral value from a type of one signness to another (for
1550 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
1551 <b>source</b> value is signed, and zero-extended if the source value is
1552 unsigned. <tt>bool</tt> values are always zero extended into either zero or
1557 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
1558 %Y = cast int 123 to bool <i>; yields bool:true</i>
1563 <!-- _______________________________________________________________________ -->
1564 </ul><a name="i_call"><h4><hr size=0>'<tt>call</tt>' Instruction</h4><ul>
1568 <result> = call <ty>* <fnptrval>(<param list>)
1573 The '<tt>call</tt>' instruction represents a simple function call.<p>
1577 This instruction requires several arguments:<p>
1580 <li>'<tt>ty</tt>': shall be the signature of the pointer to function value being
1581 invoked. The argument types must match the types implied by this signature.<p>
1583 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to be
1584 invoked. In most cases, this is a direct function invocation, but indirect
1585 <tt>call</tt>s are just as possible, calling an arbitrary pointer to function
1588 <li>'<tt>function args</tt>': argument list whose types match the function
1589 signature argument types. If the function signature indicates the function
1590 accepts a variable number of arguments, the extra arguments can be specified.
1595 The '<tt>call</tt>' instruction is used to cause control flow to transfer to a
1596 specified function, with its incoming arguments bound to the specified values.
1597 Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called function,
1598 control flow continues with the instruction after the function call, and the
1599 return value of the function is bound to the result argument. This is a simpler
1600 case of the <a href="#i_invoke">invoke</a> instruction.<p>
1604 %retval = call int %test(int %argc)
1605 call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);
1609 <!-- _______________________________________________________________________ -->
1610 </ul><a name="i_va_arg"><h4><hr size=0>'<tt>va_arg</tt>' Instruction</h4><ul>
1614 <result> = va_arg <va_list>* <arglist>, <retty>
1619 The '<tt>va_arg</tt>' instruction is used to access arguments passed through the
1620 "variable argument" area of a function call. It corresponds directly to the
1621 <tt>va_arg</tt> macro in C.<p>
1625 This instruction takes a pointer to a <tt>valist</tt> value to read a new
1626 argument from. The return type of the instruction is defined by the second
1627 argument, a type.<p>
1631 The '<tt>va_arg</tt>' instruction works just like the <tt>va_arg</tt> macro
1632 available in C. In a target-dependent way, it reads the argument indicated by
1633 the value the arglist points to, updates the arglist, then returns a value of
1634 the specified type. This instruction should be used in conjunction with the
1635 variable argument handling <a href="#int_varargs">Intrinsic Functions</a>.<p>
1637 It is legal for this instruction to be called in a function which does not take
1638 a variable number of arguments, for example, the <tt>vfprintf</tt> function.<p>
1640 <tt>va_arg</tt> is an LLVM instruction instead of an <a
1641 href="#intrinsics">intrinsic function</a> because the return type depends on an
1646 See the <a href="#int_varargs">variable argument processing</a> section.<p>
1648 <!-- *********************************************************************** -->
1649 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
1650 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
1651 <a name="intrinsics">Intrinsic Functions
1652 </b></font></td></tr></table><ul>
1653 <!-- *********************************************************************** -->
1655 LLVM supports the notion of an "intrinsic function". These functions have well
1656 known names and semantics, and are required to follow certain restrictions.
1657 Overall, these instructions represent an extension mechanism for the LLVM
1658 language that does not require changing all of the transformations in LLVM to
1659 add to the language (or the bytecode reader/writer, the parser, etc...).<p>
1661 Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix, this
1662 prefix is reserved in LLVM for intrinsic names, thus functions may not be named
1663 this. Intrinsic functions must always be external functions: you cannot define
1664 the body of intrinsic functions. Intrinsic functions may only be used in call
1665 or invoke instructions: it is illegal to take the address of an intrinsic
1666 function. Additionally, because intrinsic functions are part of the LLVM
1667 language, it is required that they all be documented here if any are added.<p>
1669 Unless an intrinsic function is target-specific, there must be a lowering pass
1670 to eliminate the intrinsic or all backends must support the intrinsic
1674 <!-- ======================================================================= -->
1675 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1676 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
1677 <a name="int_varargs">Variable Argument Handling Intrinsics
1678 </b></font></td></tr></table><ul>
1680 Variable argument support is defined in LLVM with the <a
1681 href="#i_va_arg"><tt>va_arg</tt></a> instruction and these three intrinsic
1682 functions. These function correspond almost directly to the similarly named
1683 macros defined in the <tt><stdarg.h></tt> header file.<p>
1685 All of these functions operate on arguments that use a target-specific type
1686 "<tt>va_list</tt>". The LLVM assembly language reference manual does not define
1687 what this type is, so all transformations should be prepared to handle
1688 intrinsics with any type used.<p>
1690 This example shows how the <a href="#i_va_arg"><tt>va_arg</tt></a> instruction
1691 and the variable argument handling intrinsic functions are used.<p>
1694 int %test(int %X, ...) {
1695 ; Allocate two va_list items. On this target, va_list is of type sbyte*
1699 ; Initialize variable argument processing
1700 call void (sbyte**)* %<a href="#i_va_start">llvm.va_start</a>(sbyte** %ap)
1702 ; Read a single integer argument
1703 %tmp = <a href="#i_va_arg">va_arg</a> sbyte** %ap, int
1705 ; Demonstrate usage of llvm.va_copy and llvm_va_end
1706 %apv = load sbyte** %ap
1707 call void %<a href="#i_va_copy">llvm.va_copy</a>(sbyte** %aq, sbyte* %apv)
1708 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %aq)
1710 ; Stop processing of arguments.
1711 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %ap)
1716 <!-- _______________________________________________________________________ -->
1717 </ul><a name="i_va_start"><h4><hr size=0>'<tt>llvm.va_start</tt>' Intrinsic</h4><ul>
1721 call void (va_list*)* %llvm.va_start(<va_list>* <arglist>)
1726 The '<tt>llvm.va_start</tt>' intrinsic initializes <tt>*<arglist></tt> for
1727 subsequent use by <tt><a href="#i_va_arg">va_arg</a></tt> and <tt><a
1728 href="#i_va_end">llvm.va_end</a></tt>, and must be called before either are
1733 The argument is a pointer to a <tt>va_list</tt> element to initialize.<p>
1737 The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
1738 macro available in C. In a target-dependent way, it initializes the
1739 <tt>va_list</tt> element the argument points to, so that the next call to
1740 <tt>va_arg</tt> will produce the first variable argument passed to the function.
1741 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
1742 last argument of the function, the compiler can figure that out.<p>
1745 <!-- _______________________________________________________________________ -->
1746 </ul><a name="i_va_end"><h4><hr size=0>'<tt>llvm.va_end</tt>' Intrinsic</h4><ul>
1750 call void (va_list*)* %llvm.va_end(<va_list>* <arglist>)
1755 The '<tt>llvm.va_end</tt>' intrinsic destroys <tt>*<arglist></tt> which
1756 has been initialized previously with <tt><a
1757 href="#i_va_begin">llvm.va_begin</a></tt>.<p>
1761 The argument is a pointer to a <tt>va_list</tt> element to destroy.<p>
1765 The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt> macro
1766 available in C. In a target-dependent way, it destroys the <tt>va_list</tt>
1767 that the argument points to. Calls to <a
1768 href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
1769 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly with calls
1770 to <tt>llvm.va_end</tt>.<p>
1774 <!-- _______________________________________________________________________ -->
1775 </ul><a name="i_va_copy"><h4><hr size=0>'<tt>llvm.va_copy</tt>' Intrinsic</h4><ul>
1779 call void (va_list*, va_list)* %va_copy(<va_list>* <destarglist>,
1780 <va_list> <srcarglist>)
1785 The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position from
1786 the source argument list to the destination argument list.<p>
1790 The first argument is a pointer to a <tt>va_list</tt> element to initialize.
1791 The second argument is a <tt>va_list</tt> element to copy from.<p>
1796 The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt> macro
1797 available in C. In a target-dependent way, it copies the source
1798 <tt>va_list</tt> element into the destination list. This intrinsic is necessary
1799 because the <tt><a href="i_va_begin">llvm.va_begin</a></tt> intrinsic may be
1800 arbitrarily complex and require memory allocation, for example.<p>
1803 <!-- *********************************************************************** -->
1805 <!-- *********************************************************************** -->
1810 <address><a href="mailto:sabre@nondot.org">Chris Lattner</a></address>
1811 <!-- Created: Tue Jan 23 15:19:28 CST 2001 -->
1812 <!-- hhmts start -->
1813 Last modified: Thu May 8 10:48:46 CDT 2003