1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
5 <title>LLVM Assembly Language Reference Manual</title>
6 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
7 <meta name="author" content="Chris Lattner">
8 <meta name="description"
9 content="LLVM Assembly Language Reference Manual.">
10 <link rel="stylesheet" href="llvm.css" type="text/css">
15 <div class="doc_title"> LLVM Language Reference Manual </div>
17 <li><a href="#abstract">Abstract</a></li>
18 <li><a href="#introduction">Introduction</a></li>
19 <li><a href="#identifiers">Identifiers</a></li>
20 <li><a href="#highlevel">High Level Structure</a>
22 <li><a href="#modulestructure">Module Structure</a></li>
23 <li><a href="#linkage">Linkage Types</a></li>
24 <li><a href="#globalvars">Global Variables</a></li>
25 <li><a href="#functionstructure">Function Structure</a></li>
28 <li><a href="#typesystem">Type System</a>
30 <li><a href="#t_primitive">Primitive Types</a>
32 <li><a href="#t_classifications">Type Classifications</a></li>
35 <li><a href="#t_derived">Derived Types</a>
37 <li><a href="#t_array">Array Type</a></li>
38 <li><a href="#t_function">Function Type</a></li>
39 <li><a href="#t_pointer">Pointer Type</a></li>
40 <li><a href="#t_struct">Structure Type</a></li>
41 <li><a href="#t_packed">Packed Type</a></li>
46 <li><a href="#constants">Constants</a>
48 <li><a href="#simpleconstants">Simple Constants</a>
49 <li><a href="#aggregateconstants">Aggregate Constants</a>
50 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
51 <li><a href="#undefvalues">Undefined Values</a>
52 <li><a href="#constantexprs">Constant Expressions</a>
55 <li><a href="#instref">Instruction Reference</a>
57 <li><a href="#terminators">Terminator Instructions</a>
59 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
60 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
61 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
62 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
63 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
64 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
67 <li><a href="#binaryops">Binary Operations</a>
69 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
70 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
71 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
72 <li><a href="#i_div">'<tt>div</tt>' Instruction</a></li>
73 <li><a href="#i_rem">'<tt>rem</tt>' Instruction</a></li>
74 <li><a href="#i_setcc">'<tt>set<i>cc</i></tt>' Instructions</a></li>
77 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
79 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
80 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
81 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
82 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
83 <li><a href="#i_shr">'<tt>shr</tt>' Instruction</a></li>
86 <li><a href="#memoryops">Memory Access Operations</a>
88 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
89 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
90 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
91 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
92 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
93 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
96 <li><a href="#otherops">Other Operations</a>
98 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
99 <li><a href="#i_cast">'<tt>cast .. to</tt>' Instruction</a></li>
100 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
101 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
102 <li><a href="#i_vanext">'<tt>vanext</tt>' Instruction</a></li>
103 <li><a href="#i_vaarg">'<tt>vaarg</tt>' Instruction</a></li>
108 <li><a href="#intrinsics">Intrinsic Functions</a>
110 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
112 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
113 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
114 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
117 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
119 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
120 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
121 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
124 <li><a href="#int_codegen">Code Generator Intrinsics</a>
126 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
127 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
128 <li><a href="#i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
131 <li><a href="#int_os">Operating System Intrinsics</a>
133 <li><a href="#i_readport">'<tt>llvm.readport</tt>' Intrinsic</a></li>
134 <li><a href="#i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a></li>
135 <li><a href="#i_readio">'<tt>llvm.readio</tt>' Intrinsic</a></li>
136 <li><a href="#i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a></li>
138 <li><a href="#int_libc">Standard C Library Intrinsics</a>
140 <li><a href="#i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a></li>
141 <li><a href="#i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a></li>
142 <li><a href="#i_memset">'<tt>llvm.memset</tt>' Intrinsic</a></li>
143 <li><a href="#i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a></li>
146 <li><a href="#int_debugger">Debugger intrinsics</a></li>
151 <div class="doc_author">
152 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
153 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
156 <!-- *********************************************************************** -->
157 <div class="doc_section"> <a name="abstract">Abstract </a></div>
158 <!-- *********************************************************************** -->
160 <div class="doc_text">
161 <p>This document is a reference manual for the LLVM assembly language.
162 LLVM is an SSA based representation that provides type safety,
163 low-level operations, flexibility, and the capability of representing
164 'all' high-level languages cleanly. It is the common code
165 representation used throughout all phases of the LLVM compilation
169 <!-- *********************************************************************** -->
170 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
171 <!-- *********************************************************************** -->
173 <div class="doc_text">
175 <p>The LLVM code representation is designed to be used in three
176 different forms: as an in-memory compiler IR, as an on-disk bytecode
177 representation (suitable for fast loading by a Just-In-Time compiler),
178 and as a human readable assembly language representation. This allows
179 LLVM to provide a powerful intermediate representation for efficient
180 compiler transformations and analysis, while providing a natural means
181 to debug and visualize the transformations. The three different forms
182 of LLVM are all equivalent. This document describes the human readable
183 representation and notation.</p>
185 <p>The LLVM representation aims to be a light-weight and low-level
186 while being expressive, typed, and extensible at the same time. It
187 aims to be a "universal IR" of sorts, by being at a low enough level
188 that high-level ideas may be cleanly mapped to it (similar to how
189 microprocessors are "universal IR's", allowing many source languages to
190 be mapped to them). By providing type information, LLVM can be used as
191 the target of optimizations: for example, through pointer analysis, it
192 can be proven that a C automatic variable is never accessed outside of
193 the current function... allowing it to be promoted to a simple SSA
194 value instead of a memory location.</p>
198 <!-- _______________________________________________________________________ -->
199 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
201 <div class="doc_text">
203 <p>It is important to note that this document describes 'well formed'
204 LLVM assembly language. There is a difference between what the parser
205 accepts and what is considered 'well formed'. For example, the
206 following instruction is syntactically okay, but not well formed:</p>
209 %x = <a href="#i_add">add</a> int 1, %x
212 <p>...because the definition of <tt>%x</tt> does not dominate all of
213 its uses. The LLVM infrastructure provides a verification pass that may
214 be used to verify that an LLVM module is well formed. This pass is
215 automatically run by the parser after parsing input assembly, and by
216 the optimizer before it outputs bytecode. The violations pointed out
217 by the verifier pass indicate bugs in transformation passes or input to
220 <!-- Describe the typesetting conventions here. --> </div>
222 <!-- *********************************************************************** -->
223 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
224 <!-- *********************************************************************** -->
226 <div class="doc_text">
228 <p>LLVM uses three different forms of identifiers, for different
232 <li>Named values are represented as a string of characters with a '%' prefix.
233 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
234 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
235 Identifiers which require other characters in their names can be surrounded
236 with quotes. In this way, anything except a <tt>"</tt> character can be used
239 <li>Unnamed values are represented as an unsigned numeric value with a '%'
240 prefix. For example, %12, %2, %44.</li>
242 <li>Constants, which are described in a <a href="#constants">section about
243 constants</a>, below.</li>
246 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
247 don't need to worry about name clashes with reserved words, and the set of
248 reserved words may be expanded in the future without penalty. Additionally,
249 unnamed identifiers allow a compiler to quickly come up with a temporary
250 variable without having to avoid symbol table conflicts.</p>
252 <p>Reserved words in LLVM are very similar to reserved words in other
253 languages. There are keywords for different opcodes ('<tt><a
254 href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
255 href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
256 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>', etc...),
257 and others. These reserved words cannot conflict with variable names, because
258 none of them start with a '%' character.</p>
260 <p>Here is an example of LLVM code to multiply the integer variable
261 '<tt>%X</tt>' by 8:</p>
266 %result = <a href="#i_mul">mul</a> uint %X, 8
269 <p>After strength reduction:</p>
272 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
275 <p>And the hard way:</p>
278 <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
279 <a href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
280 %result = <a href="#i_add">add</a> uint %1, %1
283 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
284 important lexical features of LLVM:</p>
288 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
291 <li>Unnamed temporaries are created when the result of a computation is not
292 assigned to a named value.</li>
294 <li>Unnamed temporaries are numbered sequentially</li>
298 <p>...and it also show a convention that we follow in this document. When
299 demonstrating instructions, we will follow an instruction with a comment that
300 defines the type and name of value produced. Comments are shown in italic
305 <!-- *********************************************************************** -->
306 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
307 <!-- *********************************************************************** -->
309 <!-- ======================================================================= -->
310 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
313 <div class="doc_text">
315 <p>LLVM programs are composed of "Module"s, each of which is a
316 translation unit of the input programs. Each module consists of
317 functions, global variables, and symbol table entries. Modules may be
318 combined together with the LLVM linker, which merges function (and
319 global variable) definitions, resolves forward declarations, and merges
320 symbol table entries. Here is an example of the "hello world" module:</p>
322 <pre><i>; Declare the string constant as a global constant...</i>
323 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
324 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
326 <i>; External declaration of the puts function</i>
327 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
329 <i>; Definition of main function</i>
330 int %main() { <i>; int()* </i>
331 <i>; Convert [13x sbyte]* to sbyte *...</i>
333 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
335 <i>; Call puts function to write out the string to stdout...</i>
337 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
339 href="#i_ret">ret</a> int 0<br>}<br></pre>
341 <p>This example is made up of a <a href="#globalvars">global variable</a>
342 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
343 function, and a <a href="#functionstructure">function definition</a>
344 for "<tt>main</tt>".</p>
346 <p>In general, a module is made up of a list of global values,
347 where both functions and global variables are global values. Global values are
348 represented by a pointer to a memory location (in this case, a pointer to an
349 array of char, and a pointer to a function), and have one of the following <a
350 href="#linkage">linkage types</a>.</p>
354 <!-- ======================================================================= -->
355 <div class="doc_subsection">
356 <a name="linkage">Linkage Types</a>
359 <div class="doc_text">
362 All Global Variables and Functions have one of the following types of linkage:
367 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
369 <dd>Global values with internal linkage are only directly accessible by
370 objects in the current module. In particular, linking code into a module with
371 an internal global value may cause the internal to be renamed as necessary to
372 avoid collisions. Because the symbol is internal to the module, all
373 references can be updated. This corresponds to the notion of the
374 '<tt>static</tt>' keyword in C, or the idea of "anonymous namespaces" in C++.
377 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
379 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
380 the twist that linking together two modules defining the same
381 <tt>linkonce</tt> globals will cause one of the globals to be discarded. This
382 is typically used to implement inline functions. Unreferenced
383 <tt>linkonce</tt> globals are allowed to be discarded.
386 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
388 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
389 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
390 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.
393 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
395 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
396 pointer to array type. When two global variables with appending linkage are
397 linked together, the two global arrays are appended together. This is the
398 LLVM, typesafe, equivalent of having the system linker append together
399 "sections" with identical names when .o files are linked.
402 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
404 <dd>If none of the above identifiers are used, the global is externally
405 visible, meaning that it participates in linkage and can be used to resolve
406 external symbol references.
410 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
411 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
412 variable and was linked with this one, one of the two would be renamed,
413 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
414 external (i.e., lacking any linkage declarations), they are accessible
415 outside of the current module. It is illegal for a function <i>declaration</i>
416 to have any linkage type other than "externally visible".</a></p>
420 <!-- ======================================================================= -->
421 <div class="doc_subsection">
422 <a name="globalvars">Global Variables</a>
425 <div class="doc_text">
427 <p>Global variables define regions of memory allocated at compilation time
428 instead of run-time. Global variables may optionally be initialized. A
429 variable may be defined as a global "constant", which indicates that the
430 contents of the variable will <b>never</b> be modified (enabling better
431 optimization, allowing the global data to be placed in the read-only section of
432 an executable, etc). Note that variables that need runtime initialization
433 cannot be marked "constant", as there is a store to the variable.</p>
436 LLVM explicitly allows <em>declarations</em> of global variables to be marked
437 constant, even if the final definition of the global is not. This capability
438 can be used to enable slightly better optimization of the program, but requires
439 the language definition to guarantee that optimizations based on the
440 'constantness' are valid for the translation units that do not include the
444 <p>As SSA values, global variables define pointer values that are in
445 scope (i.e. they dominate) all basic blocks in the program. Global
446 variables always define a pointer to their "content" type because they
447 describe a region of memory, and all memory objects in LLVM are
448 accessed through pointers.</p>
453 <!-- ======================================================================= -->
454 <div class="doc_subsection">
455 <a name="functionstructure">Functions</a>
458 <div class="doc_text">
460 <p>LLVM function definitions are composed of a (possibly empty) argument list,
461 an opening curly brace, a list of basic blocks, and a closing curly brace. LLVM
462 function declarations are defined with the "<tt>declare</tt>" keyword, a
463 function name, and a function signature.</p>
465 <p>A function definition contains a list of basic blocks, forming the CFG for
466 the function. Each basic block may optionally start with a label (giving the
467 basic block a symbol table entry), contains a list of instructions, and ends
468 with a <a href="#terminators">terminator</a> instruction (such as a branch or
469 function return).</p>
471 <p>The first basic block in program is special in two ways: it is immediately
472 executed on entrance to the function, and it is not allowed to have predecessor
473 basic blocks (i.e. there can not be any branches to the entry block of a
474 function). Because the block can have no predecessors, it also cannot have any
475 <a href="#i_phi">PHI nodes</a>.</p>
477 <p>LLVM functions are identified by their name and type signature. Hence, two
478 functions with the same name but different parameter lists or return values are
479 considered different functions, and LLVM will resolves references to each
486 <!-- *********************************************************************** -->
487 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
488 <!-- *********************************************************************** -->
490 <div class="doc_text">
492 <p>The LLVM type system is one of the most important features of the
493 intermediate representation. Being typed enables a number of
494 optimizations to be performed on the IR directly, without having to do
495 extra analyses on the side before the transformation. A strong type
496 system makes it easier to read the generated code and enables novel
497 analyses and transformations that are not feasible to perform on normal
498 three address code representations.</p>
502 <!-- ======================================================================= -->
503 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
504 <div class="doc_text">
505 <p>The primitive types are the fundamental building blocks of the LLVM
506 system. The current set of primitive types are as follows:</p>
508 <table class="layout">
513 <tr><th>Type</th><th>Description</th></tr>
514 <tr><td><tt>void</tt></td><td>No value</td></tr>
515 <tr><td><tt>ubyte</tt></td><td>Unsigned 8 bit value</td></tr>
516 <tr><td><tt>ushort</tt></td><td>Unsigned 16 bit value</td></tr>
517 <tr><td><tt>uint</tt></td><td>Unsigned 32 bit value</td></tr>
518 <tr><td><tt>ulong</tt></td><td>Unsigned 64 bit value</td></tr>
519 <tr><td><tt>float</tt></td><td>32 bit floating point value</td></tr>
520 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
527 <tr><th>Type</th><th>Description</th></tr>
528 <tr><td><tt>bool</tt></td><td>True or False value</td></tr>
529 <tr><td><tt>sbyte</tt></td><td>Signed 8 bit value</td></tr>
530 <tr><td><tt>short</tt></td><td>Signed 16 bit value</td></tr>
531 <tr><td><tt>int</tt></td><td>Signed 32 bit value</td></tr>
532 <tr><td><tt>long</tt></td><td>Signed 64 bit value</td></tr>
533 <tr><td><tt>double</tt></td><td>64 bit floating point value</td></tr>
541 <!-- _______________________________________________________________________ -->
542 <div class="doc_subsubsection"> <a name="t_classifications">Type
543 Classifications</a> </div>
544 <div class="doc_text">
545 <p>These different primitive types fall into a few useful
548 <table border="1" cellspacing="0" cellpadding="4">
550 <tr><th>Classification</th><th>Types</th></tr>
552 <td><a name="t_signed">signed</a></td>
553 <td><tt>sbyte, short, int, long, float, double</tt></td>
556 <td><a name="t_unsigned">unsigned</a></td>
557 <td><tt>ubyte, ushort, uint, ulong</tt></td>
560 <td><a name="t_integer">integer</a></td>
561 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
564 <td><a name="t_integral">integral</a></td>
565 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt>
569 <td><a name="t_floating">floating point</a></td>
570 <td><tt>float, double</tt></td>
573 <td><a name="t_firstclass">first class</a></td>
574 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long,<br>
575 float, double, <a href="#t_pointer">pointer</a>,
576 <a href="#t_packed">packed</a></tt></td>
581 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
582 most important. Values of these types are the only ones which can be
583 produced by instructions, passed as arguments, or used as operands to
584 instructions. This means that all structures and arrays must be
585 manipulated either by pointer or by component.</p>
588 <!-- ======================================================================= -->
589 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
591 <div class="doc_text">
593 <p>The real power in LLVM comes from the derived types in the system.
594 This is what allows a programmer to represent arrays, functions,
595 pointers, and other useful types. Note that these derived types may be
596 recursive: For example, it is possible to have a two dimensional array.</p>
600 <!-- _______________________________________________________________________ -->
601 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
603 <div class="doc_text">
607 <p>The array type is a very simple derived type that arranges elements
608 sequentially in memory. The array type requires a size (number of
609 elements) and an underlying data type.</p>
614 [<# elements> x <elementtype>]
617 <p>The number of elements is a constant integer value, elementtype may
618 be any type with a size.</p>
621 <table class="layout">
624 <tt>[40 x int ]</tt><br/>
625 <tt>[41 x int ]</tt><br/>
626 <tt>[40 x uint]</tt><br/>
629 Array of 40 integer values.<br/>
630 Array of 41 integer values.<br/>
631 Array of 40 unsigned integer values.<br/>
635 <p>Here are some examples of multidimensional arrays:</p>
636 <table class="layout">
639 <tt>[3 x [4 x int]]</tt><br/>
640 <tt>[12 x [10 x float]]</tt><br/>
641 <tt>[2 x [3 x [4 x uint]]]</tt><br/>
644 3x4 array integer values.<br/>
645 12x10 array of single precision floating point values.<br/>
646 2x3x4 array of unsigned integer values.<br/>
652 <!-- _______________________________________________________________________ -->
653 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
654 <div class="doc_text">
656 <p>The function type can be thought of as a function signature. It
657 consists of a return type and a list of formal parameter types.
658 Function types are usually used to build virtual function tables
659 (which are structures of pointers to functions), for indirect function
660 calls, and when defining a function.</p>
662 The return type of a function type cannot be an aggregate type.
665 <pre> <returntype> (<parameter list>)<br></pre>
666 <p>Where '<tt><parameter list></tt>' is a comma-separated list of type
667 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
668 which indicates that the function takes a variable number of arguments.
669 Variable argument functions can access their arguments with the <a
670 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
672 <table class="layout">
675 <tt>int (int)</tt> <br/>
676 <tt>float (int, int *) *</tt><br/>
677 <tt>int (sbyte *, ...)</tt><br/>
680 function taking an <tt>int</tt>, returning an <tt>int</tt><br/>
681 <a href="#t_pointer">Pointer</a> to a function that takes an
682 <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
683 returning <tt>float</tt>.<br/>
684 A vararg function that takes at least one <a href="#t_pointer">pointer</a>
685 to <tt>sbyte</tt> (signed char in C), which returns an integer. This is
686 the signature for <tt>printf</tt> in LLVM.<br/>
692 <!-- _______________________________________________________________________ -->
693 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
694 <div class="doc_text">
696 <p>The structure type is used to represent a collection of data members
697 together in memory. The packing of the field types is defined to match
698 the ABI of the underlying processor. The elements of a structure may
699 be any type that has a size.</p>
700 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
701 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
702 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
705 <pre> { <type list> }<br></pre>
707 <table class="layout">
710 <tt>{ int, int, int }</tt><br/>
711 <tt>{ float, int (int) * }</tt><br/>
714 a triple of three <tt>int</tt> values<br/>
715 A pair, where the first element is a <tt>float</tt> and the second element
716 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
717 that takes an <tt>int</tt>, returning an <tt>int</tt>.<br/>
723 <!-- _______________________________________________________________________ -->
724 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
725 <div class="doc_text">
727 <p>As in many languages, the pointer type represents a pointer or
728 reference to another object, which must live in memory.</p>
730 <pre> <type> *<br></pre>
732 <table class="layout">
735 <tt>[4x int]*</tt><br/>
736 <tt>int (int *) *</tt><br/>
739 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
740 four <tt>int</tt> values<br/>
741 A <a href="#t_pointer">pointer</a> to a <a
742 href="#t_function">function</a> that takes an <tt>int*</tt>, returning an
749 <!-- _______________________________________________________________________ -->
750 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
751 <div class="doc_text">
753 <p>A packed type is a simple derived type that represents a vector
754 of elements. Packed types are used when multiple primitive data
755 are operated in parallel using a single instruction (SIMD).
756 A packed type requires a size (number of
757 elements) and an underlying primitive data type. Packed types are
758 considered <a href="#t_firstclass">first class</a>.</p>
760 <pre> < <# elements> x <elementtype> ><br></pre>
761 <p>The number of elements is a constant integer value, elementtype may
762 be any integral or floating point type.</p>
764 <table class="layout">
767 <tt><4 x int></tt><br/>
768 <tt><8 x float></tt><br/>
769 <tt><2 x uint></tt><br/>
772 Packed vector of 4 integer values.<br/>
773 Packed vector of 8 floating-point values.<br/>
774 Packed vector of 2 unsigned integer values.<br/>
780 <!-- *********************************************************************** -->
781 <div class="doc_section"> <a name="constants">Constants</a> </div>
782 <!-- *********************************************************************** -->
784 <div class="doc_text">
786 <p>LLVM has several different basic types of constants. This section describes
787 them all and their syntax.</p>
791 <!-- ======================================================================= -->
792 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
794 <div class="doc_text">
797 <dt><b>Boolean constants</b></dt>
799 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
800 constants of the <tt><a href="#t_primitive">bool</a></tt> type.
803 <dt><b>Integer constants</b></dt>
805 <dd>Standard integers (such as '4') are constants of the <a
806 href="#t_integer">integer</a> type. Negative numbers may be used with signed
810 <dt><b>Floating point constants</b></dt>
812 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
813 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
814 notation. Floating point constants have an optional hexadecimal
815 notation (see below). Floating point constants must have a <a
816 href="#t_floating">floating point</a> type. </dd>
818 <dt><b>Null pointer constants</b></dt>
820 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
821 and must be of <a href="#t_pointer">pointer type</a>.</dd>
825 <p>The one non-intuitive notation for constants is the optional hexadecimal form
826 of floating point constants. For example, the form '<tt>double
827 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
828 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
829 (and the only time that they are generated by the disassembler) is when a
830 floating point constant must be emitted but it cannot be represented as a
831 decimal floating point number. For example, NaN's, infinities, and other
832 special values are represented in their IEEE hexadecimal format so that
833 assembly and disassembly do not cause any bits to change in the constants.</p>
837 <!-- ======================================================================= -->
838 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
841 <div class="doc_text">
844 <dt><b>Structure constants</b></dt>
846 <dd>Structure constants are represented with notation similar to structure
847 type definitions (a comma separated list of elements, surrounded by braces
848 (<tt>{}</tt>)). For example: "<tt>{ int 4, float 17.0 }</tt>". Structure
849 constants must have <a href="#t_struct">structure type</a>, and the number and
850 types of elements must match those specified by the type.
853 <dt><b>Array constants</b></dt>
855 <dd>Array constants are represented with notation similar to array type
856 definitions (a comma separated list of elements, surrounded by square brackets
857 (<tt>[]</tt>)). For example: "<tt>[ int 42, int 11, int 74 ]</tt>". Array
858 constants must have <a href="#t_array">array type</a>, and the number and
859 types of elements must match those specified by the type.
862 <dt><b>Packed constants</b></dt>
864 <dd>Packed constants are represented with notation similar to packed type
865 definitions (a comma separated list of elements, surrounded by
866 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< int 42,
867 int 11, int 74, int 100 ></tt>". Packed constants must have <a
868 href="#t_packed">packed type</a>, and the number and types of elements must
869 match those specified by the type.
872 <dt><b>Zero initialization</b></dt>
874 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
875 value to zero of <em>any</em> type, including scalar and aggregate types.
876 This is often used to avoid having to print large zero initializers (e.g. for
877 large arrays), and is always exactly equivalent to using explicit zero
884 <!-- ======================================================================= -->
885 <div class="doc_subsection">
886 <a name="globalconstants">Global Variable and Function Addresses</a>
889 <div class="doc_text">
891 <p>The addresses of <a href="#globalvars">global variables</a> and <a
892 href="#functionstructure">functions</a> are always implicitly valid (link-time)
893 constants. These constants are explicitly referenced when the <a
894 href="#identifiers">identifier for the global</a> is used and always have <a
895 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
901 %Z = global [2 x int*] [ int* %X, int* %Y ]
906 <!-- ======================================================================= -->
907 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
908 <div class="doc_text">
909 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
910 no specific value. Undefined values may be of any type, and be used anywhere
911 a constant is permitted.</p>
913 <p>Undefined values indicate to the compiler that the program is well defined
914 no matter what value is used, giving the compiler more freedom to optimize.
918 <!-- ======================================================================= -->
919 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
922 <div class="doc_text">
924 <p>Constant expressions are used to allow expressions involving other constants
925 to be used as constants. Constant expressions may be of any <a
926 href="#t_firstclass">first class</a> type, and may involve any LLVM operation
927 that does not have side effects (e.g. load and call are not supported). The
928 following is the syntax for constant expressions:</p>
931 <dt><b><tt>cast ( CST to TYPE )</tt></b></dt>
933 <dd>Cast a constant to another type.</dd>
935 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
937 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
938 constants. As with the <a href="#i_getelementptr">getelementptr</a>
939 instruction, the index list may have zero or more indexes, which are required
940 to make sense for the type of "CSTPTR".</dd>
942 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
944 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
945 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
946 binary</a> operations. The constraints on operands are the same as those for
947 the corresponding instruction (e.g. no bitwise operations on floating point
952 <!-- *********************************************************************** -->
953 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
954 <!-- *********************************************************************** -->
956 <div class="doc_text">
958 <p>The LLVM instruction set consists of several different
959 classifications of instructions: <a href="#terminators">terminator
960 instructions</a>, <a href="#binaryops">binary instructions</a>, <a
961 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
962 instructions</a>.</p>
966 <!-- ======================================================================= -->
967 <div class="doc_subsection"> <a name="terminators">Terminator
968 Instructions</a> </div>
970 <div class="doc_text">
972 <p>As mentioned <a href="#functionstructure">previously</a>, every
973 basic block in a program ends with a "Terminator" instruction, which
974 indicates which block should be executed after the current block is
975 finished. These terminator instructions typically yield a '<tt>void</tt>'
976 value: they produce control flow, not values (the one exception being
977 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
978 <p>There are six different terminator instructions: the '<a
979 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
980 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
981 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
982 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
983 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
987 <!-- _______________________________________________________________________ -->
988 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
989 Instruction</a> </div>
990 <div class="doc_text">
992 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
993 ret void <i>; Return from void function</i>
996 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
997 value) from a function, back to the caller.</p>
998 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
999 returns a value and then causes control flow, and one that just causes
1000 control flow to occur.</p>
1002 <p>The '<tt>ret</tt>' instruction may return any '<a
1003 href="#t_firstclass">first class</a>' type. Notice that a function is
1004 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1005 instruction inside of the function that returns a value that does not
1006 match the return type of the function.</p>
1008 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1009 returns back to the calling function's context. If the caller is a "<a
1010 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1011 the instruction after the call. If the caller was an "<a
1012 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1013 at the beginning "normal" of the destination block. If the instruction
1014 returns a value, that value shall set the call or invoke instruction's
1017 <pre> ret int 5 <i>; Return an integer value of 5</i>
1018 ret void <i>; Return from a void function</i>
1021 <!-- _______________________________________________________________________ -->
1022 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1023 <div class="doc_text">
1025 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1028 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1029 transfer to a different basic block in the current function. There are
1030 two forms of this instruction, corresponding to a conditional branch
1031 and an unconditional branch.</p>
1033 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1034 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
1035 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1036 value as a target.</p>
1038 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
1039 argument is evaluated. If the value is <tt>true</tt>, control flows
1040 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1041 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1043 <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
1044 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
1046 <!-- _______________________________________________________________________ -->
1047 <div class="doc_subsubsection">
1048 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1051 <div class="doc_text">
1055 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1060 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1061 several different places. It is a generalization of the '<tt>br</tt>'
1062 instruction, allowing a branch to occur to one of many possible
1068 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1069 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1070 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1071 table is not allowed to contain duplicate constant entries.</p>
1075 <p>The <tt>switch</tt> instruction specifies a table of values and
1076 destinations. When the '<tt>switch</tt>' instruction is executed, this
1077 table is searched for the given value. If the value is found, control flow is
1078 transfered to the corresponding destination; otherwise, control flow is
1079 transfered to the default destination.</p>
1081 <h5>Implementation:</h5>
1083 <p>Depending on properties of the target machine and the particular
1084 <tt>switch</tt> instruction, this instruction may be code generated in different
1085 ways. For example, it could be generated as a series of chained conditional
1086 branches or with a lookup table.</p>
1091 <i>; Emulate a conditional br instruction</i>
1092 %Val = <a href="#i_cast">cast</a> bool %value to int
1093 switch int %Val, label %truedest [int 0, label %falsedest ]
1095 <i>; Emulate an unconditional br instruction</i>
1096 switch uint 0, label %dest [ ]
1098 <i>; Implement a jump table:</i>
1099 switch uint %val, label %otherwise [ uint 0, label %onzero
1100 uint 1, label %onone
1101 uint 2, label %ontwo ]
1104 <!-- _______________________________________________________________________ -->
1105 <div class="doc_subsubsection"> <a name="i_invoke">'<tt>invoke</tt>'
1106 Instruction</a> </div>
1107 <div class="doc_text">
1109 <pre> <result> = invoke <ptr to function ty> %<function ptr val>(<function args>)<br> to label <normal label> except label <exception label><br></pre>
1111 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a
1112 specified function, with the possibility of control flow transfer to
1113 either the '<tt>normal</tt>' <tt>label</tt> label or the '<tt>exception</tt>'<tt>label</tt>.
1114 If the callee function returns with the "<tt><a href="#i_ret">ret</a></tt>"
1115 instruction, control flow will return to the "normal" label. If the
1116 callee (or any indirect callees) returns with the "<a href="#i_unwind"><tt>unwind</tt></a>"
1117 instruction, control is interrupted, and continued at the dynamically
1118 nearest "except" label.</p>
1120 <p>This instruction requires several arguments:</p>
1122 <li>'<tt>ptr to function ty</tt>': shall be the signature of the
1123 pointer to function value being invoked. In most cases, this is a
1124 direct function invocation, but indirect <tt>invoke</tt>s are just as
1125 possible, branching off an arbitrary pointer to function value. </li>
1126 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer
1127 to a function to be invoked. </li>
1128 <li>'<tt>function args</tt>': argument list whose types match the
1129 function signature argument types. If the function signature indicates
1130 the function accepts a variable number of arguments, the extra
1131 arguments can be specified. </li>
1132 <li>'<tt>normal label</tt>': the label reached when the called
1133 function executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1134 <li>'<tt>exception label</tt>': the label reached when a callee
1135 returns with the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1138 <p>This instruction is designed to operate as a standard '<tt><a
1139 href="#i_call">call</a></tt>' instruction in most regards. The
1140 primary difference is that it establishes an association with a label,
1141 which is used by the runtime library to unwind the stack.</p>
1142 <p>This instruction is used in languages with destructors to ensure
1143 that proper cleanup is performed in the case of either a <tt>longjmp</tt>
1144 or a thrown exception. Additionally, this is important for
1145 implementation of '<tt>catch</tt>' clauses in high-level languages that
1148 <pre> %retval = invoke int %Test(int 15)<br> to label %Continue<br> except label %TestCleanup <i>; {int}:retval set</i>
1153 <!-- _______________________________________________________________________ -->
1155 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1156 Instruction</a> </div>
1158 <div class="doc_text">
1167 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1168 at the first callee in the dynamic call stack which used an <a
1169 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1170 primarily used to implement exception handling.</p>
1174 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1175 immediately halt. The dynamic call stack is then searched for the first <a
1176 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1177 execution continues at the "exceptional" destination block specified by the
1178 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1179 dynamic call chain, undefined behavior results.</p>
1182 <!-- _______________________________________________________________________ -->
1184 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1185 Instruction</a> </div>
1187 <div class="doc_text">
1196 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1197 instruction is used to inform the optimizer that a particular portion of the
1198 code is not reachable. This can be used to indicate that the code after a
1199 no-return function cannot be reached, and other facts.</p>
1203 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1208 <!-- ======================================================================= -->
1209 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1210 <div class="doc_text">
1211 <p>Binary operators are used to do most of the computation in a
1212 program. They require two operands, execute an operation on them, and
1213 produce a single value. The operands might represent
1214 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1215 The result value of a binary operator is not
1216 necessarily the same type as its operands.</p>
1217 <p>There are several different binary operators:</p>
1219 <!-- _______________________________________________________________________ -->
1220 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1221 Instruction</a> </div>
1222 <div class="doc_text">
1224 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1227 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1229 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1230 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1231 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1232 Both arguments must have identical types.</p>
1234 <p>The value produced is the integer or floating point sum of the two
1237 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
1240 <!-- _______________________________________________________________________ -->
1241 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1242 Instruction</a> </div>
1243 <div class="doc_text">
1245 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1248 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1250 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1251 instruction present in most other intermediate representations.</p>
1253 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1254 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1256 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1257 Both arguments must have identical types.</p>
1259 <p>The value produced is the integer or floating point difference of
1260 the two operands.</p>
1262 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
1263 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
1266 <!-- _______________________________________________________________________ -->
1267 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1268 Instruction</a> </div>
1269 <div class="doc_text">
1271 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1274 <p>The '<tt>mul</tt>' instruction returns the product of its two
1277 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1278 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1280 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1281 Both arguments must have identical types.</p>
1283 <p>The value produced is the integer or floating point product of the
1285 <p>There is no signed vs unsigned multiplication. The appropriate
1286 action is taken based on the type of the operand.</p>
1288 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
1291 <!-- _______________________________________________________________________ -->
1292 <div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
1293 Instruction</a> </div>
1294 <div class="doc_text">
1296 <pre> <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1299 <p>The '<tt>div</tt>' instruction returns the quotient of its two
1302 <p>The two arguments to the '<tt>div</tt>' instruction must be either <a
1303 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1305 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1306 Both arguments must have identical types.</p>
1308 <p>The value produced is the integer or floating point quotient of the
1311 <pre> <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
1314 <!-- _______________________________________________________________________ -->
1315 <div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
1316 Instruction</a> </div>
1317 <div class="doc_text">
1319 <pre> <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1322 <p>The '<tt>rem</tt>' instruction returns the remainder from the
1323 division of its two operands.</p>
1325 <p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1326 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1328 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1329 Both arguments must have identical types.</p>
1331 <p>This returns the <i>remainder</i> of a division (where the result
1332 has the same sign as the divisor), not the <i>modulus</i> (where the
1333 result has the same sign as the dividend) of a value. For more
1334 information about the difference, see: <a
1335 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1338 <pre> <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1341 <!-- _______________________________________________________________________ -->
1342 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1343 Instructions</a> </div>
1344 <div class="doc_text">
1346 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1347 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1348 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1349 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1350 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1351 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1354 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1355 value based on a comparison of their two operands.</p>
1357 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1358 be of <a href="#t_firstclass">first class</a> type (it is not possible
1359 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1360 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1363 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1364 value if both operands are equal.<br>
1365 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1366 value if both operands are unequal.<br>
1367 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1368 value if the first operand is less than the second operand.<br>
1369 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1370 value if the first operand is greater than the second operand.<br>
1371 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1372 value if the first operand is less than or equal to the second operand.<br>
1373 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1374 value if the first operand is greater than or equal to the second
1377 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1378 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1379 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1380 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1381 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1382 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1385 <!-- ======================================================================= -->
1386 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1387 Operations</a> </div>
1388 <div class="doc_text">
1389 <p>Bitwise binary operators are used to do various forms of
1390 bit-twiddling in a program. They are generally very efficient
1391 instructions and can commonly be strength reduced from other
1392 instructions. They require two operands, execute an operation on them,
1393 and produce a single value. The resulting value of the bitwise binary
1394 operators is always the same type as its first operand.</p>
1396 <!-- _______________________________________________________________________ -->
1397 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1398 Instruction</a> </div>
1399 <div class="doc_text">
1401 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1404 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1405 its two operands.</p>
1407 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1408 href="#t_integral">integral</a> values. Both arguments must have
1409 identical types.</p>
1411 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1413 <div style="align: center">
1414 <table border="1" cellspacing="0" cellpadding="4">
1445 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1446 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1447 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1450 <!-- _______________________________________________________________________ -->
1451 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1452 <div class="doc_text">
1454 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1457 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1458 or of its two operands.</p>
1460 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1461 href="#t_integral">integral</a> values. Both arguments must have
1462 identical types.</p>
1464 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1466 <div style="align: center">
1467 <table border="1" cellspacing="0" cellpadding="4">
1498 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1499 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1500 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1503 <!-- _______________________________________________________________________ -->
1504 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1505 Instruction</a> </div>
1506 <div class="doc_text">
1508 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1511 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1512 or of its two operands. The <tt>xor</tt> is used to implement the
1513 "one's complement" operation, which is the "~" operator in C.</p>
1515 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1516 href="#t_integral">integral</a> values. Both arguments must have
1517 identical types.</p>
1519 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1521 <div style="align: center">
1522 <table border="1" cellspacing="0" cellpadding="4">
1554 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1555 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1556 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1557 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1560 <!-- _______________________________________________________________________ -->
1561 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1562 Instruction</a> </div>
1563 <div class="doc_text">
1565 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1568 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1569 the left a specified number of bits.</p>
1571 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
1572 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1575 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
1577 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1578 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1579 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1582 <!-- _______________________________________________________________________ -->
1583 <div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1584 Instruction</a> </div>
1585 <div class="doc_text">
1587 <pre> <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1590 <p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1591 the right a specified number of bits.</p>
1593 <p>The first argument to the '<tt>shr</tt>' instruction must be an <a
1594 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1597 <p>If the first argument is a <a href="#t_signed">signed</a> type, the
1598 most significant bit is duplicated in the newly free'd bit positions.
1599 If the first argument is unsigned, zero bits shall fill the empty
1602 <pre> <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1603 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1604 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1605 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1606 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1609 <!-- ======================================================================= -->
1610 <div class="doc_subsection"> <a name="memoryops">Memory Access
1611 Operations</a></div>
1612 <div class="doc_text">
1613 <p>A key design point of an SSA-based representation is how it
1614 represents memory. In LLVM, no memory locations are in SSA form, which
1615 makes things very simple. This section describes how to read, write,
1616 allocate, and free memory in LLVM.</p>
1618 <!-- _______________________________________________________________________ -->
1619 <div class="doc_subsubsection"> <a name="i_malloc">'<tt>malloc</tt>'
1620 Instruction</a> </div>
1621 <div class="doc_text">
1623 <pre> <result> = malloc <type>, uint <NumElements> <i>; yields {type*}:result</i>
1624 <result> = malloc <type> <i>; yields {type*}:result</i>
1627 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
1628 heap and returns a pointer to it.</p>
1630 <p>The '<tt>malloc</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1631 bytes of memory from the operating system and returns a pointer of the
1632 appropriate type to the program. The second form of the instruction is
1633 a shorter version of the first instruction that defaults to allocating
1635 <p>'<tt>type</tt>' must be a sized type.</p>
1637 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
1638 a pointer is returned.</p>
1640 <pre> %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1643 href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1644 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1645 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1648 <!-- _______________________________________________________________________ -->
1649 <div class="doc_subsubsection"> <a name="i_free">'<tt>free</tt>'
1650 Instruction</a> </div>
1651 <div class="doc_text">
1653 <pre> free <type> <value> <i>; yields {void}</i>
1656 <p>The '<tt>free</tt>' instruction returns memory back to the unused
1657 memory heap, to be reallocated in the future.</p>
1660 <p>'<tt>value</tt>' shall be a pointer value that points to a value
1661 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
1664 <p>Access to the memory pointed to by the pointer is no longer defined
1665 after this instruction executes.</p>
1667 <pre> %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1668 free [4 x ubyte]* %array
1671 <!-- _______________________________________________________________________ -->
1672 <div class="doc_subsubsection"> <a name="i_alloca">'<tt>alloca</tt>'
1673 Instruction</a> </div>
1674 <div class="doc_text">
1676 <pre> <result> = alloca <type>, uint <NumElements> <i>; yields {type*}:result</i>
1677 <result> = alloca <type> <i>; yields {type*}:result</i>
1680 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
1681 stack frame of the procedure that is live until the current function
1682 returns to its caller.</p>
1684 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1685 bytes of memory on the runtime stack, returning a pointer of the
1686 appropriate type to the program. The second form of the instruction is
1687 a shorter version of the first that defaults to allocating one element.</p>
1688 <p>'<tt>type</tt>' may be any sized type.</p>
1690 <p>Memory is allocated, a pointer is returned. '<tt>alloca</tt>'d
1691 memory is automatically released when the function returns. The '<tt>alloca</tt>'
1692 instruction is commonly used to represent automatic variables that must
1693 have an address available. When the function returns (either with the <tt><a
1694 href="#i_ret">ret</a></tt> or <tt><a href="#i_invoke">invoke</a></tt>
1695 instructions), the memory is reclaimed.</p>
1697 <pre> %ptr = alloca int <i>; yields {int*}:ptr</i>
1698 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
1701 <!-- _______________________________________________________________________ -->
1702 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
1703 Instruction</a> </div>
1704 <div class="doc_text">
1706 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
1708 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
1710 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
1711 address to load from. The pointer must point to a <a
1712 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
1713 marked as <tt>volatile</tt> then the optimizer is not allowed to modify
1714 the number or order of execution of this <tt>load</tt> with other
1715 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
1718 <p>The location of memory pointed to is loaded.</p>
1720 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1722 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1723 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1726 <!-- _______________________________________________________________________ -->
1727 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
1728 Instruction</a> </div>
1730 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1731 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1734 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
1736 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
1737 to store and an address to store it into. The type of the '<tt><pointer></tt>'
1738 operand must be a pointer to the type of the '<tt><value></tt>'
1739 operand. If the <tt>store</tt> is marked as <tt>volatile</tt> then the
1740 optimizer is not allowed to modify the number or order of execution of
1741 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
1742 href="#i_store">store</a></tt> instructions.</p>
1744 <p>The contents of memory are updated to contain '<tt><value></tt>'
1745 at the location specified by the '<tt><pointer></tt>' operand.</p>
1747 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1749 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1750 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1752 <!-- _______________________________________________________________________ -->
1753 <div class="doc_subsubsection">
1754 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
1757 <div class="doc_text">
1760 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
1766 The '<tt>getelementptr</tt>' instruction is used to get the address of a
1767 subelement of an aggregate data structure.</p>
1771 <p>This instruction takes a list of integer constants that indicate what
1772 elements of the aggregate object to index to. The actual types of the arguments
1773 provided depend on the type of the first pointer argument. The
1774 '<tt>getelementptr</tt>' instruction is used to index down through the type
1775 levels of a structure. When indexing into a structure, only <tt>uint</tt>
1776 integer constants are allowed. When indexing into an array or pointer
1777 <tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
1779 <p>For example, let's consider a C code fragment and how it gets
1780 compiled to LLVM:</p>
1794 int *foo(struct ST *s) {
1795 return &s[1].Z.B[5][13];
1799 <p>The LLVM code generated by the GCC frontend is:</p>
1802 %RT = type { sbyte, [10 x [20 x int]], sbyte }
1803 %ST = type { int, double, %RT }
1807 int* %foo(%ST* %s) {
1809 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
1816 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
1817 on the pointer type that is being index into. <a href="#t_pointer">Pointer</a>
1818 and <a href="#t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
1819 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="#t_struct">structure</a>
1820 types require <tt>uint</tt> <b>constants</b>.</p>
1822 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
1823 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
1824 }</tt>' type, a structure. The second index indexes into the third element of
1825 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
1826 sbyte }</tt>' type, another structure. The third index indexes into the second
1827 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
1828 array. The two dimensions of the array are subscripted into, yielding an
1829 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction return a pointer
1830 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
1832 <p>Note that it is perfectly legal to index partially through a
1833 structure, returning a pointer to an inner element. Because of this,
1834 the LLVM code for the given testcase is equivalent to:</p>
1837 int* "foo"(%ST* %s) {
1838 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
1839 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
1840 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
1841 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
1842 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
1848 <i>; yields [12 x ubyte]*:aptr</i>
1849 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
1853 <!-- ======================================================================= -->
1854 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
1855 <div class="doc_text">
1856 <p>The instructions in this category are the "miscellaneous"
1857 instructions, which defy better classification.</p>
1859 <!-- _______________________________________________________________________ -->
1860 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
1861 Instruction</a> </div>
1862 <div class="doc_text">
1864 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
1866 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
1867 the SSA graph representing the function.</p>
1869 <p>The type of the incoming values are specified with the first type
1870 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
1871 as arguments, with one pair for each predecessor basic block of the
1872 current block. Only values of <a href="#t_firstclass">first class</a>
1873 type may be used as the value arguments to the PHI node. Only labels
1874 may be used as the label arguments.</p>
1875 <p>There must be no non-phi instructions between the start of a basic
1876 block and the PHI instructions: i.e. PHI instructions must be first in
1879 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
1880 value specified by the parameter, depending on which basic block we
1881 came from in the last <a href="#terminators">terminator</a> instruction.</p>
1883 <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>
1886 <!-- _______________________________________________________________________ -->
1887 <div class="doc_subsubsection">
1888 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
1891 <div class="doc_text">
1896 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
1902 The '<tt>cast</tt>' instruction is used as the primitive means to convert
1903 integers to floating point, change data type sizes, and break type safety (by
1911 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
1912 class value, and a type to cast it to, which must also be a <a
1913 href="#t_firstclass">first class</a> type.
1919 This instruction follows the C rules for explicit casts when determining how the
1920 data being cast must change to fit in its new container.
1924 When casting to bool, any value that would be considered true in the context of
1925 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
1926 all else are '<tt>false</tt>'.
1930 When extending an integral value from a type of one signness to another (for
1931 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
1932 <b>source</b> value is signed, and zero-extended if the source value is
1933 unsigned. <tt>bool</tt> values are always zero extended into either zero or
1940 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
1941 %Y = cast int 123 to bool <i>; yields bool:true</i>
1945 <!-- _______________________________________________________________________ -->
1946 <div class="doc_subsubsection">
1947 <a name="i_select">'<tt>select</tt>' Instruction</a>
1950 <div class="doc_text">
1955 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
1961 The '<tt>select</tt>' instruction is used to choose one value based on a
1962 condition, without branching.
1969 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.
1975 If the boolean condition evaluates to true, the instruction returns the first
1976 value argument, otherwise it returns the second value argument.
1982 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
1990 <!-- _______________________________________________________________________ -->
1991 <div class="doc_subsubsection"> <a name="i_call">'<tt>call</tt>'
1992 Instruction</a> </div>
1993 <div class="doc_text">
1995 <pre> <result> = call <ty>* <fnptrval>(<param list>)<br></pre>
1997 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
1999 <p>This instruction requires several arguments:</p>
2002 <p>'<tt>ty</tt>': shall be the signature of the pointer to function
2003 value being invoked. The argument types must match the types implied
2004 by this signature.</p>
2007 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a
2008 function to be invoked. In most cases, this is a direct function
2009 invocation, but indirect <tt>call</tt>s are just as possible,
2010 calling an arbitrary pointer to function values.</p>
2013 <p>'<tt>function args</tt>': argument list whose types match the
2014 function signature argument types. If the function signature
2015 indicates the function accepts a variable number of arguments, the
2016 extra arguments can be specified.</p>
2020 <p>The '<tt>call</tt>' instruction is used to cause control flow to
2021 transfer to a specified function, with its incoming arguments bound to
2022 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
2023 instruction in the called function, control flow continues with the
2024 instruction after the function call, and the return value of the
2025 function is bound to the result argument. This is a simpler case of
2026 the <a href="#i_invoke">invoke</a> instruction.</p>
2028 <pre> %retval = call int %test(int %argc)<br> call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);<br></pre>
2031 <!-- _______________________________________________________________________ -->
2032 <div class="doc_subsubsection">
2033 <a name="i_vanext">'<tt>vanext</tt>' Instruction</a>
2036 <div class="doc_text">
2041 <resultarglist> = vanext <va_list> <arglist>, <argty>
2046 <p>The '<tt>vanext</tt>' instruction is used to access arguments passed
2047 through the "variable argument" area of a function call. It is used to
2048 implement the <tt>va_arg</tt> macro in C.</p>
2052 <p>This instruction takes a <tt>va_list</tt> value and the type of the
2053 argument. It returns another <tt>va_list</tt>. The actual type of
2054 <tt>va_list</tt> may be defined differently for different targets. Most targets
2055 use a <tt>va_list</tt> type of <tt>sbyte*</tt> or some other pointer type.</p>
2059 <p>The '<tt>vanext</tt>' instruction advances the specified <tt>va_list</tt>
2060 past an argument of the specified type. In conjunction with the <a
2061 href="#i_vaarg"><tt>vaarg</tt></a> instruction, it is used to implement
2062 the <tt>va_arg</tt> macro available in C. For more information, see
2063 the variable argument handling <a href="#int_varargs">Intrinsic
2066 <p>It is legal for this instruction to be called in a function which
2067 does not take a variable number of arguments, for example, the <tt>vfprintf</tt>
2070 <p><tt>vanext</tt> is an LLVM instruction instead of an <a
2071 href="#intrinsics">intrinsic function</a> because it takes a type as an
2072 argument. The type refers to the current argument in the <tt>va_list</tt>, it
2073 tells the compiler how far on the stack it needs to advance to find the next
2078 <p>See the <a href="#int_varargs">variable argument processing</a>
2083 <!-- _______________________________________________________________________ -->
2084 <div class="doc_subsubsection">
2085 <a name="i_vaarg">'<tt>vaarg</tt>' Instruction</a>
2088 <div class="doc_text">
2093 <resultval> = vaarg <va_list> <arglist>, <argty>
2098 <p>The '<tt>vaarg</tt>' instruction is used to access arguments passed through
2099 the "variable argument" area of a function call. It is used to implement the
2100 <tt>va_arg</tt> macro in C.</p>
2104 <p>This instruction takes a <tt>va_list</tt> value and the type of the
2105 argument. It returns a value of the specified argument type. Again, the actual
2106 type of <tt>va_list</tt> is target specific.</p>
2110 <p>The '<tt>vaarg</tt>' instruction loads an argument of the specified type from
2111 the specified <tt>va_list</tt>. In conjunction with the <a
2112 href="#i_vanext"><tt>vanext</tt></a> instruction, it is used to implement the
2113 <tt>va_arg</tt> macro available in C. For more information, see the variable
2114 argument handling <a href="#int_varargs">Intrinsic Functions</a>.</p>
2116 <p>It is legal for this instruction to be called in a function which does not
2117 take a variable number of arguments, for example, the <tt>vfprintf</tt>
2120 <p><tt>vaarg</tt> is an LLVM instruction instead of an <a
2121 href="#intrinsics">intrinsic function</a> because it takes an type as an
2126 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
2130 <!-- *********************************************************************** -->
2131 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
2132 <!-- *********************************************************************** -->
2134 <div class="doc_text">
2136 <p>LLVM supports the notion of an "intrinsic function". These functions have
2137 well known names and semantics, and are required to follow certain
2138 restrictions. Overall, these instructions represent an extension mechanism for
2139 the LLVM language that does not require changing all of the transformations in
2140 LLVM to add to the language (or the bytecode reader/writer, the parser,
2143 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix, this
2144 prefix is reserved in LLVM for intrinsic names, thus functions may not be named
2145 this. Intrinsic functions must always be external functions: you cannot define
2146 the body of intrinsic functions. Intrinsic functions may only be used in call
2147 or invoke instructions: it is illegal to take the address of an intrinsic
2148 function. Additionally, because intrinsic functions are part of the LLVM
2149 language, it is required that they all be documented here if any are added.</p>
2153 Adding an intrinsic to LLVM is straight-forward if it is possible to express the
2154 concept in LLVM directly (ie, code generator support is not _required_). To do
2155 this, extend the default implementation of the IntrinsicLowering class to handle
2156 the intrinsic. Code generators use this class to lower intrinsics they do not
2157 understand to raw LLVM instructions that they do.
2162 <!-- ======================================================================= -->
2163 <div class="doc_subsection">
2164 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
2167 <div class="doc_text">
2169 <p>Variable argument support is defined in LLVM with the <a
2170 href="#i_vanext"><tt>vanext</tt></a> instruction and these three
2171 intrinsic functions. These functions are related to the similarly
2172 named macros defined in the <tt><stdarg.h></tt> header file.</p>
2174 <p>All of these functions operate on arguments that use a
2175 target-specific value type "<tt>va_list</tt>". The LLVM assembly
2176 language reference manual does not define what this type is, so all
2177 transformations should be prepared to handle intrinsics with any type
2180 <p>This example shows how the <a href="#i_vanext"><tt>vanext</tt></a>
2181 instruction and the variable argument handling intrinsic functions are
2185 int %test(int %X, ...) {
2186 ; Initialize variable argument processing
2187 %ap = call sbyte* %<a href="#i_va_start">llvm.va_start</a>()
2189 ; Read a single integer argument
2190 %tmp = vaarg sbyte* %ap, int
2192 ; Advance to the next argument
2193 %ap2 = vanext sbyte* %ap, int
2195 ; Demonstrate usage of llvm.va_copy and llvm.va_end
2196 %aq = call sbyte* %<a href="#i_va_copy">llvm.va_copy</a>(sbyte* %ap2)
2197 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %aq)
2199 ; Stop processing of arguments.
2200 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte* %ap2)
2206 <!-- _______________________________________________________________________ -->
2207 <div class="doc_subsubsection">
2208 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
2212 <div class="doc_text">
2214 <pre> call <va_list> ()* %llvm.va_start()<br></pre>
2216 <p>The '<tt>llvm.va_start</tt>' intrinsic returns a new <tt><arglist></tt>
2217 for subsequent use by the variable argument intrinsics.</p>
2219 <p>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
2220 macro available in C. In a target-dependent way, it initializes and
2221 returns a <tt>va_list</tt> element, so that the next <tt>vaarg</tt>
2222 will produce the first variable argument passed to the function. Unlike
2223 the C <tt>va_start</tt> macro, this intrinsic does not need to know the
2224 last argument of the function, the compiler can figure that out.</p>
2225 <p>Note that this intrinsic function is only legal to be called from
2226 within the body of a variable argument function.</p>
2229 <!-- _______________________________________________________________________ -->
2230 <div class="doc_subsubsection">
2231 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
2234 <div class="doc_text">
2236 <pre> call void (<va_list>)* %llvm.va_end(<va_list> <arglist>)<br></pre>
2238 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
2239 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
2240 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
2242 <p>The argument is a <tt>va_list</tt> to destroy.</p>
2244 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
2245 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
2246 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
2247 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
2248 with calls to <tt>llvm.va_end</tt>.</p>
2251 <!-- _______________________________________________________________________ -->
2252 <div class="doc_subsubsection">
2253 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
2256 <div class="doc_text">
2261 call <va_list> (<va_list>)* %llvm.va_copy(<va_list> <destarglist>)
2266 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position
2267 from the source argument list to the destination argument list.</p>
2271 <p>The argument is the <tt>va_list</tt> to copy.</p>
2275 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt>
2276 macro available in C. In a target-dependent way, it copies the source
2277 <tt>va_list</tt> element into the returned list. This intrinsic is necessary
2278 because the <tt><a href="#i_va_start">llvm.va_start</a></tt> intrinsic may be
2279 arbitrarily complex and require memory allocation, for example.</p>
2283 <!-- ======================================================================= -->
2284 <div class="doc_subsection">
2285 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
2288 <div class="doc_text">
2291 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
2292 Collection</a> requires the implementation and generation of these intrinsics.
2293 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
2294 stack</a>, as well as garbage collector implementations that require <a
2295 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
2296 Front-ends for type-safe garbage collected languages should generate these
2297 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
2298 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
2302 <!-- _______________________________________________________________________ -->
2303 <div class="doc_subsubsection">
2304 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
2307 <div class="doc_text">
2312 call void (<ty>**, <ty2>*)* %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
2317 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
2318 the code generator, and allows some metadata to be associated with it.</p>
2322 <p>The first argument specifies the address of a stack object that contains the
2323 root pointer. The second pointer (which must be either a constant or a global
2324 value address) contains the meta-data to be associated with the root.</p>
2328 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
2329 location. At compile-time, the code generator generates information to allow
2330 the runtime to find the pointer at GC safe points.
2336 <!-- _______________________________________________________________________ -->
2337 <div class="doc_subsubsection">
2338 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
2341 <div class="doc_text">
2346 call sbyte* (sbyte**)* %llvm.gcread(sbyte** %Ptr)
2351 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
2352 locations, allowing garbage collector implementations that require read
2357 <p>The argument is the address to read from, which should be an address
2358 allocated from the garbage collector.</p>
2362 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
2363 instruction, but may be replaced with substantially more complex code by the
2364 garbage collector runtime, as needed.</p>
2369 <!-- _______________________________________________________________________ -->
2370 <div class="doc_subsubsection">
2371 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
2374 <div class="doc_text">
2379 call void (sbyte*, sbyte**)* %llvm.gcwrite(sbyte* %P1, sbyte** %P2)
2384 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
2385 locations, allowing garbage collector implementations that require write
2386 barriers (such as generational or reference counting collectors).</p>
2390 <p>The first argument is the reference to store, and the second is the heap
2391 location to store to.</p>
2395 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
2396 instruction, but may be replaced with substantially more complex code by the
2397 garbage collector runtime, as needed.</p>
2403 <!-- ======================================================================= -->
2404 <div class="doc_subsection">
2405 <a name="int_codegen">Code Generator Intrinsics</a>
2408 <div class="doc_text">
2410 These intrinsics are provided by LLVM to expose special features that may only
2411 be implemented with code generator support.
2416 <!-- _______________________________________________________________________ -->
2417 <div class="doc_subsubsection">
2418 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
2421 <div class="doc_text">
2425 call void* ()* %llvm.returnaddress(uint <level>)
2431 The '<tt>llvm.returnaddress</tt>' intrinsic returns a target-specific value
2432 indicating the return address of the current function or one of its callers.
2438 The argument to this intrinsic indicates which function to return the address
2439 for. Zero indicates the calling function, one indicates its caller, etc. The
2440 argument is <b>required</b> to be a constant integer value.
2446 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
2447 the return address of the specified call frame, or zero if it cannot be
2448 identified. The value returned by this intrinsic is likely to be incorrect or 0
2449 for arguments other than zero, so it should only be used for debugging purposes.
2453 Note that calling this intrinsic does not prevent function inlining or other
2454 aggressive transformations, so the value returned may not that of the obvious
2455 source-language caller.
2460 <!-- _______________________________________________________________________ -->
2461 <div class="doc_subsubsection">
2462 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
2465 <div class="doc_text">
2469 call void* ()* %llvm.frameaddress(uint <level>)
2475 The '<tt>llvm.frameaddress</tt>' intrinsic returns the target-specific frame
2476 pointer value for the specified stack frame.
2482 The argument to this intrinsic indicates which function to return the frame
2483 pointer for. Zero indicates the calling function, one indicates its caller,
2484 etc. The argument is <b>required</b> to be a constant integer value.
2490 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
2491 the frame address of the specified call frame, or zero if it cannot be
2492 identified. The value returned by this intrinsic is likely to be incorrect or 0
2493 for arguments other than zero, so it should only be used for debugging purposes.
2497 Note that calling this intrinsic does not prevent function inlining or other
2498 aggressive transformations, so the value returned may not that of the obvious
2499 source-language caller.
2503 <!-- _______________________________________________________________________ -->
2504 <div class="doc_subsubsection">
2505 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
2508 <div class="doc_text">
2512 call void (sbyte *, uint, uint)* %llvm.prefetch(sbyte * <address>,
2514 uint <locality>)
2521 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
2522 a prefetch instruction if supported, otherwise it is a noop. Prefetches have no
2523 effect on the behavior of the program, but can change its performance
2530 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
2531 determining if the fetch should be for a read (0) or write (1), and
2532 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
2533 locality, to (3) - exteremely local keep in cache. The <tt>rw</tt> and
2534 <tt>locality</tt> arguments must be constant integers.
2540 This intrinsic does not modify the behavior of the program. In particular,
2541 prefetches cannot trap and do not produce a value. On targets that support this
2542 intrinsic, the prefetch can provide hints to the processor cache for better
2549 <!-- ======================================================================= -->
2550 <div class="doc_subsection">
2551 <a name="int_os">Operating System Intrinsics</a>
2554 <div class="doc_text">
2556 These intrinsics are provided by LLVM to support the implementation of
2557 operating system level code.
2562 <!-- _______________________________________________________________________ -->
2563 <div class="doc_subsubsection">
2564 <a name="i_readport">'<tt>llvm.readport</tt>' Intrinsic</a>
2567 <div class="doc_text">
2571 call <integer type> (<integer type>)* %llvm.readport (<integer type> <address>)
2577 The '<tt>llvm.readport</tt>' intrinsic reads data from the specified hardware
2584 The argument to this intrinsic indicates the hardware I/O address from which
2585 to read the data. The address is in the hardware I/O address namespace (as
2586 opposed to being a memory location for memory mapped I/O).
2592 The '<tt>llvm.readport</tt>' intrinsic reads data from the hardware I/O port
2593 specified by <i>address</i> and returns the value. The address and return
2594 value must be integers, but the size is dependent upon the platform upon which
2595 the program is code generated. For example, on x86, the address must be an
2596 unsigned 16 bit value, and the return value must be 8, 16, or 32 bits.
2601 <!-- _______________________________________________________________________ -->
2602 <div class="doc_subsubsection">
2603 <a name="i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a>
2606 <div class="doc_text">
2610 call void (<integer type>, <integer type>)*
2611 %llvm.writeport (<integer type> <value>,
2612 <integer type> <address>)
2618 The '<tt>llvm.writeport</tt>' intrinsic writes data to the specified hardware
2625 The first argument is the value to write to the I/O port.
2629 The second argument indicates the hardware I/O address to which data should be
2630 written. The address is in the hardware I/O address namespace (as opposed to
2631 being a memory location for memory mapped I/O).
2637 The '<tt>llvm.writeport</tt>' intrinsic writes <i>value</i> to the I/O port
2638 specified by <i>address</i>. The address and value must be integers, but the
2639 size is dependent upon the platform upon which the program is code generated.
2640 For example, on x86, the address must be an unsigned 16 bit value, and the
2641 value written must be 8, 16, or 32 bits in length.
2646 <!-- _______________________________________________________________________ -->
2647 <div class="doc_subsubsection">
2648 <a name="i_readio">'<tt>llvm.readio</tt>' Intrinsic</a>
2651 <div class="doc_text">
2655 call <result> (<ty>*)* %llvm.readio (<ty> * <pointer>)
2661 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2668 The argument to this intrinsic is a pointer indicating the memory address from
2669 which to read the data. The data must be a
2670 <a href="#t_firstclass">first class</a> type.
2676 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
2677 location specified by <i>pointer</i> and returns the value. The argument must
2678 be a pointer, and the return value must be a
2679 <a href="#t_firstclass">first class</a> type. However, certain architectures
2680 may not support I/O on all first class types. For example, 32 bit processors
2681 may only support I/O on data types that are 32 bits or less.
2685 This intrinsic enforces an in-order memory model for llvm.readio and
2686 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2687 scheduled processors may execute loads and stores out of order, re-ordering at
2688 run time accesses to memory mapped I/O registers. Using these intrinsics
2689 ensures that accesses to memory mapped I/O registers occur in program order.
2694 <!-- _______________________________________________________________________ -->
2695 <div class="doc_subsubsection">
2696 <a name="i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a>
2699 <div class="doc_text">
2703 call void (<ty1>, <ty2>*)* %llvm.writeio (<ty1> <value>, <ty2> * <pointer>)
2709 The '<tt>llvm.writeio</tt>' intrinsic writes data to the specified memory
2716 The first argument is the value to write to the memory mapped I/O location.
2717 The second argument is a pointer indicating the memory address to which the
2718 data should be written.
2724 The '<tt>llvm.writeio</tt>' intrinsic writes <i>value</i> to the memory mapped
2725 I/O address specified by <i>pointer</i>. The value must be a
2726 <a href="#t_firstclass">first class</a> type. However, certain architectures
2727 may not support I/O on all first class types. For example, 32 bit processors
2728 may only support I/O on data types that are 32 bits or less.
2732 This intrinsic enforces an in-order memory model for llvm.readio and
2733 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
2734 scheduled processors may execute loads and stores out of order, re-ordering at
2735 run time accesses to memory mapped I/O registers. Using these intrinsics
2736 ensures that accesses to memory mapped I/O registers occur in program order.
2741 <!-- ======================================================================= -->
2742 <div class="doc_subsection">
2743 <a name="int_libc">Standard C Library Intrinsics</a>
2746 <div class="doc_text">
2748 LLVM provides intrinsics for a few important standard C library functions.
2749 These intrinsics allow source-language front-ends to pass information about the
2750 alignment of the pointer arguments to the code generator, providing opportunity
2751 for more efficient code generation.
2756 <!-- _______________________________________________________________________ -->
2757 <div class="doc_subsubsection">
2758 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
2761 <div class="doc_text">
2765 call void (sbyte*, sbyte*, uint, uint)* %llvm.memcpy(sbyte* <dest>, sbyte* <src>,
2766 uint <len>, uint <align>)
2772 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2773 location to the destination location.
2777 Note that, unlike the standard libc function, the <tt>llvm.memcpy</tt> intrinsic
2778 does not return a value, and takes an extra alignment argument.
2784 The first argument is a pointer to the destination, the second is a pointer to
2785 the source. The third argument is an (arbitrarily sized) integer argument
2786 specifying the number of bytes to copy, and the fourth argument is the alignment
2787 of the source and destination locations.
2791 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2792 the caller guarantees that the size of the copy is a multiple of the alignment
2793 and that both the source and destination pointers are aligned to that boundary.
2799 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
2800 location to the destination location, which are not allowed to overlap. It
2801 copies "len" bytes of memory over. If the argument is known to be aligned to
2802 some boundary, this can be specified as the fourth argument, otherwise it should
2808 <!-- _______________________________________________________________________ -->
2809 <div class="doc_subsubsection">
2810 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
2813 <div class="doc_text">
2817 call void (sbyte*, sbyte*, uint, uint)* %llvm.memmove(sbyte* <dest>, sbyte* <src>,
2818 uint <len>, uint <align>)
2824 The '<tt>llvm.memmove</tt>' intrinsic moves a block of memory from the source
2825 location to the destination location. It is similar to the '<tt>llvm.memcpy</tt>'
2826 intrinsic but allows the two memory locations to overlap.
2830 Note that, unlike the standard libc function, the <tt>llvm.memmove</tt> intrinsic
2831 does not return a value, and takes an extra alignment argument.
2837 The first argument is a pointer to the destination, the second is a pointer to
2838 the source. The third argument is an (arbitrarily sized) integer argument
2839 specifying the number of bytes to copy, and the fourth argument is the alignment
2840 of the source and destination locations.
2844 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2845 the caller guarantees that the size of the copy is a multiple of the alignment
2846 and that both the source and destination pointers are aligned to that boundary.
2852 The '<tt>llvm.memmove</tt>' intrinsic copies a block of memory from the source
2853 location to the destination location, which may overlap. It
2854 copies "len" bytes of memory over. If the argument is known to be aligned to
2855 some boundary, this can be specified as the fourth argument, otherwise it should
2861 <!-- _______________________________________________________________________ -->
2862 <div class="doc_subsubsection">
2863 <a name="i_memset">'<tt>llvm.memset</tt>' Intrinsic</a>
2866 <div class="doc_text">
2870 call void (sbyte*, ubyte, uint, uint)* %llvm.memset(sbyte* <dest>, ubyte <val>,
2871 uint <len>, uint <align>)
2877 The '<tt>llvm.memset</tt>' intrinsic fills a block of memory with a particular
2882 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
2883 does not return a value, and takes an extra alignment argument.
2889 The first argument is a pointer to the destination to fill, the second is the
2890 byte value to fill it with, the third argument is an (arbitrarily sized) integer
2891 argument specifying the number of bytes to fill, and the fourth argument is the
2892 known alignment of destination location.
2896 If the call to this intrinisic has an alignment value that is not 0 or 1, then
2897 the caller guarantees that the size of the copy is a multiple of the alignment
2898 and that the destination pointer is aligned to that boundary.
2904 The '<tt>llvm.memset</tt>' intrinsic fills "len" bytes of memory starting at the
2905 destination location. If the argument is known to be aligned to some boundary,
2906 this can be specified as the fourth argument, otherwise it should be set to 0 or
2912 <!-- _______________________________________________________________________ -->
2913 <div class="doc_subsubsection">
2914 <a name="i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a>
2917 <div class="doc_text">
2921 call bool (<float or double>, <float or double>)* %llvm.isunordered(<float or double> Val1,
2922 <float or double> Val2)
2928 The '<tt>llvm.isunordered</tt>' intrinsic returns true if either or both of the
2929 specified floating point values is a NAN.
2935 The arguments are floating point numbers of the same type.
2941 If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
2949 <!-- ======================================================================= -->
2950 <div class="doc_subsection">
2951 <a name="int_debugger">Debugger Intrinsics</a>
2954 <div class="doc_text">
2956 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
2957 are described in the <a
2958 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
2959 Debugging</a> document.
2964 <!-- *********************************************************************** -->
2967 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2968 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
2969 <a href="http://validator.w3.org/check/referer"><img
2970 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
2972 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
2973 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
2974 Last modified: $Date$