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="#callingconv">Calling Conventions</a></li>
25 <li><a href="#globalvars">Global Variables</a></li>
26 <li><a href="#functionstructure">Function Structure</a></li>
29 <li><a href="#typesystem">Type System</a>
31 <li><a href="#t_primitive">Primitive Types</a>
33 <li><a href="#t_classifications">Type Classifications</a></li>
36 <li><a href="#t_derived">Derived Types</a>
38 <li><a href="#t_array">Array Type</a></li>
39 <li><a href="#t_function">Function Type</a></li>
40 <li><a href="#t_pointer">Pointer Type</a></li>
41 <li><a href="#t_struct">Structure Type</a></li>
42 <li><a href="#t_packed">Packed Type</a></li>
43 <li><a href="#t_opaque">Opaque Type</a></li>
48 <li><a href="#constants">Constants</a>
50 <li><a href="#simpleconstants">Simple Constants</a>
51 <li><a href="#aggregateconstants">Aggregate Constants</a>
52 <li><a href="#globalconstants">Global Variable and Function Addresses</a>
53 <li><a href="#undefvalues">Undefined Values</a>
54 <li><a href="#constantexprs">Constant Expressions</a>
57 <li><a href="#instref">Instruction Reference</a>
59 <li><a href="#terminators">Terminator Instructions</a>
61 <li><a href="#i_ret">'<tt>ret</tt>' Instruction</a></li>
62 <li><a href="#i_br">'<tt>br</tt>' Instruction</a></li>
63 <li><a href="#i_switch">'<tt>switch</tt>' Instruction</a></li>
64 <li><a href="#i_invoke">'<tt>invoke</tt>' Instruction</a></li>
65 <li><a href="#i_unwind">'<tt>unwind</tt>' Instruction</a></li>
66 <li><a href="#i_unreachable">'<tt>unreachable</tt>' Instruction</a></li>
69 <li><a href="#binaryops">Binary Operations</a>
71 <li><a href="#i_add">'<tt>add</tt>' Instruction</a></li>
72 <li><a href="#i_sub">'<tt>sub</tt>' Instruction</a></li>
73 <li><a href="#i_mul">'<tt>mul</tt>' Instruction</a></li>
74 <li><a href="#i_div">'<tt>div</tt>' Instruction</a></li>
75 <li><a href="#i_rem">'<tt>rem</tt>' Instruction</a></li>
76 <li><a href="#i_setcc">'<tt>set<i>cc</i></tt>' Instructions</a></li>
79 <li><a href="#bitwiseops">Bitwise Binary Operations</a>
81 <li><a href="#i_and">'<tt>and</tt>' Instruction</a></li>
82 <li><a href="#i_or">'<tt>or</tt>' Instruction</a></li>
83 <li><a href="#i_xor">'<tt>xor</tt>' Instruction</a></li>
84 <li><a href="#i_shl">'<tt>shl</tt>' Instruction</a></li>
85 <li><a href="#i_shr">'<tt>shr</tt>' Instruction</a></li>
88 <li><a href="#memoryops">Memory Access Operations</a>
90 <li><a href="#i_malloc">'<tt>malloc</tt>' Instruction</a></li>
91 <li><a href="#i_free">'<tt>free</tt>' Instruction</a></li>
92 <li><a href="#i_alloca">'<tt>alloca</tt>' Instruction</a></li>
93 <li><a href="#i_load">'<tt>load</tt>' Instruction</a></li>
94 <li><a href="#i_store">'<tt>store</tt>' Instruction</a></li>
95 <li><a href="#i_getelementptr">'<tt>getelementptr</tt>' Instruction</a></li>
98 <li><a href="#otherops">Other Operations</a>
100 <li><a href="#i_phi">'<tt>phi</tt>' Instruction</a></li>
101 <li><a href="#i_cast">'<tt>cast .. to</tt>' Instruction</a></li>
102 <li><a href="#i_select">'<tt>select</tt>' Instruction</a></li>
103 <li><a href="#i_extractelement">'<tt>extractelement</tt>' Instruction</a></li>
104 <li><a href="#i_call">'<tt>call</tt>' Instruction</a></li>
105 <li><a href="#i_vaarg">'<tt>vaarg</tt>' Instruction</a></li>
110 <li><a href="#intrinsics">Intrinsic Functions</a>
112 <li><a href="#int_varargs">Variable Argument Handling Intrinsics</a>
114 <li><a href="#i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a></li>
115 <li><a href="#i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a></li>
116 <li><a href="#i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a></li>
119 <li><a href="#int_gc">Accurate Garbage Collection Intrinsics</a>
121 <li><a href="#i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a></li>
122 <li><a href="#i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a></li>
123 <li><a href="#i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a></li>
126 <li><a href="#int_codegen">Code Generator Intrinsics</a>
128 <li><a href="#i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a></li>
129 <li><a href="#i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a></li>
130 <li><a href="#i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a></li>
131 <li><a href="#i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a></li>
132 <li><a href="#i_readcyclecounter"><tt>llvm.readcyclecounter</tt>' Intrinsic</a></li>
135 <li><a href="#int_os">Operating System Intrinsics</a>
137 <li><a href="#i_readport">'<tt>llvm.readport</tt>' Intrinsic</a></li>
138 <li><a href="#i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a></li>
139 <li><a href="#i_readio">'<tt>llvm.readio</tt>' Intrinsic</a></li>
140 <li><a href="#i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a></li>
142 <li><a href="#int_libc">Standard C Library Intrinsics</a>
144 <li><a href="#i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a></li>
145 <li><a href="#i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a></li>
146 <li><a href="#i_memset">'<tt>llvm.memset</tt>' Intrinsic</a></li>
147 <li><a href="#i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a></li>
148 <li><a href="#i_sqrt">'<tt>llvm.sqrt</tt>' Intrinsic</a></li>
152 <li><a href="#int_count">Bit counting Intrinsics</a>
154 <li><a href="#int_ctpop">'<tt>llvm.ctpop</tt>' Intrinsic </a></li>
155 <li><a href="#int_ctlz">'<tt>llvm.ctlz</tt>' Intrinsic </a></li>
156 <li><a href="#int_cttz">'<tt>llvm.cttz</tt>' Intrinsic </a></li>
159 <li><a href="#int_debugger">Debugger intrinsics</a></li>
164 <div class="doc_author">
165 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
166 and <a href="mailto:vadve@cs.uiuc.edu">Vikram Adve</a></p>
169 <!-- *********************************************************************** -->
170 <div class="doc_section"> <a name="abstract">Abstract </a></div>
171 <!-- *********************************************************************** -->
173 <div class="doc_text">
174 <p>This document is a reference manual for the LLVM assembly language.
175 LLVM is an SSA based representation that provides type safety,
176 low-level operations, flexibility, and the capability of representing
177 'all' high-level languages cleanly. It is the common code
178 representation used throughout all phases of the LLVM compilation
182 <!-- *********************************************************************** -->
183 <div class="doc_section"> <a name="introduction">Introduction</a> </div>
184 <!-- *********************************************************************** -->
186 <div class="doc_text">
188 <p>The LLVM code representation is designed to be used in three
189 different forms: as an in-memory compiler IR, as an on-disk bytecode
190 representation (suitable for fast loading by a Just-In-Time compiler),
191 and as a human readable assembly language representation. This allows
192 LLVM to provide a powerful intermediate representation for efficient
193 compiler transformations and analysis, while providing a natural means
194 to debug and visualize the transformations. The three different forms
195 of LLVM are all equivalent. This document describes the human readable
196 representation and notation.</p>
198 <p>The LLVM representation aims to be light-weight and low-level
199 while being expressive, typed, and extensible at the same time. It
200 aims to be a "universal IR" of sorts, by being at a low enough level
201 that high-level ideas may be cleanly mapped to it (similar to how
202 microprocessors are "universal IR's", allowing many source languages to
203 be mapped to them). By providing type information, LLVM can be used as
204 the target of optimizations: for example, through pointer analysis, it
205 can be proven that a C automatic variable is never accessed outside of
206 the current function... allowing it to be promoted to a simple SSA
207 value instead of a memory location.</p>
211 <!-- _______________________________________________________________________ -->
212 <div class="doc_subsubsection"> <a name="wellformed">Well-Formedness</a> </div>
214 <div class="doc_text">
216 <p>It is important to note that this document describes 'well formed'
217 LLVM assembly language. There is a difference between what the parser
218 accepts and what is considered 'well formed'. For example, the
219 following instruction is syntactically okay, but not well formed:</p>
222 %x = <a href="#i_add">add</a> int 1, %x
225 <p>...because the definition of <tt>%x</tt> does not dominate all of
226 its uses. The LLVM infrastructure provides a verification pass that may
227 be used to verify that an LLVM module is well formed. This pass is
228 automatically run by the parser after parsing input assembly and by
229 the optimizer before it outputs bytecode. The violations pointed out
230 by the verifier pass indicate bugs in transformation passes or input to
233 <!-- Describe the typesetting conventions here. --> </div>
235 <!-- *********************************************************************** -->
236 <div class="doc_section"> <a name="identifiers">Identifiers</a> </div>
237 <!-- *********************************************************************** -->
239 <div class="doc_text">
241 <p>LLVM uses three different forms of identifiers, for different
245 <li>Named values are represented as a string of characters with a '%' prefix.
246 For example, %foo, %DivisionByZero, %a.really.long.identifier. The actual
247 regular expression used is '<tt>%[a-zA-Z$._][a-zA-Z$._0-9]*</tt>'.
248 Identifiers which require other characters in their names can be surrounded
249 with quotes. In this way, anything except a <tt>"</tt> character can be used
252 <li>Unnamed values are represented as an unsigned numeric value with a '%'
253 prefix. For example, %12, %2, %44.</li>
255 <li>Constants, which are described in a <a href="#constants">section about
256 constants</a>, below.</li>
259 <p>LLVM requires that values start with a '%' sign for two reasons: Compilers
260 don't need to worry about name clashes with reserved words, and the set of
261 reserved words may be expanded in the future without penalty. Additionally,
262 unnamed identifiers allow a compiler to quickly come up with a temporary
263 variable without having to avoid symbol table conflicts.</p>
265 <p>Reserved words in LLVM are very similar to reserved words in other
266 languages. There are keywords for different opcodes ('<tt><a
267 href="#i_add">add</a></tt>', '<tt><a href="#i_cast">cast</a></tt>', '<tt><a
268 href="#i_ret">ret</a></tt>', etc...), for primitive type names ('<tt><a
269 href="#t_void">void</a></tt>', '<tt><a href="#t_uint">uint</a></tt>', etc...),
270 and others. These reserved words cannot conflict with variable names, because
271 none of them start with a '%' character.</p>
273 <p>Here is an example of LLVM code to multiply the integer variable
274 '<tt>%X</tt>' by 8:</p>
279 %result = <a href="#i_mul">mul</a> uint %X, 8
282 <p>After strength reduction:</p>
285 %result = <a href="#i_shl">shl</a> uint %X, ubyte 3
288 <p>And the hard way:</p>
291 <a href="#i_add">add</a> uint %X, %X <i>; yields {uint}:%0</i>
292 <a href="#i_add">add</a> uint %0, %0 <i>; yields {uint}:%1</i>
293 %result = <a href="#i_add">add</a> uint %1, %1
296 <p>This last way of multiplying <tt>%X</tt> by 8 illustrates several
297 important lexical features of LLVM:</p>
301 <li>Comments are delimited with a '<tt>;</tt>' and go until the end of
304 <li>Unnamed temporaries are created when the result of a computation is not
305 assigned to a named value.</li>
307 <li>Unnamed temporaries are numbered sequentially</li>
311 <p>...and it also shows a convention that we follow in this document. When
312 demonstrating instructions, we will follow an instruction with a comment that
313 defines the type and name of value produced. Comments are shown in italic
318 <!-- *********************************************************************** -->
319 <div class="doc_section"> <a name="highlevel">High Level Structure</a> </div>
320 <!-- *********************************************************************** -->
322 <!-- ======================================================================= -->
323 <div class="doc_subsection"> <a name="modulestructure">Module Structure</a>
326 <div class="doc_text">
328 <p>LLVM programs are composed of "Module"s, each of which is a
329 translation unit of the input programs. Each module consists of
330 functions, global variables, and symbol table entries. Modules may be
331 combined together with the LLVM linker, which merges function (and
332 global variable) definitions, resolves forward declarations, and merges
333 symbol table entries. Here is an example of the "hello world" module:</p>
335 <pre><i>; Declare the string constant as a global constant...</i>
336 <a href="#identifiers">%.LC0</a> = <a href="#linkage_internal">internal</a> <a
337 href="#globalvars">constant</a> <a href="#t_array">[13 x sbyte]</a> c"hello world\0A\00" <i>; [13 x sbyte]*</i>
339 <i>; External declaration of the puts function</i>
340 <a href="#functionstructure">declare</a> int %puts(sbyte*) <i>; int(sbyte*)* </i>
342 <i>; Definition of main function</i>
343 int %main() { <i>; int()* </i>
344 <i>; Convert [13x sbyte]* to sbyte *...</i>
346 href="#i_getelementptr">getelementptr</a> [13 x sbyte]* %.LC0, long 0, long 0 <i>; sbyte*</i>
348 <i>; Call puts function to write out the string to stdout...</i>
350 href="#i_call">call</a> int %puts(sbyte* %cast210) <i>; int</i>
352 href="#i_ret">ret</a> int 0<br>}<br></pre>
354 <p>This example is made up of a <a href="#globalvars">global variable</a>
355 named "<tt>.LC0</tt>", an external declaration of the "<tt>puts</tt>"
356 function, and a <a href="#functionstructure">function definition</a>
357 for "<tt>main</tt>".</p>
359 <p>In general, a module is made up of a list of global values,
360 where both functions and global variables are global values. Global values are
361 represented by a pointer to a memory location (in this case, a pointer to an
362 array of char, and a pointer to a function), and have one of the following <a
363 href="#linkage">linkage types</a>.</p>
367 <!-- ======================================================================= -->
368 <div class="doc_subsection">
369 <a name="linkage">Linkage Types</a>
372 <div class="doc_text">
375 All Global Variables and Functions have one of the following types of linkage:
380 <dt><tt><b><a name="linkage_internal">internal</a></b></tt> </dt>
382 <dd>Global values with internal linkage are only directly accessible by
383 objects in the current module. In particular, linking code into a module with
384 an internal global value may cause the internal to be renamed as necessary to
385 avoid collisions. Because the symbol is internal to the module, all
386 references can be updated. This corresponds to the notion of the
387 '<tt>static</tt>' keyword in C, or the idea of "anonymous namespaces" in C++.
390 <dt><tt><b><a name="linkage_linkonce">linkonce</a></b></tt>: </dt>
392 <dd>"<tt>linkonce</tt>" linkage is similar to <tt>internal</tt> linkage, with
393 the twist that linking together two modules defining the same
394 <tt>linkonce</tt> globals will cause one of the globals to be discarded. This
395 is typically used to implement inline functions. Unreferenced
396 <tt>linkonce</tt> globals are allowed to be discarded.
399 <dt><tt><b><a name="linkage_weak">weak</a></b></tt>: </dt>
401 <dd>"<tt>weak</tt>" linkage is exactly the same as <tt>linkonce</tt> linkage,
402 except that unreferenced <tt>weak</tt> globals may not be discarded. This is
403 used to implement constructs in C such as "<tt>int X;</tt>" at global scope.
406 <dt><tt><b><a name="linkage_appending">appending</a></b></tt>: </dt>
408 <dd>"<tt>appending</tt>" linkage may only be applied to global variables of
409 pointer to array type. When two global variables with appending linkage are
410 linked together, the two global arrays are appended together. This is the
411 LLVM, typesafe, equivalent of having the system linker append together
412 "sections" with identical names when .o files are linked.
415 <dt><tt><b><a name="linkage_external">externally visible</a></b></tt>:</dt>
417 <dd>If none of the above identifiers are used, the global is externally
418 visible, meaning that it participates in linkage and can be used to resolve
419 external symbol references.
423 <p><a name="linkage_external">For example, since the "<tt>.LC0</tt>"
424 variable is defined to be internal, if another module defined a "<tt>.LC0</tt>"
425 variable and was linked with this one, one of the two would be renamed,
426 preventing a collision. Since "<tt>main</tt>" and "<tt>puts</tt>" are
427 external (i.e., lacking any linkage declarations), they are accessible
428 outside of the current module. It is illegal for a function <i>declaration</i>
429 to have any linkage type other than "externally visible".</a></p>
433 <!-- ======================================================================= -->
434 <div class="doc_subsection">
435 <a name="callingconv">Calling Conventions</a>
438 <div class="doc_text">
440 <p>LLVM <a href="#functionstructure">functions</a>, <a href="#i_call">calls</a>
441 and <a href="#i_invoke">invokes</a> can all have an optional calling convention
442 specified for the call. The calling convention of any pair of dynamic
443 caller/callee must match, or the behavior of the program is undefined. The
444 following calling conventions are supported by LLVM, and more may be added in
448 <dt><b>"<tt>ccc</tt>" - The C calling convention</b>:</dt>
450 <dd>This calling convention (the default if no other calling convention is
451 specified) matches the target C calling conventions. This calling convention
452 supports varargs function calls and tolerates some mismatch in the declared
453 prototype and implemented declaration of the function (as does normal C).
456 <dt><b>"<tt>fastcc</tt>" - The fast calling convention</b>:</dt>
458 <dd>This calling convention attempts to make calls as fast as possible
459 (e.g. by passing things in registers). This calling convention allows the
460 target to use whatever tricks it wants to produce fast code for the target,
461 without having to conform to an externally specified ABI. Implementations of
462 this convention should allow arbitrary tail call optimization to be supported.
463 This calling convention does not support varargs and requires the prototype of
464 all callees to exactly match the prototype of the function definition.
467 <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
469 <dd>This calling convention attempts to make code in the caller as efficient
470 as possible under the assumption that the call is not commonly executed. As
471 such, these calls often preserve all registers so that the call does not break
472 any live ranges in the caller side. This calling convention does not support
473 varargs and requires the prototype of all callees to exactly match the
474 prototype of the function definition.
477 <dt><b>"<tt>cc <<em>n</em>></tt>" - Numbered convention</b>:</dt>
479 <dd>Any calling convention may be specified by number, allowing
480 target-specific calling conventions to be used. Target specific calling
481 conventions start at 64.
485 <p>More calling conventions can be added/defined on an as-needed basis, to
486 support pascal conventions or any other well-known target-independent
491 <!-- ======================================================================= -->
492 <div class="doc_subsection">
493 <a name="globalvars">Global Variables</a>
496 <div class="doc_text">
498 <p>Global variables define regions of memory allocated at compilation time
499 instead of run-time. Global variables may optionally be initialized, may have
500 an explicit section to be placed in, and may
501 have an optional explicit alignment specified. A
502 variable may be defined as a global "constant," which indicates that the
503 contents of the variable will <b>never</b> be modified (enabling better
504 optimization, allowing the global data to be placed in the read-only section of
505 an executable, etc). Note that variables that need runtime initialization
506 cannot be marked "constant" as there is a store to the variable.</p>
509 LLVM explicitly allows <em>declarations</em> of global variables to be marked
510 constant, even if the final definition of the global is not. This capability
511 can be used to enable slightly better optimization of the program, but requires
512 the language definition to guarantee that optimizations based on the
513 'constantness' are valid for the translation units that do not include the
517 <p>As SSA values, global variables define pointer values that are in
518 scope (i.e. they dominate) all basic blocks in the program. Global
519 variables always define a pointer to their "content" type because they
520 describe a region of memory, and all memory objects in LLVM are
521 accessed through pointers.</p>
523 <p>LLVM allows an explicit section to be specified for globals. If the target
524 supports it, it will emit globals to the section specified.</p>
526 <p>An explicit alignment may be specified for a global. If not present, or if
527 the alignment is set to zero, the alignment of the global is set by the target
528 to whatever it feels convenient. If an explicit alignment is specified, the
529 global is forced to have at least that much alignment. All alignments must be
535 <!-- ======================================================================= -->
536 <div class="doc_subsection">
537 <a name="functionstructure">Functions</a>
540 <div class="doc_text">
542 <p>LLVM function definitions consist of an optional <a href="#linkage">linkage
543 type</a>, an optional <a href="#callingconv">calling convention</a>, a return
544 type, a function name, a (possibly empty) argument list, an optional section,
545 an optional alignment, an opening curly brace,
546 a list of basic blocks, and a closing curly brace. LLVM function declarations
547 are defined with the "<tt>declare</tt>" keyword, an optional <a
548 href="#callingconv">calling convention</a>, a return type, a function name,
549 a possibly empty list of arguments, and an optional alignment.</p>
551 <p>A function definition contains a list of basic blocks, forming the CFG for
552 the function. Each basic block may optionally start with a label (giving the
553 basic block a symbol table entry), contains a list of instructions, and ends
554 with a <a href="#terminators">terminator</a> instruction (such as a branch or
555 function return).</p>
557 <p>The first basic block in a program is special in two ways: it is immediately
558 executed on entrance to the function, and it is not allowed to have predecessor
559 basic blocks (i.e. there can not be any branches to the entry block of a
560 function). Because the block can have no predecessors, it also cannot have any
561 <a href="#i_phi">PHI nodes</a>.</p>
563 <p>LLVM functions are identified by their name and type signature. Hence, two
564 functions with the same name but different parameter lists or return values are
565 considered different functions, and LLVM will resolve references to each
568 <p>LLVM allows an explicit section to be specified for functions. If the target
569 supports it, it will emit functions to the section specified.</p>
571 <p>An explicit alignment may be specified for a function. If not present, or if
572 the alignment is set to zero, the alignment of the function is set by the target
573 to whatever it feels convenient. If an explicit alignment is specified, the
574 function is forced to have at least that much alignment. All alignments must be
581 <!-- *********************************************************************** -->
582 <div class="doc_section"> <a name="typesystem">Type System</a> </div>
583 <!-- *********************************************************************** -->
585 <div class="doc_text">
587 <p>The LLVM type system is one of the most important features of the
588 intermediate representation. Being typed enables a number of
589 optimizations to be performed on the IR directly, without having to do
590 extra analyses on the side before the transformation. A strong type
591 system makes it easier to read the generated code and enables novel
592 analyses and transformations that are not feasible to perform on normal
593 three address code representations.</p>
597 <!-- ======================================================================= -->
598 <div class="doc_subsection"> <a name="t_primitive">Primitive Types</a> </div>
599 <div class="doc_text">
600 <p>The primitive types are the fundamental building blocks of the LLVM
601 system. The current set of primitive types is as follows:</p>
603 <table class="layout">
608 <tr><th>Type</th><th>Description</th></tr>
609 <tr><td><tt>void</tt></td><td>No value</td></tr>
610 <tr><td><tt>ubyte</tt></td><td>Unsigned 8-bit value</td></tr>
611 <tr><td><tt>ushort</tt></td><td>Unsigned 16-bit value</td></tr>
612 <tr><td><tt>uint</tt></td><td>Unsigned 32-bit value</td></tr>
613 <tr><td><tt>ulong</tt></td><td>Unsigned 64-bit value</td></tr>
614 <tr><td><tt>float</tt></td><td>32-bit floating point value</td></tr>
615 <tr><td><tt>label</tt></td><td>Branch destination</td></tr>
622 <tr><th>Type</th><th>Description</th></tr>
623 <tr><td><tt>bool</tt></td><td>True or False value</td></tr>
624 <tr><td><tt>sbyte</tt></td><td>Signed 8-bit value</td></tr>
625 <tr><td><tt>short</tt></td><td>Signed 16-bit value</td></tr>
626 <tr><td><tt>int</tt></td><td>Signed 32-bit value</td></tr>
627 <tr><td><tt>long</tt></td><td>Signed 64-bit value</td></tr>
628 <tr><td><tt>double</tt></td><td>64-bit floating point value</td></tr>
636 <!-- _______________________________________________________________________ -->
637 <div class="doc_subsubsection"> <a name="t_classifications">Type
638 Classifications</a> </div>
639 <div class="doc_text">
640 <p>These different primitive types fall into a few useful
643 <table border="1" cellspacing="0" cellpadding="4">
645 <tr><th>Classification</th><th>Types</th></tr>
647 <td><a name="t_signed">signed</a></td>
648 <td><tt>sbyte, short, int, long, float, double</tt></td>
651 <td><a name="t_unsigned">unsigned</a></td>
652 <td><tt>ubyte, ushort, uint, ulong</tt></td>
655 <td><a name="t_integer">integer</a></td>
656 <td><tt>ubyte, sbyte, ushort, short, uint, int, ulong, long</tt></td>
659 <td><a name="t_integral">integral</a></td>
660 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long</tt>
664 <td><a name="t_floating">floating point</a></td>
665 <td><tt>float, double</tt></td>
668 <td><a name="t_firstclass">first class</a></td>
669 <td><tt>bool, ubyte, sbyte, ushort, short, uint, int, ulong, long,<br>
670 float, double, <a href="#t_pointer">pointer</a>,
671 <a href="#t_packed">packed</a></tt></td>
676 <p>The <a href="#t_firstclass">first class</a> types are perhaps the
677 most important. Values of these types are the only ones which can be
678 produced by instructions, passed as arguments, or used as operands to
679 instructions. This means that all structures and arrays must be
680 manipulated either by pointer or by component.</p>
683 <!-- ======================================================================= -->
684 <div class="doc_subsection"> <a name="t_derived">Derived Types</a> </div>
686 <div class="doc_text">
688 <p>The real power in LLVM comes from the derived types in the system.
689 This is what allows a programmer to represent arrays, functions,
690 pointers, and other useful types. Note that these derived types may be
691 recursive: For example, it is possible to have a two dimensional array.</p>
695 <!-- _______________________________________________________________________ -->
696 <div class="doc_subsubsection"> <a name="t_array">Array Type</a> </div>
698 <div class="doc_text">
702 <p>The array type is a very simple derived type that arranges elements
703 sequentially in memory. The array type requires a size (number of
704 elements) and an underlying data type.</p>
709 [<# elements> x <elementtype>]
712 <p>The number of elements is a constant integer value; elementtype may
713 be any type with a size.</p>
716 <table class="layout">
719 <tt>[40 x int ]</tt><br/>
720 <tt>[41 x int ]</tt><br/>
721 <tt>[40 x uint]</tt><br/>
724 Array of 40 integer values.<br/>
725 Array of 41 integer values.<br/>
726 Array of 40 unsigned integer values.<br/>
730 <p>Here are some examples of multidimensional arrays:</p>
731 <table class="layout">
734 <tt>[3 x [4 x int]]</tt><br/>
735 <tt>[12 x [10 x float]]</tt><br/>
736 <tt>[2 x [3 x [4 x uint]]]</tt><br/>
739 3x4 array of integer values.<br/>
740 12x10 array of single precision floating point values.<br/>
741 2x3x4 array of unsigned integer values.<br/>
746 <p>Note that 'variable sized arrays' can be implemented in LLVM with a zero
747 length array. Normally, accesses past the end of an array are undefined in
748 LLVM (e.g. it is illegal to access the 5th element of a 3 element array).
749 As a special case, however, zero length arrays are recognized to be variable
750 length. This allows implementation of 'pascal style arrays' with the LLVM
751 type "{ int, [0 x float]}", for example.</p>
755 <!-- _______________________________________________________________________ -->
756 <div class="doc_subsubsection"> <a name="t_function">Function Type</a> </div>
757 <div class="doc_text">
759 <p>The function type can be thought of as a function signature. It
760 consists of a return type and a list of formal parameter types.
761 Function types are usually used to build virtual function tables
762 (which are structures of pointers to functions), for indirect function
763 calls, and when defining a function.</p>
765 The return type of a function type cannot be an aggregate type.
768 <pre> <returntype> (<parameter list>)<br></pre>
769 <p>...where '<tt><parameter list></tt>' is a comma-separated list of type
770 specifiers. Optionally, the parameter list may include a type <tt>...</tt>,
771 which indicates that the function takes a variable number of arguments.
772 Variable argument functions can access their arguments with the <a
773 href="#int_varargs">variable argument handling intrinsic</a> functions.</p>
775 <table class="layout">
778 <tt>int (int)</tt> <br/>
779 <tt>float (int, int *) *</tt><br/>
780 <tt>int (sbyte *, ...)</tt><br/>
783 function taking an <tt>int</tt>, returning an <tt>int</tt><br/>
784 <a href="#t_pointer">Pointer</a> to a function that takes an
785 <tt>int</tt> and a <a href="#t_pointer">pointer</a> to <tt>int</tt>,
786 returning <tt>float</tt>.<br/>
787 A vararg function that takes at least one <a href="#t_pointer">pointer</a>
788 to <tt>sbyte</tt> (signed char in C), which returns an integer. This is
789 the signature for <tt>printf</tt> in LLVM.<br/>
795 <!-- _______________________________________________________________________ -->
796 <div class="doc_subsubsection"> <a name="t_struct">Structure Type</a> </div>
797 <div class="doc_text">
799 <p>The structure type is used to represent a collection of data members
800 together in memory. The packing of the field types is defined to match
801 the ABI of the underlying processor. The elements of a structure may
802 be any type that has a size.</p>
803 <p>Structures are accessed using '<tt><a href="#i_load">load</a></tt>
804 and '<tt><a href="#i_store">store</a></tt>' by getting a pointer to a
805 field with the '<tt><a href="#i_getelementptr">getelementptr</a></tt>'
808 <pre> { <type list> }<br></pre>
810 <table class="layout">
813 <tt>{ int, int, int }</tt><br/>
814 <tt>{ float, int (int) * }</tt><br/>
817 a triple of three <tt>int</tt> values<br/>
818 A pair, where the first element is a <tt>float</tt> and the second element
819 is a <a href="#t_pointer">pointer</a> to a <a href="#t_function">function</a>
820 that takes an <tt>int</tt>, returning an <tt>int</tt>.<br/>
826 <!-- _______________________________________________________________________ -->
827 <div class="doc_subsubsection"> <a name="t_pointer">Pointer Type</a> </div>
828 <div class="doc_text">
830 <p>As in many languages, the pointer type represents a pointer or
831 reference to another object, which must live in memory.</p>
833 <pre> <type> *<br></pre>
835 <table class="layout">
838 <tt>[4x int]*</tt><br/>
839 <tt>int (int *) *</tt><br/>
842 A <a href="#t_pointer">pointer</a> to <a href="#t_array">array</a> of
843 four <tt>int</tt> values<br/>
844 A <a href="#t_pointer">pointer</a> to a <a
845 href="#t_function">function</a> that takes an <tt>int*</tt>, returning an
852 <!-- _______________________________________________________________________ -->
853 <div class="doc_subsubsection"> <a name="t_packed">Packed Type</a> </div>
854 <div class="doc_text">
858 <p>A packed type is a simple derived type that represents a vector
859 of elements. Packed types are used when multiple primitive data
860 are operated in parallel using a single instruction (SIMD).
861 A packed type requires a size (number of
862 elements) and an underlying primitive data type. Vectors must have a power
863 of two length (1, 2, 4, 8, 16 ...). Packed types are
864 considered <a href="#t_firstclass">first class</a>.</p>
869 < <# elements> x <elementtype> >
872 <p>The number of elements is a constant integer value; elementtype may
873 be any integral or floating point type.</p>
877 <table class="layout">
880 <tt><4 x int></tt><br/>
881 <tt><8 x float></tt><br/>
882 <tt><2 x uint></tt><br/>
885 Packed vector of 4 integer values.<br/>
886 Packed vector of 8 floating-point values.<br/>
887 Packed vector of 2 unsigned integer values.<br/>
893 <!-- _______________________________________________________________________ -->
894 <div class="doc_subsubsection"> <a name="t_opaque">Opaque Type</a> </div>
895 <div class="doc_text">
899 <p>Opaque types are used to represent unknown types in the system. This
900 corresponds (for example) to the C notion of a foward declared structure type.
901 In LLVM, opaque types can eventually be resolved to any type (not just a
912 <table class="layout">
925 <!-- *********************************************************************** -->
926 <div class="doc_section"> <a name="constants">Constants</a> </div>
927 <!-- *********************************************************************** -->
929 <div class="doc_text">
931 <p>LLVM has several different basic types of constants. This section describes
932 them all and their syntax.</p>
936 <!-- ======================================================================= -->
937 <div class="doc_subsection"><a name="simpleconstants">Simple Constants</a></div>
939 <div class="doc_text">
942 <dt><b>Boolean constants</b></dt>
944 <dd>The two strings '<tt>true</tt>' and '<tt>false</tt>' are both valid
945 constants of the <tt><a href="#t_primitive">bool</a></tt> type.
948 <dt><b>Integer constants</b></dt>
950 <dd>Standard integers (such as '4') are constants of the <a
951 href="#t_integer">integer</a> type. Negative numbers may be used with signed
955 <dt><b>Floating point constants</b></dt>
957 <dd>Floating point constants use standard decimal notation (e.g. 123.421),
958 exponential notation (e.g. 1.23421e+2), or a more precise hexadecimal
959 notation (see below). Floating point constants must have a <a
960 href="#t_floating">floating point</a> type. </dd>
962 <dt><b>Null pointer constants</b></dt>
964 <dd>The identifier '<tt>null</tt>' is recognized as a null pointer constant
965 and must be of <a href="#t_pointer">pointer type</a>.</dd>
969 <p>The one non-intuitive notation for constants is the optional hexadecimal form
970 of floating point constants. For example, the form '<tt>double
971 0x432ff973cafa8000</tt>' is equivalent to (but harder to read than) '<tt>double
972 4.5e+15</tt>'. The only time hexadecimal floating point constants are required
973 (and the only time that they are generated by the disassembler) is when a
974 floating point constant must be emitted but it cannot be represented as a
975 decimal floating point number. For example, NaN's, infinities, and other
976 special values are represented in their IEEE hexadecimal format so that
977 assembly and disassembly do not cause any bits to change in the constants.</p>
981 <!-- ======================================================================= -->
982 <div class="doc_subsection"><a name="aggregateconstants">Aggregate Constants</a>
985 <div class="doc_text">
986 <p>Aggregate constants arise from aggregation of simple constants
987 and smaller aggregate constants.</p>
990 <dt><b>Structure constants</b></dt>
992 <dd>Structure constants are represented with notation similar to structure
993 type definitions (a comma separated list of elements, surrounded by braces
994 (<tt>{}</tt>)). For example: "<tt>{ int 4, float 17.0, int* %G }</tt>",
995 where "<tt>%G</tt>" is declared as "<tt>%G = external global int</tt>". Structure constants
996 must have <a href="#t_struct">structure type</a>, and the number and
997 types of elements must match those specified by the type.
1000 <dt><b>Array constants</b></dt>
1002 <dd>Array constants are represented with notation similar to array type
1003 definitions (a comma separated list of elements, surrounded by square brackets
1004 (<tt>[]</tt>)). For example: "<tt>[ int 42, int 11, int 74 ]</tt>". Array
1005 constants must have <a href="#t_array">array type</a>, and the number and
1006 types of elements must match those specified by the type.
1009 <dt><b>Packed constants</b></dt>
1011 <dd>Packed constants are represented with notation similar to packed type
1012 definitions (a comma separated list of elements, surrounded by
1013 less-than/greater-than's (<tt><></tt>)). For example: "<tt>< int 42,
1014 int 11, int 74, int 100 ></tt>". Packed constants must have <a
1015 href="#t_packed">packed type</a>, and the number and types of elements must
1016 match those specified by the type.
1019 <dt><b>Zero initialization</b></dt>
1021 <dd>The string '<tt>zeroinitializer</tt>' can be used to zero initialize a
1022 value to zero of <em>any</em> type, including scalar and aggregate types.
1023 This is often used to avoid having to print large zero initializers (e.g. for
1024 large arrays) and is always exactly equivalent to using explicit zero
1031 <!-- ======================================================================= -->
1032 <div class="doc_subsection">
1033 <a name="globalconstants">Global Variable and Function Addresses</a>
1036 <div class="doc_text">
1038 <p>The addresses of <a href="#globalvars">global variables</a> and <a
1039 href="#functionstructure">functions</a> are always implicitly valid (link-time)
1040 constants. These constants are explicitly referenced when the <a
1041 href="#identifiers">identifier for the global</a> is used and always have <a
1042 href="#t_pointer">pointer</a> type. For example, the following is a legal LLVM
1048 %Z = global [2 x int*] [ int* %X, int* %Y ]
1053 <!-- ======================================================================= -->
1054 <div class="doc_subsection"><a name="undefvalues">Undefined Values</a></div>
1055 <div class="doc_text">
1056 <p>The string '<tt>undef</tt>' is recognized as a type-less constant that has
1057 no specific value. Undefined values may be of any type and be used anywhere
1058 a constant is permitted.</p>
1060 <p>Undefined values indicate to the compiler that the program is well defined
1061 no matter what value is used, giving the compiler more freedom to optimize.
1065 <!-- ======================================================================= -->
1066 <div class="doc_subsection"><a name="constantexprs">Constant Expressions</a>
1069 <div class="doc_text">
1071 <p>Constant expressions are used to allow expressions involving other constants
1072 to be used as constants. Constant expressions may be of any <a
1073 href="#t_firstclass">first class</a> type and may involve any LLVM operation
1074 that does not have side effects (e.g. load and call are not supported). The
1075 following is the syntax for constant expressions:</p>
1078 <dt><b><tt>cast ( CST to TYPE )</tt></b></dt>
1080 <dd>Cast a constant to another type.</dd>
1082 <dt><b><tt>getelementptr ( CSTPTR, IDX0, IDX1, ... )</tt></b></dt>
1084 <dd>Perform the <a href="#i_getelementptr">getelementptr operation</a> on
1085 constants. As with the <a href="#i_getelementptr">getelementptr</a>
1086 instruction, the index list may have zero or more indexes, which are required
1087 to make sense for the type of "CSTPTR".</dd>
1089 <dt><b><tt>select ( COND, VAL1, VAL2 )</tt></b></dt>
1091 <dd>Perform the <a href="#i_select">select operation</a> on
1094 <dt><b><tt>extractelement ( VAL, IDX )</tt></b></dt>
1096 <dd>Perform the <a href="#i_extractelement">extractelement
1097 operation</a> on constants.
1099 <dt><b><tt>OPCODE ( LHS, RHS )</tt></b></dt>
1101 <dd>Perform the specified operation of the LHS and RHS constants. OPCODE may
1102 be any of the <a href="#binaryops">binary</a> or <a href="#bitwiseops">bitwise
1103 binary</a> operations. The constraints on operands are the same as those for
1104 the corresponding instruction (e.g. no bitwise operations on floating point
1105 values are allowed).</dd>
1109 <!-- *********************************************************************** -->
1110 <div class="doc_section"> <a name="instref">Instruction Reference</a> </div>
1111 <!-- *********************************************************************** -->
1113 <div class="doc_text">
1115 <p>The LLVM instruction set consists of several different
1116 classifications of instructions: <a href="#terminators">terminator
1117 instructions</a>, <a href="#binaryops">binary instructions</a>,
1118 <a href="#bitwiseops">bitwise binary instructions</a>, <a
1119 href="#memoryops">memory instructions</a>, and <a href="#otherops">other
1120 instructions</a>.</p>
1124 <!-- ======================================================================= -->
1125 <div class="doc_subsection"> <a name="terminators">Terminator
1126 Instructions</a> </div>
1128 <div class="doc_text">
1130 <p>As mentioned <a href="#functionstructure">previously</a>, every
1131 basic block in a program ends with a "Terminator" instruction, which
1132 indicates which block should be executed after the current block is
1133 finished. These terminator instructions typically yield a '<tt>void</tt>'
1134 value: they produce control flow, not values (the one exception being
1135 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction).</p>
1136 <p>There are six different terminator instructions: the '<a
1137 href="#i_ret"><tt>ret</tt></a>' instruction, the '<a href="#i_br"><tt>br</tt></a>'
1138 instruction, the '<a href="#i_switch"><tt>switch</tt></a>' instruction,
1139 the '<a href="#i_invoke"><tt>invoke</tt></a>' instruction, the '<a
1140 href="#i_unwind"><tt>unwind</tt></a>' instruction, and the '<a
1141 href="#i_unreachable"><tt>unreachable</tt></a>' instruction.</p>
1145 <!-- _______________________________________________________________________ -->
1146 <div class="doc_subsubsection"> <a name="i_ret">'<tt>ret</tt>'
1147 Instruction</a> </div>
1148 <div class="doc_text">
1150 <pre> ret <type> <value> <i>; Return a value from a non-void function</i>
1151 ret void <i>; Return from void function</i>
1154 <p>The '<tt>ret</tt>' instruction is used to return control flow (and a
1155 value) from a function back to the caller.</p>
1156 <p>There are two forms of the '<tt>ret</tt>' instruction: one that
1157 returns a value and then causes control flow, and one that just causes
1158 control flow to occur.</p>
1160 <p>The '<tt>ret</tt>' instruction may return any '<a
1161 href="#t_firstclass">first class</a>' type. Notice that a function is
1162 not <a href="#wellformed">well formed</a> if there exists a '<tt>ret</tt>'
1163 instruction inside of the function that returns a value that does not
1164 match the return type of the function.</p>
1166 <p>When the '<tt>ret</tt>' instruction is executed, control flow
1167 returns back to the calling function's context. If the caller is a "<a
1168 href="#i_call"><tt>call</tt></a>" instruction, execution continues at
1169 the instruction after the call. If the caller was an "<a
1170 href="#i_invoke"><tt>invoke</tt></a>" instruction, execution continues
1171 at the beginning of the "normal" destination block. If the instruction
1172 returns a value, that value shall set the call or invoke instruction's
1175 <pre> ret int 5 <i>; Return an integer value of 5</i>
1176 ret void <i>; Return from a void function</i>
1179 <!-- _______________________________________________________________________ -->
1180 <div class="doc_subsubsection"> <a name="i_br">'<tt>br</tt>' Instruction</a> </div>
1181 <div class="doc_text">
1183 <pre> br bool <cond>, label <iftrue>, label <iffalse><br> br label <dest> <i>; Unconditional branch</i>
1186 <p>The '<tt>br</tt>' instruction is used to cause control flow to
1187 transfer to a different basic block in the current function. There are
1188 two forms of this instruction, corresponding to a conditional branch
1189 and an unconditional branch.</p>
1191 <p>The conditional branch form of the '<tt>br</tt>' instruction takes a
1192 single '<tt>bool</tt>' value and two '<tt>label</tt>' values. The
1193 unconditional form of the '<tt>br</tt>' instruction takes a single '<tt>label</tt>'
1194 value as a target.</p>
1196 <p>Upon execution of a conditional '<tt>br</tt>' instruction, the '<tt>bool</tt>'
1197 argument is evaluated. If the value is <tt>true</tt>, control flows
1198 to the '<tt>iftrue</tt>' <tt>label</tt> argument. If "cond" is <tt>false</tt>,
1199 control flows to the '<tt>iffalse</tt>' <tt>label</tt> argument.</p>
1201 <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
1202 href="#i_ret">ret</a> int 1<br>IfUnequal:<br> <a href="#i_ret">ret</a> int 0<br></pre>
1204 <!-- _______________________________________________________________________ -->
1205 <div class="doc_subsubsection">
1206 <a name="i_switch">'<tt>switch</tt>' Instruction</a>
1209 <div class="doc_text">
1213 switch <intty> <value>, label <defaultdest> [ <intty> <val>, label <dest> ... ]
1218 <p>The '<tt>switch</tt>' instruction is used to transfer control flow to one of
1219 several different places. It is a generalization of the '<tt>br</tt>'
1220 instruction, allowing a branch to occur to one of many possible
1226 <p>The '<tt>switch</tt>' instruction uses three parameters: an integer
1227 comparison value '<tt>value</tt>', a default '<tt>label</tt>' destination, and
1228 an array of pairs of comparison value constants and '<tt>label</tt>'s. The
1229 table is not allowed to contain duplicate constant entries.</p>
1233 <p>The <tt>switch</tt> instruction specifies a table of values and
1234 destinations. When the '<tt>switch</tt>' instruction is executed, this
1235 table is searched for the given value. If the value is found, control flow is
1236 transfered to the corresponding destination; otherwise, control flow is
1237 transfered to the default destination.</p>
1239 <h5>Implementation:</h5>
1241 <p>Depending on properties of the target machine and the particular
1242 <tt>switch</tt> instruction, this instruction may be code generated in different
1243 ways. For example, it could be generated as a series of chained conditional
1244 branches or with a lookup table.</p>
1249 <i>; Emulate a conditional br instruction</i>
1250 %Val = <a href="#i_cast">cast</a> bool %value to int
1251 switch int %Val, label %truedest [int 0, label %falsedest ]
1253 <i>; Emulate an unconditional br instruction</i>
1254 switch uint 0, label %dest [ ]
1256 <i>; Implement a jump table:</i>
1257 switch uint %val, label %otherwise [ uint 0, label %onzero
1258 uint 1, label %onone
1259 uint 2, label %ontwo ]
1263 <!-- _______________________________________________________________________ -->
1264 <div class="doc_subsubsection">
1265 <a name="i_invoke">'<tt>invoke</tt>' Instruction</a>
1268 <div class="doc_text">
1273 <result> = invoke [<a href="#callingconv">cconv</a>] <ptr to function ty> %<function ptr val>(<function args>)
1274 to label <normal label> except label <exception label>
1279 <p>The '<tt>invoke</tt>' instruction causes control to transfer to a specified
1280 function, with the possibility of control flow transfer to either the
1281 '<tt>normal</tt>' label or the
1282 '<tt>exception</tt>' label. If the callee function returns with the
1283 "<tt><a href="#i_ret">ret</a></tt>" instruction, control flow will return to the
1284 "normal" label. If the callee (or any indirect callees) returns with the "<a
1285 href="#i_unwind"><tt>unwind</tt></a>" instruction, control is interrupted and
1286 continued at the dynamically nearest "exception" label.</p>
1290 <p>This instruction requires several arguments:</p>
1294 The optional "cconv" marker indicates which <a href="callingconv">calling
1295 convention</a> the call should use. If none is specified, the call defaults
1296 to using C calling conventions.
1298 <li>'<tt>ptr to function ty</tt>': shall be the signature of the pointer to
1299 function value being invoked. In most cases, this is a direct function
1300 invocation, but indirect <tt>invoke</tt>s are just as possible, branching off
1301 an arbitrary pointer to function value.
1304 <li>'<tt>function ptr val</tt>': An LLVM value containing a pointer to a
1305 function to be invoked. </li>
1307 <li>'<tt>function args</tt>': argument list whose types match the function
1308 signature argument types. If the function signature indicates the function
1309 accepts a variable number of arguments, the extra arguments can be
1312 <li>'<tt>normal label</tt>': the label reached when the called function
1313 executes a '<tt><a href="#i_ret">ret</a></tt>' instruction. </li>
1315 <li>'<tt>exception label</tt>': the label reached when a callee returns with
1316 the <a href="#i_unwind"><tt>unwind</tt></a> instruction. </li>
1322 <p>This instruction is designed to operate as a standard '<tt><a
1323 href="#i_call">call</a></tt>' instruction in most regards. The primary
1324 difference is that it establishes an association with a label, which is used by
1325 the runtime library to unwind the stack.</p>
1327 <p>This instruction is used in languages with destructors to ensure that proper
1328 cleanup is performed in the case of either a <tt>longjmp</tt> or a thrown
1329 exception. Additionally, this is important for implementation of
1330 '<tt>catch</tt>' clauses in high-level languages that support them.</p>
1334 %retval = invoke int %Test(int 15) to label %Continue
1335 except label %TestCleanup <i>; {int}:retval set</i>
1336 %retval = invoke <a href="#callingconv">coldcc</a> int %Test(int 15) to label %Continue
1337 except label %TestCleanup <i>; {int}:retval set</i>
1342 <!-- _______________________________________________________________________ -->
1344 <div class="doc_subsubsection"> <a name="i_unwind">'<tt>unwind</tt>'
1345 Instruction</a> </div>
1347 <div class="doc_text">
1356 <p>The '<tt>unwind</tt>' instruction unwinds the stack, continuing control flow
1357 at the first callee in the dynamic call stack which used an <a
1358 href="#i_invoke"><tt>invoke</tt></a> instruction to perform the call. This is
1359 primarily used to implement exception handling.</p>
1363 <p>The '<tt>unwind</tt>' intrinsic causes execution of the current function to
1364 immediately halt. The dynamic call stack is then searched for the first <a
1365 href="#i_invoke"><tt>invoke</tt></a> instruction on the call stack. Once found,
1366 execution continues at the "exceptional" destination block specified by the
1367 <tt>invoke</tt> instruction. If there is no <tt>invoke</tt> instruction in the
1368 dynamic call chain, undefined behavior results.</p>
1371 <!-- _______________________________________________________________________ -->
1373 <div class="doc_subsubsection"> <a name="i_unreachable">'<tt>unreachable</tt>'
1374 Instruction</a> </div>
1376 <div class="doc_text">
1385 <p>The '<tt>unreachable</tt>' instruction has no defined semantics. This
1386 instruction is used to inform the optimizer that a particular portion of the
1387 code is not reachable. This can be used to indicate that the code after a
1388 no-return function cannot be reached, and other facts.</p>
1392 <p>The '<tt>unreachable</tt>' instruction has no defined semantics.</p>
1397 <!-- ======================================================================= -->
1398 <div class="doc_subsection"> <a name="binaryops">Binary Operations</a> </div>
1399 <div class="doc_text">
1400 <p>Binary operators are used to do most of the computation in a
1401 program. They require two operands, execute an operation on them, and
1402 produce a single value. The operands might represent
1403 multiple data, as is the case with the <a href="#t_packed">packed</a> data type.
1404 The result value of a binary operator is not
1405 necessarily the same type as its operands.</p>
1406 <p>There are several different binary operators:</p>
1408 <!-- _______________________________________________________________________ -->
1409 <div class="doc_subsubsection"> <a name="i_add">'<tt>add</tt>'
1410 Instruction</a> </div>
1411 <div class="doc_text">
1413 <pre> <result> = add <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1416 <p>The '<tt>add</tt>' instruction returns the sum of its two operands.</p>
1418 <p>The two arguments to the '<tt>add</tt>' instruction must be either <a
1419 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a> values.
1420 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1421 Both arguments must have identical types.</p>
1423 <p>The value produced is the integer or floating point sum of the two
1426 <pre> <result> = add int 4, %var <i>; yields {int}:result = 4 + %var</i>
1429 <!-- _______________________________________________________________________ -->
1430 <div class="doc_subsubsection"> <a name="i_sub">'<tt>sub</tt>'
1431 Instruction</a> </div>
1432 <div class="doc_text">
1434 <pre> <result> = sub <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1437 <p>The '<tt>sub</tt>' instruction returns the difference of its two
1439 <p>Note that the '<tt>sub</tt>' instruction is used to represent the '<tt>neg</tt>'
1440 instruction present in most other intermediate representations.</p>
1442 <p>The two arguments to the '<tt>sub</tt>' instruction must be either <a
1443 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1445 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1446 Both arguments must have identical types.</p>
1448 <p>The value produced is the integer or floating point difference of
1449 the two operands.</p>
1451 <pre> <result> = sub int 4, %var <i>; yields {int}:result = 4 - %var</i>
1452 <result> = sub int 0, %val <i>; yields {int}:result = -%var</i>
1455 <!-- _______________________________________________________________________ -->
1456 <div class="doc_subsubsection"> <a name="i_mul">'<tt>mul</tt>'
1457 Instruction</a> </div>
1458 <div class="doc_text">
1460 <pre> <result> = mul <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1463 <p>The '<tt>mul</tt>' instruction returns the product of its two
1466 <p>The two arguments to the '<tt>mul</tt>' instruction must be either <a
1467 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1469 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1470 Both arguments must have identical types.</p>
1472 <p>The value produced is the integer or floating point product of the
1474 <p>There is no signed vs unsigned multiplication. The appropriate
1475 action is taken based on the type of the operand.</p>
1477 <pre> <result> = mul int 4, %var <i>; yields {int}:result = 4 * %var</i>
1480 <!-- _______________________________________________________________________ -->
1481 <div class="doc_subsubsection"> <a name="i_div">'<tt>div</tt>'
1482 Instruction</a> </div>
1483 <div class="doc_text">
1485 <pre> <result> = div <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1488 <p>The '<tt>div</tt>' instruction returns the quotient of its two
1491 <p>The two arguments to the '<tt>div</tt>' instruction must be either <a
1492 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1494 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1495 Both arguments must have identical types.</p>
1497 <p>The value produced is the integer or floating point quotient of the
1500 <pre> <result> = div int 4, %var <i>; yields {int}:result = 4 / %var</i>
1503 <!-- _______________________________________________________________________ -->
1504 <div class="doc_subsubsection"> <a name="i_rem">'<tt>rem</tt>'
1505 Instruction</a> </div>
1506 <div class="doc_text">
1508 <pre> <result> = rem <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1511 <p>The '<tt>rem</tt>' instruction returns the remainder from the
1512 division of its two operands.</p>
1514 <p>The two arguments to the '<tt>rem</tt>' instruction must be either <a
1515 href="#t_integer">integer</a> or <a href="#t_floating">floating point</a>
1517 This instruction can also take <a href="#t_packed">packed</a> versions of the values.
1518 Both arguments must have identical types.</p>
1520 <p>This returns the <i>remainder</i> of a division (where the result
1521 has the same sign as the divisor), not the <i>modulus</i> (where the
1522 result has the same sign as the dividend) of a value. For more
1523 information about the difference, see <a
1524 href="http://mathforum.org/dr.math/problems/anne.4.28.99.html">The
1527 <pre> <result> = rem int 4, %var <i>; yields {int}:result = 4 % %var</i>
1530 <!-- _______________________________________________________________________ -->
1531 <div class="doc_subsubsection"> <a name="i_setcc">'<tt>set<i>cc</i></tt>'
1532 Instructions</a> </div>
1533 <div class="doc_text">
1535 <pre> <result> = seteq <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1536 <result> = setne <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1537 <result> = setlt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1538 <result> = setgt <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1539 <result> = setle <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1540 <result> = setge <ty> <var1>, <var2> <i>; yields {bool}:result</i>
1543 <p>The '<tt>set<i>cc</i></tt>' family of instructions returns a boolean
1544 value based on a comparison of their two operands.</p>
1546 <p>The two arguments to the '<tt>set<i>cc</i></tt>' instructions must
1547 be of <a href="#t_firstclass">first class</a> type (it is not possible
1548 to compare '<tt>label</tt>'s, '<tt>array</tt>'s, '<tt>structure</tt>'
1549 or '<tt>void</tt>' values, etc...). Both arguments must have identical
1552 <p>The '<tt>seteq</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1553 value if both operands are equal.<br>
1554 The '<tt>setne</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1555 value if both operands are unequal.<br>
1556 The '<tt>setlt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1557 value if the first operand is less than the second operand.<br>
1558 The '<tt>setgt</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1559 value if the first operand is greater than the second operand.<br>
1560 The '<tt>setle</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1561 value if the first operand is less than or equal to the second operand.<br>
1562 The '<tt>setge</tt>' instruction yields a <tt>true</tt> '<tt>bool</tt>'
1563 value if the first operand is greater than or equal to the second
1566 <pre> <result> = seteq int 4, 5 <i>; yields {bool}:result = false</i>
1567 <result> = setne float 4, 5 <i>; yields {bool}:result = true</i>
1568 <result> = setlt uint 4, 5 <i>; yields {bool}:result = true</i>
1569 <result> = setgt sbyte 4, 5 <i>; yields {bool}:result = false</i>
1570 <result> = setle sbyte 4, 5 <i>; yields {bool}:result = true</i>
1571 <result> = setge sbyte 4, 5 <i>; yields {bool}:result = false</i>
1574 <!-- ======================================================================= -->
1575 <div class="doc_subsection"> <a name="bitwiseops">Bitwise Binary
1576 Operations</a> </div>
1577 <div class="doc_text">
1578 <p>Bitwise binary operators are used to do various forms of
1579 bit-twiddling in a program. They are generally very efficient
1580 instructions and can commonly be strength reduced from other
1581 instructions. They require two operands, execute an operation on them,
1582 and produce a single value. The resulting value of the bitwise binary
1583 operators is always the same type as its first operand.</p>
1585 <!-- _______________________________________________________________________ -->
1586 <div class="doc_subsubsection"> <a name="i_and">'<tt>and</tt>'
1587 Instruction</a> </div>
1588 <div class="doc_text">
1590 <pre> <result> = and <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1593 <p>The '<tt>and</tt>' instruction returns the bitwise logical and of
1594 its two operands.</p>
1596 <p>The two arguments to the '<tt>and</tt>' instruction must be <a
1597 href="#t_integral">integral</a> values. Both arguments must have
1598 identical types.</p>
1600 <p>The truth table used for the '<tt>and</tt>' instruction is:</p>
1602 <div style="align: center">
1603 <table border="1" cellspacing="0" cellpadding="4">
1634 <pre> <result> = and int 4, %var <i>; yields {int}:result = 4 & %var</i>
1635 <result> = and int 15, 40 <i>; yields {int}:result = 8</i>
1636 <result> = and int 4, 8 <i>; yields {int}:result = 0</i>
1639 <!-- _______________________________________________________________________ -->
1640 <div class="doc_subsubsection"> <a name="i_or">'<tt>or</tt>' Instruction</a> </div>
1641 <div class="doc_text">
1643 <pre> <result> = or <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1646 <p>The '<tt>or</tt>' instruction returns the bitwise logical inclusive
1647 or of its two operands.</p>
1649 <p>The two arguments to the '<tt>or</tt>' instruction must be <a
1650 href="#t_integral">integral</a> values. Both arguments must have
1651 identical types.</p>
1653 <p>The truth table used for the '<tt>or</tt>' instruction is:</p>
1655 <div style="align: center">
1656 <table border="1" cellspacing="0" cellpadding="4">
1687 <pre> <result> = or int 4, %var <i>; yields {int}:result = 4 | %var</i>
1688 <result> = or int 15, 40 <i>; yields {int}:result = 47</i>
1689 <result> = or int 4, 8 <i>; yields {int}:result = 12</i>
1692 <!-- _______________________________________________________________________ -->
1693 <div class="doc_subsubsection"> <a name="i_xor">'<tt>xor</tt>'
1694 Instruction</a> </div>
1695 <div class="doc_text">
1697 <pre> <result> = xor <ty> <var1>, <var2> <i>; yields {ty}:result</i>
1700 <p>The '<tt>xor</tt>' instruction returns the bitwise logical exclusive
1701 or of its two operands. The <tt>xor</tt> is used to implement the
1702 "one's complement" operation, which is the "~" operator in C.</p>
1704 <p>The two arguments to the '<tt>xor</tt>' instruction must be <a
1705 href="#t_integral">integral</a> values. Both arguments must have
1706 identical types.</p>
1708 <p>The truth table used for the '<tt>xor</tt>' instruction is:</p>
1710 <div style="align: center">
1711 <table border="1" cellspacing="0" cellpadding="4">
1743 <pre> <result> = xor int 4, %var <i>; yields {int}:result = 4 ^ %var</i>
1744 <result> = xor int 15, 40 <i>; yields {int}:result = 39</i>
1745 <result> = xor int 4, 8 <i>; yields {int}:result = 12</i>
1746 <result> = xor int %V, -1 <i>; yields {int}:result = ~%V</i>
1749 <!-- _______________________________________________________________________ -->
1750 <div class="doc_subsubsection"> <a name="i_shl">'<tt>shl</tt>'
1751 Instruction</a> </div>
1752 <div class="doc_text">
1754 <pre> <result> = shl <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1757 <p>The '<tt>shl</tt>' instruction returns the first operand shifted to
1758 the left a specified number of bits.</p>
1760 <p>The first argument to the '<tt>shl</tt>' instruction must be an <a
1761 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1764 <p>The value produced is <tt>var1</tt> * 2<sup><tt>var2</tt></sup>.</p>
1766 <pre> <result> = shl int 4, ubyte %var <i>; yields {int}:result = 4 << %var</i>
1767 <result> = shl int 4, ubyte 2 <i>; yields {int}:result = 16</i>
1768 <result> = shl int 1, ubyte 10 <i>; yields {int}:result = 1024</i>
1771 <!-- _______________________________________________________________________ -->
1772 <div class="doc_subsubsection"> <a name="i_shr">'<tt>shr</tt>'
1773 Instruction</a> </div>
1774 <div class="doc_text">
1776 <pre> <result> = shr <ty> <var1>, ubyte <var2> <i>; yields {ty}:result</i>
1779 <p>The '<tt>shr</tt>' instruction returns the first operand shifted to
1780 the right a specified number of bits.</p>
1782 <p>The first argument to the '<tt>shr</tt>' instruction must be an <a
1783 href="#t_integer">integer</a> type. The second argument must be an '<tt>ubyte</tt>'
1786 <p>If the first argument is a <a href="#t_signed">signed</a> type, the
1787 most significant bit is duplicated in the newly free'd bit positions.
1788 If the first argument is unsigned, zero bits shall fill the empty
1791 <pre> <result> = shr int 4, ubyte %var <i>; yields {int}:result = 4 >> %var</i>
1792 <result> = shr uint 4, ubyte 1 <i>; yields {uint}:result = 2</i>
1793 <result> = shr int 4, ubyte 2 <i>; yields {int}:result = 1</i>
1794 <result> = shr sbyte 4, ubyte 3 <i>; yields {sbyte}:result = 0</i>
1795 <result> = shr sbyte -2, ubyte 1 <i>; yields {sbyte}:result = -1</i>
1799 <!-- ======================================================================= -->
1800 <div class="doc_subsection">
1801 <a name="memoryops">Memory Access Operations</a>
1804 <div class="doc_text">
1806 <p>A key design point of an SSA-based representation is how it
1807 represents memory. In LLVM, no memory locations are in SSA form, which
1808 makes things very simple. This section describes how to read, write,
1809 allocate, and free memory in LLVM.</p>
1813 <!-- _______________________________________________________________________ -->
1814 <div class="doc_subsubsection">
1815 <a name="i_malloc">'<tt>malloc</tt>' Instruction</a>
1818 <div class="doc_text">
1823 <result> = malloc <type>[, uint <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
1828 <p>The '<tt>malloc</tt>' instruction allocates memory from the system
1829 heap and returns a pointer to it.</p>
1833 <p>The '<tt>malloc</tt>' instruction allocates
1834 <tt>sizeof(<type>)*NumElements</tt>
1835 bytes of memory from the operating system and returns a pointer of the
1836 appropriate type to the program. If "NumElements" is specified, it is the
1837 number of elements allocated. If an alignment is specified, the value result
1838 of the allocation is guaranteed to be aligned to at least that boundary. If
1839 not specified, or if zero, the target can choose to align the allocation on any
1840 convenient boundary.</p>
1842 <p>'<tt>type</tt>' must be a sized type.</p>
1846 <p>Memory is allocated using the system "<tt>malloc</tt>" function, and
1847 a pointer is returned.</p>
1852 %array = malloc [4 x ubyte ] <i>; yields {[%4 x ubyte]*}:array</i>
1854 %size = <a href="#i_add">add</a> uint 2, 2 <i>; yields {uint}:size = uint 4</i>
1855 %array1 = malloc ubyte, uint 4 <i>; yields {ubyte*}:array1</i>
1856 %array2 = malloc [12 x ubyte], uint %size <i>; yields {[12 x ubyte]*}:array2</i>
1857 %array3 = malloc int, uint 4, align 1024 <i>; yields {int*}:array3</i>
1858 %array4 = malloc int, align 1024 <i>; yields {int*}:array4</i>
1862 <!-- _______________________________________________________________________ -->
1863 <div class="doc_subsubsection">
1864 <a name="i_free">'<tt>free</tt>' Instruction</a>
1867 <div class="doc_text">
1872 free <type> <value> <i>; yields {void}</i>
1877 <p>The '<tt>free</tt>' instruction returns memory back to the unused
1878 memory heap to be reallocated in the future.</p>
1882 <p>'<tt>value</tt>' shall be a pointer value that points to a value
1883 that was allocated with the '<tt><a href="#i_malloc">malloc</a></tt>'
1888 <p>Access to the memory pointed to by the pointer is no longer defined
1889 after this instruction executes.</p>
1894 %array = <a href="#i_malloc">malloc</a> [4 x ubyte] <i>; yields {[4 x ubyte]*}:array</i>
1895 free [4 x ubyte]* %array
1899 <!-- _______________________________________________________________________ -->
1900 <div class="doc_subsubsection">
1901 <a name="i_alloca">'<tt>alloca</tt>' Instruction</a>
1904 <div class="doc_text">
1909 <result> = alloca <type>[, uint <NumElements>][, align <alignment>] <i>; yields {type*}:result</i>
1914 <p>The '<tt>alloca</tt>' instruction allocates memory on the current
1915 stack frame of the procedure that is live until the current function
1916 returns to its caller.</p>
1920 <p>The '<tt>alloca</tt>' instruction allocates <tt>sizeof(<type>)*NumElements</tt>
1921 bytes of memory on the runtime stack, returning a pointer of the
1922 appropriate type to the program. If "NumElements" is specified, it is the
1923 number of elements allocated. If an alignment is specified, the value result
1924 of the allocation is guaranteed to be aligned to at least that boundary. If
1925 not specified, or if zero, the target can choose to align the allocation on any
1926 convenient boundary.</p>
1928 <p>'<tt>type</tt>' may be any sized type.</p>
1932 <p>Memory is allocated; a pointer is returned. '<tt>alloca</tt>'d
1933 memory is automatically released when the function returns. The '<tt>alloca</tt>'
1934 instruction is commonly used to represent automatic variables that must
1935 have an address available. When the function returns (either with the <tt><a
1936 href="#i_ret">ret</a></tt> or <tt><a href="#i_unwind">unwind</a></tt>
1937 instructions), the memory is reclaimed.</p>
1942 %ptr = alloca int <i>; yields {int*}:ptr</i>
1943 %ptr = alloca int, uint 4 <i>; yields {int*}:ptr</i>
1944 %ptr = alloca int, uint 4, align 1024 <i>; yields {int*}:ptr</i>
1945 %ptr = alloca int, align 1024 <i>; yields {int*}:ptr</i>
1949 <!-- _______________________________________________________________________ -->
1950 <div class="doc_subsubsection"> <a name="i_load">'<tt>load</tt>'
1951 Instruction</a> </div>
1952 <div class="doc_text">
1954 <pre> <result> = load <ty>* <pointer><br> <result> = volatile load <ty>* <pointer><br></pre>
1956 <p>The '<tt>load</tt>' instruction is used to read from memory.</p>
1958 <p>The argument to the '<tt>load</tt>' instruction specifies the memory
1959 address from which to load. The pointer must point to a <a
1960 href="#t_firstclass">first class</a> type. If the <tt>load</tt> is
1961 marked as <tt>volatile</tt>, then the optimizer is not allowed to modify
1962 the number or order of execution of this <tt>load</tt> with other
1963 volatile <tt>load</tt> and <tt><a href="#i_store">store</a></tt>
1966 <p>The location of memory pointed to is loaded.</p>
1968 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1970 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1971 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
1974 <!-- _______________________________________________________________________ -->
1975 <div class="doc_subsubsection"> <a name="i_store">'<tt>store</tt>'
1976 Instruction</a> </div>
1978 <pre> store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1979 volatile store <ty> <value>, <ty>* <pointer> <i>; yields {void}</i>
1982 <p>The '<tt>store</tt>' instruction is used to write to memory.</p>
1984 <p>There are two arguments to the '<tt>store</tt>' instruction: a value
1985 to store and an address in which to store it. The type of the '<tt><pointer></tt>'
1986 operand must be a pointer to the type of the '<tt><value></tt>'
1987 operand. If the <tt>store</tt> is marked as <tt>volatile</tt>, then the
1988 optimizer is not allowed to modify the number or order of execution of
1989 this <tt>store</tt> with other volatile <tt>load</tt> and <tt><a
1990 href="#i_store">store</a></tt> instructions.</p>
1992 <p>The contents of memory are updated to contain '<tt><value></tt>'
1993 at the location specified by the '<tt><pointer></tt>' operand.</p>
1995 <pre> %ptr = <a href="#i_alloca">alloca</a> int <i>; yields {int*}:ptr</i>
1997 href="#i_store">store</a> int 3, int* %ptr <i>; yields {void}</i>
1998 %val = load int* %ptr <i>; yields {int}:val = int 3</i>
2000 <!-- _______________________________________________________________________ -->
2001 <div class="doc_subsubsection">
2002 <a name="i_getelementptr">'<tt>getelementptr</tt>' Instruction</a>
2005 <div class="doc_text">
2008 <result> = getelementptr <ty>* <ptrval>{, <ty> <idx>}*
2014 The '<tt>getelementptr</tt>' instruction is used to get the address of a
2015 subelement of an aggregate data structure.</p>
2019 <p>This instruction takes a list of integer constants that indicate what
2020 elements of the aggregate object to index to. The actual types of the arguments
2021 provided depend on the type of the first pointer argument. The
2022 '<tt>getelementptr</tt>' instruction is used to index down through the type
2023 levels of a structure or to a specific index in an array. When indexing into a
2024 structure, only <tt>uint</tt>
2025 integer constants are allowed. When indexing into an array or pointer,
2026 <tt>int</tt> and <tt>long</tt> indexes are allowed of any sign.</p>
2028 <p>For example, let's consider a C code fragment and how it gets
2029 compiled to LLVM:</p>
2043 int *foo(struct ST *s) {
2044 return &s[1].Z.B[5][13];
2048 <p>The LLVM code generated by the GCC frontend is:</p>
2051 %RT = type { sbyte, [10 x [20 x int]], sbyte }
2052 %ST = type { int, double, %RT }
2056 int* %foo(%ST* %s) {
2058 %reg = getelementptr %ST* %s, int 1, uint 2, uint 1, int 5, int 13
2065 <p>The index types specified for the '<tt>getelementptr</tt>' instruction depend
2066 on the pointer type that is being indexed into. <a href="#t_pointer">Pointer</a>
2067 and <a href="#t_array">array</a> types require <tt>uint</tt>, <tt>int</tt>,
2068 <tt>ulong</tt>, or <tt>long</tt> values, and <a href="#t_struct">structure</a>
2069 types require <tt>uint</tt> <b>constants</b>.</p>
2071 <p>In the example above, the first index is indexing into the '<tt>%ST*</tt>'
2072 type, which is a pointer, yielding a '<tt>%ST</tt>' = '<tt>{ int, double, %RT
2073 }</tt>' type, a structure. The second index indexes into the third element of
2074 the structure, yielding a '<tt>%RT</tt>' = '<tt>{ sbyte, [10 x [20 x int]],
2075 sbyte }</tt>' type, another structure. The third index indexes into the second
2076 element of the structure, yielding a '<tt>[10 x [20 x int]]</tt>' type, an
2077 array. The two dimensions of the array are subscripted into, yielding an
2078 '<tt>int</tt>' type. The '<tt>getelementptr</tt>' instruction returns a pointer
2079 to this element, thus computing a value of '<tt>int*</tt>' type.</p>
2081 <p>Note that it is perfectly legal to index partially through a
2082 structure, returning a pointer to an inner element. Because of this,
2083 the LLVM code for the given testcase is equivalent to:</p>
2086 int* %foo(%ST* %s) {
2087 %t1 = getelementptr %ST* %s, int 1 <i>; yields %ST*:%t1</i>
2088 %t2 = getelementptr %ST* %t1, int 0, uint 2 <i>; yields %RT*:%t2</i>
2089 %t3 = getelementptr %RT* %t2, int 0, uint 1 <i>; yields [10 x [20 x int]]*:%t3</i>
2090 %t4 = getelementptr [10 x [20 x int]]* %t3, int 0, int 5 <i>; yields [20 x int]*:%t4</i>
2091 %t5 = getelementptr [20 x int]* %t4, int 0, int 13 <i>; yields int*:%t5</i>
2096 <p>Note that it is undefined to access an array out of bounds: array and
2097 pointer indexes must always be within the defined bounds of the array type.
2098 The one exception for this rules is zero length arrays. These arrays are
2099 defined to be accessible as variable length arrays, which requires access
2100 beyond the zero'th element.</p>
2105 <i>; yields [12 x ubyte]*:aptr</i>
2106 %aptr = getelementptr {int, [12 x ubyte]}* %sptr, long 0, uint 1
2110 <!-- ======================================================================= -->
2111 <div class="doc_subsection"> <a name="otherops">Other Operations</a> </div>
2112 <div class="doc_text">
2113 <p>The instructions in this category are the "miscellaneous"
2114 instructions, which defy better classification.</p>
2116 <!-- _______________________________________________________________________ -->
2117 <div class="doc_subsubsection"> <a name="i_phi">'<tt>phi</tt>'
2118 Instruction</a> </div>
2119 <div class="doc_text">
2121 <pre> <result> = phi <ty> [ <val0>, <label0>], ...<br></pre>
2123 <p>The '<tt>phi</tt>' instruction is used to implement the φ node in
2124 the SSA graph representing the function.</p>
2126 <p>The type of the incoming values are specified with the first type
2127 field. After this, the '<tt>phi</tt>' instruction takes a list of pairs
2128 as arguments, with one pair for each predecessor basic block of the
2129 current block. Only values of <a href="#t_firstclass">first class</a>
2130 type may be used as the value arguments to the PHI node. Only labels
2131 may be used as the label arguments.</p>
2132 <p>There must be no non-phi instructions between the start of a basic
2133 block and the PHI instructions: i.e. PHI instructions must be first in
2136 <p>At runtime, the '<tt>phi</tt>' instruction logically takes on the
2137 value specified by the parameter, depending on which basic block we
2138 came from in the last <a href="#terminators">terminator</a> instruction.</p>
2140 <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>
2143 <!-- _______________________________________________________________________ -->
2144 <div class="doc_subsubsection">
2145 <a name="i_cast">'<tt>cast .. to</tt>' Instruction</a>
2148 <div class="doc_text">
2153 <result> = cast <ty> <value> to <ty2> <i>; yields ty2</i>
2159 The '<tt>cast</tt>' instruction is used as the primitive means to convert
2160 integers to floating point, change data type sizes, and break type safety (by
2168 The '<tt>cast</tt>' instruction takes a value to cast, which must be a first
2169 class value, and a type to cast it to, which must also be a <a
2170 href="#t_firstclass">first class</a> type.
2176 This instruction follows the C rules for explicit casts when determining how the
2177 data being cast must change to fit in its new container.
2181 When casting to bool, any value that would be considered true in the context of
2182 a C '<tt>if</tt>' condition is converted to the boolean '<tt>true</tt>' values,
2183 all else are '<tt>false</tt>'.
2187 When extending an integral value from a type of one signness to another (for
2188 example '<tt>sbyte</tt>' to '<tt>ulong</tt>'), the value is sign-extended if the
2189 <b>source</b> value is signed, and zero-extended if the source value is
2190 unsigned. <tt>bool</tt> values are always zero extended into either zero or
2197 %X = cast int 257 to ubyte <i>; yields ubyte:1</i>
2198 %Y = cast int 123 to bool <i>; yields bool:true</i>
2202 <!-- _______________________________________________________________________ -->
2203 <div class="doc_subsubsection">
2204 <a name="i_select">'<tt>select</tt>' Instruction</a>
2207 <div class="doc_text">
2212 <result> = select bool <cond>, <ty> <val1>, <ty> <val2> <i>; yields ty</i>
2218 The '<tt>select</tt>' instruction is used to choose one value based on a
2219 condition, without branching.
2226 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.
2232 If the boolean condition evaluates to true, the instruction returns the first
2233 value argument; otherwise, it returns the second value argument.
2239 %X = select bool true, ubyte 17, ubyte 42 <i>; yields ubyte:17</i>
2244 <!-- _______________________________________________________________________ -->
2245 <div class="doc_subsubsection">
2246 <a name="i_extractelement">'<tt>extractelement</tt>' Instruction</a>
2249 <div class="doc_text">
2254 <result> = extractelement <n x <ty>> <val>, uint <idx> <i>; yields <ty></i>
2260 The '<tt>extractelement</tt>' instruction extracts a single scalar
2261 element from a vector at a specified index.
2268 The first operand of an '<tt>extractelement</tt>' instruction is a
2269 value of <a href="#t_packed">packed</a> type. The second operand is
2270 an index indicating the position from which to extract the element.
2271 The index may be a variable.</p>
2276 The result is a scalar of the same type as the element type of
2277 <tt>val</tt>. Its value is the value at position <tt>idx</tt> of
2278 <tt>val</tt>. If <tt>idx</tt> exceeds the length of <tt>val</tt>, the
2279 results are undefined.
2285 %result = extractelement <4 x int> %vec, uint 0 <i>; yields int</i>
2290 <!-- _______________________________________________________________________ -->
2291 <div class="doc_subsubsection">
2292 <a name="i_call">'<tt>call</tt>' Instruction</a>
2295 <div class="doc_text">
2299 <result> = [tail] call [<a href="#callingconv">cconv</a>] <ty>* <fnptrval>(<param list>)
2304 <p>The '<tt>call</tt>' instruction represents a simple function call.</p>
2308 <p>This instruction requires several arguments:</p>
2312 <p>The optional "tail" marker indicates whether the callee function accesses
2313 any allocas or varargs in the caller. If the "tail" marker is present, the
2314 function call is eligible for tail call optimization. Note that calls may
2315 be marked "tail" even if they do not occur before a <a
2316 href="#i_ret"><tt>ret</tt></a> instruction.
2319 <p>The optional "cconv" marker indicates which <a href="callingconv">calling
2320 convention</a> the call should use. If none is specified, the call defaults
2321 to using C calling conventions.
2324 <p>'<tt>ty</tt>': shall be the signature of the pointer to function value
2325 being invoked. The argument types must match the types implied by this
2326 signature. This type can be omitted if the function is not varargs and
2327 if the function type does not return a pointer to a function.</p>
2330 <p>'<tt>fnptrval</tt>': An LLVM value containing a pointer to a function to
2331 be invoked. In most cases, this is a direct function invocation, but
2332 indirect <tt>call</tt>s are just as possible, calling an arbitrary pointer
2333 to function value.</p>
2336 <p>'<tt>function args</tt>': argument list whose types match the
2337 function signature argument types. All arguments must be of
2338 <a href="#t_firstclass">first class</a> type. If the function signature
2339 indicates the function accepts a variable number of arguments, the extra
2340 arguments can be specified.</p>
2346 <p>The '<tt>call</tt>' instruction is used to cause control flow to
2347 transfer to a specified function, with its incoming arguments bound to
2348 the specified values. Upon a '<tt><a href="#i_ret">ret</a></tt>'
2349 instruction in the called function, control flow continues with the
2350 instruction after the function call, and the return value of the
2351 function is bound to the result argument. This is a simpler case of
2352 the <a href="#i_invoke">invoke</a> instruction.</p>
2357 %retval = call int %test(int %argc)
2358 call int(sbyte*, ...) *%printf(sbyte* %msg, int 12, sbyte 42);
2359 %X = tail call int %foo()
2360 %Y = tail call <a href="#callingconv">fastcc</a> int %foo()
2365 <!-- _______________________________________________________________________ -->
2366 <div class="doc_subsubsection">
2367 <a name="i_vaarg">'<tt>vaarg</tt>' Instruction</a>
2370 <div class="doc_text">
2375 <resultval> = va_arg <va_list*> <arglist>, <argty>
2380 <p>The '<tt>va_arg</tt>' instruction is used to access arguments passed through
2381 the "variable argument" area of a function call. It is used to implement the
2382 <tt>va_arg</tt> macro in C.</p>
2386 <p>This instruction takes a <tt>va_list*</tt> value and the type of
2387 the argument. It returns a value of the specified argument type and
2388 increments the <tt>va_list</tt> to point to the next argument. Again, the
2389 actual type of <tt>va_list</tt> is target specific.</p>
2393 <p>The '<tt>va_arg</tt>' instruction loads an argument of the specified
2394 type from the specified <tt>va_list</tt> and causes the
2395 <tt>va_list</tt> to point to the next argument. For more information,
2396 see the variable argument handling <a href="#int_varargs">Intrinsic
2399 <p>It is legal for this instruction to be called in a function which does not
2400 take a variable number of arguments, for example, the <tt>vfprintf</tt>
2403 <p><tt>va_arg</tt> is an LLVM instruction instead of an <a
2404 href="#intrinsics">intrinsic function</a> because it takes a type as an
2409 <p>See the <a href="#int_varargs">variable argument processing</a> section.</p>
2413 <!-- *********************************************************************** -->
2414 <div class="doc_section"> <a name="intrinsics">Intrinsic Functions</a> </div>
2415 <!-- *********************************************************************** -->
2417 <div class="doc_text">
2419 <p>LLVM supports the notion of an "intrinsic function". These functions have
2420 well known names and semantics and are required to follow certain
2421 restrictions. Overall, these instructions represent an extension mechanism for
2422 the LLVM language that does not require changing all of the transformations in
2423 LLVM to add to the language (or the bytecode reader/writer, the parser,
2426 <p>Intrinsic function names must all start with an "<tt>llvm.</tt>" prefix. This
2427 prefix is reserved in LLVM for intrinsic names; thus, functions may not be named
2428 this. Intrinsic functions must always be external functions: you cannot define
2429 the body of intrinsic functions. Intrinsic functions may only be used in call
2430 or invoke instructions: it is illegal to take the address of an intrinsic
2431 function. Additionally, because intrinsic functions are part of the LLVM
2432 language, it is required that they all be documented here if any are added.</p>
2435 <p>To learn how to add an intrinsic function, please see the <a
2436 href="ExtendingLLVM.html">Extending LLVM Guide</a>.
2441 <!-- ======================================================================= -->
2442 <div class="doc_subsection">
2443 <a name="int_varargs">Variable Argument Handling Intrinsics</a>
2446 <div class="doc_text">
2448 <p>Variable argument support is defined in LLVM with the <a
2449 href="#i_vanext"><tt>vanext</tt></a> instruction and these three
2450 intrinsic functions. These functions are related to the similarly
2451 named macros defined in the <tt><stdarg.h></tt> header file.</p>
2453 <p>All of these functions operate on arguments that use a
2454 target-specific value type "<tt>va_list</tt>". The LLVM assembly
2455 language reference manual does not define what this type is, so all
2456 transformations should be prepared to handle intrinsics with any type
2459 <p>This example shows how the <a href="#i_vanext"><tt>vanext</tt></a>
2460 instruction and the variable argument handling intrinsic functions are
2464 int %test(int %X, ...) {
2465 ; Initialize variable argument processing
2467 call void %<a href="#i_va_start">llvm.va_start</a>(sbyte** %ap)
2469 ; Read a single integer argument
2470 %tmp = va_arg sbyte** %ap, int
2472 ; Demonstrate usage of llvm.va_copy and llvm.va_end
2474 call void %<a href="#i_va_copy">llvm.va_copy</a>(sbyte** %aq, sbyte** %ap)
2475 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %aq)
2477 ; Stop processing of arguments.
2478 call void %<a href="#i_va_end">llvm.va_end</a>(sbyte** %ap)
2484 <!-- _______________________________________________________________________ -->
2485 <div class="doc_subsubsection">
2486 <a name="i_va_start">'<tt>llvm.va_start</tt>' Intrinsic</a>
2490 <div class="doc_text">
2492 <pre> declare void %llvm.va_start(<va_list>* <arglist>)<br></pre>
2494 <P>The '<tt>llvm.va_start</tt>' intrinsic initializes
2495 <tt>*<arglist></tt> for subsequent use by <tt><a
2496 href="#i_va_arg">va_arg</a></tt>.</p>
2500 <P>The argument is a pointer to a <tt>va_list</tt> element to initialize.</p>
2504 <P>The '<tt>llvm.va_start</tt>' intrinsic works just like the <tt>va_start</tt>
2505 macro available in C. In a target-dependent way, it initializes the
2506 <tt>va_list</tt> element the argument points to, so that the next call to
2507 <tt>va_arg</tt> will produce the first variable argument passed to the function.
2508 Unlike the C <tt>va_start</tt> macro, this intrinsic does not need to know the
2509 last argument of the function, the compiler can figure that out.</p>
2513 <!-- _______________________________________________________________________ -->
2514 <div class="doc_subsubsection">
2515 <a name="i_va_end">'<tt>llvm.va_end</tt>' Intrinsic</a>
2518 <div class="doc_text">
2520 <pre> declare void %llvm.va_end(<va_list*> <arglist>)<br></pre>
2522 <p>The '<tt>llvm.va_end</tt>' intrinsic destroys <tt><arglist></tt>
2523 which has been initialized previously with <tt><a href="#i_va_start">llvm.va_start</a></tt>
2524 or <tt><a href="#i_va_copy">llvm.va_copy</a></tt>.</p>
2526 <p>The argument is a <tt>va_list</tt> to destroy.</p>
2528 <p>The '<tt>llvm.va_end</tt>' intrinsic works just like the <tt>va_end</tt>
2529 macro available in C. In a target-dependent way, it destroys the <tt>va_list</tt>.
2530 Calls to <a href="#i_va_start"><tt>llvm.va_start</tt></a> and <a
2531 href="#i_va_copy"><tt>llvm.va_copy</tt></a> must be matched exactly
2532 with calls to <tt>llvm.va_end</tt>.</p>
2535 <!-- _______________________________________________________________________ -->
2536 <div class="doc_subsubsection">
2537 <a name="i_va_copy">'<tt>llvm.va_copy</tt>' Intrinsic</a>
2540 <div class="doc_text">
2545 declare void %llvm.va_copy(<va_list>* <destarglist>,
2546 <va_list>* <srcarglist>)
2551 <p>The '<tt>llvm.va_copy</tt>' intrinsic copies the current argument position from
2552 the source argument list to the destination argument list.</p>
2556 <p>The first argument is a pointer to a <tt>va_list</tt> element to initialize.
2557 The second argument is a pointer to a <tt>va_list</tt> element to copy from.</p>
2562 <p>The '<tt>llvm.va_copy</tt>' intrinsic works just like the <tt>va_copy</tt> macro
2563 available in C. In a target-dependent way, it copies the source
2564 <tt>va_list</tt> element into the destination list. This intrinsic is necessary
2565 because the <tt><a href="i_va_begin">llvm.va_begin</a></tt> intrinsic may be
2566 arbitrarily complex and require memory allocation, for example.</p>
2570 <!-- ======================================================================= -->
2571 <div class="doc_subsection">
2572 <a name="int_gc">Accurate Garbage Collection Intrinsics</a>
2575 <div class="doc_text">
2578 LLVM support for <a href="GarbageCollection.html">Accurate Garbage
2579 Collection</a> requires the implementation and generation of these intrinsics.
2580 These intrinsics allow identification of <a href="#i_gcroot">GC roots on the
2581 stack</a>, as well as garbage collector implementations that require <a
2582 href="#i_gcread">read</a> and <a href="#i_gcwrite">write</a> barriers.
2583 Front-ends for type-safe garbage collected languages should generate these
2584 intrinsics to make use of the LLVM garbage collectors. For more details, see <a
2585 href="GarbageCollection.html">Accurate Garbage Collection with LLVM</a>.
2589 <!-- _______________________________________________________________________ -->
2590 <div class="doc_subsubsection">
2591 <a name="i_gcroot">'<tt>llvm.gcroot</tt>' Intrinsic</a>
2594 <div class="doc_text">
2599 declare void %llvm.gcroot(<ty>** %ptrloc, <ty2>* %metadata)
2604 <p>The '<tt>llvm.gcroot</tt>' intrinsic declares the existence of a GC root to
2605 the code generator, and allows some metadata to be associated with it.</p>
2609 <p>The first argument specifies the address of a stack object that contains the
2610 root pointer. The second pointer (which must be either a constant or a global
2611 value address) contains the meta-data to be associated with the root.</p>
2615 <p>At runtime, a call to this intrinsics stores a null pointer into the "ptrloc"
2616 location. At compile-time, the code generator generates information to allow
2617 the runtime to find the pointer at GC safe points.
2623 <!-- _______________________________________________________________________ -->
2624 <div class="doc_subsubsection">
2625 <a name="i_gcread">'<tt>llvm.gcread</tt>' Intrinsic</a>
2628 <div class="doc_text">
2633 declare sbyte* %llvm.gcread(sbyte** %Ptr)
2638 <p>The '<tt>llvm.gcread</tt>' intrinsic identifies reads of references from heap
2639 locations, allowing garbage collector implementations that require read
2644 <p>The argument is the address to read from, which should be an address
2645 allocated from the garbage collector.</p>
2649 <p>The '<tt>llvm.gcread</tt>' intrinsic has the same semantics as a load
2650 instruction, but may be replaced with substantially more complex code by the
2651 garbage collector runtime, as needed.</p>
2656 <!-- _______________________________________________________________________ -->
2657 <div class="doc_subsubsection">
2658 <a name="i_gcwrite">'<tt>llvm.gcwrite</tt>' Intrinsic</a>
2661 <div class="doc_text">
2666 declare void %llvm.gcwrite(sbyte* %P1, sbyte** %P2)
2671 <p>The '<tt>llvm.gcwrite</tt>' intrinsic identifies writes of references to heap
2672 locations, allowing garbage collector implementations that require write
2673 barriers (such as generational or reference counting collectors).</p>
2677 <p>The first argument is the reference to store, and the second is the heap
2678 location to store to.</p>
2682 <p>The '<tt>llvm.gcwrite</tt>' intrinsic has the same semantics as a store
2683 instruction, but may be replaced with substantially more complex code by the
2684 garbage collector runtime, as needed.</p>
2690 <!-- ======================================================================= -->
2691 <div class="doc_subsection">
2692 <a name="int_codegen">Code Generator Intrinsics</a>
2695 <div class="doc_text">
2697 These intrinsics are provided by LLVM to expose special features that may only
2698 be implemented with code generator support.
2703 <!-- _______________________________________________________________________ -->
2704 <div class="doc_subsubsection">
2705 <a name="i_returnaddress">'<tt>llvm.returnaddress</tt>' Intrinsic</a>
2708 <div class="doc_text">
2712 declare void* %llvm.returnaddress(uint <level>)
2718 The '<tt>llvm.returnaddress</tt>' intrinsic returns a target-specific value
2719 indicating the return address of the current function or one of its callers.
2725 The argument to this intrinsic indicates which function to return the address
2726 for. Zero indicates the calling function, one indicates its caller, etc. The
2727 argument is <b>required</b> to be a constant integer value.
2733 The '<tt>llvm.returnaddress</tt>' intrinsic either returns a pointer indicating
2734 the return address of the specified call frame, or zero if it cannot be
2735 identified. The value returned by this intrinsic is likely to be incorrect or 0
2736 for arguments other than zero, so it should only be used for debugging purposes.
2740 Note that calling this intrinsic does not prevent function inlining or other
2741 aggressive transformations, so the value returned may not be that of the obvious
2742 source-language caller.
2747 <!-- _______________________________________________________________________ -->
2748 <div class="doc_subsubsection">
2749 <a name="i_frameaddress">'<tt>llvm.frameaddress</tt>' Intrinsic</a>
2752 <div class="doc_text">
2756 declare void* %llvm.frameaddress(uint <level>)
2762 The '<tt>llvm.frameaddress</tt>' intrinsic returns the target-specific frame
2763 pointer value for the specified stack frame.
2769 The argument to this intrinsic indicates which function to return the frame
2770 pointer for. Zero indicates the calling function, one indicates its caller,
2771 etc. The argument is <b>required</b> to be a constant integer value.
2777 The '<tt>llvm.frameaddress</tt>' intrinsic either returns a pointer indicating
2778 the frame address of the specified call frame, or zero if it cannot be
2779 identified. The value returned by this intrinsic is likely to be incorrect or 0
2780 for arguments other than zero, so it should only be used for debugging purposes.
2784 Note that calling this intrinsic does not prevent function inlining or other
2785 aggressive transformations, so the value returned may not be that of the obvious
2786 source-language caller.
2790 <!-- _______________________________________________________________________ -->
2791 <div class="doc_subsubsection">
2792 <a name="i_prefetch">'<tt>llvm.prefetch</tt>' Intrinsic</a>
2795 <div class="doc_text">
2799 declare void %llvm.prefetch(sbyte * <address>,
2800 uint <rw>, uint <locality>)
2807 The '<tt>llvm.prefetch</tt>' intrinsic is a hint to the code generator to insert
2808 a prefetch instruction if supported; otherwise, it is a noop. Prefetches have
2810 effect on the behavior of the program but can change its performance
2817 <tt>address</tt> is the address to be prefetched, <tt>rw</tt> is the specifier
2818 determining if the fetch should be for a read (0) or write (1), and
2819 <tt>locality</tt> is a temporal locality specifier ranging from (0) - no
2820 locality, to (3) - extremely local keep in cache. The <tt>rw</tt> and
2821 <tt>locality</tt> arguments must be constant integers.
2827 This intrinsic does not modify the behavior of the program. In particular,
2828 prefetches cannot trap and do not produce a value. On targets that support this
2829 intrinsic, the prefetch can provide hints to the processor cache for better
2835 <!-- _______________________________________________________________________ -->
2836 <div class="doc_subsubsection">
2837 <a name="i_pcmarker">'<tt>llvm.pcmarker</tt>' Intrinsic</a>
2840 <div class="doc_text">
2844 declare void %llvm.pcmarker( uint <id> )
2851 The '<tt>llvm.pcmarker</tt>' intrinsic is a method to export a Program Counter
2853 code to simulators and other tools. The method is target specific, but it is
2854 expected that the marker will use exported symbols to transmit the PC of the marker.
2855 The marker makes no guarantees that it will remain with any specific instruction
2856 after optimizations. It is possible that the presence of a marker will inhibit
2857 optimizations. The intended use is to be inserted after optmizations to allow
2858 correlations of simulation runs.
2864 <tt>id</tt> is a numerical id identifying the marker.
2870 This intrinsic does not modify the behavior of the program. Backends that do not
2871 support this intrinisic may ignore it.
2876 <!-- _______________________________________________________________________ -->
2877 <div class="doc_subsubsection">
2878 <a name="i_readcyclecounter">'<tt>llvm.readcyclecounter</tt>' Intrinsic</a>
2881 <div class="doc_text">
2885 declare ulong %llvm.readcyclecounter( )
2892 The '<tt>llvm.readcyclecounter</tt>' intrinsic provides access to the cycle
2893 counter register (or similar low latency, high accuracy clocks) on those targets
2894 that support it. On X86, it should map to RDTSC. On Alpha, it should map to RPCC.
2895 As the backing counters overflow quickly (on the order of 9 seconds on alpha), this
2896 should only be used for small timings.
2902 When directly supported, reading the cycle counter should not modify any memory.
2903 Implementations are allowed to either return a application specific value or a
2904 system wide value. On backends without support, this is lowered to a constant 0.
2910 <!-- ======================================================================= -->
2911 <div class="doc_subsection">
2912 <a name="int_os">Operating System Intrinsics</a>
2915 <div class="doc_text">
2917 These intrinsics are provided by LLVM to support the implementation of
2918 operating system level code.
2923 <!-- _______________________________________________________________________ -->
2924 <div class="doc_subsubsection">
2925 <a name="i_readport">'<tt>llvm.readport</tt>' Intrinsic</a>
2928 <div class="doc_text">
2932 declare <integer type> %llvm.readport (<integer type> <address>)
2938 The '<tt>llvm.readport</tt>' intrinsic reads data from the specified hardware
2945 The argument to this intrinsic indicates the hardware I/O address from which
2946 to read the data. The address is in the hardware I/O address namespace (as
2947 opposed to being a memory location for memory mapped I/O).
2953 The '<tt>llvm.readport</tt>' intrinsic reads data from the hardware I/O port
2954 specified by <i>address</i> and returns the value. The address and return
2955 value must be integers, but the size is dependent upon the platform upon which
2956 the program is code generated. For example, on x86, the address must be an
2957 unsigned 16-bit value, and the return value must be 8, 16, or 32 bits.
2962 <!-- _______________________________________________________________________ -->
2963 <div class="doc_subsubsection">
2964 <a name="i_writeport">'<tt>llvm.writeport</tt>' Intrinsic</a>
2967 <div class="doc_text">
2971 call void (<integer type>, <integer type>)*
2972 %llvm.writeport (<integer type> <value>,
2973 <integer type> <address>)
2979 The '<tt>llvm.writeport</tt>' intrinsic writes data to the specified hardware
2986 The first argument is the value to write to the I/O port.
2990 The second argument indicates the hardware I/O address to which data should be
2991 written. The address is in the hardware I/O address namespace (as opposed to
2992 being a memory location for memory mapped I/O).
2998 The '<tt>llvm.writeport</tt>' intrinsic writes <i>value</i> to the I/O port
2999 specified by <i>address</i>. The address and value must be integers, but the
3000 size is dependent upon the platform upon which the program is code generated.
3001 For example, on x86, the address must be an unsigned 16-bit value, and the
3002 value written must be 8, 16, or 32 bits in length.
3007 <!-- _______________________________________________________________________ -->
3008 <div class="doc_subsubsection">
3009 <a name="i_readio">'<tt>llvm.readio</tt>' Intrinsic</a>
3012 <div class="doc_text">
3016 declare <result> %llvm.readio (<ty> * <pointer>)
3022 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
3029 The argument to this intrinsic is a pointer indicating the memory address from
3030 which to read the data. The data must be a
3031 <a href="#t_firstclass">first class</a> type.
3037 The '<tt>llvm.readio</tt>' intrinsic reads data from a memory mapped I/O
3038 location specified by <i>pointer</i> and returns the value. The argument must
3039 be a pointer, and the return value must be a
3040 <a href="#t_firstclass">first class</a> type. However, certain architectures
3041 may not support I/O on all first class types. For example, 32-bit processors
3042 may only support I/O on data types that are 32 bits or less.
3046 This intrinsic enforces an in-order memory model for llvm.readio and
3047 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
3048 scheduled processors may execute loads and stores out of order, re-ordering at
3049 run time accesses to memory mapped I/O registers. Using these intrinsics
3050 ensures that accesses to memory mapped I/O registers occur in program order.
3055 <!-- _______________________________________________________________________ -->
3056 <div class="doc_subsubsection">
3057 <a name="i_writeio">'<tt>llvm.writeio</tt>' Intrinsic</a>
3060 <div class="doc_text">
3064 declare void %llvm.writeio (<ty1> <value>, <ty2> * <pointer>)
3070 The '<tt>llvm.writeio</tt>' intrinsic writes data to the specified memory
3077 The first argument is the value to write to the memory mapped I/O location.
3078 The second argument is a pointer indicating the memory address to which the
3079 data should be written.
3085 The '<tt>llvm.writeio</tt>' intrinsic writes <i>value</i> to the memory mapped
3086 I/O address specified by <i>pointer</i>. The value must be a
3087 <a href="#t_firstclass">first class</a> type. However, certain architectures
3088 may not support I/O on all first class types. For example, 32-bit processors
3089 may only support I/O on data types that are 32 bits or less.
3093 This intrinsic enforces an in-order memory model for llvm.readio and
3094 llvm.writeio calls on machines that use dynamic scheduling. Dynamically
3095 scheduled processors may execute loads and stores out of order, re-ordering at
3096 run time accesses to memory mapped I/O registers. Using these intrinsics
3097 ensures that accesses to memory mapped I/O registers occur in program order.
3102 <!-- ======================================================================= -->
3103 <div class="doc_subsection">
3104 <a name="int_libc">Standard C Library Intrinsics</a>
3107 <div class="doc_text">
3109 LLVM provides intrinsics for a few important standard C library functions.
3110 These intrinsics allow source-language front-ends to pass information about the
3111 alignment of the pointer arguments to the code generator, providing opportunity
3112 for more efficient code generation.
3117 <!-- _______________________________________________________________________ -->
3118 <div class="doc_subsubsection">
3119 <a name="i_memcpy">'<tt>llvm.memcpy</tt>' Intrinsic</a>
3122 <div class="doc_text">
3126 declare void %llvm.memcpy(sbyte* <dest>, sbyte* <src>,
3127 uint <len>, uint <align>)
3133 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
3134 location to the destination location.
3138 Note that, unlike the standard libc function, the <tt>llvm.memcpy</tt> intrinsic
3139 does not return a value, and takes an extra alignment argument.
3145 The first argument is a pointer to the destination, the second is a pointer to
3146 the source. The third argument is an (arbitrarily sized) integer argument
3147 specifying the number of bytes to copy, and the fourth argument is the alignment
3148 of the source and destination locations.
3152 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3153 the caller guarantees that the size of the copy is a multiple of the alignment
3154 and that both the source and destination pointers are aligned to that boundary.
3160 The '<tt>llvm.memcpy</tt>' intrinsic copies a block of memory from the source
3161 location to the destination location, which are not allowed to overlap. It
3162 copies "len" bytes of memory over. If the argument is known to be aligned to
3163 some boundary, this can be specified as the fourth argument, otherwise it should
3169 <!-- _______________________________________________________________________ -->
3170 <div class="doc_subsubsection">
3171 <a name="i_memmove">'<tt>llvm.memmove</tt>' Intrinsic</a>
3174 <div class="doc_text">
3178 declare void %llvm.memmove(sbyte* <dest>, sbyte* <src>,
3179 uint <len>, uint <align>)
3185 The '<tt>llvm.memmove</tt>' intrinsic moves a block of memory from the source
3186 location to the destination location. It is similar to the '<tt>llvm.memcpy</tt>'
3187 intrinsic but allows the two memory locations to overlap.
3191 Note that, unlike the standard libc function, the <tt>llvm.memmove</tt> intrinsic
3192 does not return a value, and takes an extra alignment argument.
3198 The first argument is a pointer to the destination, the second is a pointer to
3199 the source. The third argument is an (arbitrarily sized) integer argument
3200 specifying the number of bytes to copy, and the fourth argument is the alignment
3201 of the source and destination locations.
3205 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3206 the caller guarantees that the size of the copy is a multiple of the alignment
3207 and that both the source and destination pointers are aligned to that boundary.
3213 The '<tt>llvm.memmove</tt>' intrinsic copies a block of memory from the source
3214 location to the destination location, which may overlap. It
3215 copies "len" bytes of memory over. If the argument is known to be aligned to
3216 some boundary, this can be specified as the fourth argument, otherwise it should
3222 <!-- _______________________________________________________________________ -->
3223 <div class="doc_subsubsection">
3224 <a name="i_memset">'<tt>llvm.memset</tt>' Intrinsic</a>
3227 <div class="doc_text">
3231 declare void %llvm.memset(sbyte* <dest>, ubyte <val>,
3232 uint <len>, uint <align>)
3238 The '<tt>llvm.memset</tt>' intrinsic fills a block of memory with a particular
3243 Note that, unlike the standard libc function, the <tt>llvm.memset</tt> intrinsic
3244 does not return a value, and takes an extra alignment argument.
3250 The first argument is a pointer to the destination to fill, the second is the
3251 byte value to fill it with, the third argument is an (arbitrarily sized) integer
3252 argument specifying the number of bytes to fill, and the fourth argument is the
3253 known alignment of destination location.
3257 If the call to this intrinisic has an alignment value that is not 0 or 1, then
3258 the caller guarantees that the size of the copy is a multiple of the alignment
3259 and that the destination pointer is aligned to that boundary.
3265 The '<tt>llvm.memset</tt>' intrinsic fills "len" bytes of memory starting at the
3266 destination location. If the argument is known to be aligned to some boundary,
3267 this can be specified as the fourth argument, otherwise it should be set to 0 or
3273 <!-- _______________________________________________________________________ -->
3274 <div class="doc_subsubsection">
3275 <a name="i_isunordered">'<tt>llvm.isunordered</tt>' Intrinsic</a>
3278 <div class="doc_text">
3282 declare bool %llvm.isunordered(<float or double> Val1, <float or double> Val2)
3288 The '<tt>llvm.isunordered</tt>' intrinsic returns true if either or both of the
3289 specified floating point values is a NAN.
3295 The arguments are floating point numbers of the same type.
3301 If either or both of the arguments is a SNAN or QNAN, it returns true, otherwise
3307 <!-- _______________________________________________________________________ -->
3308 <div class="doc_subsubsection">
3309 <a name="i_sqrt">'<tt>llvm.sqrt</tt>' Intrinsic</a>
3312 <div class="doc_text">
3316 declare <float or double> %llvm.sqrt(<float or double> Val)
3322 The '<tt>llvm.sqrt</tt>' intrinsic returns the sqrt of the specified operand,
3323 returning the same value as the libm '<tt>sqrt</tt>' function would. Unlike
3324 <tt>sqrt</tt> in libm, however, <tt>llvm.sqrt</tt> has undefined behavior for
3325 negative numbers (which allows for better optimization).
3331 The argument and return value are floating point numbers of the same type.
3337 This function returns the sqrt of the specified operand if it is a positive
3338 floating point number.
3342 <!-- ======================================================================= -->
3343 <div class="doc_subsection">
3344 <a name="int_count">Bit Counting Intrinsics</a>
3347 <div class="doc_text">
3349 LLVM provides intrinsics for a few important bit counting operations.
3350 These allow efficient code generation for some algorithms.
3355 <!-- _______________________________________________________________________ -->
3356 <div class="doc_subsubsection">
3357 <a name="int_ctpop">'<tt>llvm.ctpop</tt>' Intrinsic</a>
3360 <div class="doc_text">
3364 declare int %llvm.ctpop(int <src>)
3371 The '<tt>llvm.ctpop</tt>' intrinsic counts the number of ones in a variable.
3377 The only argument is the value to be counted. The argument may be of any
3378 integer type. The return type must match the argument type.
3384 The '<tt>llvm.ctpop</tt>' intrinsic counts the 1's in a variable.
3388 <!-- _______________________________________________________________________ -->
3389 <div class="doc_subsubsection">
3390 <a name="int_ctlz">'<tt>llvm.ctlz</tt>' Intrinsic</a>
3393 <div class="doc_text">
3397 declare int %llvm.ctlz(int <src>)
3404 The '<tt>llvm.ctlz</tt>' intrinsic counts the number of leading zeros in a
3411 The only argument is the value to be counted. The argument may be of any
3412 integer type. The return type must match the argument type.
3418 The '<tt>llvm.ctlz</tt>' intrinsic counts the leading (most significant) zeros
3419 in a variable. If the src == 0 then the result is the size in bits of the type
3420 of src. For example, <tt>llvm.cttz(int 2) = 30</tt>.
3426 <!-- _______________________________________________________________________ -->
3427 <div class="doc_subsubsection">
3428 <a name="int_cttz">'<tt>llvm.cttz</tt>' Intrinsic</a>
3431 <div class="doc_text">
3435 declare int %llvm.cttz(int <src>)
3442 The '<tt>llvm.cttz</tt>' intrinsic counts the number of trailing zeros.
3448 The only argument is the value to be counted. The argument may be of any
3449 integer type. The return type must match the argument type.
3455 The '<tt>llvm.cttz</tt>' intrinsic counts the trailing (least significant) zeros
3456 in a variable. If the src == 0 then the result is the size in bits of the type
3457 of src. For example, <tt>llvm.cttz(2) = 1</tt>.
3461 <!-- ======================================================================= -->
3462 <div class="doc_subsection">
3463 <a name="int_debugger">Debugger Intrinsics</a>
3466 <div class="doc_text">
3468 The LLVM debugger intrinsics (which all start with <tt>llvm.dbg.</tt> prefix),
3469 are described in the <a
3470 href="SourceLevelDebugging.html#format_common_intrinsics">LLVM Source Level
3471 Debugging</a> document.
3476 <!-- *********************************************************************** -->
3479 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
3480 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
3481 <a href="http://validator.w3.org/check/referer"><img
3482 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
3484 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
3485 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
3486 Last modified: $Date$