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_os">Operating System Intrinsics</a>
108 <li><a href="#i_readport">'<tt>llvm.readport</tt>' Intrinsic</a></li>
109 <li><a href="#i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a></li>
110 <li><a href="#i_readio">'<tt>llvm.readio</tt>' Intrinsic</a></li>
111 <li><a href="#i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a></li>
113 <li><a href="#int_libc">Standard C Library Intrinsics</a>
115 <li><a href="#i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a></li>
116 <li><a href="#i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a></li>
117 <li><a href="#i_memset">'<tt>llvm.memset</tt>' Intrinsic</a></li>
120 <li><a href="#int_debugger">Debugger intrinsics</a>
124 <div class="doc_text">
125 <p><b>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
126 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></b></p>
129 <!-- *********************************************************************** -->
130 <div class="doc_section"> <a name="abstract">Abstract </a></div>
131 <!-- *********************************************************************** -->
132 <div class="doc_text">
133 <p>This document is a reference manual for the LLVM assembly language.
134 LLVM is an SSA based representation that provides type safety,
135 low-level operations, flexibility, and the capability of representing
136 'all' high-level languages cleanly. It is the common code
137 representation used throughout all phases of the LLVM compilation
140 <!-- *********************************************************************** -->
141 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
142 <!-- *********************************************************************** -->
143 <div class="doc_text">
144 <p>The LLVM code representation is designed to be used in three
145 different forms: as an in-memory compiler IR, as an on-disk bytecode
146 representation (suitable for fast loading by a Just-In-Time compiler),
147 and as a human readable assembly language representation. This allows
148 LLVM to provide a powerful intermediate representation for efficient
149 compiler transformations and analysis, while providing a natural means
150 to debug and visualize the transformations. The three different forms
151 of LLVM are all equivalent. This document describes the human readable
152 representation and notation.</p>
153 <p>The LLVM representation aims to be a light-weight and low-level
154 while being expressive, typed, and extensible at the same time. It
155 aims to be a "universal IR" of sorts, by being at a low enough level
156 that high-level ideas may be cleanly mapped to it (similar to how
157 microprocessors are "universal IR's", allowing many source languages to
158 be mapped to them). By providing type information, LLVM can be used as
159 the target of optimizations: for example, through pointer analysis, it
160 can be proven that a C automatic variable is never accessed outside of
161 the current function... allowing it to be promoted to a simple SSA
162 value instead of a memory location.</p>
164 <!-- _______________________________________________________________________ -->
165 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
166 <div class="doc_text">
167 <p>It is important to note that this document describes 'well formed'
168 LLVM assembly language. There is a difference between what the parser
169 accepts and what is considered 'well formed'. For example, the
170 following instruction is syntactically okay, but not well formed:</p>
171 <pre> %x = <a href="#i_add">add</a> int 1, %x<br></pre>
172 <p>...because the definition of <tt>%x</tt> does not dominate all of
173 its uses. The LLVM infrastructure provides a verification pass that may
174 be used to verify that an LLVM module is well formed. This pass is
175 automatically run by the parser after parsing input assembly, and by
176 the optimizer before it outputs bytecode. The violations pointed out
177 by the verifier pass indicate bugs in transformation passes or input to
179 <!-- Describe the typesetting conventions here. --> </div>
180 <!-- *********************************************************************** -->
181 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
182 <!-- *********************************************************************** -->
183 <div class="doc_text">
184 <p>LLVM uses three different forms of identifiers, for different
187 <li>Numeric constants are represented as you would expect: 12, -3
188 123.421, etc. Floating point constants have an optional hexadecimal
190 <li>Named values are represented as a string of characters with a '%'
191 prefix. For example, %foo, %DivisionByZero,
192 %a.really.long.identifier. The actual regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
193 Identifiers which require other characters in their names can be
194 surrounded with quotes. In this way, anything except a <tt>"</tt>
195 character can be used in a name.</li>
196 <li>Unnamed values are represented as an unsigned numeric value with
197 a '%' prefix. For example, %12, %2, %44.</li>
199 <p>LLVM requires that values start with a '%' sign for two reasons:
200 Compilers don't need to worry about name clashes with reserved words,
201 and the set of reserved words may be expanded in the future without
202 penalty. Additionally, unnamed identifiers allow a compiler to quickly
203 come up with a temporary variable without having to avoid symbol table
205 <p>Reserved words in LLVM are very similar to reserved words in other
206 languages. There are keywords for different opcodes ('<tt><a
207 href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
208 href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
209 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>',
210 etc...), and others. These reserved words cannot conflict with
211 variable names, because none of them start with a '%' character.</p>
212 <p>Here is an example of LLVM code to multiply the integer variable '<tt>%X</tt>'
215 <pre> %result = <a href="#i_mul">mul</a> uint %X, 8<br></pre>
216 <p>After strength reduction:</p>
217 <pre> %result = <a href="#i_shl">shl</a> uint %X, ubyte 3<br></pre>
218 <p>And the hard way:</p>
219 <pre> <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
221 href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
223 href="#i_add">add</a> uint %1, %1<br></pre>
224 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
225 important lexical features of LLVM:</p>
227 <li>Comments are delimited with a '<tt>;</tt>' and go until the end
229 <li>Unnamed temporaries are created when the result of a computation
230 is not assigned to a named value.</li>
231 <li>Unnamed temporaries are numbered sequentially</li>
233 <p>...and it also show a convention that we follow in this document.
234 When demonstrating instructions, we will follow an instruction with a
235 comment that defines the type and name of value produced. Comments are
236 shown in italic text.</p>
237 <p>The one non-intuitive notation for constants is the optional
238 hexidecimal form of floating point constants. For example, the form '<tt>double
239 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
240 4.5e+15</tt>' which is also supported by the parser. The only time
241 hexadecimal floating point constants are useful (and the only time that
242 they are generated by the disassembler) is when an FP constant has to
243 be emitted that is not representable as a decimal floating point number
244 exactly. For example, NaN's, infinities, and other special cases are
245 represented in their IEEE hexadecimal format so that assembly and
246 disassembly do not cause any bits to change in the constants.</p>
248 <!-- *********************************************************************** -->
249 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
250 <!-- *********************************************************************** -->
251 <div class="doc_text">
252 <p>The LLVM type system is one of the most important features of the
253 intermediate representation. Being typed enables a number of
254 optimizations to be performed on the IR directly, without having to do
255 extra analyses on the side before the transformation. A strong type
256 system makes it easier to read the generated code and enables novel
257 analyses and transformations that are not feasible to perform on normal
258 three address code representations.</p>
259 <!-- The written form for the type system was heavily influenced by the
260 syntactic problems with types in the C language<sup><a
261 href="#rw_stroustrup">1</a></sup>.<p> --> </div>
262 <!-- ======================================================================= -->
263 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
264 <div class="doc_text">
265 <p>The primitive types are the fundamental building blocks of the LLVM
266 system. The current set of primitive types are as follows:</p>
268 <table border="0" style="align: center">
272 <table border="1" cellspacing="0" cellpadding="4" style="align: center">
275 <td><tt>void</tt></td>
279 <td><tt>ubyte</tt></td>
280 <td>Unsigned 8 bit value</td>
283 <td><tt>ushort</tt></td>
284 <td>Unsigned 16 bit value</td>
287 <td><tt>uint</tt></td>
288 <td>Unsigned 32 bit value</td>
291 <td><tt>ulong</tt></td>
292 <td>Unsigned 64 bit value</td>
295 <td><tt>float</tt></td>
296 <td>32 bit floating point value</td>
299 <td><tt>label</tt></td>
300 <td>Branch destination</td>
306 <table border="1" cellspacing="0" cellpadding="4">
309 <td><tt>bool</tt></td>
310 <td>True or False value</td>
313 <td><tt>sbyte</tt></td>
314 <td>Signed 8 bit value</td>
317 <td><tt>short</tt></td>
318 <td>Signed 16 bit value</td>
321 <td><tt>int</tt></td>
322 <td>Signed 32 bit value</td>
325 <td><tt>long</tt></td>
326 <td>Signed 64 bit value</td>
329 <td><tt>double</tt></td>
330 <td>64 bit floating point value</td>
340 <!-- _______________________________________________________________________ -->
341 <div class="doc_subsubsection"> <a name="t_classifications">Type
342 Classifications</a> </div>
343 <div class="doc_text">
344 <p>These different primitive types fall into a few useful
347 <table border="1" cellspacing="0" cellpadding="4">
350 <td><a name="t_signed">signed</a></td>
351 <td><tt>sbyte, short, int, long, float, double</tt></td>
354 <td><a name="t_unsigned">unsigned</a></td>
355 <td><tt>ubyte, ushort, uint, ulong</tt></td>
358 <td><a name="t_integer">integer</a></td>
359 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
362 <td><a name="t_integral">integral</a></td>
363 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
366 <td><a name="t_floating">floating point</a></td>
367 <td><tt>float, double</tt></td>
370 <td><a name="t_firstclass">first class</a></td>
371 <td><tt>bool, ubyte, sbyte, ushort, short,<br>
372 uint, int, ulong, long, float, double, <a href="#t_pointer">pointer</a></tt></td>
377 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
378 most important. Values of these types are the only ones which can be
379 produced by instructions, passed as arguments, or used as operands to
380 instructions. This means that all structures and arrays must be
381 manipulated either by pointer or by component.</p>
383 <!-- ======================================================================= -->
384 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
385 <div class="doc_text">
386 <p>The real power in LLVM comes from the derived types in the system.
387 This is what allows a programmer to represent arrays, functions,
388 pointers, and other useful types. Note that these derived types may be
389 recursive: For example, it is possible to have a two dimensional array.</p>
391 <!-- _______________________________________________________________________ -->
392 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
393 <div class="doc_text">
395 <p>The array type is a very simple derived type that arranges elements
396 sequentially in memory. The array type requires a size (number of
397 elements) and an underlying data type.</p>
399 <pre> [<# elements> x <elementtype>]<br></pre>
400 <p>The number of elements is a constant integer value, elementtype may
401 be any type with a size.</p>
403 <p> <tt>[40 x int ]</tt>: Array of 40 integer values.<br>
404 <tt>[41 x int ]</tt>: Array of 41 integer values.<br>
405 <tt>[40 x uint]</tt>: Array of 40 unsigned integer values.</p>
407 <p>Here are some examples of multidimensional arrays:</p>
409 <table border="0" cellpadding="0" cellspacing="0">
412 <td><tt>[3 x [4 x int]]</tt></td>
413 <td>: 3x4 array integer values.</td>
416 <td><tt>[12 x [10 x float]]</tt></td>
417 <td>: 12x10 array of single precision floating point values.</td>
420 <td><tt>[2 x [3 x [4 x uint]]]</tt></td>
421 <td>: 2x3x4 array of unsigned integer values.</td>
427 <!-- _______________________________________________________________________ -->
428 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
429 <div class="doc_text">
431 <p>The function type can be thought of as a function signature. It
432 consists of a return type and a list of formal parameter types.
433 Function types are usually used to build virtual function tables
434 (which are structures of pointers to functions), for indirect function
435 calls, and when defining a function.</p>
437 The return type of a function type cannot be an aggregate type.
440 <pre> <returntype> (<parameter list>)<br></pre>
441 <p>Where '<tt><parameter list></tt>' is a comma-separated list of
442 type specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
443 which indicates that the function takes a variable number of arguments.
444 Variable argument functions can access their arguments with the <a
445 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
448 <table border="0" cellpadding="0" cellspacing="0">
451 <td><tt>int (int)</tt></td>
452 <td>: function taking an <tt>int</tt>, returning an <tt>int</tt></td>
455 <td><tt>float (int, int *) *</tt></td>
456 <td>: <a href="#t_pointer">Pointer</a> to a function that takes
457 an <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
458 returning <tt>float</tt>.</td>
461 <td><tt>int (sbyte *, ...)</tt></td>
462 <td>: A vararg function that takes at least one <a
463 href="#t_pointer">pointer</a> to <tt>sbyte</tt> (signed char in C),
464 which returns an integer. This is the signature for <tt>printf</tt>
471 <!-- _______________________________________________________________________ -->
472 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
473 <div class="doc_text">
475 <p>The structure type is used to represent a collection of data members
476 together in memory. The packing of the field types is defined to match
477 the ABI of the underlying processor. The elements of a structure may
478 be any type that has a size.</p>
479 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
480 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
481 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
484 <pre> { <type list> }<br></pre>
487 <table border="0" cellpadding="0" cellspacing="0">
490 <td><tt>{ int, int, int }</tt></td>
491 <td>: a triple of three <tt>int</tt> values</td>
494 <td><tt>{ float, int (int) * }</tt></td>
495 <td>: A pair, where the first element is a <tt>float</tt> and the
496 second element is a <a href="#t_pointer">pointer</a> to a <a
497 href="t_function">function</a> that takes an <tt>int</tt>, returning
498 an <tt>int</tt>.</td>
504 <!-- _______________________________________________________________________ -->
505 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
506 <div class="doc_text">
508 <p>As in many languages, the pointer type represents a pointer or
509 reference to another object, which must live in memory.</p>
511 <pre> <type> *<br></pre>
514 <table border="0" cellpadding="0" cellspacing="0">
517 <td><tt>[4x int]*</tt></td>
518 <td>: <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a>
519 of four <tt>int</tt> values</td>
522 <td><tt>int (int *) *</tt></td>
523 <td>: A <a href="#t_pointer">pointer</a> to a <a
524 href="t_function">function</a> that takes an <tt>int</tt>, returning
525 an <tt>int</tt>.</td>
531 <!-- _______________________________________________________________________ --><!--
532 <div class="doc_subsubsection">
533 <a name="t_packed">Packed Type</a>
536 <div class="doc_text">
538 Mention/decide that packed types work with saturation or not. Maybe have a packed+saturated type in addition to just a packed type.<p>
540 Packed types should be 'nonsaturated' because standard data types are not saturated. Maybe have a saturated packed type?<p>
544 --><!-- *********************************************************************** -->
545 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
546 <!-- *********************************************************************** --><!-- ======================================================================= -->
547 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a> </div>
548 <div class="doc_text">
549 <p>LLVM programs are composed of "Module"s, each of which is a
550 translation unit of the input programs. Each module consists of
551 functions, global variables, and symbol table entries. Modules may be
552 combined together with the LLVM linker, which merges function (and
553 global variable) definitions, resolves forward declarations, and merges
554 symbol table entries. Here is an example of the "hello world" module:</p>
555 <pre><i>; Declare the string constant as a global constant...</i>
556 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
557 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
559 <i>; External declaration of the puts function</i>
560 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
562 <i>; Definition of main function</i>
563 int %main() { <i>; int()* </i>
564 <i>; Convert [13x sbyte]* to sbyte *...</i>
566 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
568 <i>; Call puts function to write out the string to stdout...</i>
570 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
572 href="#i_ret">ret</a> int 0<br>}<br></pre>
573 <p>This example is made up of a <a href="#globalvars">global variable</a>
574 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
575 function, and a <a href="#functionstructure">function definition</a>
576 for "<tt>main</tt>".</p>
577 <a name="linkage"> In general, a module is made up of a list of global
578 values, where both functions and global variables are global values.
579 Global values are represented by a pointer to a memory location (in
580 this case, a pointer to an array of char, and a pointer to a function),
581 and have one of the following linkage types:</a>
584 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
585 <dd>Global values with internal linkage are only directly accessible
586 by objects in the current module. In particular, linking code into a
587 module with an internal global value may cause the internal to be
588 renamed as necessary to avoid collisions. Because the symbol is
589 internal to the module, all references can be updated. This
590 corresponds to the notion of the '<tt>static</tt>' keyword in C, or the
591 idea of "anonymous namespaces" in C++.
594 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
595 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt>
596 linkage, with the twist that linking together two modules defining the
597 same <tt>linkonce</tt> globals will cause one of the globals to be
598 discarded. This is typically used to implement inline functions.
599 Unreferenced <tt>linkonce</tt> globals are allowed to be discarded.
602 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
603 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt>
604 linkage, except that unreferenced <tt>weak</tt> globals may not be
605 discarded. This is used to implement constructs in C such as "<tt>int
606 X;</tt>" at global scope.
609 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
610 <dd>"<tt>appending</tt>" linkage may only be applied to global
611 variables of pointer to array type. When two global variables with
612 appending linkage are linked together, the two global arrays are
613 appended together. This is the LLVM, typesafe, equivalent of having
614 the system linker append together "sections" with identical names when
618 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
619 <dd>If none of the above identifiers are used, the global is
620 externally visible, meaning that it participates in linkage and can be
621 used to resolve external symbol references.
626 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
627 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
628 variable and was linked with this one, one of the two would be renamed,
629 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
630 external (i.e., lacking any linkage declarations), they are accessible
631 outside of the current module. It is illegal for a function <i>declaration</i>
632 to have any linkage type other than "externally visible".</a></p>
635 <!-- ======================================================================= -->
636 <div class="doc_subsection">
637 <a name="globalvars">Global Variables</a>
640 <div class="doc_text">
642 <p>Global variables define regions of memory allocated at compilation
643 time instead of run-time. Global variables may optionally be
644 initialized. A variable may be defined as a global "constant", which
645 indicates that the contents of the variable will never be modified
646 (opening options for optimization).</p>
648 <p>As SSA values, global variables define pointer values that are in
649 scope (i.e. they dominate) for all basic blocks in the program. Global
650 variables always define a pointer to their "content" type because they
651 describe a region of memory, and all memory objects in LLVM are
652 accessed through pointers.</p>
657 <!-- ======================================================================= -->
658 <div class="doc_subsection">
659 <a name="functionstructure">Functions</a>
662 <div class="doc_text">
664 <p>LLVM function definitions are composed of a (possibly empty) argument list,
665 an opening curly brace, a list of basic blocks, and a closing curly brace. LLVM
666 function declarations are defined with the "<tt>declare</tt>" keyword, a
667 function name, and a function signature.</p>
669 <p>A function definition contains a list of basic blocks, forming the CFG for
670 the function. Each basic block may optionally start with a label (giving the
671 basic block a symbol table entry), contains a list of instructions, and ends
672 with a <a href="#terminators">terminator</a> instruction (such as a branch or
673 function return).</p>
675 <p>The first basic block in program is special in two ways: it is immediately
676 executed on entrance to the function, and it is not allowed to have predecessor
677 basic blocks (i.e. there can not be any branches to the entry block of a
678 function). Because the block can have no predecessors, it also cannot have any
679 <a href="#i_phi">PHI nodes</a>.</p>
681 <p>LLVM functions are identified by their name and type signature. Hence, two
682 functions with the same name but different parameter lists or return values are
683 considered different functions, and LLVM will resolves references to each
689 <!-- *********************************************************************** -->
690 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
691 <!-- *********************************************************************** -->
692 <div class="doc_text">
693 <p>The LLVM instruction set consists of several different
694 classifications of instructions: <a href="#terminators">terminator
695 instructions</a>, <a href="#binaryops">binary instructions</a>, <a
696 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
697 instructions</a>.</p>
699 <!-- ======================================================================= -->
700 <div class="doc_subsection"> <a name="terminators">Terminator
701 Instructions</a> </div>
702 <div class="doc_text">
703 <p>As mentioned <a href="#functionstructure">previously</a>, every
704 basic block in a program ends with a "Terminator" instruction, which
705 indicates which block should be executed after the current block is
706 finished. These terminator instructions typically yield a '<tt>void</tt>'
707 value: they produce control flow, not values (the one exception being
708 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
709 <p>There are five different terminator instructions: the '<a
710 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
711 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
712 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, and the '<a
713 href="#i_unwind"><tt>unwind</tt></a>' instruction.</p>
715 <!-- _______________________________________________________________________ -->
716 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
717 Instruction</a> </div>
718 <div class="doc_text">
720 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
721 ret void <i>; Return from void function</i>
724 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
725 value) from a function, back to the caller.</p>
726 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
727 returns a value and then causes control flow, and one that just causes
728 control flow to occur.</p>
730 <p>The '<tt>ret</tt>' instruction may return any '<a
731 href="#t_firstclass">first class</a>' type. Notice that a function is
732 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
733 instruction inside of the function that returns a value that does not
734 match the return type of the function.</p>
736 <p>When the '<tt>ret</tt>' instruction is executed, control flow
737 returns back to the calling function's context. If the caller is a "<a
738 href="#i_call"><tt>call</tt></a> instruction, execution continues at
739 the instruction after the call. If the caller was an "<a
740 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
741 at the beginning "normal" of the destination block. If the instruction
742 returns a value, that value shall set the call or invoke instruction's
745 <pre> ret int 5 <i>; Return an integer value of 5</i>
746 ret void <i>; Return from a void function</i>
749 <!-- _______________________________________________________________________ -->
750 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
751 <div class="doc_text">
753 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
756 <p>The '<tt>br</tt>' instruction is used to cause control flow to
757 transfer to a different basic block in the current function. There are
758 two forms of this instruction, corresponding to a conditional branch
759 and an unconditional branch.</p>
761 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
762 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
763 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
764 value as a target.</p>
766 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
767 argument is evaluated. If the value is <tt>true</tt>, control flows
768 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
769 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
771 <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
772 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
774 <!-- _______________________________________________________________________ -->
775 <div class="doc_subsubsection">
776 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
779 <div class="doc_text">
783 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
788 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
789 several different places. It is a generalization of the '<tt>br</tt>'
790 instruction, allowing a branch to occur to one of many possible
796 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
797 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
798 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
799 table is not allowed to contain duplicate constant entries.</p>
803 <p>The <tt>switch</tt> instruction specifies a table of values and
804 destinations. When the '<tt>switch</tt>' instruction is executed, this
805 table is searched for the given value. If the value is found, the
806 corresponding destination is branched to, otherwise the default value
807 it transfered to.</p>
809 <h5>Implementation:</h5>
811 <p>Depending on properties of the target machine and the particular
812 <tt>switch</tt> instruction, this instruction may be code generated in different
813 ways, for example as a series of chained conditional branches, or with a lookup
819 <i>; Emulate a conditional br instruction</i>
820 %Val = <a href="#i_cast">cast</a> bool %value to int
821 switch int %Val, label %truedest [int 0, label %falsedest ]
823 <i>; Emulate an unconditional br instruction</i>
824 switch uint 0, label %dest [ ]
826 <i>; Implement a jump table:</i>
827 switch uint %val, label %otherwise [ uint 0, label %onzero
829 uint 2, label %ontwo ]
832 <!-- _______________________________________________________________________ -->
833 <div class="doc_subsubsection"> <a name="i_invoke">'<tt>invoke</tt>'
834 Instruction</a> </div>
835 <div class="doc_text">
837 <pre> <result> = invoke <ptr to function ty> %<function ptr val>(<function args>)<br> to label <normal label> except label <exception label><br></pre>
839 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a
840 specified function, with the possibility of control flow transfer to
841 either the '<tt>normal</tt>' <tt>label</tt> label or the '<tt>exception</tt>'<tt>label</tt>.
842 If the callee function returns with the "<tt><a href="#i_ret">ret</a></tt>"
843 instruction, control flow will return to the "normal" label. If the
844 callee (or any indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
845 instruction, control is interrupted, and continued at the dynamically
846 nearest "except" label.</p>
848 <p>This instruction requires several arguments:</p>
850 <li>'<tt>ptr to function ty</tt>': shall be the signature of the
851 pointer to function value being invoked. In most cases, this is a
852 direct function invocation, but indirect <tt>invoke</tt>s are just as
853 possible, branching off an arbitrary pointer to function value. </li>
854 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer
855 to a function to be invoked. </li>
856 <li>'<tt>function args</tt>': argument list whose types match the
857 function signature argument types. If the function signature indicates
858 the function accepts a variable number of arguments, the extra
859 arguments can be specified. </li>
860 <li>'<tt>normal label</tt>': the label reached when the called
861 function executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
862 <li>'<tt>exception label</tt>': the label reached when a callee
863 returns with the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
866 <p>This instruction is designed to operate as a standard '<tt><a
867 href="#i_call">call</a></tt>' instruction in most regards. The
868 primary difference is that it establishes an association with a label,
869 which is used by the runtime library to unwind the stack.</p>
870 <p>This instruction is used in languages with destructors to ensure
871 that proper cleanup is performed in the case of either a <tt>longjmp</tt>
872 or a thrown exception. Additionally, this is important for
873 implementation of '<tt>catch</tt>' clauses in high-level languages that
876 <pre> %retval = invoke int %Test(int 15)<br> to label %Continue<br> except label %TestCleanup <i>; {int}:retval set</i>
879 <!-- _______________________________________________________________________ -->
880 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
881 Instruction</a> </div>
882 <div class="doc_text">
884 <pre> unwind<br></pre>
886 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing
887 control flow at the first callee in the dynamic call stack which used
888 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the
889 call. This is primarily used to implement exception handling.</p>
891 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current
892 function to immediately halt. The dynamic call stack is then searched
893 for the first <a href="#i_invoke"><tt>invoke</tt></a> instruction on
894 the call stack. Once found, execution continues at the "exceptional"
895 destination block specified by the <tt>invoke</tt> instruction. If
896 there is no <tt>invoke</tt> instruction in the dynamic call chain,
897 undefined behavior results.</p>
899 <!-- ======================================================================= -->
900 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
901 <div class="doc_text">
902 <p>Binary operators are used to do most of the computation in a
903 program. They require two operands, execute an operation on them, and
904 produce a single value. The result value of a binary operator is not
905 necessarily the same type as its operands.</p>
906 <p>There are several different binary operators:</p>
908 <!-- _______________________________________________________________________ -->
909 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
910 Instruction</a> </div>
911 <div class="doc_text">
913 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
916 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
918 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
919 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
920 values. Both arguments must have identical types.</p>
922 <p>The value produced is the integer or floating point sum of the two
925 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
928 <!-- _______________________________________________________________________ -->
929 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
930 Instruction</a> </div>
931 <div class="doc_text">
933 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
936 <p>The '<tt>sub</tt>' instruction returns the difference of its two
938 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
939 instruction present in most other intermediate representations.</p>
941 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
942 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
943 values. Both arguments must have identical types.</p>
945 <p>The value produced is the integer or floating point difference of
946 the two operands.</p>
948 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
949 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
952 <!-- _______________________________________________________________________ -->
953 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
954 Instruction</a> </div>
955 <div class="doc_text">
957 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
960 <p>The '<tt>mul</tt>' instruction returns the product of its two
963 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
964 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
965 values. Both arguments must have identical types.</p>
967 <p>The value produced is the integer or floating point product of the
969 <p>There is no signed vs unsigned multiplication. The appropriate
970 action is taken based on the type of the operand.</p>
972 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
975 <!-- _______________________________________________________________________ -->
976 <div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
977 Instruction</a> </div>
978 <div class="doc_text">
980 <pre> <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
983 <p>The '<tt>div</tt>' instruction returns the quotient of its two
986 <p>The two arguments to the '<tt>div</tt>' instruction must be either <a
987 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
988 values. Both arguments must have identical types.</p>
990 <p>The value produced is the integer or floating point quotient of the
993 <pre> <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
996 <!-- _______________________________________________________________________ -->
997 <div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
998 Instruction</a> </div>
999 <div class="doc_text">
1001 <pre> <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1004 <p>The '<tt>rem</tt>' instruction returns the remainder from the
1005 division of its two operands.</p>
1007 <p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1008 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1009 values. Both arguments must have identical types.</p>
1011 <p>This returns the <i>remainder</i> of a division (where the result
1012 has the same sign as the divisor), not the <i>modulus</i> (where the
1013 result has the same sign as the dividend) of a value. For more
1014 information about the difference, see: <a
1015 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1018 <pre> <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1021 <!-- _______________________________________________________________________ -->
1022 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1023 Instructions</a> </div>
1024 <div class="doc_text">
1026 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1027 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1028 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1029 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1030 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1031 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1034 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1035 value based on a comparison of their two operands.</p>
1037 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1038 be of <a href="#t_firstclass">first class</a> type (it is not possible
1039 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1040 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1043 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1044 value if both operands are equal.<br>
1045 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1046 value if both operands are unequal.<br>
1047 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1048 value if the first operand is less than the second operand.<br>
1049 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1050 value if the first operand is greater than the second operand.<br>
1051 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1052 value if the first operand is less than or equal to the second operand.<br>
1053 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1054 value if the first operand is greater than or equal to the second
1057 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1058 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1059 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1060 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1061 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1062 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1065 <!-- ======================================================================= -->
1066 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1067 Operations</a> </div>
1068 <div class="doc_text">
1069 <p>Bitwise binary operators are used to do various forms of
1070 bit-twiddling in a program. They are generally very efficient
1071 instructions, and can commonly be strength reduced from other
1072 instructions. They require two operands, execute an operation on them,
1073 and produce a single value. The resulting value of the bitwise binary
1074 operators is always the same type as its first operand.</p>
1076 <!-- _______________________________________________________________________ -->
1077 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1078 Instruction</a> </div>
1079 <div class="doc_text">
1081 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1084 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1085 its two operands.</p>
1087 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1088 href="#t_integral">integral</a> values. Both arguments must have
1089 identical types.</p>
1091 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1093 <div style="align: center">
1094 <table border="1" cellspacing="0" cellpadding="4">
1125 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1126 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1127 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1130 <!-- _______________________________________________________________________ -->
1131 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1132 <div class="doc_text">
1134 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1137 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1138 or of its two operands.</p>
1140 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1141 href="#t_integral">integral</a> values. Both arguments must have
1142 identical types.</p>
1144 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1146 <div style="align: center">
1147 <table border="1" cellspacing="0" cellpadding="4">
1178 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1179 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1180 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1183 <!-- _______________________________________________________________________ -->
1184 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1185 Instruction</a> </div>
1186 <div class="doc_text">
1188 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1191 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1192 or of its two operands. The <tt>xor</tt> is used to implement the
1193 "one's complement" operation, which is the "~" operator in C.</p>
1195 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1196 href="#t_integral">integral</a> values. Both arguments must have
1197 identical types.</p>
1199 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1201 <div style="align: center">
1202 <table border="1" cellspacing="0" cellpadding="4">
1234 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1235 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1236 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1237 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1240 <!-- _______________________________________________________________________ -->
1241 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1242 Instruction</a> </div>
1243 <div class="doc_text">
1245 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1248 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1249 the left a specified number of bits.</p>
1251 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
1252 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1255 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
1257 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1258 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1259 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1262 <!-- _______________________________________________________________________ -->
1263 <div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1264 Instruction</a> </div>
1265 <div class="doc_text">
1267 <pre> <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1270 <p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1271 the right a specified number of bits.</p>
1273 <p>The first argument to the '<tt>shr</tt>' instruction must be an <a
1274 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1277 <p>If the first argument is a <a href="#t_signed">signed</a> type, the
1278 most significant bit is duplicated in the newly free'd bit positions.
1279 If the first argument is unsigned, zero bits shall fill the empty
1282 <pre> <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1283 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1284 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1285 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1286 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1289 <!-- ======================================================================= -->
1290 <div class="doc_subsection"> <a name="memoryops">Memory Access
1291 Operations</a></div>
1292 <div class="doc_text">
1293 <p>A key design point of an SSA-based representation is how it
1294 represents memory. In LLVM, no memory locations are in SSA form, which
1295 makes things very simple. This section describes how to read, write,
1296 allocate and free memory in LLVM.</p>
1298 <!-- _______________________________________________________________________ -->
1299 <div class="doc_subsubsection"> <a name="i_malloc">'<tt>malloc</tt>'
1300 Instruction</a> </div>
1301 <div class="doc_text">
1303 <pre> <result> = malloc <type>, uint <NumElements> <i>; yields {type*}:result</i>
1304 <result> = malloc <type> <i>; yields {type*}:result</i>
1307 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
1308 heap and returns a pointer to it.</p>
1310 <p>The '<tt>malloc</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1311 bytes of memory from the operating system and returns a pointer of the
1312 appropriate type to the program. The second form of the instruction is
1313 a shorter version of the first instruction that defaults to allocating
1315 <p>'<tt>type</tt>' must be a sized type.</p>
1317 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
1318 a pointer is returned.</p>
1320 <pre> %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1323 href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1324 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1325 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1328 <!-- _______________________________________________________________________ -->
1329 <div class="doc_subsubsection"> <a name="i_free">'<tt>free</tt>'
1330 Instruction</a> </div>
1331 <div class="doc_text">
1333 <pre> free <type> <value> <i>; yields {void}</i>
1336 <p>The '<tt>free</tt>' instruction returns memory back to the unused
1337 memory heap, to be reallocated in the future.</p>
1340 <p>'<tt>value</tt>' shall be a pointer value that points to a value
1341 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
1344 <p>Access to the memory pointed to by the pointer is not longer defined
1345 after this instruction executes.</p>
1347 <pre> %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1348 free [4 x ubyte]* %array
1351 <!-- _______________________________________________________________________ -->
1352 <div class="doc_subsubsection"> <a name="i_alloca">'<tt>alloca</tt>'
1353 Instruction</a> </div>
1354 <div class="doc_text">
1356 <pre> <result> = alloca <type>, uint <NumElements> <i>; yields {type*}:result</i>
1357 <result> = alloca <type> <i>; yields {type*}:result</i>
1360 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
1361 stack frame of the procedure that is live until the current function
1362 returns to its caller.</p>
1364 <p>The the '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1365 bytes of memory on the runtime stack, returning a pointer of the
1366 appropriate type to the program. The second form of the instruction is
1367 a shorter version of the first that defaults to allocating one element.</p>
1368 <p>'<tt>type</tt>' may be any sized type.</p>
1370 <p>Memory is allocated, a pointer is returned. '<tt>alloca</tt>'d
1371 memory is automatically released when the function returns. The '<tt>alloca</tt>'
1372 instruction is commonly used to represent automatic variables that must
1373 have an address available. When the function returns (either with the <tt><a
1374 href="#i_ret">ret</a></tt> or <tt><a href="#i_invoke">invoke</a></tt>
1375 instructions), the memory is reclaimed.</p>
1377 <pre> %ptr = alloca int <i>; yields {int*}:ptr</i>
1378 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
1381 <!-- _______________________________________________________________________ -->
1382 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
1383 Instruction</a> </div>
1384 <div class="doc_text">
1386 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
1388 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
1390 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
1391 address to load from. The pointer must point to a <a
1392 href="t_firstclass">first class</a> type. If the <tt>load</tt> is
1393 marked as <tt>volatile</tt> then the optimizer is not allowed to modify
1394 the number or order of execution of this <tt>load</tt> with other
1395 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
1398 <p>The location of memory pointed to is loaded.</p>
1400 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1402 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1403 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1406 <!-- _______________________________________________________________________ -->
1407 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
1408 Instruction</a> </div>
1410 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1411 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1414 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
1416 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
1417 to store and an address to store it into. The type of the '<tt><pointer></tt>'
1418 operand must be a pointer to the type of the '<tt><value></tt>'
1419 operand. If the <tt>store</tt> is marked as <tt>volatile</tt> then the
1420 optimizer is not allowed to modify the number or order of execution of
1421 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
1422 href="#i_store">store</a></tt> instructions.</p>
1424 <p>The contents of memory are updated to contain '<tt><value></tt>'
1425 at the location specified by the '<tt><pointer></tt>' operand.</p>
1427 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1429 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1430 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1432 <!-- _______________________________________________________________________ -->
1433 <div class="doc_subsubsection">
1434 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
1437 <div class="doc_text">
1440 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
1446 The '<tt>getelementptr</tt>' instruction is used to get the address of a
1447 subelement of an aggregate data structure.</p>
1451 <p>This instruction takes a list of integer constants that indicate what
1452 elements of the aggregate object to index to. The actual types of the arguments
1453 provided depend on the type of the first pointer argument. The
1454 '<tt>getelementptr</tt>' instruction is used to index down through the type
1455 levels of a structure. When indexing into a structure, only <tt>uint</tt>
1456 integer constants are allowed. When indexing into an array or pointer
1457 <tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
1459 <p>For example, let's consider a C code fragment and how it gets
1460 compiled to LLVM:</p>
1474 int *foo(struct ST *s) {
1475 return &s[1].Z.B[5][13];
1479 <p>The LLVM code generated by the GCC frontend is:</p>
1482 %RT = type { sbyte, [10 x [20 x int]], sbyte }
1483 %ST = type { int, double, %RT }
1485 int* "foo"(%ST* %s) {
1486 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13<br>
1493 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
1494 on the pointer type that is being index into. <a href="t_pointer">Pointer</a>
1495 and <a href="t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
1496 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="t_struct">structure</a>
1497 types require <tt>uint</tt> <b>constants</b>.</p>
1499 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
1500 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
1501 }</tt>' type, a structure. The second index indexes into the third element of
1502 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
1503 sbyte }</tt>' type, another structure. The third index indexes into the second
1504 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
1505 array. The two dimensions of the array are subscripted into, yielding an
1506 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction return a pointer
1507 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
1509 <p>Note that it is perfectly legal to index partially through a
1510 structure, returning a pointer to an inner element. Because of this,
1511 the LLVM code for the given testcase is equivalent to:</p>
1514 int* "foo"(%ST* %s) {
1515 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
1516 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
1517 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
1518 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
1519 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
1525 <i>; yields [12 x ubyte]*:aptr</i>
1526 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
1530 <!-- ======================================================================= -->
1531 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
1532 <div class="doc_text">
1533 <p>The instructions in this category are the "miscellaneous"
1534 instructions, which defy better classification.</p>
1536 <!-- _______________________________________________________________________ -->
1537 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
1538 Instruction</a> </div>
1539 <div class="doc_text">
1541 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
1543 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
1544 the SSA graph representing the function.</p>
1546 <p>The type of the incoming values are specified with the first type
1547 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
1548 as arguments, with one pair for each predecessor basic block of the
1549 current block. Only values of <a href="#t_firstclass">first class</a>
1550 type may be used as the value arguments to the PHI node. Only labels
1551 may be used as the label arguments.</p>
1552 <p>There must be no non-phi instructions between the start of a basic
1553 block and the PHI instructions: i.e. PHI instructions must be first in
1556 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
1557 value specified by the parameter, depending on which basic block we
1558 came from in the last <a href="#terminators">terminator</a> instruction.</p>
1560 <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>
1563 <!-- _______________________________________________________________________ -->
1564 <div class="doc_subsubsection">
1565 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
1568 <div class="doc_text">
1573 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
1579 The '<tt>cast</tt>' instruction is used as the primitive means to convert
1580 integers to floating point, change data type sizes, and break type safety (by
1588 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
1589 class value, and a type to cast it to, which must also be a <a
1590 href="#t_firstclass">first class</a> type.
1596 This instruction follows the C rules for explicit casts when determining how the
1597 data being cast must change to fit in its new container.
1601 When casting to bool, any value that would be considered true in the context of
1602 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
1603 all else are '<tt>false</tt>'.
1607 When extending an integral value from a type of one signness to another (for
1608 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
1609 <b>source</b> value is signed, and zero-extended if the source value is
1610 unsigned. <tt>bool</tt> values are always zero extended into either zero or
1617 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
1618 %Y = cast int 123 to bool <i>; yields bool:true</i>
1622 <!-- _______________________________________________________________________ -->
1623 <div class="doc_subsubsection">
1624 <a name="i_select">'<tt>select</tt>' Instruction</a>
1627 <div class="doc_text">
1632 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
1638 The '<tt>select</tt>' instruction is used to choose one value based on a
1639 condition, without branching.
1646 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.
1652 If the boolean condition evaluates to true, the instruction returns the first
1653 value argument, otherwise it returns the second value argument.
1659 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
1667 <!-- _______________________________________________________________________ -->
1668 <div class="doc_subsubsection"> <a name="i_call">'<tt>call</tt>'
1669 Instruction</a> </div>
1670 <div class="doc_text">
1672 <pre> <result> = call <ty>* <fnptrval>(<param list>)<br></pre>
1674 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
1676 <p>This instruction requires several arguments:</p>
1679 <p>'<tt>ty</tt>': shall be the signature of the pointer to function
1680 value being invoked. The argument types must match the types implied
1681 by this signature.</p>
1684 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a
1685 function to be invoked. In most cases, this is a direct function
1686 invocation, but indirect <tt>call</tt>s are just as possible,
1687 calling an arbitrary pointer to function values.</p>
1690 <p>'<tt>function args</tt>': argument list whose types match the
1691 function signature argument types. If the function signature
1692 indicates the function accepts a variable number of arguments, the
1693 extra arguments can be specified.</p>
1697 <p>The '<tt>call</tt>' instruction is used to cause control flow to
1698 transfer to a specified function, with its incoming arguments bound to
1699 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
1700 instruction in the called function, control flow continues with the
1701 instruction after the function call, and the return value of the
1702 function is bound to the result argument. This is a simpler case of
1703 the <a href="#i_invoke">invoke</a> instruction.</p>
1705 <pre> %retval = call int %test(int %argc)<br> call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);<br></pre>
1707 <!-- _______________________________________________________________________ -->
1708 <div class="doc_subsubsection"> <a name="i_vanext">'<tt>vanext</tt>'
1709 Instruction</a> </div>
1710 <div class="doc_text">
1712 <pre> <resultarglist> = vanext <va_list> <arglist>, <argty><br></pre>
1714 <p>The '<tt>vanext</tt>' instruction is used to access arguments passed
1715 through the "variable argument" area of a function call. It is used to
1716 implement the <tt>va_arg</tt> macro in C.</p>
1718 <p>This instruction takes a <tt>valist</tt> value and the type of the
1719 argument. It returns another <tt>valist</tt>.</p>
1721 <p>The '<tt>vanext</tt>' instruction advances the specified <tt>valist</tt>
1722 past an argument of the specified type. In conjunction with the <a
1723 href="#i_vaarg"><tt>vaarg</tt></a> instruction, it is used to implement
1724 the <tt>va_arg</tt> macro available in C. For more information, see
1725 the variable argument handling <a href="#int_varargs">Intrinsic
1727 <p>It is legal for this instruction to be called in a function which
1728 does not take a variable number of arguments, for example, the <tt>vfprintf</tt>
1730 <p><tt>vanext</tt> is an LLVM instruction instead of an <a
1731 href="#intrinsics">intrinsic function</a> because it takes an type as
1734 <p>See the <a href="#int_varargs">variable argument processing</a>
1737 <!-- _______________________________________________________________________ -->
1738 <div class="doc_subsubsection"> <a name="i_vaarg">'<tt>vaarg</tt>'
1739 Instruction</a> </div>
1740 <div class="doc_text">
1742 <pre> <resultval> = vaarg <va_list> <arglist>, <argty><br></pre>
1744 <p>The '<tt>vaarg</tt>' instruction is used to access arguments passed
1745 through the "variable argument" area of a function call. It is used to
1746 implement the <tt>va_arg</tt> macro in C.</p>
1748 <p>This instruction takes a <tt>valist</tt> value and the type of the
1749 argument. It returns a value of the specified argument type.</p>
1751 <p>The '<tt>vaarg</tt>' instruction loads an argument of the specified
1752 type from the specified <tt>va_list</tt>. In conjunction with the <a
1753 href="#i_vanext"><tt>vanext</tt></a> instruction, it is used to
1754 implement the <tt>va_arg</tt> macro available in C. For more
1755 information, see the variable argument handling <a href="#int_varargs">Intrinsic
1757 <p>It is legal for this instruction to be called in a function which
1758 does not take a variable number of arguments, for example, the <tt>vfprintf</tt>
1760 <p><tt>vaarg</tt> is an LLVM instruction instead of an <a
1761 href="#intrinsics">intrinsic function</a> because it takes an type as
1764 <p>See the <a href="#int_varargs">variable argument processing</a>
1768 <!-- *********************************************************************** -->
1769 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
1770 <!-- *********************************************************************** -->
1772 <div class="doc_text">
1774 <p>LLVM supports the notion of an "intrinsic function". These functions have
1775 well known names and semantics, and are required to follow certain
1776 restrictions. Overall, these instructions represent an extension mechanism for
1777 the LLVM language that does not require changing all of the transformations in
1778 LLVM to add to the language (or the bytecode reader/writer, the parser,
1781 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix, this
1782 prefix is reserved in LLVM for intrinsic names, thus functions may not be named
1783 this. Intrinsic functions must always be external functions: you cannot define
1784 the body of intrinsic functions. Intrinsic functions may only be used in call
1785 or invoke instructions: it is illegal to take the address of an intrinsic
1786 function. Additionally, because intrinsic functions are part of the LLVM
1787 language, it is required that they all be documented here if any are added.</p>
1791 Adding an intrinsic to LLVM is straight-forward if it is possible to express the
1792 concept in LLVM directly (ie, code generator support is not _required_). To do
1793 this, extend the default implementation of the IntrinsicLowering class to handle
1794 the intrinsic. Code generators use this class to lower intrinsics they do not
1795 understand to raw LLVM instructions that they do.
1800 <!-- ======================================================================= -->
1801 <div class="doc_subsection">
1802 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
1805 <div class="doc_text">
1806 <p>Variable argument support is defined in LLVM with the <a
1807 href="#i_vanext"><tt>vanext</tt></a> instruction and these three
1808 intrinsic functions. These functions are related to the similarly
1809 named macros defined in the <tt><stdarg.h></tt> header file.</p>
1810 <p>All of these functions operate on arguments that use a
1811 target-specific value type "<tt>va_list</tt>". The LLVM assembly
1812 language reference manual does not define what this type is, so all
1813 transformations should be prepared to handle intrinsics with any type
1815 <p>This example shows how the <a href="#i_vanext"><tt>vanext</tt></a>
1816 instruction and the variable argument handling intrinsic functions are
1819 int %test(int %X, ...) {
1820 ; Initialize variable argument processing
1821 %ap = call sbyte* %<a href="#i_va_start">llvm.va_start</a>()
1823 ; Read a single integer argument
1824 %tmp = vaarg sbyte* %ap, int
1826 ; Advance to the next argument
1827 %ap2 = vanext sbyte* %ap, int
1829 ; Demonstrate usage of llvm.va_copy and llvm.va_end
1830 %aq = call sbyte* %<a href="#i_va_copy">llvm.va_copy</a>(sbyte* %ap2)
1831 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %aq)
1833 ; Stop processing of arguments.
1834 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %ap2)
1840 <!-- _______________________________________________________________________ -->
1841 <div class="doc_subsubsection">
1842 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
1846 <div class="doc_text">
1848 <pre> call va_list ()* %llvm.va_start()<br></pre>
1850 <p>The '<tt>llvm.va_start</tt>' intrinsic returns a new <tt><arglist></tt>
1851 for subsequent use by the variable argument intrinsics.</p>
1853 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
1854 macro available in C. In a target-dependent way, it initializes and
1855 returns a <tt>va_list</tt> element, so that the next <tt>vaarg</tt>
1856 will produce the first variable argument passed to the function. Unlike
1857 the C <tt>va_start</tt> macro, this intrinsic does not need to know the
1858 last argument of the function, the compiler can figure that out.</p>
1859 <p>Note that this intrinsic function is only legal to be called from
1860 within the body of a variable argument function.</p>
1863 <!-- _______________________________________________________________________ -->
1864 <div class="doc_subsubsection">
1865 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
1868 <div class="doc_text">
1870 <pre> call void (va_list)* %llvm.va_end(va_list <arglist>)<br></pre>
1872 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
1873 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
1874 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
1876 <p>The argument is a <tt>va_list</tt> to destroy.</p>
1878 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
1879 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
1880 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
1881 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
1882 with calls to <tt>llvm.va_end</tt>.</p>
1885 <!-- _______________________________________________________________________ -->
1886 <div class="doc_subsubsection">
1887 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
1890 <div class="doc_text">
1892 <pre> call va_list (va_list)* %llvm.va_copy(va_list <destarglist>)<br></pre>
1894 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument
1895 position from the source argument list to the destination argument list.</p>
1897 <p>The argument is the <tt>va_list</tt> to copy.</p>
1899 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
1900 macro available in C. In a target-dependent way, it copies the source <tt>va_list</tt>
1901 element into the returned list. This intrinsic is necessary because the <tt><a
1902 href="i_va_start">llvm.va_start</a></tt> intrinsic may be arbitrarily
1903 complex and require memory allocation, for example.</p>
1906 <!-- ======================================================================= -->
1907 <div class="doc_subsection">
1908 <a name="int_codegen">Code Generator Intrinsics</a>
1911 <div class="doc_text">
1913 These intrinsics are provided by LLVM to expose special features that may only
1914 be implemented with code generator support.
1919 <!-- _______________________________________________________________________ -->
1920 <div class="doc_subsubsection">
1921 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
1924 <div class="doc_text">
1928 call void* ()* %llvm.returnaddress(uint <level>)
1934 The '<tt>llvm.returnaddress</tt>' intrinsic returns a target-specific value
1935 indicating the return address of the current function or one of its callers.
1941 The argument to this intrinsic indicates which function to return the address
1942 for. Zero indicates the calling function, one indicates its caller, etc. The
1943 argument is <b>required</b> to be a constant integer value.
1949 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
1950 the return address of the specified call frame, or zero if it cannot be
1951 identified. The value returned by this intrinsic is likely to be incorrect or 0
1952 for arguments other than zero, so it should only be used for debugging purposes.
1956 Note that calling this intrinsic does not prevent function inlining or other
1957 aggressive transformations, so the value returned may not that of the obvious
1958 source-language caller.
1963 <!-- _______________________________________________________________________ -->
1964 <div class="doc_subsubsection">
1965 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
1968 <div class="doc_text">
1972 call void* ()* %llvm.frameaddress(uint <level>)
1978 The '<tt>llvm.frameaddress</tt>' intrinsic returns the target-specific frame
1979 pointer value for the specified stack frame.
1985 The argument to this intrinsic indicates which function to return the frame
1986 pointer for. Zero indicates the calling function, one indicates its caller,
1987 etc. The argument is <b>required</b> to be a constant integer value.
1993 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
1994 the frame address of the specified call frame, or zero if it cannot be
1995 identified. The value returned by this intrinsic is likely to be incorrect or 0
1996 for arguments other than zero, so it should only be used for debugging purposes.
2000 Note that calling this intrinsic does not prevent function inlining or other
2001 aggressive transformations, so the value returned may not that of the obvious
2002 source-language caller.
2006 <!-- ======================================================================= -->
2007 <div class="doc_subsection">
2008 <a name="int_os">Operating System Intrinsics</a>
2011 <div class="doc_text">
2013 These intrinsics are provided by LLVM to support the implementation of
2014 operating system level code.
2019 <!-- _______________________________________________________________________ -->
2020 <div class="doc_subsubsection">
2021 <a name="i_readport">'<tt>llvm.readport</tt>' Intrinsic</a>
2024 <div class="doc_text">
2028 call <integer type> (<integer type>)* %llvm.readport (<integer type> <address>)
2034 The '<tt>llvm.readport</tt>' intrinsic reads data from the specified hardware
2041 The argument to this intrinsic indicates the hardware I/O address from which
2042 to read the data. The address is in the hardware I/O address namespace (as
2043 opposed to being a memory location for memory mapped I/O).
2049 The '<tt>llvm.readport</tt>' intrinsic reads data from the hardware I/O port
2050 specified by <i>address</i> and returns the value. The address and return
2051 value must be integers, but the size is dependent upon the platform upon which
2052 the program is code generated. For example, on x86, the address must be an
2053 unsigned 16 bit value, and the return value must be 8, 16, or 32 bits.
2058 <!-- _______________________________________________________________________ -->
2059 <div class="doc_subsubsection">
2060 <a name="i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a>
2063 <div class="doc_text">
2067 call void (<integer type>, <integer type>)* %llvm.writeport (<integer type> <value>, <integer type> <address>)
2073 The '<tt>llvm.writeport</tt>' intrinsic writes data to the specified hardware
2080 The first argument is the value to write to the I/O port.
2084 The second argument indicates the hardware I/O address to which data should be
2085 written. The address is in the hardware I/O address namespace (as opposed to
2086 being a memory location for memory mapped I/O).
2092 The '<tt>llvm.writeport</tt>' intrinsic writes <i>value</i> to the I/O port
2093 specified by <i>address</i>. The address and value must be integers, but the
2094 size is dependent upon the platform upon which the program is code generated.
2095 For example, on x86, the address must be an unsigned 16 bit value, and the
2096 value written must be 8, 16, or 32 bits in length.
2101 <!-- _______________________________________________________________________ -->
2102 <div class="doc_subsubsection">
2103 <a name="i_readio">'<tt>llvm.readio</tt>' Intrinsic</a>
2106 <div class="doc_text">
2110 call <result> (<ty>*)* %llvm.readio (<ty> * <pointer>)
2116 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2123 The argument to this intrinsic is a pointer indicating the memory address from
2124 which to read the data. The data must be a
2125 <a href="#t_firstclass">first class</a> type.
2131 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2132 location specified by <i>pointer</i> and returns the value. The argument must
2133 be a pointer, and the return value must be a
2134 <a href="#t_firstclass">first class</a> type. However, certain architectures
2135 may not support I/O on all first class types. For example, 32 bit processors
2136 may only support I/O on data types that are 32 bits or less.
2140 This intrinsic enforces an in-order memory model for llvm.readio and
2141 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2142 scheduled processors may execute loads and stores out of order, re-ordering at
2143 run time accesses to memory mapped I/O registers. Using these intrinsics
2144 ensures that accesses to memory mapped I/O registers occur in program order.
2149 <!-- _______________________________________________________________________ -->
2150 <div class="doc_subsubsection">
2151 <a name="i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a>
2154 <div class="doc_text">
2158 call void (<ty1>, <ty2>*)* %llvm.writeio (<ty1> <value>, <ty2> * <pointer>)
2164 The '<tt>llvm.writeio</tt>' intrinsic writes data to the specified memory
2171 The first argument is the value to write to the memory mapped I/O location.
2172 The second argument is a pointer indicating the memory address to which the
2173 data should be written.
2179 The '<tt>llvm.writeio</tt>' intrinsic writes <i>value</i> to the memory mapped
2180 I/O address specified by <i>pointer</i>. The value must be a
2181 <a href="#t_firstclass">first class</a> type. However, certain architectures
2182 may not support I/O on all first class types. For example, 32 bit processors
2183 may only support I/O on data types that are 32 bits or less.
2187 This intrinsic enforces an in-order memory model for llvm.readio and
2188 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2189 scheduled processors may execute loads and stores out of order, re-ordering at
2190 run time accesses to memory mapped I/O registers. Using these intrinsics
2191 ensures that accesses to memory mapped I/O registers occur in program order.
2197 <!-- ======================================================================= -->
2198 <div class="doc_subsection">
2199 <a name="int_libc">Standard C Library Intrinsics</a>
2202 <div class="doc_text">
2204 LLVM provides intrinsics for a few important standard C library functions.
2205 These intrinsics allow source-language front-ends to pass information about the
2206 alignment of the pointer arguments to the code generator, providing opportunity
2207 for more efficient code generation.
2212 <!-- _______________________________________________________________________ -->
2213 <div class="doc_subsubsection">
2214 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
2217 <div class="doc_text">
2221 call void (sbyte*, sbyte*, uint, uint)* %llvm.memcpy(sbyte* <dest>, sbyte* <src>,
2222 uint <len>, uint <align>)
2228 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2229 location to the destination location.
2233 Note that, unlike the standard libc function, the <tt>llvm.memcpy</tt> intrinsic
2234 does not return a value, and takes an extra alignment argument.
2240 The first argument is a pointer to the destination, the second is a pointer to
2241 the source. The third argument is an (arbitrarily sized) integer argument
2242 specifying the number of bytes to copy, and the fourth argument is the alignment
2243 of the source and destination locations.
2247 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2248 the caller guarantees that the size of the copy is a multiple of the alignment
2249 and that both the source and destination pointers are aligned to that boundary.
2255 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2256 location to the destination location, which are not allowed to overlap. It
2257 copies "len" bytes of memory over. If the argument is known to be aligned to
2258 some boundary, this can be specified as the fourth argument, otherwise it should
2264 <!-- _______________________________________________________________________ -->
2265 <div class="doc_subsubsection">
2266 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
2269 <div class="doc_text">
2273 call void (sbyte*, sbyte*, uint, uint)* %llvm.memmove(sbyte* <dest>, sbyte* <src>,
2274 uint <len>, uint <align>)
2280 The '<tt>llvm.memmove</tt>' intrinsic moves a block of memory from the source
2281 location to the destination location. It is similar to the '<tt>llvm.memcpy</tt>'
2282 intrinsic but allows the two memory locations to overlap.
2286 Note that, unlike the standard libc function, the <tt>llvm.memmove</tt> intrinsic
2287 does not return a value, and takes an extra alignment argument.
2293 The first argument is a pointer to the destination, the second is a pointer to
2294 the source. The third argument is an (arbitrarily sized) integer argument
2295 specifying the number of bytes to copy, and the fourth argument is the alignment
2296 of the source and destination locations.
2300 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2301 the caller guarantees that the size of the copy is a multiple of the alignment
2302 and that both the source and destination pointers are aligned to that boundary.
2308 The '<tt>llvm.memmove</tt>' intrinsic copies a block of memory from the source
2309 location to the destination location, which may overlap. It
2310 copies "len" bytes of memory over. If the argument is known to be aligned to
2311 some boundary, this can be specified as the fourth argument, otherwise it should
2317 <!-- _______________________________________________________________________ -->
2318 <div class="doc_subsubsection">
2319 <a name="i_memset">'<tt>llvm.memset</tt>' Intrinsic</a>
2322 <div class="doc_text">
2326 call void (sbyte*, ubyte, uint, uint)* %llvm.memset(sbyte* <dest>, ubyte <val>,
2327 uint <len>, uint <align>)
2333 The '<tt>llvm.memset</tt>' intrinsic fills a block of memory with a particular
2338 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
2339 does not return a value, and takes an extra alignment argument.
2345 The first argument is a pointer to the destination to fill, the second is the
2346 byte value to fill it with, the third argument is an (arbitrarily sized) integer
2347 argument specifying the number of bytes to fill, and the fourth argument is the
2348 known alignment of destination location.
2352 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2353 the caller guarantees that the size of the copy is a multiple of the alignment
2354 and that the destination pointer is aligned to that boundary.
2360 The '<tt>llvm.memset</tt>' intrinsic fills "len" bytes of memory starting at the
2361 destination location. If the argument is known to be aligned to some boundary,
2362 this can be specified as the fourth argument, otherwise it should be set to 0 or
2368 <!-- ======================================================================= -->
2369 <div class="doc_subsection">
2370 <a name="int_debugger">Debugger Intrinsics</a>
2373 <div class="doc_text">
2375 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
2376 are described in the <a
2377 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
2378 Debugging</a> document.
2383 <!-- *********************************************************************** -->
2386 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2387 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
2388 <a href="http://validator.w3.org/check/referer"><img
2389 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
2391 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
2392 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
2393 Last modified: $Date$