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 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
8 <meta name="description"
9 content="LLVM Assembly Language Reference Manual.">
10 <link rel="stylesheet" href="llvm.css" type="text/css">
15 <div class="doc_title"> LLVM Language Reference Manual </div>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
20 <li><a href="#typesystem">Type System</a>
22 <li><a href="#t_primitive">Primitive Types</a>
24 <li><a href="#t_classifications">Type Classifications</a></li>
27 <li><a href="#t_derived">Derived Types</a>
29 <li><a href="#t_array">Array Type</a></li>
30 <li><a href="#t_function">Function Type</a></li>
31 <li><a href="#t_pointer">Pointer Type</a></li>
32 <li><a href="#t_struct">Structure Type</a></li>
33 <li><a href="#t_packed">Packed Type</a></li>
38 <li><a href="#highlevel">High Level Structure</a>
40 <li><a href="#modulestructure">Module Structure</a></li>
41 <li><a href="#globalvars">Global Variables</a></li>
42 <li><a href="#functionstructure">Function Structure</a></li>
45 <li><a href="#instref">Instruction Reference</a>
47 <li><a href="#terminators">Terminator Instructions</a>
49 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
50 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
51 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
52 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
53 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
54 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
57 <li><a href="#binaryops">Binary Operations</a>
59 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
60 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
61 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
62 <li><a href="#i_div">'<tt>div</tt>' Instruction</a></li>
63 <li><a href="#i_rem">'<tt>rem</tt>' Instruction</a></li>
64 <li><a href="#i_setcc">'<tt>set<i>cc</i></tt>' Instructions</a></li>
67 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
69 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
70 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
71 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
72 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
73 <li><a href="#i_shr">'<tt>shr</tt>' Instruction</a></li>
76 <li><a href="#memoryops">Memory Access Operations</a>
78 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
79 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
80 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
81 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
82 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
83 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
86 <li><a href="#otherops">Other Operations</a>
88 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
89 <li><a href="#i_cast">'<tt>cast .. to</tt>' Instruction</a></li>
90 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
91 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
92 <li><a href="#i_vanext">'<tt>vanext</tt>' Instruction</a></li>
93 <li><a href="#i_vaarg">'<tt>vaarg</tt>' Instruction</a></li>
98 <li><a href="#intrinsics">Intrinsic Functions</a>
100 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
102 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
103 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
104 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
107 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
109 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
110 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
111 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
114 <li><a href="#int_codegen">Code Generator Intrinsics</a>
116 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
117 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
120 <li><a href="#int_os">Operating System Intrinsics</a>
122 <li><a href="#i_readport">'<tt>llvm.readport</tt>' Intrinsic</a></li>
123 <li><a href="#i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a></li>
124 <li><a href="#i_readio">'<tt>llvm.readio</tt>' Intrinsic</a></li>
125 <li><a href="#i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a></li>
127 <li><a href="#int_libc">Standard C Library Intrinsics</a>
129 <li><a href="#i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a></li>
130 <li><a href="#i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a></li>
131 <li><a href="#i_memset">'<tt>llvm.memset</tt>' Intrinsic</a></li>
132 <li><a href="#i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a></li>
135 <li><a href="#int_debugger">Debugger intrinsics</a></li>
140 <div class="doc_author">
141 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
142 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
145 <!-- *********************************************************************** -->
146 <div class="doc_section"> <a name="abstract">Abstract </a></div>
147 <!-- *********************************************************************** -->
149 <div class="doc_text">
150 <p>This document is a reference manual for the LLVM assembly language.
151 LLVM is an SSA based representation that provides type safety,
152 low-level operations, flexibility, and the capability of representing
153 'all' high-level languages cleanly. It is the common code
154 representation used throughout all phases of the LLVM compilation
158 <!-- *********************************************************************** -->
159 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
160 <!-- *********************************************************************** -->
162 <div class="doc_text">
164 <p>The LLVM code representation is designed to be used in three
165 different forms: as an in-memory compiler IR, as an on-disk bytecode
166 representation (suitable for fast loading by a Just-In-Time compiler),
167 and as a human readable assembly language representation. This allows
168 LLVM to provide a powerful intermediate representation for efficient
169 compiler transformations and analysis, while providing a natural means
170 to debug and visualize the transformations. The three different forms
171 of LLVM are all equivalent. This document describes the human readable
172 representation and notation.</p>
174 <p>The LLVM representation aims to be a light-weight and low-level
175 while being expressive, typed, and extensible at the same time. It
176 aims to be a "universal IR" of sorts, by being at a low enough level
177 that high-level ideas may be cleanly mapped to it (similar to how
178 microprocessors are "universal IR's", allowing many source languages to
179 be mapped to them). By providing type information, LLVM can be used as
180 the target of optimizations: for example, through pointer analysis, it
181 can be proven that a C automatic variable is never accessed outside of
182 the current function... allowing it to be promoted to a simple SSA
183 value instead of a memory location.</p>
187 <!-- _______________________________________________________________________ -->
188 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
190 <div class="doc_text">
192 <p>It is important to note that this document describes 'well formed'
193 LLVM assembly language. There is a difference between what the parser
194 accepts and what is considered 'well formed'. For example, the
195 following instruction is syntactically okay, but not well formed:</p>
198 %x = <a href="#i_add">add</a> int 1, %x
201 <p>...because the definition of <tt>%x</tt> does not dominate all of
202 its uses. The LLVM infrastructure provides a verification pass that may
203 be used to verify that an LLVM module is well formed. This pass is
204 automatically run by the parser after parsing input assembly, and by
205 the optimizer before it outputs bytecode. The violations pointed out
206 by the verifier pass indicate bugs in transformation passes or input to
209 <!-- Describe the typesetting conventions here. --> </div>
211 <!-- *********************************************************************** -->
212 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
213 <!-- *********************************************************************** -->
215 <div class="doc_text">
217 <p>LLVM uses three different forms of identifiers, for different
221 <li>Numeric constants are represented as you would expect: 12, -3
222 123.421, etc. Floating point constants have an optional hexadecimal
224 <li>Named values are represented as a string of characters with a '%'
225 prefix. For example, %foo, %DivisionByZero,
226 %a.really.long.identifier. The actual regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
227 Identifiers which require other characters in their names can be
228 surrounded with quotes. In this way, anything except a <tt>"</tt>
229 character can be used in a name.</li>
230 <li>Unnamed values are represented as an unsigned numeric value with
231 a '%' prefix. For example, %12, %2, %44.</li>
233 <p>LLVM requires that values start with a '%' sign for two reasons:
234 Compilers don't need to worry about name clashes with reserved words,
235 and the set of reserved words may be expanded in the future without
236 penalty. Additionally, unnamed identifiers allow a compiler to quickly
237 come up with a temporary variable without having to avoid symbol table
239 <p>Reserved words in LLVM are very similar to reserved words in other
240 languages. There are keywords for different opcodes ('<tt><a
241 href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
242 href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
243 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>',
244 etc...), and others. These reserved words cannot conflict with
245 variable names, because none of them start with a '%' character.</p>
246 <p>Here is an example of LLVM code to multiply the integer variable '<tt>%X</tt>'
249 <pre> %result = <a href="#i_mul">mul</a> uint %X, 8<br></pre>
250 <p>After strength reduction:</p>
251 <pre> %result = <a href="#i_shl">shl</a> uint %X, ubyte 3<br></pre>
252 <p>And the hard way:</p>
253 <pre> <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
255 href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
257 href="#i_add">add</a> uint %1, %1<br></pre>
258 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
259 important lexical features of LLVM:</p>
261 <li>Comments are delimited with a '<tt>;</tt>' and go until the end
263 <li>Unnamed temporaries are created when the result of a computation
264 is not assigned to a named value.</li>
265 <li>Unnamed temporaries are numbered sequentially</li>
267 <p>...and it also show a convention that we follow in this document.
268 When demonstrating instructions, we will follow an instruction with a
269 comment that defines the type and name of value produced. Comments are
270 shown in italic text.</p>
271 <p>The one non-intuitive notation for constants is the optional
272 hexidecimal form of floating point constants. For example, the form '<tt>double
273 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
274 4.5e+15</tt>' which is also supported by the parser. The only time
275 hexadecimal floating point constants are useful (and the only time that
276 they are generated by the disassembler) is when an FP constant has to
277 be emitted that is not representable as a decimal floating point number
278 exactly. For example, NaN's, infinities, and other special cases are
279 represented in their IEEE hexadecimal format so that assembly and
280 disassembly do not cause any bits to change in the constants.</p>
282 <!-- *********************************************************************** -->
283 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
284 <!-- *********************************************************************** -->
285 <div class="doc_text">
286 <p>The LLVM type system is one of the most important features of the
287 intermediate representation. Being typed enables a number of
288 optimizations to be performed on the IR directly, without having to do
289 extra analyses on the side before the transformation. A strong type
290 system makes it easier to read the generated code and enables novel
291 analyses and transformations that are not feasible to perform on normal
292 three address code representations.</p>
293 <!-- The written form for the type system was heavily influenced by the
294 syntactic problems with types in the C language<sup><a
295 href="#rw_stroustrup">1</a></sup>.<p> --> </div>
296 <!-- ======================================================================= -->
297 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
298 <div class="doc_text">
299 <p>The primitive types are the fundamental building blocks of the LLVM
300 system. The current set of primitive types are as follows:</p>
302 <table border="0" style="align: center">
306 <table border="1" cellspacing="0" cellpadding="4" style="align: center">
309 <td><tt>void</tt></td>
313 <td><tt>ubyte</tt></td>
314 <td>Unsigned 8 bit value</td>
317 <td><tt>ushort</tt></td>
318 <td>Unsigned 16 bit value</td>
321 <td><tt>uint</tt></td>
322 <td>Unsigned 32 bit value</td>
325 <td><tt>ulong</tt></td>
326 <td>Unsigned 64 bit value</td>
329 <td><tt>float</tt></td>
330 <td>32 bit floating point value</td>
333 <td><tt>label</tt></td>
334 <td>Branch destination</td>
340 <table border="1" cellspacing="0" cellpadding="4">
343 <td><tt>bool</tt></td>
344 <td>True or False value</td>
347 <td><tt>sbyte</tt></td>
348 <td>Signed 8 bit value</td>
351 <td><tt>short</tt></td>
352 <td>Signed 16 bit value</td>
355 <td><tt>int</tt></td>
356 <td>Signed 32 bit value</td>
359 <td><tt>long</tt></td>
360 <td>Signed 64 bit value</td>
363 <td><tt>double</tt></td>
364 <td>64 bit floating point value</td>
374 <!-- _______________________________________________________________________ -->
375 <div class="doc_subsubsection"> <a name="t_classifications">Type
376 Classifications</a> </div>
377 <div class="doc_text">
378 <p>These different primitive types fall into a few useful
381 <table border="1" cellspacing="0" cellpadding="4">
384 <td><a name="t_signed">signed</a></td>
385 <td><tt>sbyte, short, int, long, float, double</tt></td>
388 <td><a name="t_unsigned">unsigned</a></td>
389 <td><tt>ubyte, ushort, uint, ulong</tt></td>
392 <td><a name="t_integer">integer</a></td>
393 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
396 <td><a name="t_integral">integral</a></td>
397 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt>
401 <td><a name="t_floating">floating point</a></td>
402 <td><tt>float, double</tt></td>
405 <td><a name="t_firstclass">first class</a></td>
406 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long,<br>
407 float, double, <a href="#t_pointer">pointer</a>,
408 <a href="#t_packed">packed</a></tt></td>
413 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
414 most important. Values of these types are the only ones which can be
415 produced by instructions, passed as arguments, or used as operands to
416 instructions. This means that all structures and arrays must be
417 manipulated either by pointer or by component.</p>
419 <!-- ======================================================================= -->
420 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
421 <div class="doc_text">
422 <p>The real power in LLVM comes from the derived types in the system.
423 This is what allows a programmer to represent arrays, functions,
424 pointers, and other useful types. Note that these derived types may be
425 recursive: For example, it is possible to have a two dimensional array.</p>
427 <!-- _______________________________________________________________________ -->
428 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
429 <div class="doc_text">
431 <p>The array type is a very simple derived type that arranges elements
432 sequentially in memory. The array type requires a size (number of
433 elements) and an underlying data type.</p>
435 <pre> [<# elements> x <elementtype>]<br></pre>
436 <p>The number of elements is a constant integer value, elementtype may
437 be any type with a size.</p>
439 <p> <tt>[40 x int ]</tt>: Array of 40 integer values.<br>
440 <tt>[41 x int ]</tt>: Array of 41 integer values.<br>
441 <tt>[40 x uint]</tt>: Array of 40 unsigned integer values.</p>
443 <p>Here are some examples of multidimensional arrays:</p>
445 <table border="0" cellpadding="0" cellspacing="0">
448 <td><tt>[3 x [4 x int]]</tt></td>
449 <td>: 3x4 array integer values.</td>
452 <td><tt>[12 x [10 x float]]</tt></td>
453 <td>: 12x10 array of single precision floating point values.</td>
456 <td><tt>[2 x [3 x [4 x uint]]]</tt></td>
457 <td>: 2x3x4 array of unsigned integer values.</td>
463 <!-- _______________________________________________________________________ -->
464 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
465 <div class="doc_text">
467 <p>The function type can be thought of as a function signature. It
468 consists of a return type and a list of formal parameter types.
469 Function types are usually used to build virtual function tables
470 (which are structures of pointers to functions), for indirect function
471 calls, and when defining a function.</p>
473 The return type of a function type cannot be an aggregate type.
476 <pre> <returntype> (<parameter list>)<br></pre>
477 <p>Where '<tt><parameter list></tt>' is a comma-separated list of type
478 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
479 which indicates that the function takes a variable number of arguments.
480 Variable argument functions can access their arguments with the <a
481 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
484 <table border="0" cellpadding="0" cellspacing="0">
487 <td><tt>int (int)</tt></td>
488 <td>: function taking an <tt>int</tt>, returning an <tt>int</tt></td>
491 <td><tt>float (int, int *) *</tt></td>
492 <td>: <a href="#t_pointer">Pointer</a> to a function that takes an
493 <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
494 returning <tt>float</tt>.</td>
497 <td><tt>int (sbyte *, ...)</tt></td>
498 <td>: A vararg function that takes at least one <a
499 href="#t_pointer">pointer</a> to <tt>sbyte</tt> (signed char in C), which
500 returns an integer. This is the signature for <tt>printf</tt> in
507 <!-- _______________________________________________________________________ -->
508 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
509 <div class="doc_text">
511 <p>The structure type is used to represent a collection of data members
512 together in memory. The packing of the field types is defined to match
513 the ABI of the underlying processor. The elements of a structure may
514 be any type that has a size.</p>
515 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
516 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
517 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
520 <pre> { <type list> }<br></pre>
523 <table border="0" cellpadding="0" cellspacing="0">
526 <td><tt>{ int, int, int }</tt></td>
527 <td>: a triple of three <tt>int</tt> values</td>
530 <td><tt>{ float, int (int) * }</tt></td>
531 <td>: A pair, where the first element is a <tt>float</tt> and the second
532 element is a <a href="#t_pointer">pointer</a> to a <a
533 href="#t_function">function</a> that takes an <tt>int</tt>, returning an
540 <!-- _______________________________________________________________________ -->
541 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
542 <div class="doc_text">
544 <p>As in many languages, the pointer type represents a pointer or
545 reference to another object, which must live in memory.</p>
547 <pre> <type> *<br></pre>
550 <table border="0" cellpadding="0" cellspacing="0">
553 <td><tt>[4x int]*</tt></td>
554 <td>: <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
555 four <tt>int</tt> values</td>
558 <td><tt>int (int *) *</tt></td>
559 <td>: A <a href="#t_pointer">pointer</a> to a <a
560 href="#t_function">function</a> that takes an <tt>int</tt>, returning an
566 <!-- _______________________________________________________________________ -->
567 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
568 <div class="doc_text">
570 <p>A packed type is a simple derived type that represents a vector
571 of elements. Packed types are used when multiple primitive data
572 are operated in parallel using a single instruction (SIMD).
573 A packed type requires a size (number of
574 elements) and an underlying primitive data type. Packed types are
575 considered <a href="#t_firstclass">first class</a>.</p>
577 <pre> < <# elements> x <elementtype> ><br></pre>
578 <p>The number of elements is a constant integer value, elementtype may
579 be any integral or floating point type.</p>
581 <p> <tt><4 x int></tt>: Packed vector of 4 integer values.<br>
582 <tt><8 x float></tt>: Packed vector of 8 floating-point values.<br>
583 <tt><2 x uint></tt>: Packed vector of 2 unsigned integer values.</p>
587 <!-- *********************************************************************** -->
588 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
589 <!-- *********************************************************************** -->
590 <!-- ======================================================================= -->
591 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
593 <div class="doc_text">
594 <p>LLVM programs are composed of "Module"s, each of which is a
595 translation unit of the input programs. Each module consists of
596 functions, global variables, and symbol table entries. Modules may be
597 combined together with the LLVM linker, which merges function (and
598 global variable) definitions, resolves forward declarations, and merges
599 symbol table entries. Here is an example of the "hello world" module:</p>
600 <pre><i>; Declare the string constant as a global constant...</i>
601 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
602 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
604 <i>; External declaration of the puts function</i>
605 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
607 <i>; Definition of main function</i>
608 int %main() { <i>; int()* </i>
609 <i>; Convert [13x sbyte]* to sbyte *...</i>
611 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
613 <i>; Call puts function to write out the string to stdout...</i>
615 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
617 href="#i_ret">ret</a> int 0<br>}<br></pre>
618 <p>This example is made up of a <a href="#globalvars">global variable</a>
619 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
620 function, and a <a href="#functionstructure">function definition</a>
621 for "<tt>main</tt>".</p>
622 <a name="linkage"> In general, a module is made up of a list of global
623 values, where both functions and global variables are global values.
624 Global values are represented by a pointer to a memory location (in
625 this case, a pointer to an array of char, and a pointer to a function),
626 and have one of the following linkage types:</a>
629 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
630 <dd>Global values with internal linkage are only directly accessible
631 by objects in the current module. In particular, linking code into a
632 module with an internal global value may cause the internal to be
633 renamed as necessary to avoid collisions. Because the symbol is
634 internal to the module, all references can be updated. This
635 corresponds to the notion of the '<tt>static</tt>' keyword in C, or the
636 idea of "anonymous namespaces" in C++.
639 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
640 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt>
641 linkage, with the twist that linking together two modules defining the
642 same <tt>linkonce</tt> globals will cause one of the globals to be
643 discarded. This is typically used to implement inline functions.
644 Unreferenced <tt>linkonce</tt> globals are allowed to be discarded.
647 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
648 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt>
649 linkage, except that unreferenced <tt>weak</tt> globals may not be
650 discarded. This is used to implement constructs in C such as "<tt>int
651 X;</tt>" at global scope.
654 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
655 <dd>"<tt>appending</tt>" linkage may only be applied to global
656 variables of pointer to array type. When two global variables with
657 appending linkage are linked together, the two global arrays are
658 appended together. This is the LLVM, typesafe, equivalent of having
659 the system linker append together "sections" with identical names when
663 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
664 <dd>If none of the above identifiers are used, the global is
665 externally visible, meaning that it participates in linkage and can be
666 used to resolve external symbol references.
671 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
672 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
673 variable and was linked with this one, one of the two would be renamed,
674 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
675 external (i.e., lacking any linkage declarations), they are accessible
676 outside of the current module. It is illegal for a function <i>declaration</i>
677 to have any linkage type other than "externally visible".</a></p>
680 <!-- ======================================================================= -->
681 <div class="doc_subsection">
682 <a name="globalvars">Global Variables</a>
685 <div class="doc_text">
687 <p>Global variables define regions of memory allocated at compilation
688 time instead of run-time. Global variables may optionally be
689 initialized. A variable may be defined as a global "constant", which
690 indicates that the contents of the variable will never be modified
691 (opening options for optimization).</p>
693 <p>As SSA values, global variables define pointer values that are in
694 scope (i.e. they dominate) for all basic blocks in the program. Global
695 variables always define a pointer to their "content" type because they
696 describe a region of memory, and all memory objects in LLVM are
697 accessed through pointers.</p>
702 <!-- ======================================================================= -->
703 <div class="doc_subsection">
704 <a name="functionstructure">Functions</a>
707 <div class="doc_text">
709 <p>LLVM function definitions are composed of a (possibly empty) argument list,
710 an opening curly brace, a list of basic blocks, and a closing curly brace. LLVM
711 function declarations are defined with the "<tt>declare</tt>" keyword, a
712 function name, and a function signature.</p>
714 <p>A function definition contains a list of basic blocks, forming the CFG for
715 the function. Each basic block may optionally start with a label (giving the
716 basic block a symbol table entry), contains a list of instructions, and ends
717 with a <a href="#terminators">terminator</a> instruction (such as a branch or
718 function return).</p>
720 <p>The first basic block in program is special in two ways: it is immediately
721 executed on entrance to the function, and it is not allowed to have predecessor
722 basic blocks (i.e. there can not be any branches to the entry block of a
723 function). Because the block can have no predecessors, it also cannot have any
724 <a href="#i_phi">PHI nodes</a>.</p>
726 <p>LLVM functions are identified by their name and type signature. Hence, two
727 functions with the same name but different parameter lists or return values are
728 considered different functions, and LLVM will resolves references to each
734 <!-- *********************************************************************** -->
735 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
736 <!-- *********************************************************************** -->
737 <div class="doc_text">
738 <p>The LLVM instruction set consists of several different
739 classifications of instructions: <a href="#terminators">terminator
740 instructions</a>, <a href="#binaryops">binary instructions</a>, <a
741 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
742 instructions</a>.</p>
744 <!-- ======================================================================= -->
745 <div class="doc_subsection"> <a name="terminators">Terminator
746 Instructions</a> </div>
747 <div class="doc_text">
748 <p>As mentioned <a href="#functionstructure">previously</a>, every
749 basic block in a program ends with a "Terminator" instruction, which
750 indicates which block should be executed after the current block is
751 finished. These terminator instructions typically yield a '<tt>void</tt>'
752 value: they produce control flow, not values (the one exception being
753 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
754 <p>There are five different terminator instructions: the '<a
755 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
756 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
757 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
758 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
759 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
761 <!-- _______________________________________________________________________ -->
762 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
763 Instruction</a> </div>
764 <div class="doc_text">
766 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
767 ret void <i>; Return from void function</i>
770 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
771 value) from a function, back to the caller.</p>
772 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
773 returns a value and then causes control flow, and one that just causes
774 control flow to occur.</p>
776 <p>The '<tt>ret</tt>' instruction may return any '<a
777 href="#t_firstclass">first class</a>' type. Notice that a function is
778 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
779 instruction inside of the function that returns a value that does not
780 match the return type of the function.</p>
782 <p>When the '<tt>ret</tt>' instruction is executed, control flow
783 returns back to the calling function's context. If the caller is a "<a
784 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
785 the instruction after the call. If the caller was an "<a
786 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
787 at the beginning "normal" of the destination block. If the instruction
788 returns a value, that value shall set the call or invoke instruction's
791 <pre> ret int 5 <i>; Return an integer value of 5</i>
792 ret void <i>; Return from a void function</i>
795 <!-- _______________________________________________________________________ -->
796 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
797 <div class="doc_text">
799 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
802 <p>The '<tt>br</tt>' instruction is used to cause control flow to
803 transfer to a different basic block in the current function. There are
804 two forms of this instruction, corresponding to a conditional branch
805 and an unconditional branch.</p>
807 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
808 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
809 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
810 value as a target.</p>
812 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
813 argument is evaluated. If the value is <tt>true</tt>, control flows
814 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
815 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
817 <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
818 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
820 <!-- _______________________________________________________________________ -->
821 <div class="doc_subsubsection">
822 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
825 <div class="doc_text">
829 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
834 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
835 several different places. It is a generalization of the '<tt>br</tt>'
836 instruction, allowing a branch to occur to one of many possible
842 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
843 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
844 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
845 table is not allowed to contain duplicate constant entries.</p>
849 <p>The <tt>switch</tt> instruction specifies a table of values and
850 destinations. When the '<tt>switch</tt>' instruction is executed, this
851 table is searched for the given value. If the value is found, control flow is
852 transfered to the corresponding destination; otherwise, control flow is
853 transfered to the default destination.</p>
855 <h5>Implementation:</h5>
857 <p>Depending on properties of the target machine and the particular
858 <tt>switch</tt> instruction, this instruction may be code generated in different
859 ways. For example, it could be generated as a series of chained conditional
860 branches or with a lookup table.</p>
865 <i>; Emulate a conditional br instruction</i>
866 %Val = <a href="#i_cast">cast</a> bool %value to int
867 switch int %Val, label %truedest [int 0, label %falsedest ]
869 <i>; Emulate an unconditional br instruction</i>
870 switch uint 0, label %dest [ ]
872 <i>; Implement a jump table:</i>
873 switch uint %val, label %otherwise [ uint 0, label %onzero
875 uint 2, label %ontwo ]
878 <!-- _______________________________________________________________________ -->
879 <div class="doc_subsubsection"> <a name="i_invoke">'<tt>invoke</tt>'
880 Instruction</a> </div>
881 <div class="doc_text">
883 <pre> <result> = invoke <ptr to function ty> %<function ptr val>(<function args>)<br> to label <normal label> except label <exception label><br></pre>
885 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a
886 specified function, with the possibility of control flow transfer to
887 either the '<tt>normal</tt>' <tt>label</tt> label or the '<tt>exception</tt>'<tt>label</tt>.
888 If the callee function returns with the "<tt><a href="#i_ret">ret</a></tt>"
889 instruction, control flow will return to the "normal" label. If the
890 callee (or any indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
891 instruction, control is interrupted, and continued at the dynamically
892 nearest "except" label.</p>
894 <p>This instruction requires several arguments:</p>
896 <li>'<tt>ptr to function ty</tt>': shall be the signature of the
897 pointer to function value being invoked. In most cases, this is a
898 direct function invocation, but indirect <tt>invoke</tt>s are just as
899 possible, branching off an arbitrary pointer to function value. </li>
900 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer
901 to a function to be invoked. </li>
902 <li>'<tt>function args</tt>': argument list whose types match the
903 function signature argument types. If the function signature indicates
904 the function accepts a variable number of arguments, the extra
905 arguments can be specified. </li>
906 <li>'<tt>normal label</tt>': the label reached when the called
907 function executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
908 <li>'<tt>exception label</tt>': the label reached when a callee
909 returns with the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
912 <p>This instruction is designed to operate as a standard '<tt><a
913 href="#i_call">call</a></tt>' instruction in most regards. The
914 primary difference is that it establishes an association with a label,
915 which is used by the runtime library to unwind the stack.</p>
916 <p>This instruction is used in languages with destructors to ensure
917 that proper cleanup is performed in the case of either a <tt>longjmp</tt>
918 or a thrown exception. Additionally, this is important for
919 implementation of '<tt>catch</tt>' clauses in high-level languages that
922 <pre> %retval = invoke int %Test(int 15)<br> to label %Continue<br> except label %TestCleanup <i>; {int}:retval set</i>
927 <!-- _______________________________________________________________________ -->
929 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
930 Instruction</a> </div>
932 <div class="doc_text">
941 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
942 at the first callee in the dynamic call stack which used an <a
943 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
944 primarily used to implement exception handling.</p>
948 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
949 immediately halt. The dynamic call stack is then searched for the first <a
950 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
951 execution continues at the "exceptional" destination block specified by the
952 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
953 dynamic call chain, undefined behavior results.</p>
956 <!-- _______________________________________________________________________ -->
958 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
959 Instruction</a> </div>
961 <div class="doc_text">
970 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
971 instruction is used to inform the optimizer that a particular portion of the
972 code is not reachable. This can be used to indicate that the code after a
973 no-return function cannot be reached, and other facts.</p>
977 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
982 <!-- ======================================================================= -->
983 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
984 <div class="doc_text">
985 <p>Binary operators are used to do most of the computation in a
986 program. They require two operands, execute an operation on them, and
987 produce a single value. Although, that single value might represent
988 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
989 The result value of a binary operator is not
990 necessarily the same type as its operands.</p>
991 <p>There are several different binary operators:</p>
993 <!-- _______________________________________________________________________ -->
994 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
995 Instruction</a> </div>
996 <div class="doc_text">
998 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1001 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1003 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1004 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1005 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1006 Both arguments must have identical types.</p>
1008 <p>The value produced is the integer or floating point sum of the two
1011 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
1014 <!-- _______________________________________________________________________ -->
1015 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1016 Instruction</a> </div>
1017 <div class="doc_text">
1019 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1022 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1024 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1025 instruction present in most other intermediate representations.</p>
1027 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1028 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1030 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1031 Both arguments must have identical types.</p>
1033 <p>The value produced is the integer or floating point difference of
1034 the two operands.</p>
1036 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
1037 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
1040 <!-- _______________________________________________________________________ -->
1041 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1042 Instruction</a> </div>
1043 <div class="doc_text">
1045 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1048 <p>The '<tt>mul</tt>' instruction returns the product of its two
1051 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1052 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1054 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1055 Both arguments must have identical types.</p>
1057 <p>The value produced is the integer or floating point product of the
1059 <p>There is no signed vs unsigned multiplication. The appropriate
1060 action is taken based on the type of the operand.</p>
1062 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
1065 <!-- _______________________________________________________________________ -->
1066 <div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
1067 Instruction</a> </div>
1068 <div class="doc_text">
1070 <pre> <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1073 <p>The '<tt>div</tt>' instruction returns the quotient of its two
1076 <p>The two arguments to the '<tt>div</tt>' instruction must be either <a
1077 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1079 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1080 Both arguments must have identical types.</p>
1082 <p>The value produced is the integer or floating point quotient of the
1085 <pre> <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
1088 <!-- _______________________________________________________________________ -->
1089 <div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
1090 Instruction</a> </div>
1091 <div class="doc_text">
1093 <pre> <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1096 <p>The '<tt>rem</tt>' instruction returns the remainder from the
1097 division of its two operands.</p>
1099 <p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1100 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1102 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1103 Both arguments must have identical types.</p>
1105 <p>This returns the <i>remainder</i> of a division (where the result
1106 has the same sign as the divisor), not the <i>modulus</i> (where the
1107 result has the same sign as the dividend) of a value. For more
1108 information about the difference, see: <a
1109 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1112 <pre> <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1115 <!-- _______________________________________________________________________ -->
1116 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1117 Instructions</a> </div>
1118 <div class="doc_text">
1120 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1121 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1122 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1123 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1124 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1125 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1128 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1129 value based on a comparison of their two operands.</p>
1131 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1132 be of <a href="#t_firstclass">first class</a> type (it is not possible
1133 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1134 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1137 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1138 value if both operands are equal.<br>
1139 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1140 value if both operands are unequal.<br>
1141 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1142 value if the first operand is less than the second operand.<br>
1143 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1144 value if the first operand is greater than the second operand.<br>
1145 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1146 value if the first operand is less than or equal to the second operand.<br>
1147 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1148 value if the first operand is greater than or equal to the second
1151 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1152 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1153 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1154 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1155 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1156 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1159 <!-- ======================================================================= -->
1160 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1161 Operations</a> </div>
1162 <div class="doc_text">
1163 <p>Bitwise binary operators are used to do various forms of
1164 bit-twiddling in a program. They are generally very efficient
1165 instructions, and can commonly be strength reduced from other
1166 instructions. They require two operands, execute an operation on them,
1167 and produce a single value. The resulting value of the bitwise binary
1168 operators is always the same type as its first operand.</p>
1170 <!-- _______________________________________________________________________ -->
1171 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1172 Instruction</a> </div>
1173 <div class="doc_text">
1175 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1178 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1179 its two operands.</p>
1181 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1182 href="#t_integral">integral</a> values. Both arguments must have
1183 identical types.</p>
1185 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1187 <div style="align: center">
1188 <table border="1" cellspacing="0" cellpadding="4">
1219 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1220 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1221 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1224 <!-- _______________________________________________________________________ -->
1225 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1226 <div class="doc_text">
1228 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1231 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1232 or of its two operands.</p>
1234 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1235 href="#t_integral">integral</a> values. Both arguments must have
1236 identical types.</p>
1238 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1240 <div style="align: center">
1241 <table border="1" cellspacing="0" cellpadding="4">
1272 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1273 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1274 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1277 <!-- _______________________________________________________________________ -->
1278 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1279 Instruction</a> </div>
1280 <div class="doc_text">
1282 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1285 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1286 or of its two operands. The <tt>xor</tt> is used to implement the
1287 "one's complement" operation, which is the "~" operator in C.</p>
1289 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1290 href="#t_integral">integral</a> values. Both arguments must have
1291 identical types.</p>
1293 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1295 <div style="align: center">
1296 <table border="1" cellspacing="0" cellpadding="4">
1328 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1329 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1330 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1331 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1334 <!-- _______________________________________________________________________ -->
1335 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1336 Instruction</a> </div>
1337 <div class="doc_text">
1339 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1342 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1343 the left a specified number of bits.</p>
1345 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
1346 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1349 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
1351 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1352 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1353 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1356 <!-- _______________________________________________________________________ -->
1357 <div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1358 Instruction</a> </div>
1359 <div class="doc_text">
1361 <pre> <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1364 <p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1365 the right a specified number of bits.</p>
1367 <p>The first argument to the '<tt>shr</tt>' instruction must be an <a
1368 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1371 <p>If the first argument is a <a href="#t_signed">signed</a> type, the
1372 most significant bit is duplicated in the newly free'd bit positions.
1373 If the first argument is unsigned, zero bits shall fill the empty
1376 <pre> <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1377 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1378 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1379 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1380 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1383 <!-- ======================================================================= -->
1384 <div class="doc_subsection"> <a name="memoryops">Memory Access
1385 Operations</a></div>
1386 <div class="doc_text">
1387 <p>A key design point of an SSA-based representation is how it
1388 represents memory. In LLVM, no memory locations are in SSA form, which
1389 makes things very simple. This section describes how to read, write,
1390 allocate and free memory in LLVM.</p>
1392 <!-- _______________________________________________________________________ -->
1393 <div class="doc_subsubsection"> <a name="i_malloc">'<tt>malloc</tt>'
1394 Instruction</a> </div>
1395 <div class="doc_text">
1397 <pre> <result> = malloc <type>, uint <NumElements> <i>; yields {type*}:result</i>
1398 <result> = malloc <type> <i>; yields {type*}:result</i>
1401 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
1402 heap and returns a pointer to it.</p>
1404 <p>The '<tt>malloc</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1405 bytes of memory from the operating system and returns a pointer of the
1406 appropriate type to the program. The second form of the instruction is
1407 a shorter version of the first instruction that defaults to allocating
1409 <p>'<tt>type</tt>' must be a sized type.</p>
1411 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
1412 a pointer is returned.</p>
1414 <pre> %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1417 href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1418 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1419 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1422 <!-- _______________________________________________________________________ -->
1423 <div class="doc_subsubsection"> <a name="i_free">'<tt>free</tt>'
1424 Instruction</a> </div>
1425 <div class="doc_text">
1427 <pre> free <type> <value> <i>; yields {void}</i>
1430 <p>The '<tt>free</tt>' instruction returns memory back to the unused
1431 memory heap, to be reallocated in the future.</p>
1434 <p>'<tt>value</tt>' shall be a pointer value that points to a value
1435 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
1438 <p>Access to the memory pointed to by the pointer is not longer defined
1439 after this instruction executes.</p>
1441 <pre> %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1442 free [4 x ubyte]* %array
1445 <!-- _______________________________________________________________________ -->
1446 <div class="doc_subsubsection"> <a name="i_alloca">'<tt>alloca</tt>'
1447 Instruction</a> </div>
1448 <div class="doc_text">
1450 <pre> <result> = alloca <type>, uint <NumElements> <i>; yields {type*}:result</i>
1451 <result> = alloca <type> <i>; yields {type*}:result</i>
1454 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
1455 stack frame of the procedure that is live until the current function
1456 returns to its caller.</p>
1458 <p>The the '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1459 bytes of memory on the runtime stack, returning a pointer of the
1460 appropriate type to the program. The second form of the instruction is
1461 a shorter version of the first that defaults to allocating one element.</p>
1462 <p>'<tt>type</tt>' may be any sized type.</p>
1464 <p>Memory is allocated, a pointer is returned. '<tt>alloca</tt>'d
1465 memory is automatically released when the function returns. The '<tt>alloca</tt>'
1466 instruction is commonly used to represent automatic variables that must
1467 have an address available. When the function returns (either with the <tt><a
1468 href="#i_ret">ret</a></tt> or <tt><a href="#i_invoke">invoke</a></tt>
1469 instructions), the memory is reclaimed.</p>
1471 <pre> %ptr = alloca int <i>; yields {int*}:ptr</i>
1472 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
1475 <!-- _______________________________________________________________________ -->
1476 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
1477 Instruction</a> </div>
1478 <div class="doc_text">
1480 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
1482 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
1484 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
1485 address to load from. The pointer must point to a <a
1486 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
1487 marked as <tt>volatile</tt> then the optimizer is not allowed to modify
1488 the number or order of execution of this <tt>load</tt> with other
1489 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
1492 <p>The location of memory pointed to is loaded.</p>
1494 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1496 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1497 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1500 <!-- _______________________________________________________________________ -->
1501 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
1502 Instruction</a> </div>
1504 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1505 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1508 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
1510 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
1511 to store and an address to store it into. The type of the '<tt><pointer></tt>'
1512 operand must be a pointer to the type of the '<tt><value></tt>'
1513 operand. If the <tt>store</tt> is marked as <tt>volatile</tt> then the
1514 optimizer is not allowed to modify the number or order of execution of
1515 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
1516 href="#i_store">store</a></tt> instructions.</p>
1518 <p>The contents of memory are updated to contain '<tt><value></tt>'
1519 at the location specified by the '<tt><pointer></tt>' operand.</p>
1521 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1523 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1524 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1526 <!-- _______________________________________________________________________ -->
1527 <div class="doc_subsubsection">
1528 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
1531 <div class="doc_text">
1534 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
1540 The '<tt>getelementptr</tt>' instruction is used to get the address of a
1541 subelement of an aggregate data structure.</p>
1545 <p>This instruction takes a list of integer constants that indicate what
1546 elements of the aggregate object to index to. The actual types of the arguments
1547 provided depend on the type of the first pointer argument. The
1548 '<tt>getelementptr</tt>' instruction is used to index down through the type
1549 levels of a structure. When indexing into a structure, only <tt>uint</tt>
1550 integer constants are allowed. When indexing into an array or pointer
1551 <tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
1553 <p>For example, let's consider a C code fragment and how it gets
1554 compiled to LLVM:</p>
1568 int *foo(struct ST *s) {
1569 return &s[1].Z.B[5][13];
1573 <p>The LLVM code generated by the GCC frontend is:</p>
1576 %RT = type { sbyte, [10 x [20 x int]], sbyte }
1577 %ST = type { int, double, %RT }
1581 int* %foo(%ST* %s) {
1583 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
1590 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
1591 on the pointer type that is being index into. <a href="#t_pointer">Pointer</a>
1592 and <a href="#t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
1593 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="#t_struct">structure</a>
1594 types require <tt>uint</tt> <b>constants</b>.</p>
1596 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
1597 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
1598 }</tt>' type, a structure. The second index indexes into the third element of
1599 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
1600 sbyte }</tt>' type, another structure. The third index indexes into the second
1601 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
1602 array. The two dimensions of the array are subscripted into, yielding an
1603 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction return a pointer
1604 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
1606 <p>Note that it is perfectly legal to index partially through a
1607 structure, returning a pointer to an inner element. Because of this,
1608 the LLVM code for the given testcase is equivalent to:</p>
1611 int* "foo"(%ST* %s) {
1612 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
1613 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
1614 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
1615 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
1616 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
1622 <i>; yields [12 x ubyte]*:aptr</i>
1623 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
1627 <!-- ======================================================================= -->
1628 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
1629 <div class="doc_text">
1630 <p>The instructions in this category are the "miscellaneous"
1631 instructions, which defy better classification.</p>
1633 <!-- _______________________________________________________________________ -->
1634 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
1635 Instruction</a> </div>
1636 <div class="doc_text">
1638 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
1640 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
1641 the SSA graph representing the function.</p>
1643 <p>The type of the incoming values are specified with the first type
1644 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
1645 as arguments, with one pair for each predecessor basic block of the
1646 current block. Only values of <a href="#t_firstclass">first class</a>
1647 type may be used as the value arguments to the PHI node. Only labels
1648 may be used as the label arguments.</p>
1649 <p>There must be no non-phi instructions between the start of a basic
1650 block and the PHI instructions: i.e. PHI instructions must be first in
1653 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
1654 value specified by the parameter, depending on which basic block we
1655 came from in the last <a href="#terminators">terminator</a> instruction.</p>
1657 <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>
1660 <!-- _______________________________________________________________________ -->
1661 <div class="doc_subsubsection">
1662 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
1665 <div class="doc_text">
1670 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
1676 The '<tt>cast</tt>' instruction is used as the primitive means to convert
1677 integers to floating point, change data type sizes, and break type safety (by
1685 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
1686 class value, and a type to cast it to, which must also be a <a
1687 href="#t_firstclass">first class</a> type.
1693 This instruction follows the C rules for explicit casts when determining how the
1694 data being cast must change to fit in its new container.
1698 When casting to bool, any value that would be considered true in the context of
1699 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
1700 all else are '<tt>false</tt>'.
1704 When extending an integral value from a type of one signness to another (for
1705 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
1706 <b>source</b> value is signed, and zero-extended if the source value is
1707 unsigned. <tt>bool</tt> values are always zero extended into either zero or
1714 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
1715 %Y = cast int 123 to bool <i>; yields bool:true</i>
1719 <!-- _______________________________________________________________________ -->
1720 <div class="doc_subsubsection">
1721 <a name="i_select">'<tt>select</tt>' Instruction</a>
1724 <div class="doc_text">
1729 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
1735 The '<tt>select</tt>' instruction is used to choose one value based on a
1736 condition, without branching.
1743 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.
1749 If the boolean condition evaluates to true, the instruction returns the first
1750 value argument, otherwise it returns the second value argument.
1756 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
1764 <!-- _______________________________________________________________________ -->
1765 <div class="doc_subsubsection"> <a name="i_call">'<tt>call</tt>'
1766 Instruction</a> </div>
1767 <div class="doc_text">
1769 <pre> <result> = call <ty>* <fnptrval>(<param list>)<br></pre>
1771 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
1773 <p>This instruction requires several arguments:</p>
1776 <p>'<tt>ty</tt>': shall be the signature of the pointer to function
1777 value being invoked. The argument types must match the types implied
1778 by this signature.</p>
1781 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a
1782 function to be invoked. In most cases, this is a direct function
1783 invocation, but indirect <tt>call</tt>s are just as possible,
1784 calling an arbitrary pointer to function values.</p>
1787 <p>'<tt>function args</tt>': argument list whose types match the
1788 function signature argument types. If the function signature
1789 indicates the function accepts a variable number of arguments, the
1790 extra arguments can be specified.</p>
1794 <p>The '<tt>call</tt>' instruction is used to cause control flow to
1795 transfer to a specified function, with its incoming arguments bound to
1796 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
1797 instruction in the called function, control flow continues with the
1798 instruction after the function call, and the return value of the
1799 function is bound to the result argument. This is a simpler case of
1800 the <a href="#i_invoke">invoke</a> instruction.</p>
1802 <pre> %retval = call int %test(int %argc)<br> call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);<br></pre>
1805 <!-- _______________________________________________________________________ -->
1806 <div class="doc_subsubsection">
1807 <a name="i_vanext">'<tt>vanext</tt>' Instruction</a>
1810 <div class="doc_text">
1815 <resultarglist> = vanext <va_list> <arglist>, <argty>
1820 <p>The '<tt>vanext</tt>' instruction is used to access arguments passed
1821 through the "variable argument" area of a function call. It is used to
1822 implement the <tt>va_arg</tt> macro in C.</p>
1826 <p>This instruction takes a <tt>va_list</tt> value and the type of the
1827 argument. It returns another <tt>va_list</tt>. The actual type of
1828 <tt>va_list</tt> may be defined differently for different targets. Most targets
1829 use a <tt>va_list</tt> type of <tt>sbyte*</tt> or some other pointer type.</p>
1833 <p>The '<tt>vanext</tt>' instruction advances the specified <tt>va_list</tt>
1834 past an argument of the specified type. In conjunction with the <a
1835 href="#i_vaarg"><tt>vaarg</tt></a> instruction, it is used to implement
1836 the <tt>va_arg</tt> macro available in C. For more information, see
1837 the variable argument handling <a href="#int_varargs">Intrinsic
1840 <p>It is legal for this instruction to be called in a function which
1841 does not take a variable number of arguments, for example, the <tt>vfprintf</tt>
1844 <p><tt>vanext</tt> is an LLVM instruction instead of an <a
1845 href="#intrinsics">intrinsic function</a> because it takes a type as an
1846 argument. The type refers to the current argument in the <tt>va_list</tt>, it
1847 tells the compiler how far on the stack it needs to advance to find the next
1852 <p>See the <a href="#int_varargs">variable argument processing</a>
1857 <!-- _______________________________________________________________________ -->
1858 <div class="doc_subsubsection">
1859 <a name="i_vaarg">'<tt>vaarg</tt>' Instruction</a>
1862 <div class="doc_text">
1867 <resultval> = vaarg <va_list> <arglist>, <argty>
1872 <p>The '<tt>vaarg</tt>' instruction is used to access arguments passed through
1873 the "variable argument" area of a function call. It is used to implement the
1874 <tt>va_arg</tt> macro in C.</p>
1878 <p>This instruction takes a <tt>va_list</tt> value and the type of the
1879 argument. It returns a value of the specified argument type. Again, the actual
1880 type of <tt>va_list</tt> is target specific.</p>
1884 <p>The '<tt>vaarg</tt>' instruction loads an argument of the specified type from
1885 the specified <tt>va_list</tt>. In conjunction with the <a
1886 href="#i_vanext"><tt>vanext</tt></a> instruction, it is used to implement the
1887 <tt>va_arg</tt> macro available in C. For more information, see the variable
1888 argument handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
1890 <p>It is legal for this instruction to be called in a function which does not
1891 take a variable number of arguments, for example, the <tt>vfprintf</tt>
1894 <p><tt>vaarg</tt> is an LLVM instruction instead of an <a
1895 href="#intrinsics">intrinsic function</a> because it takes an type as an
1900 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
1904 <!-- *********************************************************************** -->
1905 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
1906 <!-- *********************************************************************** -->
1908 <div class="doc_text">
1910 <p>LLVM supports the notion of an "intrinsic function". These functions have
1911 well known names and semantics, and are required to follow certain
1912 restrictions. Overall, these instructions represent an extension mechanism for
1913 the LLVM language that does not require changing all of the transformations in
1914 LLVM to add to the language (or the bytecode reader/writer, the parser,
1917 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix, this
1918 prefix is reserved in LLVM for intrinsic names, thus functions may not be named
1919 this. Intrinsic functions must always be external functions: you cannot define
1920 the body of intrinsic functions. Intrinsic functions may only be used in call
1921 or invoke instructions: it is illegal to take the address of an intrinsic
1922 function. Additionally, because intrinsic functions are part of the LLVM
1923 language, it is required that they all be documented here if any are added.</p>
1927 Adding an intrinsic to LLVM is straight-forward if it is possible to express the
1928 concept in LLVM directly (ie, code generator support is not _required_). To do
1929 this, extend the default implementation of the IntrinsicLowering class to handle
1930 the intrinsic. Code generators use this class to lower intrinsics they do not
1931 understand to raw LLVM instructions that they do.
1936 <!-- ======================================================================= -->
1937 <div class="doc_subsection">
1938 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
1941 <div class="doc_text">
1943 <p>Variable argument support is defined in LLVM with the <a
1944 href="#i_vanext"><tt>vanext</tt></a> instruction and these three
1945 intrinsic functions. These functions are related to the similarly
1946 named macros defined in the <tt><stdarg.h></tt> header file.</p>
1948 <p>All of these functions operate on arguments that use a
1949 target-specific value type "<tt>va_list</tt>". The LLVM assembly
1950 language reference manual does not define what this type is, so all
1951 transformations should be prepared to handle intrinsics with any type
1954 <p>This example shows how the <a href="#i_vanext"><tt>vanext</tt></a>
1955 instruction and the variable argument handling intrinsic functions are
1959 int %test(int %X, ...) {
1960 ; Initialize variable argument processing
1961 %ap = call sbyte* %<a href="#i_va_start">llvm.va_start</a>()
1963 ; Read a single integer argument
1964 %tmp = vaarg sbyte* %ap, int
1966 ; Advance to the next argument
1967 %ap2 = vanext sbyte* %ap, int
1969 ; Demonstrate usage of llvm.va_copy and llvm.va_end
1970 %aq = call sbyte* %<a href="#i_va_copy">llvm.va_copy</a>(sbyte* %ap2)
1971 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %aq)
1973 ; Stop processing of arguments.
1974 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %ap2)
1980 <!-- _______________________________________________________________________ -->
1981 <div class="doc_subsubsection">
1982 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
1986 <div class="doc_text">
1988 <pre> call <va_list> ()* %llvm.va_start()<br></pre>
1990 <p>The '<tt>llvm.va_start</tt>' intrinsic returns a new <tt><arglist></tt>
1991 for subsequent use by the variable argument intrinsics.</p>
1993 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
1994 macro available in C. In a target-dependent way, it initializes and
1995 returns a <tt>va_list</tt> element, so that the next <tt>vaarg</tt>
1996 will produce the first variable argument passed to the function. Unlike
1997 the C <tt>va_start</tt> macro, this intrinsic does not need to know the
1998 last argument of the function, the compiler can figure that out.</p>
1999 <p>Note that this intrinsic function is only legal to be called from
2000 within the body of a variable argument function.</p>
2003 <!-- _______________________________________________________________________ -->
2004 <div class="doc_subsubsection">
2005 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
2008 <div class="doc_text">
2010 <pre> call void (<va_list>)* %llvm.va_end(<va_list> <arglist>)<br></pre>
2012 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
2013 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
2014 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
2016 <p>The argument is a <tt>va_list</tt> to destroy.</p>
2018 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
2019 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
2020 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
2021 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
2022 with calls to <tt>llvm.va_end</tt>.</p>
2025 <!-- _______________________________________________________________________ -->
2026 <div class="doc_subsubsection">
2027 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
2030 <div class="doc_text">
2035 call <va_list> (<va_list>)* %llvm.va_copy(<va_list> <destarglist>)
2040 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
2041 from the source argument list to the destination argument list.</p>
2045 <p>The argument is the <tt>va_list</tt> to copy.</p>
2049 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
2050 macro available in C. In a target-dependent way, it copies the source
2051 <tt>va_list</tt> element into the returned list. This intrinsic is necessary
2052 because the <tt><a href="#i_va_start">llvm.va_start</a></tt> intrinsic may be
2053 arbitrarily complex and require memory allocation, for example.</p>
2057 <!-- ======================================================================= -->
2058 <div class="doc_subsection">
2059 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
2062 <div class="doc_text">
2065 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
2066 Collection</a> requires the implementation and generation of these intrinsics.
2067 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
2068 stack</a>, as well as garbage collector implementations that require <a
2069 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
2070 Front-ends for type-safe garbage collected languages should generate these
2071 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
2072 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
2076 <!-- _______________________________________________________________________ -->
2077 <div class="doc_subsubsection">
2078 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
2081 <div class="doc_text">
2086 call void (<ty>**, <ty2>*)* %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
2091 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existance of a GC root to
2092 the code generator, and allows some metadata to be associated with it.</p>
2096 <p>The first argument specifies the address of a stack object that contains the
2097 root pointer. The second pointer (which must be either a constant or a global
2098 value address) contains the meta-data to be associated with the root.</p>
2102 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
2103 location. At compile-time, the code generator generates information to allow
2104 the runtime to find the pointer at GC safe points.
2110 <!-- _______________________________________________________________________ -->
2111 <div class="doc_subsubsection">
2112 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
2115 <div class="doc_text">
2120 call sbyte* (sbyte**)* %llvm.gcread(sbyte** %Ptr)
2125 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
2126 locations, allowing garbage collector implementations that require read
2131 <p>The argument is the address to read from, which should be an address
2132 allocated from the garbage collector.</p>
2136 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
2137 instruction, but may be replaced with substantially more complex code by the
2138 garbage collector runtime, as needed.</p>
2143 <!-- _______________________________________________________________________ -->
2144 <div class="doc_subsubsection">
2145 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
2148 <div class="doc_text">
2153 call void (sbyte*, sbyte**)* %llvm.gcwrite(sbyte* %P1, sbyte** %P2)
2158 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
2159 locations, allowing garbage collector implementations that require write
2160 barriers (such as generational or reference counting collectors).</p>
2164 <p>The first argument is the reference to store, and the second is the heap
2165 location to store to.</p>
2169 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
2170 instruction, but may be replaced with substantially more complex code by the
2171 garbage collector runtime, as needed.</p>
2177 <!-- ======================================================================= -->
2178 <div class="doc_subsection">
2179 <a name="int_codegen">Code Generator Intrinsics</a>
2182 <div class="doc_text">
2184 These intrinsics are provided by LLVM to expose special features that may only
2185 be implemented with code generator support.
2190 <!-- _______________________________________________________________________ -->
2191 <div class="doc_subsubsection">
2192 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
2195 <div class="doc_text">
2199 call void* ()* %llvm.returnaddress(uint <level>)
2205 The '<tt>llvm.returnaddress</tt>' intrinsic returns a target-specific value
2206 indicating the return address of the current function or one of its callers.
2212 The argument to this intrinsic indicates which function to return the address
2213 for. Zero indicates the calling function, one indicates its caller, etc. The
2214 argument is <b>required</b> to be a constant integer value.
2220 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
2221 the return address of the specified call frame, or zero if it cannot be
2222 identified. The value returned by this intrinsic is likely to be incorrect or 0
2223 for arguments other than zero, so it should only be used for debugging purposes.
2227 Note that calling this intrinsic does not prevent function inlining or other
2228 aggressive transformations, so the value returned may not that of the obvious
2229 source-language caller.
2234 <!-- _______________________________________________________________________ -->
2235 <div class="doc_subsubsection">
2236 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
2239 <div class="doc_text">
2243 call void* ()* %llvm.frameaddress(uint <level>)
2249 The '<tt>llvm.frameaddress</tt>' intrinsic returns the target-specific frame
2250 pointer value for the specified stack frame.
2256 The argument to this intrinsic indicates which function to return the frame
2257 pointer for. Zero indicates the calling function, one indicates its caller,
2258 etc. The argument is <b>required</b> to be a constant integer value.
2264 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
2265 the frame address of the specified call frame, or zero if it cannot be
2266 identified. The value returned by this intrinsic is likely to be incorrect or 0
2267 for arguments other than zero, so it should only be used for debugging purposes.
2271 Note that calling this intrinsic does not prevent function inlining or other
2272 aggressive transformations, so the value returned may not that of the obvious
2273 source-language caller.
2277 <!-- ======================================================================= -->
2278 <div class="doc_subsection">
2279 <a name="int_os">Operating System Intrinsics</a>
2282 <div class="doc_text">
2284 These intrinsics are provided by LLVM to support the implementation of
2285 operating system level code.
2290 <!-- _______________________________________________________________________ -->
2291 <div class="doc_subsubsection">
2292 <a name="i_readport">'<tt>llvm.readport</tt>' Intrinsic</a>
2295 <div class="doc_text">
2299 call <integer type> (<integer type>)* %llvm.readport (<integer type> <address>)
2305 The '<tt>llvm.readport</tt>' intrinsic reads data from the specified hardware
2312 The argument to this intrinsic indicates the hardware I/O address from which
2313 to read the data. The address is in the hardware I/O address namespace (as
2314 opposed to being a memory location for memory mapped I/O).
2320 The '<tt>llvm.readport</tt>' intrinsic reads data from the hardware I/O port
2321 specified by <i>address</i> and returns the value. The address and return
2322 value must be integers, but the size is dependent upon the platform upon which
2323 the program is code generated. For example, on x86, the address must be an
2324 unsigned 16 bit value, and the return value must be 8, 16, or 32 bits.
2329 <!-- _______________________________________________________________________ -->
2330 <div class="doc_subsubsection">
2331 <a name="i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a>
2334 <div class="doc_text">
2338 call void (<integer type>, <integer type>)* %llvm.writeport (<integer type> <value>, <integer type> <address>)
2344 The '<tt>llvm.writeport</tt>' intrinsic writes data to the specified hardware
2351 The first argument is the value to write to the I/O port.
2355 The second argument indicates the hardware I/O address to which data should be
2356 written. The address is in the hardware I/O address namespace (as opposed to
2357 being a memory location for memory mapped I/O).
2363 The '<tt>llvm.writeport</tt>' intrinsic writes <i>value</i> to the I/O port
2364 specified by <i>address</i>. The address and value must be integers, but the
2365 size is dependent upon the platform upon which the program is code generated.
2366 For example, on x86, the address must be an unsigned 16 bit value, and the
2367 value written must be 8, 16, or 32 bits in length.
2372 <!-- _______________________________________________________________________ -->
2373 <div class="doc_subsubsection">
2374 <a name="i_readio">'<tt>llvm.readio</tt>' Intrinsic</a>
2377 <div class="doc_text">
2381 call <result> (<ty>*)* %llvm.readio (<ty> * <pointer>)
2387 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2394 The argument to this intrinsic is a pointer indicating the memory address from
2395 which to read the data. The data must be a
2396 <a href="#t_firstclass">first class</a> type.
2402 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2403 location specified by <i>pointer</i> and returns the value. The argument must
2404 be a pointer, and the return value must be a
2405 <a href="#t_firstclass">first class</a> type. However, certain architectures
2406 may not support I/O on all first class types. For example, 32 bit processors
2407 may only support I/O on data types that are 32 bits or less.
2411 This intrinsic enforces an in-order memory model for llvm.readio and
2412 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2413 scheduled processors may execute loads and stores out of order, re-ordering at
2414 run time accesses to memory mapped I/O registers. Using these intrinsics
2415 ensures that accesses to memory mapped I/O registers occur in program order.
2420 <!-- _______________________________________________________________________ -->
2421 <div class="doc_subsubsection">
2422 <a name="i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a>
2425 <div class="doc_text">
2429 call void (<ty1>, <ty2>*)* %llvm.writeio (<ty1> <value>, <ty2> * <pointer>)
2435 The '<tt>llvm.writeio</tt>' intrinsic writes data to the specified memory
2442 The first argument is the value to write to the memory mapped I/O location.
2443 The second argument is a pointer indicating the memory address to which the
2444 data should be written.
2450 The '<tt>llvm.writeio</tt>' intrinsic writes <i>value</i> to the memory mapped
2451 I/O address specified by <i>pointer</i>. The value must be a
2452 <a href="#t_firstclass">first class</a> type. However, certain architectures
2453 may not support I/O on all first class types. For example, 32 bit processors
2454 may only support I/O on data types that are 32 bits or less.
2458 This intrinsic enforces an in-order memory model for llvm.readio and
2459 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2460 scheduled processors may execute loads and stores out of order, re-ordering at
2461 run time accesses to memory mapped I/O registers. Using these intrinsics
2462 ensures that accesses to memory mapped I/O registers occur in program order.
2467 <!-- ======================================================================= -->
2468 <div class="doc_subsection">
2469 <a name="int_libc">Standard C Library Intrinsics</a>
2472 <div class="doc_text">
2474 LLVM provides intrinsics for a few important standard C library functions.
2475 These intrinsics allow source-language front-ends to pass information about the
2476 alignment of the pointer arguments to the code generator, providing opportunity
2477 for more efficient code generation.
2482 <!-- _______________________________________________________________________ -->
2483 <div class="doc_subsubsection">
2484 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
2487 <div class="doc_text">
2491 call void (sbyte*, sbyte*, uint, uint)* %llvm.memcpy(sbyte* <dest>, sbyte* <src>,
2492 uint <len>, uint <align>)
2498 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2499 location to the destination location.
2503 Note that, unlike the standard libc function, the <tt>llvm.memcpy</tt> intrinsic
2504 does not return a value, and takes an extra alignment argument.
2510 The first argument is a pointer to the destination, the second is a pointer to
2511 the source. The third argument is an (arbitrarily sized) integer argument
2512 specifying the number of bytes to copy, and the fourth argument is the alignment
2513 of the source and destination locations.
2517 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2518 the caller guarantees that the size of the copy is a multiple of the alignment
2519 and that both the source and destination pointers are aligned to that boundary.
2525 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2526 location to the destination location, which are not allowed to overlap. It
2527 copies "len" bytes of memory over. If the argument is known to be aligned to
2528 some boundary, this can be specified as the fourth argument, otherwise it should
2534 <!-- _______________________________________________________________________ -->
2535 <div class="doc_subsubsection">
2536 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
2539 <div class="doc_text">
2543 call void (sbyte*, sbyte*, uint, uint)* %llvm.memmove(sbyte* <dest>, sbyte* <src>,
2544 uint <len>, uint <align>)
2550 The '<tt>llvm.memmove</tt>' intrinsic moves a block of memory from the source
2551 location to the destination location. It is similar to the '<tt>llvm.memcpy</tt>'
2552 intrinsic but allows the two memory locations to overlap.
2556 Note that, unlike the standard libc function, the <tt>llvm.memmove</tt> intrinsic
2557 does not return a value, and takes an extra alignment argument.
2563 The first argument is a pointer to the destination, the second is a pointer to
2564 the source. The third argument is an (arbitrarily sized) integer argument
2565 specifying the number of bytes to copy, and the fourth argument is the alignment
2566 of the source and destination locations.
2570 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2571 the caller guarantees that the size of the copy is a multiple of the alignment
2572 and that both the source and destination pointers are aligned to that boundary.
2578 The '<tt>llvm.memmove</tt>' intrinsic copies a block of memory from the source
2579 location to the destination location, which may overlap. It
2580 copies "len" bytes of memory over. If the argument is known to be aligned to
2581 some boundary, this can be specified as the fourth argument, otherwise it should
2587 <!-- _______________________________________________________________________ -->
2588 <div class="doc_subsubsection">
2589 <a name="i_memset">'<tt>llvm.memset</tt>' Intrinsic</a>
2592 <div class="doc_text">
2596 call void (sbyte*, ubyte, uint, uint)* %llvm.memset(sbyte* <dest>, ubyte <val>,
2597 uint <len>, uint <align>)
2603 The '<tt>llvm.memset</tt>' intrinsic fills a block of memory with a particular
2608 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
2609 does not return a value, and takes an extra alignment argument.
2615 The first argument is a pointer to the destination to fill, the second is the
2616 byte value to fill it with, the third argument is an (arbitrarily sized) integer
2617 argument specifying the number of bytes to fill, and the fourth argument is the
2618 known alignment of destination location.
2622 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2623 the caller guarantees that the size of the copy is a multiple of the alignment
2624 and that the destination pointer is aligned to that boundary.
2630 The '<tt>llvm.memset</tt>' intrinsic fills "len" bytes of memory starting at the
2631 destination location. If the argument is known to be aligned to some boundary,
2632 this can be specified as the fourth argument, otherwise it should be set to 0 or
2638 <!-- _______________________________________________________________________ -->
2639 <div class="doc_subsubsection">
2640 <a name="i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a>
2643 <div class="doc_text">
2647 call bool (<float or double>, <float or double>)* %llvm.isunordered(<float or double> Val1,
2648 <float or double> Val2)
2654 The '<tt>llvm.isunordered</tt>' intrinsic returns true if either or both of the
2655 specified floating point values is a NAN.
2661 The arguments are floating point numbers of the same type.
2667 If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
2675 <!-- ======================================================================= -->
2676 <div class="doc_subsection">
2677 <a name="int_debugger">Debugger Intrinsics</a>
2680 <div class="doc_text">
2682 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
2683 are described in the <a
2684 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
2685 Debugging</a> document.
2690 <!-- *********************************************************************** -->
2693 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2694 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
2695 <a href="http://validator.w3.org/check/referer"><img
2696 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
2698 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
2699 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
2700 Last modified: $Date$