1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>LLVM Assembly Language Reference Manual</title>
6 <link rel="stylesheet" href="llvm.css" type="text/css">
11 <div class="doc_title"> LLVM Language Reference Manual </div>
13 <li><a href="#abstract">Abstract</a></li>
14 <li><a href="#introduction">Introduction</a></li>
15 <li><a href="#identifiers">Identifiers</a></li>
16 <li><a href="#typesystem">Type System</a>
18 <li><a href="#t_primitive">Primitive Types</a>
20 <li><a href="#t_classifications">Type Classifications</a></li>
23 <li><a href="#t_derived">Derived Types</a>
25 <li><a href="#t_array">Array Type</a></li>
26 <li><a href="#t_function">Function Type</a></li>
27 <li><a href="#t_pointer">Pointer Type</a></li>
28 <li><a href="#t_struct">Structure Type</a></li>
29 <!-- <li><a href="#t_packed" >Packed Type</a> -->
34 <li><a href="#highlevel">High Level Structure</a>
36 <li><a href="#modulestructure">Module Structure</a></li>
37 <li><a href="#globalvars">Global Variables</a></li>
38 <li><a href="#functionstructure">Function Structure</a></li>
41 <li><a href="#instref">Instruction Reference</a>
43 <li><a href="#terminators">Terminator Instructions</a>
45 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
46 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
47 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
48 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
49 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
52 <li><a href="#binaryops">Binary Operations</a>
54 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
55 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
56 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
57 <li><a href="#i_div">'<tt>div</tt>' Instruction</a></li>
58 <li><a href="#i_rem">'<tt>rem</tt>' Instruction</a></li>
59 <li><a href="#i_setcc">'<tt>set<i>cc</i></tt>' Instructions</a></li>
62 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
64 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
65 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
66 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
67 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
68 <li><a href="#i_shr">'<tt>shr</tt>' Instruction</a></li>
71 <li><a href="#memoryops">Memory Access Operations</a>
73 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
74 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
75 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
76 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
77 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
78 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
81 <li><a href="#otherops">Other Operations</a>
83 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
84 <li><a href="#i_cast">'<tt>cast .. to</tt>' Instruction</a></li>
85 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
86 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
87 <li><a href="#i_vanext">'<tt>vanext</tt>' Instruction</a></li>
88 <li><a href="#i_vaarg">'<tt>vaarg</tt>' Instruction</a></li>
93 <li><a href="#intrinsics">Intrinsic Functions</a>
95 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
97 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
98 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
99 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
102 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
104 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
105 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
106 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
109 <li><a href="#int_codegen">Code Generator Intrinsics</a>
111 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
112 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
115 <li><a href="#int_os">Operating System Intrinsics</a>
117 <li><a href="#i_readport">'<tt>llvm.readport</tt>' Intrinsic</a></li>
118 <li><a href="#i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a></li>
119 <li><a href="#i_readio">'<tt>llvm.readio</tt>' Intrinsic</a></li>
120 <li><a href="#i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a></li>
122 <li><a href="#int_libc">Standard C Library Intrinsics</a>
124 <li><a href="#i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a></li>
125 <li><a href="#i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a></li>
126 <li><a href="#i_memset">'<tt>llvm.memset</tt>' Intrinsic</a></li>
127 <li><a href="#i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a></li>
130 <li><a href="#int_debugger">Debugger intrinsics</a></li>
135 <div class="doc_author">
136 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
137 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
140 <!-- *********************************************************************** -->
141 <div class="doc_section"> <a name="abstract">Abstract </a></div>
142 <!-- *********************************************************************** -->
144 <div class="doc_text">
145 <p>This document is a reference manual for the LLVM assembly language.
146 LLVM is an SSA based representation that provides type safety,
147 low-level operations, flexibility, and the capability of representing
148 'all' high-level languages cleanly. It is the common code
149 representation used throughout all phases of the LLVM compilation
153 <!-- *********************************************************************** -->
154 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
155 <!-- *********************************************************************** -->
157 <div class="doc_text">
159 <p>The LLVM code representation is designed to be used in three
160 different forms: as an in-memory compiler IR, as an on-disk bytecode
161 representation (suitable for fast loading by a Just-In-Time compiler),
162 and as a human readable assembly language representation. This allows
163 LLVM to provide a powerful intermediate representation for efficient
164 compiler transformations and analysis, while providing a natural means
165 to debug and visualize the transformations. The three different forms
166 of LLVM are all equivalent. This document describes the human readable
167 representation and notation.</p>
169 <p>The LLVM representation aims to be a light-weight and low-level
170 while being expressive, typed, and extensible at the same time. It
171 aims to be a "universal IR" of sorts, by being at a low enough level
172 that high-level ideas may be cleanly mapped to it (similar to how
173 microprocessors are "universal IR's", allowing many source languages to
174 be mapped to them). By providing type information, LLVM can be used as
175 the target of optimizations: for example, through pointer analysis, it
176 can be proven that a C automatic variable is never accessed outside of
177 the current function... allowing it to be promoted to a simple SSA
178 value instead of a memory location.</p>
182 <!-- _______________________________________________________________________ -->
183 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
185 <div class="doc_text">
187 <p>It is important to note that this document describes 'well formed'
188 LLVM assembly language. There is a difference between what the parser
189 accepts and what is considered 'well formed'. For example, the
190 following instruction is syntactically okay, but not well formed:</p>
193 %x = <a href="#i_add">add</a> int 1, %x
196 <p>...because the definition of <tt>%x</tt> does not dominate all of
197 its uses. The LLVM infrastructure provides a verification pass that may
198 be used to verify that an LLVM module is well formed. This pass is
199 automatically run by the parser after parsing input assembly, and by
200 the optimizer before it outputs bytecode. The violations pointed out
201 by the verifier pass indicate bugs in transformation passes or input to
204 <!-- Describe the typesetting conventions here. --> </div>
206 <!-- *********************************************************************** -->
207 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
208 <!-- *********************************************************************** -->
210 <div class="doc_text">
212 <p>LLVM uses three different forms of identifiers, for different
216 <li>Numeric constants are represented as you would expect: 12, -3
217 123.421, etc. Floating point constants have an optional hexadecimal
219 <li>Named values are represented as a string of characters with a '%'
220 prefix. For example, %foo, %DivisionByZero,
221 %a.really.long.identifier. The actual regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
222 Identifiers which require other characters in their names can be
223 surrounded with quotes. In this way, anything except a <tt>"</tt>
224 character can be used in a name.</li>
225 <li>Unnamed values are represented as an unsigned numeric value with
226 a '%' prefix. For example, %12, %2, %44.</li>
228 <p>LLVM requires that values start with a '%' sign for two reasons:
229 Compilers don't need to worry about name clashes with reserved words,
230 and the set of reserved words may be expanded in the future without
231 penalty. Additionally, unnamed identifiers allow a compiler to quickly
232 come up with a temporary variable without having to avoid symbol table
234 <p>Reserved words in LLVM are very similar to reserved words in other
235 languages. There are keywords for different opcodes ('<tt><a
236 href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
237 href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
238 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>',
239 etc...), and others. These reserved words cannot conflict with
240 variable names, because none of them start with a '%' character.</p>
241 <p>Here is an example of LLVM code to multiply the integer variable '<tt>%X</tt>'
244 <pre> %result = <a href="#i_mul">mul</a> uint %X, 8<br></pre>
245 <p>After strength reduction:</p>
246 <pre> %result = <a href="#i_shl">shl</a> uint %X, ubyte 3<br></pre>
247 <p>And the hard way:</p>
248 <pre> <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
250 href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
252 href="#i_add">add</a> uint %1, %1<br></pre>
253 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
254 important lexical features of LLVM:</p>
256 <li>Comments are delimited with a '<tt>;</tt>' and go until the end
258 <li>Unnamed temporaries are created when the result of a computation
259 is not assigned to a named value.</li>
260 <li>Unnamed temporaries are numbered sequentially</li>
262 <p>...and it also show a convention that we follow in this document.
263 When demonstrating instructions, we will follow an instruction with a
264 comment that defines the type and name of value produced. Comments are
265 shown in italic text.</p>
266 <p>The one non-intuitive notation for constants is the optional
267 hexidecimal form of floating point constants. For example, the form '<tt>double
268 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
269 4.5e+15</tt>' which is also supported by the parser. The only time
270 hexadecimal floating point constants are useful (and the only time that
271 they are generated by the disassembler) is when an FP constant has to
272 be emitted that is not representable as a decimal floating point number
273 exactly. For example, NaN's, infinities, and other special cases are
274 represented in their IEEE hexadecimal format so that assembly and
275 disassembly do not cause any bits to change in the constants.</p>
277 <!-- *********************************************************************** -->
278 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
279 <!-- *********************************************************************** -->
280 <div class="doc_text">
281 <p>The LLVM type system is one of the most important features of the
282 intermediate representation. Being typed enables a number of
283 optimizations to be performed on the IR directly, without having to do
284 extra analyses on the side before the transformation. A strong type
285 system makes it easier to read the generated code and enables novel
286 analyses and transformations that are not feasible to perform on normal
287 three address code representations.</p>
288 <!-- The written form for the type system was heavily influenced by the
289 syntactic problems with types in the C language<sup><a
290 href="#rw_stroustrup">1</a></sup>.<p> --> </div>
291 <!-- ======================================================================= -->
292 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
293 <div class="doc_text">
294 <p>The primitive types are the fundamental building blocks of the LLVM
295 system. The current set of primitive types are as follows:</p>
297 <table border="0" style="align: center">
301 <table border="1" cellspacing="0" cellpadding="4" style="align: center">
304 <td><tt>void</tt></td>
308 <td><tt>ubyte</tt></td>
309 <td>Unsigned 8 bit value</td>
312 <td><tt>ushort</tt></td>
313 <td>Unsigned 16 bit value</td>
316 <td><tt>uint</tt></td>
317 <td>Unsigned 32 bit value</td>
320 <td><tt>ulong</tt></td>
321 <td>Unsigned 64 bit value</td>
324 <td><tt>float</tt></td>
325 <td>32 bit floating point value</td>
328 <td><tt>label</tt></td>
329 <td>Branch destination</td>
335 <table border="1" cellspacing="0" cellpadding="4">
338 <td><tt>bool</tt></td>
339 <td>True or False value</td>
342 <td><tt>sbyte</tt></td>
343 <td>Signed 8 bit value</td>
346 <td><tt>short</tt></td>
347 <td>Signed 16 bit value</td>
350 <td><tt>int</tt></td>
351 <td>Signed 32 bit value</td>
354 <td><tt>long</tt></td>
355 <td>Signed 64 bit value</td>
358 <td><tt>double</tt></td>
359 <td>64 bit floating point value</td>
369 <!-- _______________________________________________________________________ -->
370 <div class="doc_subsubsection"> <a name="t_classifications">Type
371 Classifications</a> </div>
372 <div class="doc_text">
373 <p>These different primitive types fall into a few useful
376 <table border="1" cellspacing="0" cellpadding="4">
379 <td><a name="t_signed">signed</a></td>
380 <td><tt>sbyte, short, int, long, float, double</tt></td>
383 <td><a name="t_unsigned">unsigned</a></td>
384 <td><tt>ubyte, ushort, uint, ulong</tt></td>
387 <td><a name="t_integer">integer</a></td>
388 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
391 <td><a name="t_integral">integral</a></td>
392 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
395 <td><a name="t_floating">floating point</a></td>
396 <td><tt>float, double</tt></td>
399 <td><a name="t_firstclass">first class</a></td>
400 <td><tt>bool, ubyte, sbyte, ushort, short,<br>
401 uint, int, ulong, long, float, double, <a href="#t_pointer">pointer</a></tt></td>
406 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
407 most important. Values of these types are the only ones which can be
408 produced by instructions, passed as arguments, or used as operands to
409 instructions. This means that all structures and arrays must be
410 manipulated either by pointer or by component.</p>
412 <!-- ======================================================================= -->
413 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
414 <div class="doc_text">
415 <p>The real power in LLVM comes from the derived types in the system.
416 This is what allows a programmer to represent arrays, functions,
417 pointers, and other useful types. Note that these derived types may be
418 recursive: For example, it is possible to have a two dimensional array.</p>
420 <!-- _______________________________________________________________________ -->
421 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
422 <div class="doc_text">
424 <p>The array type is a very simple derived type that arranges elements
425 sequentially in memory. The array type requires a size (number of
426 elements) and an underlying data type.</p>
428 <pre> [<# elements> x <elementtype>]<br></pre>
429 <p>The number of elements is a constant integer value, elementtype may
430 be any type with a size.</p>
432 <p> <tt>[40 x int ]</tt>: Array of 40 integer values.<br>
433 <tt>[41 x int ]</tt>: Array of 41 integer values.<br>
434 <tt>[40 x uint]</tt>: Array of 40 unsigned integer values.</p>
436 <p>Here are some examples of multidimensional arrays:</p>
438 <table border="0" cellpadding="0" cellspacing="0">
441 <td><tt>[3 x [4 x int]]</tt></td>
442 <td>: 3x4 array integer values.</td>
445 <td><tt>[12 x [10 x float]]</tt></td>
446 <td>: 12x10 array of single precision floating point values.</td>
449 <td><tt>[2 x [3 x [4 x uint]]]</tt></td>
450 <td>: 2x3x4 array of unsigned integer values.</td>
456 <!-- _______________________________________________________________________ -->
457 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
458 <div class="doc_text">
460 <p>The function type can be thought of as a function signature. It
461 consists of a return type and a list of formal parameter types.
462 Function types are usually used to build virtual function tables
463 (which are structures of pointers to functions), for indirect function
464 calls, and when defining a function.</p>
466 The return type of a function type cannot be an aggregate type.
469 <pre> <returntype> (<parameter list>)<br></pre>
470 <p>Where '<tt><parameter list></tt>' is a comma-separated list of
471 type specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
472 which indicates that the function takes a variable number of arguments.
473 Variable argument functions can access their arguments with the <a
474 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
477 <table border="0" cellpadding="0" cellspacing="0">
480 <td><tt>int (int)</tt></td>
481 <td>: function taking an <tt>int</tt>, returning an <tt>int</tt></td>
484 <td><tt>float (int, int *) *</tt></td>
485 <td>: <a href="#t_pointer">Pointer</a> to a function that takes
486 an <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
487 returning <tt>float</tt>.</td>
490 <td><tt>int (sbyte *, ...)</tt></td>
491 <td>: A vararg function that takes at least one <a
492 href="#t_pointer">pointer</a> to <tt>sbyte</tt> (signed char in C),
493 which returns an integer. This is the signature for <tt>printf</tt>
500 <!-- _______________________________________________________________________ -->
501 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
502 <div class="doc_text">
504 <p>The structure type is used to represent a collection of data members
505 together in memory. The packing of the field types is defined to match
506 the ABI of the underlying processor. The elements of a structure may
507 be any type that has a size.</p>
508 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
509 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
510 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
513 <pre> { <type list> }<br></pre>
516 <table border="0" cellpadding="0" cellspacing="0">
519 <td><tt>{ int, int, int }</tt></td>
520 <td>: a triple of three <tt>int</tt> values</td>
523 <td><tt>{ float, int (int) * }</tt></td>
524 <td>: A pair, where the first element is a <tt>float</tt> and the
525 second element is a <a href="#t_pointer">pointer</a> to a <a
526 href="#t_function">function</a> that takes an <tt>int</tt>, returning
527 an <tt>int</tt>.</td>
533 <!-- _______________________________________________________________________ -->
534 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
535 <div class="doc_text">
537 <p>As in many languages, the pointer type represents a pointer or
538 reference to another object, which must live in memory.</p>
540 <pre> <type> *<br></pre>
543 <table border="0" cellpadding="0" cellspacing="0">
546 <td><tt>[4x int]*</tt></td>
547 <td>: <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a>
548 of four <tt>int</tt> values</td>
551 <td><tt>int (int *) *</tt></td>
552 <td>: A <a href="#t_pointer">pointer</a> to a <a
553 href="#t_function">function</a> that takes an <tt>int</tt>, returning
554 an <tt>int</tt>.</td>
560 <!-- _______________________________________________________________________ --><!--
561 <div class="doc_subsubsection">
562 <a name="t_packed">Packed Type</a>
565 <div class="doc_text">
567 Mention/decide that packed types work with saturation or not. Maybe have a packed+saturated type in addition to just a packed type.<p>
569 Packed types should be 'nonsaturated' because standard data types are not saturated. Maybe have a saturated packed type?<p>
573 --><!-- *********************************************************************** -->
574 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
575 <!-- *********************************************************************** --><!-- ======================================================================= -->
576 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a> </div>
577 <div class="doc_text">
578 <p>LLVM programs are composed of "Module"s, each of which is a
579 translation unit of the input programs. Each module consists of
580 functions, global variables, and symbol table entries. Modules may be
581 combined together with the LLVM linker, which merges function (and
582 global variable) definitions, resolves forward declarations, and merges
583 symbol table entries. Here is an example of the "hello world" module:</p>
584 <pre><i>; Declare the string constant as a global constant...</i>
585 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
586 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
588 <i>; External declaration of the puts function</i>
589 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
591 <i>; Definition of main function</i>
592 int %main() { <i>; int()* </i>
593 <i>; Convert [13x sbyte]* to sbyte *...</i>
595 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
597 <i>; Call puts function to write out the string to stdout...</i>
599 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
601 href="#i_ret">ret</a> int 0<br>}<br></pre>
602 <p>This example is made up of a <a href="#globalvars">global variable</a>
603 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
604 function, and a <a href="#functionstructure">function definition</a>
605 for "<tt>main</tt>".</p>
606 <a name="linkage"> In general, a module is made up of a list of global
607 values, where both functions and global variables are global values.
608 Global values are represented by a pointer to a memory location (in
609 this case, a pointer to an array of char, and a pointer to a function),
610 and have one of the following linkage types:</a>
613 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
614 <dd>Global values with internal linkage are only directly accessible
615 by objects in the current module. In particular, linking code into a
616 module with an internal global value may cause the internal to be
617 renamed as necessary to avoid collisions. Because the symbol is
618 internal to the module, all references can be updated. This
619 corresponds to the notion of the '<tt>static</tt>' keyword in C, or the
620 idea of "anonymous namespaces" in C++.
623 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
624 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt>
625 linkage, with the twist that linking together two modules defining the
626 same <tt>linkonce</tt> globals will cause one of the globals to be
627 discarded. This is typically used to implement inline functions.
628 Unreferenced <tt>linkonce</tt> globals are allowed to be discarded.
631 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
632 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt>
633 linkage, except that unreferenced <tt>weak</tt> globals may not be
634 discarded. This is used to implement constructs in C such as "<tt>int
635 X;</tt>" at global scope.
638 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
639 <dd>"<tt>appending</tt>" linkage may only be applied to global
640 variables of pointer to array type. When two global variables with
641 appending linkage are linked together, the two global arrays are
642 appended together. This is the LLVM, typesafe, equivalent of having
643 the system linker append together "sections" with identical names when
647 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
648 <dd>If none of the above identifiers are used, the global is
649 externally visible, meaning that it participates in linkage and can be
650 used to resolve external symbol references.
655 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
656 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
657 variable and was linked with this one, one of the two would be renamed,
658 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
659 external (i.e., lacking any linkage declarations), they are accessible
660 outside of the current module. It is illegal for a function <i>declaration</i>
661 to have any linkage type other than "externally visible".</a></p>
664 <!-- ======================================================================= -->
665 <div class="doc_subsection">
666 <a name="globalvars">Global Variables</a>
669 <div class="doc_text">
671 <p>Global variables define regions of memory allocated at compilation
672 time instead of run-time. Global variables may optionally be
673 initialized. A variable may be defined as a global "constant", which
674 indicates that the contents of the variable will never be modified
675 (opening options for optimization).</p>
677 <p>As SSA values, global variables define pointer values that are in
678 scope (i.e. they dominate) for all basic blocks in the program. Global
679 variables always define a pointer to their "content" type because they
680 describe a region of memory, and all memory objects in LLVM are
681 accessed through pointers.</p>
686 <!-- ======================================================================= -->
687 <div class="doc_subsection">
688 <a name="functionstructure">Functions</a>
691 <div class="doc_text">
693 <p>LLVM function definitions are composed of a (possibly empty) argument list,
694 an opening curly brace, a list of basic blocks, and a closing curly brace. LLVM
695 function declarations are defined with the "<tt>declare</tt>" keyword, a
696 function name, and a function signature.</p>
698 <p>A function definition contains a list of basic blocks, forming the CFG for
699 the function. Each basic block may optionally start with a label (giving the
700 basic block a symbol table entry), contains a list of instructions, and ends
701 with a <a href="#terminators">terminator</a> instruction (such as a branch or
702 function return).</p>
704 <p>The first basic block in program is special in two ways: it is immediately
705 executed on entrance to the function, and it is not allowed to have predecessor
706 basic blocks (i.e. there can not be any branches to the entry block of a
707 function). Because the block can have no predecessors, it also cannot have any
708 <a href="#i_phi">PHI nodes</a>.</p>
710 <p>LLVM functions are identified by their name and type signature. Hence, two
711 functions with the same name but different parameter lists or return values are
712 considered different functions, and LLVM will resolves references to each
718 <!-- *********************************************************************** -->
719 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
720 <!-- *********************************************************************** -->
721 <div class="doc_text">
722 <p>The LLVM instruction set consists of several different
723 classifications of instructions: <a href="#terminators">terminator
724 instructions</a>, <a href="#binaryops">binary instructions</a>, <a
725 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
726 instructions</a>.</p>
728 <!-- ======================================================================= -->
729 <div class="doc_subsection"> <a name="terminators">Terminator
730 Instructions</a> </div>
731 <div class="doc_text">
732 <p>As mentioned <a href="#functionstructure">previously</a>, every
733 basic block in a program ends with a "Terminator" instruction, which
734 indicates which block should be executed after the current block is
735 finished. These terminator instructions typically yield a '<tt>void</tt>'
736 value: they produce control flow, not values (the one exception being
737 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
738 <p>There are five different terminator instructions: the '<a
739 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
740 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
741 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, and the '<a
742 href="#i_unwind"><tt>unwind</tt></a>' instruction.</p>
744 <!-- _______________________________________________________________________ -->
745 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
746 Instruction</a> </div>
747 <div class="doc_text">
749 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
750 ret void <i>; Return from void function</i>
753 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
754 value) from a function, back to the caller.</p>
755 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
756 returns a value and then causes control flow, and one that just causes
757 control flow to occur.</p>
759 <p>The '<tt>ret</tt>' instruction may return any '<a
760 href="#t_firstclass">first class</a>' type. Notice that a function is
761 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
762 instruction inside of the function that returns a value that does not
763 match the return type of the function.</p>
765 <p>When the '<tt>ret</tt>' instruction is executed, control flow
766 returns back to the calling function's context. If the caller is a "<a
767 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
768 the instruction after the call. If the caller was an "<a
769 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
770 at the beginning "normal" of the destination block. If the instruction
771 returns a value, that value shall set the call or invoke instruction's
774 <pre> ret int 5 <i>; Return an integer value of 5</i>
775 ret void <i>; Return from a void function</i>
778 <!-- _______________________________________________________________________ -->
779 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
780 <div class="doc_text">
782 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
785 <p>The '<tt>br</tt>' instruction is used to cause control flow to
786 transfer to a different basic block in the current function. There are
787 two forms of this instruction, corresponding to a conditional branch
788 and an unconditional branch.</p>
790 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
791 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
792 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
793 value as a target.</p>
795 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
796 argument is evaluated. If the value is <tt>true</tt>, control flows
797 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
798 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
800 <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
801 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
803 <!-- _______________________________________________________________________ -->
804 <div class="doc_subsubsection">
805 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
808 <div class="doc_text">
812 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
817 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
818 several different places. It is a generalization of the '<tt>br</tt>'
819 instruction, allowing a branch to occur to one of many possible
825 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
826 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
827 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
828 table is not allowed to contain duplicate constant entries.</p>
832 <p>The <tt>switch</tt> instruction specifies a table of values and
833 destinations. When the '<tt>switch</tt>' instruction is executed, this
834 table is searched for the given value. If the value is found, control flow is
835 transfered to the corresponding destination; otherwise, control flow is
836 transfered to the default destination.</p>
838 <h5>Implementation:</h5>
840 <p>Depending on properties of the target machine and the particular
841 <tt>switch</tt> instruction, this instruction may be code generated in different
842 ways. For example, it could be generated as a series of chained conditional
843 branches or with a lookup table.</p>
848 <i>; Emulate a conditional br instruction</i>
849 %Val = <a href="#i_cast">cast</a> bool %value to int
850 switch int %Val, label %truedest [int 0, label %falsedest ]
852 <i>; Emulate an unconditional br instruction</i>
853 switch uint 0, label %dest [ ]
855 <i>; Implement a jump table:</i>
856 switch uint %val, label %otherwise [ uint 0, label %onzero
858 uint 2, label %ontwo ]
861 <!-- _______________________________________________________________________ -->
862 <div class="doc_subsubsection"> <a name="i_invoke">'<tt>invoke</tt>'
863 Instruction</a> </div>
864 <div class="doc_text">
866 <pre> <result> = invoke <ptr to function ty> %<function ptr val>(<function args>)<br> to label <normal label> except label <exception label><br></pre>
868 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a
869 specified function, with the possibility of control flow transfer to
870 either the '<tt>normal</tt>' <tt>label</tt> label or the '<tt>exception</tt>'<tt>label</tt>.
871 If the callee function returns with the "<tt><a href="#i_ret">ret</a></tt>"
872 instruction, control flow will return to the "normal" label. If the
873 callee (or any indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
874 instruction, control is interrupted, and continued at the dynamically
875 nearest "except" label.</p>
877 <p>This instruction requires several arguments:</p>
879 <li>'<tt>ptr to function ty</tt>': shall be the signature of the
880 pointer to function value being invoked. In most cases, this is a
881 direct function invocation, but indirect <tt>invoke</tt>s are just as
882 possible, branching off an arbitrary pointer to function value. </li>
883 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer
884 to a function to be invoked. </li>
885 <li>'<tt>function args</tt>': argument list whose types match the
886 function signature argument types. If the function signature indicates
887 the function accepts a variable number of arguments, the extra
888 arguments can be specified. </li>
889 <li>'<tt>normal label</tt>': the label reached when the called
890 function executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
891 <li>'<tt>exception label</tt>': the label reached when a callee
892 returns with the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
895 <p>This instruction is designed to operate as a standard '<tt><a
896 href="#i_call">call</a></tt>' instruction in most regards. The
897 primary difference is that it establishes an association with a label,
898 which is used by the runtime library to unwind the stack.</p>
899 <p>This instruction is used in languages with destructors to ensure
900 that proper cleanup is performed in the case of either a <tt>longjmp</tt>
901 or a thrown exception. Additionally, this is important for
902 implementation of '<tt>catch</tt>' clauses in high-level languages that
905 <pre> %retval = invoke int %Test(int 15)<br> to label %Continue<br> except label %TestCleanup <i>; {int}:retval set</i>
908 <!-- _______________________________________________________________________ -->
909 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
910 Instruction</a> </div>
911 <div class="doc_text">
913 <pre> unwind<br></pre>
915 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing
916 control flow at the first callee in the dynamic call stack which used
917 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the
918 call. This is primarily used to implement exception handling.</p>
920 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current
921 function to immediately halt. The dynamic call stack is then searched
922 for the first <a href="#i_invoke"><tt>invoke</tt></a> instruction on
923 the call stack. Once found, execution continues at the "exceptional"
924 destination block specified by the <tt>invoke</tt> instruction. If
925 there is no <tt>invoke</tt> instruction in the dynamic call chain,
926 undefined behavior results.</p>
928 <!-- ======================================================================= -->
929 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
930 <div class="doc_text">
931 <p>Binary operators are used to do most of the computation in a
932 program. They require two operands, execute an operation on them, and
933 produce a single value. The result value of a binary operator is not
934 necessarily the same type as its operands.</p>
935 <p>There are several different binary operators:</p>
937 <!-- _______________________________________________________________________ -->
938 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
939 Instruction</a> </div>
940 <div class="doc_text">
942 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
945 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
947 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
948 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
949 values. Both arguments must have identical types.</p>
951 <p>The value produced is the integer or floating point sum of the two
954 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
957 <!-- _______________________________________________________________________ -->
958 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
959 Instruction</a> </div>
960 <div class="doc_text">
962 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
965 <p>The '<tt>sub</tt>' instruction returns the difference of its two
967 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
968 instruction present in most other intermediate representations.</p>
970 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
971 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
972 values. Both arguments must have identical types.</p>
974 <p>The value produced is the integer or floating point difference of
975 the two operands.</p>
977 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
978 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
981 <!-- _______________________________________________________________________ -->
982 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
983 Instruction</a> </div>
984 <div class="doc_text">
986 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
989 <p>The '<tt>mul</tt>' instruction returns the product of its two
992 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
993 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
994 values. Both arguments must have identical types.</p>
996 <p>The value produced is the integer or floating point product of the
998 <p>There is no signed vs unsigned multiplication. The appropriate
999 action is taken based on the type of the operand.</p>
1001 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
1004 <!-- _______________________________________________________________________ -->
1005 <div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
1006 Instruction</a> </div>
1007 <div class="doc_text">
1009 <pre> <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1012 <p>The '<tt>div</tt>' instruction returns the quotient of its two
1015 <p>The two arguments to the '<tt>div</tt>' instruction must be either <a
1016 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1017 values. Both arguments must have identical types.</p>
1019 <p>The value produced is the integer or floating point quotient of the
1022 <pre> <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
1025 <!-- _______________________________________________________________________ -->
1026 <div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
1027 Instruction</a> </div>
1028 <div class="doc_text">
1030 <pre> <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1033 <p>The '<tt>rem</tt>' instruction returns the remainder from the
1034 division of its two operands.</p>
1036 <p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1037 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1038 values. Both arguments must have identical types.</p>
1040 <p>This returns the <i>remainder</i> of a division (where the result
1041 has the same sign as the divisor), not the <i>modulus</i> (where the
1042 result has the same sign as the dividend) of a value. For more
1043 information about the difference, see: <a
1044 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1047 <pre> <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1050 <!-- _______________________________________________________________________ -->
1051 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1052 Instructions</a> </div>
1053 <div class="doc_text">
1055 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1056 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1057 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1058 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1059 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1060 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1063 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1064 value based on a comparison of their two operands.</p>
1066 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1067 be of <a href="#t_firstclass">first class</a> type (it is not possible
1068 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1069 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1072 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1073 value if both operands are equal.<br>
1074 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1075 value if both operands are unequal.<br>
1076 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1077 value if the first operand is less than the second operand.<br>
1078 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1079 value if the first operand is greater than the second operand.<br>
1080 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1081 value if the first operand is less than or equal to the second operand.<br>
1082 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1083 value if the first operand is greater than or equal to the second
1086 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1087 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1088 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1089 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1090 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1091 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1094 <!-- ======================================================================= -->
1095 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1096 Operations</a> </div>
1097 <div class="doc_text">
1098 <p>Bitwise binary operators are used to do various forms of
1099 bit-twiddling in a program. They are generally very efficient
1100 instructions, and can commonly be strength reduced from other
1101 instructions. They require two operands, execute an operation on them,
1102 and produce a single value. The resulting value of the bitwise binary
1103 operators is always the same type as its first operand.</p>
1105 <!-- _______________________________________________________________________ -->
1106 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1107 Instruction</a> </div>
1108 <div class="doc_text">
1110 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1113 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1114 its two operands.</p>
1116 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1117 href="#t_integral">integral</a> values. Both arguments must have
1118 identical types.</p>
1120 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1122 <div style="align: center">
1123 <table border="1" cellspacing="0" cellpadding="4">
1154 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1155 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1156 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1159 <!-- _______________________________________________________________________ -->
1160 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1161 <div class="doc_text">
1163 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1166 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1167 or of its two operands.</p>
1169 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1170 href="#t_integral">integral</a> values. Both arguments must have
1171 identical types.</p>
1173 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1175 <div style="align: center">
1176 <table border="1" cellspacing="0" cellpadding="4">
1207 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1208 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1209 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1212 <!-- _______________________________________________________________________ -->
1213 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1214 Instruction</a> </div>
1215 <div class="doc_text">
1217 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1220 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1221 or of its two operands. The <tt>xor</tt> is used to implement the
1222 "one's complement" operation, which is the "~" operator in C.</p>
1224 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1225 href="#t_integral">integral</a> values. Both arguments must have
1226 identical types.</p>
1228 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1230 <div style="align: center">
1231 <table border="1" cellspacing="0" cellpadding="4">
1263 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1264 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1265 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1266 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1269 <!-- _______________________________________________________________________ -->
1270 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1271 Instruction</a> </div>
1272 <div class="doc_text">
1274 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1277 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1278 the left a specified number of bits.</p>
1280 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
1281 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1284 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
1286 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1287 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1288 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1291 <!-- _______________________________________________________________________ -->
1292 <div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1293 Instruction</a> </div>
1294 <div class="doc_text">
1296 <pre> <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1299 <p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1300 the right a specified number of bits.</p>
1302 <p>The first argument to the '<tt>shr</tt>' instruction must be an <a
1303 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1306 <p>If the first argument is a <a href="#t_signed">signed</a> type, the
1307 most significant bit is duplicated in the newly free'd bit positions.
1308 If the first argument is unsigned, zero bits shall fill the empty
1311 <pre> <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1312 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1313 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1314 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1315 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1318 <!-- ======================================================================= -->
1319 <div class="doc_subsection"> <a name="memoryops">Memory Access
1320 Operations</a></div>
1321 <div class="doc_text">
1322 <p>A key design point of an SSA-based representation is how it
1323 represents memory. In LLVM, no memory locations are in SSA form, which
1324 makes things very simple. This section describes how to read, write,
1325 allocate and free memory in LLVM.</p>
1327 <!-- _______________________________________________________________________ -->
1328 <div class="doc_subsubsection"> <a name="i_malloc">'<tt>malloc</tt>'
1329 Instruction</a> </div>
1330 <div class="doc_text">
1332 <pre> <result> = malloc <type>, uint <NumElements> <i>; yields {type*}:result</i>
1333 <result> = malloc <type> <i>; yields {type*}:result</i>
1336 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
1337 heap and returns a pointer to it.</p>
1339 <p>The '<tt>malloc</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1340 bytes of memory from the operating system and returns a pointer of the
1341 appropriate type to the program. The second form of the instruction is
1342 a shorter version of the first instruction that defaults to allocating
1344 <p>'<tt>type</tt>' must be a sized type.</p>
1346 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
1347 a pointer is returned.</p>
1349 <pre> %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1352 href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1353 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1354 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1357 <!-- _______________________________________________________________________ -->
1358 <div class="doc_subsubsection"> <a name="i_free">'<tt>free</tt>'
1359 Instruction</a> </div>
1360 <div class="doc_text">
1362 <pre> free <type> <value> <i>; yields {void}</i>
1365 <p>The '<tt>free</tt>' instruction returns memory back to the unused
1366 memory heap, to be reallocated in the future.</p>
1369 <p>'<tt>value</tt>' shall be a pointer value that points to a value
1370 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
1373 <p>Access to the memory pointed to by the pointer is not longer defined
1374 after this instruction executes.</p>
1376 <pre> %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1377 free [4 x ubyte]* %array
1380 <!-- _______________________________________________________________________ -->
1381 <div class="doc_subsubsection"> <a name="i_alloca">'<tt>alloca</tt>'
1382 Instruction</a> </div>
1383 <div class="doc_text">
1385 <pre> <result> = alloca <type>, uint <NumElements> <i>; yields {type*}:result</i>
1386 <result> = alloca <type> <i>; yields {type*}:result</i>
1389 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
1390 stack frame of the procedure that is live until the current function
1391 returns to its caller.</p>
1393 <p>The the '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1394 bytes of memory on the runtime stack, returning a pointer of the
1395 appropriate type to the program. The second form of the instruction is
1396 a shorter version of the first that defaults to allocating one element.</p>
1397 <p>'<tt>type</tt>' may be any sized type.</p>
1399 <p>Memory is allocated, a pointer is returned. '<tt>alloca</tt>'d
1400 memory is automatically released when the function returns. The '<tt>alloca</tt>'
1401 instruction is commonly used to represent automatic variables that must
1402 have an address available. When the function returns (either with the <tt><a
1403 href="#i_ret">ret</a></tt> or <tt><a href="#i_invoke">invoke</a></tt>
1404 instructions), the memory is reclaimed.</p>
1406 <pre> %ptr = alloca int <i>; yields {int*}:ptr</i>
1407 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
1410 <!-- _______________________________________________________________________ -->
1411 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
1412 Instruction</a> </div>
1413 <div class="doc_text">
1415 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
1417 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
1419 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
1420 address to load from. The pointer must point to a <a
1421 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
1422 marked as <tt>volatile</tt> then the optimizer is not allowed to modify
1423 the number or order of execution of this <tt>load</tt> with other
1424 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
1427 <p>The location of memory pointed to is loaded.</p>
1429 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1431 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1432 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1435 <!-- _______________________________________________________________________ -->
1436 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
1437 Instruction</a> </div>
1439 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1440 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1443 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
1445 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
1446 to store and an address to store it into. The type of the '<tt><pointer></tt>'
1447 operand must be a pointer to the type of the '<tt><value></tt>'
1448 operand. If the <tt>store</tt> is marked as <tt>volatile</tt> then the
1449 optimizer is not allowed to modify the number or order of execution of
1450 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
1451 href="#i_store">store</a></tt> instructions.</p>
1453 <p>The contents of memory are updated to contain '<tt><value></tt>'
1454 at the location specified by the '<tt><pointer></tt>' operand.</p>
1456 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1458 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1459 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1461 <!-- _______________________________________________________________________ -->
1462 <div class="doc_subsubsection">
1463 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
1466 <div class="doc_text">
1469 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
1475 The '<tt>getelementptr</tt>' instruction is used to get the address of a
1476 subelement of an aggregate data structure.</p>
1480 <p>This instruction takes a list of integer constants that indicate what
1481 elements of the aggregate object to index to. The actual types of the arguments
1482 provided depend on the type of the first pointer argument. The
1483 '<tt>getelementptr</tt>' instruction is used to index down through the type
1484 levels of a structure. When indexing into a structure, only <tt>uint</tt>
1485 integer constants are allowed. When indexing into an array or pointer
1486 <tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
1488 <p>For example, let's consider a C code fragment and how it gets
1489 compiled to LLVM:</p>
1503 int *foo(struct ST *s) {
1504 return &s[1].Z.B[5][13];
1508 <p>The LLVM code generated by the GCC frontend is:</p>
1511 %RT = type { sbyte, [10 x [20 x int]], sbyte }
1512 %ST = type { int, double, %RT }
1516 int* %foo(%ST* %s) {
1518 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
1525 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
1526 on the pointer type that is being index into. <a href="#t_pointer">Pointer</a>
1527 and <a href="#t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
1528 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="#t_struct">structure</a>
1529 types require <tt>uint</tt> <b>constants</b>.</p>
1531 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
1532 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
1533 }</tt>' type, a structure. The second index indexes into the third element of
1534 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
1535 sbyte }</tt>' type, another structure. The third index indexes into the second
1536 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
1537 array. The two dimensions of the array are subscripted into, yielding an
1538 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction return a pointer
1539 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
1541 <p>Note that it is perfectly legal to index partially through a
1542 structure, returning a pointer to an inner element. Because of this,
1543 the LLVM code for the given testcase is equivalent to:</p>
1546 int* "foo"(%ST* %s) {
1547 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
1548 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
1549 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
1550 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
1551 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
1557 <i>; yields [12 x ubyte]*:aptr</i>
1558 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
1562 <!-- ======================================================================= -->
1563 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
1564 <div class="doc_text">
1565 <p>The instructions in this category are the "miscellaneous"
1566 instructions, which defy better classification.</p>
1568 <!-- _______________________________________________________________________ -->
1569 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
1570 Instruction</a> </div>
1571 <div class="doc_text">
1573 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
1575 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
1576 the SSA graph representing the function.</p>
1578 <p>The type of the incoming values are specified with the first type
1579 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
1580 as arguments, with one pair for each predecessor basic block of the
1581 current block. Only values of <a href="#t_firstclass">first class</a>
1582 type may be used as the value arguments to the PHI node. Only labels
1583 may be used as the label arguments.</p>
1584 <p>There must be no non-phi instructions between the start of a basic
1585 block and the PHI instructions: i.e. PHI instructions must be first in
1588 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
1589 value specified by the parameter, depending on which basic block we
1590 came from in the last <a href="#terminators">terminator</a> instruction.</p>
1592 <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>
1595 <!-- _______________________________________________________________________ -->
1596 <div class="doc_subsubsection">
1597 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
1600 <div class="doc_text">
1605 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
1611 The '<tt>cast</tt>' instruction is used as the primitive means to convert
1612 integers to floating point, change data type sizes, and break type safety (by
1620 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
1621 class value, and a type to cast it to, which must also be a <a
1622 href="#t_firstclass">first class</a> type.
1628 This instruction follows the C rules for explicit casts when determining how the
1629 data being cast must change to fit in its new container.
1633 When casting to bool, any value that would be considered true in the context of
1634 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
1635 all else are '<tt>false</tt>'.
1639 When extending an integral value from a type of one signness to another (for
1640 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
1641 <b>source</b> value is signed, and zero-extended if the source value is
1642 unsigned. <tt>bool</tt> values are always zero extended into either zero or
1649 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
1650 %Y = cast int 123 to bool <i>; yields bool:true</i>
1654 <!-- _______________________________________________________________________ -->
1655 <div class="doc_subsubsection">
1656 <a name="i_select">'<tt>select</tt>' Instruction</a>
1659 <div class="doc_text">
1664 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
1670 The '<tt>select</tt>' instruction is used to choose one value based on a
1671 condition, without branching.
1678 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.
1684 If the boolean condition evaluates to true, the instruction returns the first
1685 value argument, otherwise it returns the second value argument.
1691 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
1699 <!-- _______________________________________________________________________ -->
1700 <div class="doc_subsubsection"> <a name="i_call">'<tt>call</tt>'
1701 Instruction</a> </div>
1702 <div class="doc_text">
1704 <pre> <result> = call <ty>* <fnptrval>(<param list>)<br></pre>
1706 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
1708 <p>This instruction requires several arguments:</p>
1711 <p>'<tt>ty</tt>': shall be the signature of the pointer to function
1712 value being invoked. The argument types must match the types implied
1713 by this signature.</p>
1716 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a
1717 function to be invoked. In most cases, this is a direct function
1718 invocation, but indirect <tt>call</tt>s are just as possible,
1719 calling an arbitrary pointer to function values.</p>
1722 <p>'<tt>function args</tt>': argument list whose types match the
1723 function signature argument types. If the function signature
1724 indicates the function accepts a variable number of arguments, the
1725 extra arguments can be specified.</p>
1729 <p>The '<tt>call</tt>' instruction is used to cause control flow to
1730 transfer to a specified function, with its incoming arguments bound to
1731 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
1732 instruction in the called function, control flow continues with the
1733 instruction after the function call, and the return value of the
1734 function is bound to the result argument. This is a simpler case of
1735 the <a href="#i_invoke">invoke</a> instruction.</p>
1737 <pre> %retval = call int %test(int %argc)<br> call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);<br></pre>
1739 <!-- _______________________________________________________________________ -->
1740 <div class="doc_subsubsection"> <a name="i_vanext">'<tt>vanext</tt>'
1741 Instruction</a> </div>
1742 <div class="doc_text">
1744 <pre> <resultarglist> = vanext <va_list> <arglist>, <argty><br></pre>
1746 <p>The '<tt>vanext</tt>' instruction is used to access arguments passed
1747 through the "variable argument" area of a function call. It is used to
1748 implement the <tt>va_arg</tt> macro in C.</p>
1750 <p>This instruction takes a <tt>valist</tt> value and the type of the
1751 argument. It returns another <tt>valist</tt>.</p>
1753 <p>The '<tt>vanext</tt>' instruction advances the specified <tt>valist</tt>
1754 past an argument of the specified type. In conjunction with the <a
1755 href="#i_vaarg"><tt>vaarg</tt></a> instruction, it is used to implement
1756 the <tt>va_arg</tt> macro available in C. For more information, see
1757 the variable argument handling <a href="#int_varargs">Intrinsic
1759 <p>It is legal for this instruction to be called in a function which
1760 does not take a variable number of arguments, for example, the <tt>vfprintf</tt>
1762 <p><tt>vanext</tt> is an LLVM instruction instead of an <a
1763 href="#intrinsics">intrinsic function</a> because it takes an type as
1766 <p>See the <a href="#int_varargs">variable argument processing</a>
1769 <!-- _______________________________________________________________________ -->
1770 <div class="doc_subsubsection"> <a name="i_vaarg">'<tt>vaarg</tt>'
1771 Instruction</a> </div>
1772 <div class="doc_text">
1774 <pre> <resultval> = vaarg <va_list> <arglist>, <argty><br></pre>
1776 <p>The '<tt>vaarg</tt>' instruction is used to access arguments passed
1777 through the "variable argument" area of a function call. It is used to
1778 implement the <tt>va_arg</tt> macro in C.</p>
1780 <p>This instruction takes a <tt>valist</tt> value and the type of the
1781 argument. It returns a value of the specified argument type.</p>
1783 <p>The '<tt>vaarg</tt>' instruction loads an argument of the specified
1784 type from the specified <tt>va_list</tt>. In conjunction with the <a
1785 href="#i_vanext"><tt>vanext</tt></a> instruction, it is used to
1786 implement the <tt>va_arg</tt> macro available in C. For more
1787 information, see the variable argument handling <a href="#int_varargs">Intrinsic
1789 <p>It is legal for this instruction to be called in a function which
1790 does not take a variable number of arguments, for example, the <tt>vfprintf</tt>
1792 <p><tt>vaarg</tt> is an LLVM instruction instead of an <a
1793 href="#intrinsics">intrinsic function</a> because it takes an type as
1796 <p>See the <a href="#int_varargs">variable argument processing</a>
1800 <!-- *********************************************************************** -->
1801 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
1802 <!-- *********************************************************************** -->
1804 <div class="doc_text">
1806 <p>LLVM supports the notion of an "intrinsic function". These functions have
1807 well known names and semantics, and are required to follow certain
1808 restrictions. Overall, these instructions represent an extension mechanism for
1809 the LLVM language that does not require changing all of the transformations in
1810 LLVM to add to the language (or the bytecode reader/writer, the parser,
1813 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix, this
1814 prefix is reserved in LLVM for intrinsic names, thus functions may not be named
1815 this. Intrinsic functions must always be external functions: you cannot define
1816 the body of intrinsic functions. Intrinsic functions may only be used in call
1817 or invoke instructions: it is illegal to take the address of an intrinsic
1818 function. Additionally, because intrinsic functions are part of the LLVM
1819 language, it is required that they all be documented here if any are added.</p>
1823 Adding an intrinsic to LLVM is straight-forward if it is possible to express the
1824 concept in LLVM directly (ie, code generator support is not _required_). To do
1825 this, extend the default implementation of the IntrinsicLowering class to handle
1826 the intrinsic. Code generators use this class to lower intrinsics they do not
1827 understand to raw LLVM instructions that they do.
1832 <!-- ======================================================================= -->
1833 <div class="doc_subsection">
1834 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
1837 <div class="doc_text">
1839 <p>Variable argument support is defined in LLVM with the <a
1840 href="#i_vanext"><tt>vanext</tt></a> instruction and these three
1841 intrinsic functions. These functions are related to the similarly
1842 named macros defined in the <tt><stdarg.h></tt> header file.</p>
1844 <p>All of these functions operate on arguments that use a
1845 target-specific value type "<tt>va_list</tt>". The LLVM assembly
1846 language reference manual does not define what this type is, so all
1847 transformations should be prepared to handle intrinsics with any type
1850 <p>This example shows how the <a href="#i_vanext"><tt>vanext</tt></a>
1851 instruction and the variable argument handling intrinsic functions are
1855 int %test(int %X, ...) {
1856 ; Initialize variable argument processing
1857 %ap = call sbyte* %<a href="#i_va_start">llvm.va_start</a>()
1859 ; Read a single integer argument
1860 %tmp = vaarg sbyte* %ap, int
1862 ; Advance to the next argument
1863 %ap2 = vanext sbyte* %ap, int
1865 ; Demonstrate usage of llvm.va_copy and llvm.va_end
1866 %aq = call sbyte* %<a href="#i_va_copy">llvm.va_copy</a>(sbyte* %ap2)
1867 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %aq)
1869 ; Stop processing of arguments.
1870 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %ap2)
1876 <!-- _______________________________________________________________________ -->
1877 <div class="doc_subsubsection">
1878 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
1882 <div class="doc_text">
1884 <pre> call va_list ()* %llvm.va_start()<br></pre>
1886 <p>The '<tt>llvm.va_start</tt>' intrinsic returns a new <tt><arglist></tt>
1887 for subsequent use by the variable argument intrinsics.</p>
1889 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
1890 macro available in C. In a target-dependent way, it initializes and
1891 returns a <tt>va_list</tt> element, so that the next <tt>vaarg</tt>
1892 will produce the first variable argument passed to the function. Unlike
1893 the C <tt>va_start</tt> macro, this intrinsic does not need to know the
1894 last argument of the function, the compiler can figure that out.</p>
1895 <p>Note that this intrinsic function is only legal to be called from
1896 within the body of a variable argument function.</p>
1899 <!-- _______________________________________________________________________ -->
1900 <div class="doc_subsubsection">
1901 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
1904 <div class="doc_text">
1906 <pre> call void (va_list)* %llvm.va_end(va_list <arglist>)<br></pre>
1908 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
1909 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
1910 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
1912 <p>The argument is a <tt>va_list</tt> to destroy.</p>
1914 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
1915 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
1916 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
1917 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
1918 with calls to <tt>llvm.va_end</tt>.</p>
1921 <!-- _______________________________________________________________________ -->
1922 <div class="doc_subsubsection">
1923 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
1926 <div class="doc_text">
1931 call va_list (va_list)* %llvm.va_copy(va_list <destarglist>)
1936 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
1937 from the source argument list to the destination argument list.</p>
1941 <p>The argument is the <tt>va_list</tt> to copy.</p>
1945 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
1946 macro available in C. In a target-dependent way, it copies the source
1947 <tt>va_list</tt> element into the returned list. This intrinsic is necessary
1948 because the <tt><a href="#i_va_start">llvm.va_start</a></tt> intrinsic may be
1949 arbitrarily complex and require memory allocation, for example.</p>
1953 <!-- ======================================================================= -->
1954 <div class="doc_subsection">
1955 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
1958 <div class="doc_text">
1961 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
1962 Collection</a> requires the implementation and generation of these intrinsics.
1963 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
1964 stack</a>, as well as garbage collector implementations that require <a
1965 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
1966 Front-ends for type-safe garbage collected languages should generate these
1967 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
1968 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
1972 <!-- _______________________________________________________________________ -->
1973 <div class="doc_subsubsection">
1974 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
1977 <div class="doc_text">
1982 call void (<ty>**, <ty2>*)* %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
1987 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existance of a GC root to
1988 the code generator, and allows some metadata to be associated with it.</p>
1992 <p>The first argument specifies the address of a stack object that contains the
1993 root pointer. The second pointer (which must be either a constant or a global
1994 value address) contains the meta-data to be associated with the root.</p>
1998 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
1999 location. At compile-time, the code generator generates information to allow
2000 the runtime to find the pointer at GC safe points.
2006 <!-- _______________________________________________________________________ -->
2007 <div class="doc_subsubsection">
2008 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
2011 <div class="doc_text">
2016 call sbyte* (sbyte**)* %llvm.gcread(sbyte** %Ptr)
2021 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
2022 locations, allowing garbage collector implementations that require read
2027 <p>The argument is the address to read from, which should be an address
2028 allocated from the garbage collector.</p>
2032 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
2033 instruction, but may be replaced with substantially more complex code by the
2034 garbage collector runtime, as needed.</p>
2039 <!-- _______________________________________________________________________ -->
2040 <div class="doc_subsubsection">
2041 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
2044 <div class="doc_text">
2049 call void (sbyte*, sbyte**)* %llvm.gcwrite(sbyte* %P1, sbyte** %P2)
2054 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
2055 locations, allowing garbage collector implementations that require write
2056 barriers (such as generational or reference counting collectors).</p>
2060 <p>The first argument is the reference to store, and the second is the heap
2061 location to store to.</p>
2065 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
2066 instruction, but may be replaced with substantially more complex code by the
2067 garbage collector runtime, as needed.</p>
2073 <!-- ======================================================================= -->
2074 <div class="doc_subsection">
2075 <a name="int_codegen">Code Generator Intrinsics</a>
2078 <div class="doc_text">
2080 These intrinsics are provided by LLVM to expose special features that may only
2081 be implemented with code generator support.
2086 <!-- _______________________________________________________________________ -->
2087 <div class="doc_subsubsection">
2088 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
2091 <div class="doc_text">
2095 call void* ()* %llvm.returnaddress(uint <level>)
2101 The '<tt>llvm.returnaddress</tt>' intrinsic returns a target-specific value
2102 indicating the return address of the current function or one of its callers.
2108 The argument to this intrinsic indicates which function to return the address
2109 for. Zero indicates the calling function, one indicates its caller, etc. The
2110 argument is <b>required</b> to be a constant integer value.
2116 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
2117 the return address of the specified call frame, or zero if it cannot be
2118 identified. The value returned by this intrinsic is likely to be incorrect or 0
2119 for arguments other than zero, so it should only be used for debugging purposes.
2123 Note that calling this intrinsic does not prevent function inlining or other
2124 aggressive transformations, so the value returned may not that of the obvious
2125 source-language caller.
2130 <!-- _______________________________________________________________________ -->
2131 <div class="doc_subsubsection">
2132 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
2135 <div class="doc_text">
2139 call void* ()* %llvm.frameaddress(uint <level>)
2145 The '<tt>llvm.frameaddress</tt>' intrinsic returns the target-specific frame
2146 pointer value for the specified stack frame.
2152 The argument to this intrinsic indicates which function to return the frame
2153 pointer for. Zero indicates the calling function, one indicates its caller,
2154 etc. The argument is <b>required</b> to be a constant integer value.
2160 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
2161 the frame address of the specified call frame, or zero if it cannot be
2162 identified. The value returned by this intrinsic is likely to be incorrect or 0
2163 for arguments other than zero, so it should only be used for debugging purposes.
2167 Note that calling this intrinsic does not prevent function inlining or other
2168 aggressive transformations, so the value returned may not that of the obvious
2169 source-language caller.
2173 <!-- ======================================================================= -->
2174 <div class="doc_subsection">
2175 <a name="int_os">Operating System Intrinsics</a>
2178 <div class="doc_text">
2180 These intrinsics are provided by LLVM to support the implementation of
2181 operating system level code.
2186 <!-- _______________________________________________________________________ -->
2187 <div class="doc_subsubsection">
2188 <a name="i_readport">'<tt>llvm.readport</tt>' Intrinsic</a>
2191 <div class="doc_text">
2195 call <integer type> (<integer type>)* %llvm.readport (<integer type> <address>)
2201 The '<tt>llvm.readport</tt>' intrinsic reads data from the specified hardware
2208 The argument to this intrinsic indicates the hardware I/O address from which
2209 to read the data. The address is in the hardware I/O address namespace (as
2210 opposed to being a memory location for memory mapped I/O).
2216 The '<tt>llvm.readport</tt>' intrinsic reads data from the hardware I/O port
2217 specified by <i>address</i> and returns the value. The address and return
2218 value must be integers, but the size is dependent upon the platform upon which
2219 the program is code generated. For example, on x86, the address must be an
2220 unsigned 16 bit value, and the return value must be 8, 16, or 32 bits.
2225 <!-- _______________________________________________________________________ -->
2226 <div class="doc_subsubsection">
2227 <a name="i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a>
2230 <div class="doc_text">
2234 call void (<integer type>, <integer type>)* %llvm.writeport (<integer type> <value>, <integer type> <address>)
2240 The '<tt>llvm.writeport</tt>' intrinsic writes data to the specified hardware
2247 The first argument is the value to write to the I/O port.
2251 The second argument indicates the hardware I/O address to which data should be
2252 written. The address is in the hardware I/O address namespace (as opposed to
2253 being a memory location for memory mapped I/O).
2259 The '<tt>llvm.writeport</tt>' intrinsic writes <i>value</i> to the I/O port
2260 specified by <i>address</i>. The address and value must be integers, but the
2261 size is dependent upon the platform upon which the program is code generated.
2262 For example, on x86, the address must be an unsigned 16 bit value, and the
2263 value written must be 8, 16, or 32 bits in length.
2268 <!-- _______________________________________________________________________ -->
2269 <div class="doc_subsubsection">
2270 <a name="i_readio">'<tt>llvm.readio</tt>' Intrinsic</a>
2273 <div class="doc_text">
2277 call <result> (<ty>*)* %llvm.readio (<ty> * <pointer>)
2283 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2290 The argument to this intrinsic is a pointer indicating the memory address from
2291 which to read the data. The data must be a
2292 <a href="#t_firstclass">first class</a> type.
2298 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2299 location specified by <i>pointer</i> and returns the value. The argument must
2300 be a pointer, and the return value must be a
2301 <a href="#t_firstclass">first class</a> type. However, certain architectures
2302 may not support I/O on all first class types. For example, 32 bit processors
2303 may only support I/O on data types that are 32 bits or less.
2307 This intrinsic enforces an in-order memory model for llvm.readio and
2308 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2309 scheduled processors may execute loads and stores out of order, re-ordering at
2310 run time accesses to memory mapped I/O registers. Using these intrinsics
2311 ensures that accesses to memory mapped I/O registers occur in program order.
2316 <!-- _______________________________________________________________________ -->
2317 <div class="doc_subsubsection">
2318 <a name="i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a>
2321 <div class="doc_text">
2325 call void (<ty1>, <ty2>*)* %llvm.writeio (<ty1> <value>, <ty2> * <pointer>)
2331 The '<tt>llvm.writeio</tt>' intrinsic writes data to the specified memory
2338 The first argument is the value to write to the memory mapped I/O location.
2339 The second argument is a pointer indicating the memory address to which the
2340 data should be written.
2346 The '<tt>llvm.writeio</tt>' intrinsic writes <i>value</i> to the memory mapped
2347 I/O address specified by <i>pointer</i>. The value must be a
2348 <a href="#t_firstclass">first class</a> type. However, certain architectures
2349 may not support I/O on all first class types. For example, 32 bit processors
2350 may only support I/O on data types that are 32 bits or less.
2354 This intrinsic enforces an in-order memory model for llvm.readio and
2355 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2356 scheduled processors may execute loads and stores out of order, re-ordering at
2357 run time accesses to memory mapped I/O registers. Using these intrinsics
2358 ensures that accesses to memory mapped I/O registers occur in program order.
2363 <!-- ======================================================================= -->
2364 <div class="doc_subsection">
2365 <a name="int_libc">Standard C Library Intrinsics</a>
2368 <div class="doc_text">
2370 LLVM provides intrinsics for a few important standard C library functions.
2371 These intrinsics allow source-language front-ends to pass information about the
2372 alignment of the pointer arguments to the code generator, providing opportunity
2373 for more efficient code generation.
2378 <!-- _______________________________________________________________________ -->
2379 <div class="doc_subsubsection">
2380 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
2383 <div class="doc_text">
2387 call void (sbyte*, sbyte*, uint, uint)* %llvm.memcpy(sbyte* <dest>, sbyte* <src>,
2388 uint <len>, uint <align>)
2394 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2395 location to the destination location.
2399 Note that, unlike the standard libc function, the <tt>llvm.memcpy</tt> intrinsic
2400 does not return a value, and takes an extra alignment argument.
2406 The first argument is a pointer to the destination, the second is a pointer to
2407 the source. The third argument is an (arbitrarily sized) integer argument
2408 specifying the number of bytes to copy, and the fourth argument is the alignment
2409 of the source and destination locations.
2413 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2414 the caller guarantees that the size of the copy is a multiple of the alignment
2415 and that both the source and destination pointers are aligned to that boundary.
2421 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2422 location to the destination location, which are not allowed to overlap. It
2423 copies "len" bytes of memory over. If the argument is known to be aligned to
2424 some boundary, this can be specified as the fourth argument, otherwise it should
2430 <!-- _______________________________________________________________________ -->
2431 <div class="doc_subsubsection">
2432 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
2435 <div class="doc_text">
2439 call void (sbyte*, sbyte*, uint, uint)* %llvm.memmove(sbyte* <dest>, sbyte* <src>,
2440 uint <len>, uint <align>)
2446 The '<tt>llvm.memmove</tt>' intrinsic moves a block of memory from the source
2447 location to the destination location. It is similar to the '<tt>llvm.memcpy</tt>'
2448 intrinsic but allows the two memory locations to overlap.
2452 Note that, unlike the standard libc function, the <tt>llvm.memmove</tt> intrinsic
2453 does not return a value, and takes an extra alignment argument.
2459 The first argument is a pointer to the destination, the second is a pointer to
2460 the source. The third argument is an (arbitrarily sized) integer argument
2461 specifying the number of bytes to copy, and the fourth argument is the alignment
2462 of the source and destination locations.
2466 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2467 the caller guarantees that the size of the copy is a multiple of the alignment
2468 and that both the source and destination pointers are aligned to that boundary.
2474 The '<tt>llvm.memmove</tt>' intrinsic copies a block of memory from the source
2475 location to the destination location, which may overlap. It
2476 copies "len" bytes of memory over. If the argument is known to be aligned to
2477 some boundary, this can be specified as the fourth argument, otherwise it should
2483 <!-- _______________________________________________________________________ -->
2484 <div class="doc_subsubsection">
2485 <a name="i_memset">'<tt>llvm.memset</tt>' Intrinsic</a>
2488 <div class="doc_text">
2492 call void (sbyte*, ubyte, uint, uint)* %llvm.memset(sbyte* <dest>, ubyte <val>,
2493 uint <len>, uint <align>)
2499 The '<tt>llvm.memset</tt>' intrinsic fills a block of memory with a particular
2504 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
2505 does not return a value, and takes an extra alignment argument.
2511 The first argument is a pointer to the destination to fill, the second is the
2512 byte value to fill it with, the third argument is an (arbitrarily sized) integer
2513 argument specifying the number of bytes to fill, and the fourth argument is the
2514 known alignment of destination location.
2518 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2519 the caller guarantees that the size of the copy is a multiple of the alignment
2520 and that the destination pointer is aligned to that boundary.
2526 The '<tt>llvm.memset</tt>' intrinsic fills "len" bytes of memory starting at the
2527 destination location. If the argument is known to be aligned to some boundary,
2528 this can be specified as the fourth argument, otherwise it should be set to 0 or
2534 <!-- _______________________________________________________________________ -->
2535 <div class="doc_subsubsection">
2536 <a name="i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a>
2539 <div class="doc_text">
2543 call bool (<float or double>, <float or double>)* %llvm.isunordered(<float or double> Val1,
2544 <float or double> Val2)
2550 The '<tt>llvm.isunordered</tt>' intrinsic returns true if either or both of the
2551 specified floating point values is a NAN.
2557 The arguments are floating point numbers of the same type.
2563 If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
2571 <!-- ======================================================================= -->
2572 <div class="doc_subsection">
2573 <a name="int_debugger">Debugger Intrinsics</a>
2576 <div class="doc_text">
2578 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
2579 are described in the <a
2580 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
2581 Debugging</a> document.
2586 <!-- *********************************************************************** -->
2589 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2590 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
2591 <a href="http://validator.w3.org/check/referer"><img
2592 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
2594 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
2595 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
2596 Last modified: $Date$