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>
46 <li><a href="#constants">Constants</a>
48 <li><a href="#simpleconstants">Simple Constants</a>
49 <li><a href="#aggregateconstants">Aggregate Constants</a>
50 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
51 <li><a href="#undefvalues">Undefined Values</a>
52 <li><a href="#constantexprs">Constant Expressions</a>
55 <li><a href="#instref">Instruction Reference</a>
57 <li><a href="#terminators">Terminator Instructions</a>
59 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
60 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
61 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
62 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
63 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
64 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
67 <li><a href="#binaryops">Binary Operations</a>
69 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
70 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
71 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
72 <li><a href="#i_div">'<tt>div</tt>' Instruction</a></li>
73 <li><a href="#i_rem">'<tt>rem</tt>' Instruction</a></li>
74 <li><a href="#i_setcc">'<tt>set<i>cc</i></tt>' Instructions</a></li>
77 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
79 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
80 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
81 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
82 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
83 <li><a href="#i_shr">'<tt>shr</tt>' Instruction</a></li>
86 <li><a href="#memoryops">Memory Access Operations</a>
88 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
89 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
90 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
91 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
92 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
93 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
96 <li><a href="#otherops">Other Operations</a>
98 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
99 <li><a href="#i_cast">'<tt>cast .. to</tt>' Instruction</a></li>
100 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
101 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
102 <li><a href="#i_vanext">'<tt>vanext</tt>' Instruction</a></li>
103 <li><a href="#i_vaarg">'<tt>vaarg</tt>' Instruction</a></li>
108 <li><a href="#intrinsics">Intrinsic Functions</a>
110 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
112 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
113 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
114 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
117 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
119 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
120 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
121 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
124 <li><a href="#int_codegen">Code Generator Intrinsics</a>
126 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
127 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
128 <li><a href="#i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
129 <li><a href="#i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
132 <li><a href="#int_os">Operating System Intrinsics</a>
134 <li><a href="#i_readport">'<tt>llvm.readport</tt>' Intrinsic</a></li>
135 <li><a href="#i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a></li>
136 <li><a href="#i_readio">'<tt>llvm.readio</tt>' Intrinsic</a></li>
137 <li><a href="#i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a></li>
139 <li><a href="#int_libc">Standard C Library Intrinsics</a>
141 <li><a href="#i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a></li>
142 <li><a href="#i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a></li>
143 <li><a href="#i_memset">'<tt>llvm.memset</tt>' Intrinsic</a></li>
144 <li><a href="#i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a></li>
147 <li><a href="#int_debugger">Debugger intrinsics</a></li>
152 <div class="doc_author">
153 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
154 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
157 <!-- *********************************************************************** -->
158 <div class="doc_section"> <a name="abstract">Abstract </a></div>
159 <!-- *********************************************************************** -->
161 <div class="doc_text">
162 <p>This document is a reference manual for the LLVM assembly language.
163 LLVM is an SSA based representation that provides type safety,
164 low-level operations, flexibility, and the capability of representing
165 'all' high-level languages cleanly. It is the common code
166 representation used throughout all phases of the LLVM compilation
170 <!-- *********************************************************************** -->
171 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
172 <!-- *********************************************************************** -->
174 <div class="doc_text">
176 <p>The LLVM code representation is designed to be used in three
177 different forms: as an in-memory compiler IR, as an on-disk bytecode
178 representation (suitable for fast loading by a Just-In-Time compiler),
179 and as a human readable assembly language representation. This allows
180 LLVM to provide a powerful intermediate representation for efficient
181 compiler transformations and analysis, while providing a natural means
182 to debug and visualize the transformations. The three different forms
183 of LLVM are all equivalent. This document describes the human readable
184 representation and notation.</p>
186 <p>The LLVM representation aims to be a light-weight and low-level
187 while being expressive, typed, and extensible at the same time. It
188 aims to be a "universal IR" of sorts, by being at a low enough level
189 that high-level ideas may be cleanly mapped to it (similar to how
190 microprocessors are "universal IR's", allowing many source languages to
191 be mapped to them). By providing type information, LLVM can be used as
192 the target of optimizations: for example, through pointer analysis, it
193 can be proven that a C automatic variable is never accessed outside of
194 the current function... allowing it to be promoted to a simple SSA
195 value instead of a memory location.</p>
199 <!-- _______________________________________________________________________ -->
200 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
202 <div class="doc_text">
204 <p>It is important to note that this document describes 'well formed'
205 LLVM assembly language. There is a difference between what the parser
206 accepts and what is considered 'well formed'. For example, the
207 following instruction is syntactically okay, but not well formed:</p>
210 %x = <a href="#i_add">add</a> int 1, %x
213 <p>...because the definition of <tt>%x</tt> does not dominate all of
214 its uses. The LLVM infrastructure provides a verification pass that may
215 be used to verify that an LLVM module is well formed. This pass is
216 automatically run by the parser after parsing input assembly, and by
217 the optimizer before it outputs bytecode. The violations pointed out
218 by the verifier pass indicate bugs in transformation passes or input to
221 <!-- Describe the typesetting conventions here. --> </div>
223 <!-- *********************************************************************** -->
224 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
225 <!-- *********************************************************************** -->
227 <div class="doc_text">
229 <p>LLVM uses three different forms of identifiers, for different
233 <li>Named values are represented as a string of characters with a '%' prefix.
234 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
235 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
236 Identifiers which require other characters in their names can be surrounded
237 with quotes. In this way, anything except a <tt>"</tt> character can be used
240 <li>Unnamed values are represented as an unsigned numeric value with a '%'
241 prefix. For example, %12, %2, %44.</li>
243 <li>Constants, which are described in a <a href="#constants">section about
244 constants</a>, below.</li>
247 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
248 don't need to worry about name clashes with reserved words, and the set of
249 reserved words may be expanded in the future without penalty. Additionally,
250 unnamed identifiers allow a compiler to quickly come up with a temporary
251 variable without having to avoid symbol table conflicts.</p>
253 <p>Reserved words in LLVM are very similar to reserved words in other
254 languages. There are keywords for different opcodes ('<tt><a
255 href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
256 href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
257 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>', etc...),
258 and others. These reserved words cannot conflict with variable names, because
259 none of them start with a '%' character.</p>
261 <p>Here is an example of LLVM code to multiply the integer variable
262 '<tt>%X</tt>' by 8:</p>
267 %result = <a href="#i_mul">mul</a> uint %X, 8
270 <p>After strength reduction:</p>
273 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
276 <p>And the hard way:</p>
279 <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
280 <a href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
281 %result = <a href="#i_add">add</a> uint %1, %1
284 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
285 important lexical features of LLVM:</p>
289 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
292 <li>Unnamed temporaries are created when the result of a computation is not
293 assigned to a named value.</li>
295 <li>Unnamed temporaries are numbered sequentially</li>
299 <p>...and it also show a convention that we follow in this document. When
300 demonstrating instructions, we will follow an instruction with a comment that
301 defines the type and name of value produced. Comments are shown in italic
306 <!-- *********************************************************************** -->
307 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
308 <!-- *********************************************************************** -->
310 <!-- ======================================================================= -->
311 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
314 <div class="doc_text">
316 <p>LLVM programs are composed of "Module"s, each of which is a
317 translation unit of the input programs. Each module consists of
318 functions, global variables, and symbol table entries. Modules may be
319 combined together with the LLVM linker, which merges function (and
320 global variable) definitions, resolves forward declarations, and merges
321 symbol table entries. Here is an example of the "hello world" module:</p>
323 <pre><i>; Declare the string constant as a global constant...</i>
324 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
325 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
327 <i>; External declaration of the puts function</i>
328 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
330 <i>; Definition of main function</i>
331 int %main() { <i>; int()* </i>
332 <i>; Convert [13x sbyte]* to sbyte *...</i>
334 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
336 <i>; Call puts function to write out the string to stdout...</i>
338 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
340 href="#i_ret">ret</a> int 0<br>}<br></pre>
342 <p>This example is made up of a <a href="#globalvars">global variable</a>
343 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
344 function, and a <a href="#functionstructure">function definition</a>
345 for "<tt>main</tt>".</p>
347 <p>In general, a module is made up of a list of global values,
348 where both functions and global variables are global values. Global values are
349 represented by a pointer to a memory location (in this case, a pointer to an
350 array of char, and a pointer to a function), and have one of the following <a
351 href="#linkage">linkage types</a>.</p>
355 <!-- ======================================================================= -->
356 <div class="doc_subsection">
357 <a name="linkage">Linkage Types</a>
360 <div class="doc_text">
363 All Global Variables and Functions have one of the following types of linkage:
368 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
370 <dd>Global values with internal linkage are only directly accessible by
371 objects in the current module. In particular, linking code into a module with
372 an internal global value may cause the internal to be renamed as necessary to
373 avoid collisions. Because the symbol is internal to the module, all
374 references can be updated. This corresponds to the notion of the
375 '<tt>static</tt>' keyword in C, or the idea of "anonymous namespaces" in C++.
378 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
380 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
381 the twist that linking together two modules defining the same
382 <tt>linkonce</tt> globals will cause one of the globals to be discarded. This
383 is typically used to implement inline functions. Unreferenced
384 <tt>linkonce</tt> globals are allowed to be discarded.
387 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
389 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
390 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
391 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.
394 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
396 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
397 pointer to array type. When two global variables with appending linkage are
398 linked together, the two global arrays are appended together. This is the
399 LLVM, typesafe, equivalent of having the system linker append together
400 "sections" with identical names when .o files are linked.
403 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
405 <dd>If none of the above identifiers are used, the global is externally
406 visible, meaning that it participates in linkage and can be used to resolve
407 external symbol references.
411 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
412 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
413 variable and was linked with this one, one of the two would be renamed,
414 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
415 external (i.e., lacking any linkage declarations), they are accessible
416 outside of the current module. It is illegal for a function <i>declaration</i>
417 to have any linkage type other than "externally visible".</a></p>
421 <!-- ======================================================================= -->
422 <div class="doc_subsection">
423 <a name="globalvars">Global Variables</a>
426 <div class="doc_text">
428 <p>Global variables define regions of memory allocated at compilation time
429 instead of run-time. Global variables may optionally be initialized. A
430 variable may be defined as a global "constant", which indicates that the
431 contents of the variable will <b>never</b> be modified (enabling better
432 optimization, allowing the global data to be placed in the read-only section of
433 an executable, etc). Note that variables that need runtime initialization
434 cannot be marked "constant", as there is a store to the variable.</p>
437 LLVM explicitly allows <em>declarations</em> of global variables to be marked
438 constant, even if the final definition of the global is not. This capability
439 can be used to enable slightly better optimization of the program, but requires
440 the language definition to guarantee that optimizations based on the
441 'constantness' are valid for the translation units that do not include the
445 <p>As SSA values, global variables define pointer values that are in
446 scope (i.e. they dominate) all basic blocks in the program. Global
447 variables always define a pointer to their "content" type because they
448 describe a region of memory, and all memory objects in LLVM are
449 accessed through pointers.</p>
454 <!-- ======================================================================= -->
455 <div class="doc_subsection">
456 <a name="functionstructure">Functions</a>
459 <div class="doc_text">
461 <p>LLVM function definitions are composed of a (possibly empty) argument list,
462 an opening curly brace, a list of basic blocks, and a closing curly brace. LLVM
463 function declarations are defined with the "<tt>declare</tt>" keyword, a
464 function name, and a function signature.</p>
466 <p>A function definition contains a list of basic blocks, forming the CFG for
467 the function. Each basic block may optionally start with a label (giving the
468 basic block a symbol table entry), contains a list of instructions, and ends
469 with a <a href="#terminators">terminator</a> instruction (such as a branch or
470 function return).</p>
472 <p>The first basic block in program is special in two ways: it is immediately
473 executed on entrance to the function, and it is not allowed to have predecessor
474 basic blocks (i.e. there can not be any branches to the entry block of a
475 function). Because the block can have no predecessors, it also cannot have any
476 <a href="#i_phi">PHI nodes</a>.</p>
478 <p>LLVM functions are identified by their name and type signature. Hence, two
479 functions with the same name but different parameter lists or return values are
480 considered different functions, and LLVM will resolve references to each
487 <!-- *********************************************************************** -->
488 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
489 <!-- *********************************************************************** -->
491 <div class="doc_text">
493 <p>The LLVM type system is one of the most important features of the
494 intermediate representation. Being typed enables a number of
495 optimizations to be performed on the IR directly, without having to do
496 extra analyses on the side before the transformation. A strong type
497 system makes it easier to read the generated code and enables novel
498 analyses and transformations that are not feasible to perform on normal
499 three address code representations.</p>
503 <!-- ======================================================================= -->
504 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
505 <div class="doc_text">
506 <p>The primitive types are the fundamental building blocks of the LLVM
507 system. The current set of primitive types is as follows:</p>
509 <table class="layout">
514 <tr><th>Type</th><th>Description</th></tr>
515 <tr><td><tt>void</tt></td><td>No value</td></tr>
516 <tr><td><tt>ubyte</tt></td><td>Unsigned 8 bit value</td></tr>
517 <tr><td><tt>ushort</tt></td><td>Unsigned 16 bit value</td></tr>
518 <tr><td><tt>uint</tt></td><td>Unsigned 32 bit value</td></tr>
519 <tr><td><tt>ulong</tt></td><td>Unsigned 64 bit value</td></tr>
520 <tr><td><tt>float</tt></td><td>32 bit floating point value</td></tr>
521 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
528 <tr><th>Type</th><th>Description</th></tr>
529 <tr><td><tt>bool</tt></td><td>True or False value</td></tr>
530 <tr><td><tt>sbyte</tt></td><td>Signed 8 bit value</td></tr>
531 <tr><td><tt>short</tt></td><td>Signed 16 bit value</td></tr>
532 <tr><td><tt>int</tt></td><td>Signed 32 bit value</td></tr>
533 <tr><td><tt>long</tt></td><td>Signed 64 bit value</td></tr>
534 <tr><td><tt>double</tt></td><td>64 bit floating point value</td></tr>
542 <!-- _______________________________________________________________________ -->
543 <div class="doc_subsubsection"> <a name="t_classifications">Type
544 Classifications</a> </div>
545 <div class="doc_text">
546 <p>These different primitive types fall into a few useful
549 <table border="1" cellspacing="0" cellpadding="4">
551 <tr><th>Classification</th><th>Types</th></tr>
553 <td><a name="t_signed">signed</a></td>
554 <td><tt>sbyte, short, int, long, float, double</tt></td>
557 <td><a name="t_unsigned">unsigned</a></td>
558 <td><tt>ubyte, ushort, uint, ulong</tt></td>
561 <td><a name="t_integer">integer</a></td>
562 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
565 <td><a name="t_integral">integral</a></td>
566 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt>
570 <td><a name="t_floating">floating point</a></td>
571 <td><tt>float, double</tt></td>
574 <td><a name="t_firstclass">first class</a></td>
575 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long,<br>
576 float, double, <a href="#t_pointer">pointer</a>,
577 <a href="#t_packed">packed</a></tt></td>
582 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
583 most important. Values of these types are the only ones which can be
584 produced by instructions, passed as arguments, or used as operands to
585 instructions. This means that all structures and arrays must be
586 manipulated either by pointer or by component.</p>
589 <!-- ======================================================================= -->
590 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
592 <div class="doc_text">
594 <p>The real power in LLVM comes from the derived types in the system.
595 This is what allows a programmer to represent arrays, functions,
596 pointers, and other useful types. Note that these derived types may be
597 recursive: For example, it is possible to have a two dimensional array.</p>
601 <!-- _______________________________________________________________________ -->
602 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
604 <div class="doc_text">
608 <p>The array type is a very simple derived type that arranges elements
609 sequentially in memory. The array type requires a size (number of
610 elements) and an underlying data type.</p>
615 [<# elements> x <elementtype>]
618 <p>The number of elements is a constant integer value, elementtype may
619 be any type with a size.</p>
622 <table class="layout">
625 <tt>[40 x int ]</tt><br/>
626 <tt>[41 x int ]</tt><br/>
627 <tt>[40 x uint]</tt><br/>
630 Array of 40 integer values.<br/>
631 Array of 41 integer values.<br/>
632 Array of 40 unsigned integer values.<br/>
636 <p>Here are some examples of multidimensional arrays:</p>
637 <table class="layout">
640 <tt>[3 x [4 x int]]</tt><br/>
641 <tt>[12 x [10 x float]]</tt><br/>
642 <tt>[2 x [3 x [4 x uint]]]</tt><br/>
645 3x4 array integer values.<br/>
646 12x10 array of single precision floating point values.<br/>
647 2x3x4 array of unsigned integer values.<br/>
653 <!-- _______________________________________________________________________ -->
654 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
655 <div class="doc_text">
657 <p>The function type can be thought of as a function signature. It
658 consists of a return type and a list of formal parameter types.
659 Function types are usually used to build virtual function tables
660 (which are structures of pointers to functions), for indirect function
661 calls, and when defining a function.</p>
663 The return type of a function type cannot be an aggregate type.
666 <pre> <returntype> (<parameter list>)<br></pre>
667 <p>Where '<tt><parameter list></tt>' is a comma-separated list of type
668 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
669 which indicates that the function takes a variable number of arguments.
670 Variable argument functions can access their arguments with the <a
671 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
673 <table class="layout">
676 <tt>int (int)</tt> <br/>
677 <tt>float (int, int *) *</tt><br/>
678 <tt>int (sbyte *, ...)</tt><br/>
681 function taking an <tt>int</tt>, returning an <tt>int</tt><br/>
682 <a href="#t_pointer">Pointer</a> to a function that takes an
683 <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
684 returning <tt>float</tt>.<br/>
685 A vararg function that takes at least one <a href="#t_pointer">pointer</a>
686 to <tt>sbyte</tt> (signed char in C), which returns an integer. This is
687 the signature for <tt>printf</tt> in LLVM.<br/>
693 <!-- _______________________________________________________________________ -->
694 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
695 <div class="doc_text">
697 <p>The structure type is used to represent a collection of data members
698 together in memory. The packing of the field types is defined to match
699 the ABI of the underlying processor. The elements of a structure may
700 be any type that has a size.</p>
701 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
702 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
703 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
706 <pre> { <type list> }<br></pre>
708 <table class="layout">
711 <tt>{ int, int, int }</tt><br/>
712 <tt>{ float, int (int) * }</tt><br/>
715 a triple of three <tt>int</tt> values<br/>
716 A pair, where the first element is a <tt>float</tt> and the second element
717 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
718 that takes an <tt>int</tt>, returning an <tt>int</tt>.<br/>
724 <!-- _______________________________________________________________________ -->
725 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
726 <div class="doc_text">
728 <p>As in many languages, the pointer type represents a pointer or
729 reference to another object, which must live in memory.</p>
731 <pre> <type> *<br></pre>
733 <table class="layout">
736 <tt>[4x int]*</tt><br/>
737 <tt>int (int *) *</tt><br/>
740 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
741 four <tt>int</tt> values<br/>
742 A <a href="#t_pointer">pointer</a> to a <a
743 href="#t_function">function</a> that takes an <tt>int*</tt>, returning an
750 <!-- _______________________________________________________________________ -->
751 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
752 <div class="doc_text">
754 <p>A packed type is a simple derived type that represents a vector
755 of elements. Packed types are used when multiple primitive data
756 are operated in parallel using a single instruction (SIMD).
757 A packed type requires a size (number of
758 elements) and an underlying primitive data type. Packed types are
759 considered <a href="#t_firstclass">first class</a>.</p>
761 <pre> < <# elements> x <elementtype> ><br></pre>
762 <p>The number of elements is a constant integer value, elementtype may
763 be any integral or floating point type.</p>
765 <table class="layout">
768 <tt><4 x int></tt><br/>
769 <tt><8 x float></tt><br/>
770 <tt><2 x uint></tt><br/>
773 Packed vector of 4 integer values.<br/>
774 Packed vector of 8 floating-point values.<br/>
775 Packed vector of 2 unsigned integer values.<br/>
781 <!-- *********************************************************************** -->
782 <div class="doc_section"> <a name="constants">Constants</a> </div>
783 <!-- *********************************************************************** -->
785 <div class="doc_text">
787 <p>LLVM has several different basic types of constants. This section describes
788 them all and their syntax.</p>
792 <!-- ======================================================================= -->
793 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
795 <div class="doc_text">
798 <dt><b>Boolean constants</b></dt>
800 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
801 constants of the <tt><a href="#t_primitive">bool</a></tt> type.
804 <dt><b>Integer constants</b></dt>
806 <dd>Standard integers (such as '4') are constants of the <a
807 href="#t_integer">integer</a> type. Negative numbers may be used with signed
811 <dt><b>Floating point constants</b></dt>
813 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
814 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
815 notation. Floating point constants have an optional hexadecimal
816 notation (see below). Floating point constants must have a <a
817 href="#t_floating">floating point</a> type. </dd>
819 <dt><b>Null pointer constants</b></dt>
821 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
822 and must be of <a href="#t_pointer">pointer type</a>.</dd>
826 <p>The one non-intuitive notation for constants is the optional hexadecimal form
827 of floating point constants. For example, the form '<tt>double
828 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
829 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
830 (and the only time that they are generated by the disassembler) is when a
831 floating point constant must be emitted but it cannot be represented as a
832 decimal floating point number. For example, NaN's, infinities, and other
833 special values are represented in their IEEE hexadecimal format so that
834 assembly and disassembly do not cause any bits to change in the constants.</p>
838 <!-- ======================================================================= -->
839 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
842 <div class="doc_text">
843 <p>Aggregate constants arise from aggregation of simple constants
844 and smaller aggregate constants.</p>
847 <dt><b>Structure constants</b></dt>
849 <dd>Structure constants are represented with notation similar to structure
850 type definitions (a comma separated list of elements, surrounded by braces
851 (<tt>{}</tt>)). For example: "<tt>{ int 4, float 17.0, int* %G }</tt>",
852 where "<tt>%G</tt>" is declared as "<tt>%G = external global int</tt>". Structure constants
853 must have <a href="#t_struct">structure type</a>, and the number and
854 types of elements must match those specified by the type.
857 <dt><b>Array constants</b></dt>
859 <dd>Array constants are represented with notation similar to array type
860 definitions (a comma separated list of elements, surrounded by square brackets
861 (<tt>[]</tt>)). For example: "<tt>[ int 42, int 11, int 74 ]</tt>". Array
862 constants must have <a href="#t_array">array type</a>, and the number and
863 types of elements must match those specified by the type.
866 <dt><b>Packed constants</b></dt>
868 <dd>Packed constants are represented with notation similar to packed type
869 definitions (a comma separated list of elements, surrounded by
870 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< int 42,
871 int 11, int 74, int 100 ></tt>". Packed constants must have <a
872 href="#t_packed">packed type</a>, and the number and types of elements must
873 match those specified by the type.
876 <dt><b>Zero initialization</b></dt>
878 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
879 value to zero of <em>any</em> type, including scalar and aggregate types.
880 This is often used to avoid having to print large zero initializers (e.g. for
881 large arrays), and is always exactly equivalent to using explicit zero
888 <!-- ======================================================================= -->
889 <div class="doc_subsection">
890 <a name="globalconstants">Global Variable and Function Addresses</a>
893 <div class="doc_text">
895 <p>The addresses of <a href="#globalvars">global variables</a> and <a
896 href="#functionstructure">functions</a> are always implicitly valid (link-time)
897 constants. These constants are explicitly referenced when the <a
898 href="#identifiers">identifier for the global</a> is used and always have <a
899 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
905 %Z = global [2 x int*] [ int* %X, int* %Y ]
910 <!-- ======================================================================= -->
911 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
912 <div class="doc_text">
913 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
914 no specific value. Undefined values may be of any type, and be used anywhere
915 a constant is permitted.</p>
917 <p>Undefined values indicate to the compiler that the program is well defined
918 no matter what value is used, giving the compiler more freedom to optimize.
922 <!-- ======================================================================= -->
923 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
926 <div class="doc_text">
928 <p>Constant expressions are used to allow expressions involving other constants
929 to be used as constants. Constant expressions may be of any <a
930 href="#t_firstclass">first class</a> type, and may involve any LLVM operation
931 that does not have side effects (e.g. load and call are not supported). The
932 following is the syntax for constant expressions:</p>
935 <dt><b><tt>cast ( CST to TYPE )</tt></b></dt>
937 <dd>Cast a constant to another type.</dd>
939 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
941 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
942 constants. As with the <a href="#i_getelementptr">getelementptr</a>
943 instruction, the index list may have zero or more indexes, which are required
944 to make sense for the type of "CSTPTR".</dd>
946 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
948 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
949 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
950 binary</a> operations. The constraints on operands are the same as those for
951 the corresponding instruction (e.g. no bitwise operations on floating point
956 <!-- *********************************************************************** -->
957 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
958 <!-- *********************************************************************** -->
960 <div class="doc_text">
962 <p>The LLVM instruction set consists of several different
963 classifications of instructions: <a href="#terminators">terminator
964 instructions</a>, <a href="#binaryops">binary instructions</a>, <a
965 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
966 instructions</a>.</p>
970 <!-- ======================================================================= -->
971 <div class="doc_subsection"> <a name="terminators">Terminator
972 Instructions</a> </div>
974 <div class="doc_text">
976 <p>As mentioned <a href="#functionstructure">previously</a>, every
977 basic block in a program ends with a "Terminator" instruction, which
978 indicates which block should be executed after the current block is
979 finished. These terminator instructions typically yield a '<tt>void</tt>'
980 value: they produce control flow, not values (the one exception being
981 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
982 <p>There are six different terminator instructions: the '<a
983 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
984 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
985 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
986 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
987 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
991 <!-- _______________________________________________________________________ -->
992 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
993 Instruction</a> </div>
994 <div class="doc_text">
996 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
997 ret void <i>; Return from void function</i>
1000 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1001 value) from a function, back to the caller.</p>
1002 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1003 returns a value and then causes control flow, and one that just causes
1004 control flow to occur.</p>
1006 <p>The '<tt>ret</tt>' instruction may return any '<a
1007 href="#t_firstclass">first class</a>' type. Notice that a function is
1008 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1009 instruction inside of the function that returns a value that does not
1010 match the return type of the function.</p>
1012 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1013 returns back to the calling function's context. If the caller is a "<a
1014 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1015 the instruction after the call. If the caller was an "<a
1016 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1017 at the beginning "normal" of the destination block. If the instruction
1018 returns a value, that value shall set the call or invoke instruction's
1021 <pre> ret int 5 <i>; Return an integer value of 5</i>
1022 ret void <i>; Return from a void function</i>
1025 <!-- _______________________________________________________________________ -->
1026 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1027 <div class="doc_text">
1029 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1032 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1033 transfer to a different basic block in the current function. There are
1034 two forms of this instruction, corresponding to a conditional branch
1035 and an unconditional branch.</p>
1037 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1038 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
1039 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1040 value as a target.</p>
1042 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
1043 argument is evaluated. If the value is <tt>true</tt>, control flows
1044 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1045 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1047 <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
1048 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
1050 <!-- _______________________________________________________________________ -->
1051 <div class="doc_subsubsection">
1052 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1055 <div class="doc_text">
1059 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1064 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1065 several different places. It is a generalization of the '<tt>br</tt>'
1066 instruction, allowing a branch to occur to one of many possible
1072 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1073 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1074 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1075 table is not allowed to contain duplicate constant entries.</p>
1079 <p>The <tt>switch</tt> instruction specifies a table of values and
1080 destinations. When the '<tt>switch</tt>' instruction is executed, this
1081 table is searched for the given value. If the value is found, control flow is
1082 transfered to the corresponding destination; otherwise, control flow is
1083 transfered to the default destination.</p>
1085 <h5>Implementation:</h5>
1087 <p>Depending on properties of the target machine and the particular
1088 <tt>switch</tt> instruction, this instruction may be code generated in different
1089 ways. For example, it could be generated as a series of chained conditional
1090 branches or with a lookup table.</p>
1095 <i>; Emulate a conditional br instruction</i>
1096 %Val = <a href="#i_cast">cast</a> bool %value to int
1097 switch int %Val, label %truedest [int 0, label %falsedest ]
1099 <i>; Emulate an unconditional br instruction</i>
1100 switch uint 0, label %dest [ ]
1102 <i>; Implement a jump table:</i>
1103 switch uint %val, label %otherwise [ uint 0, label %onzero
1104 uint 1, label %onone
1105 uint 2, label %ontwo ]
1108 <!-- _______________________________________________________________________ -->
1109 <div class="doc_subsubsection"> <a name="i_invoke">'<tt>invoke</tt>'
1110 Instruction</a> </div>
1111 <div class="doc_text">
1113 <pre> <result> = invoke <ptr to function ty> %<function ptr val>(<function args>)<br> to label <normal label> except label <exception label><br></pre>
1115 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a
1116 specified function, with the possibility of control flow transfer to
1117 either the '<tt>normal</tt>' <tt>label</tt> label or the '<tt>exception</tt>'<tt>label</tt>.
1118 If the callee function returns with the "<tt><a href="#i_ret">ret</a></tt>"
1119 instruction, control flow will return to the "normal" label. If the
1120 callee (or any indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
1121 instruction, control is interrupted, and continued at the dynamically
1122 nearest "except" label.</p>
1124 <p>This instruction requires several arguments:</p>
1126 <li>'<tt>ptr to function ty</tt>': shall be the signature of the
1127 pointer to function value being invoked. In most cases, this is a
1128 direct function invocation, but indirect <tt>invoke</tt>s are just as
1129 possible, branching off an arbitrary pointer to function value. </li>
1130 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer
1131 to a function to be invoked. </li>
1132 <li>'<tt>function args</tt>': argument list whose types match the
1133 function signature argument types. If the function signature indicates
1134 the function accepts a variable number of arguments, the extra
1135 arguments can be specified. </li>
1136 <li>'<tt>normal label</tt>': the label reached when the called
1137 function executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1138 <li>'<tt>exception label</tt>': the label reached when a callee
1139 returns with the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1142 <p>This instruction is designed to operate as a standard '<tt><a
1143 href="#i_call">call</a></tt>' instruction in most regards. The
1144 primary difference is that it establishes an association with a label,
1145 which is used by the runtime library to unwind the stack.</p>
1146 <p>This instruction is used in languages with destructors to ensure
1147 that proper cleanup is performed in the case of either a <tt>longjmp</tt>
1148 or a thrown exception. Additionally, this is important for
1149 implementation of '<tt>catch</tt>' clauses in high-level languages that
1152 <pre> %retval = invoke int %Test(int 15)<br> to label %Continue<br> except label %TestCleanup <i>; {int}:retval set</i>
1157 <!-- _______________________________________________________________________ -->
1159 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1160 Instruction</a> </div>
1162 <div class="doc_text">
1171 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1172 at the first callee in the dynamic call stack which used an <a
1173 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1174 primarily used to implement exception handling.</p>
1178 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1179 immediately halt. The dynamic call stack is then searched for the first <a
1180 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1181 execution continues at the "exceptional" destination block specified by the
1182 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1183 dynamic call chain, undefined behavior results.</p>
1186 <!-- _______________________________________________________________________ -->
1188 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1189 Instruction</a> </div>
1191 <div class="doc_text">
1200 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1201 instruction is used to inform the optimizer that a particular portion of the
1202 code is not reachable. This can be used to indicate that the code after a
1203 no-return function cannot be reached, and other facts.</p>
1207 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1212 <!-- ======================================================================= -->
1213 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1214 <div class="doc_text">
1215 <p>Binary operators are used to do most of the computation in a
1216 program. They require two operands, execute an operation on them, and
1217 produce a single value. The operands might represent
1218 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1219 The result value of a binary operator is not
1220 necessarily the same type as its operands.</p>
1221 <p>There are several different binary operators:</p>
1223 <!-- _______________________________________________________________________ -->
1224 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1225 Instruction</a> </div>
1226 <div class="doc_text">
1228 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1231 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1233 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1234 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1235 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1236 Both arguments must have identical types.</p>
1238 <p>The value produced is the integer or floating point sum of the two
1241 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
1244 <!-- _______________________________________________________________________ -->
1245 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1246 Instruction</a> </div>
1247 <div class="doc_text">
1249 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1252 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1254 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1255 instruction present in most other intermediate representations.</p>
1257 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1258 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1260 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1261 Both arguments must have identical types.</p>
1263 <p>The value produced is the integer or floating point difference of
1264 the two operands.</p>
1266 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
1267 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
1270 <!-- _______________________________________________________________________ -->
1271 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1272 Instruction</a> </div>
1273 <div class="doc_text">
1275 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1278 <p>The '<tt>mul</tt>' instruction returns the product of its two
1281 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1282 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1284 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1285 Both arguments must have identical types.</p>
1287 <p>The value produced is the integer or floating point product of the
1289 <p>There is no signed vs unsigned multiplication. The appropriate
1290 action is taken based on the type of the operand.</p>
1292 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
1295 <!-- _______________________________________________________________________ -->
1296 <div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
1297 Instruction</a> </div>
1298 <div class="doc_text">
1300 <pre> <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1303 <p>The '<tt>div</tt>' instruction returns the quotient of its two
1306 <p>The two arguments to the '<tt>div</tt>' instruction must be either <a
1307 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1309 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1310 Both arguments must have identical types.</p>
1312 <p>The value produced is the integer or floating point quotient of the
1315 <pre> <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
1318 <!-- _______________________________________________________________________ -->
1319 <div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
1320 Instruction</a> </div>
1321 <div class="doc_text">
1323 <pre> <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1326 <p>The '<tt>rem</tt>' instruction returns the remainder from the
1327 division of its two operands.</p>
1329 <p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1330 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1332 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1333 Both arguments must have identical types.</p>
1335 <p>This returns the <i>remainder</i> of a division (where the result
1336 has the same sign as the divisor), not the <i>modulus</i> (where the
1337 result has the same sign as the dividend) of a value. For more
1338 information about the difference, see: <a
1339 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1342 <pre> <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1345 <!-- _______________________________________________________________________ -->
1346 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1347 Instructions</a> </div>
1348 <div class="doc_text">
1350 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1351 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1352 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1353 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1354 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1355 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1358 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1359 value based on a comparison of their two operands.</p>
1361 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1362 be of <a href="#t_firstclass">first class</a> type (it is not possible
1363 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1364 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1367 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1368 value if both operands are equal.<br>
1369 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1370 value if both operands are unequal.<br>
1371 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1372 value if the first operand is less than the second operand.<br>
1373 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1374 value if the first operand is greater than the second operand.<br>
1375 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1376 value if the first operand is less than or equal to the second operand.<br>
1377 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1378 value if the first operand is greater than or equal to the second
1381 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1382 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1383 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1384 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1385 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1386 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1389 <!-- ======================================================================= -->
1390 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1391 Operations</a> </div>
1392 <div class="doc_text">
1393 <p>Bitwise binary operators are used to do various forms of
1394 bit-twiddling in a program. They are generally very efficient
1395 instructions and can commonly be strength reduced from other
1396 instructions. They require two operands, execute an operation on them,
1397 and produce a single value. The resulting value of the bitwise binary
1398 operators is always the same type as its first operand.</p>
1400 <!-- _______________________________________________________________________ -->
1401 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1402 Instruction</a> </div>
1403 <div class="doc_text">
1405 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1408 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1409 its two operands.</p>
1411 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1412 href="#t_integral">integral</a> values. Both arguments must have
1413 identical types.</p>
1415 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1417 <div style="align: center">
1418 <table border="1" cellspacing="0" cellpadding="4">
1449 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1450 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1451 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1454 <!-- _______________________________________________________________________ -->
1455 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1456 <div class="doc_text">
1458 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1461 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1462 or of its two operands.</p>
1464 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1465 href="#t_integral">integral</a> values. Both arguments must have
1466 identical types.</p>
1468 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1470 <div style="align: center">
1471 <table border="1" cellspacing="0" cellpadding="4">
1502 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1503 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1504 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1507 <!-- _______________________________________________________________________ -->
1508 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1509 Instruction</a> </div>
1510 <div class="doc_text">
1512 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1515 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1516 or of its two operands. The <tt>xor</tt> is used to implement the
1517 "one's complement" operation, which is the "~" operator in C.</p>
1519 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1520 href="#t_integral">integral</a> values. Both arguments must have
1521 identical types.</p>
1523 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1525 <div style="align: center">
1526 <table border="1" cellspacing="0" cellpadding="4">
1558 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1559 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1560 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1561 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1564 <!-- _______________________________________________________________________ -->
1565 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1566 Instruction</a> </div>
1567 <div class="doc_text">
1569 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1572 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1573 the left a specified number of bits.</p>
1575 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
1576 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1579 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
1581 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1582 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1583 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1586 <!-- _______________________________________________________________________ -->
1587 <div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1588 Instruction</a> </div>
1589 <div class="doc_text">
1591 <pre> <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1594 <p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1595 the right a specified number of bits.</p>
1597 <p>The first argument to the '<tt>shr</tt>' instruction must be an <a
1598 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1601 <p>If the first argument is a <a href="#t_signed">signed</a> type, the
1602 most significant bit is duplicated in the newly free'd bit positions.
1603 If the first argument is unsigned, zero bits shall fill the empty
1606 <pre> <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1607 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1608 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1609 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1610 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1613 <!-- ======================================================================= -->
1614 <div class="doc_subsection"> <a name="memoryops">Memory Access
1615 Operations</a></div>
1616 <div class="doc_text">
1617 <p>A key design point of an SSA-based representation is how it
1618 represents memory. In LLVM, no memory locations are in SSA form, which
1619 makes things very simple. This section describes how to read, write,
1620 allocate, and free memory in LLVM.</p>
1622 <!-- _______________________________________________________________________ -->
1623 <div class="doc_subsubsection"> <a name="i_malloc">'<tt>malloc</tt>'
1624 Instruction</a> </div>
1625 <div class="doc_text">
1627 <pre> <result> = malloc <type>, uint <NumElements> <i>; yields {type*}:result</i>
1628 <result> = malloc <type> <i>; yields {type*}:result</i>
1631 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
1632 heap and returns a pointer to it.</p>
1634 <p>The '<tt>malloc</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1635 bytes of memory from the operating system and returns a pointer of the
1636 appropriate type to the program. The second form of the instruction is
1637 a shorter version of the first instruction that defaults to allocating
1639 <p>'<tt>type</tt>' must be a sized type.</p>
1641 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
1642 a pointer is returned.</p>
1644 <pre> %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1647 href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1648 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1649 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1652 <!-- _______________________________________________________________________ -->
1653 <div class="doc_subsubsection"> <a name="i_free">'<tt>free</tt>'
1654 Instruction</a> </div>
1655 <div class="doc_text">
1657 <pre> free <type> <value> <i>; yields {void}</i>
1660 <p>The '<tt>free</tt>' instruction returns memory back to the unused
1661 memory heap, to be reallocated in the future.</p>
1664 <p>'<tt>value</tt>' shall be a pointer value that points to a value
1665 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
1668 <p>Access to the memory pointed to by the pointer is no longer defined
1669 after this instruction executes.</p>
1671 <pre> %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1672 free [4 x ubyte]* %array
1675 <!-- _______________________________________________________________________ -->
1676 <div class="doc_subsubsection"> <a name="i_alloca">'<tt>alloca</tt>'
1677 Instruction</a> </div>
1678 <div class="doc_text">
1680 <pre> <result> = alloca <type>, uint <NumElements> <i>; yields {type*}:result</i>
1681 <result> = alloca <type> <i>; yields {type*}:result</i>
1684 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
1685 stack frame of the procedure that is live until the current function
1686 returns to its caller.</p>
1688 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1689 bytes of memory on the runtime stack, returning a pointer of the
1690 appropriate type to the program. The second form of the instruction is
1691 a shorter version of the first that defaults to allocating one element.</p>
1692 <p>'<tt>type</tt>' may be any sized type.</p>
1694 <p>Memory is allocated, a pointer is returned. '<tt>alloca</tt>'d
1695 memory is automatically released when the function returns. The '<tt>alloca</tt>'
1696 instruction is commonly used to represent automatic variables that must
1697 have an address available. When the function returns (either with the <tt><a
1698 href="#i_ret">ret</a></tt> or <tt><a href="#i_invoke">invoke</a></tt>
1699 instructions), the memory is reclaimed.</p>
1701 <pre> %ptr = alloca int <i>; yields {int*}:ptr</i>
1702 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
1705 <!-- _______________________________________________________________________ -->
1706 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
1707 Instruction</a> </div>
1708 <div class="doc_text">
1710 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
1712 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
1714 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
1715 address to load from. The pointer must point to a <a
1716 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
1717 marked as <tt>volatile</tt> then the optimizer is not allowed to modify
1718 the number or order of execution of this <tt>load</tt> with other
1719 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
1722 <p>The location of memory pointed to is loaded.</p>
1724 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1726 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1727 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1730 <!-- _______________________________________________________________________ -->
1731 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
1732 Instruction</a> </div>
1734 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1735 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1738 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
1740 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
1741 to store and an address to store it into. The type of the '<tt><pointer></tt>'
1742 operand must be a pointer to the type of the '<tt><value></tt>'
1743 operand. If the <tt>store</tt> is marked as <tt>volatile</tt> then the
1744 optimizer is not allowed to modify the number or order of execution of
1745 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
1746 href="#i_store">store</a></tt> instructions.</p>
1748 <p>The contents of memory are updated to contain '<tt><value></tt>'
1749 at the location specified by the '<tt><pointer></tt>' operand.</p>
1751 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1753 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1754 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1756 <!-- _______________________________________________________________________ -->
1757 <div class="doc_subsubsection">
1758 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
1761 <div class="doc_text">
1764 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
1770 The '<tt>getelementptr</tt>' instruction is used to get the address of a
1771 subelement of an aggregate data structure.</p>
1775 <p>This instruction takes a list of integer constants that indicate what
1776 elements of the aggregate object to index to. The actual types of the arguments
1777 provided depend on the type of the first pointer argument. The
1778 '<tt>getelementptr</tt>' instruction is used to index down through the type
1779 levels of a structure. When indexing into a structure, only <tt>uint</tt>
1780 integer constants are allowed. When indexing into an array or pointer
1781 <tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
1783 <p>For example, let's consider a C code fragment and how it gets
1784 compiled to LLVM:</p>
1798 int *foo(struct ST *s) {
1799 return &s[1].Z.B[5][13];
1803 <p>The LLVM code generated by the GCC frontend is:</p>
1806 %RT = type { sbyte, [10 x [20 x int]], sbyte }
1807 %ST = type { int, double, %RT }
1811 int* %foo(%ST* %s) {
1813 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
1820 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
1821 on the pointer type that is being index into. <a href="#t_pointer">Pointer</a>
1822 and <a href="#t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
1823 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="#t_struct">structure</a>
1824 types require <tt>uint</tt> <b>constants</b>.</p>
1826 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
1827 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
1828 }</tt>' type, a structure. The second index indexes into the third element of
1829 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
1830 sbyte }</tt>' type, another structure. The third index indexes into the second
1831 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
1832 array. The two dimensions of the array are subscripted into, yielding an
1833 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction return a pointer
1834 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
1836 <p>Note that it is perfectly legal to index partially through a
1837 structure, returning a pointer to an inner element. Because of this,
1838 the LLVM code for the given testcase is equivalent to:</p>
1841 int* %foo(%ST* %s) {
1842 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
1843 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
1844 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
1845 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
1846 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
1852 <i>; yields [12 x ubyte]*:aptr</i>
1853 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
1857 <!-- ======================================================================= -->
1858 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
1859 <div class="doc_text">
1860 <p>The instructions in this category are the "miscellaneous"
1861 instructions, which defy better classification.</p>
1863 <!-- _______________________________________________________________________ -->
1864 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
1865 Instruction</a> </div>
1866 <div class="doc_text">
1868 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
1870 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
1871 the SSA graph representing the function.</p>
1873 <p>The type of the incoming values are specified with the first type
1874 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
1875 as arguments, with one pair for each predecessor basic block of the
1876 current block. Only values of <a href="#t_firstclass">first class</a>
1877 type may be used as the value arguments to the PHI node. Only labels
1878 may be used as the label arguments.</p>
1879 <p>There must be no non-phi instructions between the start of a basic
1880 block and the PHI instructions: i.e. PHI instructions must be first in
1883 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
1884 value specified by the parameter, depending on which basic block we
1885 came from in the last <a href="#terminators">terminator</a> instruction.</p>
1887 <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>
1890 <!-- _______________________________________________________________________ -->
1891 <div class="doc_subsubsection">
1892 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
1895 <div class="doc_text">
1900 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
1906 The '<tt>cast</tt>' instruction is used as the primitive means to convert
1907 integers to floating point, change data type sizes, and break type safety (by
1915 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
1916 class value, and a type to cast it to, which must also be a <a
1917 href="#t_firstclass">first class</a> type.
1923 This instruction follows the C rules for explicit casts when determining how the
1924 data being cast must change to fit in its new container.
1928 When casting to bool, any value that would be considered true in the context of
1929 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
1930 all else are '<tt>false</tt>'.
1934 When extending an integral value from a type of one signness to another (for
1935 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
1936 <b>source</b> value is signed, and zero-extended if the source value is
1937 unsigned. <tt>bool</tt> values are always zero extended into either zero or
1944 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
1945 %Y = cast int 123 to bool <i>; yields bool:true</i>
1949 <!-- _______________________________________________________________________ -->
1950 <div class="doc_subsubsection">
1951 <a name="i_select">'<tt>select</tt>' Instruction</a>
1954 <div class="doc_text">
1959 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
1965 The '<tt>select</tt>' instruction is used to choose one value based on a
1966 condition, without branching.
1973 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.
1979 If the boolean condition evaluates to true, the instruction returns the first
1980 value argument, otherwise it returns the second value argument.
1986 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
1994 <!-- _______________________________________________________________________ -->
1995 <div class="doc_subsubsection"> <a name="i_call">'<tt>call</tt>'
1996 Instruction</a> </div>
1997 <div class="doc_text">
1999 <pre> <result> = call <ty>* <fnptrval>(<param list>)<br></pre>
2001 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
2003 <p>This instruction requires several arguments:</p>
2006 <p>'<tt>ty</tt>': shall be the signature of the pointer to function
2007 value being invoked. The argument types must match the types implied
2008 by this signature.</p>
2011 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a
2012 function to be invoked. In most cases, this is a direct function
2013 invocation, but indirect <tt>call</tt>s are just as possible,
2014 calling an arbitrary pointer to function values.</p>
2017 <p>'<tt>function args</tt>': argument list whose types match the
2018 function signature argument types. If the function signature
2019 indicates the function accepts a variable number of arguments, the
2020 extra arguments can be specified.</p>
2024 <p>The '<tt>call</tt>' instruction is used to cause control flow to
2025 transfer to a specified function, with its incoming arguments bound to
2026 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
2027 instruction in the called function, control flow continues with the
2028 instruction after the function call, and the return value of the
2029 function is bound to the result argument. This is a simpler case of
2030 the <a href="#i_invoke">invoke</a> instruction.</p>
2032 <pre> %retval = call int %test(int %argc)<br> call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);<br></pre>
2035 <!-- _______________________________________________________________________ -->
2036 <div class="doc_subsubsection">
2037 <a name="i_vanext">'<tt>vanext</tt>' Instruction</a>
2040 <div class="doc_text">
2045 <resultarglist> = vanext <va_list> <arglist>, <argty>
2050 <p>The '<tt>vanext</tt>' instruction is used to access arguments passed
2051 through the "variable argument" area of a function call. It is used to
2052 implement the <tt>va_arg</tt> macro in C.</p>
2056 <p>This instruction takes a <tt>va_list</tt> value and the type of the
2057 argument. It returns another <tt>va_list</tt>. The actual type of
2058 <tt>va_list</tt> may be defined differently for different targets. Most targets
2059 use a <tt>va_list</tt> type of <tt>sbyte*</tt> or some other pointer type.</p>
2063 <p>The '<tt>vanext</tt>' instruction advances the specified <tt>va_list</tt>
2064 past an argument of the specified type. In conjunction with the <a
2065 href="#i_vaarg"><tt>vaarg</tt></a> instruction, it is used to implement
2066 the <tt>va_arg</tt> macro available in C. For more information, see
2067 the variable argument handling <a href="#int_varargs">Intrinsic
2070 <p>It is legal for this instruction to be called in a function which
2071 does not take a variable number of arguments, for example, the <tt>vfprintf</tt>
2074 <p><tt>vanext</tt> is an LLVM instruction instead of an <a
2075 href="#intrinsics">intrinsic function</a> because it takes a type as an
2076 argument. The type refers to the current argument in the <tt>va_list</tt>, it
2077 tells the compiler how far on the stack it needs to advance to find the next
2082 <p>See the <a href="#int_varargs">variable argument processing</a>
2087 <!-- _______________________________________________________________________ -->
2088 <div class="doc_subsubsection">
2089 <a name="i_vaarg">'<tt>vaarg</tt>' Instruction</a>
2092 <div class="doc_text">
2097 <resultval> = vaarg <va_list> <arglist>, <argty>
2102 <p>The '<tt>vaarg</tt>' instruction is used to access arguments passed through
2103 the "variable argument" area of a function call. It is used to implement the
2104 <tt>va_arg</tt> macro in C.</p>
2108 <p>This instruction takes a <tt>va_list</tt> value and the type of the
2109 argument. It returns a value of the specified argument type. Again, the actual
2110 type of <tt>va_list</tt> is target specific.</p>
2114 <p>The '<tt>vaarg</tt>' instruction loads an argument of the specified type from
2115 the specified <tt>va_list</tt>. In conjunction with the <a
2116 href="#i_vanext"><tt>vanext</tt></a> instruction, it is used to implement the
2117 <tt>va_arg</tt> macro available in C. For more information, see the variable
2118 argument handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
2120 <p>It is legal for this instruction to be called in a function which does not
2121 take a variable number of arguments, for example, the <tt>vfprintf</tt>
2124 <p><tt>vaarg</tt> is an LLVM instruction instead of an <a
2125 href="#intrinsics">intrinsic function</a> because it takes an type as an
2130 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
2134 <!-- *********************************************************************** -->
2135 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
2136 <!-- *********************************************************************** -->
2138 <div class="doc_text">
2140 <p>LLVM supports the notion of an "intrinsic function". These functions have
2141 well known names and semantics, and are required to follow certain
2142 restrictions. Overall, these instructions represent an extension mechanism for
2143 the LLVM language that does not require changing all of the transformations in
2144 LLVM to add to the language (or the bytecode reader/writer, the parser,
2147 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix, this
2148 prefix is reserved in LLVM for intrinsic names, thus functions may not be named
2149 this. Intrinsic functions must always be external functions: you cannot define
2150 the body of intrinsic functions. Intrinsic functions may only be used in call
2151 or invoke instructions: it is illegal to take the address of an intrinsic
2152 function. Additionally, because intrinsic functions are part of the LLVM
2153 language, it is required that they all be documented here if any are added.</p>
2157 Adding an intrinsic to LLVM is straight-forward if it is possible to express the
2158 concept in LLVM directly (ie, code generator support is not _required_). To do
2159 this, extend the default implementation of the IntrinsicLowering class to handle
2160 the intrinsic. Code generators use this class to lower intrinsics they do not
2161 understand to raw LLVM instructions that they do.
2166 <!-- ======================================================================= -->
2167 <div class="doc_subsection">
2168 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
2171 <div class="doc_text">
2173 <p>Variable argument support is defined in LLVM with the <a
2174 href="#i_vanext"><tt>vanext</tt></a> instruction and these three
2175 intrinsic functions. These functions are related to the similarly
2176 named macros defined in the <tt><stdarg.h></tt> header file.</p>
2178 <p>All of these functions operate on arguments that use a
2179 target-specific value type "<tt>va_list</tt>". The LLVM assembly
2180 language reference manual does not define what this type is, so all
2181 transformations should be prepared to handle intrinsics with any type
2184 <p>This example shows how the <a href="#i_vanext"><tt>vanext</tt></a>
2185 instruction and the variable argument handling intrinsic functions are
2189 int %test(int %X, ...) {
2190 ; Initialize variable argument processing
2191 %ap = call sbyte* %<a href="#i_va_start">llvm.va_start</a>()
2193 ; Read a single integer argument
2194 %tmp = vaarg sbyte* %ap, int
2196 ; Advance to the next argument
2197 %ap2 = vanext sbyte* %ap, int
2199 ; Demonstrate usage of llvm.va_copy and llvm.va_end
2200 %aq = call sbyte* %<a href="#i_va_copy">llvm.va_copy</a>(sbyte* %ap2)
2201 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %aq)
2203 ; Stop processing of arguments.
2204 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %ap2)
2210 <!-- _______________________________________________________________________ -->
2211 <div class="doc_subsubsection">
2212 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
2216 <div class="doc_text">
2218 <pre> call <va_list> ()* %llvm.va_start()<br></pre>
2220 <p>The '<tt>llvm.va_start</tt>' intrinsic returns a new <tt><arglist></tt>
2221 for subsequent use by the variable argument intrinsics.</p>
2223 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
2224 macro available in C. In a target-dependent way, it initializes and
2225 returns a <tt>va_list</tt> element, so that the next <tt>vaarg</tt>
2226 will produce the first variable argument passed to the function. Unlike
2227 the C <tt>va_start</tt> macro, this intrinsic does not need to know the
2228 last argument of the function, the compiler can figure that out.</p>
2229 <p>Note that this intrinsic function is only legal to be called from
2230 within the body of a variable argument function.</p>
2233 <!-- _______________________________________________________________________ -->
2234 <div class="doc_subsubsection">
2235 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
2238 <div class="doc_text">
2240 <pre> call void (<va_list>)* %llvm.va_end(<va_list> <arglist>)<br></pre>
2242 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
2243 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
2244 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
2246 <p>The argument is a <tt>va_list</tt> to destroy.</p>
2248 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
2249 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
2250 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
2251 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
2252 with calls to <tt>llvm.va_end</tt>.</p>
2255 <!-- _______________________________________________________________________ -->
2256 <div class="doc_subsubsection">
2257 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
2260 <div class="doc_text">
2265 call <va_list> (<va_list>)* %llvm.va_copy(<va_list> <destarglist>)
2270 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
2271 from the source argument list to the destination argument list.</p>
2275 <p>The argument is the <tt>va_list</tt> to copy.</p>
2279 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
2280 macro available in C. In a target-dependent way, it copies the source
2281 <tt>va_list</tt> element into the returned list. This intrinsic is necessary
2282 because the <tt><a href="#i_va_start">llvm.va_start</a></tt> intrinsic may be
2283 arbitrarily complex and require memory allocation, for example.</p>
2287 <!-- ======================================================================= -->
2288 <div class="doc_subsection">
2289 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
2292 <div class="doc_text">
2295 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
2296 Collection</a> requires the implementation and generation of these intrinsics.
2297 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
2298 stack</a>, as well as garbage collector implementations that require <a
2299 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
2300 Front-ends for type-safe garbage collected languages should generate these
2301 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
2302 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
2306 <!-- _______________________________________________________________________ -->
2307 <div class="doc_subsubsection">
2308 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
2311 <div class="doc_text">
2316 call void (<ty>**, <ty2>*)* %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
2321 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
2322 the code generator, and allows some metadata to be associated with it.</p>
2326 <p>The first argument specifies the address of a stack object that contains the
2327 root pointer. The second pointer (which must be either a constant or a global
2328 value address) contains the meta-data to be associated with the root.</p>
2332 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
2333 location. At compile-time, the code generator generates information to allow
2334 the runtime to find the pointer at GC safe points.
2340 <!-- _______________________________________________________________________ -->
2341 <div class="doc_subsubsection">
2342 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
2345 <div class="doc_text">
2350 call sbyte* (sbyte**)* %llvm.gcread(sbyte** %Ptr)
2355 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
2356 locations, allowing garbage collector implementations that require read
2361 <p>The argument is the address to read from, which should be an address
2362 allocated from the garbage collector.</p>
2366 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
2367 instruction, but may be replaced with substantially more complex code by the
2368 garbage collector runtime, as needed.</p>
2373 <!-- _______________________________________________________________________ -->
2374 <div class="doc_subsubsection">
2375 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
2378 <div class="doc_text">
2383 call void (sbyte*, sbyte**)* %llvm.gcwrite(sbyte* %P1, sbyte** %P2)
2388 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
2389 locations, allowing garbage collector implementations that require write
2390 barriers (such as generational or reference counting collectors).</p>
2394 <p>The first argument is the reference to store, and the second is the heap
2395 location to store to.</p>
2399 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
2400 instruction, but may be replaced with substantially more complex code by the
2401 garbage collector runtime, as needed.</p>
2407 <!-- ======================================================================= -->
2408 <div class="doc_subsection">
2409 <a name="int_codegen">Code Generator Intrinsics</a>
2412 <div class="doc_text">
2414 These intrinsics are provided by LLVM to expose special features that may only
2415 be implemented with code generator support.
2420 <!-- _______________________________________________________________________ -->
2421 <div class="doc_subsubsection">
2422 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
2425 <div class="doc_text">
2429 call void* ()* %llvm.returnaddress(uint <level>)
2435 The '<tt>llvm.returnaddress</tt>' intrinsic returns a target-specific value
2436 indicating the return address of the current function or one of its callers.
2442 The argument to this intrinsic indicates which function to return the address
2443 for. Zero indicates the calling function, one indicates its caller, etc. The
2444 argument is <b>required</b> to be a constant integer value.
2450 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
2451 the return address of the specified call frame, or zero if it cannot be
2452 identified. The value returned by this intrinsic is likely to be incorrect or 0
2453 for arguments other than zero, so it should only be used for debugging purposes.
2457 Note that calling this intrinsic does not prevent function inlining or other
2458 aggressive transformations, so the value returned may not be that of the obvious
2459 source-language caller.
2464 <!-- _______________________________________________________________________ -->
2465 <div class="doc_subsubsection">
2466 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
2469 <div class="doc_text">
2473 call void* ()* %llvm.frameaddress(uint <level>)
2479 The '<tt>llvm.frameaddress</tt>' intrinsic returns the target-specific frame
2480 pointer value for the specified stack frame.
2486 The argument to this intrinsic indicates which function to return the frame
2487 pointer for. Zero indicates the calling function, one indicates its caller,
2488 etc. The argument is <b>required</b> to be a constant integer value.
2494 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
2495 the frame address of the specified call frame, or zero if it cannot be
2496 identified. The value returned by this intrinsic is likely to be incorrect or 0
2497 for arguments other than zero, so it should only be used for debugging purposes.
2501 Note that calling this intrinsic does not prevent function inlining or other
2502 aggressive transformations, so the value returned may not be that of the obvious
2503 source-language caller.
2507 <!-- _______________________________________________________________________ -->
2508 <div class="doc_subsubsection">
2509 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
2512 <div class="doc_text">
2516 call void (sbyte *, uint, uint)* %llvm.prefetch(sbyte * <address>,
2518 uint <locality>)
2525 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
2526 a prefetch instruction if supported, otherwise it is a noop. Prefetches have no
2527 effect on the behavior of the program, but can change its performance
2534 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
2535 determining if the fetch should be for a read (0) or write (1), and
2536 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
2537 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
2538 <tt>locality</tt> arguments must be constant integers.
2544 This intrinsic does not modify the behavior of the program. In particular,
2545 prefetches cannot trap and do not produce a value. On targets that support this
2546 intrinsic, the prefetch can provide hints to the processor cache for better
2552 <!-- _______________________________________________________________________ -->
2553 <div class="doc_subsubsection">
2554 <a name="i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
2557 <div class="doc_text">
2561 call void (uint)* %llvm.pcmarker( uint <id> )
2568 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a PC in a region of
2569 code to simulators and other tools. The method is target specific, but it is
2570 expected that the marker will use exported symbols to transmit the PC of the marker.
2571 The marker makes no guaranties that it will remain with any specific instruction
2572 after optimizations. It is possible that the presense of a marker will inhibit
2573 optimizations. The intended use is to be inserted after optmizations to allow
2574 corrolations of simulation runs.
2580 <tt>id</tt> is a numerical id identifying the marker.
2586 This intrinsic does not modify the behavior of the program. Backends that do not
2587 support this intrinisic may ignore it.
2593 <!-- ======================================================================= -->
2594 <div class="doc_subsection">
2595 <a name="int_os">Operating System Intrinsics</a>
2598 <div class="doc_text">
2600 These intrinsics are provided by LLVM to support the implementation of
2601 operating system level code.
2606 <!-- _______________________________________________________________________ -->
2607 <div class="doc_subsubsection">
2608 <a name="i_readport">'<tt>llvm.readport</tt>' Intrinsic</a>
2611 <div class="doc_text">
2615 call <integer type> (<integer type>)* %llvm.readport (<integer type> <address>)
2621 The '<tt>llvm.readport</tt>' intrinsic reads data from the specified hardware
2628 The argument to this intrinsic indicates the hardware I/O address from which
2629 to read the data. The address is in the hardware I/O address namespace (as
2630 opposed to being a memory location for memory mapped I/O).
2636 The '<tt>llvm.readport</tt>' intrinsic reads data from the hardware I/O port
2637 specified by <i>address</i> and returns the value. The address and return
2638 value must be integers, but the size is dependent upon the platform upon which
2639 the program is code generated. For example, on x86, the address must be an
2640 unsigned 16 bit value, and the return value must be 8, 16, or 32 bits.
2645 <!-- _______________________________________________________________________ -->
2646 <div class="doc_subsubsection">
2647 <a name="i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a>
2650 <div class="doc_text">
2654 call void (<integer type>, <integer type>)*
2655 %llvm.writeport (<integer type> <value>,
2656 <integer type> <address>)
2662 The '<tt>llvm.writeport</tt>' intrinsic writes data to the specified hardware
2669 The first argument is the value to write to the I/O port.
2673 The second argument indicates the hardware I/O address to which data should be
2674 written. The address is in the hardware I/O address namespace (as opposed to
2675 being a memory location for memory mapped I/O).
2681 The '<tt>llvm.writeport</tt>' intrinsic writes <i>value</i> to the I/O port
2682 specified by <i>address</i>. The address and value must be integers, but the
2683 size is dependent upon the platform upon which the program is code generated.
2684 For example, on x86, the address must be an unsigned 16 bit value, and the
2685 value written must be 8, 16, or 32 bits in length.
2690 <!-- _______________________________________________________________________ -->
2691 <div class="doc_subsubsection">
2692 <a name="i_readio">'<tt>llvm.readio</tt>' Intrinsic</a>
2695 <div class="doc_text">
2699 call <result> (<ty>*)* %llvm.readio (<ty> * <pointer>)
2705 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2712 The argument to this intrinsic is a pointer indicating the memory address from
2713 which to read the data. The data must be a
2714 <a href="#t_firstclass">first class</a> type.
2720 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2721 location specified by <i>pointer</i> and returns the value. The argument must
2722 be a pointer, and the return value must be a
2723 <a href="#t_firstclass">first class</a> type. However, certain architectures
2724 may not support I/O on all first class types. For example, 32 bit processors
2725 may only support I/O on data types that are 32 bits or less.
2729 This intrinsic enforces an in-order memory model for llvm.readio and
2730 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2731 scheduled processors may execute loads and stores out of order, re-ordering at
2732 run time accesses to memory mapped I/O registers. Using these intrinsics
2733 ensures that accesses to memory mapped I/O registers occur in program order.
2738 <!-- _______________________________________________________________________ -->
2739 <div class="doc_subsubsection">
2740 <a name="i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a>
2743 <div class="doc_text">
2747 call void (<ty1>, <ty2>*)* %llvm.writeio (<ty1> <value>, <ty2> * <pointer>)
2753 The '<tt>llvm.writeio</tt>' intrinsic writes data to the specified memory
2760 The first argument is the value to write to the memory mapped I/O location.
2761 The second argument is a pointer indicating the memory address to which the
2762 data should be written.
2768 The '<tt>llvm.writeio</tt>' intrinsic writes <i>value</i> to the memory mapped
2769 I/O address specified by <i>pointer</i>. The value must be a
2770 <a href="#t_firstclass">first class</a> type. However, certain architectures
2771 may not support I/O on all first class types. For example, 32 bit processors
2772 may only support I/O on data types that are 32 bits or less.
2776 This intrinsic enforces an in-order memory model for llvm.readio and
2777 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2778 scheduled processors may execute loads and stores out of order, re-ordering at
2779 run time accesses to memory mapped I/O registers. Using these intrinsics
2780 ensures that accesses to memory mapped I/O registers occur in program order.
2785 <!-- ======================================================================= -->
2786 <div class="doc_subsection">
2787 <a name="int_libc">Standard C Library Intrinsics</a>
2790 <div class="doc_text">
2792 LLVM provides intrinsics for a few important standard C library functions.
2793 These intrinsics allow source-language front-ends to pass information about the
2794 alignment of the pointer arguments to the code generator, providing opportunity
2795 for more efficient code generation.
2800 <!-- _______________________________________________________________________ -->
2801 <div class="doc_subsubsection">
2802 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
2805 <div class="doc_text">
2809 call void (sbyte*, sbyte*, uint, uint)* %llvm.memcpy(sbyte* <dest>, sbyte* <src>,
2810 uint <len>, uint <align>)
2816 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2817 location to the destination location.
2821 Note that, unlike the standard libc function, the <tt>llvm.memcpy</tt> intrinsic
2822 does not return a value, and takes an extra alignment argument.
2828 The first argument is a pointer to the destination, the second is a pointer to
2829 the source. The third argument is an (arbitrarily sized) integer argument
2830 specifying the number of bytes to copy, and the fourth argument is the alignment
2831 of the source and destination locations.
2835 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2836 the caller guarantees that the size of the copy is a multiple of the alignment
2837 and that both the source and destination pointers are aligned to that boundary.
2843 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2844 location to the destination location, which are not allowed to overlap. It
2845 copies "len" bytes of memory over. If the argument is known to be aligned to
2846 some boundary, this can be specified as the fourth argument, otherwise it should
2852 <!-- _______________________________________________________________________ -->
2853 <div class="doc_subsubsection">
2854 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
2857 <div class="doc_text">
2861 call void (sbyte*, sbyte*, uint, uint)* %llvm.memmove(sbyte* <dest>, sbyte* <src>,
2862 uint <len>, uint <align>)
2868 The '<tt>llvm.memmove</tt>' intrinsic moves a block of memory from the source
2869 location to the destination location. It is similar to the '<tt>llvm.memcpy</tt>'
2870 intrinsic but allows the two memory locations to overlap.
2874 Note that, unlike the standard libc function, the <tt>llvm.memmove</tt> intrinsic
2875 does not return a value, and takes an extra alignment argument.
2881 The first argument is a pointer to the destination, the second is a pointer to
2882 the source. The third argument is an (arbitrarily sized) integer argument
2883 specifying the number of bytes to copy, and the fourth argument is the alignment
2884 of the source and destination locations.
2888 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2889 the caller guarantees that the size of the copy is a multiple of the alignment
2890 and that both the source and destination pointers are aligned to that boundary.
2896 The '<tt>llvm.memmove</tt>' intrinsic copies a block of memory from the source
2897 location to the destination location, which may overlap. It
2898 copies "len" bytes of memory over. If the argument is known to be aligned to
2899 some boundary, this can be specified as the fourth argument, otherwise it should
2905 <!-- _______________________________________________________________________ -->
2906 <div class="doc_subsubsection">
2907 <a name="i_memset">'<tt>llvm.memset</tt>' Intrinsic</a>
2910 <div class="doc_text">
2914 call void (sbyte*, ubyte, uint, uint)* %llvm.memset(sbyte* <dest>, ubyte <val>,
2915 uint <len>, uint <align>)
2921 The '<tt>llvm.memset</tt>' intrinsic fills a block of memory with a particular
2926 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
2927 does not return a value, and takes an extra alignment argument.
2933 The first argument is a pointer to the destination to fill, the second is the
2934 byte value to fill it with, the third argument is an (arbitrarily sized) integer
2935 argument specifying the number of bytes to fill, and the fourth argument is the
2936 known alignment of destination location.
2940 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2941 the caller guarantees that the size of the copy is a multiple of the alignment
2942 and that the destination pointer is aligned to that boundary.
2948 The '<tt>llvm.memset</tt>' intrinsic fills "len" bytes of memory starting at the
2949 destination location. If the argument is known to be aligned to some boundary,
2950 this can be specified as the fourth argument, otherwise it should be set to 0 or
2956 <!-- _______________________________________________________________________ -->
2957 <div class="doc_subsubsection">
2958 <a name="i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a>
2961 <div class="doc_text">
2965 call bool (<float or double>, <float or double>)* %llvm.isunordered(<float or double> Val1,
2966 <float or double> Val2)
2972 The '<tt>llvm.isunordered</tt>' intrinsic returns true if either or both of the
2973 specified floating point values is a NAN.
2979 The arguments are floating point numbers of the same type.
2985 If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
2993 <!-- ======================================================================= -->
2994 <div class="doc_subsection">
2995 <a name="int_debugger">Debugger Intrinsics</a>
2998 <div class="doc_text">
3000 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
3001 are described in the <a
3002 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
3003 Debugging</a> document.
3008 <!-- *********************************************************************** -->
3011 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
3012 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
3013 <a href="http://validator.w3.org/check/referer"><img
3014 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
3016 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
3017 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
3018 Last modified: $Date$