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 Assembly 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="#unaryops">Unary Operations</a>
45 <li><a href="#i_not" >'<tt>not</tt>' Instruction</a>
47 <li><a href="#binaryops">Binary Operations</a>
49 <li><a href="#i_add" >'<tt>add</tt>' Instruction</a>
50 <li><a href="#i_sub" >'<tt>sub</tt>' Instruction</a>
51 <li><a href="#i_mul" >'<tt>mul</tt>' Instruction</a>
52 <li><a href="#i_div" >'<tt>div</tt>' Instruction</a>
53 <li><a href="#i_rem" >'<tt>rem</tt>' Instruction</a>
54 <li><a href="#i_setcc">'<tt>set<i>cc</i></tt>' Instructions</a>
56 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
58 <li><a href="#i_and">'<tt>and</tt>' Instruction</a>
59 <li><a href="#i_or" >'<tt>or</tt>' Instruction</a>
60 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a>
61 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a>
62 <li><a href="#i_shr">'<tt>shr</tt>' Instruction</a>
64 <li><a href="#memoryops">Memory Access Operations</a>
66 <li><a href="#i_malloc" >'<tt>malloc</tt>' Instruction</a>
67 <li><a href="#i_free" >'<tt>free</tt>' Instruction</a>
68 <li><a href="#i_alloca" >'<tt>alloca</tt>' Instruction</a>
69 <li><a href="#i_load" >'<tt>load</tt>' Instruction</a>
70 <li><a href="#i_store" >'<tt>store</tt>' Instruction</a>
71 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
73 <li><a href="#otherops">Other Operations</a>
75 <li><a href="#i_phi" >'<tt>phi</tt>' Instruction</a>
76 <li><a href="#i_cast">'<tt>cast .. to</tt>' Instruction</a>
77 <li><a href="#i_call" >'<tt>call</tt>' Instruction</a>
81 <li><a href="#related">Related Work</a>
86 <!-- *********************************************************************** -->
87 <p><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
88 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
89 <a name="abstract">Abstract
90 </b></font></td></tr></table><ul>
91 <!-- *********************************************************************** -->
94 This document describes the LLVM assembly language. LLVM is an SSA based
95 representation that is a useful midlevel IR, providing type safety, low level
96 operations, flexibility, and the capability of representing 'all' high level
103 <!-- *********************************************************************** -->
104 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
105 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
106 <a name="introduction">Introduction
107 </b></font></td></tr></table><ul>
108 <!-- *********************************************************************** -->
110 The LLVM code representation is designed to be used in three different forms: as
111 an in-memory compiler IR, as an on-disk bytecode representation, suitable for
112 fast loading by a dynamic compiler, and as a human readable assembly language
113 representation. This allows LLVM to provide a powerful intermediate
114 representation for efficient compiler transformations and analysis, while
115 providing a natural means to debug and visualize the transformations. The three
116 different forms of LLVM are all equivalent. This document describes the human
117 readable representation and notation.<p>
119 The LLVM representation aims to be a light weight and low level while being
120 expressive, type safe, and extensible at the same time. It aims to be a
121 "universal IR" of sorts, by being at a low enough level that high level ideas
122 may be cleanly mapped to it (similar to how microprocessors are "universal
123 IR's", allowing many source languages to be mapped to them). By providing type
124 safety, LLVM can be used as the target of optimizations: for example, through
125 pointer analysis, it can be proven that a C automatic variable is never accessed
126 outside of the current function... allowing it to be promoted to a simple SSA
127 value instead of a memory location.<p>
129 <!-- _______________________________________________________________________ -->
130 </ul><a name="wellformed"><h4><hr size=0>Well Formedness</h4><ul>
132 It is important to note that this document describes 'well formed' llvm assembly
133 language. There is a difference between what the parser accepts and what is
134 considered 'well formed'. For example, the following instruction is
135 syntactically okay, but not well formed:<p>
138 %x = <a href="#i_add">add</a> int 1, %x
141 ...because only a <tt><a href="#i_phi">phi</a></tt> node may refer to itself.
142 The LLVM api provides a verification pass (created by the
143 <tt>createVerifierPass</tt> function) that may be used to verify that an LLVM
144 module is well formed. This pass is automatically run by the parser after
145 parsing input assembly, and by the optimizer before it outputs bytecode. The
146 violations pointed out by the verifier pass indicate bugs in transformation
147 passes or input to the parser.<p>
149 Describe the typesetting conventions here.
152 <!-- *********************************************************************** -->
153 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
154 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
155 <a name="identifiers">Identifiers
156 </b></font></td></tr></table><ul>
157 <!-- *********************************************************************** -->
159 LLVM uses three different forms of identifiers, for different purposes:<p>
162 <li>Numeric constants are represented as you would expect: 12, -3 123.421, etc. Floating point constants have an optional hexidecimal notation.
163 <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>'.
164 <li>Unnamed values are represented as an unsigned numeric value with a '%' prefix. For example, %12, %2, %44.
167 LLVM requires the values start with a '%' sign for two reasons: Compilers don't
168 need to worry about name clashes with reserved words, and the set of reserved
169 words may be expanded in the future without penalty. Additionally, unnamed
170 identifiers allow a compiler to quickly come up with a temporary variable
171 without having to avoid symbol table conflicts.<p>
173 Reserved words in LLVM are very similar to reserved words in other languages.
174 There are keywords for different opcodes ('<tt><a href="#i_add">add</a></tt>',
175 '<tt><a href="#i_cast">cast</a></tt>', '<tt><a href="#i_ret">ret</a></tt>',
176 etc...), for primitive type names ('<tt><a href="#t_void">void</a></tt>',
177 '<tt><a href="#t_uint">uint</a></tt>', etc...), and others. These reserved
178 words cannot conflict with variable names, because none of them start with a '%'
181 Here is an example of LLVM code to multiply the integer variable '<tt>%X</tt>'
186 %result = <a href="#i_mul">mul</a> uint %X, 8
189 After strength reduction:
191 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
196 <a href="#i_add">add</a> uint %X, %X <i>; yields {int}:%0</i>
197 <a href="#i_add">add</a> uint %0, %0 <i>; yields {int}:%1</i>
198 %result = <a href="#i_add">add</a> uint %1, %1
201 This last way of multiplying <tt>%X</tt> by 8 illustrates several important lexical features of LLVM:<p>
204 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of line.
205 <li>Unnamed temporaries are created when the result of a computation is not
206 assigned to a named value.
207 <li>Unnamed temporaries are numbered sequentially
210 ...and it also show a convention that we follow in this document. When
211 demonstrating instructions, we will follow an instruction with a comment that
212 defines the type and name of value produced. Comments are shown in italic
215 The one unintuitive notation for constants is the optional hexidecimal form of
216 floating point constants. For example, the form '<tt>double
217 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
218 4.5e+15</tt>' which is also supported by the parser. The only time hexadecimal
219 floating point constants are useful (and the only time that they are generated
220 by the disassembler) is when an FP constant has to be emitted that is not
221 representable as a decimal floating point number exactly. For example, NaN's,
222 infinities, and other special cases are represented in their IEEE hexadecimal
223 format so that assembly and disassembly do not cause any bits to change in the
227 <!-- *********************************************************************** -->
228 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
229 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
230 <a name="typesystem">Type System
231 </b></font></td></tr></table><ul>
232 <!-- *********************************************************************** -->
234 The LLVM type system is one of the most important features of the intermediate
235 representation. Being strongly typed enables a number of optimizations to be
236 performed on the IR directly, without having to do extra analyses on the side
237 before the transformation. A strong type system makes it easier to read the
238 generated code and enables novel analyses and transformations that are not
239 feasible to perform on normal three address code representations.<p>
241 The written form for the type system was heavily influenced by the syntactic
242 problems with types in the C language<sup><a
243 href="#rw_stroustrup">1</a></sup>.<p>
247 <!-- ======================================================================= -->
248 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
249 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
250 <a name="t_primitive">Primitive Types
251 </b></font></td></tr></table><ul>
253 The primitive types are the fundemental building blocks of the LLVM system. The
254 current set of primitive types are as follows:<p>
256 <table border=0 align=center><tr><td>
258 <table border=1 cellspacing=0 cellpadding=4 align=center>
259 <tr><td><tt>void</tt></td> <td>No value</td></tr>
260 <tr><td><tt>ubyte</tt></td> <td>Unsigned 8 bit value</td></tr>
261 <tr><td><tt>ushort</tt></td><td>Unsigned 16 bit value</td></tr>
262 <tr><td><tt>uint</tt></td> <td>Unsigned 32 bit value</td></tr>
263 <tr><td><tt>ulong</tt></td> <td>Unsigned 64 bit value</td></tr>
264 <tr><td><tt>float</tt></td> <td>32 bit floating point value</td></tr>
265 <tr><td><tt>label</tt></td> <td>Branch destination</td></tr>
270 <table border=1 cellspacing=0 cellpadding=4 align=center>
271 <tr><td><tt>bool</tt></td> <td>True or False value</td></tr>
272 <tr><td><tt>sbyte</tt></td> <td>Signed 8 bit value</td></tr>
273 <tr><td><tt>short</tt></td> <td>Signed 16 bit value</td></tr>
274 <tr><td><tt>int</tt></td> <td>Signed 32 bit value</td></tr>
275 <tr><td><tt>long</tt></td> <td>Signed 64 bit value</td></tr>
276 <tr><td><tt>double</tt></td><td>64 bit floating point value</td></tr>
279 </td></tr></table><p>
283 <!-- _______________________________________________________________________ -->
284 </ul><a name="t_classifications"><h4><hr size=0>Type Classifications</h4><ul>
286 These different primitive types fall into a few useful classifications:<p>
288 <table border=1 cellspacing=0 cellpadding=4 align=center>
289 <tr><td><a name="t_signed">signed</td> <td><tt>sbyte, short, int, long, float, double</tt></td></tr>
290 <tr><td><a name="t_unsigned">unsigned</td><td><tt>ubyte, ushort, uint, ulong</tt></td></tr>
291 <tr><td><a name="t_integral">integral</td><td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td></tr>
292 <tr><td><a name="t_floating">floating point</td><td><tt>float, double</tt></td></tr>
293 <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>
300 <!-- ======================================================================= -->
301 </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>
302 <a name="t_derived">Derived Types
303 </b></font></td></tr></table><ul>
305 The real power in LLVM comes from the derived types in the system. This is what
306 allows a programmer to represent arrays, functions, pointers, and other useful
307 types. Note that these derived types may be recursive: For example, it is
308 possible to have a two dimensional array.<p>
312 <!-- _______________________________________________________________________ -->
313 </ul><a name="t_array"><h4><hr size=0>Array Type</h4><ul>
317 The array type is a very simple derived type that arranges elements sequentially
318 in memory. The array type requires a size (number of elements) and an
319 underlying data type.<p>
323 [<# elements> x <elementtype>]
326 The number of elements is a constant integer value, elementtype may be any type
331 <tt>[40 x int ]</tt>: Array of 40 integer values.<br>
332 <tt>[41 x int ]</tt>: Array of 41 integer values.<br>
333 <tt>[40 x uint]</tt>: Array of 40 unsigned integer values.<p>
336 Here are some examples of multidimensional arrays:<p>
338 <table border=0 cellpadding=0 cellspacing=0>
339 <tr><td><tt>[3 x [4 x int]]</tt></td><td>: 3x4 array integer values.</td></tr>
340 <tr><td><tt>[12 x [10 x float]]</tt></td><td>: 2x10 array of single precision floating point values.</td></tr>
341 <tr><td><tt>[2 x [3 x [4 x uint]]]</tt></td><td>: 2x3x4 array of unsigned integer values.</td></tr>
346 <!-- _______________________________________________________________________ -->
347 </ul><a name="t_function"><h4><hr size=0>Function Type</h4><ul>
351 The function type can be thought of as a function signature. It consists of a
352 return type and a list of formal parameter types. Function types are usually
353 used when to build virtual function tables (which are structures of pointers to
354 functions), for indirect function calls, and when defining a function.<p>
358 <returntype> (<parameter list>)
361 Where '<tt><parameter list></tt>' is a comma seperated list of type
362 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
363 which indicates that the function takes a variable number of arguments. Note
364 that there currently is no way to define a function in LLVM that takes a
365 variable number of arguments, but it is possible to <b>call</b> a function that
370 <table border=0 cellpadding=0 cellspacing=0>
372 <tr><td><tt>int (int)</tt></td><td>: function taking an <tt>int</tt>, returning
373 an <tt>int</tt></td></tr>
375 <tr><td><tt>float (int, int *) *</tt></td><td>: <a href="#t_pointer">Pointer</a>
376 to a function that takes an <tt>int</tt> and a <a href="#t_pointer">pointer</a>
377 to <tt>int</tt>, returning <tt>float</tt>.</td></tr>
379 <tr><td><tt>int (sbyte *, ...)</tt></td><td>: A vararg function that takes at
380 least one <a href="#t_pointer">pointer</a> to <tt>sbyte</tt> (signed char in C),
381 which returns an integer. This is the signature for <tt>printf</tt> in
389 <!-- _______________________________________________________________________ -->
390 </ul><a name="t_struct"><h4><hr size=0>Structure Type</h4><ul>
394 The structure type is used to represent a collection of data members together in
395 memory. Although the runtime is allowed to lay out the data members any way
396 that it would like, they are guaranteed to be "close" to each other.<p>
398 Structures are accessed using '<tt><a href="#i_load">load</a></tt> and '<tt><a
399 href="#i_store">store</a></tt>' by getting a pointer to a field with the '<tt><a
400 href="#i_getelementptr">getelementptr</a></tt>' instruction.<p>
404 { <type list> }
409 <table border=0 cellpadding=0 cellspacing=0>
411 <tr><td><tt>{ int, int, int }</tt></td><td>: a triple of three <tt>int</tt>
414 <tr><td><tt>{ float, int (int *) * }</tt></td><td>: A pair, where the first
415 element is a <tt>float</tt> and the second element is a <a
416 href="#t_pointer">pointer</a> to a <a href="t_function">function</a> that takes
417 an <tt>int</tt>, returning an <tt>int</tt>.</td></tr>
422 <!-- _______________________________________________________________________ -->
423 </ul><a name="t_pointer"><h4><hr size=0>Pointer Type</h4><ul>
427 As in many languages, the pointer type represents a pointer or reference to
428 another object, which must live in memory.<p>
437 <table border=0 cellpadding=0 cellspacing=0>
439 <tr><td><tt>[4x int]*</tt></td><td>: <a href="#t_pointer">pointer</a> to <a
440 href="#t_array">array</a> of four <tt>int</tt> values</td></tr>
442 <tr><td><tt>int (int *) *</tt></td><td>: A <a href="#t_pointer">pointer</a> to a
443 <a href="t_function">function</a> that takes an <tt>int</tt>, returning an
444 <tt>int</tt>.</td></tr>
450 <!-- _______________________________________________________________________ -->
452 </ul><a name="t_packed"><h4><hr size=0>Packed Type</h4><ul>
454 Mention/decide that packed types work with saturation or not. Maybe have a packed+saturated type in addition to just a packed type.<p>
456 Packed types should be 'nonsaturated' because standard data types are not saturated. Maybe have a saturated packed type?<p>
461 <!-- *********************************************************************** -->
462 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
463 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
464 <a name="highlevel">High Level Structure
465 </b></font></td></tr></table><ul>
466 <!-- *********************************************************************** -->
469 <!-- ======================================================================= -->
470 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
471 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
472 <a name="modulestructure">Module Structure
473 </b></font></td></tr></table><ul>
475 LLVM programs are composed of "Module"s, each of which is a translation unit of
476 the input programs. Each module consists of functions, global variables, and
477 symbol table entries. Modules may be combined together with the LLVM linker,
478 which merges function (and global variable) definitions, resolves forward
479 declarations, and merges symbol table entries. Here is an example of the "hello world" module:<p>
482 <i>; Declare the string constant as a global constant...</i>
483 <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>
485 <i>; Forward declaration of puts</i>
486 <a href="#functionstructure">declare</a> int "puts"(sbyte*) <i>; int(sbyte*)* </i>
488 <i>; Definition of main function</i>
489 int "main"() { <i>; int()* </i>
490 <i>; Convert [13x sbyte]* to sbyte *...</i>
491 %cast210 = <a href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, uint 0, uint 0 <i>; sbyte*</i>
493 <i>; Call puts function to write out the string to stdout...</i>
494 <a href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
495 <a href="#i_ret">ret</a> int 0
499 This example is made up of a <a href="#globalvars">global variable</a> named
500 "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>" function, and a
501 <a href="#functionstructure">function definition</a> for "<tt>main</tt>".<p>
503 <a name="linkage_decl">
504 In general, a module is made up of a list of global values, where both functions
505 and global variables are global values. Global values are represented by a
506 pointer to a memory location (in this case, a pointer to an array of char, and a
507 pointer to a function), and can be either "internal" or externally accessible
508 (which corresponds to the static keyword in C, when used at function scope).<p>
510 For example, since the "<tt>.LC0</tt>" variable is defined to be internal, if
511 another module defined a "<tt>.LC0</tt>" variable and was linked with this one,
512 one of the two would be renamed, preventing a collision. Since "<tt>main</tt>"
513 and "<tt>puts</tt>" are external (lacking "<tt>internal</tt>" declarations),
514 they are accessible outside of the current module. It is illegal for a function
515 declaration to be "<tt>internal</tt>".<p>
518 <!-- ======================================================================= -->
519 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
520 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
521 <a name="globalvars">Global Variables
522 </b></font></td></tr></table><ul>
524 Global variables define regions of memory allocated at compilation time instead
525 of runtime. Global variables, may optionally be initialized. A variable may be
526 defined as a global "constant", which indicates that the contents of the
527 variable will never be modified (opening options for optimization). Constants
528 must always have an initial value.<p>
530 As SSA values, global variables define pointer values that are in scope in
531 (i.e. they dominate) all basic blocks in the program. Global variables always
532 define a pointer to their "content" type because they describe a region of
533 memory, and all memory objects in LLVM are accessed through pointers.<p>
537 <!-- ======================================================================= -->
538 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
539 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
540 <a name="functionstructure">Function Structure
541 </b></font></td></tr></table><ul>
543 LLVM functions definitions are composed of a (possibly empty) argument list, an
544 opening curly brace, a list of basic blocks, and a closing curly brace. LLVM
545 function declarations are defined with the "<tt>declare</tt>" keyword, a
546 function name and a function signature.<p>
548 A function definition contains a list of basic blocks, forming the CFG for the
549 function. Each basic block may optionally start with a label (giving the basic
550 block a symbol table entry), contains a list of instructions, and ends with a <a
551 href="#terminators">terminator</a> instruction (such as a branch or function
554 The first basic block in program is special in two ways: it is immediately
555 executed on entrance to the function, and it is not allowed to have predecessor
556 basic blocks (i.e. there can not be any branches to the entry block of a
560 <!-- *********************************************************************** -->
561 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
562 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
563 <a name="instref">Instruction Reference
564 </b></font></td></tr></table><ul>
565 <!-- *********************************************************************** -->
567 The LLVM instruction set consists of several different classifications of
568 instructions: <a href="#terminators">terminator instructions</a>, a <a
569 href="#unaryops">unary instruction</a>, <a href="#binaryops">binary
570 instructions</a>, <a href="#memoryops">memory instructions</a>, and <a
571 href="#otherops">other instructions</a>.<p>
574 <!-- ======================================================================= -->
575 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
576 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
577 <a name="terminators">Terminator Instructions
578 </b></font></td></tr></table><ul>
580 As mentioned <a href="#functionstructure">previously</a>, every basic block in a
581 program ends with a "Terminator" instruction, which indicates where control flow
582 should go now that this basic block has been completely executed. These
583 terminator instructions typically yield a '<tt>void</tt>' value: they produce
584 control flow, not values (the one exception being the '<a
585 href="#i_invoke"><tt>invoke</tt></a>' instruction).<p>
587 There are four different terminator instructions: the '<a
588 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a
589 href="#i_br"><tt>br</tt></a>' instruction, the '<a
590 href="#i_switch"><tt>switch</tt></a>' instruction, and the '<a
591 href="#i_invoke"><tt>invoke</tt></a>' instruction.<p>
594 <!-- _______________________________________________________________________ -->
595 </ul><a name="i_ret"><h4><hr size=0>'<tt>ret</tt>' Instruction</h4><ul>
599 ret <type> <value> <i>; Return a value from a non-void function</i>
600 ret void <i>; Return from void function</i>
605 The '<tt>ret</tt>' instruction is used to return control flow (and a value) from
606 a function, back to the caller.<p>
608 There are two forms of the '<tt>ret</tt>' instructruction: one that returns a
609 value and then causes control flow, and one that just causes control flow to
614 The '<tt>ret</tt>' instruction may return any '<a href="#t_firstclass">first
615 class</a>' type. Notice that a function is not <a href="#wellformed">well
616 formed</a> if there exists a '<tt>ret</tt>' instruction inside of the function
617 that returns a value that does not match the return type of the function.<p>
621 When the '<tt>ret</tt>' instruction is executed, control flow returns back to
622 the calling function's context. If the instruction returns a value, that value
623 shall be propogated into the calling function's data space.<p>
627 ret int 5 <i>; Return an integer value of 5</i>
628 ret void <i>; Return from a void function</i>
632 <!-- _______________________________________________________________________ -->
633 </ul><a name="i_br"><h4><hr size=0>'<tt>br</tt>' Instruction</h4><ul>
637 br bool <cond>, label <iftrue>, label <iffalse>
638 br label <dest> <i>; Unconditional branch</i>
643 The '<tt>br</tt>' instruction is used to cause control flow to transfer to a
644 different basic block in the current function. There are two forms of this
645 instruction, corresponding to a conditional branch and an unconditional
650 The conditional branch form of the '<tt>br</tt>' instruction takes a single
651 '<tt>bool</tt>' value and two '<tt>label</tt>' values. The unconditional form
652 of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>' value as a
657 Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
658 argument is evaluated. If the value is <tt>true</tt>, control flows to the
659 '<tt>iftrue</tt>' '<tt>label</tt>' argument. If "cond" is <tt>false</tt>,
660 control flows to the '<tt>iffalse</tt>' '<tt>label</tt>' argument.<p>
665 %cond = <a href="#i_setcc">seteq</a> int %a, %b
666 br bool %cond, label %IfEqual, label %IfUnequal
668 <a href="#i_ret">ret</a> int 1
670 <a href="#i_ret">ret</a> int 0
674 <!-- _______________________________________________________________________ -->
675 </ul><a name="i_switch"><h4><hr size=0>'<tt>switch</tt>' Instruction</h4><ul>
679 <i>; Definitions for lookup indirect branch</i>
680 %switchtype = type [<anysize> x { uint, label }]
682 <i>; Lookup indirect branch</i>
683 switch uint <value>, label <defaultdest>, %switchtype <switchtable>
685 <i>; Indexed indirect branch</i>
686 switch uint <idxvalue>, label <defaultdest>, [<anysize> x label] <desttable>
691 The '<tt>switch</tt>' instruction is used to transfer control flow to one of
692 several different places. It is a generalization of the '<tt>br</tt>'
693 instruction, allowing a branch to occur to one of many possible destinations.<p>
695 The '<tt>switch</tt>' statement supports two different styles of indirect
696 branching: lookup branching and indexed branching. Lookup branching is
697 generally useful if the values to switch on are spread far appart, where index
698 branching is useful if the values to switch on are generally dense.<p>
700 The two different forms of the '<tt>switch</tt>' statement are simple hints to
701 the underlying implementation. For example, the compiler may choose to
702 implement a small indirect branch table as a series of predicated comparisons:
703 if it is faster for the target architecture.<p>
707 The lookup form of the '<tt>switch</tt>' instruction uses three parameters: a
708 '<tt>uint</tt>' comparison value '<tt>value</tt>', a default '<tt>label</tt>'
709 destination, and an array of pairs of comparison value constants and
710 '<tt>label</tt>'s. The sized array must be a constant value.<p>
712 The indexed form of the '<tt>switch</tt>' instruction uses three parameters: an
713 '<tt>uint</tt>' index value, a default '<tt>label</tt>' and a sized array of
714 '<tt>label</tt>'s. The '<tt>dests</tt>' array must be a constant array.
718 The lookup style switch statement specifies a table of values and destinations.
719 When the '<tt>switch</tt>' instruction is executed, this table is searched for
720 the given value. If the value is found, the corresponding destination is
723 The index branch form simply looks up a label element directly in a table and
726 In either case, the compiler knows the static size of the array, because it is
727 provided as part of the constant values type.<p>
731 <i>; Emulate a conditional br instruction</i>
732 %Val = <a href="#i_cast">cast</a> bool %value to uint
733 switch uint %Val, label %truedest, [1 x label] [label %falsedest ]
735 <i>; Emulate an unconditional br instruction</i>
736 switch uint 0, label %dest, [ 0 x label] [ ]
738 <i>; Implement a jump table:</i>
739 switch uint %val, label %otherwise, [3 x label] [ label %onzero,
747 <!-- _______________________________________________________________________ -->
748 </ul><a name="i_invoke"><h4><hr size=0>'<tt>invoke</tt>' Instruction</h4><ul>
752 <result> = invoke <ptr to function ty> %<function ptr val>(<function args>)
753 to label <normal label> except label <exception label>
758 The '<tt>invoke</tt>' instruction is used to cause control flow to transfer to a
759 specified function, with the possibility of control flow transfer to either the
760 '<tt>normal label</tt>' label or the '<tt>exception label</tt>'. The '<tt><a
761 href="#i_call">call</a></tt>' instruction is closely related, but guarantees
762 that control flow either never returns from the called function, or that it
763 returns to the instruction succeeding the '<tt><a href="#i_call">call</a></tt>'
768 This instruction requires several arguments:<p>
771 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
772 function value being invoked. In most cases, this is a direct function
773 invocation, but indirect <tt>invoke</tt>'s are just as possible, branching off
774 an arbitrary pointer to function value.<p>
776 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
777 function to be invoked.
779 <li>'<tt>function args</tt>': argument list whose types match the function
780 signature argument types. If the function signature indicates the function
781 accepts a variable number of arguments, the extra arguments can be specified.
783 <li>'<tt>normal label</tt>': the label reached when the called function executes
784 a '<tt><a href="#i_ret">ret</a></tt>' instruction.
786 <li>'<tt>exception label</tt>': the label reached when an exception is thrown.
791 This instruction is designed to operate as a standard '<tt><a
792 href="#i_call">call</a></tt>' instruction in most regards. The primary
793 difference is that it associates a label with the function invocation that may
794 be accessed via the runtime library provided by the execution environment. This
795 instruction is used in languages with destructors to ensure that proper cleanup
796 is performed in the case of either a <tt>longjmp</tt> or a thrown exception.
797 Additionally, this is important for implementation of '<tt>catch</tt>' clauses
798 in high-level languages that support them.<p>
800 For a more comprehensive explanation of this instruction look in the llvm/docs/2001-05-18-ExceptionHandling.txt document.<p>
804 %retval = invoke int %Test(int 15)
805 to label %Continue except label %TestCleanup <i>; {int}:retval set</i>
810 <!-- ======================================================================= -->
811 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
812 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
813 <a name="unaryops">Unary Operations
814 </b></font></td></tr></table><ul>
816 Unary operators are used to do a simple operation to a single value.<p>
818 There is only one unary operator: the '<a href="#i_not"><tt>not</tt></a>' instruction.<p>
821 <!-- _______________________________________________________________________ -->
822 </ul><a name="i_not"><h4><hr size=0>'<tt>not</tt>' Instruction</h4><ul>
826 <result> = not <ty> <var> <i>; yields {ty}:result</i>
830 The '<tt>not</tt>' instruction returns the bitwise complement of its operand.<p>
833 The single argument to '<tt>not</tt>' must be of of <a href="#t_integral">integral</a> or bool type.<p>
836 <h5>Semantics:</h5> The '<tt>not</tt>' instruction returns the bitwise
837 complement (AKA ones complement) of an <a href="#t_integral">integral</a>
841 <result> = xor bool true, <var> <i>; yields {bool}:result</i>
846 %x = not int 1 <i>; {int}:x is now equal to -2</i>
847 %x = not bool true <i>; {bool}:x is now equal to false</i>
852 <!-- ======================================================================= -->
853 </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>
854 <a name="binaryops">Binary Operations
855 </b></font></td></tr></table><ul>
857 Binary operators are used to do most of the computation in a program. They
858 require two operands, execute an operation on them, and produce a single value.
859 The result value of a binary operator is not neccesarily the same type as its
862 There are several different binary operators:<p>
865 <!-- _______________________________________________________________________ -->
866 </ul><a name="i_add"><h4><hr size=0>'<tt>add</tt>' Instruction</h4><ul>
870 <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
874 The '<tt>add</tt>' instruction returns the sum of its two operands.<p>
877 The two arguments to the '<tt>add</tt>' instruction must be either <a href="#t_integral">integral</a> or <a href="#t_floating">floating point</a> values. Both arguments must have identical types.<p>
884 <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
888 <!-- _______________________________________________________________________ -->
889 </ul><a name="i_sub"><h4><hr size=0>'<tt>sub</tt>' Instruction</h4><ul>
893 <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
898 The '<tt>sub</tt>' instruction returns the difference of its two operands.<p>
900 Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
901 instruction present in most other intermediate representations.<p>
905 The two arguments to the '<tt>sub</tt>' instruction must be either <a
906 href="#t_integral">integral</a> or <a href="#t_floating">floating point</a>
907 values. Both arguments must have identical types.<p>
914 <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
915 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
918 <!-- _______________________________________________________________________ -->
919 </ul><a name="i_mul"><h4><hr size=0>'<tt>mul</tt>' Instruction</h4><ul>
923 <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
927 The '<tt>mul</tt>' instruction returns the product of its two operands.<p>
930 The two arguments to the '<tt>mul</tt>' instruction must be either <a href="#t_integral">integral</a> or <a href="#t_floating">floating point</a> values. Both arguments must have identical types.<p>
935 There is no signed vs unsigned multiplication. The appropriate action is taken
936 based on the type of the operand. <p>
941 <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
945 <!-- _______________________________________________________________________ -->
946 </ul><a name="i_div"><h4><hr size=0>'<tt>div</tt>' Instruction</h4><ul>
950 <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
955 The '<tt>div</tt>' instruction returns the quotient of its two operands.<p>
959 The two arguments to the '<tt>div</tt>' instruction must be either <a
960 href="#t_integral">integral</a> or <a href="#t_floating">floating point</a>
961 values. Both arguments must have identical types.<p>
968 <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
972 <!-- _______________________________________________________________________ -->
973 </ul><a name="i_rem"><h4><hr size=0>'<tt>rem</tt>' Instruction</h4><ul>
977 <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
981 The '<tt>rem</tt>' instruction returns the remainder from the division of its two operands.<p>
984 The two arguments to the '<tt>rem</tt>' instruction must be either <a href="#t_integral">integral</a> or <a href="#t_floating">floating point</a> values. Both arguments must have identical types.<p>
988 This returns the <i>remainder</i> of a division (where the result has the same
989 sign as the divisor), not the <i>modulus</i> (where the result has the same sign
990 as the dividend) of a value. For more information about the difference, see: <a
991 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The Math
998 <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1002 <!-- _______________________________________________________________________ -->
1003 </ul><a name="i_setcc"><h4><hr size=0>'<tt>set<i>cc</i></tt>' Instructions</h4><ul>
1007 <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1008 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1009 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1010 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1011 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1012 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1015 <h5>Overview:</h5> The '<tt>set<i>cc</i></tt>' family of instructions returns a
1016 boolean value based on a comparison of their two operands.<p>
1018 <h5>Arguments:</h5> The two arguments to the '<tt>set<i>cc</i></tt>'
1019 instructions must be of <a href="#t_firstclass">first class</a> or <a
1020 href="#t_pointer">pointer</a> type (it is not possible to compare
1021 '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>' or '<tt>void</tt>'
1022 values, etc...). Both arguments must have identical types.<p>
1024 The '<tt>setlt</tt>', '<tt>setgt</tt>', '<tt>setle</tt>', and '<tt>setge</tt>'
1025 instructions do not operate on '<tt>bool</tt>' typed arguments.<p>
1029 The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
1030 both operands are equal.<br>
1032 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
1033 both operands are unequal.<br>
1035 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
1036 the first operand is less than the second operand.<br>
1038 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
1039 the first operand is greater than the second operand.<br>
1041 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
1042 the first operand is less than or equal to the second operand.<br>
1044 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>' value if
1045 the first operand is greater than or equal to the second operand.<p>
1049 <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1050 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1051 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1052 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1053 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1054 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1059 <!-- ======================================================================= -->
1060 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1061 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
1062 <a name="bitwiseops">Bitwise Binary Operations
1063 </b></font></td></tr></table><ul>
1065 Bitwise binary operators are used to do various forms of bit-twiddling in a
1066 program. They are generally very efficient instructions, and can commonly be
1067 strength reduced from other instructions. They require two operands, execute an
1068 operation on them, and produce a single value. The resulting value of the
1069 bitwise binary operators is always the same type as its first operand.<p>
1071 <!-- _______________________________________________________________________ -->
1072 </ul><a name="i_and"><h4><hr size=0>'<tt>and</tt>' Instruction</h4><ul>
1076 <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1080 The '<tt>and</tt>' instruction returns the bitwise logical and of its two operands.<p>
1084 The two arguments to the '<tt>and</tt>' instruction must be either <a
1085 href="#t_integral">integral</a> or <tt>bool</tt> values. Both arguments must
1086 have identical types.<p>
1095 <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1096 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1097 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1102 <!-- _______________________________________________________________________ -->
1103 </ul><a name="i_or"><h4><hr size=0>'<tt>or</tt>' Instruction</h4><ul>
1107 <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1110 <h5>Overview:</h5> The '<tt>or</tt>' instruction returns the bitwise logical
1111 inclusive or of its two operands.<p>
1115 The two arguments to the '<tt>or</tt>' instruction must be either <a
1116 href="#t_integral">integral</a> or <tt>bool</tt> values. Both arguments must
1117 have identical types.<p>
1126 <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1127 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1128 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1132 <!-- _______________________________________________________________________ -->
1133 </ul><a name="i_xor"><h4><hr size=0>'<tt>xor</tt>' Instruction</h4><ul>
1137 <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1142 The '<tt>xor</tt>' instruction returns the bitwise logical exclusive or of its
1147 The two arguments to the '<tt>xor</tt>' instruction must be either <a
1148 href="#t_integral">integral</a> or <tt>bool</tt> values. Both arguments must
1149 have identical types.<p>
1158 <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1159 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1160 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1164 <!-- _______________________________________________________________________ -->
1165 </ul><a name="i_shl"><h4><hr size=0>'<tt>shl</tt>' Instruction</h4><ul>
1169 <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1174 The '<tt>shl</tt>' instruction returns the first operand shifted to the left a
1175 specified number of bits.
1179 The first argument to the '<tt>shl</tt>' instruction must be an <a
1180 href="#t_integral">integral</a> type. The second argument must be an
1181 '<tt>ubyte</tt>' type.<p>
1184 ... 0 bits are shifted into the emptied bit positions...<p>
1189 <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1190 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1191 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1195 <!-- _______________________________________________________________________ -->
1196 </ul><a name="i_shr"><h4><hr size=0>'<tt>shr</tt>' Instruction</h4><ul>
1201 <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1205 The '<tt>shr</tt>' instruction returns the first operand shifted to the right a specified number of bits.
1208 The first argument to the '<tt>shr</tt>' instruction must be an <a href="#t_integral">integral</a> type. The second argument must be an '<tt>ubyte</tt>' type.<p>
1211 ... if the first argument is a <a href="#t_signed">signed</a> type, the most significant bit is duplicated in the newly free'd bit positions. If the first argument is unsigned, zeros shall fill the empty positions...<p>
1215 <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1216 <result> = shr int 4, ubyte 1 <i>; yields {int}:result = 2</i>
1217 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1218 <result> = shr int 4, ubyte 3 <i>; yields {int}:result = 0</i>
1225 <!-- ======================================================================= -->
1226 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1227 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
1228 <a name="memoryops">Memory Access Operations
1229 </b></font></td></tr></table><ul>
1231 Accessing memory in SSA form is, well, sticky at best. This section describes how to read, write, allocate and free memory in LLVM.<p>
1234 <!-- _______________________________________________________________________ -->
1235 </ul><a name="i_malloc"><h4><hr size=0>'<tt>malloc</tt>' Instruction</h4><ul>
1239 <result> = malloc <type>, uint <NumElements> <i>; yields {type*}:result</i>
1240 <result> = malloc <type> <i>; yields {type*}:result</i>
1244 The '<tt>malloc</tt>' instruction allocates memory from the system heap and returns a pointer to it.<p>
1248 The the '<tt>malloc</tt>' instruction allocates
1249 <tt>sizeof(<type>)*NumElements</tt> bytes of memory from the operating
1250 system, and returns a pointer of the appropriate type to the program. The
1251 second form of the instruction is a shorter version of the first instruction
1252 that defaults to allocating one element.<p>
1254 '<tt>type</tt>' must be a sized type<p>
1257 Memory is allocated, a pointer is returned.<p>
1261 %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1263 %size = <a href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1264 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1265 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1269 <!-- _______________________________________________________________________ -->
1270 </ul><a name="i_free"><h4><hr size=0>'<tt>free</tt>' Instruction</h4><ul>
1274 free <type> <value> <i>; yields {void}</i>
1279 The '<tt>free</tt>' instruction returns memory back to the unused memory heap, to be reallocated in the future.<p>
1284 '<tt>value</tt>' shall be a pointer value that points to a value that was
1285 allocated with the '<tt><a href="#i_malloc">malloc</a></tt>' instruction.<p>
1290 Access to the memory pointed to by the pointer is not longer defined after this instruction executes.<p>
1294 %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1295 free [4 x ubyte]* %array
1299 <!-- _______________________________________________________________________ -->
1300 </ul><a name="i_alloca"><h4><hr size=0>'<tt>alloca</tt>' Instruction</h4><ul>
1304 <result> = alloca <type>, uint <NumElements> <i>; yields {type*}:result</i>
1305 <result> = alloca <type> <i>; yields {type*}:result</i>
1310 The '<tt>alloca</tt>' instruction allocates memory on the current stack frame of
1311 the procedure that is live until the current function returns to its caller.<p>
1315 The the '<tt>alloca</tt>' instruction allocates
1316 <tt>sizeof(<type>)*NumElements</tt> bytes of memory on the runtime stack,
1317 returning a pointer of the appropriate type to the program. The second form of
1318 the instruction is a shorter version of the first that defaults to allocating
1321 '<tt>type</tt>' may be any sized type.<p>
1325 Memory is allocated, a pointer is returned. '<tt>alloca</tt>'d memory is
1326 automatically released when the function returns. The '<tt>alloca</tt>'
1327 instruction is commonly used to represent automatic variables that must have an
1328 address available, as well as spilled variables.<p>
1332 %ptr = alloca int <i>; yields {int*}:ptr</i>
1333 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
1337 <!-- _______________________________________________________________________ -->
1338 </ul><a name="i_load"><h4><hr size=0>'<tt>load</tt>' Instruction</h4><ul>
1342 <result> = load <ty>* <pointer>
1346 The '<tt>load</tt>' instruction is used to read from memory.<p>
1350 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>
1354 The location of memory pointed to is loaded.
1358 %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1359 <a href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1360 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1366 <!-- _______________________________________________________________________ -->
1367 </ul><a name="i_store"><h4><hr size=0>'<tt>store</tt>' Instruction</h4><ul>
1371 store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1375 The '<tt>store</tt>' instruction is used to write to memory.<p>
1379 There are two arguments to the '<tt>store</tt>' instruction: a value to store
1380 and an address to store it into. The type of the '<tt><pointer></tt>'
1381 operand must be a pointer to the type of the '<tt><value></tt>'
1384 <h5>Semantics:</h5> The contents of memory are updated to contain
1385 '<tt><value></tt>' at the location specified by the
1386 '<tt><pointer></tt>' operand.<p>
1390 %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1391 <a href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1392 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1398 <!-- _______________________________________________________________________ -->
1399 </ul><a name="i_getelementptr"><h4><hr size=0>'<tt>getelementptr</tt>' Instruction</h4><ul>
1403 <result> = getelementptr <ty>* <ptrval>{, uint <aidx>|, ubyte <sidx>}*
1408 The '<tt>getelementptr</tt>' instruction is used to get the address of a
1409 subelement of an aggregate data structure.<p>
1413 This instruction takes a list of <tt>uint</tt> values and <tt>ubyte</tt>
1414 constants that indicate what form of addressing to perform. The actual types of
1415 the arguments provided depend on the type of the first pointer argument. The
1416 '<tt>getelementptr</tt>' instruction is used to index down through the type
1417 levels of a structure.<p>
1419 For example, lets consider a C code fragment and how it gets compiled to
1434 int *foo(struct ST *s) {
1435 return &s[1].Z.B[5][13];
1439 The LLVM code generated by the GCC frontend is:
1442 %RT = type { sbyte, [10 x [20 x int]], sbyte }
1443 %ST = type { int, double, %RT }
1445 int* "foo"(%ST* %s) {
1446 %reg = getelementptr %ST* %s, uint 1, ubyte 2, ubyte 1, uint 5, uint 13
1453 The index types specified for the '<tt>getelementptr</tt>' instruction depend on
1454 the pointer type that is being index into. <a href="t_pointer">Pointer</a> and
1455 <a href="t_array">array</a> types require '<tt>uint</tt>' values, and <a
1456 href="t_struct">structure</a> types require '<tt>ubyte</tt>'
1457 <b>constants</b>.<p>
1459 In the example above, the first index is indexing into the '<tt>%ST*</tt>' type,
1460 which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT }</tt>'
1461 type, a structure. The second index indexes into the third element of the
1462 structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]], sbyte
1463 }</tt>' type, another structure. The third index indexes into the second
1464 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
1465 array. The two dimensions of the array are subscripted into, yielding an
1466 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction return a pointer
1467 to this element, thus yielding a '<tt>int*</tt>' type.<p>
1469 Note that it is perfectly legal to index partially through a structure,
1470 returning a pointer to an inner element. Because of this, the LLVM code for the
1471 given testcase is equivalent to:<p>
1474 int* "foo"(%ST* %s) {
1475 %t1 = getelementptr %ST* %s , uint 1 <i>; yields %ST*:%t1</i>
1476 %t2 = getelementptr %ST* %t1, uint 0, ubyte 2 <i>; yields %RT*:%t2</i>
1477 %t3 = getelementptr %RT* %t2, uint 0, ubyte 1 <i>; yields [10 x [20 x int]]*:%t3</i>
1478 %t4 = getelementptr [10 x [20 x int]]* %t3, uint 0, uint 5 <i>; yields [20 x int]*:%t4</i>
1479 %t5 = getelementptr [20 x int]* %t4, uint 0, uint 13 <i>; yields int*:%t5</i>
1488 <i>; yields {[12 x ubyte]*}:aptr</i>
1489 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, uint 0, ubyte 1
1494 <!-- ======================================================================= -->
1495 </ul><table width="100%" bgcolor="#441188" border=0 cellpadding=4 cellspacing=0>
1496 <tr><td> </td><td width="100%"> <font color="#EEEEFF" face="Georgia,Palatino"><b>
1497 <a name="otherops">Other Operations
1498 </b></font></td></tr></table><ul>
1500 The instructions in this catagory are the "miscellaneous" functions, that defy better classification.<p>
1503 <!-- _______________________________________________________________________ -->
1504 </ul><a name="i_phi"><h4><hr size=0>'<tt>phi</tt>' Instruction</h4><ul>
1508 <result> = phi <ty> [ <val0>, <label0>], ...
1513 The '<tt>phi</tt>' instruction is used to implement the φ node in the SSA
1514 graph representing the function.<p>
1518 The type of the incoming values are specified with the first type field. After
1519 this, the '<tt>phi</tt>' instruction takes a list of pairs as arguments, with
1520 one pair for each predecessor basic block of the current block.<p>
1522 There must be no non-phi instructions between the start of a basic block and the
1523 PHI instructions: i.e. PHI instructions must be first in a basic block.<p>
1527 At runtime, the '<tt>phi</tt>' instruction logically takes on the value
1528 specified by the parameter, depending on which basic block we came from in the
1529 last <a href="#terminators">terminator</a> instruction.<p>
1534 Loop: ; Infinite loop that counts from 0 on up...
1535 %indvar = phi uint [ 0, %LoopHeader ], [ %nextindvar, %Loop ]
1536 %nextindvar = add uint %indvar, 1
1541 <!-- _______________________________________________________________________ -->
1542 </ul><a name="i_cast"><h4><hr size=0>'<tt>cast .. to</tt>' Instruction</h4><ul>
1546 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
1551 The '<tt>cast</tt>' instruction is used as the primitive means to convert
1552 integers to floating point, change data type sizes, and break type safety (by
1553 casting pointers).<p>
1557 The '<tt>cast</tt>' instruction takes a value to case, which must be a first
1558 class value, and a type to cast it to, which must also be a first class type.<p>
1562 This instruction follows the C rules for explicit casts when determining how the
1563 data being cast must change to fit in its new container.<p>
1565 When casting to bool, any value that would be considered true in the context of a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values, all else are '<tt>false</tt>'.<p>
1569 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
1570 %Y = cast int 123 to bool <i>; yields bool::true</i>
1575 <!-- _______________________________________________________________________ -->
1576 </ul><a name="i_call"><h4><hr size=0>'<tt>call</tt>' Instruction</h4><ul>
1580 <result> = call <ty>* <fnptrval>(<param list>)
1585 The '<tt>call</tt>' instruction represents a simple function call.<p>
1589 This instruction requires several arguments:<p>
1592 <li>'<tt>ty</tt>': shall be the signature of the pointer to function value being
1593 invoked. The argument types must match the types implied by this signature.<p>
1595 <li>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to be
1596 invoked. In most cases, this is a direct function invocation, but indirect
1597 <tt>call</tt>'s are just as possible, calling an arbitrary pointer to function
1600 <li>'<tt>function args</tt>': argument list whose types match the function
1601 signature argument types. If the function signature indicates the function
1602 accepts a variable number of arguments, the extra arguments can be specified.
1607 The '<tt>call</tt>' instruction is used to cause control flow to transfer to a
1608 specified function, with its incoming arguments bound to the specified values.
1609 Upon a '<tt><a href="#i_ret">ret</a></tt>' instruction in the called function,
1610 control flow continues with the instruction after the function call, and the
1611 return value of the function is bound to the result argument. This is a simpler
1612 case of the <a href="#i_invoke">invoke</a> instruction.<p>
1616 %retval = call int %test(int %argc)
1617 call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);
1623 <!x- *********************************************************************** -x>
1624 </ul><table width="100%" bgcolor="#330077" border=0 cellpadding=4 cellspacing=0>
1625 <tr><td align=center><font color="#EEEEFF" size=+2 face="Georgia,Palatino"><b>
1626 <a name="related">Related Work
1627 </b></font></td></tr></table><ul>
1628 <!x- *********************************************************************** -x>
1631 Codesigned virtual machines.<p>
1634 <a name="rw_safetsa">
1636 <DD>Description here<p>
1639 <dt><a href="http://www.javasoft.com">Java</a>
1640 <DD>Desciption here<p>
1643 <dt><a href="http://www.microsoft.com/net">Microsoft .net</a>
1644 <DD>Desciption here<p>
1646 <a name="rw_gccrtl">
1647 <dt><a href="http://www.math.umn.edu/systems_guide/gcc-2.95.1/gcc_15.html">GNU RTL Intermediate Representation</a>
1648 <DD>Desciption here<p>
1651 <dt><a href="http://developer.intel.com/design/ia-64/index.htm">IA64 Architecture & Instruction Set</a>
1652 <DD>Desciption here<p>
1655 <dt><a href="http://www-cs-faculty.stanford.edu/~knuth/mmix-news.html">MMIX Instruction Set</a>
1656 <DD>Desciption here<p>
1658 <a name="rw_stroustrup">
1659 <dt><a href="http://www.research.att.com/~bs/devXinterview.html">"Interview With Bjarne Stroustrup"</a>
1660 <DD>This interview influenced the design and thought process behind LLVM in several ways, most notably the way that derived types are written in text format. See the question that starts with "you defined the C declarator syntax as an experiment that failed".<p>
1663 <!x- _______________________________________________________________________ -x>
1664 </ul><a name="rw_vectorization"><h3><hr size=0>Vectorized Architectures</h3><ul>
1667 <a name="rw_intel_simd">
1668 <dt>Intel MMX, MMX2, SSE, SSE2
1669 <DD>Description here<p>
1671 <a name="rw_amd_simd">
1672 <dt><a href="http://www.nondot.org/~sabre/os/H1ChipFeatures/3DNow!TechnologyManual.pdf">AMD 3Dnow!, 3Dnow! 2</a>
1673 <DD>Desciption here<p>
1675 <a name="rw_sun_simd">
1676 <dt><a href="http://www.nondot.org/~sabre/os/H1ChipFeatures/VISInstructionSetUsersManual.pdf">Sun VIS ISA</a>
1677 <DD>Desciption here<p>
1679 <a name="rw_powerpc_simd">
1681 <DD>Desciption here<p>
1690 <!-- *********************************************************************** -->
1692 <!-- *********************************************************************** -->
1697 <address><a href="mailto:sabre@nondot.org">Chris Lattner</a></address>
1698 <!-- Created: Tue Jan 23 15:19:28 CST 2001 -->
1699 <!-- hhmts start -->
1700 Last modified: Mon May 6 17:07:42 CDT 2002