1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>LLVM Assembly Language Reference Manual</title>
6 <link rel="stylesheet" href="llvm.css" type="text/css">
9 <div class="doc_title"> LLVM Language Reference Manual </div>
11 <li><a href="#abstract">Abstract</a></li>
12 <li><a href="#introduction">Introduction</a></li>
13 <li><a href="#identifiers">Identifiers</a></li>
14 <li><a href="#typesystem">Type System</a>
16 <li><a href="#t_primitive">Primitive Types</a>
18 <li><a href="#t_classifications">Type Classifications</a></li>
21 <li><a href="#t_derived">Derived Types</a>
23 <li><a href="#t_array">Array Type</a></li>
24 <li><a href="#t_function">Function Type</a></li>
25 <li><a href="#t_pointer">Pointer Type</a></li>
26 <li><a href="#t_struct">Structure Type</a></li>
27 <!-- <li><a href="#t_packed" >Packed Type</a> -->
32 <li><a href="#highlevel">High Level Structure</a>
34 <li><a href="#modulestructure">Module Structure</a></li>
35 <li><a href="#globalvars">Global Variables</a></li>
36 <li><a href="#functionstructure">Function Structure</a></li>
39 <li><a href="#instref">Instruction Reference</a>
41 <li><a href="#terminators">Terminator Instructions</a>
43 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
44 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
45 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
46 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
47 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
50 <li><a href="#binaryops">Binary Operations</a>
52 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
53 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
54 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
55 <li><a href="#i_div">'<tt>div</tt>' Instruction</a></li>
56 <li><a href="#i_rem">'<tt>rem</tt>' Instruction</a></li>
57 <li><a href="#i_setcc">'<tt>set<i>cc</i></tt>' Instructions</a></li>
60 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
62 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
63 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
64 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
65 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
66 <li><a href="#i_shr">'<tt>shr</tt>' Instruction</a></li>
69 <li><a href="#memoryops">Memory Access Operations</a>
71 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
72 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
73 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
74 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
75 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
76 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
79 <li><a href="#otherops">Other Operations</a>
81 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
82 <li><a href="#i_cast">'<tt>cast .. to</tt>' Instruction</a></li>
83 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
84 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
85 <li><a href="#i_vanext">'<tt>vanext</tt>' Instruction</a></li>
86 <li><a href="#i_vaarg">'<tt>vaarg</tt>' Instruction</a></li>
91 <li><a href="#intrinsics">Intrinsic Functions</a>
93 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
95 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
96 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
97 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
100 <li><a href="#int_codegen">Code Generator Intrinsics</a>
102 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
103 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
106 <li><a href="#int_libc">Standard C Library Intrinsics</a>
108 <li><a href="#i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a></li>
109 <li><a href="#i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a></li>
110 <li><a href="#i_memset">'<tt>llvm.memset</tt>' Intrinsic</a></li>
113 <li><a href="#int_debugger">Debugger intrinsics</a>
117 <div class="doc_text">
118 <p><b>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
119 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></b></p>
122 <!-- *********************************************************************** -->
123 <div class="doc_section"> <a name="abstract">Abstract </a></div>
124 <!-- *********************************************************************** -->
125 <div class="doc_text">
126 <p>This document is a reference manual for the LLVM assembly language.
127 LLVM is an SSA based representation that provides type safety,
128 low-level operations, flexibility, and the capability of representing
129 'all' high-level languages cleanly. It is the common code
130 representation used throughout all phases of the LLVM compilation
133 <!-- *********************************************************************** -->
134 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
135 <!-- *********************************************************************** -->
136 <div class="doc_text">
137 <p>The LLVM code representation is designed to be used in three
138 different forms: as an in-memory compiler IR, as an on-disk bytecode
139 representation (suitable for fast loading by a Just-In-Time compiler),
140 and as a human readable assembly language representation. This allows
141 LLVM to provide a powerful intermediate representation for efficient
142 compiler transformations and analysis, while providing a natural means
143 to debug and visualize the transformations. The three different forms
144 of LLVM are all equivalent. This document describes the human readable
145 representation and notation.</p>
146 <p>The LLVM representation aims to be a light-weight and low-level
147 while being expressive, typed, and extensible at the same time. It
148 aims to be a "universal IR" of sorts, by being at a low enough level
149 that high-level ideas may be cleanly mapped to it (similar to how
150 microprocessors are "universal IR's", allowing many source languages to
151 be mapped to them). By providing type information, LLVM can be used as
152 the target of optimizations: for example, through pointer analysis, it
153 can be proven that a C automatic variable is never accessed outside of
154 the current function... allowing it to be promoted to a simple SSA
155 value instead of a memory location.</p>
157 <!-- _______________________________________________________________________ -->
158 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
159 <div class="doc_text">
160 <p>It is important to note that this document describes 'well formed'
161 LLVM assembly language. There is a difference between what the parser
162 accepts and what is considered 'well formed'. For example, the
163 following instruction is syntactically okay, but not well formed:</p>
164 <pre> %x = <a href="#i_add">add</a> int 1, %x<br></pre>
165 <p>...because the definition of <tt>%x</tt> does not dominate all of
166 its uses. The LLVM infrastructure provides a verification pass that may
167 be used to verify that an LLVM module is well formed. This pass is
168 automatically run by the parser after parsing input assembly, and by
169 the optimizer before it outputs bytecode. The violations pointed out
170 by the verifier pass indicate bugs in transformation passes or input to
172 <!-- Describe the typesetting conventions here. --> </div>
173 <!-- *********************************************************************** -->
174 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
175 <!-- *********************************************************************** -->
176 <div class="doc_text">
177 <p>LLVM uses three different forms of identifiers, for different
180 <li>Numeric constants are represented as you would expect: 12, -3
181 123.421, etc. Floating point constants have an optional hexidecimal
183 <li>Named values are represented as a string of characters with a '%'
184 prefix. For example, %foo, %DivisionByZero,
185 %a.really.long.identifier. The actual regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
186 Identifiers which require other characters in their names can be
187 surrounded with quotes. In this way, anything except a <tt>"</tt>
188 character can be used in a name.</li>
189 <li>Unnamed values are represented as an unsigned numeric value with
190 a '%' prefix. For example, %12, %2, %44.</li>
192 <p>LLVM requires that values start with a '%' sign for two reasons:
193 Compilers don't need to worry about name clashes with reserved words,
194 and the set of reserved words may be expanded in the future without
195 penalty. Additionally, unnamed identifiers allow a compiler to quickly
196 come up with a temporary variable without having to avoid symbol table
198 <p>Reserved words in LLVM are very similar to reserved words in other
199 languages. There are keywords for different opcodes ('<tt><a
200 href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
201 href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
202 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>',
203 etc...), and others. These reserved words cannot conflict with
204 variable names, because none of them start with a '%' character.</p>
205 <p>Here is an example of LLVM code to multiply the integer variable '<tt>%X</tt>'
208 <pre> %result = <a href="#i_mul">mul</a> uint %X, 8<br></pre>
209 <p>After strength reduction:</p>
210 <pre> %result = <a href="#i_shl">shl</a> uint %X, ubyte 3<br></pre>
211 <p>And the hard way:</p>
212 <pre> <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
214 href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
216 href="#i_add">add</a> uint %1, %1<br></pre>
217 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
218 important lexical features of LLVM:</p>
220 <li>Comments are delimited with a '<tt>;</tt>' and go until the end
222 <li>Unnamed temporaries are created when the result of a computation
223 is not assigned to a named value.</li>
224 <li>Unnamed temporaries are numbered sequentially</li>
226 <p>...and it also show a convention that we follow in this document.
227 When demonstrating instructions, we will follow an instruction with a
228 comment that defines the type and name of value produced. Comments are
229 shown in italic text.</p>
230 <p>The one non-intuitive notation for constants is the optional
231 hexidecimal form of floating point constants. For example, the form '<tt>double
232 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
233 4.5e+15</tt>' which is also supported by the parser. The only time
234 hexadecimal floating point constants are useful (and the only time that
235 they are generated by the disassembler) is when an FP constant has to
236 be emitted that is not representable as a decimal floating point number
237 exactly. For example, NaN's, infinities, and other special cases are
238 represented in their IEEE hexadecimal format so that assembly and
239 disassembly do not cause any bits to change in the constants.</p>
241 <!-- *********************************************************************** -->
242 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
243 <!-- *********************************************************************** -->
244 <div class="doc_text">
245 <p>The LLVM type system is one of the most important features of the
246 intermediate representation. Being typed enables a number of
247 optimizations to be performed on the IR directly, without having to do
248 extra analyses on the side before the transformation. A strong type
249 system makes it easier to read the generated code and enables novel
250 analyses and transformations that are not feasible to perform on normal
251 three address code representations.</p>
252 <!-- The written form for the type system was heavily influenced by the
253 syntactic problems with types in the C language<sup><a
254 href="#rw_stroustrup">1</a></sup>.<p> --> </div>
255 <!-- ======================================================================= -->
256 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
257 <div class="doc_text">
258 <p>The primitive types are the fundemental building blocks of the LLVM
259 system. The current set of primitive types are as follows:</p>
261 <table border="0" style="align: center">
265 <table border="1" cellspacing="0" cellpadding="4" style="align: center">
268 <td><tt>void</tt></td>
272 <td><tt>ubyte</tt></td>
273 <td>Unsigned 8 bit value</td>
276 <td><tt>ushort</tt></td>
277 <td>Unsigned 16 bit value</td>
280 <td><tt>uint</tt></td>
281 <td>Unsigned 32 bit value</td>
284 <td><tt>ulong</tt></td>
285 <td>Unsigned 64 bit value</td>
288 <td><tt>float</tt></td>
289 <td>32 bit floating point value</td>
292 <td><tt>label</tt></td>
293 <td>Branch destination</td>
299 <table border="1" cellspacing="0" cellpadding="4">
302 <td><tt>bool</tt></td>
303 <td>True or False value</td>
306 <td><tt>sbyte</tt></td>
307 <td>Signed 8 bit value</td>
310 <td><tt>short</tt></td>
311 <td>Signed 16 bit value</td>
314 <td><tt>int</tt></td>
315 <td>Signed 32 bit value</td>
318 <td><tt>long</tt></td>
319 <td>Signed 64 bit value</td>
322 <td><tt>double</tt></td>
323 <td>64 bit floating point value</td>
333 <!-- _______________________________________________________________________ -->
334 <div class="doc_subsubsection"> <a name="t_classifications">Type
335 Classifications</a> </div>
336 <div class="doc_text">
337 <p>These different primitive types fall into a few useful
340 <table border="1" cellspacing="0" cellpadding="4">
343 <td><a name="t_signed">signed</a></td>
344 <td><tt>sbyte, short, int, long, float, double</tt></td>
347 <td><a name="t_unsigned">unsigned</a></td>
348 <td><tt>ubyte, ushort, uint, ulong</tt></td>
351 <td><a name="t_integer">integer</a></td>
352 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
355 <td><a name="t_integral">integral</a></td>
356 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
359 <td><a name="t_floating">floating point</a></td>
360 <td><tt>float, double</tt></td>
363 <td><a name="t_firstclass">first class</a></td>
364 <td><tt>bool, ubyte, sbyte, ushort, short,<br>
365 uint, int, ulong, long, float, double, <a href="#t_pointer">pointer</a></tt></td>
370 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
371 most important. Values of these types are the only ones which can be
372 produced by instructions, passed as arguments, or used as operands to
373 instructions. This means that all structures and arrays must be
374 manipulated either by pointer or by component.</p>
376 <!-- ======================================================================= -->
377 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
378 <div class="doc_text">
379 <p>The real power in LLVM comes from the derived types in the system.
380 This is what allows a programmer to represent arrays, functions,
381 pointers, and other useful types. Note that these derived types may be
382 recursive: For example, it is possible to have a two dimensional array.</p>
384 <!-- _______________________________________________________________________ -->
385 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
386 <div class="doc_text">
388 <p>The array type is a very simple derived type that arranges elements
389 sequentially in memory. The array type requires a size (number of
390 elements) and an underlying data type.</p>
392 <pre> [<# elements> x <elementtype>]<br></pre>
393 <p>The number of elements is a constant integer value, elementtype may
394 be any type with a size.</p>
396 <p> <tt>[40 x int ]</tt>: Array of 40 integer values.<br>
397 <tt>[41 x int ]</tt>: Array of 41 integer values.<br>
398 <tt>[40 x uint]</tt>: Array of 40 unsigned integer values.</p>
400 <p>Here are some examples of multidimensional arrays:</p>
402 <table border="0" cellpadding="0" cellspacing="0">
405 <td><tt>[3 x [4 x int]]</tt></td>
406 <td>: 3x4 array integer values.</td>
409 <td><tt>[12 x [10 x float]]</tt></td>
410 <td>: 12x10 array of single precision floating point values.</td>
413 <td><tt>[2 x [3 x [4 x uint]]]</tt></td>
414 <td>: 2x3x4 array of unsigned integer values.</td>
420 <!-- _______________________________________________________________________ -->
421 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
422 <div class="doc_text">
424 <p>The function type can be thought of as a function signature. It
425 consists of a return type and a list of formal parameter types.
426 Function types are usually used to build virtual function tables
427 (which are structures of pointers to functions), for indirect function
428 calls, and when defining a function.</p>
430 The return type of a function type cannot be an aggregate type.
433 <pre> <returntype> (<parameter list>)<br></pre>
434 <p>Where '<tt><parameter list></tt>' is a comma-separated list of
435 type specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
436 which indicates that the function takes a variable number of arguments.
437 Variable argument functions can access their arguments with the <a
438 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
441 <table border="0" cellpadding="0" cellspacing="0">
444 <td><tt>int (int)</tt></td>
445 <td>: function taking an <tt>int</tt>, returning an <tt>int</tt></td>
448 <td><tt>float (int, int *) *</tt></td>
449 <td>: <a href="#t_pointer">Pointer</a> to a function that takes
450 an <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
451 returning <tt>float</tt>.</td>
454 <td><tt>int (sbyte *, ...)</tt></td>
455 <td>: A vararg function that takes at least one <a
456 href="#t_pointer">pointer</a> to <tt>sbyte</tt> (signed char in C),
457 which returns an integer. This is the signature for <tt>printf</tt>
464 <!-- _______________________________________________________________________ -->
465 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
466 <div class="doc_text">
468 <p>The structure type is used to represent a collection of data members
469 together in memory. The packing of the field types is defined to match
470 the ABI of the underlying processor. The elements of a structure may
471 be any type that has a size.</p>
472 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
473 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
474 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
477 <pre> { <type list> }<br></pre>
480 <table border="0" cellpadding="0" cellspacing="0">
483 <td><tt>{ int, int, int }</tt></td>
484 <td>: a triple of three <tt>int</tt> values</td>
487 <td><tt>{ float, int (int) * }</tt></td>
488 <td>: A pair, where the first element is a <tt>float</tt> and the
489 second element is a <a href="#t_pointer">pointer</a> to a <a
490 href="t_function">function</a> that takes an <tt>int</tt>, returning
491 an <tt>int</tt>.</td>
497 <!-- _______________________________________________________________________ -->
498 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
499 <div class="doc_text">
501 <p>As in many languages, the pointer type represents a pointer or
502 reference to another object, which must live in memory.</p>
504 <pre> <type> *<br></pre>
507 <table border="0" cellpadding="0" cellspacing="0">
510 <td><tt>[4x int]*</tt></td>
511 <td>: <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a>
512 of four <tt>int</tt> values</td>
515 <td><tt>int (int *) *</tt></td>
516 <td>: A <a href="#t_pointer">pointer</a> to a <a
517 href="t_function">function</a> that takes an <tt>int</tt>, returning
518 an <tt>int</tt>.</td>
524 <!-- _______________________________________________________________________ --><!--
525 <div class="doc_subsubsection">
526 <a name="t_packed">Packed Type</a>
529 <div class="doc_text">
531 Mention/decide that packed types work with saturation or not. Maybe have a packed+saturated type in addition to just a packed type.<p>
533 Packed types should be 'nonsaturated' because standard data types are not saturated. Maybe have a saturated packed type?<p>
537 --><!-- *********************************************************************** -->
538 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
539 <!-- *********************************************************************** --><!-- ======================================================================= -->
540 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a> </div>
541 <div class="doc_text">
542 <p>LLVM programs are composed of "Module"s, each of which is a
543 translation unit of the input programs. Each module consists of
544 functions, global variables, and symbol table entries. Modules may be
545 combined together with the LLVM linker, which merges function (and
546 global variable) definitions, resolves forward declarations, and merges
547 symbol table entries. Here is an example of the "hello world" module:</p>
548 <pre><i>; Declare the string constant as a global constant...</i>
549 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
550 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
552 <i>; External declaration of the puts function</i>
553 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
555 <i>; Definition of main function</i>
556 int %main() { <i>; int()* </i>
557 <i>; Convert [13x sbyte]* to sbyte *...</i>
559 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
561 <i>; Call puts function to write out the string to stdout...</i>
563 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
565 href="#i_ret">ret</a> int 0<br>}<br></pre>
566 <p>This example is made up of a <a href="#globalvars">global variable</a>
567 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
568 function, and a <a href="#functionstructure">function definition</a>
569 for "<tt>main</tt>".</p>
570 <a name="linkage"> In general, a module is made up of a list of global
571 values, where both functions and global variables are global values.
572 Global values are represented by a pointer to a memory location (in
573 this case, a pointer to an array of char, and a pointer to a function),
574 and have one of the following linkage types:</a>
577 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
578 <dd>Global values with internal linkage are only directly accessible
579 by objects in the current module. In particular, linking code into a
580 module with an internal global value may cause the internal to be
581 renamed as necessary to avoid collisions. Because the symbol is
582 internal to the module, all references can be updated. This
583 corresponds to the notion of the '<tt>static</tt>' keyword in C, or the
584 idea of "anonymous namespaces" in C++.
587 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
588 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt>
589 linkage, with the twist that linking together two modules defining the
590 same <tt>linkonce</tt> globals will cause one of the globals to be
591 discarded. This is typically used to implement inline functions.
592 Unreferenced <tt>linkonce</tt> globals are allowed to be discarded.
595 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
596 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt>
597 linkage, except that unreferenced <tt>weak</tt> globals may not be
598 discarded. This is used to implement constructs in C such as "<tt>int
599 X;</tt>" at global scope.
602 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
603 <dd>"<tt>appending</tt>" linkage may only be applied to global
604 variables of pointer to array type. When two global variables with
605 appending linkage are linked together, the two global arrays are
606 appended together. This is the LLVM, typesafe, equivalent of having
607 the system linker append together "sections" with identical names when
611 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
612 <dd>If none of the above identifiers are used, the global is
613 externally visible, meaning that it participates in linkage and can be
614 used to resolve external symbol references.
619 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
620 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
621 variable and was linked with this one, one of the two would be renamed,
622 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
623 external (i.e., lacking any linkage declarations), they are accessible
624 outside of the current module. It is illegal for a function <i>declaration</i>
625 to have any linkage type other than "externally visible".</a></p>
628 <!-- ======================================================================= -->
629 <div class="doc_subsection">
630 <a name="globalvars">Global Variables</a>
633 <div class="doc_text">
635 <p>Global variables define regions of memory allocated at compilation
636 time instead of run-time. Global variables may optionally be
637 initialized. A variable may be defined as a global "constant", which
638 indicates that the contents of the variable will never be modified
639 (opening options for optimization).</p>
641 <p>As SSA values, global variables define pointer values that are in
642 scope (i.e. they dominate) for all basic blocks in the program. Global
643 variables always define a pointer to their "content" type because they
644 describe a region of memory, and all memory objects in LLVM are
645 accessed through pointers.</p>
650 <!-- ======================================================================= -->
651 <div class="doc_subsection">
652 <a name="functionstructure">Functions</a>
655 <div class="doc_text">
657 <p>LLVM function definitions are composed of a (possibly empty) argument list,
658 an opening curly brace, a list of basic blocks, and a closing curly brace. LLVM
659 function declarations are defined with the "<tt>declare</tt>" keyword, a
660 function name, and a function signature.</p>
662 <p>A function definition contains a list of basic blocks, forming the CFG for
663 the function. Each basic block may optionally start with a label (giving the
664 basic block a symbol table entry), contains a list of instructions, and ends
665 with a <a href="#terminators">terminator</a> instruction (such as a branch or
666 function return).</p>
668 <p>The first basic block in program is special in two ways: it is immediately
669 executed on entrance to the function, and it is not allowed to have predecessor
670 basic blocks (i.e. there can not be any branches to the entry block of a
671 function). Because the block can have no predecessors, it also cannot have any
672 <a href="#i_phi">PHI nodes</a>.</p>
674 <p>LLVM functions are identified by their name and type signature. Hence, two
675 functions with the same name but different parameter lists or return values are
676 considered different functions, and LLVM will resolves references to each
682 <!-- *********************************************************************** -->
683 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
684 <!-- *********************************************************************** -->
685 <div class="doc_text">
686 <p>The LLVM instruction set consists of several different
687 classifications of instructions: <a href="#terminators">terminator
688 instructions</a>, <a href="#binaryops">binary instructions</a>, <a
689 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
690 instructions</a>.</p>
692 <!-- ======================================================================= -->
693 <div class="doc_subsection"> <a name="terminators">Terminator
694 Instructions</a> </div>
695 <div class="doc_text">
696 <p>As mentioned <a href="#functionstructure">previously</a>, every
697 basic block in a program ends with a "Terminator" instruction, which
698 indicates which block should be executed after the current block is
699 finished. These terminator instructions typically yield a '<tt>void</tt>'
700 value: they produce control flow, not values (the one exception being
701 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
702 <p>There are five different terminator instructions: the '<a
703 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
704 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
705 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, and the '<a
706 href="#i_unwind"><tt>unwind</tt></a>' instruction.</p>
708 <!-- _______________________________________________________________________ -->
709 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
710 Instruction</a> </div>
711 <div class="doc_text">
713 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
714 ret void <i>; Return from void function</i>
717 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
718 value) from a function, back to the caller.</p>
719 <p>There are two forms of the '<tt>ret</tt>' instructruction: one that
720 returns a value and then causes control flow, and one that just causes
721 control flow to occur.</p>
723 <p>The '<tt>ret</tt>' instruction may return any '<a
724 href="#t_firstclass">first class</a>' type. Notice that a function is
725 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
726 instruction inside of the function that returns a value that does not
727 match the return type of the function.</p>
729 <p>When the '<tt>ret</tt>' instruction is executed, control flow
730 returns back to the calling function's context. If the caller is a "<a
731 href="#i_call"><tt>call</tt></a> instruction, execution continues at
732 the instruction after the call. If the caller was an "<a
733 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
734 at the beginning "normal" of the destination block. If the instruction
735 returns a value, that value shall set the call or invoke instruction's
738 <pre> ret int 5 <i>; Return an integer value of 5</i>
739 ret void <i>; Return from a void function</i>
742 <!-- _______________________________________________________________________ -->
743 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
744 <div class="doc_text">
746 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
749 <p>The '<tt>br</tt>' instruction is used to cause control flow to
750 transfer to a different basic block in the current function. There are
751 two forms of this instruction, corresponding to a conditional branch
752 and an unconditional branch.</p>
754 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
755 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
756 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
757 value as a target.</p>
759 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
760 argument is evaluated. If the value is <tt>true</tt>, control flows
761 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
762 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
764 <pre>Test:<br> %cond = <a href="#i_setcc">seteq</a> int %a, %b<br> br bool %cond, label %IfEqual, label %IfUnequal<br>IfEqual:<br> <a
765 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
767 <!-- _______________________________________________________________________ -->
768 <div class="doc_subsubsection">
769 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
772 <div class="doc_text">
776 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
781 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
782 several different places. It is a generalization of the '<tt>br</tt>'
783 instruction, allowing a branch to occur to one of many possible
789 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
790 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
791 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
792 table is not allowed to contain duplicate constant entries.</p>
796 <p>The <tt>switch</tt> instruction specifies a table of values and
797 destinations. When the '<tt>switch</tt>' instruction is executed, this
798 table is searched for the given value. If the value is found, the
799 corresponding destination is branched to, otherwise the default value
800 it transfered to.</p>
802 <h5>Implementation:</h5>
804 <p>Depending on properties of the target machine and the particular
805 <tt>switch</tt> instruction, this instruction may be code generated in different
806 ways, for example as a series of chained conditional branches, or with a lookup
812 <i>; Emulate a conditional br instruction</i>
813 %Val = <a href="#i_cast">cast</a> bool %value to int
814 switch int %Val, label %truedest [int 0, label %falsedest ]
816 <i>; Emulate an unconditional br instruction</i>
817 switch uint 0, label %dest [ ]
819 <i>; Implement a jump table:</i>
820 switch uint %val, label %otherwise [ uint 0, label %onzero
822 uint 2, label %ontwo ]
825 <!-- _______________________________________________________________________ -->
826 <div class="doc_subsubsection"> <a name="i_invoke">'<tt>invoke</tt>'
827 Instruction</a> </div>
828 <div class="doc_text">
830 <pre> <result> = invoke <ptr to function ty> %<function ptr val>(<function args>)<br> to label <normal label> except label <exception label><br></pre>
832 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a
833 specified function, with the possibility of control flow transfer to
834 either the '<tt>normal</tt>' <tt>label</tt> label or the '<tt>exception</tt>'<tt>label</tt>.
835 If the callee function returns with the "<tt><a href="#i_ret">ret</a></tt>"
836 instruction, control flow will return to the "normal" label. If the
837 callee (or any indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
838 instruction, control is interrupted, and continued at the dynamically
839 nearest "except" label.</p>
841 <p>This instruction requires several arguments:</p>
843 <li>'<tt>ptr to function ty</tt>': shall be the signature of the
844 pointer to function value being invoked. In most cases, this is a
845 direct function invocation, but indirect <tt>invoke</tt>s are just as
846 possible, branching off an arbitrary pointer to function value. </li>
847 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer
848 to a function to be invoked. </li>
849 <li>'<tt>function args</tt>': argument list whose types match the
850 function signature argument types. If the function signature indicates
851 the function accepts a variable number of arguments, the extra
852 arguments can be specified. </li>
853 <li>'<tt>normal label</tt>': the label reached when the called
854 function executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
855 <li>'<tt>exception label</tt>': the label reached when a callee
856 returns with the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
859 <p>This instruction is designed to operate as a standard '<tt><a
860 href="#i_call">call</a></tt>' instruction in most regards. The
861 primary difference is that it establishes an association with a label,
862 which is used by the runtime library to unwind the stack.</p>
863 <p>This instruction is used in languages with destructors to ensure
864 that proper cleanup is performed in the case of either a <tt>longjmp</tt>
865 or a thrown exception. Additionally, this is important for
866 implementation of '<tt>catch</tt>' clauses in high-level languages that
869 <pre> %retval = invoke int %Test(int 15)<br> to label %Continue<br> except label %TestCleanup <i>; {int}:retval set</i>
872 <!-- _______________________________________________________________________ -->
873 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
874 Instruction</a> </div>
875 <div class="doc_text">
877 <pre> unwind<br></pre>
879 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing
880 control flow at the first callee in the dynamic call stack which used
881 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the
882 call. This is primarily used to implement exception handling.</p>
884 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current
885 function to immediately halt. The dynamic call stack is then searched
886 for the first <a href="#i_invoke"><tt>invoke</tt></a> instruction on
887 the call stack. Once found, execution continues at the "exceptional"
888 destination block specified by the <tt>invoke</tt> instruction. If
889 there is no <tt>invoke</tt> instruction in the dynamic call chain,
890 undefined behavior results.</p>
892 <!-- ======================================================================= -->
893 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
894 <div class="doc_text">
895 <p>Binary operators are used to do most of the computation in a
896 program. They require two operands, execute an operation on them, and
897 produce a single value. The result value of a binary operator is not
898 necessarily the same type as its operands.</p>
899 <p>There are several different binary operators:</p>
901 <!-- _______________________________________________________________________ -->
902 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
903 Instruction</a> </div>
904 <div class="doc_text">
906 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
909 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
911 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
912 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
913 values. Both arguments must have identical types.</p>
915 <p>The value produced is the integer or floating point sum of the two
918 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
921 <!-- _______________________________________________________________________ -->
922 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
923 Instruction</a> </div>
924 <div class="doc_text">
926 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
929 <p>The '<tt>sub</tt>' instruction returns the difference of its two
931 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
932 instruction present in most other intermediate representations.</p>
934 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
935 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
936 values. Both arguments must have identical types.</p>
938 <p>The value produced is the integer or floating point difference of
939 the two operands.</p>
941 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
942 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
945 <!-- _______________________________________________________________________ -->
946 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
947 Instruction</a> </div>
948 <div class="doc_text">
950 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
953 <p>The '<tt>mul</tt>' instruction returns the product of its two
956 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
957 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
958 values. Both arguments must have identical types.</p>
960 <p>The value produced is the integer or floating point product of the
962 <p>There is no signed vs unsigned multiplication. The appropriate
963 action is taken based on the type of the operand.</p>
965 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
968 <!-- _______________________________________________________________________ -->
969 <div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
970 Instruction</a> </div>
971 <div class="doc_text">
973 <pre> <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
976 <p>The '<tt>div</tt>' instruction returns the quotient of its two
979 <p>The two arguments to the '<tt>div</tt>' instruction must be either <a
980 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
981 values. Both arguments must have identical types.</p>
983 <p>The value produced is the integer or floating point quotient of the
986 <pre> <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
989 <!-- _______________________________________________________________________ -->
990 <div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
991 Instruction</a> </div>
992 <div class="doc_text">
994 <pre> <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
997 <p>The '<tt>rem</tt>' instruction returns the remainder from the
998 division of its two operands.</p>
1000 <p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1001 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1002 values. Both arguments must have identical types.</p>
1004 <p>This returns the <i>remainder</i> of a division (where the result
1005 has the same sign as the divisor), not the <i>modulus</i> (where the
1006 result has the same sign as the dividend) of a value. For more
1007 information about the difference, see: <a
1008 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1011 <pre> <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1014 <!-- _______________________________________________________________________ -->
1015 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1016 Instructions</a> </div>
1017 <div class="doc_text">
1019 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1020 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1021 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1022 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1023 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1024 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1027 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1028 value based on a comparison of their two operands.</p>
1030 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1031 be of <a href="#t_firstclass">first class</a> type (it is not possible
1032 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1033 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1036 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1037 value if both operands are equal.<br>
1038 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1039 value if both operands are unequal.<br>
1040 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1041 value if the first operand is less than the second operand.<br>
1042 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1043 value if the first operand is greater than the second operand.<br>
1044 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1045 value if the first operand is less than or equal to the second operand.<br>
1046 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1047 value if the first operand is greater than or equal to the second
1050 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1051 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1052 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1053 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1054 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1055 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1058 <!-- ======================================================================= -->
1059 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1060 Operations</a> </div>
1061 <div class="doc_text">
1062 <p>Bitwise binary operators are used to do various forms of
1063 bit-twiddling in a program. They are generally very efficient
1064 instructions, and can commonly be strength reduced from other
1065 instructions. They require two operands, execute an operation on them,
1066 and produce a single value. The resulting value of the bitwise binary
1067 operators is always the same type as its first operand.</p>
1069 <!-- _______________________________________________________________________ -->
1070 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1071 Instruction</a> </div>
1072 <div class="doc_text">
1074 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1077 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1078 its two operands.</p>
1080 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1081 href="#t_integral">integral</a> values. Both arguments must have
1082 identical types.</p>
1084 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1086 <div style="align: center">
1087 <table border="1" cellspacing="0" cellpadding="4">
1118 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1119 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1120 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1123 <!-- _______________________________________________________________________ -->
1124 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1125 <div class="doc_text">
1127 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1130 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1131 or of its two operands.</p>
1133 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1134 href="#t_integral">integral</a> values. Both arguments must have
1135 identical types.</p>
1137 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1139 <div style="align: center">
1140 <table border="1" cellspacing="0" cellpadding="4">
1171 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1172 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1173 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1176 <!-- _______________________________________________________________________ -->
1177 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1178 Instruction</a> </div>
1179 <div class="doc_text">
1181 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1184 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1185 or of its two operands. The <tt>xor</tt> is used to implement the
1186 "one's complement" operation, which is the "~" operator in C.</p>
1188 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1189 href="#t_integral">integral</a> values. Both arguments must have
1190 identical types.</p>
1192 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1194 <div style="align: center">
1195 <table border="1" cellspacing="0" cellpadding="4">
1227 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1228 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1229 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1230 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1233 <!-- _______________________________________________________________________ -->
1234 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1235 Instruction</a> </div>
1236 <div class="doc_text">
1238 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1241 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1242 the left a specified number of bits.</p>
1244 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
1245 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1248 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
1250 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1251 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1252 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1255 <!-- _______________________________________________________________________ -->
1256 <div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1257 Instruction</a> </div>
1258 <div class="doc_text">
1260 <pre> <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1263 <p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1264 the right a specified number of bits.</p>
1266 <p>The first argument to the '<tt>shr</tt>' instruction must be an <a
1267 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1270 <p>If the first argument is a <a href="#t_signed">signed</a> type, the
1271 most significant bit is duplicated in the newly free'd bit positions.
1272 If the first argument is unsigned, zero bits shall fill the empty
1275 <pre> <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1276 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1277 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1278 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1279 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1282 <!-- ======================================================================= -->
1283 <div class="doc_subsection"> <a name="memoryops">Memory Access
1284 Operations</a></div>
1285 <div class="doc_text">
1286 <p>A key design point of an SSA-based representation is how it
1287 represents memory. In LLVM, no memory locations are in SSA form, which
1288 makes things very simple. This section describes how to read, write,
1289 allocate and free memory in LLVM.</p>
1291 <!-- _______________________________________________________________________ -->
1292 <div class="doc_subsubsection"> <a name="i_malloc">'<tt>malloc</tt>'
1293 Instruction</a> </div>
1294 <div class="doc_text">
1296 <pre> <result> = malloc <type>, uint <NumElements> <i>; yields {type*}:result</i>
1297 <result> = malloc <type> <i>; yields {type*}:result</i>
1300 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
1301 heap and returns a pointer to it.</p>
1303 <p>The '<tt>malloc</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1304 bytes of memory from the operating system and returns a pointer of the
1305 appropriate type to the program. The second form of the instruction is
1306 a shorter version of the first instruction that defaults to allocating
1308 <p>'<tt>type</tt>' must be a sized type.</p>
1310 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
1311 a pointer is returned.</p>
1313 <pre> %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1316 href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1317 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1318 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1321 <!-- _______________________________________________________________________ -->
1322 <div class="doc_subsubsection"> <a name="i_free">'<tt>free</tt>'
1323 Instruction</a> </div>
1324 <div class="doc_text">
1326 <pre> free <type> <value> <i>; yields {void}</i>
1329 <p>The '<tt>free</tt>' instruction returns memory back to the unused
1330 memory heap, to be reallocated in the future.</p>
1333 <p>'<tt>value</tt>' shall be a pointer value that points to a value
1334 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
1337 <p>Access to the memory pointed to by the pointer is not longer defined
1338 after this instruction executes.</p>
1340 <pre> %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1341 free [4 x ubyte]* %array
1344 <!-- _______________________________________________________________________ -->
1345 <div class="doc_subsubsection"> <a name="i_alloca">'<tt>alloca</tt>'
1346 Instruction</a> </div>
1347 <div class="doc_text">
1349 <pre> <result> = alloca <type>, uint <NumElements> <i>; yields {type*}:result</i>
1350 <result> = alloca <type> <i>; yields {type*}:result</i>
1353 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
1354 stack frame of the procedure that is live until the current function
1355 returns to its caller.</p>
1357 <p>The the '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1358 bytes of memory on the runtime stack, returning a pointer of the
1359 appropriate type to the program. The second form of the instruction is
1360 a shorter version of the first that defaults to allocating one element.</p>
1361 <p>'<tt>type</tt>' may be any sized type.</p>
1363 <p>Memory is allocated, a pointer is returned. '<tt>alloca</tt>'d
1364 memory is automatically released when the function returns. The '<tt>alloca</tt>'
1365 instruction is commonly used to represent automatic variables that must
1366 have an address available. When the function returns (either with the <tt><a
1367 href="#i_ret">ret</a></tt> or <tt><a href="#i_invoke">invoke</a></tt>
1368 instructions), the memory is reclaimed.</p>
1370 <pre> %ptr = alloca int <i>; yields {int*}:ptr</i>
1371 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
1374 <!-- _______________________________________________________________________ -->
1375 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
1376 Instruction</a> </div>
1377 <div class="doc_text">
1379 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
1381 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
1383 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
1384 address to load from. The pointer must point to a <a
1385 href="t_firstclass">first class</a> type. If the <tt>load</tt> is
1386 marked as <tt>volatile</tt> then the optimizer is not allowed to modify
1387 the number or order of execution of this <tt>load</tt> with other
1388 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
1391 <p>The location of memory pointed to is loaded.</p>
1393 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1395 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1396 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1399 <!-- _______________________________________________________________________ -->
1400 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
1401 Instruction</a> </div>
1403 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1404 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1407 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
1409 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
1410 to store and an address to store it into. The type of the '<tt><pointer></tt>'
1411 operand must be a pointer to the type of the '<tt><value></tt>'
1412 operand. If the <tt>store</tt> is marked as <tt>volatile</tt> then the
1413 optimizer is not allowed to modify the number or order of execution of
1414 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
1415 href="#i_store">store</a></tt> instructions.</p>
1417 <p>The contents of memory are updated to contain '<tt><value></tt>'
1418 at the location specified by the '<tt><pointer></tt>' operand.</p>
1420 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1422 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1423 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1425 <!-- _______________________________________________________________________ -->
1426 <div class="doc_subsubsection"> <a name="i_getelementptr">'<tt>getelementptr</tt>'
1427 Instruction</a> </div>
1428 <div class="doc_text">
1430 <pre> <result> = getelementptr <ty>* <ptrval>{, long <aidx>|, ubyte <sidx>}*<br></pre>
1432 <p>The '<tt>getelementptr</tt>' instruction is used to get the address
1433 of a subelement of an aggregate data structure.</p>
1435 <p>This instruction takes a list of <tt>long</tt> values and <tt>ubyte</tt>
1436 constants that indicate what form of addressing to perform. The actual
1437 types of the arguments provided depend on the type of the first pointer
1438 argument. The '<tt>getelementptr</tt>' instruction is used to index
1439 down through the type levels of a structure.</p>
1440 <p>For example, let's consider a C code fragment and how it gets
1441 compiled to LLVM:</p>
1442 <pre>struct RT {<br> char A;<br> int B[10][20];<br> char C;<br>};<br>struct ST {<br> int X;<br> double Y;<br> struct RT Z;<br>};<br><br>int *foo(struct ST *s) {<br> return &s[1].Z.B[5][13];<br>}<br></pre>
1443 <p>The LLVM code generated by the GCC frontend is:</p>
1444 <pre>%RT = type { sbyte, [10 x [20 x int]], sbyte }<br>%ST = type { int, double, %RT }<br><br>int* "foo"(%ST* %s) {<br> %reg = getelementptr %ST* %s, long 1, ubyte 2, ubyte 1, long 5, long 13<br> ret int* %reg<br>}<br></pre>
1446 <p>The index types specified for the '<tt>getelementptr</tt>'
1447 instruction depend on the pointer type that is being index into. <a
1448 href="t_pointer">Pointer</a> and <a href="t_array">array</a> types
1449 require '<tt>long</tt>' values, and <a href="t_struct">structure</a>
1450 types require '<tt>ubyte</tt>' <b>constants</b>.</p>
1451 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
1452 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int,
1453 double, %RT }</tt>' type, a structure. The second index indexes into
1454 the third element of the structure, yielding a '<tt>%RT</tt>' = '<tt>{
1455 sbyte, [10 x [20 x int]], sbyte }</tt>' type, another structure. The
1456 third index indexes into the second element of the structure, yielding
1457 a '<tt>[10 x [20 x int]]</tt>' type, an array. The two dimensions of
1458 the array are subscripted into, yielding an '<tt>int</tt>' type. The '<tt>getelementptr</tt>'
1459 instruction return a pointer to this element, thus yielding a '<tt>int*</tt>'
1461 <p>Note that it is perfectly legal to index partially through a
1462 structure, returning a pointer to an inner element. Because of this,
1463 the LLVM code for the given testcase is equivalent to:</p>
1464 <pre>int* "foo"(%ST* %s) {<br> %t1 = getelementptr %ST* %s , long 1 <i>; yields %ST*:%t1</i>
1465 %t2 = getelementptr %ST* %t1, long 0, ubyte 2 <i>; yields %RT*:%t2</i>
1466 %t3 = getelementptr %RT* %t2, long 0, ubyte 1 <i>; yields [10 x [20 x int]]*:%t3</i>
1467 %t4 = getelementptr [10 x [20 x int]]* %t3, long 0, long 5 <i>; yields [20 x int]*:%t4</i>
1468 %t5 = getelementptr [20 x int]* %t4, long 0, long 13 <i>; yields int*:%t5</i>
1473 <pre> <i>; yields [12 x ubyte]*:aptr</i>
1474 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, ubyte 1<br></pre>
1475 <h5> Note To The Novice:</h5>
1476 When using indexing into global arrays with the '<tt>getelementptr</tt>'
1477 instruction, you must remember that the </div>
1478 <!-- ======================================================================= -->
1479 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
1480 <div class="doc_text">
1481 <p>The instructions in this catagory are the "miscellaneous"
1482 instructions, which defy better classification.</p>
1484 <!-- _______________________________________________________________________ -->
1485 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
1486 Instruction</a> </div>
1487 <div class="doc_text">
1489 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
1491 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
1492 the SSA graph representing the function.</p>
1494 <p>The type of the incoming values are specified with the first type
1495 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
1496 as arguments, with one pair for each predecessor basic block of the
1497 current block. Only values of <a href="#t_firstclass">first class</a>
1498 type may be used as the value arguments to the PHI node. Only labels
1499 may be used as the label arguments.</p>
1500 <p>There must be no non-phi instructions between the start of a basic
1501 block and the PHI instructions: i.e. PHI instructions must be first in
1504 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
1505 value specified by the parameter, depending on which basic block we
1506 came from in the last <a href="#terminators">terminator</a> instruction.</p>
1508 <pre>Loop: ; Infinite loop that counts from 0 on up...<br> %indvar = phi uint [ 0, %LoopHeader ], [ %nextindvar, %Loop ]<br> %nextindvar = add uint %indvar, 1<br> br label %Loop<br></pre>
1511 <!-- _______________________________________________________________________ -->
1512 <div class="doc_subsubsection">
1513 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
1516 <div class="doc_text">
1521 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
1527 The '<tt>cast</tt>' instruction is used as the primitive means to convert
1528 integers to floating point, change data type sizes, and break type safety (by
1536 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
1537 class value, and a type to cast it to, which must also be a <a
1538 href="#t_firstclass">first class</a> type.
1544 This instruction follows the C rules for explicit casts when determining how the
1545 data being cast must change to fit in its new container.
1549 When casting to bool, any value that would be considered true in the context of
1550 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
1551 all else are '<tt>false</tt>'.
1555 When extending an integral value from a type of one signness to another (for
1556 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
1557 <b>source</b> value is signed, and zero-extended if the source value is
1558 unsigned. <tt>bool</tt> values are always zero extended into either zero or
1565 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
1566 %Y = cast int 123 to bool <i>; yields bool:true</i>
1570 <!-- _______________________________________________________________________ -->
1571 <div class="doc_subsubsection">
1572 <a name="i_select">'<tt>select</tt>' Instruction</a>
1575 <div class="doc_text">
1580 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
1586 The '<tt>select</tt>' instruction is used to choose one value based on a
1587 condition, without branching.
1594 The '<tt>select</tt>' instruction requires a boolean value indicating the condition, and two values of the same <a href="#t_firstclass">first class</a> type.
1600 If the boolean condition evaluates to true, the instruction returns the first
1601 value argument, otherwise it returns the second value argument.
1607 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
1615 <!-- _______________________________________________________________________ -->
1616 <div class="doc_subsubsection"> <a name="i_call">'<tt>call</tt>'
1617 Instruction</a> </div>
1618 <div class="doc_text">
1620 <pre> <result> = call <ty>* <fnptrval>(<param list>)<br></pre>
1622 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
1624 <p>This instruction requires several arguments:</p>
1627 <p>'<tt>ty</tt>': shall be the signature of the pointer to function
1628 value being invoked. The argument types must match the types implied
1629 by this signature.</p>
1632 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a
1633 function to be invoked. In most cases, this is a direct function
1634 invocation, but indirect <tt>call</tt>s are just as possible,
1635 calling an arbitrary pointer to function values.</p>
1638 <p>'<tt>function args</tt>': argument list whose types match the
1639 function signature argument types. If the function signature
1640 indicates the function accepts a variable number of arguments, the
1641 extra arguments can be specified.</p>
1645 <p>The '<tt>call</tt>' instruction is used to cause control flow to
1646 transfer to a specified function, with its incoming arguments bound to
1647 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
1648 instruction in the called function, control flow continues with the
1649 instruction after the function call, and the return value of the
1650 function is bound to the result argument. This is a simpler case of
1651 the <a href="#i_invoke">invoke</a> instruction.</p>
1653 <pre> %retval = call int %test(int %argc)<br> call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);<br></pre>
1655 <!-- _______________________________________________________________________ -->
1656 <div class="doc_subsubsection"> <a name="i_vanext">'<tt>vanext</tt>'
1657 Instruction</a> </div>
1658 <div class="doc_text">
1660 <pre> <resultarglist> = vanext <va_list> <arglist>, <argty><br></pre>
1662 <p>The '<tt>vanext</tt>' instruction is used to access arguments passed
1663 through the "variable argument" area of a function call. It is used to
1664 implement the <tt>va_arg</tt> macro in C.</p>
1666 <p>This instruction takes a <tt>valist</tt> value and the type of the
1667 argument. It returns another <tt>valist</tt>.</p>
1669 <p>The '<tt>vanext</tt>' instruction advances the specified <tt>valist</tt>
1670 past an argument of the specified type. In conjunction with the <a
1671 href="#i_vaarg"><tt>vaarg</tt></a> instruction, it is used to implement
1672 the <tt>va_arg</tt> macro available in C. For more information, see
1673 the variable argument handling <a href="#int_varargs">Intrinsic
1675 <p>It is legal for this instruction to be called in a function which
1676 does not take a variable number of arguments, for example, the <tt>vfprintf</tt>
1678 <p><tt>vanext</tt> is an LLVM instruction instead of an <a
1679 href="#intrinsics">intrinsic function</a> because it takes an type as
1682 <p>See the <a href="#int_varargs">variable argument processing</a>
1685 <!-- _______________________________________________________________________ -->
1686 <div class="doc_subsubsection"> <a name="i_vaarg">'<tt>vaarg</tt>'
1687 Instruction</a> </div>
1688 <div class="doc_text">
1690 <pre> <resultval> = vaarg <va_list> <arglist>, <argty><br></pre>
1692 <p>The '<tt>vaarg</tt>' instruction is used to access arguments passed
1693 through the "variable argument" area of a function call. It is used to
1694 implement the <tt>va_arg</tt> macro in C.</p>
1696 <p>This instruction takes a <tt>valist</tt> value and the type of the
1697 argument. It returns a value of the specified argument type.</p>
1699 <p>The '<tt>vaarg</tt>' instruction loads an argument of the specified
1700 type from the specified <tt>va_list</tt>. In conjunction with the <a
1701 href="#i_vanext"><tt>vanext</tt></a> instruction, it is used to
1702 implement the <tt>va_arg</tt> macro available in C. For more
1703 information, see the variable argument handling <a href="#int_varargs">Intrinsic
1705 <p>It is legal for this instruction to be called in a function which
1706 does not take a variable number of arguments, for example, the <tt>vfprintf</tt>
1708 <p><tt>vaarg</tt> is an LLVM instruction instead of an <a
1709 href="#intrinsics">intrinsic function</a> because it takes an type as
1712 <p>See the <a href="#int_varargs">variable argument processing</a>
1716 <!-- *********************************************************************** -->
1717 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
1718 <!-- *********************************************************************** -->
1720 <div class="doc_text">
1722 <p>LLVM supports the notion of an "intrinsic function". These functions have
1723 well known names and semantics, and are required to follow certain
1724 restrictions. Overall, these instructions represent an extension mechanism for
1725 the LLVM language that does not require changing all of the transformations in
1726 LLVM to add to the language (or the bytecode reader/writer, the parser,
1729 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix, this
1730 prefix is reserved in LLVM for intrinsic names, thus functions may not be named
1731 this. Intrinsic functions must always be external functions: you cannot define
1732 the body of intrinsic functions. Intrinsic functions may only be used in call
1733 or invoke instructions: it is illegal to take the address of an intrinsic
1734 function. Additionally, because intrinsic functions are part of the LLVM
1735 language, it is required that they all be documented here if any are added.</p>
1739 Adding an intrinsic to LLVM is straight-forward if it is possible to express the
1740 concept in LLVM directly (ie, code generator support is not _required_). To do
1741 this, extend the default implementation of the IntrinsicLowering class to handle
1742 the intrinsic. Code generators use this class to lower intrinsics they do not
1743 understand to raw LLVM instructions that they do.
1748 <!-- ======================================================================= -->
1749 <div class="doc_subsection">
1750 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
1753 <div class="doc_text">
1754 <p>Variable argument support is defined in LLVM with the <a
1755 href="#i_vanext"><tt>vanext</tt></a> instruction and these three
1756 intrinsic functions. These functions are related to the similarly
1757 named macros defined in the <tt><stdarg.h></tt> header file.</p>
1758 <p>All of these functions operate on arguments that use a
1759 target-specific value type "<tt>va_list</tt>". The LLVM assembly
1760 language reference manual does not define what this type is, so all
1761 transformations should be prepared to handle intrinsics with any type
1763 <p>This example shows how the <a href="#i_vanext"><tt>vanext</tt></a>
1764 instruction and the variable argument handling intrinsic functions are
1767 int %test(int %X, ...) {
1768 ; Initialize variable argument processing
1769 %ap = call sbyte* %<a href="#i_va_start">llvm.va_start</a>()
1771 ; Read a single integer argument
1772 %tmp = vaarg sbyte* %ap, int
1774 ; Advance to the next argument
1775 %ap2 = vanext sbyte* %ap, int
1777 ; Demonstrate usage of llvm.va_copy and llvm.va_end
1778 %aq = call sbyte* %<a href="#i_va_copy">llvm.va_copy</a>(sbyte* %ap2)
1779 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %aq)
1781 ; Stop processing of arguments.
1782 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %ap2)
1788 <!-- _______________________________________________________________________ -->
1789 <div class="doc_subsubsection">
1790 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
1794 <div class="doc_text">
1796 <pre> call va_list ()* %llvm.va_start()<br></pre>
1798 <p>The '<tt>llvm.va_start</tt>' intrinsic returns a new <tt><arglist></tt>
1799 for subsequent use by the variable argument intrinsics.</p>
1801 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
1802 macro available in C. In a target-dependent way, it initializes and
1803 returns a <tt>va_list</tt> element, so that the next <tt>vaarg</tt>
1804 will produce the first variable argument passed to the function. Unlike
1805 the C <tt>va_start</tt> macro, this intrinsic does not need to know the
1806 last argument of the function, the compiler can figure that out.</p>
1807 <p>Note that this intrinsic function is only legal to be called from
1808 within the body of a variable argument function.</p>
1811 <!-- _______________________________________________________________________ -->
1812 <div class="doc_subsubsection">
1813 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
1816 <div class="doc_text">
1818 <pre> call void (va_list)* %llvm.va_end(va_list <arglist>)<br></pre>
1820 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
1821 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
1822 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
1824 <p>The argument is a <tt>va_list</tt> to destroy.</p>
1826 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
1827 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
1828 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
1829 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
1830 with calls to <tt>llvm.va_end</tt>.</p>
1833 <!-- _______________________________________________________________________ -->
1834 <div class="doc_subsubsection">
1835 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
1838 <div class="doc_text">
1840 <pre> call va_list (va_list)* %llvm.va_copy(va_list <destarglist>)<br></pre>
1842 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument
1843 position from the source argument list to the destination argument list.</p>
1845 <p>The argument is the <tt>va_list</tt> to copy.</p>
1847 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
1848 macro available in C. In a target-dependent way, it copies the source <tt>va_list</tt>
1849 element into the returned list. This intrinsic is necessary because the <tt><a
1850 href="i_va_start">llvm.va_start</a></tt> intrinsic may be arbitrarily
1851 complex and require memory allocation, for example.</p>
1854 <!-- ======================================================================= -->
1855 <div class="doc_subsection">
1856 <a name="int_codegen">Code Generator Intrinsics</a>
1859 <div class="doc_text">
1861 These intrinsics are provided by LLVM to expose special features that may only
1862 be implemented with code generator support.
1867 <!-- _______________________________________________________________________ -->
1868 <div class="doc_subsubsection">
1869 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
1872 <div class="doc_text">
1876 call void* ()* %llvm.returnaddress(uint <level>)
1882 The '<tt>llvm.returnaddress</tt>' intrinsic returns a target-specific value
1883 indicating the return address of the current function or one of its callers.
1889 The argument to this intrinsic indicates which function to return the address
1890 for. Zero indicates the calling function, one indicates its caller, etc. The
1891 argument is <b>required</b> to be a constant integer value.
1897 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
1898 the return address of the specified call frame, or zero if it cannot be
1899 identified. The value returned by this intrinsic is likely to be incorrect or 0
1900 for arguments other than zero, so it should only be used for debugging purposes.
1904 Note that calling this intrinsic does not prevent function inlining or other
1905 aggressive transformations, so the value returned may not that of the obvious
1906 source-language caller.
1911 <!-- _______________________________________________________________________ -->
1912 <div class="doc_subsubsection">
1913 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
1916 <div class="doc_text">
1920 call void* ()* %llvm.frameaddress(uint <level>)
1926 The '<tt>llvm.frameaddress</tt>' intrinsic returns the target-specific frame
1927 pointer value for the specified stack frame.
1933 The argument to this intrinsic indicates which function to return the frame
1934 pointer for. Zero indicates the calling function, one indicates its caller,
1935 etc. The argument is <b>required</b> to be a constant integer value.
1941 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
1942 the frame address of the specified call frame, or zero if it cannot be
1943 identified. The value returned by this intrinsic is likely to be incorrect or 0
1944 for arguments other than zero, so it should only be used for debugging purposes.
1948 Note that calling this intrinsic does not prevent function inlining or other
1949 aggressive transformations, so the value returned may not that of the obvious
1950 source-language caller.
1955 <!-- ======================================================================= -->
1956 <div class="doc_subsection">
1957 <a name="int_libc">Standard C Library Intrinsics</a>
1960 <div class="doc_text">
1962 LLVM provides intrinsics for a few important standard C library functions.
1963 These intrinsics allow source-language front-ends to pass information about the
1964 alignment of the pointer arguments to the code generator, providing opportunity
1965 for more efficient code generation.
1970 <!-- _______________________________________________________________________ -->
1971 <div class="doc_subsubsection">
1972 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
1975 <div class="doc_text">
1979 call void (sbyte*, sbyte*, uint, uint)* %llvm.memcpy(sbyte* <dest>, sbyte* <src>,
1980 uint <len>, uint <align>)
1986 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
1987 location to the destination location.
1991 Note that, unlike the standard libc function, the <tt>llvm.memcpy</tt> intrinsic
1992 does not return a value, and takes an extra alignment argument.
1998 The first argument is a pointer to the destination, the second is a pointer to
1999 the source. The third argument is an (arbitrarily sized) integer argument
2000 specifying the number of bytes to copy, and the fourth argument is the alignment
2001 of the source and destination locations.
2005 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2006 the caller guarantees that the size of the copy is a multiple of the alignment
2007 and that both the source and destination pointers are aligned to that boundary.
2013 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2014 location to the destination location, which are not allowed to overlap. It
2015 copies "len" bytes of memory over. If the argument is known to be aligned to
2016 some boundary, this can be specified as the fourth argument, otherwise it should
2022 <!-- _______________________________________________________________________ -->
2023 <div class="doc_subsubsection">
2024 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
2027 <div class="doc_text">
2031 call void (sbyte*, sbyte*, uint, uint)* %llvm.memmove(sbyte* <dest>, sbyte* <src>,
2032 uint <len>, uint <align>)
2038 The '<tt>llvm.memmove</tt>' intrinsic moves a block of memory from the source
2039 location to the destination location. It is similar to the '<tt>llvm.memcpy</tt>'
2040 intrinsic but allows the two memory locations to overlap.
2044 Note that, unlike the standard libc function, the <tt>llvm.memmove</tt> intrinsic
2045 does not return a value, and takes an extra alignment argument.
2051 The first argument is a pointer to the destination, the second is a pointer to
2052 the source. The third argument is an (arbitrarily sized) integer argument
2053 specifying the number of bytes to copy, and the fourth argument is the alignment
2054 of the source and destination locations.
2058 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2059 the caller guarantees that the size of the copy is a multiple of the alignment
2060 and that both the source and destination pointers are aligned to that boundary.
2066 The '<tt>llvm.memmove</tt>' intrinsic copies a block of memory from the source
2067 location to the destination location, which may overlap. It
2068 copies "len" bytes of memory over. If the argument is known to be aligned to
2069 some boundary, this can be specified as the fourth argument, otherwise it should
2075 <!-- _______________________________________________________________________ -->
2076 <div class="doc_subsubsection">
2077 <a name="i_memset">'<tt>llvm.memset</tt>' Intrinsic</a>
2080 <div class="doc_text">
2084 call void (sbyte*, ubyte, uint, uint)* %llvm.memset(sbyte* <dest>, ubyte <val>,
2085 uint <len>, uint <align>)
2091 The '<tt>llvm.memset</tt>' intrinsic fills a block of memory with a particular
2096 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
2097 does not return a value, and takes an extra alignment argument.
2103 The first argument is a pointer to the destination to fill, the second is the
2104 byte value to fill it with, the third argument is an (arbitrarily sized) integer
2105 argument specifying the number of bytes to fill, and the fourth argument is the
2106 known alignment of destination location.
2110 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2111 the caller guarantees that the size of the copy is a multiple of the alignment
2112 and that the destination pointer is aligned to that boundary.
2118 The '<tt>llvm.memset</tt>' intrinsic fills "len" bytes of memory starting at the
2119 destination location. If the argument is known to be aligned to some boundary,
2120 this can be specified as the fourth argument, otherwise it should be set to 0 or
2126 <!-- ======================================================================= -->
2127 <div class="doc_subsection">
2128 <a name="int_debugger">Debugger Intrinsics</a>
2131 <div class="doc_text">
2133 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
2134 are described in the <a
2135 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
2136 Debugging</a> document.
2141 <!-- *********************************************************************** -->
2144 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2145 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
2146 <a href="http://validator.w3.org/check/referer"><img
2147 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
2149 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
2150 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
2151 Last modified: $Date$