1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>LLVM Assembly Language Reference Manual</title>
6 <link rel="stylesheet" href="llvm.css" type="text/css">
9 <div class="doc_title"> LLVM Language Reference Manual </div>
11 <li><a href="#abstract">Abstract</a></li>
12 <li><a href="#introduction">Introduction</a></li>
13 <li><a href="#identifiers">Identifiers</a></li>
14 <li><a href="#typesystem">Type System</a>
16 <li><a href="#t_primitive">Primitive Types</a>
18 <li><a href="#t_classifications">Type Classifications</a></li>
21 <li><a href="#t_derived">Derived Types</a>
23 <li><a href="#t_array">Array Type</a></li>
24 <li><a href="#t_function">Function Type</a></li>
25 <li><a href="#t_pointer">Pointer Type</a></li>
26 <li><a href="#t_struct">Structure Type</a></li>
27 <!-- <li><a href="#t_packed" >Packed Type</a> -->
32 <li><a href="#highlevel">High Level Structure</a>
34 <li><a href="#modulestructure">Module Structure</a></li>
35 <li><a href="#globalvars">Global Variables</a></li>
36 <li><a href="#functionstructure">Function Structure</a></li>
39 <li><a href="#instref">Instruction Reference</a>
41 <li><a href="#terminators">Terminator Instructions</a>
43 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
44 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
45 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
46 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
47 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
50 <li><a href="#binaryops">Binary Operations</a>
52 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
53 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
54 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
55 <li><a href="#i_div">'<tt>div</tt>' Instruction</a></li>
56 <li><a href="#i_rem">'<tt>rem</tt>' Instruction</a></li>
57 <li><a href="#i_setcc">'<tt>set<i>cc</i></tt>' Instructions</a></li>
60 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
62 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
63 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
64 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
65 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
66 <li><a href="#i_shr">'<tt>shr</tt>' Instruction</a></li>
69 <li><a href="#memoryops">Memory Access Operations</a>
71 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
72 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
73 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
74 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
75 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
76 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
79 <li><a href="#otherops">Other Operations</a>
81 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
82 <li><a href="#i_cast">'<tt>cast .. to</tt>' Instruction</a></li>
83 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
84 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
85 <li><a href="#i_vanext">'<tt>vanext</tt>' Instruction</a></li>
86 <li><a href="#i_vaarg">'<tt>vaarg</tt>' Instruction</a></li>
91 <li><a href="#intrinsics">Intrinsic Functions</a>
93 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
95 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
96 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
97 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
100 <li><a href="#int_codegen">Code Generator Intrinsics</a>
102 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
103 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
106 <li><a href="#int_os">Operating System Intrinsics</a>
108 <li><a href="#i_readport">'<tt>llvm.readport</tt>' Intrinsic</a></li>
109 <li><a href="#i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a></li>
111 <li><a href="#int_libc">Standard C Library Intrinsics</a>
113 <li><a href="#i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a></li>
114 <li><a href="#i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a></li>
115 <li><a href="#i_memset">'<tt>llvm.memset</tt>' Intrinsic</a></li>
118 <li><a href="#int_debugger">Debugger intrinsics</a>
122 <div class="doc_text">
123 <p><b>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
124 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></b></p>
127 <!-- *********************************************************************** -->
128 <div class="doc_section"> <a name="abstract">Abstract </a></div>
129 <!-- *********************************************************************** -->
130 <div class="doc_text">
131 <p>This document is a reference manual for the LLVM assembly language.
132 LLVM is an SSA based representation that provides type safety,
133 low-level operations, flexibility, and the capability of representing
134 'all' high-level languages cleanly. It is the common code
135 representation used throughout all phases of the LLVM compilation
138 <!-- *********************************************************************** -->
139 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
140 <!-- *********************************************************************** -->
141 <div class="doc_text">
142 <p>The LLVM code representation is designed to be used in three
143 different forms: as an in-memory compiler IR, as an on-disk bytecode
144 representation (suitable for fast loading by a Just-In-Time compiler),
145 and as a human readable assembly language representation. This allows
146 LLVM to provide a powerful intermediate representation for efficient
147 compiler transformations and analysis, while providing a natural means
148 to debug and visualize the transformations. The three different forms
149 of LLVM are all equivalent. This document describes the human readable
150 representation and notation.</p>
151 <p>The LLVM representation aims to be a light-weight and low-level
152 while being expressive, typed, and extensible at the same time. It
153 aims to be a "universal IR" of sorts, by being at a low enough level
154 that high-level ideas may be cleanly mapped to it (similar to how
155 microprocessors are "universal IR's", allowing many source languages to
156 be mapped to them). By providing type information, LLVM can be used as
157 the target of optimizations: for example, through pointer analysis, it
158 can be proven that a C automatic variable is never accessed outside of
159 the current function... allowing it to be promoted to a simple SSA
160 value instead of a memory location.</p>
162 <!-- _______________________________________________________________________ -->
163 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
164 <div class="doc_text">
165 <p>It is important to note that this document describes 'well formed'
166 LLVM assembly language. There is a difference between what the parser
167 accepts and what is considered 'well formed'. For example, the
168 following instruction is syntactically okay, but not well formed:</p>
169 <pre> %x = <a href="#i_add">add</a> int 1, %x<br></pre>
170 <p>...because the definition of <tt>%x</tt> does not dominate all of
171 its uses. The LLVM infrastructure provides a verification pass that may
172 be used to verify that an LLVM module is well formed. This pass is
173 automatically run by the parser after parsing input assembly, and by
174 the optimizer before it outputs bytecode. The violations pointed out
175 by the verifier pass indicate bugs in transformation passes or input to
177 <!-- Describe the typesetting conventions here. --> </div>
178 <!-- *********************************************************************** -->
179 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
180 <!-- *********************************************************************** -->
181 <div class="doc_text">
182 <p>LLVM uses three different forms of identifiers, for different
185 <li>Numeric constants are represented as you would expect: 12, -3
186 123.421, etc. Floating point constants have an optional hexadecimal
188 <li>Named values are represented as a string of characters with a '%'
189 prefix. For example, %foo, %DivisionByZero,
190 %a.really.long.identifier. The actual regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
191 Identifiers which require other characters in their names can be
192 surrounded with quotes. In this way, anything except a <tt>"</tt>
193 character can be used in a name.</li>
194 <li>Unnamed values are represented as an unsigned numeric value with
195 a '%' prefix. For example, %12, %2, %44.</li>
197 <p>LLVM requires that values start with a '%' sign for two reasons:
198 Compilers don't need to worry about name clashes with reserved words,
199 and the set of reserved words may be expanded in the future without
200 penalty. Additionally, unnamed identifiers allow a compiler to quickly
201 come up with a temporary variable without having to avoid symbol table
203 <p>Reserved words in LLVM are very similar to reserved words in other
204 languages. There are keywords for different opcodes ('<tt><a
205 href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
206 href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
207 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>',
208 etc...), and others. These reserved words cannot conflict with
209 variable names, because none of them start with a '%' character.</p>
210 <p>Here is an example of LLVM code to multiply the integer variable '<tt>%X</tt>'
213 <pre> %result = <a href="#i_mul">mul</a> uint %X, 8<br></pre>
214 <p>After strength reduction:</p>
215 <pre> %result = <a href="#i_shl">shl</a> uint %X, ubyte 3<br></pre>
216 <p>And the hard way:</p>
217 <pre> <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
219 href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
221 href="#i_add">add</a> uint %1, %1<br></pre>
222 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
223 important lexical features of LLVM:</p>
225 <li>Comments are delimited with a '<tt>;</tt>' and go until the end
227 <li>Unnamed temporaries are created when the result of a computation
228 is not assigned to a named value.</li>
229 <li>Unnamed temporaries are numbered sequentially</li>
231 <p>...and it also show a convention that we follow in this document.
232 When demonstrating instructions, we will follow an instruction with a
233 comment that defines the type and name of value produced. Comments are
234 shown in italic text.</p>
235 <p>The one non-intuitive notation for constants is the optional
236 hexidecimal form of floating point constants. For example, the form '<tt>double
237 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
238 4.5e+15</tt>' which is also supported by the parser. The only time
239 hexadecimal floating point constants are useful (and the only time that
240 they are generated by the disassembler) is when an FP constant has to
241 be emitted that is not representable as a decimal floating point number
242 exactly. For example, NaN's, infinities, and other special cases are
243 represented in their IEEE hexadecimal format so that assembly and
244 disassembly do not cause any bits to change in the constants.</p>
246 <!-- *********************************************************************** -->
247 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
248 <!-- *********************************************************************** -->
249 <div class="doc_text">
250 <p>The LLVM type system is one of the most important features of the
251 intermediate representation. Being typed enables a number of
252 optimizations to be performed on the IR directly, without having to do
253 extra analyses on the side before the transformation. A strong type
254 system makes it easier to read the generated code and enables novel
255 analyses and transformations that are not feasible to perform on normal
256 three address code representations.</p>
257 <!-- The written form for the type system was heavily influenced by the
258 syntactic problems with types in the C language<sup><a
259 href="#rw_stroustrup">1</a></sup>.<p> --> </div>
260 <!-- ======================================================================= -->
261 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
262 <div class="doc_text">
263 <p>The primitive types are the fundamental building blocks of the LLVM
264 system. The current set of primitive types are as follows:</p>
266 <table border="0" style="align: center">
270 <table border="1" cellspacing="0" cellpadding="4" style="align: center">
273 <td><tt>void</tt></td>
277 <td><tt>ubyte</tt></td>
278 <td>Unsigned 8 bit value</td>
281 <td><tt>ushort</tt></td>
282 <td>Unsigned 16 bit value</td>
285 <td><tt>uint</tt></td>
286 <td>Unsigned 32 bit value</td>
289 <td><tt>ulong</tt></td>
290 <td>Unsigned 64 bit value</td>
293 <td><tt>float</tt></td>
294 <td>32 bit floating point value</td>
297 <td><tt>label</tt></td>
298 <td>Branch destination</td>
304 <table border="1" cellspacing="0" cellpadding="4">
307 <td><tt>bool</tt></td>
308 <td>True or False value</td>
311 <td><tt>sbyte</tt></td>
312 <td>Signed 8 bit value</td>
315 <td><tt>short</tt></td>
316 <td>Signed 16 bit value</td>
319 <td><tt>int</tt></td>
320 <td>Signed 32 bit value</td>
323 <td><tt>long</tt></td>
324 <td>Signed 64 bit value</td>
327 <td><tt>double</tt></td>
328 <td>64 bit floating point value</td>
338 <!-- _______________________________________________________________________ -->
339 <div class="doc_subsubsection"> <a name="t_classifications">Type
340 Classifications</a> </div>
341 <div class="doc_text">
342 <p>These different primitive types fall into a few useful
345 <table border="1" cellspacing="0" cellpadding="4">
348 <td><a name="t_signed">signed</a></td>
349 <td><tt>sbyte, short, int, long, float, double</tt></td>
352 <td><a name="t_unsigned">unsigned</a></td>
353 <td><tt>ubyte, ushort, uint, ulong</tt></td>
356 <td><a name="t_integer">integer</a></td>
357 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
360 <td><a name="t_integral">integral</a></td>
361 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
364 <td><a name="t_floating">floating point</a></td>
365 <td><tt>float, double</tt></td>
368 <td><a name="t_firstclass">first class</a></td>
369 <td><tt>bool, ubyte, sbyte, ushort, short,<br>
370 uint, int, ulong, long, float, double, <a href="#t_pointer">pointer</a></tt></td>
375 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
376 most important. Values of these types are the only ones which can be
377 produced by instructions, passed as arguments, or used as operands to
378 instructions. This means that all structures and arrays must be
379 manipulated either by pointer or by component.</p>
381 <!-- ======================================================================= -->
382 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
383 <div class="doc_text">
384 <p>The real power in LLVM comes from the derived types in the system.
385 This is what allows a programmer to represent arrays, functions,
386 pointers, and other useful types. Note that these derived types may be
387 recursive: For example, it is possible to have a two dimensional array.</p>
389 <!-- _______________________________________________________________________ -->
390 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
391 <div class="doc_text">
393 <p>The array type is a very simple derived type that arranges elements
394 sequentially in memory. The array type requires a size (number of
395 elements) and an underlying data type.</p>
397 <pre> [<# elements> x <elementtype>]<br></pre>
398 <p>The number of elements is a constant integer value, elementtype may
399 be any type with a size.</p>
401 <p> <tt>[40 x int ]</tt>: Array of 40 integer values.<br>
402 <tt>[41 x int ]</tt>: Array of 41 integer values.<br>
403 <tt>[40 x uint]</tt>: Array of 40 unsigned integer values.</p>
405 <p>Here are some examples of multidimensional arrays:</p>
407 <table border="0" cellpadding="0" cellspacing="0">
410 <td><tt>[3 x [4 x int]]</tt></td>
411 <td>: 3x4 array integer values.</td>
414 <td><tt>[12 x [10 x float]]</tt></td>
415 <td>: 12x10 array of single precision floating point values.</td>
418 <td><tt>[2 x [3 x [4 x uint]]]</tt></td>
419 <td>: 2x3x4 array of unsigned integer values.</td>
425 <!-- _______________________________________________________________________ -->
426 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
427 <div class="doc_text">
429 <p>The function type can be thought of as a function signature. It
430 consists of a return type and a list of formal parameter types.
431 Function types are usually used to build virtual function tables
432 (which are structures of pointers to functions), for indirect function
433 calls, and when defining a function.</p>
435 The return type of a function type cannot be an aggregate type.
438 <pre> <returntype> (<parameter list>)<br></pre>
439 <p>Where '<tt><parameter list></tt>' is a comma-separated list of
440 type specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
441 which indicates that the function takes a variable number of arguments.
442 Variable argument functions can access their arguments with the <a
443 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
446 <table border="0" cellpadding="0" cellspacing="0">
449 <td><tt>int (int)</tt></td>
450 <td>: function taking an <tt>int</tt>, returning an <tt>int</tt></td>
453 <td><tt>float (int, int *) *</tt></td>
454 <td>: <a href="#t_pointer">Pointer</a> to a function that takes
455 an <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
456 returning <tt>float</tt>.</td>
459 <td><tt>int (sbyte *, ...)</tt></td>
460 <td>: A vararg function that takes at least one <a
461 href="#t_pointer">pointer</a> to <tt>sbyte</tt> (signed char in C),
462 which returns an integer. This is the signature for <tt>printf</tt>
469 <!-- _______________________________________________________________________ -->
470 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
471 <div class="doc_text">
473 <p>The structure type is used to represent a collection of data members
474 together in memory. The packing of the field types is defined to match
475 the ABI of the underlying processor. The elements of a structure may
476 be any type that has a size.</p>
477 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
478 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
479 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
482 <pre> { <type list> }<br></pre>
485 <table border="0" cellpadding="0" cellspacing="0">
488 <td><tt>{ int, int, int }</tt></td>
489 <td>: a triple of three <tt>int</tt> values</td>
492 <td><tt>{ float, int (int) * }</tt></td>
493 <td>: A pair, where the first element is a <tt>float</tt> and the
494 second element is a <a href="#t_pointer">pointer</a> to a <a
495 href="t_function">function</a> that takes an <tt>int</tt>, returning
496 an <tt>int</tt>.</td>
502 <!-- _______________________________________________________________________ -->
503 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
504 <div class="doc_text">
506 <p>As in many languages, the pointer type represents a pointer or
507 reference to another object, which must live in memory.</p>
509 <pre> <type> *<br></pre>
512 <table border="0" cellpadding="0" cellspacing="0">
515 <td><tt>[4x int]*</tt></td>
516 <td>: <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a>
517 of four <tt>int</tt> values</td>
520 <td><tt>int (int *) *</tt></td>
521 <td>: A <a href="#t_pointer">pointer</a> to a <a
522 href="t_function">function</a> that takes an <tt>int</tt>, returning
523 an <tt>int</tt>.</td>
529 <!-- _______________________________________________________________________ --><!--
530 <div class="doc_subsubsection">
531 <a name="t_packed">Packed Type</a>
534 <div class="doc_text">
536 Mention/decide that packed types work with saturation or not. Maybe have a packed+saturated type in addition to just a packed type.<p>
538 Packed types should be 'nonsaturated' because standard data types are not saturated. Maybe have a saturated packed type?<p>
542 --><!-- *********************************************************************** -->
543 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
544 <!-- *********************************************************************** --><!-- ======================================================================= -->
545 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a> </div>
546 <div class="doc_text">
547 <p>LLVM programs are composed of "Module"s, each of which is a
548 translation unit of the input programs. Each module consists of
549 functions, global variables, and symbol table entries. Modules may be
550 combined together with the LLVM linker, which merges function (and
551 global variable) definitions, resolves forward declarations, and merges
552 symbol table entries. Here is an example of the "hello world" module:</p>
553 <pre><i>; Declare the string constant as a global constant...</i>
554 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
555 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
557 <i>; External declaration of the puts function</i>
558 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
560 <i>; Definition of main function</i>
561 int %main() { <i>; int()* </i>
562 <i>; Convert [13x sbyte]* to sbyte *...</i>
564 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
566 <i>; Call puts function to write out the string to stdout...</i>
568 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
570 href="#i_ret">ret</a> int 0<br>}<br></pre>
571 <p>This example is made up of a <a href="#globalvars">global variable</a>
572 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
573 function, and a <a href="#functionstructure">function definition</a>
574 for "<tt>main</tt>".</p>
575 <a name="linkage"> In general, a module is made up of a list of global
576 values, where both functions and global variables are global values.
577 Global values are represented by a pointer to a memory location (in
578 this case, a pointer to an array of char, and a pointer to a function),
579 and have one of the following linkage types:</a>
582 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
583 <dd>Global values with internal linkage are only directly accessible
584 by objects in the current module. In particular, linking code into a
585 module with an internal global value may cause the internal to be
586 renamed as necessary to avoid collisions. Because the symbol is
587 internal to the module, all references can be updated. This
588 corresponds to the notion of the '<tt>static</tt>' keyword in C, or the
589 idea of "anonymous namespaces" in C++.
592 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
593 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt>
594 linkage, with the twist that linking together two modules defining the
595 same <tt>linkonce</tt> globals will cause one of the globals to be
596 discarded. This is typically used to implement inline functions.
597 Unreferenced <tt>linkonce</tt> globals are allowed to be discarded.
600 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
601 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt>
602 linkage, except that unreferenced <tt>weak</tt> globals may not be
603 discarded. This is used to implement constructs in C such as "<tt>int
604 X;</tt>" at global scope.
607 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
608 <dd>"<tt>appending</tt>" linkage may only be applied to global
609 variables of pointer to array type. When two global variables with
610 appending linkage are linked together, the two global arrays are
611 appended together. This is the LLVM, typesafe, equivalent of having
612 the system linker append together "sections" with identical names when
616 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
617 <dd>If none of the above identifiers are used, the global is
618 externally visible, meaning that it participates in linkage and can be
619 used to resolve external symbol references.
624 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
625 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
626 variable and was linked with this one, one of the two would be renamed,
627 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
628 external (i.e., lacking any linkage declarations), they are accessible
629 outside of the current module. It is illegal for a function <i>declaration</i>
630 to have any linkage type other than "externally visible".</a></p>
633 <!-- ======================================================================= -->
634 <div class="doc_subsection">
635 <a name="globalvars">Global Variables</a>
638 <div class="doc_text">
640 <p>Global variables define regions of memory allocated at compilation
641 time instead of run-time. Global variables may optionally be
642 initialized. A variable may be defined as a global "constant", which
643 indicates that the contents of the variable will never be modified
644 (opening options for optimization).</p>
646 <p>As SSA values, global variables define pointer values that are in
647 scope (i.e. they dominate) for all basic blocks in the program. Global
648 variables always define a pointer to their "content" type because they
649 describe a region of memory, and all memory objects in LLVM are
650 accessed through pointers.</p>
655 <!-- ======================================================================= -->
656 <div class="doc_subsection">
657 <a name="functionstructure">Functions</a>
660 <div class="doc_text">
662 <p>LLVM function definitions are composed of a (possibly empty) argument list,
663 an opening curly brace, a list of basic blocks, and a closing curly brace. LLVM
664 function declarations are defined with the "<tt>declare</tt>" keyword, a
665 function name, and a function signature.</p>
667 <p>A function definition contains a list of basic blocks, forming the CFG for
668 the function. Each basic block may optionally start with a label (giving the
669 basic block a symbol table entry), contains a list of instructions, and ends
670 with a <a href="#terminators">terminator</a> instruction (such as a branch or
671 function return).</p>
673 <p>The first basic block in program is special in two ways: it is immediately
674 executed on entrance to the function, and it is not allowed to have predecessor
675 basic blocks (i.e. there can not be any branches to the entry block of a
676 function). Because the block can have no predecessors, it also cannot have any
677 <a href="#i_phi">PHI nodes</a>.</p>
679 <p>LLVM functions are identified by their name and type signature. Hence, two
680 functions with the same name but different parameter lists or return values are
681 considered different functions, and LLVM will resolves references to each
687 <!-- *********************************************************************** -->
688 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
689 <!-- *********************************************************************** -->
690 <div class="doc_text">
691 <p>The LLVM instruction set consists of several different
692 classifications of instructions: <a href="#terminators">terminator
693 instructions</a>, <a href="#binaryops">binary instructions</a>, <a
694 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
695 instructions</a>.</p>
697 <!-- ======================================================================= -->
698 <div class="doc_subsection"> <a name="terminators">Terminator
699 Instructions</a> </div>
700 <div class="doc_text">
701 <p>As mentioned <a href="#functionstructure">previously</a>, every
702 basic block in a program ends with a "Terminator" instruction, which
703 indicates which block should be executed after the current block is
704 finished. These terminator instructions typically yield a '<tt>void</tt>'
705 value: they produce control flow, not values (the one exception being
706 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
707 <p>There are five different terminator instructions: the '<a
708 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
709 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
710 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, and the '<a
711 href="#i_unwind"><tt>unwind</tt></a>' instruction.</p>
713 <!-- _______________________________________________________________________ -->
714 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
715 Instruction</a> </div>
716 <div class="doc_text">
718 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
719 ret void <i>; Return from void function</i>
722 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
723 value) from a function, back to the caller.</p>
724 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
725 returns a value and then causes control flow, and one that just causes
726 control flow to occur.</p>
728 <p>The '<tt>ret</tt>' instruction may return any '<a
729 href="#t_firstclass">first class</a>' type. Notice that a function is
730 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
731 instruction inside of the function that returns a value that does not
732 match the return type of the function.</p>
734 <p>When the '<tt>ret</tt>' instruction is executed, control flow
735 returns back to the calling function's context. If the caller is a "<a
736 href="#i_call"><tt>call</tt></a> instruction, execution continues at
737 the instruction after the call. If the caller was an "<a
738 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
739 at the beginning "normal" of the destination block. If the instruction
740 returns a value, that value shall set the call or invoke instruction's
743 <pre> ret int 5 <i>; Return an integer value of 5</i>
744 ret void <i>; Return from a void function</i>
747 <!-- _______________________________________________________________________ -->
748 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
749 <div class="doc_text">
751 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
754 <p>The '<tt>br</tt>' instruction is used to cause control flow to
755 transfer to a different basic block in the current function. There are
756 two forms of this instruction, corresponding to a conditional branch
757 and an unconditional branch.</p>
759 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
760 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
761 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
762 value as a target.</p>
764 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
765 argument is evaluated. If the value is <tt>true</tt>, control flows
766 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
767 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
769 <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
770 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
772 <!-- _______________________________________________________________________ -->
773 <div class="doc_subsubsection">
774 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
777 <div class="doc_text">
781 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
786 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
787 several different places. It is a generalization of the '<tt>br</tt>'
788 instruction, allowing a branch to occur to one of many possible
794 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
795 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
796 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
797 table is not allowed to contain duplicate constant entries.</p>
801 <p>The <tt>switch</tt> instruction specifies a table of values and
802 destinations. When the '<tt>switch</tt>' instruction is executed, this
803 table is searched for the given value. If the value is found, the
804 corresponding destination is branched to, otherwise the default value
805 it transfered to.</p>
807 <h5>Implementation:</h5>
809 <p>Depending on properties of the target machine and the particular
810 <tt>switch</tt> instruction, this instruction may be code generated in different
811 ways, for example as a series of chained conditional branches, or with a lookup
817 <i>; Emulate a conditional br instruction</i>
818 %Val = <a href="#i_cast">cast</a> bool %value to int
819 switch int %Val, label %truedest [int 0, label %falsedest ]
821 <i>; Emulate an unconditional br instruction</i>
822 switch uint 0, label %dest [ ]
824 <i>; Implement a jump table:</i>
825 switch uint %val, label %otherwise [ uint 0, label %onzero
827 uint 2, label %ontwo ]
830 <!-- _______________________________________________________________________ -->
831 <div class="doc_subsubsection"> <a name="i_invoke">'<tt>invoke</tt>'
832 Instruction</a> </div>
833 <div class="doc_text">
835 <pre> <result> = invoke <ptr to function ty> %<function ptr val>(<function args>)<br> to label <normal label> except label <exception label><br></pre>
837 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a
838 specified function, with the possibility of control flow transfer to
839 either the '<tt>normal</tt>' <tt>label</tt> label or the '<tt>exception</tt>'<tt>label</tt>.
840 If the callee function returns with the "<tt><a href="#i_ret">ret</a></tt>"
841 instruction, control flow will return to the "normal" label. If the
842 callee (or any indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
843 instruction, control is interrupted, and continued at the dynamically
844 nearest "except" label.</p>
846 <p>This instruction requires several arguments:</p>
848 <li>'<tt>ptr to function ty</tt>': shall be the signature of the
849 pointer to function value being invoked. In most cases, this is a
850 direct function invocation, but indirect <tt>invoke</tt>s are just as
851 possible, branching off an arbitrary pointer to function value. </li>
852 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer
853 to a function to be invoked. </li>
854 <li>'<tt>function args</tt>': argument list whose types match the
855 function signature argument types. If the function signature indicates
856 the function accepts a variable number of arguments, the extra
857 arguments can be specified. </li>
858 <li>'<tt>normal label</tt>': the label reached when the called
859 function executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
860 <li>'<tt>exception label</tt>': the label reached when a callee
861 returns with the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
864 <p>This instruction is designed to operate as a standard '<tt><a
865 href="#i_call">call</a></tt>' instruction in most regards. The
866 primary difference is that it establishes an association with a label,
867 which is used by the runtime library to unwind the stack.</p>
868 <p>This instruction is used in languages with destructors to ensure
869 that proper cleanup is performed in the case of either a <tt>longjmp</tt>
870 or a thrown exception. Additionally, this is important for
871 implementation of '<tt>catch</tt>' clauses in high-level languages that
874 <pre> %retval = invoke int %Test(int 15)<br> to label %Continue<br> except label %TestCleanup <i>; {int}:retval set</i>
877 <!-- _______________________________________________________________________ -->
878 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
879 Instruction</a> </div>
880 <div class="doc_text">
882 <pre> unwind<br></pre>
884 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing
885 control flow at the first callee in the dynamic call stack which used
886 an <a href="#i_invoke"><tt>invoke</tt></a> instruction to perform the
887 call. This is primarily used to implement exception handling.</p>
889 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current
890 function to immediately halt. The dynamic call stack is then searched
891 for the first <a href="#i_invoke"><tt>invoke</tt></a> instruction on
892 the call stack. Once found, execution continues at the "exceptional"
893 destination block specified by the <tt>invoke</tt> instruction. If
894 there is no <tt>invoke</tt> instruction in the dynamic call chain,
895 undefined behavior results.</p>
897 <!-- ======================================================================= -->
898 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
899 <div class="doc_text">
900 <p>Binary operators are used to do most of the computation in a
901 program. They require two operands, execute an operation on them, and
902 produce a single value. The result value of a binary operator is not
903 necessarily the same type as its operands.</p>
904 <p>There are several different binary operators:</p>
906 <!-- _______________________________________________________________________ -->
907 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
908 Instruction</a> </div>
909 <div class="doc_text">
911 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
914 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
916 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
917 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
918 values. Both arguments must have identical types.</p>
920 <p>The value produced is the integer or floating point sum of the two
923 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
926 <!-- _______________________________________________________________________ -->
927 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
928 Instruction</a> </div>
929 <div class="doc_text">
931 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
934 <p>The '<tt>sub</tt>' instruction returns the difference of its two
936 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
937 instruction present in most other intermediate representations.</p>
939 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
940 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
941 values. Both arguments must have identical types.</p>
943 <p>The value produced is the integer or floating point difference of
944 the two operands.</p>
946 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
947 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
950 <!-- _______________________________________________________________________ -->
951 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
952 Instruction</a> </div>
953 <div class="doc_text">
955 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
958 <p>The '<tt>mul</tt>' instruction returns the product of its two
961 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
962 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
963 values. Both arguments must have identical types.</p>
965 <p>The value produced is the integer or floating point product of the
967 <p>There is no signed vs unsigned multiplication. The appropriate
968 action is taken based on the type of the operand.</p>
970 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
973 <!-- _______________________________________________________________________ -->
974 <div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
975 Instruction</a> </div>
976 <div class="doc_text">
978 <pre> <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
981 <p>The '<tt>div</tt>' instruction returns the quotient of its two
984 <p>The two arguments to the '<tt>div</tt>' instruction must be either <a
985 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
986 values. Both arguments must have identical types.</p>
988 <p>The value produced is the integer or floating point quotient of the
991 <pre> <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
994 <!-- _______________________________________________________________________ -->
995 <div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
996 Instruction</a> </div>
997 <div class="doc_text">
999 <pre> <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1002 <p>The '<tt>rem</tt>' instruction returns the remainder from the
1003 division of its two operands.</p>
1005 <p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1006 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1007 values. Both arguments must have identical types.</p>
1009 <p>This returns the <i>remainder</i> of a division (where the result
1010 has the same sign as the divisor), not the <i>modulus</i> (where the
1011 result has the same sign as the dividend) of a value. For more
1012 information about the difference, see: <a
1013 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1016 <pre> <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1019 <!-- _______________________________________________________________________ -->
1020 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1021 Instructions</a> </div>
1022 <div class="doc_text">
1024 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1025 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1026 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1027 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1028 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1029 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1032 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1033 value based on a comparison of their two operands.</p>
1035 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1036 be of <a href="#t_firstclass">first class</a> type (it is not possible
1037 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1038 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1041 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1042 value if both operands are equal.<br>
1043 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1044 value if both operands are unequal.<br>
1045 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1046 value if the first operand is less than the second operand.<br>
1047 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1048 value if the first operand is greater than the second operand.<br>
1049 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1050 value if the first operand is less than or equal to the second operand.<br>
1051 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1052 value if the first operand is greater than or equal to the second
1055 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1056 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1057 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1058 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1059 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1060 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1063 <!-- ======================================================================= -->
1064 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1065 Operations</a> </div>
1066 <div class="doc_text">
1067 <p>Bitwise binary operators are used to do various forms of
1068 bit-twiddling in a program. They are generally very efficient
1069 instructions, and can commonly be strength reduced from other
1070 instructions. They require two operands, execute an operation on them,
1071 and produce a single value. The resulting value of the bitwise binary
1072 operators is always the same type as its first operand.</p>
1074 <!-- _______________________________________________________________________ -->
1075 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1076 Instruction</a> </div>
1077 <div class="doc_text">
1079 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1082 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1083 its two operands.</p>
1085 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1086 href="#t_integral">integral</a> values. Both arguments must have
1087 identical types.</p>
1089 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1091 <div style="align: center">
1092 <table border="1" cellspacing="0" cellpadding="4">
1123 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1124 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1125 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1128 <!-- _______________________________________________________________________ -->
1129 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1130 <div class="doc_text">
1132 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1135 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1136 or of its two operands.</p>
1138 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1139 href="#t_integral">integral</a> values. Both arguments must have
1140 identical types.</p>
1142 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1144 <div style="align: center">
1145 <table border="1" cellspacing="0" cellpadding="4">
1176 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1177 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1178 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1181 <!-- _______________________________________________________________________ -->
1182 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1183 Instruction</a> </div>
1184 <div class="doc_text">
1186 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1189 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1190 or of its two operands. The <tt>xor</tt> is used to implement the
1191 "one's complement" operation, which is the "~" operator in C.</p>
1193 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1194 href="#t_integral">integral</a> values. Both arguments must have
1195 identical types.</p>
1197 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1199 <div style="align: center">
1200 <table border="1" cellspacing="0" cellpadding="4">
1232 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1233 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1234 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1235 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1238 <!-- _______________________________________________________________________ -->
1239 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1240 Instruction</a> </div>
1241 <div class="doc_text">
1243 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1246 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1247 the left a specified number of bits.</p>
1249 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
1250 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1253 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
1255 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1256 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1257 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1260 <!-- _______________________________________________________________________ -->
1261 <div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1262 Instruction</a> </div>
1263 <div class="doc_text">
1265 <pre> <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1268 <p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1269 the right a specified number of bits.</p>
1271 <p>The first argument to the '<tt>shr</tt>' instruction must be an <a
1272 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1275 <p>If the first argument is a <a href="#t_signed">signed</a> type, the
1276 most significant bit is duplicated in the newly free'd bit positions.
1277 If the first argument is unsigned, zero bits shall fill the empty
1280 <pre> <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1281 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1282 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1283 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1284 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1287 <!-- ======================================================================= -->
1288 <div class="doc_subsection"> <a name="memoryops">Memory Access
1289 Operations</a></div>
1290 <div class="doc_text">
1291 <p>A key design point of an SSA-based representation is how it
1292 represents memory. In LLVM, no memory locations are in SSA form, which
1293 makes things very simple. This section describes how to read, write,
1294 allocate and free memory in LLVM.</p>
1296 <!-- _______________________________________________________________________ -->
1297 <div class="doc_subsubsection"> <a name="i_malloc">'<tt>malloc</tt>'
1298 Instruction</a> </div>
1299 <div class="doc_text">
1301 <pre> <result> = malloc <type>, uint <NumElements> <i>; yields {type*}:result</i>
1302 <result> = malloc <type> <i>; yields {type*}:result</i>
1305 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
1306 heap and returns a pointer to it.</p>
1308 <p>The '<tt>malloc</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1309 bytes of memory from the operating system and returns a pointer of the
1310 appropriate type to the program. The second form of the instruction is
1311 a shorter version of the first instruction that defaults to allocating
1313 <p>'<tt>type</tt>' must be a sized type.</p>
1315 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
1316 a pointer is returned.</p>
1318 <pre> %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1321 href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1322 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1323 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1326 <!-- _______________________________________________________________________ -->
1327 <div class="doc_subsubsection"> <a name="i_free">'<tt>free</tt>'
1328 Instruction</a> </div>
1329 <div class="doc_text">
1331 <pre> free <type> <value> <i>; yields {void}</i>
1334 <p>The '<tt>free</tt>' instruction returns memory back to the unused
1335 memory heap, to be reallocated in the future.</p>
1338 <p>'<tt>value</tt>' shall be a pointer value that points to a value
1339 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
1342 <p>Access to the memory pointed to by the pointer is not longer defined
1343 after this instruction executes.</p>
1345 <pre> %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1346 free [4 x ubyte]* %array
1349 <!-- _______________________________________________________________________ -->
1350 <div class="doc_subsubsection"> <a name="i_alloca">'<tt>alloca</tt>'
1351 Instruction</a> </div>
1352 <div class="doc_text">
1354 <pre> <result> = alloca <type>, uint <NumElements> <i>; yields {type*}:result</i>
1355 <result> = alloca <type> <i>; yields {type*}:result</i>
1358 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
1359 stack frame of the procedure that is live until the current function
1360 returns to its caller.</p>
1362 <p>The the '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1363 bytes of memory on the runtime stack, returning a pointer of the
1364 appropriate type to the program. The second form of the instruction is
1365 a shorter version of the first that defaults to allocating one element.</p>
1366 <p>'<tt>type</tt>' may be any sized type.</p>
1368 <p>Memory is allocated, a pointer is returned. '<tt>alloca</tt>'d
1369 memory is automatically released when the function returns. The '<tt>alloca</tt>'
1370 instruction is commonly used to represent automatic variables that must
1371 have an address available. When the function returns (either with the <tt><a
1372 href="#i_ret">ret</a></tt> or <tt><a href="#i_invoke">invoke</a></tt>
1373 instructions), the memory is reclaimed.</p>
1375 <pre> %ptr = alloca int <i>; yields {int*}:ptr</i>
1376 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
1379 <!-- _______________________________________________________________________ -->
1380 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
1381 Instruction</a> </div>
1382 <div class="doc_text">
1384 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
1386 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
1388 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
1389 address to load from. The pointer must point to a <a
1390 href="t_firstclass">first class</a> type. If the <tt>load</tt> is
1391 marked as <tt>volatile</tt> then the optimizer is not allowed to modify
1392 the number or order of execution of this <tt>load</tt> with other
1393 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
1396 <p>The location of memory pointed to is loaded.</p>
1398 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1400 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1401 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1404 <!-- _______________________________________________________________________ -->
1405 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
1406 Instruction</a> </div>
1408 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1409 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1412 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
1414 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
1415 to store and an address to store it into. The type of the '<tt><pointer></tt>'
1416 operand must be a pointer to the type of the '<tt><value></tt>'
1417 operand. If the <tt>store</tt> is marked as <tt>volatile</tt> then the
1418 optimizer is not allowed to modify the number or order of execution of
1419 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
1420 href="#i_store">store</a></tt> instructions.</p>
1422 <p>The contents of memory are updated to contain '<tt><value></tt>'
1423 at the location specified by the '<tt><pointer></tt>' operand.</p>
1425 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1427 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1428 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1430 <!-- _______________________________________________________________________ -->
1431 <div class="doc_subsubsection">
1432 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
1435 <div class="doc_text">
1438 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
1444 The '<tt>getelementptr</tt>' instruction is used to get the address of a
1445 subelement of an aggregate data structure.</p>
1449 <p>This instruction takes a list of integer constants that indicate what
1450 elements of the aggregate object to index to. The actual types of the arguments
1451 provided depend on the type of the first pointer argument. The
1452 '<tt>getelementptr</tt>' instruction is used to index down through the type
1453 levels of a structure. When indexing into a structure, only <tt>uint</tt>
1454 integer constants are allowed. When indexing into an array or pointer
1455 <tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
1457 <p>For example, let's consider a C code fragment and how it gets
1458 compiled to LLVM:</p>
1472 int *foo(struct ST *s) {
1473 return &s[1].Z.B[5][13];
1477 <p>The LLVM code generated by the GCC frontend is:</p>
1480 %RT = type { sbyte, [10 x [20 x int]], sbyte }
1481 %ST = type { int, double, %RT }
1483 int* "foo"(%ST* %s) {
1484 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13<br>
1491 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
1492 on the pointer type that is being index into. <a href="t_pointer">Pointer</a>
1493 and <a href="t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
1494 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="t_struct">structure</a>
1495 types require <tt>uint</tt> <b>constants</b>.</p>
1497 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
1498 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
1499 }</tt>' type, a structure. The second index indexes into the third element of
1500 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
1501 sbyte }</tt>' type, another structure. The third index indexes into the second
1502 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
1503 array. The two dimensions of the array are subscripted into, yielding an
1504 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction return a pointer
1505 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
1507 <p>Note that it is perfectly legal to index partially through a
1508 structure, returning a pointer to an inner element. Because of this,
1509 the LLVM code for the given testcase is equivalent to:</p>
1512 int* "foo"(%ST* %s) {
1513 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
1514 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
1515 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
1516 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
1517 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
1523 <i>; yields [12 x ubyte]*:aptr</i>
1524 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
1528 <!-- ======================================================================= -->
1529 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
1530 <div class="doc_text">
1531 <p>The instructions in this category are the "miscellaneous"
1532 instructions, which defy better classification.</p>
1534 <!-- _______________________________________________________________________ -->
1535 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
1536 Instruction</a> </div>
1537 <div class="doc_text">
1539 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
1541 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
1542 the SSA graph representing the function.</p>
1544 <p>The type of the incoming values are specified with the first type
1545 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
1546 as arguments, with one pair for each predecessor basic block of the
1547 current block. Only values of <a href="#t_firstclass">first class</a>
1548 type may be used as the value arguments to the PHI node. Only labels
1549 may be used as the label arguments.</p>
1550 <p>There must be no non-phi instructions between the start of a basic
1551 block and the PHI instructions: i.e. PHI instructions must be first in
1554 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
1555 value specified by the parameter, depending on which basic block we
1556 came from in the last <a href="#terminators">terminator</a> instruction.</p>
1558 <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>
1561 <!-- _______________________________________________________________________ -->
1562 <div class="doc_subsubsection">
1563 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
1566 <div class="doc_text">
1571 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
1577 The '<tt>cast</tt>' instruction is used as the primitive means to convert
1578 integers to floating point, change data type sizes, and break type safety (by
1586 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
1587 class value, and a type to cast it to, which must also be a <a
1588 href="#t_firstclass">first class</a> type.
1594 This instruction follows the C rules for explicit casts when determining how the
1595 data being cast must change to fit in its new container.
1599 When casting to bool, any value that would be considered true in the context of
1600 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
1601 all else are '<tt>false</tt>'.
1605 When extending an integral value from a type of one signness to another (for
1606 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
1607 <b>source</b> value is signed, and zero-extended if the source value is
1608 unsigned. <tt>bool</tt> values are always zero extended into either zero or
1615 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
1616 %Y = cast int 123 to bool <i>; yields bool:true</i>
1620 <!-- _______________________________________________________________________ -->
1621 <div class="doc_subsubsection">
1622 <a name="i_select">'<tt>select</tt>' Instruction</a>
1625 <div class="doc_text">
1630 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
1636 The '<tt>select</tt>' instruction is used to choose one value based on a
1637 condition, without branching.
1644 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.
1650 If the boolean condition evaluates to true, the instruction returns the first
1651 value argument, otherwise it returns the second value argument.
1657 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
1665 <!-- _______________________________________________________________________ -->
1666 <div class="doc_subsubsection"> <a name="i_call">'<tt>call</tt>'
1667 Instruction</a> </div>
1668 <div class="doc_text">
1670 <pre> <result> = call <ty>* <fnptrval>(<param list>)<br></pre>
1672 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
1674 <p>This instruction requires several arguments:</p>
1677 <p>'<tt>ty</tt>': shall be the signature of the pointer to function
1678 value being invoked. The argument types must match the types implied
1679 by this signature.</p>
1682 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a
1683 function to be invoked. In most cases, this is a direct function
1684 invocation, but indirect <tt>call</tt>s are just as possible,
1685 calling an arbitrary pointer to function values.</p>
1688 <p>'<tt>function args</tt>': argument list whose types match the
1689 function signature argument types. If the function signature
1690 indicates the function accepts a variable number of arguments, the
1691 extra arguments can be specified.</p>
1695 <p>The '<tt>call</tt>' instruction is used to cause control flow to
1696 transfer to a specified function, with its incoming arguments bound to
1697 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
1698 instruction in the called function, control flow continues with the
1699 instruction after the function call, and the return value of the
1700 function is bound to the result argument. This is a simpler case of
1701 the <a href="#i_invoke">invoke</a> instruction.</p>
1703 <pre> %retval = call int %test(int %argc)<br> call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);<br></pre>
1705 <!-- _______________________________________________________________________ -->
1706 <div class="doc_subsubsection"> <a name="i_vanext">'<tt>vanext</tt>'
1707 Instruction</a> </div>
1708 <div class="doc_text">
1710 <pre> <resultarglist> = vanext <va_list> <arglist>, <argty><br></pre>
1712 <p>The '<tt>vanext</tt>' instruction is used to access arguments passed
1713 through the "variable argument" area of a function call. It is used to
1714 implement the <tt>va_arg</tt> macro in C.</p>
1716 <p>This instruction takes a <tt>valist</tt> value and the type of the
1717 argument. It returns another <tt>valist</tt>.</p>
1719 <p>The '<tt>vanext</tt>' instruction advances the specified <tt>valist</tt>
1720 past an argument of the specified type. In conjunction with the <a
1721 href="#i_vaarg"><tt>vaarg</tt></a> instruction, it is used to implement
1722 the <tt>va_arg</tt> macro available in C. For more information, see
1723 the variable argument handling <a href="#int_varargs">Intrinsic
1725 <p>It is legal for this instruction to be called in a function which
1726 does not take a variable number of arguments, for example, the <tt>vfprintf</tt>
1728 <p><tt>vanext</tt> is an LLVM instruction instead of an <a
1729 href="#intrinsics">intrinsic function</a> because it takes an type as
1732 <p>See the <a href="#int_varargs">variable argument processing</a>
1735 <!-- _______________________________________________________________________ -->
1736 <div class="doc_subsubsection"> <a name="i_vaarg">'<tt>vaarg</tt>'
1737 Instruction</a> </div>
1738 <div class="doc_text">
1740 <pre> <resultval> = vaarg <va_list> <arglist>, <argty><br></pre>
1742 <p>The '<tt>vaarg</tt>' instruction is used to access arguments passed
1743 through the "variable argument" area of a function call. It is used to
1744 implement the <tt>va_arg</tt> macro in C.</p>
1746 <p>This instruction takes a <tt>valist</tt> value and the type of the
1747 argument. It returns a value of the specified argument type.</p>
1749 <p>The '<tt>vaarg</tt>' instruction loads an argument of the specified
1750 type from the specified <tt>va_list</tt>. In conjunction with the <a
1751 href="#i_vanext"><tt>vanext</tt></a> instruction, it is used to
1752 implement the <tt>va_arg</tt> macro available in C. For more
1753 information, see the variable argument handling <a href="#int_varargs">Intrinsic
1755 <p>It is legal for this instruction to be called in a function which
1756 does not take a variable number of arguments, for example, the <tt>vfprintf</tt>
1758 <p><tt>vaarg</tt> is an LLVM instruction instead of an <a
1759 href="#intrinsics">intrinsic function</a> because it takes an type as
1762 <p>See the <a href="#int_varargs">variable argument processing</a>
1766 <!-- *********************************************************************** -->
1767 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
1768 <!-- *********************************************************************** -->
1770 <div class="doc_text">
1772 <p>LLVM supports the notion of an "intrinsic function". These functions have
1773 well known names and semantics, and are required to follow certain
1774 restrictions. Overall, these instructions represent an extension mechanism for
1775 the LLVM language that does not require changing all of the transformations in
1776 LLVM to add to the language (or the bytecode reader/writer, the parser,
1779 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix, this
1780 prefix is reserved in LLVM for intrinsic names, thus functions may not be named
1781 this. Intrinsic functions must always be external functions: you cannot define
1782 the body of intrinsic functions. Intrinsic functions may only be used in call
1783 or invoke instructions: it is illegal to take the address of an intrinsic
1784 function. Additionally, because intrinsic functions are part of the LLVM
1785 language, it is required that they all be documented here if any are added.</p>
1789 Adding an intrinsic to LLVM is straight-forward if it is possible to express the
1790 concept in LLVM directly (ie, code generator support is not _required_). To do
1791 this, extend the default implementation of the IntrinsicLowering class to handle
1792 the intrinsic. Code generators use this class to lower intrinsics they do not
1793 understand to raw LLVM instructions that they do.
1798 <!-- ======================================================================= -->
1799 <div class="doc_subsection">
1800 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
1803 <div class="doc_text">
1804 <p>Variable argument support is defined in LLVM with the <a
1805 href="#i_vanext"><tt>vanext</tt></a> instruction and these three
1806 intrinsic functions. These functions are related to the similarly
1807 named macros defined in the <tt><stdarg.h></tt> header file.</p>
1808 <p>All of these functions operate on arguments that use a
1809 target-specific value type "<tt>va_list</tt>". The LLVM assembly
1810 language reference manual does not define what this type is, so all
1811 transformations should be prepared to handle intrinsics with any type
1813 <p>This example shows how the <a href="#i_vanext"><tt>vanext</tt></a>
1814 instruction and the variable argument handling intrinsic functions are
1817 int %test(int %X, ...) {
1818 ; Initialize variable argument processing
1819 %ap = call sbyte* %<a href="#i_va_start">llvm.va_start</a>()
1821 ; Read a single integer argument
1822 %tmp = vaarg sbyte* %ap, int
1824 ; Advance to the next argument
1825 %ap2 = vanext sbyte* %ap, int
1827 ; Demonstrate usage of llvm.va_copy and llvm.va_end
1828 %aq = call sbyte* %<a href="#i_va_copy">llvm.va_copy</a>(sbyte* %ap2)
1829 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %aq)
1831 ; Stop processing of arguments.
1832 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %ap2)
1838 <!-- _______________________________________________________________________ -->
1839 <div class="doc_subsubsection">
1840 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
1844 <div class="doc_text">
1846 <pre> call va_list ()* %llvm.va_start()<br></pre>
1848 <p>The '<tt>llvm.va_start</tt>' intrinsic returns a new <tt><arglist></tt>
1849 for subsequent use by the variable argument intrinsics.</p>
1851 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
1852 macro available in C. In a target-dependent way, it initializes and
1853 returns a <tt>va_list</tt> element, so that the next <tt>vaarg</tt>
1854 will produce the first variable argument passed to the function. Unlike
1855 the C <tt>va_start</tt> macro, this intrinsic does not need to know the
1856 last argument of the function, the compiler can figure that out.</p>
1857 <p>Note that this intrinsic function is only legal to be called from
1858 within the body of a variable argument function.</p>
1861 <!-- _______________________________________________________________________ -->
1862 <div class="doc_subsubsection">
1863 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
1866 <div class="doc_text">
1868 <pre> call void (va_list)* %llvm.va_end(va_list <arglist>)<br></pre>
1870 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
1871 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
1872 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
1874 <p>The argument is a <tt>va_list</tt> to destroy.</p>
1876 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
1877 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
1878 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
1879 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
1880 with calls to <tt>llvm.va_end</tt>.</p>
1883 <!-- _______________________________________________________________________ -->
1884 <div class="doc_subsubsection">
1885 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
1888 <div class="doc_text">
1890 <pre> call va_list (va_list)* %llvm.va_copy(va_list <destarglist>)<br></pre>
1892 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument
1893 position from the source argument list to the destination argument list.</p>
1895 <p>The argument is the <tt>va_list</tt> to copy.</p>
1897 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
1898 macro available in C. In a target-dependent way, it copies the source <tt>va_list</tt>
1899 element into the returned list. This intrinsic is necessary because the <tt><a
1900 href="i_va_start">llvm.va_start</a></tt> intrinsic may be arbitrarily
1901 complex and require memory allocation, for example.</p>
1904 <!-- ======================================================================= -->
1905 <div class="doc_subsection">
1906 <a name="int_codegen">Code Generator Intrinsics</a>
1909 <div class="doc_text">
1911 These intrinsics are provided by LLVM to expose special features that may only
1912 be implemented with code generator support.
1917 <!-- _______________________________________________________________________ -->
1918 <div class="doc_subsubsection">
1919 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
1922 <div class="doc_text">
1926 call void* ()* %llvm.returnaddress(uint <level>)
1932 The '<tt>llvm.returnaddress</tt>' intrinsic returns a target-specific value
1933 indicating the return address of the current function or one of its callers.
1939 The argument to this intrinsic indicates which function to return the address
1940 for. Zero indicates the calling function, one indicates its caller, etc. The
1941 argument is <b>required</b> to be a constant integer value.
1947 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
1948 the return address of the specified call frame, or zero if it cannot be
1949 identified. The value returned by this intrinsic is likely to be incorrect or 0
1950 for arguments other than zero, so it should only be used for debugging purposes.
1954 Note that calling this intrinsic does not prevent function inlining or other
1955 aggressive transformations, so the value returned may not that of the obvious
1956 source-language caller.
1961 <!-- _______________________________________________________________________ -->
1962 <div class="doc_subsubsection">
1963 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
1966 <div class="doc_text">
1970 call void* ()* %llvm.frameaddress(uint <level>)
1976 The '<tt>llvm.frameaddress</tt>' intrinsic returns the target-specific frame
1977 pointer value for the specified stack frame.
1983 The argument to this intrinsic indicates which function to return the frame
1984 pointer for. Zero indicates the calling function, one indicates its caller,
1985 etc. The argument is <b>required</b> to be a constant integer value.
1991 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
1992 the frame address of the specified call frame, or zero if it cannot be
1993 identified. The value returned by this intrinsic is likely to be incorrect or 0
1994 for arguments other than zero, so it should only be used for debugging purposes.
1998 Note that calling this intrinsic does not prevent function inlining or other
1999 aggressive transformations, so the value returned may not that of the obvious
2000 source-language caller.
2004 <!-- ======================================================================= -->
2005 <div class="doc_subsection">
2006 <a name="int_os">Operating System Intrinsics</a>
2009 <div class="doc_text">
2011 These intrinsics are provided by LLVM to support the implementation of
2012 operating system level code.
2016 <!-- _______________________________________________________________________ -->
2017 <div class="doc_subsubsection">
2018 <a name="i_readport">'<tt>llvm.readport</tt>' Intrinsic</a>
2021 <div class="doc_text">
2025 call <integer type> (<integer type>)* %llvm.readport (<integer type> <address>)
2031 The '<tt>llvm.readport</tt>' intrinsic reads data from the specified hardware
2038 The argument to this intrinsic indicates the hardware I/O address from which
2039 to read the data. The address is in the hardware I/O address namespace (as
2040 opposed to being a memory location for memory mapped I/O).
2046 The '<tt>llvm.readport</tt>' intrinsic reads data from the hardware I/O port
2047 specified by <i>address</i> and returns the value. The address and return
2048 value must be integers, but the size is dependent upon the platform upon which
2049 the program is code generated. For example, on x86, the address must be an
2050 unsigned 16 bit value, and the return value must be 8, 16, or 32 bits.
2055 <!-- _______________________________________________________________________ -->
2056 <div class="doc_subsubsection">
2057 <a name="i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a>
2060 <div class="doc_text">
2064 call void (<integer type>, <integer type>)* %llvm.writeport (<integer type> <value>, <integer type> <address>)
2070 The '<tt>llvm.writeport</tt>' intrinsic writes data to the specified hardware
2077 The first argument to this intrinsic indicates the hardware I/O address to
2078 which data should be written. The address is in the hardware I/O address
2079 namespace (as opposed to being a memory location for memory mapped I/O).
2083 The second argument is the value to write to the I/O port.
2089 The '<tt>llvm.writeport</tt>' intrinsic writes <i>value</i> to the I/O port
2090 specified by <i>address</i>. The address and value must be integers, but the
2091 size is dependent upon the platform upon which the program is code generated.
2092 For example, on x86, the address must be an unsigned 16 bit value, and the
2093 value written must be 8, 16, or 32 bits in length.
2098 <!-- ======================================================================= -->
2099 <div class="doc_subsection">
2100 <a name="int_libc">Standard C Library Intrinsics</a>
2103 <div class="doc_text">
2105 LLVM provides intrinsics for a few important standard C library functions.
2106 These intrinsics allow source-language front-ends to pass information about the
2107 alignment of the pointer arguments to the code generator, providing opportunity
2108 for more efficient code generation.
2113 <!-- _______________________________________________________________________ -->
2114 <div class="doc_subsubsection">
2115 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
2118 <div class="doc_text">
2122 call void (sbyte*, sbyte*, uint, uint)* %llvm.memcpy(sbyte* <dest>, sbyte* <src>,
2123 uint <len>, uint <align>)
2129 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2130 location to the destination location.
2134 Note that, unlike the standard libc function, the <tt>llvm.memcpy</tt> intrinsic
2135 does not return a value, and takes an extra alignment argument.
2141 The first argument is a pointer to the destination, the second is a pointer to
2142 the source. The third argument is an (arbitrarily sized) integer argument
2143 specifying the number of bytes to copy, and the fourth argument is the alignment
2144 of the source and destination locations.
2148 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2149 the caller guarantees that the size of the copy is a multiple of the alignment
2150 and that both the source and destination pointers are aligned to that boundary.
2156 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2157 location to the destination location, which are not allowed to overlap. It
2158 copies "len" bytes of memory over. If the argument is known to be aligned to
2159 some boundary, this can be specified as the fourth argument, otherwise it should
2165 <!-- _______________________________________________________________________ -->
2166 <div class="doc_subsubsection">
2167 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
2170 <div class="doc_text">
2174 call void (sbyte*, sbyte*, uint, uint)* %llvm.memmove(sbyte* <dest>, sbyte* <src>,
2175 uint <len>, uint <align>)
2181 The '<tt>llvm.memmove</tt>' intrinsic moves a block of memory from the source
2182 location to the destination location. It is similar to the '<tt>llvm.memcpy</tt>'
2183 intrinsic but allows the two memory locations to overlap.
2187 Note that, unlike the standard libc function, the <tt>llvm.memmove</tt> intrinsic
2188 does not return a value, and takes an extra alignment argument.
2194 The first argument is a pointer to the destination, the second is a pointer to
2195 the source. The third argument is an (arbitrarily sized) integer argument
2196 specifying the number of bytes to copy, and the fourth argument is the alignment
2197 of the source and destination locations.
2201 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2202 the caller guarantees that the size of the copy is a multiple of the alignment
2203 and that both the source and destination pointers are aligned to that boundary.
2209 The '<tt>llvm.memmove</tt>' intrinsic copies a block of memory from the source
2210 location to the destination location, which may overlap. It
2211 copies "len" bytes of memory over. If the argument is known to be aligned to
2212 some boundary, this can be specified as the fourth argument, otherwise it should
2218 <!-- _______________________________________________________________________ -->
2219 <div class="doc_subsubsection">
2220 <a name="i_memset">'<tt>llvm.memset</tt>' Intrinsic</a>
2223 <div class="doc_text">
2227 call void (sbyte*, ubyte, uint, uint)* %llvm.memset(sbyte* <dest>, ubyte <val>,
2228 uint <len>, uint <align>)
2234 The '<tt>llvm.memset</tt>' intrinsic fills a block of memory with a particular
2239 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
2240 does not return a value, and takes an extra alignment argument.
2246 The first argument is a pointer to the destination to fill, the second is the
2247 byte value to fill it with, the third argument is an (arbitrarily sized) integer
2248 argument specifying the number of bytes to fill, and the fourth argument is the
2249 known alignment of destination location.
2253 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2254 the caller guarantees that the size of the copy is a multiple of the alignment
2255 and that the destination pointer is aligned to that boundary.
2261 The '<tt>llvm.memset</tt>' intrinsic fills "len" bytes of memory starting at the
2262 destination location. If the argument is known to be aligned to some boundary,
2263 this can be specified as the fourth argument, otherwise it should be set to 0 or
2269 <!-- ======================================================================= -->
2270 <div class="doc_subsection">
2271 <a name="int_debugger">Debugger Intrinsics</a>
2274 <div class="doc_text">
2276 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
2277 are described in the <a
2278 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
2279 Debugging</a> document.
2284 <!-- *********************************************************************** -->
2287 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2288 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
2289 <a href="http://validator.w3.org/check/referer"><img
2290 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
2292 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
2293 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
2294 Last modified: $Date$