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>
130 <li><a href="#int_os">Operating System Intrinsics</a>
132 <li><a href="#i_readport">'<tt>llvm.readport</tt>' Intrinsic</a></li>
133 <li><a href="#i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a></li>
134 <li><a href="#i_readio">'<tt>llvm.readio</tt>' Intrinsic</a></li>
135 <li><a href="#i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a></li>
137 <li><a href="#int_libc">Standard C Library Intrinsics</a>
139 <li><a href="#i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a></li>
140 <li><a href="#i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a></li>
141 <li><a href="#i_memset">'<tt>llvm.memset</tt>' Intrinsic</a></li>
142 <li><a href="#i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a></li>
145 <li><a href="#int_debugger">Debugger intrinsics</a></li>
150 <div class="doc_author">
151 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
152 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
155 <!-- *********************************************************************** -->
156 <div class="doc_section"> <a name="abstract">Abstract </a></div>
157 <!-- *********************************************************************** -->
159 <div class="doc_text">
160 <p>This document is a reference manual for the LLVM assembly language.
161 LLVM is an SSA based representation that provides type safety,
162 low-level operations, flexibility, and the capability of representing
163 'all' high-level languages cleanly. It is the common code
164 representation used throughout all phases of the LLVM compilation
168 <!-- *********************************************************************** -->
169 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
170 <!-- *********************************************************************** -->
172 <div class="doc_text">
174 <p>The LLVM code representation is designed to be used in three
175 different forms: as an in-memory compiler IR, as an on-disk bytecode
176 representation (suitable for fast loading by a Just-In-Time compiler),
177 and as a human readable assembly language representation. This allows
178 LLVM to provide a powerful intermediate representation for efficient
179 compiler transformations and analysis, while providing a natural means
180 to debug and visualize the transformations. The three different forms
181 of LLVM are all equivalent. This document describes the human readable
182 representation and notation.</p>
184 <p>The LLVM representation aims to be a light-weight and low-level
185 while being expressive, typed, and extensible at the same time. It
186 aims to be a "universal IR" of sorts, by being at a low enough level
187 that high-level ideas may be cleanly mapped to it (similar to how
188 microprocessors are "universal IR's", allowing many source languages to
189 be mapped to them). By providing type information, LLVM can be used as
190 the target of optimizations: for example, through pointer analysis, it
191 can be proven that a C automatic variable is never accessed outside of
192 the current function... allowing it to be promoted to a simple SSA
193 value instead of a memory location.</p>
197 <!-- _______________________________________________________________________ -->
198 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
200 <div class="doc_text">
202 <p>It is important to note that this document describes 'well formed'
203 LLVM assembly language. There is a difference between what the parser
204 accepts and what is considered 'well formed'. For example, the
205 following instruction is syntactically okay, but not well formed:</p>
208 %x = <a href="#i_add">add</a> int 1, %x
211 <p>...because the definition of <tt>%x</tt> does not dominate all of
212 its uses. The LLVM infrastructure provides a verification pass that may
213 be used to verify that an LLVM module is well formed. This pass is
214 automatically run by the parser after parsing input assembly, and by
215 the optimizer before it outputs bytecode. The violations pointed out
216 by the verifier pass indicate bugs in transformation passes or input to
219 <!-- Describe the typesetting conventions here. --> </div>
221 <!-- *********************************************************************** -->
222 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
223 <!-- *********************************************************************** -->
225 <div class="doc_text">
227 <p>LLVM uses three different forms of identifiers, for different
231 <li>Named values are represented as a string of characters with a '%' prefix.
232 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
233 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
234 Identifiers which require other characters in their names can be surrounded
235 with quotes. In this way, anything except a <tt>"</tt> character can be used
238 <li>Unnamed values are represented as an unsigned numeric value with a '%'
239 prefix. For example, %12, %2, %44.</li>
241 <li>Constants, which are described in a <a href="#constants">section about
242 constants</a>, below.</li>
245 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
246 don't need to worry about name clashes with reserved words, and the set of
247 reserved words may be expanded in the future without penalty. Additionally,
248 unnamed identifiers allow a compiler to quickly come up with a temporary
249 variable without having to avoid symbol table conflicts.</p>
251 <p>Reserved words in LLVM are very similar to reserved words in other
252 languages. There are keywords for different opcodes ('<tt><a
253 href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
254 href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
255 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>', etc...),
256 and others. These reserved words cannot conflict with variable names, because
257 none of them start with a '%' character.</p>
259 <p>Here is an example of LLVM code to multiply the integer variable
260 '<tt>%X</tt>' by 8:</p>
265 %result = <a href="#i_mul">mul</a> uint %X, 8
268 <p>After strength reduction:</p>
271 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
274 <p>And the hard way:</p>
277 <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
278 <a href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
279 %result = <a href="#i_add">add</a> uint %1, %1
282 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
283 important lexical features of LLVM:</p>
287 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
290 <li>Unnamed temporaries are created when the result of a computation is not
291 assigned to a named value.</li>
293 <li>Unnamed temporaries are numbered sequentially</li>
297 <p>...and it also show a convention that we follow in this document. When
298 demonstrating instructions, we will follow an instruction with a comment that
299 defines the type and name of value produced. Comments are shown in italic
304 <!-- *********************************************************************** -->
305 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
306 <!-- *********************************************************************** -->
308 <!-- ======================================================================= -->
309 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
312 <div class="doc_text">
314 <p>LLVM programs are composed of "Module"s, each of which is a
315 translation unit of the input programs. Each module consists of
316 functions, global variables, and symbol table entries. Modules may be
317 combined together with the LLVM linker, which merges function (and
318 global variable) definitions, resolves forward declarations, and merges
319 symbol table entries. Here is an example of the "hello world" module:</p>
321 <pre><i>; Declare the string constant as a global constant...</i>
322 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
323 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
325 <i>; External declaration of the puts function</i>
326 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
328 <i>; Definition of main function</i>
329 int %main() { <i>; int()* </i>
330 <i>; Convert [13x sbyte]* to sbyte *...</i>
332 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
334 <i>; Call puts function to write out the string to stdout...</i>
336 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
338 href="#i_ret">ret</a> int 0<br>}<br></pre>
340 <p>This example is made up of a <a href="#globalvars">global variable</a>
341 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
342 function, and a <a href="#functionstructure">function definition</a>
343 for "<tt>main</tt>".</p>
345 <p>In general, a module is made up of a list of global values,
346 where both functions and global variables are global values. Global values are
347 represented by a pointer to a memory location (in this case, a pointer to an
348 array of char, and a pointer to a function), and have one of the following <a
349 href="#linkage">linkage types</a>.</p>
353 <!-- ======================================================================= -->
354 <div class="doc_subsection">
355 <a name="linkage">Linkage Types</a>
358 <div class="doc_text">
361 All Global Variables and Functions have one of the following types of linkage:
366 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
368 <dd>Global values with internal linkage are only directly accessible by
369 objects in the current module. In particular, linking code into a module with
370 an internal global value may cause the internal to be renamed as necessary to
371 avoid collisions. Because the symbol is internal to the module, all
372 references can be updated. This corresponds to the notion of the
373 '<tt>static</tt>' keyword in C, or the idea of "anonymous namespaces" in C++.
376 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
378 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
379 the twist that linking together two modules defining the same
380 <tt>linkonce</tt> globals will cause one of the globals to be discarded. This
381 is typically used to implement inline functions. Unreferenced
382 <tt>linkonce</tt> globals are allowed to be discarded.
385 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
387 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
388 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
389 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.
392 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
394 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
395 pointer to array type. When two global variables with appending linkage are
396 linked together, the two global arrays are appended together. This is the
397 LLVM, typesafe, equivalent of having the system linker append together
398 "sections" with identical names when .o files are linked.
401 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
403 <dd>If none of the above identifiers are used, the global is externally
404 visible, meaning that it participates in linkage and can be used to resolve
405 external symbol references.
409 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
410 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
411 variable and was linked with this one, one of the two would be renamed,
412 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
413 external (i.e., lacking any linkage declarations), they are accessible
414 outside of the current module. It is illegal for a function <i>declaration</i>
415 to have any linkage type other than "externally visible".</a></p>
419 <!-- ======================================================================= -->
420 <div class="doc_subsection">
421 <a name="globalvars">Global Variables</a>
424 <div class="doc_text">
426 <p>Global variables define regions of memory allocated at compilation
427 time instead of run-time. Global variables may optionally be
428 initialized. A variable may be defined as a global "constant", which
429 indicates that the contents of the variable will never be modified
430 (enabling better optimization, allowing the global data to be placed in the
431 read-only section of an executable, etc).</p>
433 <p>As SSA values, global variables define pointer values that are in
434 scope (i.e. they dominate) all basic blocks in the program. Global
435 variables always define a pointer to their "content" type because they
436 describe a region of memory, and all memory objects in LLVM are
437 accessed through pointers.</p>
442 <!-- ======================================================================= -->
443 <div class="doc_subsection">
444 <a name="functionstructure">Functions</a>
447 <div class="doc_text">
449 <p>LLVM function definitions are composed of a (possibly empty) argument list,
450 an opening curly brace, a list of basic blocks, and a closing curly brace. LLVM
451 function declarations are defined with the "<tt>declare</tt>" keyword, a
452 function name, and a function signature.</p>
454 <p>A function definition contains a list of basic blocks, forming the CFG for
455 the function. Each basic block may optionally start with a label (giving the
456 basic block a symbol table entry), contains a list of instructions, and ends
457 with a <a href="#terminators">terminator</a> instruction (such as a branch or
458 function return).</p>
460 <p>The first basic block in program is special in two ways: it is immediately
461 executed on entrance to the function, and it is not allowed to have predecessor
462 basic blocks (i.e. there can not be any branches to the entry block of a
463 function). Because the block can have no predecessors, it also cannot have any
464 <a href="#i_phi">PHI nodes</a>.</p>
466 <p>LLVM functions are identified by their name and type signature. Hence, two
467 functions with the same name but different parameter lists or return values are
468 considered different functions, and LLVM will resolves references to each
475 <!-- *********************************************************************** -->
476 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
477 <!-- *********************************************************************** -->
479 <div class="doc_text">
481 <p>The LLVM type system is one of the most important features of the
482 intermediate representation. Being typed enables a number of
483 optimizations to be performed on the IR directly, without having to do
484 extra analyses on the side before the transformation. A strong type
485 system makes it easier to read the generated code and enables novel
486 analyses and transformations that are not feasible to perform on normal
487 three address code representations.</p>
491 <!-- ======================================================================= -->
492 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
493 <div class="doc_text">
494 <p>The primitive types are the fundamental building blocks of the LLVM
495 system. The current set of primitive types are as follows:</p>
497 <table class="layout">
502 <tr><th>Type</th><th>Description</th></tr>
503 <tr><td><tt>void</tt></td><td>No value</td></tr>
504 <tr><td><tt>ubyte</tt></td><td>Unsigned 8 bit value</td></tr>
505 <tr><td><tt>ushort</tt></td><td>Unsigned 16 bit value</td></tr>
506 <tr><td><tt>uint</tt></td><td>Unsigned 32 bit value</td></tr>
507 <tr><td><tt>ulong</tt></td><td>Unsigned 64 bit value</td></tr>
508 <tr><td><tt>float</tt></td><td>32 bit floating point value</td></tr>
509 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
516 <tr><th>Type</th><th>Description</th></tr>
517 <tr><td><tt>bool</tt></td><td>True or False value</td></tr>
518 <tr><td><tt>sbyte</tt></td><td>Signed 8 bit value</td></tr>
519 <tr><td><tt>short</tt></td><td>Signed 16 bit value</td></tr>
520 <tr><td><tt>int</tt></td><td>Signed 32 bit value</td></tr>
521 <tr><td><tt>long</tt></td><td>Signed 64 bit value</td></tr>
522 <tr><td><tt>double</tt></td><td>64 bit floating point value</td></tr>
530 <!-- _______________________________________________________________________ -->
531 <div class="doc_subsubsection"> <a name="t_classifications">Type
532 Classifications</a> </div>
533 <div class="doc_text">
534 <p>These different primitive types fall into a few useful
537 <table border="1" cellspacing="0" cellpadding="4">
539 <tr><th>Classification</th><th>Types</th></tr>
541 <td><a name="t_signed">signed</a></td>
542 <td><tt>sbyte, short, int, long, float, double</tt></td>
545 <td><a name="t_unsigned">unsigned</a></td>
546 <td><tt>ubyte, ushort, uint, ulong</tt></td>
549 <td><a name="t_integer">integer</a></td>
550 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
553 <td><a name="t_integral">integral</a></td>
554 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt>
558 <td><a name="t_floating">floating point</a></td>
559 <td><tt>float, double</tt></td>
562 <td><a name="t_firstclass">first class</a></td>
563 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long,<br>
564 float, double, <a href="#t_pointer">pointer</a>,
565 <a href="#t_packed">packed</a></tt></td>
570 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
571 most important. Values of these types are the only ones which can be
572 produced by instructions, passed as arguments, or used as operands to
573 instructions. This means that all structures and arrays must be
574 manipulated either by pointer or by component.</p>
577 <!-- ======================================================================= -->
578 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
580 <div class="doc_text">
582 <p>The real power in LLVM comes from the derived types in the system.
583 This is what allows a programmer to represent arrays, functions,
584 pointers, and other useful types. Note that these derived types may be
585 recursive: For example, it is possible to have a two dimensional array.</p>
589 <!-- _______________________________________________________________________ -->
590 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
592 <div class="doc_text">
596 <p>The array type is a very simple derived type that arranges elements
597 sequentially in memory. The array type requires a size (number of
598 elements) and an underlying data type.</p>
603 [<# elements> x <elementtype>]
606 <p>The number of elements is a constant integer value, elementtype may
607 be any type with a size.</p>
610 <table class="layout">
613 <tt>[40 x int ]</tt><br/>
614 <tt>[41 x int ]</tt><br/>
615 <tt>[40 x uint]</tt><br/>
618 Array of 40 integer values.<br/>
619 Array of 41 integer values.<br/>
620 Array of 40 unsigned integer values.<br/>
624 <p>Here are some examples of multidimensional arrays:</p>
625 <table class="layout">
628 <tt>[3 x [4 x int]]</tt><br/>
629 <tt>[12 x [10 x float]]</tt><br/>
630 <tt>[2 x [3 x [4 x uint]]]</tt><br/>
633 3x4 array integer values.<br/>
634 12x10 array of single precision floating point values.<br/>
635 2x3x4 array of unsigned integer values.<br/>
641 <!-- _______________________________________________________________________ -->
642 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
643 <div class="doc_text">
645 <p>The function type can be thought of as a function signature. It
646 consists of a return type and a list of formal parameter types.
647 Function types are usually used to build virtual function tables
648 (which are structures of pointers to functions), for indirect function
649 calls, and when defining a function.</p>
651 The return type of a function type cannot be an aggregate type.
654 <pre> <returntype> (<parameter list>)<br></pre>
655 <p>Where '<tt><parameter list></tt>' is a comma-separated list of type
656 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
657 which indicates that the function takes a variable number of arguments.
658 Variable argument functions can access their arguments with the <a
659 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
661 <table class="layout">
664 <tt>int (int)</tt> <br/>
665 <tt>float (int, int *) *</tt><br/>
666 <tt>int (sbyte *, ...)</tt><br/>
669 function taking an <tt>int</tt>, returning an <tt>int</tt><br/>
670 <a href="#t_pointer">Pointer</a> to a function that takes an
671 <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
672 returning <tt>float</tt>.<br/>
673 A vararg function that takes at least one <a href="#t_pointer">pointer</a>
674 to <tt>sbyte</tt> (signed char in C), which returns an integer. This is
675 the signature for <tt>printf</tt> in LLVM.<br/>
681 <!-- _______________________________________________________________________ -->
682 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
683 <div class="doc_text">
685 <p>The structure type is used to represent a collection of data members
686 together in memory. The packing of the field types is defined to match
687 the ABI of the underlying processor. The elements of a structure may
688 be any type that has a size.</p>
689 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
690 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
691 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
694 <pre> { <type list> }<br></pre>
696 <table class="layout">
699 <tt>{ int, int, int }</tt><br/>
700 <tt>{ float, int (int) * }</tt><br/>
703 a triple of three <tt>int</tt> values<br/>
704 A pair, where the first element is a <tt>float</tt> and the second element
705 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
706 that takes an <tt>int</tt>, returning an <tt>int</tt>.<br/>
712 <!-- _______________________________________________________________________ -->
713 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
714 <div class="doc_text">
716 <p>As in many languages, the pointer type represents a pointer or
717 reference to another object, which must live in memory.</p>
719 <pre> <type> *<br></pre>
721 <table class="layout">
724 <tt>[4x int]*</tt><br/>
725 <tt>int (int *) *</tt><br/>
728 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
729 four <tt>int</tt> values<br/>
730 A <a href="#t_pointer">pointer</a> to a <a
731 href="#t_function">function</a> that takes an <tt>int</tt>, returning an
738 <!-- _______________________________________________________________________ -->
739 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
740 <div class="doc_text">
742 <p>A packed type is a simple derived type that represents a vector
743 of elements. Packed types are used when multiple primitive data
744 are operated in parallel using a single instruction (SIMD).
745 A packed type requires a size (number of
746 elements) and an underlying primitive data type. Packed types are
747 considered <a href="#t_firstclass">first class</a>.</p>
749 <pre> < <# elements> x <elementtype> ><br></pre>
750 <p>The number of elements is a constant integer value, elementtype may
751 be any integral or floating point type.</p>
753 <table class="layout">
756 <tt><4 x int></tt><br/>
757 <tt><8 x float></tt><br/>
758 <tt><2 x uint></tt><br/>
761 Packed vector of 4 integer values.<br/>
762 Packed vector of 8 floating-point values.<br/>
763 Packed vector of 2 unsigned integer values.<br/>
769 <!-- *********************************************************************** -->
770 <div class="doc_section"> <a name="constants">Constants</a> </div>
771 <!-- *********************************************************************** -->
773 <div class="doc_text">
775 <p>LLVM has several different basic types of constants. This section describes
776 them all and their syntax.</p>
780 <!-- ======================================================================= -->
781 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
783 <div class="doc_text">
786 <dt><b>Boolean constants</b></dt>
788 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
789 constants of the <tt><a href="#t_primitive">bool</a></tt> type.
792 <dt><b>Integer constants</b></dt>
794 <dd>Standard integers (such as '4') are constants of the <a
795 href="#t_integer">integer</a> type. Negative numbers may be used with signed
799 <dt><b>Floating point constants</b></dt>
801 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
802 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
803 notation. Floating point constants have an optional hexadecimal
804 notation (see below). Floating point constants must have a <a
805 href="#t_floating">floating point</a> type. </dd>
807 <dt><b>Null pointer constants</b></dt>
809 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant,
810 and must be of <a href="#t_pointer">pointer type</a>.</dd>
814 <p>The one non-intuitive notation for constants is the optional hexidecimal form
815 of floating point constants. For example, the form '<tt>double
816 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
817 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
818 (and the only time that they are generated by the disassembler) is when a
819 floating point constant must be emitted but it cannot be represented as a
820 decimal floating point number. For example, NaN's, infinities, and other
821 special values are represented in their IEEE hexadecimal format so that
822 assembly and disassembly do not cause any bits to change in the constants.</p>
826 <!-- ======================================================================= -->
827 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
830 <div class="doc_text">
833 <dt><b>Structure constants</b></dt>
835 <dd>Structure constants are represented with notation similar to structure
836 type definitions (a comma separated list of elements, surrounded by braces
837 (<tt>{}</tt>). For example: "<tt>{ int 4, float 17.0 }</tt>". Structure
838 constants must have <a href="#t_struct">structure type</a>, and the number and
839 types of elements must match those specified by the type.
842 <dt><b>Array constants</b></dt>
844 <dd>Array constants are represented with notation similar to array type
845 definitions (a comma separated list of elements, surrounded by square brackets
846 (<tt>[]</tt>). For example: "<tt>[ int 42, int 11, int 74 ]</tt>". Array
847 constants must have <a href="#t_array">array type</a>, and the number and
848 types of elements must match those specified by the type.
851 <dt><b>Packed constants</b></dt>
853 <dd>Packed constants are represented with notation similar to packed type
854 definitions (a comma separated list of elements, surrounded by
855 less-than/greater-than's (<tt><></tt>). For example: "<tt>< int 42,
856 int 11, int 74, int 100 ></tt>". Packed constants must have <a
857 href="#t_packed">packed type</a>, and the number and types of elements must
858 match those specified by the type.
861 <dt><b>Zero initialization</b></dt>
863 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
864 value to zero of <em>any</em> type, including scalar and aggregate types.
865 This is often used to avoid having to print large zero initializers (e.g. for
866 large arrays), and is always exactly equivalent to using explicit zero
873 <!-- ======================================================================= -->
874 <div class="doc_subsection">
875 <a name="globalconstants">Global Variable and Function Addresses</a>
878 <div class="doc_text">
880 <p>The addresses of <a href="#globalvars">global variables</a> and <a
881 href="#functionstructure">functions</a> are always implicitly valid (link-time)
882 constants. These constants explicitly referenced when the <a
883 href="#identifiers">identifier for the global</a> is used, and always have <a
884 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
890 %Z = global [2 x int*] [ int* %X, int* %Y ]
895 <!-- ======================================================================= -->
896 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a>
899 <div class="doc_text">
901 <p>The string '<tt>undef</tt>' is recognized as a filler that has no specified
902 value. Undefined values may be of any type, and be used anywhere a constant
905 <p>Undefined values are used to indicate the compiler that the program is well
906 defined no matter what value is used, giving it more freedom.</p>
910 <!-- ======================================================================= -->
911 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
914 <div class="doc_text">
916 <p>Constant expressions are used to allow expressions involving other constants
917 to be used as constants. Constant expressions may be of any <a
918 href="#t_firstclass">first class</a> type, and may involve any LLVM operation
919 that does not have side effects (e.g. load and call are not supported). The
920 following is the syntax for constant expressions:</p>
923 <dt><b><tt>cast ( CST to TYPE )</tt></b></dt>
925 <dd>Cast a constant to another type.</dd>
927 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
929 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
930 constants. As with the <a href="#i_getelementptr">getelementptr</a>
931 instruction, the index list may have zero or more indexes, which are required
932 to make sense for the type of "CSTPTR".</dd>
934 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
936 <dd>Perform the specied operation of the LHS and RHS constants. OPCODE may be
937 any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
938 binary</a> operations. The constraints on operands are the same as those for
939 the corresponding instruction (e.g. no bitwise operations on floating point
946 <!-- *********************************************************************** -->
947 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
948 <!-- *********************************************************************** -->
950 <div class="doc_text">
952 <p>The LLVM instruction set consists of several different
953 classifications of instructions: <a href="#terminators">terminator
954 instructions</a>, <a href="#binaryops">binary instructions</a>, <a
955 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
956 instructions</a>.</p>
960 <!-- ======================================================================= -->
961 <div class="doc_subsection"> <a name="terminators">Terminator
962 Instructions</a> </div>
964 <div class="doc_text">
966 <p>As mentioned <a href="#functionstructure">previously</a>, every
967 basic block in a program ends with a "Terminator" instruction, which
968 indicates which block should be executed after the current block is
969 finished. These terminator instructions typically yield a '<tt>void</tt>'
970 value: they produce control flow, not values (the one exception being
971 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
973 <p>There are five different terminator instructions: the '<a
974 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
975 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
976 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
977 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
978 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
982 <!-- _______________________________________________________________________ -->
983 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
984 Instruction</a> </div>
985 <div class="doc_text">
987 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
988 ret void <i>; Return from void function</i>
991 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
992 value) from a function, back to the caller.</p>
993 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
994 returns a value and then causes control flow, and one that just causes
995 control flow to occur.</p>
997 <p>The '<tt>ret</tt>' instruction may return any '<a
998 href="#t_firstclass">first class</a>' type. Notice that a function is
999 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1000 instruction inside of the function that returns a value that does not
1001 match the return type of the function.</p>
1003 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1004 returns back to the calling function's context. If the caller is a "<a
1005 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1006 the instruction after the call. If the caller was an "<a
1007 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1008 at the beginning "normal" of the destination block. If the instruction
1009 returns a value, that value shall set the call or invoke instruction's
1012 <pre> ret int 5 <i>; Return an integer value of 5</i>
1013 ret void <i>; Return from a void function</i>
1016 <!-- _______________________________________________________________________ -->
1017 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1018 <div class="doc_text">
1020 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1023 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1024 transfer to a different basic block in the current function. There are
1025 two forms of this instruction, corresponding to a conditional branch
1026 and an unconditional branch.</p>
1028 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1029 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
1030 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1031 value as a target.</p>
1033 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
1034 argument is evaluated. If the value is <tt>true</tt>, control flows
1035 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1036 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1038 <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
1039 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
1041 <!-- _______________________________________________________________________ -->
1042 <div class="doc_subsubsection">
1043 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1046 <div class="doc_text">
1050 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1055 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1056 several different places. It is a generalization of the '<tt>br</tt>'
1057 instruction, allowing a branch to occur to one of many possible
1063 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1064 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1065 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1066 table is not allowed to contain duplicate constant entries.</p>
1070 <p>The <tt>switch</tt> instruction specifies a table of values and
1071 destinations. When the '<tt>switch</tt>' instruction is executed, this
1072 table is searched for the given value. If the value is found, control flow is
1073 transfered to the corresponding destination; otherwise, control flow is
1074 transfered to the default destination.</p>
1076 <h5>Implementation:</h5>
1078 <p>Depending on properties of the target machine and the particular
1079 <tt>switch</tt> instruction, this instruction may be code generated in different
1080 ways. For example, it could be generated as a series of chained conditional
1081 branches or with a lookup table.</p>
1086 <i>; Emulate a conditional br instruction</i>
1087 %Val = <a href="#i_cast">cast</a> bool %value to int
1088 switch int %Val, label %truedest [int 0, label %falsedest ]
1090 <i>; Emulate an unconditional br instruction</i>
1091 switch uint 0, label %dest [ ]
1093 <i>; Implement a jump table:</i>
1094 switch uint %val, label %otherwise [ uint 0, label %onzero
1095 uint 1, label %onone
1096 uint 2, label %ontwo ]
1099 <!-- _______________________________________________________________________ -->
1100 <div class="doc_subsubsection"> <a name="i_invoke">'<tt>invoke</tt>'
1101 Instruction</a> </div>
1102 <div class="doc_text">
1104 <pre> <result> = invoke <ptr to function ty> %<function ptr val>(<function args>)<br> to label <normal label> except label <exception label><br></pre>
1106 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a
1107 specified function, with the possibility of control flow transfer to
1108 either the '<tt>normal</tt>' <tt>label</tt> label or the '<tt>exception</tt>'<tt>label</tt>.
1109 If the callee function returns with the "<tt><a href="#i_ret">ret</a></tt>"
1110 instruction, control flow will return to the "normal" label. If the
1111 callee (or any indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
1112 instruction, control is interrupted, and continued at the dynamically
1113 nearest "except" label.</p>
1115 <p>This instruction requires several arguments:</p>
1117 <li>'<tt>ptr to function ty</tt>': shall be the signature of the
1118 pointer to function value being invoked. In most cases, this is a
1119 direct function invocation, but indirect <tt>invoke</tt>s are just as
1120 possible, branching off an arbitrary pointer to function value. </li>
1121 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer
1122 to a function to be invoked. </li>
1123 <li>'<tt>function args</tt>': argument list whose types match the
1124 function signature argument types. If the function signature indicates
1125 the function accepts a variable number of arguments, the extra
1126 arguments can be specified. </li>
1127 <li>'<tt>normal label</tt>': the label reached when the called
1128 function executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1129 <li>'<tt>exception label</tt>': the label reached when a callee
1130 returns with the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1133 <p>This instruction is designed to operate as a standard '<tt><a
1134 href="#i_call">call</a></tt>' instruction in most regards. The
1135 primary difference is that it establishes an association with a label,
1136 which is used by the runtime library to unwind the stack.</p>
1137 <p>This instruction is used in languages with destructors to ensure
1138 that proper cleanup is performed in the case of either a <tt>longjmp</tt>
1139 or a thrown exception. Additionally, this is important for
1140 implementation of '<tt>catch</tt>' clauses in high-level languages that
1143 <pre> %retval = invoke int %Test(int 15)<br> to label %Continue<br> except label %TestCleanup <i>; {int}:retval set</i>
1148 <!-- _______________________________________________________________________ -->
1150 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1151 Instruction</a> </div>
1153 <div class="doc_text">
1162 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1163 at the first callee in the dynamic call stack which used an <a
1164 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1165 primarily used to implement exception handling.</p>
1169 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1170 immediately halt. The dynamic call stack is then searched for the first <a
1171 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1172 execution continues at the "exceptional" destination block specified by the
1173 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1174 dynamic call chain, undefined behavior results.</p>
1177 <!-- _______________________________________________________________________ -->
1179 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1180 Instruction</a> </div>
1182 <div class="doc_text">
1191 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1192 instruction is used to inform the optimizer that a particular portion of the
1193 code is not reachable. This can be used to indicate that the code after a
1194 no-return function cannot be reached, and other facts.</p>
1198 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1203 <!-- ======================================================================= -->
1204 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1205 <div class="doc_text">
1206 <p>Binary operators are used to do most of the computation in a
1207 program. They require two operands, execute an operation on them, and
1208 produce a single value. Although, that single value might represent
1209 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1210 The result value of a binary operator is not
1211 necessarily the same type as its operands.</p>
1212 <p>There are several different binary operators:</p>
1214 <!-- _______________________________________________________________________ -->
1215 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1216 Instruction</a> </div>
1217 <div class="doc_text">
1219 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1222 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1224 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1225 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1226 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1227 Both arguments must have identical types.</p>
1229 <p>The value produced is the integer or floating point sum of the two
1232 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
1235 <!-- _______________________________________________________________________ -->
1236 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1237 Instruction</a> </div>
1238 <div class="doc_text">
1240 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1243 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1245 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1246 instruction present in most other intermediate representations.</p>
1248 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1249 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1251 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1252 Both arguments must have identical types.</p>
1254 <p>The value produced is the integer or floating point difference of
1255 the two operands.</p>
1257 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
1258 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
1261 <!-- _______________________________________________________________________ -->
1262 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1263 Instruction</a> </div>
1264 <div class="doc_text">
1266 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1269 <p>The '<tt>mul</tt>' instruction returns the product of its two
1272 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1273 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1275 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1276 Both arguments must have identical types.</p>
1278 <p>The value produced is the integer or floating point product of the
1280 <p>There is no signed vs unsigned multiplication. The appropriate
1281 action is taken based on the type of the operand.</p>
1283 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
1286 <!-- _______________________________________________________________________ -->
1287 <div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
1288 Instruction</a> </div>
1289 <div class="doc_text">
1291 <pre> <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1294 <p>The '<tt>div</tt>' instruction returns the quotient of its two
1297 <p>The two arguments to the '<tt>div</tt>' instruction must be either <a
1298 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1300 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1301 Both arguments must have identical types.</p>
1303 <p>The value produced is the integer or floating point quotient of the
1306 <pre> <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
1309 <!-- _______________________________________________________________________ -->
1310 <div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
1311 Instruction</a> </div>
1312 <div class="doc_text">
1314 <pre> <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1317 <p>The '<tt>rem</tt>' instruction returns the remainder from the
1318 division of its two operands.</p>
1320 <p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1321 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1323 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1324 Both arguments must have identical types.</p>
1326 <p>This returns the <i>remainder</i> of a division (where the result
1327 has the same sign as the divisor), not the <i>modulus</i> (where the
1328 result has the same sign as the dividend) of a value. For more
1329 information about the difference, see: <a
1330 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1333 <pre> <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1336 <!-- _______________________________________________________________________ -->
1337 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1338 Instructions</a> </div>
1339 <div class="doc_text">
1341 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1342 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1343 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1344 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1345 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1346 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1349 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1350 value based on a comparison of their two operands.</p>
1352 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1353 be of <a href="#t_firstclass">first class</a> type (it is not possible
1354 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1355 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1358 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1359 value if both operands are equal.<br>
1360 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1361 value if both operands are unequal.<br>
1362 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1363 value if the first operand is less than the second operand.<br>
1364 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1365 value if the first operand is greater than the second operand.<br>
1366 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1367 value if the first operand is less than or equal to the second operand.<br>
1368 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1369 value if the first operand is greater than or equal to the second
1372 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1373 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1374 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1375 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1376 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1377 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1380 <!-- ======================================================================= -->
1381 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1382 Operations</a> </div>
1383 <div class="doc_text">
1384 <p>Bitwise binary operators are used to do various forms of
1385 bit-twiddling in a program. They are generally very efficient
1386 instructions, and can commonly be strength reduced from other
1387 instructions. They require two operands, execute an operation on them,
1388 and produce a single value. The resulting value of the bitwise binary
1389 operators is always the same type as its first operand.</p>
1391 <!-- _______________________________________________________________________ -->
1392 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1393 Instruction</a> </div>
1394 <div class="doc_text">
1396 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1399 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1400 its two operands.</p>
1402 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1403 href="#t_integral">integral</a> values. Both arguments must have
1404 identical types.</p>
1406 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1408 <div style="align: center">
1409 <table border="1" cellspacing="0" cellpadding="4">
1440 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1441 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1442 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1445 <!-- _______________________________________________________________________ -->
1446 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1447 <div class="doc_text">
1449 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1452 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1453 or of its two operands.</p>
1455 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1456 href="#t_integral">integral</a> values. Both arguments must have
1457 identical types.</p>
1459 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1461 <div style="align: center">
1462 <table border="1" cellspacing="0" cellpadding="4">
1493 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1494 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1495 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1498 <!-- _______________________________________________________________________ -->
1499 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1500 Instruction</a> </div>
1501 <div class="doc_text">
1503 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1506 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1507 or of its two operands. The <tt>xor</tt> is used to implement the
1508 "one's complement" operation, which is the "~" operator in C.</p>
1510 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1511 href="#t_integral">integral</a> values. Both arguments must have
1512 identical types.</p>
1514 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1516 <div style="align: center">
1517 <table border="1" cellspacing="0" cellpadding="4">
1549 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1550 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1551 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1552 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1555 <!-- _______________________________________________________________________ -->
1556 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1557 Instruction</a> </div>
1558 <div class="doc_text">
1560 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1563 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1564 the left a specified number of bits.</p>
1566 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
1567 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1570 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
1572 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1573 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1574 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1577 <!-- _______________________________________________________________________ -->
1578 <div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1579 Instruction</a> </div>
1580 <div class="doc_text">
1582 <pre> <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1585 <p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1586 the right a specified number of bits.</p>
1588 <p>The first argument to the '<tt>shr</tt>' instruction must be an <a
1589 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1592 <p>If the first argument is a <a href="#t_signed">signed</a> type, the
1593 most significant bit is duplicated in the newly free'd bit positions.
1594 If the first argument is unsigned, zero bits shall fill the empty
1597 <pre> <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1598 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1599 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1600 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1601 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1604 <!-- ======================================================================= -->
1605 <div class="doc_subsection"> <a name="memoryops">Memory Access
1606 Operations</a></div>
1607 <div class="doc_text">
1608 <p>A key design point of an SSA-based representation is how it
1609 represents memory. In LLVM, no memory locations are in SSA form, which
1610 makes things very simple. This section describes how to read, write,
1611 allocate and free memory in LLVM.</p>
1613 <!-- _______________________________________________________________________ -->
1614 <div class="doc_subsubsection"> <a name="i_malloc">'<tt>malloc</tt>'
1615 Instruction</a> </div>
1616 <div class="doc_text">
1618 <pre> <result> = malloc <type>, uint <NumElements> <i>; yields {type*}:result</i>
1619 <result> = malloc <type> <i>; yields {type*}:result</i>
1622 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
1623 heap and returns a pointer to it.</p>
1625 <p>The '<tt>malloc</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1626 bytes of memory from the operating system and returns a pointer of the
1627 appropriate type to the program. The second form of the instruction is
1628 a shorter version of the first instruction that defaults to allocating
1630 <p>'<tt>type</tt>' must be a sized type.</p>
1632 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
1633 a pointer is returned.</p>
1635 <pre> %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1638 href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1639 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1640 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1643 <!-- _______________________________________________________________________ -->
1644 <div class="doc_subsubsection"> <a name="i_free">'<tt>free</tt>'
1645 Instruction</a> </div>
1646 <div class="doc_text">
1648 <pre> free <type> <value> <i>; yields {void}</i>
1651 <p>The '<tt>free</tt>' instruction returns memory back to the unused
1652 memory heap, to be reallocated in the future.</p>
1655 <p>'<tt>value</tt>' shall be a pointer value that points to a value
1656 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
1659 <p>Access to the memory pointed to by the pointer is not longer defined
1660 after this instruction executes.</p>
1662 <pre> %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1663 free [4 x ubyte]* %array
1666 <!-- _______________________________________________________________________ -->
1667 <div class="doc_subsubsection"> <a name="i_alloca">'<tt>alloca</tt>'
1668 Instruction</a> </div>
1669 <div class="doc_text">
1671 <pre> <result> = alloca <type>, uint <NumElements> <i>; yields {type*}:result</i>
1672 <result> = alloca <type> <i>; yields {type*}:result</i>
1675 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
1676 stack frame of the procedure that is live until the current function
1677 returns to its caller.</p>
1679 <p>The the '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1680 bytes of memory on the runtime stack, returning a pointer of the
1681 appropriate type to the program. The second form of the instruction is
1682 a shorter version of the first that defaults to allocating one element.</p>
1683 <p>'<tt>type</tt>' may be any sized type.</p>
1685 <p>Memory is allocated, a pointer is returned. '<tt>alloca</tt>'d
1686 memory is automatically released when the function returns. The '<tt>alloca</tt>'
1687 instruction is commonly used to represent automatic variables that must
1688 have an address available. When the function returns (either with the <tt><a
1689 href="#i_ret">ret</a></tt> or <tt><a href="#i_invoke">invoke</a></tt>
1690 instructions), the memory is reclaimed.</p>
1692 <pre> %ptr = alloca int <i>; yields {int*}:ptr</i>
1693 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
1696 <!-- _______________________________________________________________________ -->
1697 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
1698 Instruction</a> </div>
1699 <div class="doc_text">
1701 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
1703 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
1705 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
1706 address to load from. The pointer must point to a <a
1707 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
1708 marked as <tt>volatile</tt> then the optimizer is not allowed to modify
1709 the number or order of execution of this <tt>load</tt> with other
1710 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
1713 <p>The location of memory pointed to is loaded.</p>
1715 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1717 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1718 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1721 <!-- _______________________________________________________________________ -->
1722 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
1723 Instruction</a> </div>
1725 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1726 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1729 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
1731 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
1732 to store and an address to store it into. The type of the '<tt><pointer></tt>'
1733 operand must be a pointer to the type of the '<tt><value></tt>'
1734 operand. If the <tt>store</tt> is marked as <tt>volatile</tt> then the
1735 optimizer is not allowed to modify the number or order of execution of
1736 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
1737 href="#i_store">store</a></tt> instructions.</p>
1739 <p>The contents of memory are updated to contain '<tt><value></tt>'
1740 at the location specified by the '<tt><pointer></tt>' operand.</p>
1742 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1744 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1745 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1747 <!-- _______________________________________________________________________ -->
1748 <div class="doc_subsubsection">
1749 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
1752 <div class="doc_text">
1755 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
1761 The '<tt>getelementptr</tt>' instruction is used to get the address of a
1762 subelement of an aggregate data structure.</p>
1766 <p>This instruction takes a list of integer constants that indicate what
1767 elements of the aggregate object to index to. The actual types of the arguments
1768 provided depend on the type of the first pointer argument. The
1769 '<tt>getelementptr</tt>' instruction is used to index down through the type
1770 levels of a structure. When indexing into a structure, only <tt>uint</tt>
1771 integer constants are allowed. When indexing into an array or pointer
1772 <tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
1774 <p>For example, let's consider a C code fragment and how it gets
1775 compiled to LLVM:</p>
1789 int *foo(struct ST *s) {
1790 return &s[1].Z.B[5][13];
1794 <p>The LLVM code generated by the GCC frontend is:</p>
1797 %RT = type { sbyte, [10 x [20 x int]], sbyte }
1798 %ST = type { int, double, %RT }
1802 int* %foo(%ST* %s) {
1804 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
1811 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
1812 on the pointer type that is being index into. <a href="#t_pointer">Pointer</a>
1813 and <a href="#t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
1814 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="#t_struct">structure</a>
1815 types require <tt>uint</tt> <b>constants</b>.</p>
1817 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
1818 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
1819 }</tt>' type, a structure. The second index indexes into the third element of
1820 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
1821 sbyte }</tt>' type, another structure. The third index indexes into the second
1822 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
1823 array. The two dimensions of the array are subscripted into, yielding an
1824 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction return a pointer
1825 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
1827 <p>Note that it is perfectly legal to index partially through a
1828 structure, returning a pointer to an inner element. Because of this,
1829 the LLVM code for the given testcase is equivalent to:</p>
1832 int* "foo"(%ST* %s) {
1833 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
1834 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
1835 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
1836 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
1837 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
1843 <i>; yields [12 x ubyte]*:aptr</i>
1844 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
1848 <!-- ======================================================================= -->
1849 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
1850 <div class="doc_text">
1851 <p>The instructions in this category are the "miscellaneous"
1852 instructions, which defy better classification.</p>
1854 <!-- _______________________________________________________________________ -->
1855 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
1856 Instruction</a> </div>
1857 <div class="doc_text">
1859 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
1861 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
1862 the SSA graph representing the function.</p>
1864 <p>The type of the incoming values are specified with the first type
1865 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
1866 as arguments, with one pair for each predecessor basic block of the
1867 current block. Only values of <a href="#t_firstclass">first class</a>
1868 type may be used as the value arguments to the PHI node. Only labels
1869 may be used as the label arguments.</p>
1870 <p>There must be no non-phi instructions between the start of a basic
1871 block and the PHI instructions: i.e. PHI instructions must be first in
1874 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
1875 value specified by the parameter, depending on which basic block we
1876 came from in the last <a href="#terminators">terminator</a> instruction.</p>
1878 <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>
1881 <!-- _______________________________________________________________________ -->
1882 <div class="doc_subsubsection">
1883 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
1886 <div class="doc_text">
1891 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
1897 The '<tt>cast</tt>' instruction is used as the primitive means to convert
1898 integers to floating point, change data type sizes, and break type safety (by
1906 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
1907 class value, and a type to cast it to, which must also be a <a
1908 href="#t_firstclass">first class</a> type.
1914 This instruction follows the C rules for explicit casts when determining how the
1915 data being cast must change to fit in its new container.
1919 When casting to bool, any value that would be considered true in the context of
1920 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
1921 all else are '<tt>false</tt>'.
1925 When extending an integral value from a type of one signness to another (for
1926 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
1927 <b>source</b> value is signed, and zero-extended if the source value is
1928 unsigned. <tt>bool</tt> values are always zero extended into either zero or
1935 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
1936 %Y = cast int 123 to bool <i>; yields bool:true</i>
1940 <!-- _______________________________________________________________________ -->
1941 <div class="doc_subsubsection">
1942 <a name="i_select">'<tt>select</tt>' Instruction</a>
1945 <div class="doc_text">
1950 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
1956 The '<tt>select</tt>' instruction is used to choose one value based on a
1957 condition, without branching.
1964 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.
1970 If the boolean condition evaluates to true, the instruction returns the first
1971 value argument, otherwise it returns the second value argument.
1977 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
1985 <!-- _______________________________________________________________________ -->
1986 <div class="doc_subsubsection"> <a name="i_call">'<tt>call</tt>'
1987 Instruction</a> </div>
1988 <div class="doc_text">
1990 <pre> <result> = call <ty>* <fnptrval>(<param list>)<br></pre>
1992 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
1994 <p>This instruction requires several arguments:</p>
1997 <p>'<tt>ty</tt>': shall be the signature of the pointer to function
1998 value being invoked. The argument types must match the types implied
1999 by this signature.</p>
2002 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a
2003 function to be invoked. In most cases, this is a direct function
2004 invocation, but indirect <tt>call</tt>s are just as possible,
2005 calling an arbitrary pointer to function values.</p>
2008 <p>'<tt>function args</tt>': argument list whose types match the
2009 function signature argument types. If the function signature
2010 indicates the function accepts a variable number of arguments, the
2011 extra arguments can be specified.</p>
2015 <p>The '<tt>call</tt>' instruction is used to cause control flow to
2016 transfer to a specified function, with its incoming arguments bound to
2017 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
2018 instruction in the called function, control flow continues with the
2019 instruction after the function call, and the return value of the
2020 function is bound to the result argument. This is a simpler case of
2021 the <a href="#i_invoke">invoke</a> instruction.</p>
2023 <pre> %retval = call int %test(int %argc)<br> call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);<br></pre>
2026 <!-- _______________________________________________________________________ -->
2027 <div class="doc_subsubsection">
2028 <a name="i_vanext">'<tt>vanext</tt>' Instruction</a>
2031 <div class="doc_text">
2036 <resultarglist> = vanext <va_list> <arglist>, <argty>
2041 <p>The '<tt>vanext</tt>' instruction is used to access arguments passed
2042 through the "variable argument" area of a function call. It is used to
2043 implement the <tt>va_arg</tt> macro in C.</p>
2047 <p>This instruction takes a <tt>va_list</tt> value and the type of the
2048 argument. It returns another <tt>va_list</tt>. The actual type of
2049 <tt>va_list</tt> may be defined differently for different targets. Most targets
2050 use a <tt>va_list</tt> type of <tt>sbyte*</tt> or some other pointer type.</p>
2054 <p>The '<tt>vanext</tt>' instruction advances the specified <tt>va_list</tt>
2055 past an argument of the specified type. In conjunction with the <a
2056 href="#i_vaarg"><tt>vaarg</tt></a> instruction, it is used to implement
2057 the <tt>va_arg</tt> macro available in C. For more information, see
2058 the variable argument handling <a href="#int_varargs">Intrinsic
2061 <p>It is legal for this instruction to be called in a function which
2062 does not take a variable number of arguments, for example, the <tt>vfprintf</tt>
2065 <p><tt>vanext</tt> is an LLVM instruction instead of an <a
2066 href="#intrinsics">intrinsic function</a> because it takes a type as an
2067 argument. The type refers to the current argument in the <tt>va_list</tt>, it
2068 tells the compiler how far on the stack it needs to advance to find the next
2073 <p>See the <a href="#int_varargs">variable argument processing</a>
2078 <!-- _______________________________________________________________________ -->
2079 <div class="doc_subsubsection">
2080 <a name="i_vaarg">'<tt>vaarg</tt>' Instruction</a>
2083 <div class="doc_text">
2088 <resultval> = vaarg <va_list> <arglist>, <argty>
2093 <p>The '<tt>vaarg</tt>' instruction is used to access arguments passed through
2094 the "variable argument" area of a function call. It is used to implement the
2095 <tt>va_arg</tt> macro in C.</p>
2099 <p>This instruction takes a <tt>va_list</tt> value and the type of the
2100 argument. It returns a value of the specified argument type. Again, the actual
2101 type of <tt>va_list</tt> is target specific.</p>
2105 <p>The '<tt>vaarg</tt>' instruction loads an argument of the specified type from
2106 the specified <tt>va_list</tt>. In conjunction with the <a
2107 href="#i_vanext"><tt>vanext</tt></a> instruction, it is used to implement the
2108 <tt>va_arg</tt> macro available in C. For more information, see the variable
2109 argument handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
2111 <p>It is legal for this instruction to be called in a function which does not
2112 take a variable number of arguments, for example, the <tt>vfprintf</tt>
2115 <p><tt>vaarg</tt> is an LLVM instruction instead of an <a
2116 href="#intrinsics">intrinsic function</a> because it takes an type as an
2121 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
2125 <!-- *********************************************************************** -->
2126 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
2127 <!-- *********************************************************************** -->
2129 <div class="doc_text">
2131 <p>LLVM supports the notion of an "intrinsic function". These functions have
2132 well known names and semantics, and are required to follow certain
2133 restrictions. Overall, these instructions represent an extension mechanism for
2134 the LLVM language that does not require changing all of the transformations in
2135 LLVM to add to the language (or the bytecode reader/writer, the parser,
2138 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix, this
2139 prefix is reserved in LLVM for intrinsic names, thus functions may not be named
2140 this. Intrinsic functions must always be external functions: you cannot define
2141 the body of intrinsic functions. Intrinsic functions may only be used in call
2142 or invoke instructions: it is illegal to take the address of an intrinsic
2143 function. Additionally, because intrinsic functions are part of the LLVM
2144 language, it is required that they all be documented here if any are added.</p>
2148 Adding an intrinsic to LLVM is straight-forward if it is possible to express the
2149 concept in LLVM directly (ie, code generator support is not _required_). To do
2150 this, extend the default implementation of the IntrinsicLowering class to handle
2151 the intrinsic. Code generators use this class to lower intrinsics they do not
2152 understand to raw LLVM instructions that they do.
2157 <!-- ======================================================================= -->
2158 <div class="doc_subsection">
2159 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
2162 <div class="doc_text">
2164 <p>Variable argument support is defined in LLVM with the <a
2165 href="#i_vanext"><tt>vanext</tt></a> instruction and these three
2166 intrinsic functions. These functions are related to the similarly
2167 named macros defined in the <tt><stdarg.h></tt> header file.</p>
2169 <p>All of these functions operate on arguments that use a
2170 target-specific value type "<tt>va_list</tt>". The LLVM assembly
2171 language reference manual does not define what this type is, so all
2172 transformations should be prepared to handle intrinsics with any type
2175 <p>This example shows how the <a href="#i_vanext"><tt>vanext</tt></a>
2176 instruction and the variable argument handling intrinsic functions are
2180 int %test(int %X, ...) {
2181 ; Initialize variable argument processing
2182 %ap = call sbyte* %<a href="#i_va_start">llvm.va_start</a>()
2184 ; Read a single integer argument
2185 %tmp = vaarg sbyte* %ap, int
2187 ; Advance to the next argument
2188 %ap2 = vanext sbyte* %ap, int
2190 ; Demonstrate usage of llvm.va_copy and llvm.va_end
2191 %aq = call sbyte* %<a href="#i_va_copy">llvm.va_copy</a>(sbyte* %ap2)
2192 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %aq)
2194 ; Stop processing of arguments.
2195 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %ap2)
2201 <!-- _______________________________________________________________________ -->
2202 <div class="doc_subsubsection">
2203 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
2207 <div class="doc_text">
2209 <pre> call <va_list> ()* %llvm.va_start()<br></pre>
2211 <p>The '<tt>llvm.va_start</tt>' intrinsic returns a new <tt><arglist></tt>
2212 for subsequent use by the variable argument intrinsics.</p>
2214 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
2215 macro available in C. In a target-dependent way, it initializes and
2216 returns a <tt>va_list</tt> element, so that the next <tt>vaarg</tt>
2217 will produce the first variable argument passed to the function. Unlike
2218 the C <tt>va_start</tt> macro, this intrinsic does not need to know the
2219 last argument of the function, the compiler can figure that out.</p>
2220 <p>Note that this intrinsic function is only legal to be called from
2221 within the body of a variable argument function.</p>
2224 <!-- _______________________________________________________________________ -->
2225 <div class="doc_subsubsection">
2226 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
2229 <div class="doc_text">
2231 <pre> call void (<va_list>)* %llvm.va_end(<va_list> <arglist>)<br></pre>
2233 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
2234 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
2235 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
2237 <p>The argument is a <tt>va_list</tt> to destroy.</p>
2239 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
2240 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
2241 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
2242 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
2243 with calls to <tt>llvm.va_end</tt>.</p>
2246 <!-- _______________________________________________________________________ -->
2247 <div class="doc_subsubsection">
2248 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
2251 <div class="doc_text">
2256 call <va_list> (<va_list>)* %llvm.va_copy(<va_list> <destarglist>)
2261 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
2262 from the source argument list to the destination argument list.</p>
2266 <p>The argument is the <tt>va_list</tt> to copy.</p>
2270 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
2271 macro available in C. In a target-dependent way, it copies the source
2272 <tt>va_list</tt> element into the returned list. This intrinsic is necessary
2273 because the <tt><a href="#i_va_start">llvm.va_start</a></tt> intrinsic may be
2274 arbitrarily complex and require memory allocation, for example.</p>
2278 <!-- ======================================================================= -->
2279 <div class="doc_subsection">
2280 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
2283 <div class="doc_text">
2286 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
2287 Collection</a> requires the implementation and generation of these intrinsics.
2288 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
2289 stack</a>, as well as garbage collector implementations that require <a
2290 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
2291 Front-ends for type-safe garbage collected languages should generate these
2292 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
2293 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
2297 <!-- _______________________________________________________________________ -->
2298 <div class="doc_subsubsection">
2299 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
2302 <div class="doc_text">
2307 call void (<ty>**, <ty2>*)* %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
2312 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existance of a GC root to
2313 the code generator, and allows some metadata to be associated with it.</p>
2317 <p>The first argument specifies the address of a stack object that contains the
2318 root pointer. The second pointer (which must be either a constant or a global
2319 value address) contains the meta-data to be associated with the root.</p>
2323 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
2324 location. At compile-time, the code generator generates information to allow
2325 the runtime to find the pointer at GC safe points.
2331 <!-- _______________________________________________________________________ -->
2332 <div class="doc_subsubsection">
2333 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
2336 <div class="doc_text">
2341 call sbyte* (sbyte**)* %llvm.gcread(sbyte** %Ptr)
2346 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
2347 locations, allowing garbage collector implementations that require read
2352 <p>The argument is the address to read from, which should be an address
2353 allocated from the garbage collector.</p>
2357 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
2358 instruction, but may be replaced with substantially more complex code by the
2359 garbage collector runtime, as needed.</p>
2364 <!-- _______________________________________________________________________ -->
2365 <div class="doc_subsubsection">
2366 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
2369 <div class="doc_text">
2374 call void (sbyte*, sbyte**)* %llvm.gcwrite(sbyte* %P1, sbyte** %P2)
2379 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
2380 locations, allowing garbage collector implementations that require write
2381 barriers (such as generational or reference counting collectors).</p>
2385 <p>The first argument is the reference to store, and the second is the heap
2386 location to store to.</p>
2390 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
2391 instruction, but may be replaced with substantially more complex code by the
2392 garbage collector runtime, as needed.</p>
2398 <!-- ======================================================================= -->
2399 <div class="doc_subsection">
2400 <a name="int_codegen">Code Generator Intrinsics</a>
2403 <div class="doc_text">
2405 These intrinsics are provided by LLVM to expose special features that may only
2406 be implemented with code generator support.
2411 <!-- _______________________________________________________________________ -->
2412 <div class="doc_subsubsection">
2413 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
2416 <div class="doc_text">
2420 call void* ()* %llvm.returnaddress(uint <level>)
2426 The '<tt>llvm.returnaddress</tt>' intrinsic returns a target-specific value
2427 indicating the return address of the current function or one of its callers.
2433 The argument to this intrinsic indicates which function to return the address
2434 for. Zero indicates the calling function, one indicates its caller, etc. The
2435 argument is <b>required</b> to be a constant integer value.
2441 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
2442 the return address of the specified call frame, or zero if it cannot be
2443 identified. The value returned by this intrinsic is likely to be incorrect or 0
2444 for arguments other than zero, so it should only be used for debugging purposes.
2448 Note that calling this intrinsic does not prevent function inlining or other
2449 aggressive transformations, so the value returned may not that of the obvious
2450 source-language caller.
2455 <!-- _______________________________________________________________________ -->
2456 <div class="doc_subsubsection">
2457 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
2460 <div class="doc_text">
2464 call void* ()* %llvm.frameaddress(uint <level>)
2470 The '<tt>llvm.frameaddress</tt>' intrinsic returns the target-specific frame
2471 pointer value for the specified stack frame.
2477 The argument to this intrinsic indicates which function to return the frame
2478 pointer for. Zero indicates the calling function, one indicates its caller,
2479 etc. The argument is <b>required</b> to be a constant integer value.
2485 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
2486 the frame address of the specified call frame, or zero if it cannot be
2487 identified. The value returned by this intrinsic is likely to be incorrect or 0
2488 for arguments other than zero, so it should only be used for debugging purposes.
2492 Note that calling this intrinsic does not prevent function inlining or other
2493 aggressive transformations, so the value returned may not that of the obvious
2494 source-language caller.
2498 <!-- ======================================================================= -->
2499 <div class="doc_subsection">
2500 <a name="int_os">Operating System Intrinsics</a>
2503 <div class="doc_text">
2505 These intrinsics are provided by LLVM to support the implementation of
2506 operating system level code.
2511 <!-- _______________________________________________________________________ -->
2512 <div class="doc_subsubsection">
2513 <a name="i_readport">'<tt>llvm.readport</tt>' Intrinsic</a>
2516 <div class="doc_text">
2520 call <integer type> (<integer type>)* %llvm.readport (<integer type> <address>)
2526 The '<tt>llvm.readport</tt>' intrinsic reads data from the specified hardware
2533 The argument to this intrinsic indicates the hardware I/O address from which
2534 to read the data. The address is in the hardware I/O address namespace (as
2535 opposed to being a memory location for memory mapped I/O).
2541 The '<tt>llvm.readport</tt>' intrinsic reads data from the hardware I/O port
2542 specified by <i>address</i> and returns the value. The address and return
2543 value must be integers, but the size is dependent upon the platform upon which
2544 the program is code generated. For example, on x86, the address must be an
2545 unsigned 16 bit value, and the return value must be 8, 16, or 32 bits.
2550 <!-- _______________________________________________________________________ -->
2551 <div class="doc_subsubsection">
2552 <a name="i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a>
2555 <div class="doc_text">
2559 call void (<integer type>, <integer type>)*
2560 %llvm.writeport (<integer type> <value>,
2561 <integer type> <address>)
2567 The '<tt>llvm.writeport</tt>' intrinsic writes data to the specified hardware
2574 The first argument is the value to write to the I/O port.
2578 The second argument indicates the hardware I/O address to which data should be
2579 written. The address is in the hardware I/O address namespace (as opposed to
2580 being a memory location for memory mapped I/O).
2586 The '<tt>llvm.writeport</tt>' intrinsic writes <i>value</i> to the I/O port
2587 specified by <i>address</i>. The address and value must be integers, but the
2588 size is dependent upon the platform upon which the program is code generated.
2589 For example, on x86, the address must be an unsigned 16 bit value, and the
2590 value written must be 8, 16, or 32 bits in length.
2595 <!-- _______________________________________________________________________ -->
2596 <div class="doc_subsubsection">
2597 <a name="i_readio">'<tt>llvm.readio</tt>' Intrinsic</a>
2600 <div class="doc_text">
2604 call <result> (<ty>*)* %llvm.readio (<ty> * <pointer>)
2610 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2617 The argument to this intrinsic is a pointer indicating the memory address from
2618 which to read the data. The data must be a
2619 <a href="#t_firstclass">first class</a> type.
2625 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2626 location specified by <i>pointer</i> and returns the value. The argument must
2627 be a pointer, and the return value must be a
2628 <a href="#t_firstclass">first class</a> type. However, certain architectures
2629 may not support I/O on all first class types. For example, 32 bit processors
2630 may only support I/O on data types that are 32 bits or less.
2634 This intrinsic enforces an in-order memory model for llvm.readio and
2635 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2636 scheduled processors may execute loads and stores out of order, re-ordering at
2637 run time accesses to memory mapped I/O registers. Using these intrinsics
2638 ensures that accesses to memory mapped I/O registers occur in program order.
2643 <!-- _______________________________________________________________________ -->
2644 <div class="doc_subsubsection">
2645 <a name="i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a>
2648 <div class="doc_text">
2652 call void (<ty1>, <ty2>*)* %llvm.writeio (<ty1> <value>, <ty2> * <pointer>)
2658 The '<tt>llvm.writeio</tt>' intrinsic writes data to the specified memory
2665 The first argument is the value to write to the memory mapped I/O location.
2666 The second argument is a pointer indicating the memory address to which the
2667 data should be written.
2673 The '<tt>llvm.writeio</tt>' intrinsic writes <i>value</i> to the memory mapped
2674 I/O address specified by <i>pointer</i>. The value must be a
2675 <a href="#t_firstclass">first class</a> type. However, certain architectures
2676 may not support I/O on all first class types. For example, 32 bit processors
2677 may only support I/O on data types that are 32 bits or less.
2681 This intrinsic enforces an in-order memory model for llvm.readio and
2682 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2683 scheduled processors may execute loads and stores out of order, re-ordering at
2684 run time accesses to memory mapped I/O registers. Using these intrinsics
2685 ensures that accesses to memory mapped I/O registers occur in program order.
2690 <!-- ======================================================================= -->
2691 <div class="doc_subsection">
2692 <a name="int_libc">Standard C Library Intrinsics</a>
2695 <div class="doc_text">
2697 LLVM provides intrinsics for a few important standard C library functions.
2698 These intrinsics allow source-language front-ends to pass information about the
2699 alignment of the pointer arguments to the code generator, providing opportunity
2700 for more efficient code generation.
2705 <!-- _______________________________________________________________________ -->
2706 <div class="doc_subsubsection">
2707 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
2710 <div class="doc_text">
2714 call void (sbyte*, sbyte*, uint, uint)* %llvm.memcpy(sbyte* <dest>, sbyte* <src>,
2715 uint <len>, uint <align>)
2721 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2722 location to the destination location.
2726 Note that, unlike the standard libc function, the <tt>llvm.memcpy</tt> intrinsic
2727 does not return a value, and takes an extra alignment argument.
2733 The first argument is a pointer to the destination, the second is a pointer to
2734 the source. The third argument is an (arbitrarily sized) integer argument
2735 specifying the number of bytes to copy, and the fourth argument is the alignment
2736 of the source and destination locations.
2740 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2741 the caller guarantees that the size of the copy is a multiple of the alignment
2742 and that both the source and destination pointers are aligned to that boundary.
2748 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2749 location to the destination location, which are not allowed to overlap. It
2750 copies "len" bytes of memory over. If the argument is known to be aligned to
2751 some boundary, this can be specified as the fourth argument, otherwise it should
2757 <!-- _______________________________________________________________________ -->
2758 <div class="doc_subsubsection">
2759 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
2762 <div class="doc_text">
2766 call void (sbyte*, sbyte*, uint, uint)* %llvm.memmove(sbyte* <dest>, sbyte* <src>,
2767 uint <len>, uint <align>)
2773 The '<tt>llvm.memmove</tt>' intrinsic moves a block of memory from the source
2774 location to the destination location. It is similar to the '<tt>llvm.memcpy</tt>'
2775 intrinsic but allows the two memory locations to overlap.
2779 Note that, unlike the standard libc function, the <tt>llvm.memmove</tt> intrinsic
2780 does not return a value, and takes an extra alignment argument.
2786 The first argument is a pointer to the destination, the second is a pointer to
2787 the source. The third argument is an (arbitrarily sized) integer argument
2788 specifying the number of bytes to copy, and the fourth argument is the alignment
2789 of the source and destination locations.
2793 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2794 the caller guarantees that the size of the copy is a multiple of the alignment
2795 and that both the source and destination pointers are aligned to that boundary.
2801 The '<tt>llvm.memmove</tt>' intrinsic copies a block of memory from the source
2802 location to the destination location, which may overlap. It
2803 copies "len" bytes of memory over. If the argument is known to be aligned to
2804 some boundary, this can be specified as the fourth argument, otherwise it should
2810 <!-- _______________________________________________________________________ -->
2811 <div class="doc_subsubsection">
2812 <a name="i_memset">'<tt>llvm.memset</tt>' Intrinsic</a>
2815 <div class="doc_text">
2819 call void (sbyte*, ubyte, uint, uint)* %llvm.memset(sbyte* <dest>, ubyte <val>,
2820 uint <len>, uint <align>)
2826 The '<tt>llvm.memset</tt>' intrinsic fills a block of memory with a particular
2831 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
2832 does not return a value, and takes an extra alignment argument.
2838 The first argument is a pointer to the destination to fill, the second is the
2839 byte value to fill it with, the third argument is an (arbitrarily sized) integer
2840 argument specifying the number of bytes to fill, and the fourth argument is the
2841 known alignment of destination location.
2845 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2846 the caller guarantees that the size of the copy is a multiple of the alignment
2847 and that the destination pointer is aligned to that boundary.
2853 The '<tt>llvm.memset</tt>' intrinsic fills "len" bytes of memory starting at the
2854 destination location. If the argument is known to be aligned to some boundary,
2855 this can be specified as the fourth argument, otherwise it should be set to 0 or
2861 <!-- _______________________________________________________________________ -->
2862 <div class="doc_subsubsection">
2863 <a name="i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a>
2866 <div class="doc_text">
2870 call bool (<float or double>, <float or double>)* %llvm.isunordered(<float or double> Val1,
2871 <float or double> Val2)
2877 The '<tt>llvm.isunordered</tt>' intrinsic returns true if either or both of the
2878 specified floating point values is a NAN.
2884 The arguments are floating point numbers of the same type.
2890 If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
2898 <!-- ======================================================================= -->
2899 <div class="doc_subsection">
2900 <a name="int_debugger">Debugger Intrinsics</a>
2903 <div class="doc_text">
2905 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
2906 are described in the <a
2907 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
2908 Debugging</a> document.
2913 <!-- *********************************************************************** -->
2916 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2917 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
2918 <a href="http://validator.w3.org/check/referer"><img
2919 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
2921 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
2922 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
2923 Last modified: $Date$