1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>LLVM Assembly Language Reference Manual</title>
6 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
8 <meta name="description"
9 content="LLVM Assembly Language Reference Manual.">
10 <link rel="stylesheet" href="llvm.css" type="text/css">
15 <div class="doc_title"> LLVM Language Reference Manual </div>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
20 <li><a href="#highlevel">High Level Structure</a>
22 <li><a href="#modulestructure">Module Structure</a></li>
23 <li><a href="#linkage">Linkage Types</a></li>
24 <li><a href="#globalvars">Global Variables</a></li>
25 <li><a href="#functionstructure">Function Structure</a></li>
28 <li><a href="#typesystem">Type System</a>
30 <li><a href="#t_primitive">Primitive Types</a>
32 <li><a href="#t_classifications">Type Classifications</a></li>
35 <li><a href="#t_derived">Derived Types</a>
37 <li><a href="#t_array">Array Type</a></li>
38 <li><a href="#t_function">Function Type</a></li>
39 <li><a href="#t_pointer">Pointer Type</a></li>
40 <li><a href="#t_struct">Structure Type</a></li>
41 <li><a href="#t_packed">Packed Type</a></li>
42 <li><a href="#t_opaque">Opaque Type</a></li>
47 <li><a href="#constants">Constants</a>
49 <li><a href="#simpleconstants">Simple Constants</a>
50 <li><a href="#aggregateconstants">Aggregate Constants</a>
51 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
52 <li><a href="#undefvalues">Undefined Values</a>
53 <li><a href="#constantexprs">Constant Expressions</a>
56 <li><a href="#instref">Instruction Reference</a>
58 <li><a href="#terminators">Terminator Instructions</a>
60 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
61 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
62 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
63 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
64 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
65 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
68 <li><a href="#binaryops">Binary Operations</a>
70 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
71 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
72 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
73 <li><a href="#i_div">'<tt>div</tt>' Instruction</a></li>
74 <li><a href="#i_rem">'<tt>rem</tt>' Instruction</a></li>
75 <li><a href="#i_setcc">'<tt>set<i>cc</i></tt>' Instructions</a></li>
78 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
80 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
81 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
82 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
83 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
84 <li><a href="#i_shr">'<tt>shr</tt>' Instruction</a></li>
87 <li><a href="#memoryops">Memory Access Operations</a>
89 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
90 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
91 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
92 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
93 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
94 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
97 <li><a href="#otherops">Other Operations</a>
99 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
100 <li><a href="#i_cast">'<tt>cast .. to</tt>' Instruction</a></li>
101 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
102 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
103 <li><a href="#i_vanext">'<tt>vanext</tt>' Instruction</a></li>
104 <li><a href="#i_vaarg">'<tt>vaarg</tt>' Instruction</a></li>
109 <li><a href="#intrinsics">Intrinsic Functions</a>
111 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
113 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
114 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
115 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
118 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
120 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
121 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
122 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
125 <li><a href="#int_codegen">Code Generator Intrinsics</a>
127 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
128 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
129 <li><a href="#i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
130 <li><a href="#i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
133 <li><a href="#int_os">Operating System Intrinsics</a>
135 <li><a href="#i_readport">'<tt>llvm.readport</tt>' Intrinsic</a></li>
136 <li><a href="#i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a></li>
137 <li><a href="#i_readio">'<tt>llvm.readio</tt>' Intrinsic</a></li>
138 <li><a href="#i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a></li>
140 <li><a href="#int_libc">Standard C Library Intrinsics</a>
142 <li><a href="#i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a></li>
143 <li><a href="#i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a></li>
144 <li><a href="#i_memset">'<tt>llvm.memset</tt>' Intrinsic</a></li>
145 <li><a href="#i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a></li>
148 <li><a href="#int_debugger">Debugger intrinsics</a></li>
153 <div class="doc_author">
154 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
155 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
158 <!-- *********************************************************************** -->
159 <div class="doc_section"> <a name="abstract">Abstract </a></div>
160 <!-- *********************************************************************** -->
162 <div class="doc_text">
163 <p>This document is a reference manual for the LLVM assembly language.
164 LLVM is an SSA based representation that provides type safety,
165 low-level operations, flexibility, and the capability of representing
166 'all' high-level languages cleanly. It is the common code
167 representation used throughout all phases of the LLVM compilation
171 <!-- *********************************************************************** -->
172 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
173 <!-- *********************************************************************** -->
175 <div class="doc_text">
177 <p>The LLVM code representation is designed to be used in three
178 different forms: as an in-memory compiler IR, as an on-disk bytecode
179 representation (suitable for fast loading by a Just-In-Time compiler),
180 and as a human readable assembly language representation. This allows
181 LLVM to provide a powerful intermediate representation for efficient
182 compiler transformations and analysis, while providing a natural means
183 to debug and visualize the transformations. The three different forms
184 of LLVM are all equivalent. This document describes the human readable
185 representation and notation.</p>
187 <p>The LLVM representation aims to be a light-weight and low-level
188 while being expressive, typed, and extensible at the same time. It
189 aims to be a "universal IR" of sorts, by being at a low enough level
190 that high-level ideas may be cleanly mapped to it (similar to how
191 microprocessors are "universal IR's", allowing many source languages to
192 be mapped to them). By providing type information, LLVM can be used as
193 the target of optimizations: for example, through pointer analysis, it
194 can be proven that a C automatic variable is never accessed outside of
195 the current function... allowing it to be promoted to a simple SSA
196 value instead of a memory location.</p>
200 <!-- _______________________________________________________________________ -->
201 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
203 <div class="doc_text">
205 <p>It is important to note that this document describes 'well formed'
206 LLVM assembly language. There is a difference between what the parser
207 accepts and what is considered 'well formed'. For example, the
208 following instruction is syntactically okay, but not well formed:</p>
211 %x = <a href="#i_add">add</a> int 1, %x
214 <p>...because the definition of <tt>%x</tt> does not dominate all of
215 its uses. The LLVM infrastructure provides a verification pass that may
216 be used to verify that an LLVM module is well formed. This pass is
217 automatically run by the parser after parsing input assembly, and by
218 the optimizer before it outputs bytecode. The violations pointed out
219 by the verifier pass indicate bugs in transformation passes or input to
222 <!-- Describe the typesetting conventions here. --> </div>
224 <!-- *********************************************************************** -->
225 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
226 <!-- *********************************************************************** -->
228 <div class="doc_text">
230 <p>LLVM uses three different forms of identifiers, for different
234 <li>Named values are represented as a string of characters with a '%' prefix.
235 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
236 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
237 Identifiers which require other characters in their names can be surrounded
238 with quotes. In this way, anything except a <tt>"</tt> character can be used
241 <li>Unnamed values are represented as an unsigned numeric value with a '%'
242 prefix. For example, %12, %2, %44.</li>
244 <li>Constants, which are described in a <a href="#constants">section about
245 constants</a>, below.</li>
248 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
249 don't need to worry about name clashes with reserved words, and the set of
250 reserved words may be expanded in the future without penalty. Additionally,
251 unnamed identifiers allow a compiler to quickly come up with a temporary
252 variable without having to avoid symbol table conflicts.</p>
254 <p>Reserved words in LLVM are very similar to reserved words in other
255 languages. There are keywords for different opcodes ('<tt><a
256 href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
257 href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
258 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>', etc...),
259 and others. These reserved words cannot conflict with variable names, because
260 none of them start with a '%' character.</p>
262 <p>Here is an example of LLVM code to multiply the integer variable
263 '<tt>%X</tt>' by 8:</p>
268 %result = <a href="#i_mul">mul</a> uint %X, 8
271 <p>After strength reduction:</p>
274 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
277 <p>And the hard way:</p>
280 <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
281 <a href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
282 %result = <a href="#i_add">add</a> uint %1, %1
285 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
286 important lexical features of LLVM:</p>
290 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
293 <li>Unnamed temporaries are created when the result of a computation is not
294 assigned to a named value.</li>
296 <li>Unnamed temporaries are numbered sequentially</li>
300 <p>...and it also show a convention that we follow in this document. When
301 demonstrating instructions, we will follow an instruction with a comment that
302 defines the type and name of value produced. Comments are shown in italic
307 <!-- *********************************************************************** -->
308 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
309 <!-- *********************************************************************** -->
311 <!-- ======================================================================= -->
312 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
315 <div class="doc_text">
317 <p>LLVM programs are composed of "Module"s, each of which is a
318 translation unit of the input programs. Each module consists of
319 functions, global variables, and symbol table entries. Modules may be
320 combined together with the LLVM linker, which merges function (and
321 global variable) definitions, resolves forward declarations, and merges
322 symbol table entries. Here is an example of the "hello world" module:</p>
324 <pre><i>; Declare the string constant as a global constant...</i>
325 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
326 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
328 <i>; External declaration of the puts function</i>
329 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
331 <i>; Definition of main function</i>
332 int %main() { <i>; int()* </i>
333 <i>; Convert [13x sbyte]* to sbyte *...</i>
335 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
337 <i>; Call puts function to write out the string to stdout...</i>
339 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
341 href="#i_ret">ret</a> int 0<br>}<br></pre>
343 <p>This example is made up of a <a href="#globalvars">global variable</a>
344 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
345 function, and a <a href="#functionstructure">function definition</a>
346 for "<tt>main</tt>".</p>
348 <p>In general, a module is made up of a list of global values,
349 where both functions and global variables are global values. Global values are
350 represented by a pointer to a memory location (in this case, a pointer to an
351 array of char, and a pointer to a function), and have one of the following <a
352 href="#linkage">linkage types</a>.</p>
356 <!-- ======================================================================= -->
357 <div class="doc_subsection">
358 <a name="linkage">Linkage Types</a>
361 <div class="doc_text">
364 All Global Variables and Functions have one of the following types of linkage:
369 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
371 <dd>Global values with internal linkage are only directly accessible by
372 objects in the current module. In particular, linking code into a module with
373 an internal global value may cause the internal to be renamed as necessary to
374 avoid collisions. Because the symbol is internal to the module, all
375 references can be updated. This corresponds to the notion of the
376 '<tt>static</tt>' keyword in C, or the idea of "anonymous namespaces" in C++.
379 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
381 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
382 the twist that linking together two modules defining the same
383 <tt>linkonce</tt> globals will cause one of the globals to be discarded. This
384 is typically used to implement inline functions. Unreferenced
385 <tt>linkonce</tt> globals are allowed to be discarded.
388 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
390 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
391 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
392 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.
395 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
397 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
398 pointer to array type. When two global variables with appending linkage are
399 linked together, the two global arrays are appended together. This is the
400 LLVM, typesafe, equivalent of having the system linker append together
401 "sections" with identical names when .o files are linked.
404 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
406 <dd>If none of the above identifiers are used, the global is externally
407 visible, meaning that it participates in linkage and can be used to resolve
408 external symbol references.
412 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
413 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
414 variable and was linked with this one, one of the two would be renamed,
415 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
416 external (i.e., lacking any linkage declarations), they are accessible
417 outside of the current module. It is illegal for a function <i>declaration</i>
418 to have any linkage type other than "externally visible".</a></p>
422 <!-- ======================================================================= -->
423 <div class="doc_subsection">
424 <a name="globalvars">Global Variables</a>
427 <div class="doc_text">
429 <p>Global variables define regions of memory allocated at compilation time
430 instead of run-time. Global variables may optionally be initialized. A
431 variable may be defined as a global "constant", which indicates that the
432 contents of the variable will <b>never</b> be modified (enabling better
433 optimization, allowing the global data to be placed in the read-only section of
434 an executable, etc). Note that variables that need runtime initialization
435 cannot be marked "constant", as there is a store to the variable.</p>
438 LLVM explicitly allows <em>declarations</em> of global variables to be marked
439 constant, even if the final definition of the global is not. This capability
440 can be used to enable slightly better optimization of the program, but requires
441 the language definition to guarantee that optimizations based on the
442 'constantness' are valid for the translation units that do not include the
446 <p>As SSA values, global variables define pointer values that are in
447 scope (i.e. they dominate) all basic blocks in the program. Global
448 variables always define a pointer to their "content" type because they
449 describe a region of memory, and all memory objects in LLVM are
450 accessed through pointers.</p>
455 <!-- ======================================================================= -->
456 <div class="doc_subsection">
457 <a name="functionstructure">Functions</a>
460 <div class="doc_text">
462 <p>LLVM function definitions are composed of a (possibly empty) argument list,
463 an opening curly brace, a list of basic blocks, and a closing curly brace. LLVM
464 function declarations are defined with the "<tt>declare</tt>" keyword, a
465 function name, and a function signature.</p>
467 <p>A function definition contains a list of basic blocks, forming the CFG for
468 the function. Each basic block may optionally start with a label (giving the
469 basic block a symbol table entry), contains a list of instructions, and ends
470 with a <a href="#terminators">terminator</a> instruction (such as a branch or
471 function return).</p>
473 <p>The first basic block in program is special in two ways: it is immediately
474 executed on entrance to the function, and it is not allowed to have predecessor
475 basic blocks (i.e. there can not be any branches to the entry block of a
476 function). Because the block can have no predecessors, it also cannot have any
477 <a href="#i_phi">PHI nodes</a>.</p>
479 <p>LLVM functions are identified by their name and type signature. Hence, two
480 functions with the same name but different parameter lists or return values are
481 considered different functions, and LLVM will resolve references to each
488 <!-- *********************************************************************** -->
489 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
490 <!-- *********************************************************************** -->
492 <div class="doc_text">
494 <p>The LLVM type system is one of the most important features of the
495 intermediate representation. Being typed enables a number of
496 optimizations to be performed on the IR directly, without having to do
497 extra analyses on the side before the transformation. A strong type
498 system makes it easier to read the generated code and enables novel
499 analyses and transformations that are not feasible to perform on normal
500 three address code representations.</p>
504 <!-- ======================================================================= -->
505 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
506 <div class="doc_text">
507 <p>The primitive types are the fundamental building blocks of the LLVM
508 system. The current set of primitive types is as follows:</p>
510 <table class="layout">
515 <tr><th>Type</th><th>Description</th></tr>
516 <tr><td><tt>void</tt></td><td>No value</td></tr>
517 <tr><td><tt>ubyte</tt></td><td>Unsigned 8-bit value</td></tr>
518 <tr><td><tt>ushort</tt></td><td>Unsigned 16-bit value</td></tr>
519 <tr><td><tt>uint</tt></td><td>Unsigned 32-bit value</td></tr>
520 <tr><td><tt>ulong</tt></td><td>Unsigned 64-bit value</td></tr>
521 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
522 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
529 <tr><th>Type</th><th>Description</th></tr>
530 <tr><td><tt>bool</tt></td><td>True or False value</td></tr>
531 <tr><td><tt>sbyte</tt></td><td>Signed 8-bit value</td></tr>
532 <tr><td><tt>short</tt></td><td>Signed 16-bit value</td></tr>
533 <tr><td><tt>int</tt></td><td>Signed 32-bit value</td></tr>
534 <tr><td><tt>long</tt></td><td>Signed 64-bit value</td></tr>
535 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
543 <!-- _______________________________________________________________________ -->
544 <div class="doc_subsubsection"> <a name="t_classifications">Type
545 Classifications</a> </div>
546 <div class="doc_text">
547 <p>These different primitive types fall into a few useful
550 <table border="1" cellspacing="0" cellpadding="4">
552 <tr><th>Classification</th><th>Types</th></tr>
554 <td><a name="t_signed">signed</a></td>
555 <td><tt>sbyte, short, int, long, float, double</tt></td>
558 <td><a name="t_unsigned">unsigned</a></td>
559 <td><tt>ubyte, ushort, uint, ulong</tt></td>
562 <td><a name="t_integer">integer</a></td>
563 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
566 <td><a name="t_integral">integral</a></td>
567 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt>
571 <td><a name="t_floating">floating point</a></td>
572 <td><tt>float, double</tt></td>
575 <td><a name="t_firstclass">first class</a></td>
576 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long,<br>
577 float, double, <a href="#t_pointer">pointer</a>,
578 <a href="#t_packed">packed</a></tt></td>
583 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
584 most important. Values of these types are the only ones which can be
585 produced by instructions, passed as arguments, or used as operands to
586 instructions. This means that all structures and arrays must be
587 manipulated either by pointer or by component.</p>
590 <!-- ======================================================================= -->
591 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
593 <div class="doc_text">
595 <p>The real power in LLVM comes from the derived types in the system.
596 This is what allows a programmer to represent arrays, functions,
597 pointers, and other useful types. Note that these derived types may be
598 recursive: For example, it is possible to have a two dimensional array.</p>
602 <!-- _______________________________________________________________________ -->
603 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
605 <div class="doc_text">
609 <p>The array type is a very simple derived type that arranges elements
610 sequentially in memory. The array type requires a size (number of
611 elements) and an underlying data type.</p>
616 [<# elements> x <elementtype>]
619 <p>The number of elements is a constant integer value, elementtype may
620 be any type with a size.</p>
623 <table class="layout">
626 <tt>[40 x int ]</tt><br/>
627 <tt>[41 x int ]</tt><br/>
628 <tt>[40 x uint]</tt><br/>
631 Array of 40 integer values.<br/>
632 Array of 41 integer values.<br/>
633 Array of 40 unsigned integer values.<br/>
637 <p>Here are some examples of multidimensional arrays:</p>
638 <table class="layout">
641 <tt>[3 x [4 x int]]</tt><br/>
642 <tt>[12 x [10 x float]]</tt><br/>
643 <tt>[2 x [3 x [4 x uint]]]</tt><br/>
646 3x4 array integer values.<br/>
647 12x10 array of single precision floating point values.<br/>
648 2x3x4 array of unsigned integer values.<br/>
654 <!-- _______________________________________________________________________ -->
655 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
656 <div class="doc_text">
658 <p>The function type can be thought of as a function signature. It
659 consists of a return type and a list of formal parameter types.
660 Function types are usually used to build virtual function tables
661 (which are structures of pointers to functions), for indirect function
662 calls, and when defining a function.</p>
664 The return type of a function type cannot be an aggregate type.
667 <pre> <returntype> (<parameter list>)<br></pre>
668 <p>Where '<tt><parameter list></tt>' is a comma-separated list of type
669 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
670 which indicates that the function takes a variable number of arguments.
671 Variable argument functions can access their arguments with the <a
672 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
674 <table class="layout">
677 <tt>int (int)</tt> <br/>
678 <tt>float (int, int *) *</tt><br/>
679 <tt>int (sbyte *, ...)</tt><br/>
682 function taking an <tt>int</tt>, returning an <tt>int</tt><br/>
683 <a href="#t_pointer">Pointer</a> to a function that takes an
684 <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
685 returning <tt>float</tt>.<br/>
686 A vararg function that takes at least one <a href="#t_pointer">pointer</a>
687 to <tt>sbyte</tt> (signed char in C), which returns an integer. This is
688 the signature for <tt>printf</tt> in LLVM.<br/>
694 <!-- _______________________________________________________________________ -->
695 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
696 <div class="doc_text">
698 <p>The structure type is used to represent a collection of data members
699 together in memory. The packing of the field types is defined to match
700 the ABI of the underlying processor. The elements of a structure may
701 be any type that has a size.</p>
702 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
703 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
704 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
707 <pre> { <type list> }<br></pre>
709 <table class="layout">
712 <tt>{ int, int, int }</tt><br/>
713 <tt>{ float, int (int) * }</tt><br/>
716 a triple of three <tt>int</tt> values<br/>
717 A pair, where the first element is a <tt>float</tt> and the second element
718 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
719 that takes an <tt>int</tt>, returning an <tt>int</tt>.<br/>
725 <!-- _______________________________________________________________________ -->
726 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
727 <div class="doc_text">
729 <p>As in many languages, the pointer type represents a pointer or
730 reference to another object, which must live in memory.</p>
732 <pre> <type> *<br></pre>
734 <table class="layout">
737 <tt>[4x int]*</tt><br/>
738 <tt>int (int *) *</tt><br/>
741 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
742 four <tt>int</tt> values<br/>
743 A <a href="#t_pointer">pointer</a> to a <a
744 href="#t_function">function</a> that takes an <tt>int*</tt>, returning an
751 <!-- _______________________________________________________________________ -->
752 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
753 <div class="doc_text">
757 <p>A packed type is a simple derived type that represents a vector
758 of elements. Packed types are used when multiple primitive data
759 are operated in parallel using a single instruction (SIMD).
760 A packed type requires a size (number of
761 elements) and an underlying primitive data type. Packed types are
762 considered <a href="#t_firstclass">first class</a>.</p>
767 < <# elements> x <elementtype> >
770 <p>The number of elements is a constant integer value, elementtype may
771 be any integral or floating point type.</p>
775 <table class="layout">
778 <tt><4 x int></tt><br/>
779 <tt><8 x float></tt><br/>
780 <tt><2 x uint></tt><br/>
783 Packed vector of 4 integer values.<br/>
784 Packed vector of 8 floating-point values.<br/>
785 Packed vector of 2 unsigned integer values.<br/>
791 <!-- _______________________________________________________________________ -->
792 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
793 <div class="doc_text">
797 <p>Opaque types are used to represent unknown types in the system. This
798 corresponds (for example) to the C notion of a foward declared structure type.
799 In LLVM, opaque types can eventually be resolved to any type (not just a
810 <table class="layout">
823 <!-- *********************************************************************** -->
824 <div class="doc_section"> <a name="constants">Constants</a> </div>
825 <!-- *********************************************************************** -->
827 <div class="doc_text">
829 <p>LLVM has several different basic types of constants. This section describes
830 them all and their syntax.</p>
834 <!-- ======================================================================= -->
835 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
837 <div class="doc_text">
840 <dt><b>Boolean constants</b></dt>
842 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
843 constants of the <tt><a href="#t_primitive">bool</a></tt> type.
846 <dt><b>Integer constants</b></dt>
848 <dd>Standard integers (such as '4') are constants of the <a
849 href="#t_integer">integer</a> type. Negative numbers may be used with signed
853 <dt><b>Floating point constants</b></dt>
855 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
856 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
857 notation. Floating point constants have an optional hexadecimal
858 notation (see below). Floating point constants must have a <a
859 href="#t_floating">floating point</a> type. </dd>
861 <dt><b>Null pointer constants</b></dt>
863 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
864 and must be of <a href="#t_pointer">pointer type</a>.</dd>
868 <p>The one non-intuitive notation for constants is the optional hexadecimal form
869 of floating point constants. For example, the form '<tt>double
870 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
871 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
872 (and the only time that they are generated by the disassembler) is when a
873 floating point constant must be emitted but it cannot be represented as a
874 decimal floating point number. For example, NaN's, infinities, and other
875 special values are represented in their IEEE hexadecimal format so that
876 assembly and disassembly do not cause any bits to change in the constants.</p>
880 <!-- ======================================================================= -->
881 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
884 <div class="doc_text">
885 <p>Aggregate constants arise from aggregation of simple constants
886 and smaller aggregate constants.</p>
889 <dt><b>Structure constants</b></dt>
891 <dd>Structure constants are represented with notation similar to structure
892 type definitions (a comma separated list of elements, surrounded by braces
893 (<tt>{}</tt>)). For example: "<tt>{ int 4, float 17.0, int* %G }</tt>",
894 where "<tt>%G</tt>" is declared as "<tt>%G = external global int</tt>". Structure constants
895 must have <a href="#t_struct">structure type</a>, and the number and
896 types of elements must match those specified by the type.
899 <dt><b>Array constants</b></dt>
901 <dd>Array constants are represented with notation similar to array type
902 definitions (a comma separated list of elements, surrounded by square brackets
903 (<tt>[]</tt>)). For example: "<tt>[ int 42, int 11, int 74 ]</tt>". Array
904 constants must have <a href="#t_array">array type</a>, and the number and
905 types of elements must match those specified by the type.
908 <dt><b>Packed constants</b></dt>
910 <dd>Packed constants are represented with notation similar to packed type
911 definitions (a comma separated list of elements, surrounded by
912 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< int 42,
913 int 11, int 74, int 100 ></tt>". Packed constants must have <a
914 href="#t_packed">packed type</a>, and the number and types of elements must
915 match those specified by the type.
918 <dt><b>Zero initialization</b></dt>
920 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
921 value to zero of <em>any</em> type, including scalar and aggregate types.
922 This is often used to avoid having to print large zero initializers (e.g. for
923 large arrays), and is always exactly equivalent to using explicit zero
930 <!-- ======================================================================= -->
931 <div class="doc_subsection">
932 <a name="globalconstants">Global Variable and Function Addresses</a>
935 <div class="doc_text">
937 <p>The addresses of <a href="#globalvars">global variables</a> and <a
938 href="#functionstructure">functions</a> are always implicitly valid (link-time)
939 constants. These constants are explicitly referenced when the <a
940 href="#identifiers">identifier for the global</a> is used and always have <a
941 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
947 %Z = global [2 x int*] [ int* %X, int* %Y ]
952 <!-- ======================================================================= -->
953 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
954 <div class="doc_text">
955 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
956 no specific value. Undefined values may be of any type, and be used anywhere
957 a constant is permitted.</p>
959 <p>Undefined values indicate to the compiler that the program is well defined
960 no matter what value is used, giving the compiler more freedom to optimize.
964 <!-- ======================================================================= -->
965 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
968 <div class="doc_text">
970 <p>Constant expressions are used to allow expressions involving other constants
971 to be used as constants. Constant expressions may be of any <a
972 href="#t_firstclass">first class</a> type, and may involve any LLVM operation
973 that does not have side effects (e.g. load and call are not supported). The
974 following is the syntax for constant expressions:</p>
977 <dt><b><tt>cast ( CST to TYPE )</tt></b></dt>
979 <dd>Cast a constant to another type.</dd>
981 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
983 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
984 constants. As with the <a href="#i_getelementptr">getelementptr</a>
985 instruction, the index list may have zero or more indexes, which are required
986 to make sense for the type of "CSTPTR".</dd>
988 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
990 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
991 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
992 binary</a> operations. The constraints on operands are the same as those for
993 the corresponding instruction (e.g. no bitwise operations on floating point
998 <!-- *********************************************************************** -->
999 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1000 <!-- *********************************************************************** -->
1002 <div class="doc_text">
1004 <p>The LLVM instruction set consists of several different
1005 classifications of instructions: <a href="#terminators">terminator
1006 instructions</a>, <a href="#binaryops">binary instructions</a>, <a
1007 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1008 instructions</a>.</p>
1012 <!-- ======================================================================= -->
1013 <div class="doc_subsection"> <a name="terminators">Terminator
1014 Instructions</a> </div>
1016 <div class="doc_text">
1018 <p>As mentioned <a href="#functionstructure">previously</a>, every
1019 basic block in a program ends with a "Terminator" instruction, which
1020 indicates which block should be executed after the current block is
1021 finished. These terminator instructions typically yield a '<tt>void</tt>'
1022 value: they produce control flow, not values (the one exception being
1023 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1024 <p>There are six different terminator instructions: the '<a
1025 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1026 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1027 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1028 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1029 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1033 <!-- _______________________________________________________________________ -->
1034 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1035 Instruction</a> </div>
1036 <div class="doc_text">
1038 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1039 ret void <i>; Return from void function</i>
1042 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1043 value) from a function, back to the caller.</p>
1044 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1045 returns a value and then causes control flow, and one that just causes
1046 control flow to occur.</p>
1048 <p>The '<tt>ret</tt>' instruction may return any '<a
1049 href="#t_firstclass">first class</a>' type. Notice that a function is
1050 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1051 instruction inside of the function that returns a value that does not
1052 match the return type of the function.</p>
1054 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1055 returns back to the calling function's context. If the caller is a "<a
1056 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1057 the instruction after the call. If the caller was an "<a
1058 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1059 at the beginning "normal" of the destination block. If the instruction
1060 returns a value, that value shall set the call or invoke instruction's
1063 <pre> ret int 5 <i>; Return an integer value of 5</i>
1064 ret void <i>; Return from a void function</i>
1067 <!-- _______________________________________________________________________ -->
1068 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1069 <div class="doc_text">
1071 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1074 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1075 transfer to a different basic block in the current function. There are
1076 two forms of this instruction, corresponding to a conditional branch
1077 and an unconditional branch.</p>
1079 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1080 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
1081 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1082 value as a target.</p>
1084 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
1085 argument is evaluated. If the value is <tt>true</tt>, control flows
1086 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1087 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1089 <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
1090 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
1092 <!-- _______________________________________________________________________ -->
1093 <div class="doc_subsubsection">
1094 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1097 <div class="doc_text">
1101 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1106 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1107 several different places. It is a generalization of the '<tt>br</tt>'
1108 instruction, allowing a branch to occur to one of many possible
1114 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1115 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1116 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1117 table is not allowed to contain duplicate constant entries.</p>
1121 <p>The <tt>switch</tt> instruction specifies a table of values and
1122 destinations. When the '<tt>switch</tt>' instruction is executed, this
1123 table is searched for the given value. If the value is found, control flow is
1124 transfered to the corresponding destination; otherwise, control flow is
1125 transfered to the default destination.</p>
1127 <h5>Implementation:</h5>
1129 <p>Depending on properties of the target machine and the particular
1130 <tt>switch</tt> instruction, this instruction may be code generated in different
1131 ways. For example, it could be generated as a series of chained conditional
1132 branches or with a lookup table.</p>
1137 <i>; Emulate a conditional br instruction</i>
1138 %Val = <a href="#i_cast">cast</a> bool %value to int
1139 switch int %Val, label %truedest [int 0, label %falsedest ]
1141 <i>; Emulate an unconditional br instruction</i>
1142 switch uint 0, label %dest [ ]
1144 <i>; Implement a jump table:</i>
1145 switch uint %val, label %otherwise [ uint 0, label %onzero
1146 uint 1, label %onone
1147 uint 2, label %ontwo ]
1150 <!-- _______________________________________________________________________ -->
1151 <div class="doc_subsubsection"> <a name="i_invoke">'<tt>invoke</tt>'
1152 Instruction</a> </div>
1153 <div class="doc_text">
1155 <pre> <result> = invoke <ptr to function ty> %<function ptr val>(<function args>)<br> to label <normal label> except label <exception label><br></pre>
1157 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a
1158 specified function, with the possibility of control flow transfer to
1159 either the '<tt>normal</tt>' <tt>label</tt> label or the '<tt>exception</tt>'<tt>label</tt>.
1160 If the callee function returns with the "<tt><a href="#i_ret">ret</a></tt>"
1161 instruction, control flow will return to the "normal" label. If the
1162 callee (or any indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
1163 instruction, control is interrupted, and continued at the dynamically
1164 nearest "except" label.</p>
1166 <p>This instruction requires several arguments:</p>
1168 <li>'<tt>ptr to function ty</tt>': shall be the signature of the
1169 pointer to function value being invoked. In most cases, this is a
1170 direct function invocation, but indirect <tt>invoke</tt>s are just as
1171 possible, branching off an arbitrary pointer to function value. </li>
1172 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer
1173 to a function to be invoked. </li>
1174 <li>'<tt>function args</tt>': argument list whose types match the
1175 function signature argument types. If the function signature indicates
1176 the function accepts a variable number of arguments, the extra
1177 arguments can be specified. </li>
1178 <li>'<tt>normal label</tt>': the label reached when the called
1179 function executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1180 <li>'<tt>exception label</tt>': the label reached when a callee
1181 returns with the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1184 <p>This instruction is designed to operate as a standard '<tt><a
1185 href="#i_call">call</a></tt>' instruction in most regards. The
1186 primary difference is that it establishes an association with a label,
1187 which is used by the runtime library to unwind the stack.</p>
1188 <p>This instruction is used in languages with destructors to ensure
1189 that proper cleanup is performed in the case of either a <tt>longjmp</tt>
1190 or a thrown exception. Additionally, this is important for
1191 implementation of '<tt>catch</tt>' clauses in high-level languages that
1194 <pre> %retval = invoke int %Test(int 15)<br> to label %Continue<br> except label %TestCleanup <i>; {int}:retval set</i>
1199 <!-- _______________________________________________________________________ -->
1201 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1202 Instruction</a> </div>
1204 <div class="doc_text">
1213 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1214 at the first callee in the dynamic call stack which used an <a
1215 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1216 primarily used to implement exception handling.</p>
1220 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1221 immediately halt. The dynamic call stack is then searched for the first <a
1222 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1223 execution continues at the "exceptional" destination block specified by the
1224 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1225 dynamic call chain, undefined behavior results.</p>
1228 <!-- _______________________________________________________________________ -->
1230 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1231 Instruction</a> </div>
1233 <div class="doc_text">
1242 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1243 instruction is used to inform the optimizer that a particular portion of the
1244 code is not reachable. This can be used to indicate that the code after a
1245 no-return function cannot be reached, and other facts.</p>
1249 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1254 <!-- ======================================================================= -->
1255 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1256 <div class="doc_text">
1257 <p>Binary operators are used to do most of the computation in a
1258 program. They require two operands, execute an operation on them, and
1259 produce a single value. The operands might represent
1260 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1261 The result value of a binary operator is not
1262 necessarily the same type as its operands.</p>
1263 <p>There are several different binary operators:</p>
1265 <!-- _______________________________________________________________________ -->
1266 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1267 Instruction</a> </div>
1268 <div class="doc_text">
1270 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1273 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1275 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1276 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1277 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1278 Both arguments must have identical types.</p>
1280 <p>The value produced is the integer or floating point sum of the two
1283 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
1286 <!-- _______________________________________________________________________ -->
1287 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1288 Instruction</a> </div>
1289 <div class="doc_text">
1291 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1294 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1296 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1297 instruction present in most other intermediate representations.</p>
1299 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1300 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1302 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1303 Both arguments must have identical types.</p>
1305 <p>The value produced is the integer or floating point difference of
1306 the two operands.</p>
1308 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
1309 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
1312 <!-- _______________________________________________________________________ -->
1313 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1314 Instruction</a> </div>
1315 <div class="doc_text">
1317 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1320 <p>The '<tt>mul</tt>' instruction returns the product of its two
1323 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1324 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1326 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1327 Both arguments must have identical types.</p>
1329 <p>The value produced is the integer or floating point product of the
1331 <p>There is no signed vs unsigned multiplication. The appropriate
1332 action is taken based on the type of the operand.</p>
1334 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
1337 <!-- _______________________________________________________________________ -->
1338 <div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
1339 Instruction</a> </div>
1340 <div class="doc_text">
1342 <pre> <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1345 <p>The '<tt>div</tt>' instruction returns the quotient of its two
1348 <p>The two arguments to the '<tt>div</tt>' instruction must be either <a
1349 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1351 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1352 Both arguments must have identical types.</p>
1354 <p>The value produced is the integer or floating point quotient of the
1357 <pre> <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
1360 <!-- _______________________________________________________________________ -->
1361 <div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
1362 Instruction</a> </div>
1363 <div class="doc_text">
1365 <pre> <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1368 <p>The '<tt>rem</tt>' instruction returns the remainder from the
1369 division of its two operands.</p>
1371 <p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1372 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1374 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1375 Both arguments must have identical types.</p>
1377 <p>This returns the <i>remainder</i> of a division (where the result
1378 has the same sign as the divisor), not the <i>modulus</i> (where the
1379 result has the same sign as the dividend) of a value. For more
1380 information about the difference, see: <a
1381 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1384 <pre> <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1387 <!-- _______________________________________________________________________ -->
1388 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1389 Instructions</a> </div>
1390 <div class="doc_text">
1392 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1393 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1394 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1395 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1396 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1397 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1400 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1401 value based on a comparison of their two operands.</p>
1403 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1404 be of <a href="#t_firstclass">first class</a> type (it is not possible
1405 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1406 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1409 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1410 value if both operands are equal.<br>
1411 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1412 value if both operands are unequal.<br>
1413 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1414 value if the first operand is less than the second operand.<br>
1415 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1416 value if the first operand is greater than the second operand.<br>
1417 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1418 value if the first operand is less than or equal to the second operand.<br>
1419 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1420 value if the first operand is greater than or equal to the second
1423 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1424 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1425 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1426 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1427 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1428 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1431 <!-- ======================================================================= -->
1432 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1433 Operations</a> </div>
1434 <div class="doc_text">
1435 <p>Bitwise binary operators are used to do various forms of
1436 bit-twiddling in a program. They are generally very efficient
1437 instructions and can commonly be strength reduced from other
1438 instructions. They require two operands, execute an operation on them,
1439 and produce a single value. The resulting value of the bitwise binary
1440 operators is always the same type as its first operand.</p>
1442 <!-- _______________________________________________________________________ -->
1443 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1444 Instruction</a> </div>
1445 <div class="doc_text">
1447 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1450 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1451 its two operands.</p>
1453 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1454 href="#t_integral">integral</a> values. Both arguments must have
1455 identical types.</p>
1457 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1459 <div style="align: center">
1460 <table border="1" cellspacing="0" cellpadding="4">
1491 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1492 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1493 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1496 <!-- _______________________________________________________________________ -->
1497 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1498 <div class="doc_text">
1500 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1503 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1504 or of its two operands.</p>
1506 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1507 href="#t_integral">integral</a> values. Both arguments must have
1508 identical types.</p>
1510 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1512 <div style="align: center">
1513 <table border="1" cellspacing="0" cellpadding="4">
1544 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1545 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1546 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1549 <!-- _______________________________________________________________________ -->
1550 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1551 Instruction</a> </div>
1552 <div class="doc_text">
1554 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1557 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1558 or of its two operands. The <tt>xor</tt> is used to implement the
1559 "one's complement" operation, which is the "~" operator in C.</p>
1561 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1562 href="#t_integral">integral</a> values. Both arguments must have
1563 identical types.</p>
1565 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1567 <div style="align: center">
1568 <table border="1" cellspacing="0" cellpadding="4">
1600 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1601 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1602 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1603 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1606 <!-- _______________________________________________________________________ -->
1607 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1608 Instruction</a> </div>
1609 <div class="doc_text">
1611 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1614 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1615 the left a specified number of bits.</p>
1617 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
1618 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1621 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
1623 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1624 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1625 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1628 <!-- _______________________________________________________________________ -->
1629 <div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1630 Instruction</a> </div>
1631 <div class="doc_text">
1633 <pre> <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1636 <p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1637 the right a specified number of bits.</p>
1639 <p>The first argument to the '<tt>shr</tt>' instruction must be an <a
1640 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1643 <p>If the first argument is a <a href="#t_signed">signed</a> type, the
1644 most significant bit is duplicated in the newly free'd bit positions.
1645 If the first argument is unsigned, zero bits shall fill the empty
1648 <pre> <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1649 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1650 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1651 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1652 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1655 <!-- ======================================================================= -->
1656 <div class="doc_subsection"> <a name="memoryops">Memory Access
1657 Operations</a></div>
1658 <div class="doc_text">
1659 <p>A key design point of an SSA-based representation is how it
1660 represents memory. In LLVM, no memory locations are in SSA form, which
1661 makes things very simple. This section describes how to read, write,
1662 allocate, and free memory in LLVM.</p>
1664 <!-- _______________________________________________________________________ -->
1665 <div class="doc_subsubsection"> <a name="i_malloc">'<tt>malloc</tt>'
1666 Instruction</a> </div>
1667 <div class="doc_text">
1669 <pre> <result> = malloc <type>, uint <NumElements> <i>; yields {type*}:result</i>
1670 <result> = malloc <type> <i>; yields {type*}:result</i>
1673 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
1674 heap and returns a pointer to it.</p>
1676 <p>The '<tt>malloc</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1677 bytes of memory from the operating system and returns a pointer of the
1678 appropriate type to the program. The second form of the instruction is
1679 a shorter version of the first instruction that defaults to allocating
1681 <p>'<tt>type</tt>' must be a sized type.</p>
1683 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
1684 a pointer is returned.</p>
1686 <pre> %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1689 href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1690 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1691 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1694 <!-- _______________________________________________________________________ -->
1695 <div class="doc_subsubsection"> <a name="i_free">'<tt>free</tt>'
1696 Instruction</a> </div>
1697 <div class="doc_text">
1699 <pre> free <type> <value> <i>; yields {void}</i>
1702 <p>The '<tt>free</tt>' instruction returns memory back to the unused
1703 memory heap, to be reallocated in the future.</p>
1706 <p>'<tt>value</tt>' shall be a pointer value that points to a value
1707 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
1710 <p>Access to the memory pointed to by the pointer is no longer defined
1711 after this instruction executes.</p>
1713 <pre> %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1714 free [4 x ubyte]* %array
1717 <!-- _______________________________________________________________________ -->
1718 <div class="doc_subsubsection"> <a name="i_alloca">'<tt>alloca</tt>'
1719 Instruction</a> </div>
1720 <div class="doc_text">
1722 <pre> <result> = alloca <type>, uint <NumElements> <i>; yields {type*}:result</i>
1723 <result> = alloca <type> <i>; yields {type*}:result</i>
1726 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
1727 stack frame of the procedure that is live until the current function
1728 returns to its caller.</p>
1730 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1731 bytes of memory on the runtime stack, returning a pointer of the
1732 appropriate type to the program. The second form of the instruction is
1733 a shorter version of the first that defaults to allocating one element.</p>
1734 <p>'<tt>type</tt>' may be any sized type.</p>
1736 <p>Memory is allocated, a pointer is returned. '<tt>alloca</tt>'d
1737 memory is automatically released when the function returns. The '<tt>alloca</tt>'
1738 instruction is commonly used to represent automatic variables that must
1739 have an address available. When the function returns (either with the <tt><a
1740 href="#i_ret">ret</a></tt> or <tt><a href="#i_invoke">invoke</a></tt>
1741 instructions), the memory is reclaimed.</p>
1743 <pre> %ptr = alloca int <i>; yields {int*}:ptr</i>
1744 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
1747 <!-- _______________________________________________________________________ -->
1748 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
1749 Instruction</a> </div>
1750 <div class="doc_text">
1752 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
1754 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
1756 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
1757 address to load from. The pointer must point to a <a
1758 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
1759 marked as <tt>volatile</tt> then the optimizer is not allowed to modify
1760 the number or order of execution of this <tt>load</tt> with other
1761 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
1764 <p>The location of memory pointed to is loaded.</p>
1766 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1768 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1769 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1772 <!-- _______________________________________________________________________ -->
1773 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
1774 Instruction</a> </div>
1776 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1777 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1780 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
1782 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
1783 to store and an address to store it into. The type of the '<tt><pointer></tt>'
1784 operand must be a pointer to the type of the '<tt><value></tt>'
1785 operand. If the <tt>store</tt> is marked as <tt>volatile</tt> then the
1786 optimizer is not allowed to modify the number or order of execution of
1787 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
1788 href="#i_store">store</a></tt> instructions.</p>
1790 <p>The contents of memory are updated to contain '<tt><value></tt>'
1791 at the location specified by the '<tt><pointer></tt>' operand.</p>
1793 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1795 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1796 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1798 <!-- _______________________________________________________________________ -->
1799 <div class="doc_subsubsection">
1800 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
1803 <div class="doc_text">
1806 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
1812 The '<tt>getelementptr</tt>' instruction is used to get the address of a
1813 subelement of an aggregate data structure.</p>
1817 <p>This instruction takes a list of integer constants that indicate what
1818 elements of the aggregate object to index to. The actual types of the arguments
1819 provided depend on the type of the first pointer argument. The
1820 '<tt>getelementptr</tt>' instruction is used to index down through the type
1821 levels of a structure. When indexing into a structure, only <tt>uint</tt>
1822 integer constants are allowed. When indexing into an array or pointer
1823 <tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
1825 <p>For example, let's consider a C code fragment and how it gets
1826 compiled to LLVM:</p>
1840 int *foo(struct ST *s) {
1841 return &s[1].Z.B[5][13];
1845 <p>The LLVM code generated by the GCC frontend is:</p>
1848 %RT = type { sbyte, [10 x [20 x int]], sbyte }
1849 %ST = type { int, double, %RT }
1853 int* %foo(%ST* %s) {
1855 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
1862 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
1863 on the pointer type that is being index into. <a href="#t_pointer">Pointer</a>
1864 and <a href="#t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
1865 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="#t_struct">structure</a>
1866 types require <tt>uint</tt> <b>constants</b>.</p>
1868 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
1869 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
1870 }</tt>' type, a structure. The second index indexes into the third element of
1871 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
1872 sbyte }</tt>' type, another structure. The third index indexes into the second
1873 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
1874 array. The two dimensions of the array are subscripted into, yielding an
1875 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction return a pointer
1876 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
1878 <p>Note that it is perfectly legal to index partially through a
1879 structure, returning a pointer to an inner element. Because of this,
1880 the LLVM code for the given testcase is equivalent to:</p>
1883 int* %foo(%ST* %s) {
1884 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
1885 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
1886 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
1887 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
1888 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
1894 <i>; yields [12 x ubyte]*:aptr</i>
1895 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
1899 <!-- ======================================================================= -->
1900 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
1901 <div class="doc_text">
1902 <p>The instructions in this category are the "miscellaneous"
1903 instructions, which defy better classification.</p>
1905 <!-- _______________________________________________________________________ -->
1906 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
1907 Instruction</a> </div>
1908 <div class="doc_text">
1910 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
1912 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
1913 the SSA graph representing the function.</p>
1915 <p>The type of the incoming values are specified with the first type
1916 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
1917 as arguments, with one pair for each predecessor basic block of the
1918 current block. Only values of <a href="#t_firstclass">first class</a>
1919 type may be used as the value arguments to the PHI node. Only labels
1920 may be used as the label arguments.</p>
1921 <p>There must be no non-phi instructions between the start of a basic
1922 block and the PHI instructions: i.e. PHI instructions must be first in
1925 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
1926 value specified by the parameter, depending on which basic block we
1927 came from in the last <a href="#terminators">terminator</a> instruction.</p>
1929 <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>
1932 <!-- _______________________________________________________________________ -->
1933 <div class="doc_subsubsection">
1934 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
1937 <div class="doc_text">
1942 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
1948 The '<tt>cast</tt>' instruction is used as the primitive means to convert
1949 integers to floating point, change data type sizes, and break type safety (by
1957 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
1958 class value, and a type to cast it to, which must also be a <a
1959 href="#t_firstclass">first class</a> type.
1965 This instruction follows the C rules for explicit casts when determining how the
1966 data being cast must change to fit in its new container.
1970 When casting to bool, any value that would be considered true in the context of
1971 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
1972 all else are '<tt>false</tt>'.
1976 When extending an integral value from a type of one signness to another (for
1977 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
1978 <b>source</b> value is signed, and zero-extended if the source value is
1979 unsigned. <tt>bool</tt> values are always zero extended into either zero or
1986 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
1987 %Y = cast int 123 to bool <i>; yields bool:true</i>
1991 <!-- _______________________________________________________________________ -->
1992 <div class="doc_subsubsection">
1993 <a name="i_select">'<tt>select</tt>' Instruction</a>
1996 <div class="doc_text">
2001 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
2007 The '<tt>select</tt>' instruction is used to choose one value based on a
2008 condition, without branching.
2015 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.
2021 If the boolean condition evaluates to true, the instruction returns the first
2022 value argument, otherwise it returns the second value argument.
2028 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
2036 <!-- _______________________________________________________________________ -->
2037 <div class="doc_subsubsection"> <a name="i_call">'<tt>call</tt>'
2038 Instruction</a> </div>
2039 <div class="doc_text">
2041 <pre> <result> = call <ty>* <fnptrval>(<param list>)<br></pre>
2043 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
2045 <p>This instruction requires several arguments:</p>
2048 <p>'<tt>ty</tt>': shall be the signature of the pointer to function
2049 value being invoked. The argument types must match the types implied
2050 by this signature.</p>
2053 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a
2054 function to be invoked. In most cases, this is a direct function
2055 invocation, but indirect <tt>call</tt>s are just as possible,
2056 calling an arbitrary pointer to function values.</p>
2059 <p>'<tt>function args</tt>': argument list whose types match the
2060 function signature argument types. All arguments must be of
2061 <a href="#t_firstclass">first class</a> type. If the function signature
2062 indicates the function accepts a variable number of arguments, the extra
2063 arguments can be specified.</p>
2067 <p>The '<tt>call</tt>' instruction is used to cause control flow to
2068 transfer to a specified function, with its incoming arguments bound to
2069 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
2070 instruction in the called function, control flow continues with the
2071 instruction after the function call, and the return value of the
2072 function is bound to the result argument. This is a simpler case of
2073 the <a href="#i_invoke">invoke</a> instruction.</p>
2075 <pre> %retval = call int %test(int %argc)<br> call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);<br></pre>
2078 <!-- _______________________________________________________________________ -->
2079 <div class="doc_subsubsection">
2080 <a name="i_vanext">'<tt>vanext</tt>' Instruction</a>
2083 <div class="doc_text">
2088 <resultarglist> = vanext <va_list> <arglist>, <argty>
2093 <p>The '<tt>vanext</tt>' instruction is used to access arguments passed
2094 through the "variable argument" area of a function call. It is used to
2095 implement the <tt>va_arg</tt> macro in C.</p>
2099 <p>This instruction takes a <tt>va_list</tt> value and the type of the
2100 argument. It returns another <tt>va_list</tt>. The actual type of
2101 <tt>va_list</tt> may be defined differently for different targets. Most targets
2102 use a <tt>va_list</tt> type of <tt>sbyte*</tt> or some other pointer type.</p>
2106 <p>The '<tt>vanext</tt>' instruction advances the specified <tt>va_list</tt>
2107 past an argument of the specified type. In conjunction with the <a
2108 href="#i_vaarg"><tt>vaarg</tt></a> instruction, it is used to implement
2109 the <tt>va_arg</tt> macro available in C. For more information, see
2110 the variable argument handling <a href="#int_varargs">Intrinsic
2113 <p>It is legal for this instruction to be called in a function which
2114 does not take a variable number of arguments, for example, the <tt>vfprintf</tt>
2117 <p><tt>vanext</tt> is an LLVM instruction instead of an <a
2118 href="#intrinsics">intrinsic function</a> because it takes a type as an
2119 argument. The type refers to the current argument in the <tt>va_list</tt>, it
2120 tells the compiler how far on the stack it needs to advance to find the next
2125 <p>See the <a href="#int_varargs">variable argument processing</a>
2130 <!-- _______________________________________________________________________ -->
2131 <div class="doc_subsubsection">
2132 <a name="i_vaarg">'<tt>vaarg</tt>' Instruction</a>
2135 <div class="doc_text">
2140 <resultval> = vaarg <va_list> <arglist>, <argty>
2145 <p>The '<tt>vaarg</tt>' instruction is used to access arguments passed through
2146 the "variable argument" area of a function call. It is used to implement the
2147 <tt>va_arg</tt> macro in C.</p>
2151 <p>This instruction takes a <tt>va_list</tt> value and the type of the
2152 argument. It returns a value of the specified argument type. Again, the actual
2153 type of <tt>va_list</tt> is target specific.</p>
2157 <p>The '<tt>vaarg</tt>' instruction loads an argument of the specified type from
2158 the specified <tt>va_list</tt>. In conjunction with the <a
2159 href="#i_vanext"><tt>vanext</tt></a> instruction, it is used to implement the
2160 <tt>va_arg</tt> macro available in C. For more information, see the variable
2161 argument handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
2163 <p>It is legal for this instruction to be called in a function which does not
2164 take a variable number of arguments, for example, the <tt>vfprintf</tt>
2167 <p><tt>vaarg</tt> is an LLVM instruction instead of an <a
2168 href="#intrinsics">intrinsic function</a> because it takes an type as an
2173 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
2177 <!-- *********************************************************************** -->
2178 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
2179 <!-- *********************************************************************** -->
2181 <div class="doc_text">
2183 <p>LLVM supports the notion of an "intrinsic function". These functions have
2184 well known names and semantics, and are required to follow certain
2185 restrictions. Overall, these instructions represent an extension mechanism for
2186 the LLVM language that does not require changing all of the transformations in
2187 LLVM to add to the language (or the bytecode reader/writer, the parser,
2190 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix, this
2191 prefix is reserved in LLVM for intrinsic names, thus functions may not be named
2192 this. Intrinsic functions must always be external functions: you cannot define
2193 the body of intrinsic functions. Intrinsic functions may only be used in call
2194 or invoke instructions: it is illegal to take the address of an intrinsic
2195 function. Additionally, because intrinsic functions are part of the LLVM
2196 language, it is required that they all be documented here if any are added.</p>
2200 Adding an intrinsic to LLVM is straight-forward if it is possible to express the
2201 concept in LLVM directly (ie, code generator support is not _required_). To do
2202 this, extend the default implementation of the IntrinsicLowering class to handle
2203 the intrinsic. Code generators use this class to lower intrinsics they do not
2204 understand to raw LLVM instructions that they do.
2209 <!-- ======================================================================= -->
2210 <div class="doc_subsection">
2211 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
2214 <div class="doc_text">
2216 <p>Variable argument support is defined in LLVM with the <a
2217 href="#i_vanext"><tt>vanext</tt></a> instruction and these three
2218 intrinsic functions. These functions are related to the similarly
2219 named macros defined in the <tt><stdarg.h></tt> header file.</p>
2221 <p>All of these functions operate on arguments that use a
2222 target-specific value type "<tt>va_list</tt>". The LLVM assembly
2223 language reference manual does not define what this type is, so all
2224 transformations should be prepared to handle intrinsics with any type
2227 <p>This example shows how the <a href="#i_vanext"><tt>vanext</tt></a>
2228 instruction and the variable argument handling intrinsic functions are
2232 int %test(int %X, ...) {
2233 ; Initialize variable argument processing
2234 %ap = call sbyte* %<a href="#i_va_start">llvm.va_start</a>()
2236 ; Read a single integer argument
2237 %tmp = vaarg sbyte* %ap, int
2239 ; Advance to the next argument
2240 %ap2 = vanext sbyte* %ap, int
2242 ; Demonstrate usage of llvm.va_copy and llvm.va_end
2243 %aq = call sbyte* %<a href="#i_va_copy">llvm.va_copy</a>(sbyte* %ap2)
2244 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %aq)
2246 ; Stop processing of arguments.
2247 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %ap2)
2253 <!-- _______________________________________________________________________ -->
2254 <div class="doc_subsubsection">
2255 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
2259 <div class="doc_text">
2261 <pre> declare <va_list> %llvm.va_start()<br></pre>
2263 <p>The '<tt>llvm.va_start</tt>' intrinsic returns a new <tt><arglist></tt>
2264 for subsequent use by the variable argument intrinsics.</p>
2266 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
2267 macro available in C. In a target-dependent way, it initializes and
2268 returns a <tt>va_list</tt> element, so that the next <tt>vaarg</tt>
2269 will produce the first variable argument passed to the function. Unlike
2270 the C <tt>va_start</tt> macro, this intrinsic does not need to know the
2271 last argument of the function, the compiler can figure that out.</p>
2272 <p>Note that this intrinsic function is only legal to be called from
2273 within the body of a variable argument function.</p>
2276 <!-- _______________________________________________________________________ -->
2277 <div class="doc_subsubsection">
2278 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
2281 <div class="doc_text">
2283 <pre> declare void %llvm.va_end(<va_list> <arglist>)<br></pre>
2285 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
2286 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
2287 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
2289 <p>The argument is a <tt>va_list</tt> to destroy.</p>
2291 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
2292 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
2293 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
2294 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
2295 with calls to <tt>llvm.va_end</tt>.</p>
2298 <!-- _______________________________________________________________________ -->
2299 <div class="doc_subsubsection">
2300 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
2303 <div class="doc_text">
2308 declare <va_list> %llvm.va_copy(<va_list> <destarglist>)
2313 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
2314 from the source argument list to the destination argument list.</p>
2318 <p>The argument is the <tt>va_list</tt> to copy.</p>
2322 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
2323 macro available in C. In a target-dependent way, it copies the source
2324 <tt>va_list</tt> element into the returned list. This intrinsic is necessary
2325 because the <tt><a href="#i_va_start">llvm.va_start</a></tt> intrinsic may be
2326 arbitrarily complex and require memory allocation, for example.</p>
2330 <!-- ======================================================================= -->
2331 <div class="doc_subsection">
2332 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
2335 <div class="doc_text">
2338 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
2339 Collection</a> requires the implementation and generation of these intrinsics.
2340 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
2341 stack</a>, as well as garbage collector implementations that require <a
2342 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
2343 Front-ends for type-safe garbage collected languages should generate these
2344 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
2345 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
2349 <!-- _______________________________________________________________________ -->
2350 <div class="doc_subsubsection">
2351 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
2354 <div class="doc_text">
2359 declare void %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
2364 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
2365 the code generator, and allows some metadata to be associated with it.</p>
2369 <p>The first argument specifies the address of a stack object that contains the
2370 root pointer. The second pointer (which must be either a constant or a global
2371 value address) contains the meta-data to be associated with the root.</p>
2375 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
2376 location. At compile-time, the code generator generates information to allow
2377 the runtime to find the pointer at GC safe points.
2383 <!-- _______________________________________________________________________ -->
2384 <div class="doc_subsubsection">
2385 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
2388 <div class="doc_text">
2393 declare sbyte* %llvm.gcread(sbyte** %Ptr)
2398 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
2399 locations, allowing garbage collector implementations that require read
2404 <p>The argument is the address to read from, which should be an address
2405 allocated from the garbage collector.</p>
2409 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
2410 instruction, but may be replaced with substantially more complex code by the
2411 garbage collector runtime, as needed.</p>
2416 <!-- _______________________________________________________________________ -->
2417 <div class="doc_subsubsection">
2418 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
2421 <div class="doc_text">
2426 declare void %llvm.gcwrite(sbyte* %P1, sbyte** %P2)
2431 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
2432 locations, allowing garbage collector implementations that require write
2433 barriers (such as generational or reference counting collectors).</p>
2437 <p>The first argument is the reference to store, and the second is the heap
2438 location to store to.</p>
2442 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
2443 instruction, but may be replaced with substantially more complex code by the
2444 garbage collector runtime, as needed.</p>
2450 <!-- ======================================================================= -->
2451 <div class="doc_subsection">
2452 <a name="int_codegen">Code Generator Intrinsics</a>
2455 <div class="doc_text">
2457 These intrinsics are provided by LLVM to expose special features that may only
2458 be implemented with code generator support.
2463 <!-- _______________________________________________________________________ -->
2464 <div class="doc_subsubsection">
2465 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
2468 <div class="doc_text">
2472 declare void* %llvm.returnaddress(uint <level>)
2478 The '<tt>llvm.returnaddress</tt>' intrinsic returns a target-specific value
2479 indicating the return address of the current function or one of its callers.
2485 The argument to this intrinsic indicates which function to return the address
2486 for. Zero indicates the calling function, one indicates its caller, etc. The
2487 argument is <b>required</b> to be a constant integer value.
2493 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
2494 the return address of the specified call frame, or zero if it cannot be
2495 identified. The value returned by this intrinsic is likely to be incorrect or 0
2496 for arguments other than zero, so it should only be used for debugging purposes.
2500 Note that calling this intrinsic does not prevent function inlining or other
2501 aggressive transformations, so the value returned may not be that of the obvious
2502 source-language caller.
2507 <!-- _______________________________________________________________________ -->
2508 <div class="doc_subsubsection">
2509 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
2512 <div class="doc_text">
2516 declare void* %llvm.frameaddress(uint <level>)
2522 The '<tt>llvm.frameaddress</tt>' intrinsic returns the target-specific frame
2523 pointer value for the specified stack frame.
2529 The argument to this intrinsic indicates which function to return the frame
2530 pointer for. Zero indicates the calling function, one indicates its caller,
2531 etc. The argument is <b>required</b> to be a constant integer value.
2537 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
2538 the frame address of the specified call frame, or zero if it cannot be
2539 identified. The value returned by this intrinsic is likely to be incorrect or 0
2540 for arguments other than zero, so it should only be used for debugging purposes.
2544 Note that calling this intrinsic does not prevent function inlining or other
2545 aggressive transformations, so the value returned may not be that of the obvious
2546 source-language caller.
2550 <!-- _______________________________________________________________________ -->
2551 <div class="doc_subsubsection">
2552 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
2555 <div class="doc_text">
2559 declare void %llvm.prefetch(sbyte * <address>,
2560 uint <rw>, uint <locality>)
2567 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
2568 a prefetch instruction if supported, otherwise it is a noop. Prefetches have no
2569 effect on the behavior of the program, but can change its performance
2576 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
2577 determining if the fetch should be for a read (0) or write (1), and
2578 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
2579 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
2580 <tt>locality</tt> arguments must be constant integers.
2586 This intrinsic does not modify the behavior of the program. In particular,
2587 prefetches cannot trap and do not produce a value. On targets that support this
2588 intrinsic, the prefetch can provide hints to the processor cache for better
2594 <!-- _______________________________________________________________________ -->
2595 <div class="doc_subsubsection">
2596 <a name="i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
2599 <div class="doc_text">
2603 declare void %llvm.pcmarker( uint <id> )
2610 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a PC in a region of
2611 code to simulators and other tools. The method is target specific, but it is
2612 expected that the marker will use exported symbols to transmit the PC of the marker.
2613 The marker makes no guaranties that it will remain with any specific instruction
2614 after optimizations. It is possible that the presense of a marker will inhibit
2615 optimizations. The intended use is to be inserted after optmizations to allow
2616 corrolations of simulation runs.
2622 <tt>id</tt> is a numerical id identifying the marker.
2628 This intrinsic does not modify the behavior of the program. Backends that do not
2629 support this intrinisic may ignore it.
2635 <!-- ======================================================================= -->
2636 <div class="doc_subsection">
2637 <a name="int_os">Operating System Intrinsics</a>
2640 <div class="doc_text">
2642 These intrinsics are provided by LLVM to support the implementation of
2643 operating system level code.
2648 <!-- _______________________________________________________________________ -->
2649 <div class="doc_subsubsection">
2650 <a name="i_readport">'<tt>llvm.readport</tt>' Intrinsic</a>
2653 <div class="doc_text">
2657 declare <integer type> %llvm.readport (<integer type> <address>)
2663 The '<tt>llvm.readport</tt>' intrinsic reads data from the specified hardware
2670 The argument to this intrinsic indicates the hardware I/O address from which
2671 to read the data. The address is in the hardware I/O address namespace (as
2672 opposed to being a memory location for memory mapped I/O).
2678 The '<tt>llvm.readport</tt>' intrinsic reads data from the hardware I/O port
2679 specified by <i>address</i> and returns the value. The address and return
2680 value must be integers, but the size is dependent upon the platform upon which
2681 the program is code generated. For example, on x86, the address must be an
2682 unsigned 16-bit value, and the return value must be 8, 16, or 32 bits.
2687 <!-- _______________________________________________________________________ -->
2688 <div class="doc_subsubsection">
2689 <a name="i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a>
2692 <div class="doc_text">
2696 call void (<integer type>, <integer type>)*
2697 %llvm.writeport (<integer type> <value>,
2698 <integer type> <address>)
2704 The '<tt>llvm.writeport</tt>' intrinsic writes data to the specified hardware
2711 The first argument is the value to write to the I/O port.
2715 The second argument indicates the hardware I/O address to which data should be
2716 written. The address is in the hardware I/O address namespace (as opposed to
2717 being a memory location for memory mapped I/O).
2723 The '<tt>llvm.writeport</tt>' intrinsic writes <i>value</i> to the I/O port
2724 specified by <i>address</i>. The address and value must be integers, but the
2725 size is dependent upon the platform upon which the program is code generated.
2726 For example, on x86, the address must be an unsigned 16-bit value, and the
2727 value written must be 8, 16, or 32 bits in length.
2732 <!-- _______________________________________________________________________ -->
2733 <div class="doc_subsubsection">
2734 <a name="i_readio">'<tt>llvm.readio</tt>' Intrinsic</a>
2737 <div class="doc_text">
2741 declare <result> %llvm.readio (<ty> * <pointer>)
2747 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2754 The argument to this intrinsic is a pointer indicating the memory address from
2755 which to read the data. The data must be a
2756 <a href="#t_firstclass">first class</a> type.
2762 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2763 location specified by <i>pointer</i> and returns the value. The argument must
2764 be a pointer, and the return value must be a
2765 <a href="#t_firstclass">first class</a> type. However, certain architectures
2766 may not support I/O on all first class types. For example, 32-bit processors
2767 may only support I/O on data types that are 32 bits or less.
2771 This intrinsic enforces an in-order memory model for llvm.readio and
2772 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2773 scheduled processors may execute loads and stores out of order, re-ordering at
2774 run time accesses to memory mapped I/O registers. Using these intrinsics
2775 ensures that accesses to memory mapped I/O registers occur in program order.
2780 <!-- _______________________________________________________________________ -->
2781 <div class="doc_subsubsection">
2782 <a name="i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a>
2785 <div class="doc_text">
2789 declare void %llvm.writeio (<ty1> <value>, <ty2> * <pointer>)
2795 The '<tt>llvm.writeio</tt>' intrinsic writes data to the specified memory
2802 The first argument is the value to write to the memory mapped I/O location.
2803 The second argument is a pointer indicating the memory address to which the
2804 data should be written.
2810 The '<tt>llvm.writeio</tt>' intrinsic writes <i>value</i> to the memory mapped
2811 I/O address specified by <i>pointer</i>. The value must be a
2812 <a href="#t_firstclass">first class</a> type. However, certain architectures
2813 may not support I/O on all first class types. For example, 32-bit processors
2814 may only support I/O on data types that are 32 bits or less.
2818 This intrinsic enforces an in-order memory model for llvm.readio and
2819 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2820 scheduled processors may execute loads and stores out of order, re-ordering at
2821 run time accesses to memory mapped I/O registers. Using these intrinsics
2822 ensures that accesses to memory mapped I/O registers occur in program order.
2827 <!-- ======================================================================= -->
2828 <div class="doc_subsection">
2829 <a name="int_libc">Standard C Library Intrinsics</a>
2832 <div class="doc_text">
2834 LLVM provides intrinsics for a few important standard C library functions.
2835 These intrinsics allow source-language front-ends to pass information about the
2836 alignment of the pointer arguments to the code generator, providing opportunity
2837 for more efficient code generation.
2842 <!-- _______________________________________________________________________ -->
2843 <div class="doc_subsubsection">
2844 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
2847 <div class="doc_text">
2851 declare void %llvm.memcpy(sbyte* <dest>, sbyte* <src>,
2852 uint <len>, uint <align>)
2858 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2859 location to the destination location.
2863 Note that, unlike the standard libc function, the <tt>llvm.memcpy</tt> intrinsic
2864 does not return a value, and takes an extra alignment argument.
2870 The first argument is a pointer to the destination, the second is a pointer to
2871 the source. The third argument is an (arbitrarily sized) integer argument
2872 specifying the number of bytes to copy, and the fourth argument is the alignment
2873 of the source and destination locations.
2877 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2878 the caller guarantees that the size of the copy is a multiple of the alignment
2879 and that both the source and destination pointers are aligned to that boundary.
2885 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2886 location to the destination location, which are not allowed to overlap. It
2887 copies "len" bytes of memory over. If the argument is known to be aligned to
2888 some boundary, this can be specified as the fourth argument, otherwise it should
2894 <!-- _______________________________________________________________________ -->
2895 <div class="doc_subsubsection">
2896 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
2899 <div class="doc_text">
2903 declare void %llvm.memmove(sbyte* <dest>, sbyte* <src>,
2904 uint <len>, uint <align>)
2910 The '<tt>llvm.memmove</tt>' intrinsic moves a block of memory from the source
2911 location to the destination location. It is similar to the '<tt>llvm.memcpy</tt>'
2912 intrinsic but allows the two memory locations to overlap.
2916 Note that, unlike the standard libc function, the <tt>llvm.memmove</tt> intrinsic
2917 does not return a value, and takes an extra alignment argument.
2923 The first argument is a pointer to the destination, the second is a pointer to
2924 the source. The third argument is an (arbitrarily sized) integer argument
2925 specifying the number of bytes to copy, and the fourth argument is the alignment
2926 of the source and destination locations.
2930 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2931 the caller guarantees that the size of the copy is a multiple of the alignment
2932 and that both the source and destination pointers are aligned to that boundary.
2938 The '<tt>llvm.memmove</tt>' intrinsic copies a block of memory from the source
2939 location to the destination location, which may overlap. It
2940 copies "len" bytes of memory over. If the argument is known to be aligned to
2941 some boundary, this can be specified as the fourth argument, otherwise it should
2947 <!-- _______________________________________________________________________ -->
2948 <div class="doc_subsubsection">
2949 <a name="i_memset">'<tt>llvm.memset</tt>' Intrinsic</a>
2952 <div class="doc_text">
2956 declare void %llvm.memset(sbyte* <dest>, ubyte <val>,
2957 uint <len>, uint <align>)
2963 The '<tt>llvm.memset</tt>' intrinsic fills a block of memory with a particular
2968 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
2969 does not return a value, and takes an extra alignment argument.
2975 The first argument is a pointer to the destination to fill, the second is the
2976 byte value to fill it with, the third argument is an (arbitrarily sized) integer
2977 argument specifying the number of bytes to fill, and the fourth argument is the
2978 known alignment of destination location.
2982 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2983 the caller guarantees that the size of the copy is a multiple of the alignment
2984 and that the destination pointer is aligned to that boundary.
2990 The '<tt>llvm.memset</tt>' intrinsic fills "len" bytes of memory starting at the
2991 destination location. If the argument is known to be aligned to some boundary,
2992 this can be specified as the fourth argument, otherwise it should be set to 0 or
2998 <!-- _______________________________________________________________________ -->
2999 <div class="doc_subsubsection">
3000 <a name="i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a>
3003 <div class="doc_text">
3007 declare bool %llvm.isunordered(<float or double> Val1, <float or double> Val2)
3013 The '<tt>llvm.isunordered</tt>' intrinsic returns true if either or both of the
3014 specified floating point values is a NAN.
3020 The arguments are floating point numbers of the same type.
3026 If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
3034 <!-- ======================================================================= -->
3035 <div class="doc_subsection">
3036 <a name="int_debugger">Debugger Intrinsics</a>
3039 <div class="doc_text">
3041 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
3042 are described in the <a
3043 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
3044 Debugging</a> document.
3049 <!-- *********************************************************************** -->
3052 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
3053 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
3054 <a href="http://validator.w3.org/check/referer"><img
3055 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
3057 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
3058 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
3059 Last modified: $Date$
projects / oota-llvm.git / blob